The Complete IMS HALDB Guide

Front cover
The Complete IMS

HALDB Guide
All You Need to Know to Manage HALDBs
Examine how to migrate databases to
HALDB
Explore the benefits of using

HALDB
Learn how to administer

and modify HALDBs
Jouko Jäntti
Cornelia Hallmen
Raymond Keung
Rich Lewis
ibm.com/redbooks
International Technical Support Organization
The Complete IMS HALDB Guide

June 2003
SG24-6945-00
Note: Before using this information and the product it supports, read the information in
“Notices” on page xi.
First Edition (June 2003)
This edition applies to IMS Version 7 (program number 5655-B01) and IMS Version 8 (program
number 5655-C56) or later for use with the OS/390 or z/OS operating system.
© Copyright International Business Machines Corporation 2003. All rights reserved.

Note to U.S. Government Users Restricted Rights -- Use, duplication or disclosure restricted by GSA ADP
Schedule Contract with IBM Corp.
Contents
Notices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xi
Trademarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xii
Preface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xiii
The team that wrote this redbook . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xiii
Become a published author . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xiv
Comments welcome . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xv
Part 1. HALDB overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
Chapter 1. HALDB introduction and structure . . . . . . . . . . . . . . . . . . . . . . . 3

1.1 An introduction to High Availability Large Databases . . . . . . . . . . . . . . . . . 4
1.2 Features and benefits . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.3 Candidates for HALDB . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.4 HALDB definition process . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.4.1 Database Recovery Control (DBRC) . . . . . . . . . . . . . . . . . . . . . . . . . 9
1.5 DL/I processing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
1.6 Logical relationships with HALDB . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
1.7 Partition selection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
1.7.1 Partition selection using key ranges . . . . . . . . . . . . . . . . . . . . . . . . . 10
1.7.2 Partition selection using a partition selection exit routine . . . . . . . . . 11
1.8 Database structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
1.8.1 Partition ID and reorganization number . . . . . . . . . . . . . . . . . . . . . . 11
1.8.2 Segment structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
1.8.3 Extended pointer set (EPS). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
1.8.4 Indirect list data set (ILDS) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
1.8.5 Number of data sets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
1.8.6 Special considerations for secondary indexes . . . . . . . . . . . . . . . . . 16
1.9 Naming conventions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
1.9.1 Partition names . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
1.9.2 DD names . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
1.9.3 Data set names . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
Chapter 2. Defining HALDB databases . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21

2.1 Overview of HALDB definition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
2.1.1 Design the logical structure of the database . . . . . . . . . . . . . . . . . . . 22
2.1.2 Implement the logical structure with the DBDGEN process . . . . . . . 22
2.1.3 Determine the partitioning scheme . . . . . . . . . . . . . . . . . . . . . . . . . . 23
2.1.4 Create the partitioning scheme . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
© Copyright IBM Corp. 2003. All rights reserved. iii

2.1.5 Database exit routines . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
2.1.6 System definition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
2.1.7 Buffer pools . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
2.1.8 Dynamic allocation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
2.2 DBDGEN differences for HALDB . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
Chapter 3. HALDB and DBRC . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31

3.1 RECON records for HALDB . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
3.1.1 Master database record . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
3.1.2 Partition database record . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
3.1.3 Partition record . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
3.1.4 Partition DBDS record . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
3.2 Dynamic allocation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
3.3 Authorization processing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
3.4 Partition initialization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
3.5 DBRC commands . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
3.5.1 INIT commands . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
3.5.2 CHANGE commands . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
3.5.3 DELETE commands . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
3.5.4 LIST commands . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
3.5.5 GENJCL commands . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
3.6 DBRC groups for HALDB . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
3.6.1 Change accumulation groups . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
3.6.2 Database data set groups . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
3.6.3 Database groups . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
3.6.4 Recovery groups . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
3.7 Use of database names in DBRC commands . . . . . . . . . . . . . . . . . . . . . . 43
3.7.1 Commands that require a master database name . . . . . . . . . . . . . . 43
3.7.2 Commands that require a partition name . . . . . . . . . . . . . . . . . . . . . 43
3.7.3 Commands that allow a master database or a partition name . . . . . 44
3.7.4 DBRC commands that are not allowed with HALDB. . . . . . . . . . . . . 44
Chapter 4. Partition Definition utility . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45

4.1 Using the PDU. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
4.1.1 Configuring the PDU . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
4.1.2 Selecting a database. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
4.1.3 Setting HALDB master DBD parameters . . . . . . . . . . . . . . . . . . . . . 51
4.1.4 Setting processing options and global partition information . . . . . . . 52
4.1.5 Creating your partitions manually . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
4.1.6 Creating your partitions automatically . . . . . . . . . . . . . . . . . . . . . . . . 57
4.1.7 Changing partition definitions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
4.1.8 Deleting definitions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
Chapter 5. Batch definition of HALDB. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
iv The Complete IMS HALDB Guide All You Need to Know to Manage HALDBs
5.1 Using the batch interface. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
5.2 DBRC initialization commands for HALDB . . . . . . . . . . . . . . . . . . . . . . . . 62
5.2.1 INIT.DB . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
5.2.2 INIT.PART . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
5.3 DBRC change commands for HALDB. . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
5.3.1 CHANGE.DB . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
5.3.2 CHANGE.PART . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
5.3.3 CHANGE.DBDS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
5.4 DBRC delete commands for HALDB. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
5.4.1 DELETE.DB . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
5.4.2 DELETE.PART . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
Chapter 6. Partition initialization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73

6.1 Partition initialization function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
6.2 DBRC flags used with partition initialization . . . . . . . . . . . . . . . . . . . . . . . 75
6.2.1 Partition initialization required (PINIT) flag . . . . . . . . . . . . . . . . . . . . 75
6.2.2 Image copy needed flag . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
6.3 Database Prereorganization utility . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
6.4 Database Partition Data Set Initialization utility . . . . . . . . . . . . . . . . . . . . . 76
6.4.1 Unconditional partition initialization . . . . . . . . . . . . . . . . . . . . . . . . . . 78
Chapter 7. Partition selection. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79

7.1 Choosing the type of partition selection . . . . . . . . . . . . . . . . . . . . . . . . . . 80
7.1.1 Key range partitioning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
7.1.2 Partitioning with an exit routine . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
7.2 Writing partition selection exit routines . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
7.2.1 Partition selection exit routine functions . . . . . . . . . . . . . . . . . . . . . . 82
7.2.2 Information passed to and from the exit routine . . . . . . . . . . . . . . . . 84
7.2.3 Sample exit routine (DFSPSE00) . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
Part 2. Migration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
Chapter 8. Migration from non-HALDB to HALDB . . . . . . . . . . . . . . . . . . . 89

8.1 General migration considerations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90
8.1.1 DBD changes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90
8.2 HALDB Migration Aid utility (DFSMAID0) . . . . . . . . . . . . . . . . . . . . . . . . . 93
8.2.1 Reports . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
8.2.2 Adjusting the sizes from the reports . . . . . . . . . . . . . . . . . . . . . . . . . 94
8.2.3 Control statements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94
8.3 Migrating simple databases. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
8.3.1 Unloading the existing database . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
8.3.2 Saving existing database information . . . . . . . . . . . . . . . . . . . . . . . . 97
8.3.3 Changing the DBD . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
8.3.4 Deleting database information from the RECONs . . . . . . . . . . . . . . 97
Contents v
8.3.5 Defining the partitions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
8.3.6 Allocating database data sets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
8.3.7 Initializing the partitions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99
8.3.8 Reloading the HALDB database . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99
8.3.9 Image copying the database data sets . . . . . . . . . . . . . . . . . . . . . . 100
8.4 ILDS creation options . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100
8.4.1 HD Reload with no control statement . . . . . . . . . . . . . . . . . . . . . . . 100
8.4.2 HD Reload with an ILDSMULTI control statement . . . . . . . . . . . . . 100
8.4.3 HD Reload with a NOILDS control statement . . . . . . . . . . . . . . . . . 101
8.5 Migrating databases with secondary indexes . . . . . . . . . . . . . . . . . . . . . 101
8.5.1 Changing DBDs for secondary indexes . . . . . . . . . . . . . . . . . . . . . 102
8.5.2 Estimating the sizes of secondary index partitions . . . . . . . . . . . . . 107
8.5.3 Unloading databases with secondary indexes . . . . . . . . . . . . . . . . 109
8.5.4 Allocating secondary index data sets . . . . . . . . . . . . . . . . . . . . . . . 111
8.5.5 Loading databases and their secondary indexes . . . . . . . . . . . . . . 111
8.5.6 Secondary index pointers after the migration . . . . . . . . . . . . . . . . . 113
8.5.7 Using MIGRATE=YES with secondary indexes . . . . . . . . . . . . . . . 113
8.6 Migrating databases with logical relationships . . . . . . . . . . . . . . . . . . . . 114
8.6.1 Changing DBDs with logical relationships. . . . . . . . . . . . . . . . . . . . 114
8.6.2 Changing DBDs with logical relationships. . . . . . . . . . . . . . . . . . . . 115
8.6.3 Unloading logically related databases for migration . . . . . . . . . . . . 118
8.6.4 Loading logically related databases for migration . . . . . . . . . . . . . . 119
8.6.5 Logical relationship pointers after the migration . . . . . . . . . . . . . . . 119
8.7 Migrating from PDB or PDF. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 120
8.7.1 Partition selection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 120
8.7.2 DBD changes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 120
8.7.3 Using standard IMS utilities for the migration . . . . . . . . . . . . . . . . . 121
8.7.4 Using application programs for the migration . . . . . . . . . . . . . . . . . 121
8.7.5 Using PSU Unload and application programs for the migration . . . 122
8.8 Migrating from user partitioning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123
8.8.1 Using application programs for the migration . . . . . . . . . . . . . . . . . 123
8.8.2 Using HD Unload and application programs for the migration . . . . 125
8.8.3 Warning on the use of HD Reload . . . . . . . . . . . . . . . . . . . . . . . . . 126
8.9 Fallback to non-HALDB . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 126
8.9.1 Fallback with secondary indexes . . . . . . . . . . . . . . . . . . . . . . . . . . 127
8.9.2 Fallback with logical relationships . . . . . . . . . . . . . . . . . . . . . . . . . . 127
Chapter 9. Migration examples . . . . . . . . . . . . . . . . . . . ....... ...... . 129

9.1 Migration of HIDAM database to PHIDAM . . . . . . . . . ....... ...... . 130
9.1.1 Image copy of non-HALDB database data sets . ....... ...... . 130
9.1.2 Running the Migration Aid utility . . . . . . . . . . . . . ....... ...... . 130
9.1.3 Unloading the HIDAM database . . . . . . . . . . . . . ....... ...... . 136
9.1.4 Saving database information. . . . . . . . . . . . . . . . ....... ...... . 137
vi The Complete IMS HALDB Guide All You Need to Know to Manage HALDBs
9.1.5 Deleting the database from the RECONs . . . . . . . . . . . . . . . . . . . . 137
9.1.6 Changing the DBD and running the DBDGEN . . . . . . . . . . . . . . . . 138
9.1.7 Defining partitions with PDU . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139
9.1.8 Allocate HALDB data sets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139
9.1.9 Initializing HALDB partitions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 140
9.1.10 Loading HALDB database. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 141
9.1.11 Image copy HALDB data sets . . . . . . . . . . . . . . . . . . . . . . . . . . . . 142
9.1.12 ACBGEN . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 142
9.2 Migration of a database with a secondary index . . . . . . . . . . . . . . . . . . . 142
9.3 Migrating a PDB database . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 147
Chapter 10. Using the HALDB Conversion and Maintenance Aid . . . . . 151
10.1 HALDB Conversion and Maintenance Aid product . . . . . . . . . . . . . . . . 152
10.1.1 The conversion process overview . . . . . . . . . . . . . . . . . . . . . . . . . 153
10.2 Activating the tool . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 154
10.2.1 Modifying the startup CLIST . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 155
10.2.2 Allocating the environment description data set . . . . . . . . . . . . . . 155
10.3 Migration example with HALDB Conversion tool. . . . . . . . . . . . . . . . . . 155
10.3.1 Define an IMS environment. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 156
10.3.2 Define a project . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 164
10.3.3 Process a conversion project . . . . . . . . . . . . . . . . . . . . . . . . . . . . 168
10.4 Additional functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 176
Part 3. Application considerations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 181
Chapter 11. Application considerations . . . . . . . . . . . . . . . . . . . . . . . . . . 183

11.1 Initial loads . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 184
11.1.1 PHDAM considerations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 184
11.1.2 PHIDAM considerations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 184
11.1.3 Loading partitions in parallel . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 185
11.1.4 Initial loads of databases with secondary indexes . . . . . . . . . . . . 186
11.1.5 Initial loads of databases with logical relationships . . . . . . . . . . . . 188
11.2 Processing partitions in parallel. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 189
11.2.1 DBRC authorization of partitions . . . . . . . . . . . . . . . . . . . . . . . . . . 189
11.2.2 Parallel processing within an online system . . . . . . . . . . . . . . . . . 190
11.2.3 Parallel processing with block level data sharing . . . . . . . . . . . . . 190
11.2.4 Parallel processing without block level data sharing . . . . . . . . . . . 190
11.2.5 Program modifications for parallel processing . . . . . . . . . . . . . . . 192
11.3 Restricting a PCB to one partition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 193
11.3.1 Processing a partition sequentially . . . . . . . . . . . . . . . . . . . . . . . . 194
11.3.2 Requesting segments outside of the partition . . . . . . . . . . . . . . . . 194
11.3.3 Secondary index and logical relationship considerations . . . . . . . 195
11.4 Handling unavailable partitions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 197
11.4.1 Database PCB status code priming . . . . . . . . . . . . . . . . . . . . . . . 198
Contents vii
11.5 Processing secondary indexes as databases . . . . . . . . . . . . . . . . . . . . 198
11.5.1 Secondary indexes with /SX fields . . . . . . . . . . . . . . . . . . . . . . . . 199
11.5.2 Secondary indexes with symbolic pointers . . . . . . . . . . . . . . . . . . 202
11.6 Handling test environments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 205
11.6.1 DBRC registration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 205
11.6.2 Testing with non-HALDB databases . . . . . . . . . . . . . . . . . . . . . . . 206
11.7 Copying databases to different environments . . . . . . . . . . . . . . . . . . . . 207
11.7.1 The partition ID . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 207
11.7.2 Using image copies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 207
11.7.3 Using unloads . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 208
11.7.4 Copying part of the data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 208
11.7.5 Copying partition definitions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 208
11.7.6 Timestamp recoveries to a copy . . . . . . . . . . . . . . . . . . . . . . . . . . 208
11.7.7 Planned enhancement for copying partition definitions. . . . . . . . . 209
Part 4. Administration. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 211
Chapter 12. HALDB online commands . . . . . . . . . . . . . . . . . . . . . . . . . . . 213

12.1 Online commands with HALDB . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 214
12.2 /DISPLAY command . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 214
12.2.1 Display of a HALDB master database. . . . . . . . . . . . . . . . . . . . . . 215
12.2.2 Display of a HALDB partition. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 215
12.3 /DBRECOVERY command . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 217
12.3.1 DBR of a HALDB master database . . . . . . . . . . . . . . . . . . . . . . . . 217
12.3.2 DBR of a HALDB partition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 217
12.4 /DBDUMP command . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 218
12.4.1 DBD of a HALDB master database . . . . . . . . . . . . . . . . . . . . . . . . 218
12.4.2 DBD of a HALDB partition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 219
12.5 /START command . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 219
12.5.1 Start of a HALDB master database . . . . . . . . . . . . . . . . . . . . . . . . 219
12.5.2 Start of a HALDB partition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 219
12.6 /STOP command . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 220
12.6.1 Stop of a HALDB master database . . . . . . . . . . . . . . . . . . . . . . . . 220
12.6.2 Stop of a HALDB partition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 220
12.7 /LOCK command . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 220
12.7.1 Lock of a HALDB master database . . . . . . . . . . . . . . . . . . . . . . . . 220
12.7.2 Lock of a HALDB partition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 221
12.8 /UNLOCK command . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 221
12.8.1 Unlock of a HALDB master database . . . . . . . . . . . . . . . . . . . . . . 221
12.8.2 Unlock of a HALDB partition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 221
12.9 Using the ALL keyword with commands . . . . . . . . . . . . . . . . . . . . . . . . 221
12.10 Commands while databases are in use . . . . . . . . . . . . . . . . . . . . . . . 221
12.11 Command examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 222
viii The Complete IMS HALDB Guide All You Need to Know to Manage HALDBs
Chapter 13. Backup and recovery . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 227
13.1 Backup and recovery overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 228
13.2 Image copies for HALDB data sets . . . . . . . . . . . . . . . . . . . . . . . . . . . . 228
13.2.1 Database Image Copy utility (DFSUDMP0) example . . . . . . . . . . 229
13.2.2 Database Image Copy 2 utility (DFSUDMT0) example. . . . . . . . . 229
13.2.3 Using GENJCL to create image copy JCL . . . . . . . . . . . . . . . . . . 230
13.3 Recoveries of HALDB data sets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 230
13.3.1 Change Accumulation utility with HALDB . . . . . . . . . . . . . . . . . . . 231
13.3.2 Database Recovery utility and ORS with HALDB . . . . . . . . . . . . . 231
13.4 Rebuilding ILDS and PHIDAM index data sets . . . . . . . . . . . . . . . . . . . 236
13.4.1 Performance of the Index/ILDS Rebuild utility . . . . . . . . . . . . . . . 237
13.4.2 Using GENJCL.USER to generate Index/ILDS Rebuild JCL . . . . 238
13.5 Recovering HALDB secondary index data sets . . . . . . . . . . . . . . . . . . 239
13.5.1 Using IMS Index Builder for recoveries . . . . . . . . . . . . . . . . . . . . . 241
13.6 Batch Backout utility with HALDB . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 241
Chapter 14. HALDB reorganization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 243

14.1 The reorganization process . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 244
14.1.1 Non-HALDB reorganization overview . . . . . . . . . . . . . . . . . . . . . . 244
14.1.2 HALDB reorganization overview . . . . . . . . . . . . . . . . . . . . . . . . . . 245
14.1.3 Options for reorganization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 246
14.1.4 Unloading HALDB partitions and databases . . . . . . . . . . . . . . . . . 246
14.1.5 Reallocating HALDB database data sets . . . . . . . . . . . . . . . . . . . 247
14.1.6 Reloading HALDB partitions and databases . . . . . . . . . . . . . . . . . 249
14.1.7 Reorganizing HALDB secondary indexes . . . . . . . . . . . . . . . . . . . 251
14.2 The self-healing pointer process . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 251
14.2.1 The elements of the self-healing process . . . . . . . . . . . . . . . . . . . 251
14.2.2 Finding target segments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 253
14.2.3 Healing pointers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 254
Chapter 15. Changing existing HALDB databases. . . . . . . . . . . . . . . . . . 257

15.1 Changing database definitions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 258
15.1.1 Changes not requiring unload and reload of the database . . . . . . 258
15.1.2 Changes requiring unload and reload of the database . . . . . . . . . 258
15.1.3 Using online change with HALDB . . . . . . . . . . . . . . . . . . . . . . . . . 258
15.1.4 Adding a secondary index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 259
15.2 Changing partition definitions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 259
15.2.1 Changing high key definitions . . . . . . . . . . . . . . . . . . . . . . . . . . . . 260
15.2.2 Making changes with a partition selection exit routine . . . . . . . . . 265
15.2.3 Changing partition names . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 267
15.2.4 Partition ID numbers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 267
15.3 Changing secondary indexes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 268
15.4 Changing logically related databases . . . . . . . . . . . . . . . . . . . . . . . . . . 269
Contents ix
Chapter 16. Using IMS tools for HALDB . . . . . . . . . . . . . . . . . . . . . . . . . . 271
16.1 IMS High Performance Unload . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 272
16.2 IMS High Performance Load . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 273
16.3 IMS Index Builder . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 275
16.4 IMS Image Copy Extensions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 276
Part 5. Appendixes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 279
Appendix A. Sample GENJCL.USER skeletons for HALDB allocation . 281

Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 282
Creation of skeletal JCL . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 282
Index skeletal JCL member. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 283
ILDS skeletal JCL member . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 284
VSAM data skeletal JCL member . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 284
OSAM data skeletal JCL member . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 285
Defaults member . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 286
Definition of DBDS groups . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 286
Execution of GENJCL.USER commands . . . . . . . . . . . . . . . . . . . . . . . . . . . . 288
Execution of allocation jobs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 289
Appendix B. HALDB functions added by APARs . . . . . . . . . . . . . . . . . . . 291

HALDB functions delivered by APARs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 292
Appendix C. Summary of HALDB utilities . . . . . . . . . . . . . . . . . . . . . . . . 293

IMS database utilities for HALDBs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 294
Abbreviations and acronyms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 297
Related publications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 299

IBM Redbooks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 299
Other publications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 299
Online resources . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 300
How to get IBM Redbooks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 300
Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 301
x The Complete IMS HALDB Guide All You Need to Know to Manage HALDBs
Notices
This information was developed for products and services offered in the U.S.A.
IBM may not offer the products, services, or features discussed in this document in other countries. Consult
your local IBM representative for information on the products and services currently available in your area.
Any reference to an IBM product, program, or service is not intended to state or imply that only that IBM
product, program, or service may be used. Any functionally equivalent product, program, or service that
does not infringe any IBM intellectual property right may be used instead. However, it is the user's
responsibility to evaluate and verify the operation of any non-IBM product, program, or service.
IBM may have patents or pending patent applications covering subject matter described in this document.
The furnishing of this document does not give you any license to these patents. You can send license
inquiries, in writing, to:
IBM Director of Licensing, IBM Corporation, North Castle Drive Armonk, NY 10504-1785 U.S.A.
The following paragraph does not apply to the United Kingdom or any other country where such provisions
are inconsistent with local law: INTERNATIONAL BUSINESS MACHINES CORPORATION PROVIDES
THIS PUBLICATION "AS IS" WITHOUT WARRANTY OF ANY KIND, EITHER EXPRESS OR IMPLIED,
INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF NON-INFRINGEMENT,
MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE. Some states do not allow disclaimer
of express or implied warranties in certain transactions, therefore, this statement may not apply to you.
This information could include technical inaccuracies or typographical errors. Changes are periodically made
to the information herein; these changes will be incorporated in new editions of the publication. IBM may
make improvements and/or changes in the product(s) and/or the program(s) described in this publication at
any time without notice.
Any references in this information to non-IBM Web sites are provided for convenience only and do not in any
manner serve as an endorsement of those Web sites. The materials at those Web sites are not part of the
materials for this IBM product and use of those Web sites is at your own risk.
IBM may use or distribute any of the information you supply in any way it believes appropriate without
incurring any obligation to you.
Information concerning non-IBM products was obtained from the suppliers of those products, their published
announcements or other publicly available sources. IBM has not tested those products and cannot confirm
the accuracy of performance, compatibility or any other claims related to non-IBM products. Questions on
the capabilities of non-IBM products should be addressed to the suppliers of those products.
This information contains examples of data and reports used in daily business operations. To illustrate them
as completely as possible, the examples include the names of individuals, companies, brands, and products.
All of these names are fictitious and any similarity to the names and addresses used by an actual business
enterprise is entirely coincidental.
COPYRIGHT LICENSE:
This information contains sample application programs in source language, which illustrates programming
techniques on various operating platforms. You may copy, modify, and distribute these sample programs in
any form without payment to IBM, for the purposes of developing, using, marketing or distributing application
programs conforming to the application programming interface for the operating platform for which the
sample programs are written. These examples have not been thoroughly tested under all conditions. IBM,
therefore, cannot guarantee or imply reliability, serviceability, or function of these programs. You may copy,
modify, and distribute these sample programs in any form without payment to IBM for the purposes of
developing, using, marketing, or distributing application programs conforming to IBM's application
programming interfaces.
© Copyright IBM Corp. 2003. All rights reserved. xi

Trademarks
The following terms are trademarks of the International Business Machines Corporation in the United States,
other countries, or both:
CICS® Language Environment® Sequent®

FICON™ MVS™ SP1®
IBM® OS/390® SP™
ibm.com® Redbooks (logo) ™FICON™ ™
IMS/ESA® Redbooks™ WebSphere®
IMS™ S/390® z/OS™
The following terms are trademarks of other companies:
Partitioned Database Facility (PDF) is a trademark of Neon Enterprise Software, Inc. in the United States,
other countries, or both.
ActionMedia, LANDesk, MMX, Pentium and ProShare are trademarks of Intel Corporation in the United
States, other countries, or both.
Microsoft, Windows, Windows NT, and the Windows logo are trademarks of Microsoft Corporation in the
United States, other countries, or both.
Java and all Java-based trademarks and logos are trademarks or registered trademarks of Sun
Microsystems, Inc. in the United States, other countries, or both.
C-bus is a trademark of Corollary, Inc. in the United States, other countries, or both.
UNIX is a registered trademark of The Open Group in the United States and other countries.
SET, SET Secure Electronic Transaction, and the SET Logo are trademarks owned by SET Secure
Electronic Transaction LLC.
Other company, product, and service names may be trademarks or service marks of others.
xii The Complete IMS HALDB Guide All You Need to Know to Manage HALDBs
Preface
This IBM Redbook describes the High Availability Large Database (HALDB)
capability available with IMS™. IMS HALDB was introduced with IMS Version 7.
It allows IMS databases to grow to almost unlimited sizes while providing
increased availability. This book updates IMS Version 7 High Availability Large
Database Guide, SG24-5751, as well as adds topics that were not covered in the
previous book.
This publication provides a broad explanation of HALDB and its uses. Specific
areas covered include:
򐂰 HALDB overview, definition, and structure
򐂰 Migration from non-HALDB databases
򐂰 Application considerations
򐂰 HALDB database administration
This publication documents our hands-on experience in a test environment. It

includes migration and administration examples. Some IBM Data Management
Tools for IMS are also discussed in this book. Special emphasis is given to the
IMS HALDB Conversion and Maintenance Aid product. Examples of the use of
these tools are provided.
The team that wrote this redbook

This redbook was produced by a team of specialists from around the world
working at the International Technical Support Organization, San Jose Center.
Jouko Jäntti is a Project Leader specializing in IMS with the IBM International
Technical Support Organization, San Jose Center. He holds a Bachelors degree
in Business Information Technology from Helsinki Business Polytechnic, Finland.
Before joining the ITSO in September 2001, Jouko worked as an Advisory IT
Specialist at IBM Global Services, Finland. Jouko has been working on several
e-business projects with customers as a specialist in IMS, WebSphere®, and
UNIX on the OS/390® platform. He has also been responsible for the local IMS
support for Finnish IMS customers. Prior to joining IBM in 1997, he worked as a
Systems Programmer and Transaction Management Specialist in a large Finnish
bank for 13 years, and was responsible for the bank's IMS systems.
Cornelia Hallmen is an IMS instructor in Munich, Germany. She has over 11

years of experience as an IMS Systems Programmer in a large German bank
© Copyright IBM Corp. 2003. All rights reserved. xiii

and was responsible for its IMS systems, especially for high availability and for
the implementation of new versions and functions. She joined IBM in 2000 as an
instructor for IMS databases.
Raymond Keung is a Systems Analyst/IMS Database Administrator in Toronto,

Canada. He has five years of experience in IMS. He has worked at IBM for 14
years. His areas of expertise include technical support for IMS upgrades and
software installation for database conversions. He is also responsible for
performing IMS database administrative tasks and maintaining data and
structural integrity of production databases.
Rich Lewis is a Senior Consulting IT Specialist with IBM Advanced Technical

Support, Americas, located in Dallas, Texas. He has provided IMS technical
support and consulting services as a member of the Dallas Systems Center for
over 20 years. His areas of expertise include IMS database design, performance,
and availability. He has developed and taught IMS classes for IBM and is a
frequent speaker at IMS technical conferences, user group meetings, road
shows, and seminars.
Thanks to the following people for their contributions to this project:
Rich Conway, Bob Haimowitz, Julie Czubik

International Technical Support Organization, Poughkeepsie Center
Rose Levin, Becky Bard, Harley Beier, Claudia Ho, John Langley, Rick
Long, Elmer Rohde, Pete Sadler, Pat Schroeck, Larry Sullivan, Vern Watts
IBM Silicon Valley Laboratory, San Jose, USA
Christian Koeppen
Peck-Software, USA
Rick Long, Ken Daugherty, Jim Kelly, Geoff Nicholls, Norbert Theurich
The authors of IMS Version 7 High Availability Large Database Guide,
SG24-5751
Become a published author

Join us for a two- to six-week residency program! Help write an IBM Redbook
dealing with specific products or solutions, while getting hands-on experience
with leading-edge technologies. You'll team with IBM technical professionals,
Business Partners and/or customers.
Your efforts will help increase product acceptance and customer satisfaction. As
a bonus, you'll develop a network of contacts in IBM development labs, and
increase your productivity and marketability.
xiv The Complete IMS HALDB Guide All You Need to Know to Manage HALDBs
Find out more about the residency program, browse the residency index, and
apply online at:
ibm.com®/redbooks/residencies.html
Comments welcome
Your comments are important to us!
We want our Redbooks™ to be as helpful as possible. Send us your comments

about this or other Redbooks in one of the following ways:
򐂰 Use the online Contact us review redbook form found at:
ibm.com/redbooks
򐂰 Send your comments in an Internet note to:
[email protected]
򐂰 Mail your comments to:
IBM Corporation, International Technical Support Organization
Dept. QXXE Building 80-E2
650 Harry Road
San Jose, California 95120-6099
Preface xv
xvi The Complete IMS HALDB Guide All You Need to Know to Manage HALDBs
Part 1
Part 1 HALDB
overview
In this part we discuss the benefits of using HALDB databases. We identify which
databases are candidates for migration to HALDB. We describe the structure of
HALDB databases and their partitions. We explain how you define HALDB
databases and how to initialize their partitions. We describe the ways in which
HALDB can assign database records to partitions.
© Copyright IBM Corp. 2003. All rights reserved. 1

2 The Complete IMS HALDB Guide All You Need to Know to Manage HALDBs
1
Chapter 1. HALDB introduction and

structure
High Availability Large Database (HALDB) was introduced with IMS Version 7 to
extend the functions provided by IMS full-function databases. As the name
suggests, HALDB is designed to increase the capacity and availability of
full-function databases. HALDB has three database types:
PHDAM Partitioned HDAM
PHIDAM Partitioned HIDAM
PSINDEX Partitioned secondary index
HALDB support includes the following:

򐂰 OSAM and VSAM (ESDS and KSDS) access methods
򐂰 Logical relationships and secondary indexes
򐂰 Data sharing
򐂰 Remote Site Recovery (RSR)
򐂰 Extended Recovery Facility (XRF)
򐂰 Online change
򐂰 OSAM sequential buffering
򐂰 IMS Monitor
򐂰 IMS Performance Analyzer product
This chapter provides an overview of HALDB and its structure.

1.1 An introduction to High Availability Large Databases
The capacity of a non-HALDB full-function database is limited by the number of
data sets allowed in a database and their maximum sizes. The capacity of a data
set is either 4 GB if it is VSAM or 8 GB for OSAM. Up to ten data sets are
allowed. All segments of the same type must be in the same data set. So, when
one data set is full, the database is full.
HALDB removes this limit by removing the restriction that all occurrences of the
same segment type must be in the same data set. HALDB groups database
records into sets of partitions. The hierarchic structure is maintained within a
partition and a whole database record always resides in a single partition.
Partitions may be very large. Each partition may have up to ten data sets. A
database may have up to 1001 partitions, so segments in a HALDB database
may be stored in up to 10,010 data sets. Each HALDB data set may be up to 4
GB, so a HALDB database may contain over 40 terabytes of data.
HALDB also has availability and manageability benefits. Partitions may be

managed, allocated, authorized, and allocated independently. Some partitions
may be available to an online system while others are being processed by
utilities. This increases the flexibility available in handling databases.
HALDB has several benefits:

Capacity It allows a single database to be very large.
Manageability It allows for the division of a database into small partitions
that are easier to manage than the entire database.
Availability It allows different subsystems or utilities access to
manage partitions independently.
Usability It provides for increased use of online processes for
definitions.
Compatibility It provides application programming compatibility with
non-HALDB databases.
These benefits are delivered with application compatibility. Only in rare cases are
application programming changes required when a database is migrated to
HALDB.
1.2 Features and benefits
HALDB addresses a number of limitations associated with non-HALDB
full-function databases. Non-HALDB database limitations include:
򐂰 Storage capacity for each database is limited by the maximum size of its data
sets. These sizes are 4 GB for VSAM and 8 GB for OSAM.
򐂰 Limited options are available for increasing the storage capacity of a single
database.
򐂰 Availability of the database is not granular. The entire database is either
available or unavailable.
򐂰 Secondary index databases with direct pointers must be rebuilt when their
indexed databases are reorganized.
򐂰 Logical relationships require updates to pointers in related databases when a
database is reorganized. Utility executions are required to make these
updates.
򐂰 Increasing the size of a database requires the reorganization of the entire
database. This may require an extended outage.
Each of these limitations is specifically addressed by HALDB while maintaining

application compatibility.
HALDB increases the storage capacity of full-function databases by providing the

ability to partition each database. Each database can have up to 1001 partitions,
and each partition can have up to 10 data set groups, each of which can be up to
4 GB. Application programs access the database without needing to understand
the partitioning of the database.
HALDB improves the manageability of databases. Large databases can be split

into several partitions with small data sets. These smaller data sets may be
easier to handle. This is particularly true for databases that previously required
multi-volume data sets. You can create multiple small partitions such that each
data set fits on a single volume. This can simplify database data set
management.
Availability of data can also be improved with HALDB. Allocation, authorization,

reorganization, and recovery can be done independently by partition. In the
event that a partition is unavailable, programs can continue to access the
partitions that are available. A status code can be returned to a program to
indicate that data are presently unavailable. In this way, much of the database
can remain available even when parts of it are offline. Similarly, reorganization of
the database can be run independently for each partition. Only those partitions of
the database that are being reorganized are unavailable. Online systems and
batch jobs may continue to access the rest of the partitions.
Chapter 1. HALDB introduction and structure 5

Secondary indexes can be managed independently from their indexed
databases. Logically related databases can be managed independently from
each other. The reorganization of a partition or database does not require
changes to secondary indexes or logically related databases. HALDB’s
self-healing process for these pointers provides this independence.
1.3 Candidates for HALDB

HALDB can be used to replace any existing HDAM or HIDAM database and its
secondary indexes. The migration process for these databases is similar to an
ordinary reorganization process. The database is unloaded into a sequential
data set, the database is redefined as a HALDB, and the database is reloaded as
a HALDB database. Application programs can then access the new HALDB
database. In almost all cases, no changes are required to existing application
programs.
The following list categorizes databases that are candidates for conversion to
HALDB:
򐂰 Very large databases
The most obvious candidates for HALDB are the databases that are
approaching the 4 GB (VSAM) or 8 GB (OSAM) data set limitation. Their
conversion to HALDB provides for future growth and makes them more
manageable.
򐂰 Previously partitioned databases
Some of these large databases already may have been partitioned with user
partitioning or by products like IMS/ESA® Partition Support Product (PDB).
These databases are also candidates for conversion to HALDB.
򐂰 Databases benefitting from improved reorganization processes
HALDB partitioning provides faster reorganizations. This means that the
reorganizations may be done more frequently. The self-healing pointer
process for secondary indexes and logical relationship eliminates many of the
utilities associated with the reorganization of non-HALDB databases. This
simplifies their reorganization.
򐂰 Databases benefitting from more free space
The virtually unlimited size of HALDB databases allows users to specify
increased free space. This may improve database performance and reduce or
eliminate the need to reorganize them.
򐂰 Databases benefitting from parallel processing
Parallel execution of application programs may access the same database. A
conversion of the database to HALDB simplifies this processing.
There are advantages to using HALDB even with only one partition:
򐂰 The reorganization process for databases with secondary indexes and logical
relationships is shortened and simplified.
򐂰 PHIDAM databases are more easily managed than HIDAM databases. The
primary index of a PHIDAM database is no longer a separate database.
򐂰 If the database grows, space may be added by simply adding partitions.
Future IMS database enhancements

HALDB support is the basis for future enhancements to IMS full-function
databases. Conversions to HALDB may provide both immediate benefits and
additional future benefits.
1.4 HALDB definition process

A HALDB definition consists of two distinct entities, the master database
definition and its associated partition definitions. The HALDB master database
definition is created by the ordinary DBDGEN process. This definition is input to
the partition definition process. This is shown in Figure 1-1 on page 8.

DBD NAME=MYDB,...
Master DATASET ...
Database SEGMENT NAME=SEG1,...
FIELD NAME=FIELD1,...
FIELD NAME=FIELD2,...
Partition
Partition DATASET(s)
Information
Partition attributes set by PDU utility
High-key = AA
Information
Partition stored in the Recon
High-key = BB
Information
Partition
Partition
High-key = CC
Information
High-key = DD
Figure 1-1 Changes to the database definition with HALDB
The definition of partitions is external to DBDGEN. DBDGEN is used to define

the organization of the database including its hierarchic structure, data set group
boundaries, pointer options, logical relationships, and secondary indexes. It does
not generate the partition definitions or data set information. The HALDB
Partition Definition utility (PDU) is an ISPF interface to DBRC. It provides the
mechanism for the definition of partitions and data set characteristics. It stores
these definitions in the DBRC RECONs. Alternatively, DBRC batch commands
can be used to define the partitions and data sets.
Both databases and partitions may be controlled and displayed with online
commands. Commands used with non-HALDB databases, such as /START,
/STOP, and /DISPLAY, are also used with HALDB databases and partitions.
See Figure 1-2 on page 9 for an illustration of a database before and after
conversion to HALDB.
DB Name
DB Name = MAGIC
MAGIC HALDB Master DB
HALDB DB Name
Name = MAGIC
MAGIC
TYPE=HDAM
TYPE= HDAM TYPE=PHDAM
TYPE= PHDAM
Partitions: PART1,
Partitions: PART1, PART2,
PART2, PART3
PART3
M M
A G A G
I C I C
DS1 DS2
A B ILDS
PART1 A B ILDS
A B ILDS
PART2
PART3
Figure 1-2 IMS database before and after conversion to HALDB
1.4.1 Database Recovery Control (DBRC)

Database Recovery Control (DBRC) is required for HALDB. HALDB databases
must be registered in the DBRC recovery control data set (RECON) either by use
of PDU or DBRC batch commands. DBRC information is also used for dynamic
allocation. DFSMDA members are not used.
From DBRC’s point of view, a HALDB is a master database consisting of partition

databases.
1.5 DL/I processing

The application programming interface for HALDB is the same as that used with
non-HALDB full-function databases. In general, programs that process a
non-HALDB will operate the same when the database is converted to HALDB.
There are three exceptions. First, if you make a partition unavailable, the
program will receive an unavailable data condition. This condition can exist with
non-HALDB databases, but for other reasons. Second, if you process a
secondary index as a database, the I/O area may differ. This depends on the

definition of the secondary index. Processing of the indexed database does not
change. Third, you cannot insert logical children with an application load program
(PROCOPT=L). This would only affect you if you created a new logically related
database. We explain how to handle these changes in Chapter 11, “Application
considerations” on page 183.
No other changes to the external characteristics of the DL/I application

programming interface were introduced for HALDB.
1.6 Logical relationships with HALDB

Logical relationships are supported with HALDB in the same way as they are with
non-HALDB DL/I databases, with a single exception: Bidirectional logical
relationships in a HALDB database must be implemented with physical pairing.
When loading a new partitioned database that contains logical relationships, the
logical child segments may not be loaded as part of the load step. Logical
children are added by normal update processing after the database has been
loaded.
1.7 Partition selection

HALDB assigns database records to partitions based on the key of the root
segment. When retrieving or inserting a database record, HALDB must select the
correct partition. This process is called partition selection. Partition selection
determines where root segments are placed and the order in which partitions are
processed.
Partition selection is based either on the key ranges or a partition selection exit
routine. For batch, BMP, and JBP programs, a PCB can be restricted to access a
single partition.
1.7.1 Partition selection using key ranges

Partition selection can be done using key ranges for each partition. These ranges
are created by assigning a high key to each partition. A root segment is assigned
to the partition with the lowest high key, which is also no higher than the key of
the root, for example, if there are three partitions with high keys of 1000, 2000,
and 3000. Keys of 1001 through 1500 are assigned to the partition with a high
key of 2000. Partitions are ordered in the order of their high keys.
1.7.2 Partition selection using a partition selection exit routine
Installations that prefer to select partitions by some criteria other than key ranges
may use a partition selection exit routine. The routine assigns records to
partitions based on the key of the root segment. It also determines the order in
which sequential processing accesses the partitions. We explain the
requirements and possibilities for these exit routines in 7.2, “Writing partition
selection exit routines” on page 82.
1.8 Database structure

There are three HALDB database types:
򐂰 PHDAM: Partitioned HDAM
򐂰 PHIDAM: Partitioned HIDAM
򐂰 PSINDEX: Partitioned secondary index
The primary index of a PHIDAM database is not separately defined. Its definition
is included with the definition of the database. Each partition of a PHIDAM
database has its own index data set.
Each partition of a PHDAM or PHIDAM has an indirect list data set (ILDS). The
ILDS is used for the self-healing pointer process. This process eliminates the
need to rebuild secondary indexes or resolve logical relationships when
databases are reorganized. We describe the process in 14.2, “The self-healing
pointer process” on page 251. It is only used for logical relationships and
secondary indexes.
HALDB uses direct pointers, such as physical child and physical twin pointers,
within PHDAM and PHIDAM databases. These are the same pointers that are
used with non-HALDB databases.
1.8.1 Partition ID and reorganization number

Each partition in a database is assigned a partition ID when it is defined. The
partition ID is a 2-byte number that is incremented with the definition of each new
partition in a database. Partition IDs are not reused.
Each partition in a PHDAM or PHIDAM has a reorganization number. This

number is incremented when the partition is reorganized.
The partition ID and reorganization number are stored in the first data set of each
partition of the PHDAM and PHIDAM databases. This is the data set in which
root segments are stored. They are stored in the field of the first bit map block
that would otherwise hold the free space element anchor point (FSEAP).

1.8.2 Segment structure
HALDB segments consist of two parts, prefix and data. This is the same as
non-HALDB databases. However, HALDB prefixes have some additional fields.
PHDAM and PHIDAM

The prefix for PHDAM and PHIDAM segments includes an indirect list key (ILK).
The ILK is assigned to a segment when it is created. Its value never changes
during the life of the segment. The ILK is 8 bytes and has the following data:
򐂰 The relative byte address (RBA) of the segment at its creation time (4 bytes)
򐂰 The partition ID at creation time (2 bytes)
򐂰 The partition reorganization number at creation time (2 bytes)
Figure 1-3 shows the format of a HDAM or HIDAM segment and that of a
PHDAM or PHIDAM segment. The ILK is the last part of the prefix in the HALDB
segments. These HALDB segments always include a physical parent pointer in
the prefix of segments other than roots. Some non-HALDB segments also have
physical parent pointers. They are required in non-HALDB segments only when
the segment is the target of a logical relationship or secondary index or the
parent of such a segment. They may also be included by specifying the pointer in
the DBDGEN.
HDAM or HIDAM segment (not a logical child)

SC DB Counters and pointers Data
|-----------------------Prefix----------------------------|
PHDAM or PHIDAM segment (not a logical child)

SC DB Counters and pointers ILK Data
|-----------------------------Prefix----------------------------------|
SC - segment code
DB - delete byte
ILK - indirect list key
Figure 1-3 HDAM, HIDAM, PHDAM, and PHIDAM segments
Figure 1-4 on page 13 shows the format of HDAM, HIDAM, PHDAM, and
PHIDAM logical child segments. HALDB logical child segments have an
extended pointer set (EPS) in their prefix. This replaces the RBA pointer, which
non-HALDB logical child segments have among their pointers. HALDB logical
child segments always have their logical parent’s concatenated key stored in the
data area. This is optional with non-HALDB logical child segments.
HDAM or HIDAM logical child segment

SC DB Counters and pointers Data
|-----------------------Prefix----------------------------|
PHDAM or PHIDAM logical child segment

SC DB Counters and pointers EPS ILK Data including LPCK
|--------------------------------------------Prefix----------------------------------------------|
SC - segment code
DB - delete byte
LPCK - logical parent's concatenated key
EPS - extended pointer set
ILK - indirect list key
Figure 1-4 HDAM, HIDAM, PHDAM, and PHIDAM logical child segments
Secondary indexes
The prefix of non-HALDB secondary index segments has a four-byte RBA when
direct pointers are used. The prefix of HALDB secondary index segments
includes an EPS and the key of the target’s root segment. Figure 1-5 shows the
format non-HALDB (INDEX) secondary index segments and HALDB (PSINDEX)
secondary index segments.
Secondary INDEX segment

DB Pointer Index Data
|-------Prefix---------|
PSINDEX segment
DB EPS Root Key Index Data
|------------------------Prefix----------------------------------|
DB - delete byte
EPS - extended pointer set
Figure 1-5 Secondary index segments

1.8.3 Extended pointer set (EPS)
Each logical child or secondary index segment contains an extended pointer set
(EPS). The EPS contains the pointer for the logical relationship or secondary
index entry. With bidirectional logical relationships, the EPS contains pointers to
the logical parent and to the logical child pair. It replaces the direct pointers or
symbolic pointers that are used with non-HALDB databases. They allow you to
reorganize a database without updating logical relationships or rebuilding
secondary indexes. We explain how EPSs are used in14.2, “The self-healing
pointer process” on page 251. The EPS is 28 bytes and contains the following
fields:
򐂰 Reorganization number of target partition
򐂰 Partition ID of target partition
򐂰 RBA pointer to logical parent or secondary index target
򐂰 RBA pointer to paired logical child for bidirectional logical relationships
򐂰 Indirect list key (ILK) of target segment
򐂰 Database record lock ID of the target
1.8.4 Indirect list data set (ILDS)

The ILDS is a VSAM KSDS data set. It is the repository for indirect pointers.
These pointers eliminate the need for updating logical relationship or secondary
index pointers after a reorganization. An ILDS contains indirect list entries (ILEs),
which are composed of keys and data. The data parts of ILEs contain direct
pointers to the target segments.
Each PHDAM or PHIDAM partition has an ILDS. The ILDS is required even when
there are no secondary indexes or logical relationships. Indirect list entries only
exist for segments involved in inter-record pointing. These are target segments of
secondary indexes or logical relationships. Secondary indexes do not have an
ILDS because they never contain any target segments.
Indirect list entries are created or updated only by reorganization reload. They
are not created or updated by non-reload processing. An initial load program
(PROCOPT=L) does not create any ILEs. After an initial load, the ILDS is empty.
If you insert a target segment during normal processing (PROCOPT=I or A), no
ILE will be created either. ILEs are not needed in these cases because inserts of
target segments create accurate pointers in the EPSs of the secondary index
entries or logical child segments.
The key to the ILE is a 9-byte field consisting of the indirect list key and the
segment code of the target segment. The ILK is also stored in the segment prefix
of the target segment. The next field contains flags (1 byte). ILE data contains an
old and a new copy (20 bytes each) of information needed for the self-healing
pointer process.
These are:
򐂰 Partition ID
򐂰 Reorganization number
򐂰 RBA of logical parent or secondary index target segment
򐂰 Database record lock ID for target segment
򐂰 RBA of paired logical child when physical pairing is used
1.8.5 Number of data sets

The number of data sets per partition depends on the database organization.
򐂰 PSINDEX
A PSINDEX may contain one to 1001 partitions. Each partition contains only
one data set, so a PSINDEX may have one to 1001 data sets.
򐂰 PHDAM
A PHDAM database may contain one to 1001 partitions. Each partition has an
ILDS and one to 10 data sets for data set groups, so a partition has two to 11
data sets and a PHDAM database may have two to 11,011 data sets.
򐂰 PHIDAM
A PHIDAM database may contain one to 1001 partitions. Each partition has
an ILDS, a primary index data set, and 1 to 10 data sets for data set groups,
so a partition has three to 12 data sets and a PHIDAM database may have
three to 12,012 data sets.
The data sets for data set groups are called data data sets. These data sets
contain the segment data for PHDAM and PHIDAM databases.
Figure 1-6 on page 16 shows the data sets for the different types of HALDB
databases.

PHIDAM PHDAM PSINDEX
Index ... ...
1-1001 partitions
ILDS ... ...
... ... up to 10
Data data sets
... ...
1-1001 partitions 1-1001 partitions
Figure 1-6 Data sets in HALDB
The maximum data set size is 4 GB. This applies to both OSAM and VSAM data
sets.
1.8.6 Special considerations for secondary indexes

A secondary index database must be a PSINDEX if the target database is a
HALDB. The secondary index itself may have one or more partitions.
The key ranges for the partitions of a secondary index are not likely to match the
key ranges of its target database. Partition selection for the secondary index is
based on its key. Partition selection for the target database is based on its key.
Usually, these keys are unrelated. In Figure 1-7 on page 17 the target database
is partitioned on last name and the secondary index is partitioned on employee
number. You will notice that both partitions of the secondary index have target
segments in each of the partitions of the target database.
This will be significant for timestamp recoveries of partitions. A timestamp

recovery of a partition in the target database would require a timestamp recovery
of all of the partitions in the secondary index to the same time. This would require
a timestamp recovery of the other partitions in the target database to this same
time. That is, you will not be able to do a timestamp recovery of an individual
partition for most databases that have secondary indexes.
Partition in secondary index may point to multiple

partitions in target database
Secondary Index
000-499 500-999
A-J K-P Q-Z
Target Database
Figure 1-7 Partitioned secondary index
A partition selection exit could be used to partition the secondary index along the
same boundaries of the main database. There has to be an appropriate key,
because the selection of a partition always depends on the root key. The exit has
the possibility to use just parts of the root key to define the correct partition.
Miscellaneous requirements and considerations

There are some other requirements and considerations for secondary indexes:
򐂰 Symbolic pointing is not allowed.
򐂰 Secondary indexes must have unique keys. /SX and /CK fields may be used
to provide uniqueness.
򐂰 Shared secondary indexes are not supported.
򐂰 The reorganization of a target database does not require the rebuilding of its
secondary indexes.
1.9 Naming conventions

HALDB databases may have many partitions and even more data sets. Naming
conventions for DD names and data set names are used to make their
management simpler.

1.9.1 Partition names
Each partition is given a partition name that is one to seven bytes in length.
These names must be unique among the database names, partition names, and
Fast Path area names that are registered in a set of RECONs. A partition name
may be used to describe the data in the partition, but the name should be chosen
carefully. If you add or delete partitions or modify their boundaries, data may be
moved from one partition to another. This may make the assignment of
meaningful names difficult. There is no way to change the name of an existing
partition.
1.9.2 DD names
DD names are constructed by appending a one-byte suffix to the partition name.
The suffix indicates the function of the data set.
For PHIDAM or PHDAM databases, these are:

򐂰 Suffixes A through J are used for data data sets.
This is the same value that is defined in the DSGROUP statement in the DBD.
If there is no DSGROUP specified, the default is A.
򐂰 Suffix L is used for the indirect list data set (ILDS).
򐂰 Suffix X is used for the primary index data set.
For PSINDEX databases, there is only one data set in a partition: Suffix A is used
for the secondary index data set.
Thus, a partition in a PHIDAM database with partition name PART1 would have
DD names of PART1A for its first data data set, PART1B through PART1J for any
successive data data sets, PART1L for its ILDS, and PART1X for its primary
index data set.
A secondary index partition with partition name PARSI would have a DD name of
PARSIA for its data set.
1.9.3 Data set names

Each partition is given a data set name prefix. It may be up to 37 characters. IMS
appends seven characters to this prefix to form the data set name. The
IMS-supplied portion contains a one-character alphabetic suffix to indicate the
function of the data set and a five-digit partition ID number. The appendix of
.A00001 indicates the first data data set in the partition with partition ID 1. The
alphabetic character is the same one used as the suffix for the DD name of the
data set.
There is no required correlation between the partition name and the names of its
data sets. You may assign the same data set name prefix for every partition in a
database. Since partition IDs are unique, data sets in different partitions will have
different data set names.
Some examples of databases that we use later in this publication are the
Customer database (CUSTDB) and the New Order database (NORDDB). In our
example the Customer database is PHDAM and has two partitions, and the New
Order database is PHIDAM and has two partitions. Table 1-1 shows the
corresponding DD and data set names. The example used may not be a practical
one for a real environment, as the partition names and DD names must be
unique within their RECON.
Table 1-1 Customer and order databases - DSN table

Master DBD Partition DD name Partition ID Data set name
name name
CUSTDB CUST1 CUST1A 1 IMSPSA.IM0A.CUSTDB.A00001

CUST1L IMSPSA.IM0A.CUSTDB.L00001
CUSTDB CUST2 CUST2A 2 IMSPSA.IM0A.CUSTDB.A00002

CUST2L IMSPSA.IM0A.CUSTDB.L00002
NORDDB NORD1 NORD1A 1 IMSPSA.IM0A.NORDDB.A00001

NORD1L IMSPSA.IM0A.NORDDB.L00001
NORD1X IMSPSA.IM0A.NORDDB.X00001
NORDDB NORD2 NORD2A 2- IMSPSA.IM0A.NORDDB.A00002




2
Chapter 2. Defining HALDB databases

This chapter gives an overview of the definition tasks for HALDB. The process
you follow to implement a HALDB is similar to implementing a full-function
non-HALDB database.
This chapter describes system definition, buffer pool assignment, definition of

database exit routines, dynamic allocation, and DBDGEN differences for HALDB
databases.

2.1 Overview of HALDB definition
This section briefly describes the steps you need to follow for defining a HALDB
database. A HALDB structure consists of two distinct entities, the master
database structure and its associated partition structures. The master structure
is defined and generated with the DBDGEN process. It is basically unchanged
from non-HALDB methods, so your first step is to design the logical structure of
your database and implement it with the DBDGEN process. This defines the
structure of the master database. A HALDB database contains one or more
partitions. You must determine the number and characteristics of your partitions
and implement them. The definition process for partitions is external to the
DBDGEN process. It uses either the Partition Definition utility (PDU) or DBRC
commands.
After ACBGEN and system definition for the master database, assignment of
data sets to buffer pools also must be done.
Some of the differences between HALDB and non-HALDB definitions are

covered in more detail Chapter 8, “Migration from non-HALDB to HALDB” on
page 89.
2.1.1 Design the logical structure of the database

You need to determine your segment structures, the relationships between
segments, segment layouts, and, if appropriate, logical relationships and
secondary indexes. You follow the same process here as you would for a
non-HALDB database.
2.1.2 Implement the logical structure with the DBDGEN process

The DBDGEN creates the HALDB master DBD.
The master DBD provides the structure for DL/I processing, as well as a model
for creating the partitions.
The logical structure of the database includes the relationship between the
segments, the pointers used, the database access method (PHDAM or
PHIDAM), the segment layouts, the database physical data set organization
(VSAM or OSAM), and, if appropriate, any secondary indexes or logical
relationships. You should also decide whether you need data set groups. One
reason for data set groups is now obsolete. You do not need them to resolve size
problems, but they could still be helpful for resolving potential performance
problems.
Other functions of DBDGEN for non-HALDB databases, such as defining the
physical characteristics of the data sets and their DDNAMEs, are not performed
for HALDB. You have to use other tasks to define them.
2.1.3 Determine the partitioning scheme

You must determine your partitioning scheme. The questions you need to answer
here include:
򐂰 How many partitions do you need in your database?
򐂰 How much DASD storage space is required for each partition?
򐂰 How much free space do you need in your database? Do the free space
parameters need to be different for each partition?
򐂰 What key is to be used to determine the split between partitions?
򐂰 Do you need a partition selection exit?
򐂰 For PHDAM, do you want all of your partitions to use the same randomizing
parameters, or do you want to have randomizing parameters specific for each
partition?
򐂰 Do you want to automatically or manually define your partitions? This decision
influences the names of the partitions. You cannot change these names
afterwards.
The HALDB Migration Aid utility (DFSMAID0) may help you in determining the
answers to some of these questions.
You have to decide how many partitions you want. Usually this is determined by
two considerations: The size of the database and the time required for the
reorganization of the database. Partitions may be reorganized independently and
in parallel. Since smaller partitions may be reorganized faster, defining many
small partitions allows you to reorganize the database in a shorter time.
2.1.4 Create the partitioning scheme

HALDB partition definitions are stored in DBRC RECONs. When defining
partitions, you must have update authority for the RECON data sets.
From DBRC’s point of view, a HALDB is a master database consisting of partition

databases.
Chapter 2. Defining HALDB databases 23

You have two ways of defining partitions:
򐂰 Partition Definition utility (PDU)
This utility is invoked under ISPF. It reads the DBDLIB and registers the
database. You can use it only after having performed a DBDGEN.
There are panels in which you can specify the physical characteristics, such
as high keys, randomizer values, and block sizes, as well as some DBRC
options, such as skeletal JCL members and maximum number of image
copies. Parameters for which default values were defined in DBDGEN are
included on some panels. The output of the definition is placed in DBRC’s
RECON data sets.
For more information about the PDU see Chapter 4, “Partition Definition
utility” on page 45.
򐂰 DBRC batch commands
You may issue the INIT.DB and INIT.PART commands to register the master
database and define its partitions.
For more information about DBRC batch commands see Chapter 5, “Batch
definition of HALDB” on page 61.
When using key range partitioning we recommend that you specify a value of
x’FF’ for the last partition. This ensures that all keys may be assigned to a
partition. If the last partition (the partition with the highest key specified) has a
key value other than x’FF’s, any attempt to access or insert a database record
with a key higher than the high key specified results in an FM status code for the
call. Application programs written for non-HALDB databases are unlikely to
expect this status code.
2.1.5 Database exit routines

With the exception of PHDAM randomizing routines, database exit routines are
defined for the whole database, not just for a partition. These exit routines are:
򐂰 Data capture exit routine
򐂰 Data conversion user exit routine
򐂰 secondary index database maintenance exit routine
򐂰 Segment edit/compression exit routine
These exit routines are not passed any information about partitions. Their actions
cannot be tailored for individual partitions.
2.1.6 System definition
The DATABASE macro statement is used to define physical databases for an
IMS online system. One DATABASE macro must be specified for each PHDAM,
PHIDAM, and PSINDEX database. Unlike HIDAM primary indexes, PHIDAM
primary indexes are not defined explicitly and DATABASE macros are not used in
the system definition. Instead, they are implicitly defined by the DATABASE
macro for the PHIDAM database.
2.1.7 Buffer pools

HALDB database data sets are assigned to buffer subpools with DFSVSMxx
members for online systems and the DFSVSAMP DD data sets for batch jobs
and utilities. This is the same method that is used for non-HALDB databases.
Buffer subpools may be used for both HALDB and non-HALDB data sets. These
subpools are defined with IOBF statements for OSAM and VSRBF statements
for VSAM. They are not changed for HALDB support.
DBD statements are used to assign data sets to specific subpools. Example 2-1
shows the DBD statement format. The dbdname value is used to specify a
database or partition. For non-HALDB data sets, the data set identifier is a
number. This is the data set number within the database specified by the
dbdname value. For HALDB data sets, the data set identifier is a letter. This is the
letter associated with the type of data set. It is the same letter that is used for the
suffix in the data set’s DDNAME. A through J are used for data set groups. X is
used for PHIDAM primary indexes. L is used for ILDSs. A is used for secondary
indexes. The id value identifies the database subpool to which the data set is
assigned.
If a master database name is specified on a DBD statement, it applies to all

partitions in the database. If a partition name is specified, it applies only to that
partition. Statements specifying a partition name override any statements
specifying the master database name of the partition.
Example 2-1 DBD control statement for buffer pool assignment

DBD=dbdname(data set identifier,id)
Buffer pools for ILDSs are always required when processing the PHDAM and
PHIDAM databases. This is true even when the ILDSs have no entries in them.

2.1.8 Dynamic allocation
Dynamic allocation of HALDB data sets uses DBRC RECON information. Since
all HALDB databases are registered in the RECONs, this information always
exists. DFSMDA members are not used with HALDB data sets.
Dynamic allocation for HALDB data sets cannot be overridden. DD statements

should never be used for HALDB data sets. If they are present and they do not
match with the information in the RECON, the opening of the data sets fails.
The NODYNALLOC statement in DFSVSAMP specifications for batch jobs is

ignored for HALDB data sets. For non-HALDB data sets it is used to disable
dynamic allocation.
2.2 DBDGEN differences for HALDB

The database definition statements for HALDB databases are slightly different
from non-HALDB databases. The four main things that are different in HALDB
definitions are:
򐂰 ACCESS operands PHDAM, PHIDAM, and PSINDEX are used.
򐂰 DDNAME and data set information are not part of DBDGEN for HALDB. The
database definition is purely for defining the hierarchical structure and
relationships of the data. DATASET statements are not allowed for HALDB.
򐂰 The primary index of a PHIDAM database is automatically included in the
generation of the PHIDAM database. DBDGEN cannot be performed for the
primary index.
򐂰 Since DATASET statements are not allowed in HALDB database definitions,
another method is used for defining multiple data set groups. They are
defined with the DSGROUP parameter on SEGM statements. This parameter
is not used with non-HALDB databases.
DBDGEN does not define individual partitions. This is done using the HALDB
Partition Definition utility or DBRC batch commands. See Chapter 4, “Partition
Definition utility” on page 45, and Chapter 5, “Batch definition of HALDB” on
page 61.
Definitions specified in the DBDGEN source will generally apply to all partitions.
The possible exception is the RMNAME parameter on the DBD statement.
RMNAME must be specified in the DBD for PHDAM databases, but it may be
overridden. Overrides may specify different values for different partitions.
The PSNAME parameter on the DBD statement specifies the name of a partition
selection exit routine. It may be overridden by the PDU or DBRC commands. The
override may specify a different name or may specify that key range partitioning
is to be used. Similarly, if the PSNAME parameter is not specified in the DBD, it
may be specified by the PDU or DBRC commands.
Note: Partition selection is done prior to invoking the randomizing module.

Randomizing modules only select locations within partitions.
Table 2-1 shows the parameters that are changed or added to DBDGEN
statements.
Table 2-1 DBDGEN changes

Statement Keyword Description of change
DBD NAME No change.
ACCESS Add operands PHDAM, PHIDAM, and PSINDEX.
INDEX operand is not used for primary index of PHIDAM.
Other operands are unchanged.
RMNAME (mod,anch,rbn,bytes) - Default values for partitions. RBN is high

RBN per partition.
PSNAME Optional parameter. The partition selection exit module name is

(new) the operand.
A user provided partition selection routine is not used with key
range partitioning. PSNAME may be overridden by PDU or
DBRC commands.
AREA/ Not allowed for HALDB.

DATASET Partitions are defined outside of DBDGEN.

SEGM PARENT VIRTUAL/PHYSICAL: Either may be specified, but only

physical used. That is, either specification results in the logical
parent’s concatenated key being physically stored in the logical
child segment.
POINTER HIER and HIERBWD are not allowed. TWIN, TWINBWD, or

NOTWIN must be specified PHDAM and PHIDAM databases.
For root segments in PHIDAM databases, only NOTWIN or
TWINBWD are allowed.
All dependent segments have a physical parent pointer.
If the segment is a logical child, PAIRED and LPARNT are

assumed. LTWIN and LTWINBWD are not allowed.
SOURCE Not allowed for PHDAM or PHIDAM. Only physical pairing is

supported for bidirectional logical relationships.
DSGROUP This keyword is used to specify multiple data set groups for
(new) partitioned databases. The format is DSGROUP=c, where c is
the letters A through J. Up to 10 data set groups can be
specified. The DSGROUP value corresponds to the values
used in the DDNAME and data set naming conventions for
HALDB.
The default for every segment is A (first data set in each

partition). If specified on the root segment, it must be
DSGROUP=A.
Gaps in the A–J sequence are not allowed. If DSGROUP=C is

specified on a SEGM statement, there must also be at least one
SEGM statement with DSGROUP=B and one with A.
LCHILD LCHILD statements are not used for the primary index of a
PHIDAM database. These indexes are automatically generated
by PHIDAM definitions.
POINTER PTR=INDX is required for LCHILD statement following the

target segment in indexed PHDAM or PHIDAM DBD.
PTR=SNGL is required for the LCHILD statement in PSINDEX
DBD. It is the default value for this LCHILD statement
PTR=NONE is required for the LCHILD statement following a
logical parent segment. It is the default for this LCHILD
statement.
PAIR Unchanged from non-HALDB. HALDB does not support virtual

pairing, so paired segment is physically paired.
INDEX Not needed for a primary index of PHIDAM, otherwise

unchanged.
RKSIZE For PSINDEX only, specifies the length of the root segment key
(new) of the target database.
RULES Not used: It does not apply to physical pairing.
FIELD Unchanged.
Note: /CK means the same as with non-HALDB. /SX creates an
8-byte ILK instead of 4-byte RBA.
XDFLD CONST For shared secondary indexes. Not allowed for HALDB.
We provide examples of the needed DBD changes later in this publication.

Example 9-8 on page 138 shows the DBD source for a HIDAM database, and
Example 9-9 on page 139 shows the DBD source after the database has
converted to a PHIDAM.

3
Chapter 3. HALDB and DBRC

All HALDB databases are registered in the DBRC RECONs. Dynamic allocation
always uses RECON information. In this chapter we explain the RECON records
used for HALDB and the DBRC commands used to add, delete, modify, and list
HALDB information. We provide sample listings of HALDB information from the
RECONs.

3.1 RECON records for HALDB
HALDB databases are described in four types of RECON records. These records
are:
򐂰 Master database record
򐂰 Partition database record
򐂰 Partition record
򐂰 Partition database data set record
Each HALDB database has one master database record. There is a partition
database (DB) record and a partition (PART) record for each partition. There is a
partition database data set (DBDS) record for each database data set. We
describe these records and their uses in the following sections.
3.1.1 Master database record

Each HALDB database has a master database record. This is a new type of
database record. There is a flag in the record to indicate that it is a HALDB
master database record. This record is created either by the HALDB Partition
Definition utility or an INIT.DB command that specifies TYPHALDB.
The record stores information that applies to the entire database. Among the
data are:
򐂰 Partition selection exit routine name if one is used
򐂰 Number of partitions defined
򐂰 Change version number
򐂰 Highest allocated partition ID number
򐂰 Data set group count
򐂰 Global DMB number
򐂰 Share level
򐂰 RSR global service group and tracking level
򐂰 Recoverable or not recoverable
Authorization information is not stored in this record.
The record may be mapped with the DSPDBHRC macro in ADFSMAC and
SDFSMAC.
The key of the record is composed of:

򐂰 HALDB database name (bytes 1–8)
򐂰 Zeros (bytes 9–16)
򐂰 x’18’ (byte 17)
Example 3-1 on page 39 shows a listing of a master database record.
3.1.2 Partition database record

Each HALDB partition has a partition database record. This is a new type of
database record. There is a flag in the record to indicate that it is a HALDB
partition database record. This record is created either by the HALDB Partition
Definition utility when a partition is defined or by an INIT.PART command.
The record stores information that applies to the partition. Among the data are:
򐂰 Master database name
򐂰 Change version number for the partition
򐂰 Partition init needed flag
򐂰 Authorization flags including prohibit further authorization and read only
򐂰 Global DMB number, which is copied from the master database record
򐂰 Share level, which is copied from the master database record
򐂰 RSR global service group and tracking level, which are copied from the
master database record
򐂰 Recoverable or not recoverable, which is copied from the master database
record
Authorization information is stored in this record. The record may be mapped

with the DSPDBHRC macro in ADFSMAC and SDFSMAC.

򐂰 Partition name (bytes 1–8)
򐂰 x’18’ (byte 17)
Example 3-2 on page 39 shows a listing for a partition.
3.1.3 Partition record

Each HALDB partition has a partition record. This is a new type of RECON
record. This record is created either by the HALDB Partition Definition utility
when a partition is defined, or by an INIT.PART command. It is created when the
partition database record is created.
The record stores information that applies to the partition. Among the data are:
򐂰 Data set name prefix
Chapter 3. HALDB and DBRC 33

򐂰 Partition ID
򐂰 Partition high key or partition string
򐂰 Length of partition high key or partition string
򐂰 Free space information
򐂰 Data set group count
򐂰 Partition init needed flag (PINIT flag)
򐂰 Previous and next partition names if using key range partitioning
򐂰 Randomization specifications for PHDAM
򐂰 Data set group block sizes for OSAM
The record may be mapped with the DSPPTNRC macro in ADFSMAC and
SDFSMAC.

򐂰 HALDB master database name (bytes 1–8)
򐂰 x’19’ (byte 17)
3.1.4 Partition DBDS record

Each partition data set has a partition DBDS record. This is a new type of DBDS
record. This record is created either by the HALDB Partition Definition utility
when a partition is defined or by an INIT.PART command. It is created when the
partition database data sets are defined. There are records for all partition data
sets including ILDSs and PHIDAM primary index data sets.
The record stores information that applies to the partition data set. Among the
data are:
򐂰 Data set name
򐂰 Skeletal JCL members
򐂰 Image copy information
򐂰 Change accumulation information
򐂰 Flags and counters for image copy needed, recovery needed, and so on
The record may be mapped with the DSPDSHRC macro in ADFSMAC and
SDFSMAC.

򐂰 DDName (bytes 9–16)
򐂰 x’20’ (byte 17)
3.2 Dynamic allocation
The information in the RECONs is used for dynamic allocation of HALDB
database data sets. The data set names and DD names from partition DBDS
records are used to provide this information. DFSMDA members are never used
with HALDB.
3.3 Authorization processing

Authorization for HALDB is done by partition, not database. The share level is set
for an entire database, but individual partitions are authorized. Authorization
occurs when a partition is first used. For example, a batch program that reads a
database sequentially will authorize partitions one at a time. This occurs when
each partition is first accessed. Authorization information is kept in the partition
database records for each partition. The master database record contains the
share level for the database but does not have information about any
authorizations.
Since authorizations are by partition, not an entire database, different partitions

may be authorized to different subsystems or utilities at the same time. This
allows you to reorganize partitions in parallel as explained in Chapter 14,
“HALDB reorganization” on page 243. It also allows you to execute non-data
sharing batch programs against different partitions of the same database, as
explained in 11.2, “Processing partitions in parallel” on page 189.
3.4 Partition initialization

Partitions must be initialized before they may be loaded. This is explained in
Chapter 6, “Partition initialization” on page 73. The initialization required state for
each partition is kept in the partition database record and the partition record for
each partition. The state may be turned off or on with the CHANGE.DB
command. This is explained in 6.2.1, “Partition initialization required (PINIT) flag”
on page 75.
3.5 DBRC commands

You may use DBRC commands to define, delete, or modify information about
HALDB databases and partitions. These commands are an alternative to the
HALDB Partition Definition utility. You also may use DBRC commands to list
RECON information about HALDB databases and partitions and to generate JCL
for use with HALDB databases and partitions.

3.5.1 INIT commands
You may use INIT commands to define a HALDB database and its partitions.
INIT.DB
The INIT.DB command can be used to register the master database with DBRC.
It defines the DBD name of the master database, its sharing level, the type of
database, whether this database is partitioned by key ranges or a selection exit,
and whether it is recoverable.
INIT.PART
The INIT.PART command can be used to register a partition after the master
database is already registered with DBRC. It creates the HALDB partition
structure, that is a partition record, a partition database record, and one or more
partition DBDS records according to the DBD specification.
We explain the use of INIT.DB and INIT.PART commands in 5.2, “DBRC

initialization commands for HALDB” on page 62. INIT.DBDS commands are not
used for registering HALDB database data sets.
3.5.2 CHANGE commands

You may use the CHANGE.DB, CHANGE.PART, and CHANGE.DBDS
commands to modify information about HALDB databases, partitions, and data
sets.
CHANGE.DB
The CHANGE.DB command may specify either a HALDB master database or
partition.
The master database name may be used to change the following:

򐂰 Share level
򐂰 Recoverable attribute
򐂰 Partition selection exit usage or routine name
򐂰 Image-copied required attribute for non-recoverable databases
򐂰 RSR specifications
򐂰 Partition initialization required status
When the master database name is used, the command applies to all partitions
and data sets in the database.
A partition name may be used to change the following:

򐂰 The prohibit further authorization flag
򐂰 The backout needed flag
򐂰 The read only authorization flag
򐂰 The authorization of the partition to a subsystem
򐂰 The partition init needed (PINIT) flag
When the partition name is used, the command applies only to the partition.
CHANGE.PART
You may use the CHANGE.PART command to change partition definitions. Both
the master database name and the partition name are specified in the command.
The master database name and partition name cannot be changed with this
command. The command may be used to change the following attributes for a
partition:
򐂰 The high key value when using key range partitioning
򐂰 The partition string when using a partition selection exit routine
򐂰 The data set name prefix
򐂰 The PHDAM randomization parameters
򐂰 The free space specifications
򐂰 The OSAM block sizes
򐂰 The GENMAX value
򐂰 The RECOVPD value
򐂰 The image copy REUSE attribute
򐂰 The skeletal JCL member names
CHANGE.DBDS
You may use the CHANGE.DBDS command to change information about data
sets in a HALDB partition. Specify the partition name, not the master database
name, in the DBD parameter. The command may not be used to change data set
names or DD names. These are determined by the HALDB naming conventions.
The command may be used to change the following attributes in data sets other
than ILDSs and PHIDAM primary indexes:
򐂰 The GENMAX value
򐂰 The RECOVPD value
򐂰 The image copy REUSE attribute
򐂰 The skeletal JCL member names
򐂰 The image copy needed flag
򐂰 The recovery needed flag
򐂰 Error queue elements (EQEs)
For ILDSs and PHIDAM primary indexes the only attributes that may be changed
are:
򐂰 The recovery needed flag
򐂰 Error queue elements (EQEs)

We explain the use of CHANGE.DB, CHANGE.PART, and CHANGE.DBDS
commands in 5.3, “DBRC change commands for HALDB” on page 68.
3.5.3 DELETE commands

You may use DELETE.DB and DELETE.PART commands to delete HALDB
databases and partitions from the RECONs. DELETE.DBDS commands are not
allowed for HALDB database data sets.
DELETE.DB
You may use the DELETE.DB command to delete a HALDB master database, all
of its partitions, and associated allocation, image copy, recovery, and
reorganization records from the RECONs. You cannot specify a partition name
on the DELETE.DB command. Use the DELETE.PART command to delete a
partition from the RECONs.
DELETE.PART
You may use the DELETE.PART command to delete a partition and associated
allocation, image copy, recovery, reorganization, and DBDS records from the
RECONs.
We explain the use of DELETE.DB and DELETE.PART commands in 5.4,

“DBRC delete commands for HALDB” on page 71. DELETE.DBDS commands
are not used to delete HALDB database data sets.
3.5.4 LIST commands

You may use LIST.DB, LIST.DBDS, and LIST.HISTORY commands to list
information about HALDB databases, partitions, and data sets from the
RECONs.
LIST.DB
You may use the LIST.DB command to list HALDB database, partition, and data
set information. You may specify the TYPHALDB parameter to list information
about all HALDB databases. Alternatively, you may specify either a HALDB
master database name or a partition name in the DBD parameter of the
command. If you use the master database name, information from all partitions is
listed.
LIST.DBDS
You may use the LIST.DBDS command to list information about HALDB
database data sets. You may specify either a master database name or a
partition name in the DBD parameter. If you specify a DDN parameter, you must
specify the data sets’ partition name in the DBD parameter.
Sample listings
Example 3-1 shows a listing of a master database record.
Example 3-1 LIST.DB output for master database

DB
DBD=CUSTDB DMB#=31 CHANGE#=5 TYPE=HALDB
SHARE LEVEL=3 GSGNAME=**NULL**
PSNAME=**NULL** DBORG=PHDAM DSORG=OSAM CURRENT PARTITION ID=00004
FLAGS: COUNTERS:
RECOVERABLE =YES PARTITIONS =4
DATA SET GROUP MEMBERS =1
Example 3-2 shows a listing for a partition. The partition high key record is listed
twice. The second listing is in hexadecimal. Even though the key is only 10 bytes,
it is listed as if it were 32 bytes. The last 22 bytes are a conversion of blanks to
hexadecimal. The hexadecimal listings of partition high keys or partition strings
are always done so that remaining bytes on a line are listed as X’40’.
Example 3-2 LIST.DB output for a partition

DB
DBD=CUSTDB1 MASTER DB=CUSTDB CHANGE#=2 TYPE=PART
USID=0000000001 AUTHORIZED USID=0000000000 HARD USID=0000000000
RECEIVE USID=0000000000 RECEIVE NEEDED USID=0000000000
DBRCVGRP=**NULL**
DSN PREFIX=IMSPSA.IM0A.CUSTDB PARTITION ID=00001
PREVIOUS PARTITION=**NULL** NEXT PARTITION=CUSTDB2
OLRIMSID=**NULL** ACTIVE DBDS=A-J M-V EXIST=NO
RANDOMIZER:
NAME=DFSHDC40 ANCHOR=11 HIGH BLOCK#=15000 BYTES=1
FREE SPACE:
FREE BLOCK FREQ FACTOR=0 FREE SPACE PERCENTAGE=0
PARTITION HIGH KEY/STRING (CHAR): (LENGTH=10 )

0005999999
PARTITION HIGH KEY/STRING (HEX):
F0F0F0F5F9F9F9F9F9F940404040404040404040404040404040404040404040
OSAM BLOCK SIZE:

A = 8192
FLAGS: COUNTERS:
BACKOUT NEEDED =OFF RECOVERY NEEDED COUNT =0

READ ONLY =OFF IMAGE COPY NEEDED COUNT =0
PROHIBIT AUTHORIZATION=OFF AUTHORIZED SUBSYSTEMS =0
HELD AUTHORIZATION STATE=0
EEQE COUNT =0
TRACKING SUSPENDED =NO RECEIVE REQUIRED COUNT =0
OFR REQUIRED =NO
PARTITION INIT NEEDED =NO
ONLINE REORG ACTIVE =NO
PARTITION DISABLED =NO
Example 3-3 shows a listing of a HALDB database data set. ALLOC and REORG
information is not shown but would appear for most data sets. Even though the
database is PHDAM, the DBORG value is listed as HDAM. The DBORG value for
PHIDAM database data sets is listed as HIDAM. The DBORG value for ILDS,
PHIDAM primary index, and secondary index data sets is always INDEX.
Example 3-3 LIST.DB or LIST.DBDS output for a HALDB database data set
DBDS
DSN=IMSPSA.IM0A.CUSTDB.A00001 TYPE=PART
DBD=CUSTDB1 DDN=CUSTDB1A DSID=001 DBORG=HDAM DSORG=OSAM
CAGRP=**NULL** GENMAX=2 IC AVAIL=0 IC USED=1 DSSN=00000000
NOREUSE RECOVPD=0 OTHER DDN=**NULL**
DEFLTJCL=**NULL** ICJCL=ICJCL OICJCL=OICJCL RECOVJCL=RECOVJCL
RECVJCL=RECVJCL
FLAGS: COUNTERS:
IC NEEDED =OFF
RECOV NEEDED =OFF
RECEIVE NEEDED =OFF EEQE COUNT =0
--------------------------------------------------------------------------
IMAGE
RUN = 2003.073 16:26:23.3 -05:00 * RECORD COUNT =15000
STOP = 0000.000 00:00:00.0 +00:00 BATCH USID=0000000001
IC1
DSN=IMSPSA.IC1.CUSTDB1.CUSTDB1A.D03073.T143654 FILE SEQ=0001
UNIT=3390 VOLS DEF=0001 VOLS USED=0001
VOLSER=TOTIMM
3.5.5 GENJCL commands

You may use GENJCL commands to generate JCL for HALDB databases,
partitions, and data sets. The commands that are used with non-HALDB
databases are used with HALDB. You may specify either a HALDB master
database or a partition name in the DBD parameter. If you specify a DDN
parameter, you must specify a partition name in the DBD parameter.
All GENJCL commands, other than GENJCL.USER, process data for DDNAMEs
ending A through J. They do not process data for ILDSs (L) and PHIDAM primary
indexes (X). ILDSs and PHIDAM primary indexes do not use standard recovery
techniques. There are no image copies for them. They have no database change
after image log records. They are not included in Change Accumulation
processing. Thus, they are not applicable to GENJCL.IC, GENJCL.OIC,
GENJCL.RECOV, GENJCL.CA, and GENJCL.RECEIVE processing.
GENJCL.USER
You may use GENJCL.USER to process any HALDB database, partition, or data
set. This includes ILDSs and PHIDAM primary indexes.
Generating Index/ILDS Rebuild utility JCL

IMS supplies a sample skeletal JCL member that you may use to generate JCL
for the Index/ILDS Rebuild utility (DFSPREC0). The sample member name is
DSPUPJCL in ADFSISRC and SDFSISRC. DBRC installation places skeletal
JCL members into an IMS procedure library (PROCLIB). When using
DSPUPJCL you specify the master database name, the partition name, and the
type of recovery as user keys. The format of the command is shown in
Figure 3-1.
GENJCL.USER MEMBER(DSPUPJCL)
USERKEYS((%MDBNAME,'masterdbname'),
(%DBNAME,'partitionname'),
(%RCVTYP,'ILE | INDEX | BOTH'))
Figure 3-1 GENJCL.USER command format for DFSPREC0
Example 3-4 shows an example of a GENJCL.USER command. It will generate

the INDEX/ILDS Rebuild utility JCL for the recovery of the PHIDAM index of
partition NORDDB2 in database NORDDB.
Example 3-4 GENJCL.USER command to rebuild the index of partition NORDDB2

GENJCL.USER MEMBER(DSPUPJCL) -
USERKEYS((%MDBNAME,’NORDDB’),(%DBNAME,’NORDDB2’),(%RCVTYP,’INDEX’))

Symbolic keywords
IMS supplies symbolic keywords and symbolic keyword values for use with
HALDB. The %mdbname symbolic keyword is used for the master database
name. The %dbname symbolic keyword is used for the partition name. The
%dbtype symbolic keyword has the following values for HALDB database data
sets:
PDATA Data set types A through J (data DBDS and PSINDEX)
PILDS Data set type L (ILDS)
PINDEX Data set type X (PHIDAM primary index)
3.6 DBRC groups for HALDB

HALDB databases, partitions, and data sets may be members of DBRC groups.
3.6.1 Change accumulation groups

When defining a change accumulation group, you must specify the partition
name and DDNAME for HALDB database data sets. ILDSs and PHIDAM primary
indexes cannot be included in a change accumulation group.
3.6.2 Database data set groups

When defining a DBDS group, you must specify the partition name and
DDNAME for HALDB database data sets. ILDSs and PHIDAM primary indexes
may be included in a DBDS group.
3.6.3 Database groups

When defining a database group, you may specify master database names,
partition names, or both.
3.6.4 Recovery groups

When defining a recovery group, you must specify a master database name. You
cannot specify a partition name. When a database belongs to a recovery group,
all of its partitions belong to the group.
3.7 Use of database names in DBRC commands
Many DBRC commands include the specification of a database name. Some of
these commands require the use of a master database name. Some require the
use of a partition name. Some allow either to be used.
3.7.1 Commands that require a master database name

The following commands require that the specified database name be a master
database:
򐂰 INIT.DB
򐂰 DELETE.DB
򐂰 CHANGE.DB with the SHARELVL, NONRECOV, RECOVABL, PARTSEL,
HIKEY, ICREQ, NOICREQ, GSGNAME, NOTCOVER, RCVTRACK, or
DBTRACK parameters
򐂰 INIT.DBDSGRP with the RECOVGRP parameter
򐂰 CHANGE.DBDSGRP with the ADDRECOV or DELRECOV parameter
3.7.2 Commands that require a partition name

The following commands require that the specified database name be a partition
name:
򐂰 CHANGE.DBDS
򐂰 CHANGE.DB with the AUTH, NOAUTH, BACKOUT, NOBACK, READON,
READOFF, or UNAUTH parameters
򐂰 LIST.DBDS with the DDN parameter
򐂰 LIST.HISTORY with the DDN parameter
򐂰 INIT.CAGRP
򐂰 CHANGE.CAGRP
򐂰 INIT.DBDSGRP with the MEMBERS parameter
򐂰 CHANGE.DBDSGRP with the ADDMEM or DELMEM parameter
򐂰 INIT.IC
򐂰 CHANGE.IC
򐂰 CHANGE.UIC
򐂰 NOTIFY.IC
򐂰 NOTIFY.UIC

򐂰 DELETE.IC
򐂰 DELETE.UIC
򐂰 NOTIFY.REORG
򐂰 DELETE.REORG
򐂰 NOTIFY.RECOV
򐂰 DELETE.RECOV
򐂰 NOTIFY.BKOUT
򐂰 CHANGE.BKOUT
򐂰 NOTIFY.ALLOC
򐂰 DELETE.ALLOC
3.7.3 Commands that allow a master database or a partition name

The following commands allow the specified database name to be either a
master database name or a partition name:
򐂰 CHANGE.DB with PINIT or NOPINIT parameters
򐂰 LIST.DB
򐂰 LIST.DBDS without the DDN parameter
򐂰 LIST.HISTORY without the DDN parameter
򐂰 GENJCL.xxx
򐂰 INIT.DBDSGRP with the DBGRP parameter
򐂰 CHANGE.DBDSGRP with the ADDDB or DELDB parameter
򐂰 LIST.DBDSGRP ALL
3.7.4 DBRC commands that are not allowed with HALDB

The following commands are not allowed with HALDB:
򐂰 INIT.DBDS
򐂰 DELETE.DBDS
4
Chapter 4. Partition Definition utility

This chapter describes how you can define, change, and delete your HALDB
databases and partitions after you have performed DBDGEN by using the
Partition Definition utility.
This chapter provides an overview of the utility. It includes security

considerations. It also covers the configuration of the utility, the selection of
databases, and the definition of their partitions. You may choose between two
methods of defining partitions, automatic or manual. We also explain how to
change and delete these definitions.
You can use both the HALDB Partition Definition utility and commands to
manage your HALDBs. Chapter 5, “Batch definition of HALDB” on page 61,
provides information about the alternative DBRC commands that may be used
for definitions.
Refer to manuals IMS Version 8: Administration Guide: Database Manager,

SC27-1283, and IMS Version 8: Utilities Reference: Database and Transaction
Manager, SC27-1308, for more information on using the PDU.

4.1 Using the PDU
The HALDB Partition Definition utility (PDU) is an ISPF application that allows
the database administrator to register the HALDB master database and to
manage its partitions.
The utility is accessed by logging on to TSO and starting ISPF after the dialog
data sets have been made available to the TSO user. You can add the data sets
to a LOGON procedure or use TSO commands to allocate them. You can use the
TSOLIB command to add data sets to the STEPLIB. Table 4-1 shows which file
names and data sets need to be allocated. Be sure to use your own high level
qualifiers.
Table 4-1 File names and data sets to allocate

File DD name Sample data set name Disposition
STEPLIB IVPEXE81.SDFSRESL N/A
SYSPROC IVPEXE81.SDFSEXEC SHR
ISPMLIB IVPEXE81.SDFSMLIB SHR
ISPPLIB IVPEXE81.SDFSPLIB SHR
ISPTLIB IVPEXE81.SDFSTLIB SHR
IMS IVPEXE81.DBDLIB SHR
If you use a logon procedure, you must log on again and specify logon with the
new procedure. If you use allocation commands, they must be issued outside of
ISPF. After you allocate the data sets and restart ISPF, restart the Install/IVP
dialog, return to this task description, and continue with the remaining steps.
Start the HALDB Partition Definition utility from the ISPF command line by
issuing the following command:
TSO %DFSHALDB
The utility consists of several panels and programs that perform various actions
on the database and its partitions. The PDU updates the RECON data sets. You
must be careful to properly back up your installation’s RECONs so you do not
lose the information about your HALDB databases. As all changes are made
directly to the RECON and not held anywhere else it is important to understand
that the changes take effect immediately. There is no record kept of any changes
made. It is important that you protect critical information.
You specify the database and the type of action to perform on the first panel.
Actions include define, modify, and view. The succeeding panels guide you
through the processes.
Note: The PDU has minimal impact on online IMS subsystems with regard to
RECON contention. The RECON is only reserved for the time it takes to
process a DBRC request. It is not held for the duration of the utility execution.
You must have control authority on the RECON data sets to be able to use the
PDU. IMS Version 8 introduces additional security support for DBRC commands
when entered using the DBRC utility or using PDU. Commands may be secured
at the command verb, resource type, and resource name level. Table 4-2 shows
the equivalent DBRC commands that need to be protected. For example, the
HALDB QUERY command is protected by the LIST.DB master database
command.
Table 4-2 DBRC command authorization with IMS Version 8

HALDB request Master or partition Equivalent DBRC command
Query Master LIST.DB DBD(master db)
Set Master INIT.DB DBD(master db)
Set Partition INIT.PART DBD(master db)
Change Master CHANGE.DB DBD(master db)
Change Partition CHANGE.PART DBD(master db)
Delete Master DELETE.DB DBD(master db)
Delete Partition DELETE.PART DBD(master db)
4.1.1 Configuring the PDU

The first time you use the PDU, you must configure it for you. Configuration
means the setting of DBDLIBs and RECONs you want to use. You may have
several configurations in the PDU with different data sets. Figure 4-1 on page 48
shows the initial PDU panel.
Chapter 4. Partition Definition utility 47

Figure 4-1 PDU primary panel
Type a 7 in the Option field to indicate that you want to do configuration, and
press Enter. Figure 4-2 shows the Configurations panel.
Figure 4-2 PDU configuration selection panel
You must type a slash (/) character in the Act field of the blank line to reach the
Configuration Details panel, as shown in Figure 4-3 on page 49. On this panel,
enter the configuration name with a brief description. We recommend that you
specify a meaningful name, so you can easily associate the RECON data sets
used in this configuration. Enter the appropriate RECON and DBDLIB data sets
(up to 10) for this configuration and press Enter.
Figure 4-3 PDU Configuration Details panel
Upon returning to the PDU main panel, your new configuration becomes your
default configuration. When you are maintaining multiple configurations in the
PDU, you can change your active configuration at any time by entering the
Configurations selection panel and typing a slash (/) in the Act field of your
desired configuration.
If you choose Option 6 you may see the DBDLIB allocated, and if you enter ONLY
RECON in the command line, the output shows you the allocated RECON data
sets. Typing ONLY shows you all allocated data sets.
4.1.2 Selecting a database

Prior to defining your partitions, you must run the DBDGEN utility to define the
master database. The PDU does not allow you to proceed until the DBDGEN has
been run.
After you select an active configuration, you need to select a database. In the
main panel type a 1 in the Option field.

Figure 4-4 PDU database selection
You may either select a database directly from the main panel (as shown in
Figure 4-4), or enter a filter in the Database name field. Filter characters are
standard ISPF filter characters: Percent (%) for matching single characters and
asterisk (*) for matching multiple characters. A single asterisk presents the entire
list of DBDLIB members (Figure 4-5).
If you use a filter in the Database name field, you will be presented with a
standard ISPF library listing. You may select a HALDB database from the list by
typing in a slash (/) in the selection field.
Figure 4-5 Database selection list
You cannot select a non-HALDB database from this list. Attempting to do so

takes you back to the main panel.
4.1.3 Setting HALDB master DBD parameters
After making a selection, you can start defining the information that pertains to
the entire HALDB master database. This is performed in the Partitioned
Database Information panel, as shown in Figure 4-6 on page 52.
We can see that this database has been defined as a PHIDAM database with
one data set group and no partition selection exit. The defaults of no RSR
parameters, share level 0, and recoverable are also shown. You may change
most parameters. You cannot change the database organization or the number
of data set groups. If you change a value that is also in the DBD, such as the
partition selection exit routine name, it is not updated in the DBDLIB. If you do so,
your DBD could have definitions for your database that differ from those actually
being used. The values in PDU override those in DBD.
The message about database information for NORDDB means that it is not yet
registered in the RECONs. The information shown is taken from the DBD (and
also contains some default values).
After you have pressed Enter, the PDU registers the HALDB master database in
the RECONs, but no partition information is available yet.

Figure 4-6 Defining HALDB master DBD information
After this panel is processed, the RECON listing shows the master database
without partitions (see Example 5-4 on page 64).
4.1.4 Setting processing options and global partition information

Next, you need to decide whether you want the PDU to automatically name and
create your partitions, or whether you wish to perform this task manually. You
may also set some options that apply to all of your partitions. Refer to the
Partition Default Information panel (Figure 4-7 on page 53) where these options
are set.
Figure 4-7 Setting partition defaults
On the right side of the panel you see More: +, which means that there are
further default settings. Press F8 to scroll and you get the panel shown in
Figure 4-8 on page 54. This shows you the defaults. Regardless of your choice
for the type of definition, you can set the free space parameters for this database.
The values will be used in all of the partitions when they are created. You may
also change the DBRC options to be used. For a PHDAM database, you are
provided with the opportunity to set global randomizing parameters for your
partitions. The defaults for the randomizer are taken from the DBD.

Figure 4-8 Scrolling partition defaults
4.1.5 Creating your partitions manually

To create your partitions manually, enter NO in the Automatic definition field. Then
provide the partition name and data set name prefix. These are shown in
Figure 4-9 on page 55. The DDNAMEs and data set names will be created
according to the naming conventions.
If you want to use the defaults defined on the previous panel, that is all you have
to do. Otherwise, specify individual free-space information, randomizing values
for PHDAM partitions, and your selected DBRC options.
Figure 4-9 Manual definition of partitions
When you press Enter, you will be presented with the Change Partition panel, as
shown in Figure 4-10 on page 56. On this panel you can name your individual
partitions as well as define the high key values for them. As you can see, some
information is propagated forward from previous panels. The partition name and
the data set name prefix were copied forward. You can rename them at this time.
In Figure 4-10 on page 56 the key is numeric. You could also have a key such as
cz. The partition high key field is case sensitive, so the value cz is different from
the value CZ. The values entered in this field are not translated automatically
from lower case to upper case.
High key and partition string values may be specified as either character or
hexadecimal. The default is character. To specify a hexadecimal value you must
specify it explicitly. For example, you could specify a value of x’0084C68E0000’
for a six-byte key.
The partition ID is generated by the PDU and is used as part of the database
data set name. You cannot change it. When you press Enter you receive a
message informing you that the partition was added successfully. Then the
Change Partition panel is shown with the partition ID. To create another partition,
you may just change the value in the Partition name field and enter a new high
key. Then the next partition is defined. To stop defining further partitions press
F12.

On your last partition, specify a value in the partition high key field that is a higher
value than any expected key value for your database. This prevents the return of
an FM status code to a program that attempts to insert or retrieve a root segment
with a key value higher than all of your defined partition high keys.
Figure 4-10 Change Partition panel
As you complete your partition definitions, the Database Partitions panel, as

shown in Figure 4-11 on page 57, lists what you have defined. You can update
your partition definitions by entering a slash (/) in the appropriate Act field and
pressing Enter.
At this time, the RECONs have also been updated with the partition information.
Example 5-8 on page 67 shows the first part of the RECON listing.
Figure 4-11 Database partitions listing
4.1.6 Creating your partitions automatically

If the number of partitions you need to define is small, creating them manually
may be sufficient. You define the partitions serially, perhaps setting different
attributes for each individual partition. Partition names may be easily controlled
and defined. However, if you have a larger number of partitions, this process is
rather tedious and prone to errors. You have the option of automatically creating
your partitions. Then all partitions have the same attributes in the beginning, but
you could change them afterwards. You cannot change partition names or
partition IDs.
Before you can start an automatic definition you need to create an input file
containing the partition high key values for each partition. This input file must
have the same number of records as partitions you wish to define, and the
partition high key values must start in column 1 of each record. Figure 4-12 is an
ISPF browse of an input key’s data set, which is used to define four partitions
corresponding to the four records containing the high key. The last line contains
the value x’FF’. It creates a high key of all x’FF’ bytes.
Figure 4-12 Sample input keys data set

The data set must be allocated with RECFM=VB and a maximum LRECL of 540.
You may define it as a PDS and then specify the appropriate member.
To automatically define the same database defined in 4.1.5, “Creating your

partitions manually” on page 54, your entries on the Partition Default Information
panel are slightly different. Figure 4-13 shows the Partition Default Information
panel for automatic definitions.
You now enter YES in the Automatic definition field. The data set containing the
partition high key values is entered in the Input dataset field. Then you specify
the pattern for the partition name. The resulting partition names differ from the
previously entered partition names. In our example, we enter NORDD%%. The value
may be up to seven characters, including % signs. The % signs are replaced with
alphanumeric values in collating sequence. For the first partition, each % sign is
replaced with an A. Further partitions get B to Z followed by 0 to 9. If you have
more than 36 partitions, you must use more than one % sign. In our example, the
names of the partitions will be NORDDAA, NORDDAB, NORDDAC, and
NORDDAD. As before, you must enter the data set name prefix you want to use.
Figure 4-13 Partition defaults for automatic definition
Press Enter, and the PDU defines the partitions for you. There will be a pop-up
window showing you the status of the definition process. This includes the
number of partitions defined and the elapsed time. Figure 4-14 on page 59
shows the results with the message that four partitions have been added
successfully. Note that the partition names generated are different than the
manually created names used before. The information in the RECONs is
equivalent to what we generated in our manual definitions. Only the partition
names are different.
Figure 4-14 Automatic partition definition results
4.1.7 Changing partition definitions

After initially defining your database, you may discover that your partition
definitions were either incorrect or incomplete. If you have not already loaded
your database, changing the definitions is very easy. From the Database
Partitions panel, which shows you the list of partitions defined, you can select a
partition to change by placing a slash (/) in the appropriate Act field and pressing
Enter. Alternatively, you could select a partition with a slash, move the cursor
under Edit or File, and then press Enter. This opens a window (see Figure 4-15
on page 60 for the Edit pull-down window). You may then select the option you
wish. By using this window you may also change the attributes of all partitions
simultaneously. In this case do not select a partition by entering a slash before
you open the window.

Figure 4-15 Using pull-down window
You can also manage individual partitions with line commands:

򐂰 To open a partition place an O in the Act field of the selected partition.
򐂰 To open its data set groups use a G.
򐂰 To copy the partition use a C.
򐂰 To delete the partition use a D.
If you have to change your partitions after you have already loaded your
database, see “Changing existing HALDB databases” on page 257 for further
information.
4.1.8 Deleting definitions

You may delete partition definitions by using line commands or the pull-down
window. A Confirm Deletion panel is presented to you with two options. The
difference is only applicable to requests for the deletion of more than one
partition. Option 1 shows you the confirm panel for every partition you want to
delete. Option 2 shows the panel only once for all partitions. Deleting a partition
removes all information from the RECON.
Note: If you want to add the partition again, the PDU will not reuse the
partition ID of the previously deleted partition. The PDU will assign the next
sequential partition ID and your data set names will include the new partition
ID.
To delete the partitioned database use Option 3 of the main menu. You get the
Delete Database Information panel. There you have to type a slash (/) and press
Enter. Then all information about the HALDB database and all its partitions is
deleted from the RECONs. Therefore, handle this command with care.
5
Chapter 5. Batch definition of HALDB

This chapter describes how you can define, change, and delete your HALDB
databases and partitions after you have performed DBDGEN by using DBRC
batch commands. The commands are INIT, CHANGE, and DELETE. We
describe these commands and give examples of RECON listings.

5.1 Using the batch interface
Instead of using the Partition Definition utility you may use DBRC batch
commands to register HALDB databases and define their partitions. This
capability was added by APARs. For the INIT commands, IMS Version 7 must
include PQ35893. The INIT commands are included in the base for IMS Version
8. For the DELETE and CHANGE commands, Version 7 requires PQ52858, and
Version 8 requires PQ59888.
These batch commands have the same functions as the PDU. With DBRC,
however, you have to identify partitions one at a time. There is no automatic
definition. You can use any combination of PDU and DBRC commands. When
using DBRC batch INIT commands, you must include an IMS DD statement in
your job stream. This statement identifies the DBDLIB. Otherwise, you get an
error message as shown in Example 5-1.
Example 5-1 Error message for missing DBDLIB

DSP0011I IMS DD STATEMENT INVALID OR MISSING
DSP0209I PROCESSING TERMINATED WITH CONDITION CODE = 12
Most, but not all, DBRC commands that are used with non-HALDB databases
are also used with HALDB.
5.2 DBRC initialization commands for HALDB

You can use INIT.DB and INIT.PART DBRC commands for registering HALDB
databases with DBRC.
5.2.1 INIT.DB
Prior to registering your database, you must run DBDGEN to define the master
database. In Example 5-2 a DBDGEN for TESTDB has not been done. The
results of an INIT.DB command for TESTDB are shown.
Example 5-2 Error message if no DBDGEN has been run

DSP1050I DATA BASE NOT DEFINED IN DBD LIBRARY
DSP1050I DBD=TESTDB
You may use the INIT.DB command to register a database with DBRC. A
database must be registered with DBRC before you can define its partitions.
The INIT.DB command registers the master database with DBRC. It defines the
DBD name of the master database, its sharing level, the type of database,
whether this database is partitioned by key ranges or a selection exit, and
whether it is recoverable. For further keywords see Table 5-1.
Table 5-1 Keywords with INIT.DB

HALDB keywords Other valid keywords
TYPHALDB DBD
HIKEY | PARTSEL DBTRACK | RCVTRACK
GSGNAME
ICREQ | NOICREQ
NONRECOV | RECOVABL
SHARELVL(0 | 1 | 2 | 3)
TYPHALDB
TYPHALDB is a keyword that you must use to describe the HALDB databases. It
is mutually exclusive with the keywords TYPEIMS for non-HALDB full-function
databases and TYPEFP for Fast Path databases.
HIKEY/PARTSEL
These mutually exclusive optional keywords identify whether the database uses
a partition selection exit or high key values. This setting determines whether the
KEYSTRNG parameter on the INIT.PART command defines a partition selection
exit string or a high key value. These parameters are only valid with TYPHALDB.
HIKEY specifies that this HALDB uses high key values (see Example 5-3).
PARTSEL(pgmname) specifies the name of the partition selection exit routine.
Example 5-3 Registering a HALDB database using key ranges

INIT.DB DBD(NORDDB) TYPHALDB HIKEY SHARELVL(3) +
RECOVABL
If you omit both parameters, the default is taken from the DBD. If a partition
selection exit routine is specified in the DBD, it is used. If it is not specified in the
DBD, key range partitioning (HIKEY) is used.
Parameters specified with the INIT.DB command override the specifications in

the DBD.
If you list the RECON after issuing the INIT.DB command, you get the output for
your database. Example 5-4 on page 64 shows the results of the command from
Example 5-3. You see the master database defined, but no partitions yet. To
specify the partitions you have to use another command.
Chapter 5. Batch definition of HALDB 63

Example 5-4 LIST.RECON
DBD=NORDDB DMB#=16 CHANGE#=0 TYPE=HALDB
PSNAME=**NULL** DBORG=PHIDAM DSORG=OSAM CURRENT PARTITION ID=00000
FLAGS: COUNTERS:
---------------------------------------------------------------------------
DSP0060I NO PARTITIONS REGISTERED FOR THE DATABASE
DSP0060I DBDNAME=NORDDB
5.2.2 INIT.PART
You can use the INIT.PART command only if the database is already registered
with DBRC. The INIT.PART command registers a partition. It creates the HALDB
partition structure (a partition record, a partition database record, and one or
more DBDS records according to the DBD specification). The INIT.PART fails if
the database is being used by the PDU (Example 5-5).
Example 5-5 Error message when HALDB is being used by PDU

DSP1052I DATABASE NORDDB IS IN USE BY HALDB PARTITION DEFINITION UTILITY
Most parameters of the INIT.PART command apply to all the partition DBDSs
created as a result of this command. An example is the ICJCL skeleton. This
differs from the PDU, where these parameters may be specified separately for
each partition DBDS being created. These parameters could be changed later by
using the CHANGE.DBDS command (or PDU).
Restriction: You can not use an INIT.DBDS command for HALDB data sets.
Data sets are always defined by the INIT.PART.
Table 5-2 on page 65 shows valid keywords for the INIT.PART command.
Table 5-2 Keywords for INIT.PART
HALDB keywords Keywords with the same meaning as in
INIT.DBDS for non-HALDB
DBD GENMAX
PART DEFLTJCL
KEYSTRNG ICJCL, OICJCL
DSNPREFX NOREUSE / REUSE
RANDOMZR, ANCHOR, HIBLOCK and RECOVJCL
BYTES, if PHDAM RECOVPD
FBFF and FSPF RECCVJCL
BLOCKSZE, if OSAM
DBD
DBD(name) is a required parameter. It identifies the master database.
PART
PART(name) is a required parameter. It identifies the HALDB partition name. The
name must be alphanumeric, up to seven characters long, with the first character
being alphabetic.
KEYSTRNG
KEYSTRNG is an optional parameter you use to specify a partition high key
value or a partition selection string for use by a partition selection exit. It may be
a character value up to 256 characters long or a hexadecimal value up to 512
characters long.
Character values must be alphanumeric with no embedded blanks or commas

unless the string is enclosed by single quotes. They will be folded to upper case
unless enclosed in single quotes. Hexadecimal values must be enclosed by
single quotes and preceded by the letter X (see Example 5-6).
If no partition selection exit is specified in the HALDB master definition,

KEYSTRNG defines the partition high key and is required. The length of the high
key must be equal to or less than the root key length. If you define a smaller
length, the high key value is padded with hexadecimal ’FF’s up to the defined root
key length.
If a partition selection exit is defined, KEYSTRNG defines a partition selection

string, which is passed to the partition selection routine. Your exit may or may not
require a partition selection string. If required, you determine its contents. It may
be to 256 bytes.
Example 5-6 Definition of KEYSTRNG

KEYSTRNG(00060300002443)

or in hex values:
KEYSTRNG(X’F0F0F0F6F0F3F0F0F0F0F2F4F4F34’)
DSNPREFX
This parameter is required. It specifies the data set name prefix of the partition’s
data sets. It may be up to 37 characters.
PHDAM
If your database is PHDAM, you may specify the randomizer module name with
RANDOMZR and the values for ANCHOR, HIBLOCK, and BYTES, which
correspond to the specifications in a DBD RMNAME parameter where
RMNAME=(module name, number of root anchor points, number of control
intervals in root addressable area, number of bytes).
If you omit these specifications here, they are obtained from the DBD. If the
values in DBD and INIT.PART are different, the values specified by INIT.PART are
used.
Free space
You may enter your specifications for space that is to be left as free space during
load or reorganization by using the keywords FBFF (every nth control interval or
block is left as free space) and FSPF (free space percentage value per interval or
block). The specification applies to all data set groups in the partition.
Example 5-7 defines two partitions named NORD1 and NORD2 in master
database NORDDB. We specify their high key values, data set name prefixes,
block sizes, and free space values.
Example 5-7 INIT.PART

INIT.PART DBD(NORDDB) PART(NORD1) KEYSTRNG(00060300002443) +
DSNPREFX(IMSPSA.IM0A.NORDDB) BLOCKSZE(4096) FBFF(5) FSPF(25)
INIT.PART DBD(NORDDB) PART(NORD2) KEYSTRNG(00110500002902) +
DSNPREFX(IMSPSA.IM0A.NORDDB) BLOCKSZE(4096) FBFF(5) FSPF(25)
After these commands, a LIST.RECON shows that the master database now has
two partitions. It lists information about the partitions. Example 5-8 on page 67
shows only the first partition. The master database listing now indicates that the
number of partitions is 2.
A database record for the NORD1 partition has been added with TYPE=PART.
The DBD name is the partition name. The master database name is also shown.
You may check the data set name prefix, free space, and high key specifications.
The list also includes the partition ID, the previous partition, and the following
partition. DBRC has the partition init needed flag for HALDB partitions. It has
automatically been set to YES.
Example 5-8 RECON listing after INIT.PART (part 1)

DB
DBD=NORDDB DMB#=18 CHANGE#=3 TYPE=HALDB
PSNAME=**NULL** DBORG=PHIDAM DSORG=OSAM CURRENT PARTITION ID=00002
FLAGS: COUNTERS:
---------------------------------------------------------------------------
DBD=NORD1 MASTER DB=NORDDB CHANGE#=2 TYPE=PART
USID=0000000001 AUTHORIZED USID=0000000000 HARD USID=0000000000
RECEIVE USID=0000000000 RECEIVE NEEDED USID=0000000000
DBRCVGRP=**NULL**
DSN PREFIX=IMSPSA.IM0A.NORDDB PARTITION ID=00001
PREVIOUS PARTITION=**NULL** NEXT PARTITION=NORD2
OLRIMSID=**NULL** ACTIVE DBDS=A-J M-V EXIST=NO
FREE SPACE:
FREE BLOCK FREQ FACTOR=5 FREE SPACE PERCENTAGE=25
PARTITION HIGH KEY/STRING (CHAR): (LENGTH=14 )

00060300002443
PARTITION HIGH KEY/STRING (HEX):
F0F0F0F6F0F3F0F0F0F0F2F4F4F3404040404040404040404040404040404040
OSAM BLOCK SIZE:

A = 4096
FLAGS: COUNTERS:
BACKOUT NEEDED =OFF RECOVERY NEEDED COUNT =0
READ ONLY =OFF IMAGE COPY NEEDED COUNT =0
PROHIBIT AUTHORIZATION=OFF AUTHORIZED SUBSYSTEMS =0
HELD AUTHORIZATION STATE=0
EEQE COUNT =0
TRACKING SUSPENDED =NO RECEIVE REQUIRED COUNT =0
OFR REQUIRED =NO
PARTITION INIT NEEDED =YES
ONLINE REORG ACTIVE =NO
PARTITION DISABLED =NO
The INIT.PART command also creates all of the DBDS records for your partitions.
Certain values have automatically been created. Example 5-9 on page 68 lists
the DBDS records.
򐂰

The DBDS record includes the following information:
򐂰 The data set name field (DSN) is composed of the data set name prefix field
followed by the suffix .A00001. The suffix is the concatenation of dot, the
DSGROUP parameter in the DBD source (A, if you have only one data set
group), and the generated 5-byte partition ID.
򐂰 The DBD field is the partition name you chose.
򐂰 The DD name field (DDN) is the concatenation of the partition name and the
DSGROUP value. A partition name must be unique in a set of RECONs.
In the listing two additional DBDS records also appear. The first is for the indirect
list data set (ILDS). It is suffixed by .L00001. The second is for the PHIDAM
primary index and is suffixed by .X00001. You may not change these suffixes.
Example 5-9 RECON listing after partition definition (part 2)

--------------------------------------------------------------------------
DBDS
DSN=IMSPSA.IM0A.NORDDB.A00001 TYPE=PART
DBD=NORD1 DDN=NORD1A DSID=001 DBORG=HIDAM DSORG=OSAM
CAGRP=**NULL** GENMAX=2 IC AVAIL=0 IC USED=0 DSSN=00000000
NOREUSE RECOVPD=0 OTHER DDN=**NULL**
DEFLTJCL=**NULL** ICJCL=ICJCL OICJCL=OICJCL RECOVJCL=RECOVJCL
RECVJCL=ICRCVJCL
FLAGS: COUNTERS:
IC NEEDED =OFF
RECOV NEEDED =OFF
RECEIVE NEEDED =OFF EEQE COUNT =0
--------------------------------------------------------------------------
DSN=IMSPSA.IM0A.NORDDB.L00001 TYPE=PART
DBD=NORD1 DDN=NORD1L DSID=003 DBORG=INDEX DSORG=VSAM
FLAGS: COUNTERS:
RECOV NEEDED =OFF EEQE COUNT =0
--------------------------------------------------------------------------
DBDS
DSN=IMSPSA.IM0A.NORDDB.X00001 TYPE=PART
DBD=NORD1 DDN=NORD1X DSID=005 DBORG=INDEX DSORG=VSAM
FLAGS: COUNTERS:
RECOV NEEDED =OFF EEQE COUNT =0
--------------------------------------------------------------------------
5.3 DBRC change commands for HALDB

You can use the CHANGE.DB, CHANGE.PART and CHANGE.DBDS DBRC
commands for changing the information related to HALDB databases in RECON.
5.3.1 CHANGE.DB
HALDB support for the CHANGE.DB command was added to IMS Version 7 by
APAR PQ52858 and to IMS Version 8 by APAR PQ59888. The CHANGE.DB
command is used to change the information about a database or partition. You
may use the CHANGE.DB command with the DBD for the master database
name as well as with the DBD for a partition database name. These are shown in
Example 5-10.
Example 5-10 CHANGE.DB for master and partition database

CHANGE.DB DBD(NORDDB) NONRECOV
CHANGE.DB DBD(NORD1) PINIT
In the first case, changes apply to all partitions. In the second, changes only
apply to the specified partition. The keywords you may use depend on whether
you specify the master or partition database, as shown in Table 5-3.
Table 5-3 Keywords for CHANGE.DB

Valid keywords for: Master Partition
AUTH I NOAUTH NO YES

BACKOUT I NOBACK NO YES
READON I READOFF NO YES
SHARELVL YES NO
NONRECOV I RECOVABL YES NO
TYPEFP I TYPEIMS NO NO
GSGNAME I NOTCOVER YES NO
RCVTRACK I DBTRACK YES NO
UNAUTH NO YES
PINIT I NOPINIT YES YES
PARTSEL I HIKEY YES NO
ICREQ I NOICREQ YES NO
5.3.2 CHANGE.PART
The CHANGE.PART command was added to IMS Version 7 by APAR PQ52858
and to IMS Version 8 by APAR PQ59888. You can use the CHANGE.PART
command to change the attributes of a partition. The changes apply to all the
DBDSs of the partition. This command may change all of the attributes that you
may set with an INIT.PART command (see “INIT.PART” on page 64). The
following example changes the data set name prefix of all DBDSs of partition
NORD1 to IMSPSA.IM0B.NORDDB (Example 5-11 on page 70).

Example 5-11 CHANGE.PART to change the DSN prefix
CHANGE.PART DBD(NORDDB) PART(NORD1) DSNPREFX(IMSPSA.IM0B.NORDDB)
We can change some attributes, such as skeletal JCL member names, for a
single DBDS of a partition by using a CHANGE.DBDS command. It would be a
good practice to always do a LIST.PART before the CHANGE.PART command.
Note that if you change the randomizer parameters in any way, you should have
unloaded the partition first. If you have not, you are not able to read the partition.
The change of the randomizer parameters turns the partition init needed flag on
and the partition is unreadable.
5.3.3 CHANGE.DBDS
Use the CHANGE.DBDS command to change some of the attributes of a single
DBDS.
The information is stored in a DBDS record in RECON. You cannot change the
data set name or DD name with this command. Keywords DDNNEW or DSN are
not allowed for HALDB data sets. The naming conventions apply to all DBDSs of
a partition and cannot be overridden for a single DBDS.
You have to distinguish between data data sets and index or ILDS data sets
when using other keywords.
For ILDS and index data sets only, the keywords listed in Example 5-12 are valid.
Example 5-12 Valid keywords for ILDS and PHIDAM primary index
ADDEQE | DELEQE
RECOV | NORECOV
This means that you may only change the information about error queue
elements and the recovery needed status for PHIDAM primary index and ILDS
data sets.
For data data sets you can use more keywords. They are listed in Example 5-13.
Example 5-13 Valid keywords for data data sets in HALDB partitions
ADDEQE | DELEQE
DEFLTJCL | NODEFLT
GENMAX
ICJCL
ICON | ICOFF
NOREUSE | REUSE
OICJCL
RECOV | NORECOV
RECOVJCL
RECOVPD
RECVJCL
Example 5-14 is an example of changing the status of a data set from recovery
needed to recovery needed off.
Example 5-14 CHANGE.DBDS

CHANGE.DBDS DBD(NORD1) DDN(NORD1A) NORECOV
5.4 DBRC delete commands for HALDB

You can use DELETE.DB and DELETE.PART DBRC commands for deleting
HALDB databases from RECON.
5.4.1 DELETE.DB
HALDB support for the DELETE.DB command was added to IMS Version 7 by
APAR PQ52858 and to IMS Version 8 by APAR PQ59888. Use a DELETE.DB
command to delete a database from the RECONs. The command also deletes all
information related to the database that has been recorded in the RECONs.
Example 5-15 shows the deletion of database NORDDB and its related
information from the RECONs. A partition name cannot be specified. If the
database is a HALDB database, the master and all its partitions are deleted. If
the database or any of its partitions is in use, the command fails, and none of the
RECON records are deleted.
Example 5-15 Delete a database

DELETE.DB DBD(NORDDB)
5.4.2 DELETE.PART
The DELETE.DB command was added to IMS Version 7 by APAR PQ52858 and
to IMS Version 8 by APAR PQ59888. Use a DELETE.PART command to delete a
partition from the RECONs. The command also deletes all information related to
that partition that has been recorded in the RECONs. Note that if you delete a
partition, you will not be able to use the recovery utility to put the partition back.

Example 5-16 shows the deletion of partition NORD1 and its related information
from the RECONs. If the partition or any other partition affected by the deletion is
authorized, the command fails and none of the RECON records are deleted.
Example 5-16 Delete a partition

DELETE.PART DBD(NORDDB) PART(NORD1)
Restriction: You cannot use a DELETE.DBDS command to delete a partition

DBDS.
6
Chapter 6. Partition initialization

All HALDB partitions must be initialized before they are used. This chapter
describes the partition initialization function and the utilities that initialize
partitions. We provide examples showing the use of these utilities.

6.1 Partition initialization function
HALDB partitions must be initialized before they may be used. This function does
not exist with non-HALDB databases. Initialization is done either by the Database
Prereorganization utility or the Database Partition Data Set Initialization utility.
Alternatively, the IBM High Performance Load tool may be used to do partition
initialization. Initialization makes a partition usable but does not place any
database segments in the partition. After initialization a partition is empty.
When creating a new HALDB database or migrating a non-HALDB database to

HALDB, you may create one or more partitions that contain no data. For
example, you might have a partition for root keys ranging from 4,000,001 through
5,000,000, but not have any roots in this range. Later, a batch program or online
transaction might attempt to get or insert a root in this range. Partition
initialization ensures that the partition may be processed by such a program.
Partition initialization is required for a partition in the following three cases:

򐂰 The partition is loaded by an initial load program. An initial load program uses
PROCOPT=L.
򐂰 The partition is loaded by HD Reload when migrating from a non-HALDB
database.
򐂰 The partition is accessed by an application program before any segments are
loaded in it by initial load or a reload utility.
Initialization is not required for a normal reorganization reload. A partition does

not have to be initialized before a non-migration use of HD Reload loads
segments in the partition. This is true even when the partition’s data sets have
been deleted and redefined after the unload. On the other hand, initialization is
required for partitions with no data.
Partition initialization is used with all HALDB databases including secondary

indexes (PSINDEX).
Partition initialization writes the partition ID number and the initial reorganization
number in PHDAM and PHIDAM partitions. The initial reorganization number is
one. The numbers are written in the first four bytes of the first block of the first
data set. This is the bit map block. The first four bytes are used for the DBRC
usage indicator in non-HALDB databases. Since DBRC registration is always
required with HALDB, the DBRC usage indicator function is not used with
HALDB. For PHDAM partitions, a dummy record is written and deleted. For
PHIDAM partitions, the high keys (all X’FF’s) record is written. This record is
stored in every PHIDAM partition. For PSINDEX partitions, one record is written
and then deleted. These actions make the high-used-RBA non-zero.
6.2 DBRC flags used with partition initialization
Partition initialization uses two flags in the RECONs. These are the partition init
needed and the image copy needed flags.
6.2.1 Partition initialization required (PINIT) flag

A partition in a HALDB database may have a partition init needed (PINIT) flag
set in the DBRC RECONs. The flag indicates that the partition requires
initialization. When either the Database Prereorganization utility or the Partition
Initialization utility is run against the database, it initializes only those partitions
with their PINIT flags set. There is one exception to this. We discuss this
exception in 6.4.1, “Unconditional partition initialization” on page 78.
The PINIT flag is set in five ways:

򐂰 The flag is set when a partition is defined.
򐂰 The flag is set when explicitly turned on by a DBRC CHANGE.DB command.
The command specifies the PINIT keyword. Example 6-1 shows the
command to turn on the flag for all partitions in the PEOPLE database.
Example 6-2 shows the command to turn on the flag for only the PEO01
partition in the PEOPLE database.
򐂰 The flag is set when making changes to partition boundaries such that data
may be moved either into the partition or out of the partition. This only applies
to databases that use key range partitioning.
򐂰 The flag for a partition is set when its partition string is changed. This only
applies to databases using a partition selection exit routine.
򐂰 The flags for all partitions are set when the partition selection exit routine
name is changed. This only applies to databases using a partition selection
exit routine.
The setting of the flag is shown in a listing of the partition database record in the
RECONs as PARTITION INIT NEEDED =NO or PARTITION INIT NEEDED
=YES.
Example 6-1 Setting the PINIT flag for master database PEOPLE
CHANGE.DB DBD(PEOPLE) PINIT
Example 6-2 Setting the PINIT flag for partition PEO01

CHANGE.DB DBD(PEO01) PINIT
Chapter 6. Partition initialization 75

6.2.2 Image copy needed flag
Partition initialization sets the image copy needed flag for each partition that is
initialized. The image copy needed flag does not prevent authorization for initial
loads. It does prevent authorizations for updates. If you initialize but do not
initially load a partition, it must be image copied before any subsystem will be
given authorization to update it.
6.3 Database Prereorganization utility

The Database Prereorganization utility (DFSURPR0) may be used to initialize
partitions. This function was added to the utility in IMS Version 7.
As when used with non-HALDB databases, the utility also creates a control data
set (CDS). The CDS is not required by other utilities when they process HALDB
databases. With HALDB the CDS is merely used to disable any attempted use of
the Database Scan and Database Prefix Resolution utilities. Nevertheless, the
DFSURCDS DD statement is required with the Prereorganization utility.
The DBIL, DBR, or DBM control statement is used to indicate which database or
databases should have their partitions initialized. All of these control statements
do the same initialization for HALDB partitions. We used the DBIL statement in
our testing. You may specify multiple databases on the statement. Database
names must be eight bytes, left justified, and separated by commas.
Example 6-3 shows the control statement we used to initialize the partitions in
databases PEOPLE and PEOSKSI. Since PEOPLE is only six bytes, we placed
two blanks before the following comma.
Example 6-3 Prereorganization DBIL= control statement

//SYSIN DD *
DBIL=PEOPLE ,PEOSKSI
/*
A DFSVSAMP DD statement with definitions for buffer pools is required when

running Prereorganization to initialize partitions. The buffer pools must contain
buffers large enough to hold the OSAM blocks or VSAM CIs that are written
during the process. Large pools are not advantageous.
6.4 Database Partition Data Set Initialization utility

The Database Partition Data Set Initialization utility (DFSUPNT0) may be used to
initialize partitions. It is an alternative to the Database Prereorganization utility.
There are two ways to execute this utility. It may use a utility (ULU) region under
the control of DFSRRC00 or it may execute as a stand-alone program.
Example 6-4 shows the JCL that may be used to execute as a utility region.
Example 6-5 shows the JCL that may be used to execute as a stand-alone
program. When executed as a utility region, the utility can initialize the partitions
for only a single database. The database is specified as the third subparameter
in the PARM specification. In Example 6-4 the partitions in database PEOPLE
are initialized.
Example 6-4 Using a ULU region with DFSUPNT0

//PARTINIT EXEC PGM=DFSRRC00,REGION=1024K,
// PARM='ULU,DFSUPNT0,PEOPLE,,,,,,,,,,,Y,N'
//DFSRESLB DD DSN=IMSPSA.IMS0.SDFSRESL,DISP=SHR
//IMS DD DISP=SHR,DSN=JOUKO3.HALDB.DBDLIB
//RECON1 DD DISP=SHR,DSN=IMSPSA.IM0A.RECON1
//DFSVSAMP DD *
VSRBF=8192,10
IOBF=(4096,10)
/*
//SYSPRINT DD SYSOUT=*
//SYSUDUMP DD SYSOUT=*
//SYSIN DD DUMMY
When executed as stand-alone, the databases to be initialized must be specified

in the SYSIN data set. Multiple databases may be initialized with one execution.
This is shown in Example 6-5 where the partitions in both PEOPLE and
PEOSKSI are initialized.
Example 6-5 Using a stand-alone region with DFSUPNT0

//PARTINIT EXEC PGM=DFSUPNT0,REGION=1024K
//DFSVSAMP DD *
VSRBF=8192,10
IOBF=(4096,10)
/*
//SYSIN DD *
PEOPLE
PEOSKSI
Chapter 6. Partition initialization 77

/*
A DFSVSAMP DD statement with definitions for buffer pools is required when

running the Database Partition Data Set Initialization utility. The buffer pools must
contain buffers large enough to hold the OSAM blocks or VSAM CIs that are
written during the process. Large pools are not advantageous
This utility does not create the control data set (CDS) that the Prereorganization
utility creates. This is not a problem since the CDS is not used with HALDB.
6.4.1 Unconditional partition initialization

The Database Partition Data Set Initialization utility may be used to initialize
partitions even when their partition init needed (PINIT) flag is not set. This
function is called unconditional partition initialization. It was added to IMS Version
7 by APAR PQ49638 and to IMS Version 8 by PQ55002. It allows a user to delete
and define all of the data sets in a database and then initialize them without
having to issue any DBRC CHANGE.DB commands.
As mentioned in 6.1, “Partition initialization function” on page 74, initialization is

not required for a normal reorganization reload if each partition contains data.
This is true even when the partitions’ data sets have been deleted and redefined
after the unload. Partition initialization is required for empty partitions. In these
cases, unconditional partition initialization may be useful.
Unconditional partition initialization is invoked by including an INITALL control

statement after a DFSOVRDS DD statement. Example 6-6 shows these
statements.
Example 6-6 DFSOVRDS DD statement with the INITALL control statement

//DFSOVRDS DD *
INITALL
/*
Unconditional partition initialization is not supported by the Database

Prereorganization utility (DFSURPR0). This capability is only provided by the
Database Partition Data Set Initialization utility.
7
Chapter 7. Partition selection

In this chapter we explain your choices for partition selection. We describe the
advantages and disadvantages for using key range partitioning versus the use of
partition selection exit routines. We provide guidance on writing partition
selection exit routines.

7.1 Choosing the type of partition selection
HALDB allows you to choose between two types of partition selection. Key range
partitioning assigns a range of keys to each partition. Alternatively, you may write
a partition selection exit routine that uses its own criteria for assigning keys to
partitions. In both types, only the key of the root segment is used to assign a
record to a partition. Both types may be used with any HALDB database type.
These are PHDAM, PHIDAM, and PSINDEX.
7.1.1 Key range partitioning

Key range partitioning is usually preferred. With key range partitioning, a high key
is assigned to each partition. The partition holds all records with keys higher than
the previous partition and up to the value of its high key.
Key range partitioning is the simpler method to implement because you do not
have to write an exit routine. You only have to assign a high key to each partition.
In PHIDAM and PSINDEX databases, the records are in key sequence across
the entire database. This makes them compatible with HIDAM and non-HALDB
secondary index databases. When a partition selection exit routine is used for
PHIDAM or PSINDEX databases, records are in key sequence within a partition,
but not across partitions. This would make these databases incompatible with
HIDAM and non-HALDB secondary index databases. Applications that require
database records to be in key sequence would not function correctly.
IMS is aware of the impact of the addition of partitions, their deletion, or changes
in their boundaries. IMS can determine which partitions require initialization after
these changes. Figure 7-1 illustrates this. In the figure, the change of the high
key for partition B from 40000 to 50000 would require the movement of records
from partition C to partition B. Both of these partitions would have to be unloaded
before the change. Both would have to be initialized and reloaded. IMS sets the
partition init needed flag for both partitions when the high key for partition B is
changed. This is not true when you use an exit routine. The logic of the exit
routine determines which partitions require initialization after these changes. You
are responsible for understanding this and initializing the correct partitions.
A B C D
High key = 20000 High key = 40000 High key = 70000 High key = high values
Figure 7-1 Key range partitioning
We recommend that you assign a high values key to a partition. If you do not and
if an application attempts to insert a root with a key higher than the highest key
defined, the call will receive an FM status code. Most applications are not coded
to handle this status code. A high values key is one with X’FF’ for each byte.
There is a case where key range partitioning will require the addition of partitions
or changes to their key ranges over time. This occurs when records that are
added to the database have increasing keys. Examples include keys created by
incrementing the previous key and keys based on dates or times. As records are
added, you will need to split the last partition into multiple partitions. As records
are deleted, you may want to consolidate partitions containing lower values.
7.1.2 Partitioning with an exit routine

You may use a partition selection exit routine to assign records to partitions. Only
the key of the root segment is available from the database record for the exit
routine. Since writing and maintaining the exit routine increases your
responsibilities, we recommend that you use an exit routine only if key range
partitioning does not meet your needs.
An exit routine would allow you to assign records to a partition based on a value
other than the high order part of the key. Figure 7-2 illustrates a use of an exit
routine. In this figure a warehouse code is in the fifth and sixth bytes of the
eight-byte key. The partition selection exit routine assigns records to partitions
based on the warehouse code. The installation plans on processing data by
warehouse. By isolating data for a warehouse in one partition, they can simplify
the processing and improve its performance.
xxxx01xx xxxx02xx xxxx03xx xxxx04xx
Key format = xxxxnnxx

warehouse code
Figure 7-2 Partitioning with an exit routine
The HALDB Conversion and Maintenance Aid product includes an exit routine
that will do this kind of partition selection. If you use it, you do not have to write
the exit routine. You only have to specify what part of the key is to be used and
the values for each partition.
Chapter 7. Partition selection 81

Another use of a partition selection exit routine is to isolate some database
records because of their special characteristics. For example, the sizes of most
records in a PHDAM database may be fairly uniform, but others may be much
larger. These unusually large records could otherwise cause space use problems
within partitions. If the keys of these large records are known, an exit routine
could recognize their keys and place them in a partition with different space
characteristics. The partition might have many fewer records spread across the
same amount of space. It might even have its own specialized randomization
routine.
In the last paragraph in 7.1.1, “Key range partitioning” on page 80, we discuss a
database whose key values continued to increase. With key range partitioning
this could present a problem. You would have to continually adjust the partitions.
An exit routine might solve this problem. If there is no need for key sequence
processing, an exit routine could distribute the records across partitions evenly.
For example, it might assign the records based on the low order part of the key.
Then each partition would always receive a consistent percentage of the records.
Databases whose records must be processed in key sequence must use key
range partitioning. Typically, a partition selection exit routine would assign
records such that moving from one partition to the next might cause a decrement
in the key value. The keys are sequential within the partitions but not from one
partition to the next. This may have an impact on the application if it expects the
keys to always increase. Records in HIDAM databases and non-HALDB
secondary indexes are always in key sequence. When they are migrated to
PHIDAM or PSINDEX do not use a partition selection exit routine unless you are
sure that key sequence processing is not required.
7.2 Writing partition selection exit routines

You may write a partition selection exit routine to select partitions for database
records in a HALDB database. The exit routine name must match what you
specify in the Partition Definition utility or the DBRC INIT.DB command. The exit
routine provides several functions. Documentation on writing the exit routine can
be found in the manuals IMS Version 7 Customization Guide, SC26-9427, and
IMS Version 8 Customization Guide, SC27-1294.
7.2.1 Partition selection exit routine functions

The partition selection routine functions are divided into two categories. The first
category is partition selection. The functions in partition selection are:
򐂰 Select target partition.
򐂰 Select first partition.
򐂰 Select next partition.
The second category of functions is for partition structure management

awareness. These functions are:
򐂰 Partition structure initialization
򐂰 Partition structure termination
򐂰 Partition structure modification
Partition selection functions

When a program issues a call, IMS determines which of these functions is
required and invokes the exit routine for the function.
Select target partition

This function is used for calls qualified on the root segment. The root key is
passed to the exit routine. The routine selects the partition where the key is
assigned.
Select first partition

This function is used for unqualified calls with no previous position in the
database. The routine selects the first partition.
Select next partition

This function is used when IMS reaches the end of a partition. IMS passes the
partition ID of the last partition processed to the exit routine. The exit either
selects the next partition or indicates that the end of the database has been
reached.
Partition structure management awareness functions

These functions allows the exit routine to adjust to changes in the environment,
such as the addition or deletion of partitions.
Partition structure initialization

This function is invoked when a database is first referenced. The exit routine can
do any initialization processing it requires. For example, it could load a module or
acquire storage.
Partition structure termination

This function is invoked when a database is no longer used. For example, when a
/DBR command is issued for the database, the exit is driven for this function. The
exit routine can do any termination processing it requires. For example, it could
delete any modules it has loaded or release any storage it has required.

Partition structure modification
This function is invoked when DBRC authorization determines that the database
definition has been changed. Examples of these changes include adding a new
partition, deleting a partition, changing a partition string value, or changing the
randomization parameters for a PHDAM partition. The exit routine can make any
necessary adjustments due to the change.
7.2.2 Information passed to and from the exit routine

The information passed to the exit routine depends on the function invoked.
Information is passed in two areas. Register 1 points to a parameter list that
contains the addresses of these areas.
The Exit Communication Area contains an indication of the function requested,

provides the input data, and contains storage for the exit routine to indicate its
chosen action. For example, a select target partition request includes the key of
the root segment. In this case, the exit routine would place the selected partition
ID in the feed back field. This area is mapped by the DFSPECA DSECT, which is
documented in the HALDB Partition Selection Exit Routine chapter of the IMS
Version 8 Customization Guide, SC27-1294, and included in the DFSPSE00
sample program.
The second area contains the Partition Definition Area prefix and entries. The
prefix contains a pointer to the first entry and the number of entries. It also
contains five words of storage that are available for your use. There is an entry
for each partition. They are in adjacent storage locations. Each entry contains
information about a partition including its name, partition ID, partition string
length, and partition string address. The DFSPDA DSECT maps the prefix and
the DFSPDAE DSECT maps an entry. These DSECTs can be found in IMS
Version 8 Customization Guide, SC27-1294, and they are also included in the
DFSPSE00 sample program.
7.2.3 Sample exit routine (DFSPSE00)

IMS provides a sample partition selection exit routine (DFSPSE00). In IMS
Version 7 it is in the ADFSSRC library. In IMS Version 8 it is in the ADFSSMPL
and SDFSSMPL libraries. This sample is not intended for actual use. It
implements key range partitioning by using the partition string for each partition
as its high key. If you want this type of partitioning there is no need to have an exit
routine. The sample demonstrates the interface and control blocks used by the
exit. We highly recommend that you examine this sample before writing your own
exit routine.
User modification to the sample exit routine
The sample could be the basis for writing a routine to support other types of
partitioning. For example, if you wish to implement partitioning as shown in
Figure 7-2 on page 81, you can modify the sample. The modification needs to
compare the warehouse code field in the key with the partition strings. Each
warehouse code would be the partition string for its partition. Example 7-1 shows
the modification to DFSPSE00 that we made to implement this partitioning
scheme.
The modification is in the select target partition function of the routine that is at
label PSETRGT. The modification is surprisingly easy. It requires only the
addition of one instruction. This is a load address (LA) instruction to change the
address for the compare. Since the field with the warehouse code is at offset four
in the key, we add four bytes to the key address. We do not have to modify the
length of the compare. The module bases the compare length on the size of the
largest partition string, not the size of the key as it is defined in the DBD. This is
what our partitioning scheme needs. When defining the partitions for use with
this exit routine, we specify two-byte partition strings. Each string is a two-byte
warehouse code.
Example 7-1 Modification to DFSPSE00

*********************************************************************** 04340000
* SELECT TARGET PARTITION * 04350000
* * 04360000
* THE PARTITION EXIT COMMUNICATION AREA (PECKEY) SPECIFIES * 04370000
* THE CURRENT DL/I CALL KEY. THIS KEY IS USED TO SELECT THE * 04380000
* TARGET PARTITION. THE PARTITION DEFINED WITH A HIGH KEY * 04390000
* EQUAL TO OR GREATER THAN THE CURRENT DL/I CALL KEY IS * 04400000
* SELECTED. * 04410000
* * 04420000
* PARTITION ID RETURNED TO IMS FOR SELECTED PARTITION * 04430000
*********************************************************************** 04440000
PSETRGT DS 0H 04450000
LTR R7,R7 PKT EXIST 04460000
BZ PSEE804 NO, ERROR 04470000
L R0,PDAUSR3 NUMBER OF PKT ENTRIES 04480000
L R2,PECKEY DL/I CALL KEY ADDRESS 04490000
********************************************************************** 04491001
* PSEWAREH MODIFICATION 04491101
* BEGIN MODIFICATION TO SELECT PARTITION ON SUBSET OF KEY 04491202
* SUBSET OF KEY IS AT OFFSET 4 04491302
* 04491402
LA R2,4(,R2) 04491501
* 04491602
* THE IC INSTRUCTION BELOW INSERTS THE PARTITION STRING LENGTH 04491702
* IN THE EX INSTRUCTION WHICH IS USED TO DO THE COMPARE 04491802
* OUR PARTITION DEFINITION USES ONLY TWO BYTES FOR THE STRINGS 04491902
* END OF MODIFICATION TO SELECT PARTITION ON SUBSET OF KEY 04492001
********************************************************************** 04493001

SR R3,R3 04500000
SPACE 04510000
PSET3000 DS 0H 04520000
IC R3,PKTKEYL KEY LENGTH -1 04530000
EX R3,PTKEY COMPARE PART HI-KEY 04540000
BNH PSET4000 PARTITION FOUND 04550000
AR R7,R6 NEXT PKT ENTRY 04560000
BCT R0,PSET3000 CHECK NEXT ENTRY 04570000
B PSEE402 ERROR 04580000
SPACE 04590000
PSET4000 DS 0H 04600000
LH R1,PKTID TARGET PARTITION ID 04610000
STH R1,PECFDB2 . 04620000
SPACE 04630000
*********************************************************************** 04640000
* SELECT TARGET PARTITION PROCESSING COMPLETE * 04650000
* RETURN TO IMS WITH PECRC SET TO ZERO * 04660000
*********************************************************************** 04670000
B PSEEXIT RETURN TO IMS 04680000
DFSPSE00 contains a write to operator (WTO) instruction in its partition

structure initialization function that is at label PSEINIT. This could be useful in a
test environment, but should be eliminated before the exit routine is used in a
production environment. The instruction may either be deleted from the code or
converted to a comment.
Part 2
Part 2 Migration
In this part we describe the process used to migrate databases to HALDB. We
include secondary indexes and logically related databases. We provide
examples of the migration of databases.
The HALDB Conversion and Maintenance Aid is a product that simplifies the
migration process and the maintenance of HALDB databases. We describe this
tool and show you how to use it.

8
Chapter 8. Migration from non-HALDB

to HALDB
In this chapter we describe the process for migrating a current non-HALDB
database to HALDB. We describe the different processes used for migrating
databases with logical relationships or secondary indexes and those that use
PDB. We show you how to migrate user-partitioned databases.

8.1 General migration considerations
A HDAM or HIDAM database may be migrated to HALDB with the following
steps:
򐂰 Unload of the existing database using HD Unload and specifying
MIGRATE=YES
򐂰 DBDGEN as a HALDB database
򐂰 Deletion of database information from the RECONs
򐂰 Definition of the partitions
򐂰 Allocation of database data sets
򐂰 Partition initialization
򐂰 Reload as a HALDB database using HD Reload
򐂰 Image copy of the database data sets
Secondary indexes are created when their indexed database is migrated. They
are created with the following steps:
򐂰 DBDGEN as a HALDB secondary index
򐂰 Deletion of secondary index database information from the RECONs
򐂰 Definition of partitions
򐂰 Sort of output file from the unload of the indexed database
򐂰 Load of the secondary index using HD Reload
The migration of databases without logical relationships or secondary indexes is

explained in 8.3, “Migrating simple databases” on page 96.
Changed and additional steps for the migration of databases with secondary
indexes are explained in 8.5, “Migrating databases with secondary indexes” on
page 101.
If databases are logically related, they must be migrated together. Changed

steps for the migration of these databases are explained in 8.6, “Migrating
databases with logical relationships” on page 114.
8.1.1 DBD changes

You must modify the DBD for any database that you migrate to HALDB. This
section explains the modifications that are required.
HIDAM index DBDs
There is no DBD for a PHIDAM primary index. When a HIDAM database is
migrated to PHIDAM, the DBD for the HIDAM Index is discarded. IMS gets the
information required to generate the PHIDAM primary index from the PHIDAM
DBD.
DBD statement
There are the following three changes to the DBD statement for HALDB:
򐂰 There are new values for the ACCESS parameter. They are PHDAM,
PHIDAM, and PSINDEX. The value should be changed to the appropriate
value.
򐂰 The optional PSNAME parameter has been added. It is used to specify a
default partition selection exit routine. If you are not going to use the exit
routine, do not specify this parameter.
򐂰 The RMNAME parameter has a slightly different meaning. With PHDAM
databases it is optional. It is used to define default randomizer values for
partitions. These values may be changed for individual partitions when they
are defined.
DATASET statement
The DATASET statement is not used in HALDB DBDs. The entities it defines for
non-HALDB databases are handled differently with HALDB as follows:
򐂰 DDNAMEs are created by the definition of partitions.
򐂰 Data set groups are defined by the DSGROUP parameter on SEGM
statements.
򐂰 Free space and OSAM block size specifications are made when partitions are
defined.
򐂰 The SCAN parameter is not specified with HALDB DBDs. HALDB operates as
if SCAN=0 were specified. This is also the recommended value for
SEGM statement
There are the following three changes to the SEGM statement for HALDB:
򐂰 The valid specifications for the PTR parameter have changed.
HALDB does not use hierarchic pointers. Any use of HIER, H, HIERBWD, or
HB keywords with the PTR parameter must be changed to TWIN, T,
TWINBWD, or TB.
PHIDAM does not support the use of twin forward only (TWIN or T) pointers
for root segments. If you have HIDAM roots using twin forward only pointers,
Chapter 8. Migration from non-HALDB to HALDB 91

the keyword for the PTR parameter on the SEGM statement should be
changed to NOTWIN, NT, TWINBWD, or TB.
HALDB does not use symbolic pointers. Changes required for logical
relationships are explained in “Changing pointer options for logical
relationships” on page 115.
򐂰 The specification of data set groups for HALDBs is done with the DSGROUP
parameter.
You may not want to maintain multiple data set groups. If you do want multiple
data set groups, you must specify the DSGROUP parameter on the SEGM
statement for any segment that is not in the first data set group. The valid
values for DSGROUP are the letters A through J. A is the first data set group,
B is the second, and J is the tenth.
򐂰 The BYTES parameter for a secondary index segment must be increased by
four bytes if a /SX field is used as a subsequence field. We explain this further
in “/SX subsequence fields” on page 102.
LCHILD statement
The LCHILD statement is not used to define the primary index in a HIDAM DBD.
When converting a HIDAM DBD to PHIDAM, this LCHILD statement should be
deleted.
When converting a secondary index DBD to a HALDB PSINDEX DBD, there are
three possible changes required:
򐂰 If PTR=SYMB is specified in a secondary index, it must be changed to
PTR=SNGL or omitted. PTR=SNGL is the default and only valid specification
for PSINDEX LCHILD statements.
򐂰 If PTR=SYMB is specified in a HDAM or HIDAM indexed database, it must be
changed to PTR=INDX.
򐂰 An RKSIZE parameter must be specified on the LCHILD statement in the
secondary index DBD. This is the size of the root’s key in the target database.
When converting a logical relationship using symbolic pointing, you must omit the
PTR=SYMB specification on the LCHILD statement for the logical relationship.
The changes for LCHILD statements used to define logical relationships using
virtual pairing are explained in “Making changes to the DBDs for physical pairing”
on page 116.
XDFLD statement
HALDB does not support shared secondary indexes. If the CONST parameter is
specified in an XDFLD statement of an indexed database, it must be deleted. The
CONST parameter is used to specify the character associated with a shared
secondary index. Each HALDB secondary index must be stored in its own
secondary index database. Separate PSINDEX databases must be defined for
each shared secondary index being converted to HALDB.
FIELD statement
The BYTES parameter of the sequence field for a secondary index segment
must be increased by four bytes if a /SX field is used as a subsequence field. We
explain this further in “/SX subsequence fields” on page 102.
8.2 HALDB Migration Aid utility (DFSMAID0)

The HALDB Migration Aid utility (DFSMAID0) may be used to understand
partitioning requirements for a database. The utility reads an existing database
and reports on the sizes and boundaries of potential HALDB partitions. The
existing database may be HDAM, HIDAM, PHDAM, PHIDAM, a non-HALDB
secondary index, or a HALDB secondary index (PSINDEX). HDAM, HIDAM, and
secondary index databases are read to understand the requirements for a
migration. PHDAM, PHIDAM, and PSINDEX databases are read to understand
the requirements for repartitioning an existing HALDB database.
The utility reports for non-HALDB secondary indexes must be used with care.
Some of the information reported is not accurate. We explain the special
considerations for using DFSMAID0 with secondary indexes in 8.5.2, “Estimating
the sizes of secondary index partitions” on page 107.
You choose the type of analysis that the utility does. There are three options.
First, you may specify the high key values for potential partitions. The utility
reports the space required for each of these partitions. Second, you may specify
the maximum number of bytes of segment data stored in each partition. The
utility reports the number of partitions required and their high keys. Third, you
may specify the number of equal-sized partitions to be used. The utility reports
the space required and the high key for each partition. In all cases, the utility
reports the total amount of space required for all segments in the database.
8.2.1 Reports
There is a separate report for each partition and one for the database.
Example 8-1 on page 94 shows a typical report for a partition. The lowest key
and highest key in the partition are reported in both character (EBCDIC) and
hexadecimal formats. Each segment type in the database is listed. The
segments column reports the number of segments. The bytes column reports the
number of bytes those segments require in HALDB. The prefix-incr column

reports the increase in sizes for the segments due to growth in the size of their
prefixes. The length-incr column reports the increase reports on the space
required to store the new physical logical children when converting from virtual
pairing. If the database being analyzed does not contain virtual logical children,
the values in this column will be zeros. The last line of the report contains the
values for all segments combined.
Example 8-1 Sample DFSMAID0 report for a partition

partition 1 :
minimum key =
+0000 d2c1c1f1 f1f2f3f4 |KAA11234 |
maximum key =
+0000 d2f2f3f9 f9f2f3f4 |K2399234 |
segments bytes prefix-incr length-incr

1) 'PRODUCT ' 31567 4040576 252536 0
2) 'INVENT' 103781 8094918 830248 0
3) 'ORDQTY' 171182 10955648 1369456 0
4) 'MFGSPECS' 51115 10938610 408920 0
SUM) 357645 34029752 2861160 0
8.2.2 Adjusting the sizes from the reports

You must add to the sizes of the partitions reported by the utility when allocating
space. The values in the reports do not include space for PHDAM root anchor
points (RAPs), bit maps, free space element anchor points (FSEAPs), or free
space. Typically, space for RAPs, bit maps, and FSEAPs is not significant. Free
space may be. In fact, you may want to increase the amount of free space in your
database. With more free space, you may be able to decrease the frequency of
required reorganizations or eliminate them.
We explain how to use the report with secondary indexes in 8.5.2, “Estimating
the sizes of secondary index partitions” on page 107.
8.2.3 Control statements

SYSIN control statements are used to choose the type of analysis.
KR control statement
KR control statements are used for key range analysis. You specify a statement
for each partition except one with high values. When doing key range analysis,
the utility always assumes that a partition with high values is to be included. Keys
may be specified as either character or hexadecimal. Example 8-2 shows the
specification of keys in both formats. This produces reports for four partitions.
Example 8-2 Key range control statements

KR=C’2500000’
KR=C’5000000’
KR=X’F7F5F0F0F0F0F0F0’
MAX control statement

A MAX control statement is used for analyzing the number of partitions required
if each is limited to the number of bytes of segment data specified. Only one MAX
control statement may be specified. Values up to 2147483647 may be used.
Higher values are reduced to 2147483647. Example 8-3 shows the specification
of a maximum size of 250,000,000 bytes. You should include a SAMPLE control
statement when using MAX to analyze large databases.
The MAX statement should not be used when reading a non-HALDB secondary
index. The number of bytes reported for secondary indexes is inaccurate,
therefore, the MAX statement would provide misleading information.
Example 8-3 Maximum bytes control statement

MAX=250000000
NBR control statement

A NBR control statement is used for analyzing the size and high keys for a
specified number of equally sized partitions. Only one NBR control statement
may be specified. Values from one to 1001 may be specified. Example 8-4 shows
the specification of 12 partitions. You should include a SAMPLE control
statement when using NBR to analyze large databases.
Example 8-4 Number of partitions control statement

NBR=12
SAMPLE control statement

A SAMPLE control statement may be used with a MAX or NBR control
statement. It improves the efficiency of the execution by analyzing a subset of the
records in the database. For large databases, it may significantly reduce the time

required for execution by reducing the amount of data stored and sorted. Sample
values of 20,000 are reasonable for most analyses. Example 8-5 shows the
specification of a sample size of 20,000 database records. SAMPLE values up to
699049 are valid. If you specify a SAMPLE statement, it must be the first control
statement.
Example 8-5 Sample control statement

SAMPLE=20000
The SAMPLE control statement has an optional SEED parameter. It is used to

change the records that are sampled. Normally, you do not need to specify a
seed. If you want to verify an analysis that used a SAMPLE statement, you may
run it again with a different SEED value and compare the reports. The default for
SEED is 1.
8.3 Migrating simple databases

Databases that have no secondary indexes or logical relationships may be
migrated one at a time or in groups. The steps for the migration are:
򐂰 Unload of the existing database using HD Unload and specifying
MIGRATE=YES
򐂰 DBDGEN as a HALDB database
򐂰 Delete database information from the RECONs
򐂰 Definition of the partitions
򐂰 Partition initialization
򐂰 Reload as a HALDB database using HD Reload
When you migrate a database, you will probably want to keep the same database
name. This avoids the need to change PSBs that reference the database. In our
examples and explanations we have assumed that the same name will be used
for the HALDB.
8.3.1 Unloading the existing database

The HD Reorganization Unload utility (DFSURGU0) is used to unload the
existing HDAM or HIDAM database. The unload must use the DBD that defines
the non-HALDB database. The MIGRATE=YES control statement must be
specified. With this statement, HD Unload creates a HALDB prefix for each
unloaded segment in its output file. The prefix includes a generated ILK to be
associated with the segment. OSAM sequential buffering should be used with
OSAM database data sets. This will typically improve the performance of the
unload.
8.3.2 Saving existing database information

Before changing the DBD or deleting information from the RECONs, we
recommend that you save your DBD source and back up your RECONs. If you
need to fall back after attempting a migration, you will need this information.
8.3.3 Changing the DBD

The DBD for your database must be changed to a HALDB database organization
type. You might have to change other DBD specifications. These considerations
are explained in 8.1.1, “DBD changes” on page 90.
8.3.4 Deleting database information from the RECONs

The database information in the RECONs must be deleted. This must be done
before the HALDB database is registered. This registration occurs when the
partitions are defined. The deletion may be done with a DBRC DELETE.DB
command. Figure 8-6 shows the command we used to delete information about
the NORDDB database. If you are migrating a HIDAM database, you will want to
delete the HIDAM index information eventually. This does not have to be done
now. The index for a PHIDAM database is generated automatically and does not
use the HIDAM index database name.
Example 8-6 DBRC command to delete a database from the RECONs

After the DELETE.DB command is processed, the information in the RECONs

about the non-HALDB database is deleted. You cannot use the RECONs to
restore this version of the database. You may want to copy the RECONs before
issuing the command. Then, if necessary, you could use the copy with a
GENJCL.RECOV command and the old DBD to restore the database to its
non-HALDB state.
8.3.5 Defining the partitions

The partitions are defined either with the Partition Definition utility (PDU) or with
DBRC INIT commands. The use of the PDU is explained in Chapter 4, “Partition

Definition utility” on page 45. The use of DBRC INIT commands is explained in
Chapter 5, “Batch definition of HALDB” on page 61. Before defining the partitions
you will need to decide on your partition boundaries. The HALDB Migration Aid
utility (DFSMAIDO) can provide useful information to help make these decisions.
Its use is explained in 8.2, “HALDB Migration Aid utility (DFSMAID0)” on
page 93.
8.3.6 Allocating database data sets

You are responsible for allocating the database data sets used by the databases.
The data set names are created from the data set name prefix that you specify
when you define a partition and the data set name algorithm. This is explained in
1.9, “Naming conventions” on page 17. You must use these names when you
allocate your data sets.
You may be able to use the information generated by the HALDB Migration Aid
utility (DFSMAID0) to assist you in sizing the data sets. It reports the number of
bytes of segment data in each partition. You will need additional space for
segments to be added, free space, and bit maps.
When migrating to HALDB you may want to increase your free space
parameters. Non-HALDB databases sometimes have suboptimal free space
values due to data set size limitations. Additional free space may allow you to
reorganize your databases or partitions less frequently. In some cases, additional
free space could eliminate the need for reorganizations.
Each PHDAM or PHIDAM partition includes an ILDS. This is required even when
there are no secondary indexes or logical relationships. In this case, you can
allocate a very small amount of space for the data set, such as one track.
All ILDSs have fixed-length 50-byte records. Keys are nine bytes at zero offset.
Example 8-7 shows IDCAMS DEFINE statements to allocate an ILDS.
Example 8-7 Sample DEFINE for an ILDS

DEFINE CLUSTER( -
NAME(JOUKO3.HALDB.DB.PEOPLE.L00001) -
INDEXED -
CYL(1 1) -
RECORDSIZE(50 50) -
SHAREOPTIONS(3 3) -
SPEED -
KEY(9,0) -
FREESPACE(10,10) -
CONTROLINTERVALSIZE(8192) -
VOLUMES(TOTIMN) -
)
You must specify REUSE on the DEFINE for all HALDB VSAM data sets other
than ILDSs.
DBDGEN output is useful when allocating PHIDAM primary indexes. The output
of DBDGEN for the PHIDAM database reports required parameters for the
IDCAMS define. This is shown following the label RECOMMENDED VSAM
DEFINE CLUSTER PARAMETERS. The required parameters are the KEY
values, RECORDSIZE values, INDEXED, and REUSE. Example 8-8 shows
IDCAMS DEFINE statements to allocate a PHIDAM primary index data set.
Example 8-8 Sample DEFINE for a PHIDAM primary index data set
DEFINE CLUSTER( -
NAME(JOUKO3.HALDB.DB.RZL.X00001) -
INDEXED -
CYL(10 5) -
RECORDSIZE(20 20) -
SHAREOPTIONS(3 3) -
REUSE -
KEY(14,5) -
FREESPACE(25,10) -
VOLUMES(TOTIMN) -
)
Appendix A, “Sample GENJCL.USER skeletons for HALDB allocation” on

page 281, explains how DBRC GENJCL.USER may be used to create jobs to
allocate database data sets.
8.3.7 Initializing the partitions

The partitions in the HALDB database must be initialized. Either the Database
Prereorganization utility or the HALDB Database Data Set Initialization utility may
be used. We recommend the use of the Prereorganization utility. Partition
initialization is explained in Chapter 6, “Partition initialization” on page 73.
8.3.8 Reloading the HALDB database

The HD Reorganization Reload utility (DFSURGL0) is used to load the HALDB
database. The input to Reload is the output from the Unload utility. When there
are no secondary indexes or logical relationships defined, the Reload utility does
not update the ILDS. Nevertheless, an ILDS must be available for each partition.

Note: Message IEC161I 152-061 is issued for ILDSs because these data sets
are empty when opened. This does not indicate a problem.
8.3.9 Image copying the database data sets

The image copy needed flag is set for each database data set other than ILDSs
and PHIDAM primary indexes. Image copies should be taken of the flagged data
sets.
8.4 ILDS creation options

HD Reload creates an entry in the ILDS for a partition for each segment that is
the target of a secondary index or logical relationship pointer. These entries are
used to heal pointers when they are first used after the load. HD Reload has
three options for the creation of these ILDS entries: They may be created as the
target segments are loaded; they may be created in a multithread process after
all target segments have been loaded; and they may be created outside of the
execution of HD Reload.
8.4.1 HD Reload with no control statement

When you do not specify a control statement in the SYSIN data for HD Reload,
an ILDS entry is created when a target of a secondary index or logical
relationship is inserted in the database. These updates could be slow for three
reasons. First, there may be many CI and CA splits. The inserts in the ILDS are
done in VSAM update mode. When a CI or CA is filled, it must be split by VSAM.
Second, they may be random. The ILDS keys are based on the ILK, which is
based on the location of the target segment in the non-HALDB. Third, this is a
single thread process. These writes are not overlapped with other reload
processing.
8.4.2 HD Reload with an ILDSMULTI control statement

The ILDSMULTI control statement was added to IMS Version 7 by APAR
PQ36991 and to IMS Version 8 by PQ54227. When you specify an ILDSMULTI
control statement in the SYSIN data, HD Reload uses a multi-thread process.
ILDS entries are not written when target segments are loaded. Instead, they are
saved in data spaces. When HD Reload has completed the load of the database,
it sorts the ILDS entries by key within partition. It creates a separate process for
each partition. Each process writes the ILDS entries in VSAM load mode. This
process can be much more efficient than creating the entries as the targets are
loaded. First, the writes of the ILDS entries to different partitions are overlapped.
Second, the writes are sequential. Third, the writes are done in load mode. This
avoids the overhead of CI and CA splits. There is an additional benefit of load
mode: Free space may be created. Future reorganizations could benefit from
having this free space for their updates to the ILDSs.
Do not specify ILDSMULTI if the database has no logical relationships or

secondary indexes defined.
The ILDSMULTI control statement may be used only during migrations. It is

invalid if the input file was not created by an HD Unload execution using either
MIGRATE=YES or MIGRATX=YES.
8.4.3 HD Reload with a NOILDS control statement

The NOILDS control statement was added to IMS Version 7 by APAR PQ36991
and to IMS Version 8 by PQ54227. When you specify a NOILDS control
statement in the SYSIN data, HD Reload does not create entries in the ILDSs.
They must be created by a separate process. The HALDB Index/ILDS Rebuild
utility (DFSPREC0) is used to do this. Separate executions of DFSPREC0 are
required for each partition. These executions may be done in parallel and on
different machines. Obviously, the use of NOILDS improves the performance of
HD Reload since it does not have to write the ILDS entries. On the other hand,
DFSPREC0 must read the partitions to create the ILDSs.
The NOILDS control statement is not limited to migrations. It may be used for any
execution of HD Reload.
8.5 Migrating databases with secondary indexes

IMS does not support HALDB secondary indexes into non-HALDB databases or
non-HALDB secondary indexes into HALDB databases. When you migrate a
database that has secondary indexes, you must also migrate any secondary
indexes that it has.
There are two standard ways to migrate databases with secondary indexes. First,
you can migrate the database and create the HALDB secondary indexes from
the migrated source segments. This method uses the MIGRATX=YES control
statement with HD Unload. We recommend this method. Second, you can
migrate the database and migrate the secondary indexes. This technique
requires that each secondary index be unloaded. The MIGRATE=YES control
statement is used for the unload of the indexed database and the unload of each
secondary index. This method has two disadvantages. Each secondary index
must be read. When reading the secondary index, the indexed database also

must be read. These extra reads can substantially elongate the migration
process. For this reason, we recommend that you not use this method. We only
explain the use of the first method.
There is an alternative to the two standard migration methods. This requires the
use of the IMS Index Builder tool. This tool can build your secondary indexes
from the indexed database. You do not have to migrate the secondary indexes
when you use Index Builder. Instead, you can migrate the indexed database and
then use Index Builder to create the secondary indexes.
8.5.1 Changing DBDs for secondary indexes

You must change the DBDs for your secondary index databases before they are
loaded as HALDB secondary indexes. You may also need to change the DBDs of
some of your indexed databases. We explain all changes required for the
migration to HALDB in 8.1.1, “DBD changes” on page 90. Specific changes for
secondary indexes are explained here.
PSINDEX must be specified on the ACCESS parameter of the DBD statement

for the secondary index.
RKSIZE must be specified on the LCHILD statement of the secondary index.
/SX subsequence fields

If you use a /SX field to create unique keys for a secondary index, you must
adjust your secondary index DBD. In non-HALDB secondary indexes a /SX field
contains a four-byte RBA. In a HALDB secondary index a /SX field contains a
eight-byte ILK. The larger size of the field requires two changes.
򐂰 The BYTES parameter on the SEGM statement in the secondary index DBD
must be increased by four.
򐂰 The BYTES parameter on the FIELD statement for the sequence field in the
secondary index DBD must be increased by four.
Changes are not required to the XDFLD statement or the /SX FIELD statement in
the indexed DBD. The BYTES parameter on the /SX FIELD statement is not
required and is ignored if it is specified.
Example 8-9 shows the DBD for an indexed HDAM database with a /SX field
defined.
Example 8-9 HDAM DBD with /SX field

DBD NAME=VEHICLE,ACCESS=(HDAM,OSAM), X
RMNAME=(DFSHDC40,2,500,)
DATASET DD1=VEHICLE1,BLOCK=1648,SCAN=0
SEGM NAME=AUTO,BYTES=54,PTR=TB
FIELD NAME=(ID,SEQ,U),BYTES=10,START=1
FIELD NAME=MAKE,BYTES=20,START=11
FIELD NAME=MODEL,BYTES=20,START=31
FIELD NAME=YEAR,BYTES=4,START=51
FIELD NAME=/SX1
LCHILD NAME=(MAKEMOD,VEHSI),PTR=INDX
XDFLD NAME=MMIDX,SRCH=(MAKE,MODEL),SUBSEQ=/SX1
DBDGEN
FINISH
END
Example 8-10 shows the DBD for a non-HALDB secondary index that uses the
/SX field defined in Example 8-9 on page 102.
Example 8-10 Non-HALDB secondary index DBD using a /SX field

DBD NAME=VEHSI,ACCESS=INDEX
DATASET DD1=VEHSI1,DEVICE=3390,SIZE=8192
SEGM NAME=MAKEMOD,BYTES=44,PARENT=0
FIELD NAME=(NAMES,SEQ,U),BYTES=44,START=1
LCHILD NAME=(AUTO,VEHICLE),INDEX=MMIDX
DBDGEN
FINISH
END
Example 8-11 shows the DBD from Example 8-9 on page 102 after it has been
converted to HALDB. The only changes are those made for all HDAM DBDs
when they are converted to PHDAM. That is, the ACCESS parameter on the
DBD STATEMENT is changed to PHDAM and the DATASET statement is
deleted.
Example 8-11 PHDAM DBD with /SX field

DBD NAME=VEHICLE,ACCESS=(PHDAM,OSAM), X
RMNAME=(DFSHDC40,2,500,)
SEGM NAME=AUTO,BYTES=54,PTR=TB
FIELD NAME=(ID,SEQ,U),BYTES=10,START=1
FIELD NAME=MAKE,BYTES=20,START=11
FIELD NAME=MODEL,BYTES=20,START=31
FIELD NAME=YEAR,BYTES=4,START=51
FIELD NAME=/SX1
LCHILD NAME=(MAKEMOD,VEHSI),PTR=INDX
XDFLD NAME=MMIDX,SRCH=(MAKE,MODEL),SUBSEQ=/SX1
DBDGEN
FINISH

END
Example 8-12 shows the DBD from Example 8-10 on page 103 after it has been
converted to HALDB. As with other secondary index DBD conversions, the
ACCESS parameter on the DBD statement is changed to PSINDEX, the
DATASET statement is deleted, and the RKSIZE parameter is added to the
LCHILD statement. Since a /SX field is used as a subsequence field, the BYTES
parameters on the SEGM and FIELD statements are incremented by four.
Example 8-12 HALDB secondary index DBD using a /SX field

DBD NAME=VEHSI,ACCESS=PSINDEX
SEGM NAME=MAKEMOD,BYTES=48,PARENT=0
FIELD NAME=(NAMES,SEQ,U),BYTES=48,START=1
LCHILD NAME=(AUTO,VEHICLE),INDEX=MMIDX,RKSIZE=10
DBDGEN
FINISH
END
Symbolic pointers
You cannot have symbolic pointers in a HALDB secondary index. If you currently
have symbolic pointers, you must eliminate them. Symbolic pointers are defined
with a PTR=SYMB parameter on the LCHILD statements in the non-HALDB
secondary index and the indexed database DBDs. You must change the
PTR=SYMB specification in the indexed database DBD to PTR=INDX and either
omit the PTR parameter in the secondary index DBD or specify PTR=SNGL in it.
An illustration of these changes is shown in the following examples.
Example 8-13 shows the non-HALDB secondary index DBD with symbolic
pointing defined. Example 8-14 on page 105 shows the indexed HDAM DBD.
Example 8-15 on page 105 shows the secondary index DBD after it has been
converted to HALDB. PTR=SYMB is deleted from the LCHILD statement.
RKSIZE is added to it. Symbolic pointers are stored in the data area of index
segments. Since it is not in the HALDB secondary index, the BYTES parameter
on the SEGM segment is reduced by the size of the symbolic pointer.
Example 8-16 on page 105 shows the indexed database DBD after it has been
converted to HALDB. PTR=INDX is specified on the LCHILD segment for the
secondary index relationship.
Example 8-13 DBD for secondary index using symbolic pointing

DBD NAME=CONTRSI,ACCESS=INDEX
DATASET DD1=CONTSI,DEVICE=3390,SIZE=8192
SEGM NAME=CONTR,BYTES=26,PARENT=0
FIELD NAME=(CONTRNUM,SEQ,U),BYTES=8,START=1,TYPE=C
LCHILD NAME=(CONTRACT,ENGAGEM),INDEX=CONTRIDX,PTR=SYMB
DBDGEN
FINISH
END
Example 8-14 DBD for indexed database with symbolic pointing

DBD NAME=ENGAGEM,ACCESS=HDAM,RMNAME=(DFSHDC40,1,500,824)
DATASET DD1=ENGAHDAM,BLOCK=1648,SCAN=0
SEGM NAME=CLIENT,BYTES=100,PTR=TWIN
FIELD NAME=(CLNUM,SEQ,U),BYTES=10,START=1,TYPE=C
SEGM NAME=CONTRACT,PARENT=CLIENT,BYTES=60,PTR=TWIN
FIELD NAME=(CONTRNO,SEQ,U),BYTES=8,START=1,TYPE=C
LCHILD NAME=(CONTR,CONTRSI),PTR=SYMB
XDFLD NAME=CONTRIDX,SRCH=CONTRNO
DBDGEN
FINISH
END
Example 8-15 DBD for HALDB secondary index

DBD NAME=CONTRSI,ACCESS=PSINDEX
LCHILD NAME=(CONTRACT,ENGAGEM),INDEX=CONTRIDX,RKSIZE=10
DBDGEN
FINISH
END
Example 8-16 DBD for HALDB indexed database

DBD NAME=ENGAGEM,ACCESS=PHDAM,RMNAME=(DFSHDC40,1,500,824)
LCHILD NAME=(CONTR,CONTRSI),PTR=INDX
DBDGEN
FINISH
END
If you have applications that process a secondary index using symbolic pointers
as a database, you will want to make other changes to your DBDs. We explain
these changes in 11.5.2, “Secondary indexes with symbolic pointers” on
page 202.

Non-unique keys
HALDB secondary indexes must have unique keys. Non-HALDB secondary
indexes may have non-unique keys. Non-unique keys are specified with an M as
the third subparameter of the NAME parameter for the key field in a secondary
index DBD. Non-unique keys require an overflow data set to hold duplicate keys.
This is specified in the OVFLW parameter of the DBD DATASET statement. If
your non-HALDB secondary index DBD has these definitions, they must be
changed as part of the conversion. Example 8-17 shows a non-HALDB
secondary index DBD with non-unique keys. Example 8-18 shows the DBD for
the secondary index after it has been converted to HALDB. Since the index key
must be unique, you will probably need to add a subsequence field. Typically, you
would do this by adding a /SX system-related field to the XDFLD statement in the
indexed database DBD. This would add eight bytes to the length of the sequence
field in the index. In the examples, the BYTES parameter in the FIELD statement
is increased by eight for this conversion.
Example 8-17 Non-HALDB secondary index DBD with non-unique keys

DBD NAME=XSI3,ACCESS=INDEX
DATASET DD1=XSI301,OVFLW=XSI302
SEGM NAME=XSNAM,BYTES=6,PARENT=0
FIELD NAME=(XSNAME,SEQ,M),START=1,BYTES=6
LCHILD NAME=(PERF,XPER01),INDEX=NAMX1,POINTER=SNGL
DBDGEN
FINISH
END
Example 8-18 HALDB secondary index DBD with unique keys

DBD NAME=XSI3,ACCESS=PSINDEX
SEGM NAME=XSNAM,BYTES=14,PARENT=0
FIELD NAME=(XSNAME,SEQ,U),START=1,BYTES=14
LCHILD NAME=(PERF,XPER01),INDEX=NAMX1,POINTER=SNGL,RKSIZE=12
DBDGEN
FINISH
END
Shared secondary indexes

You cannot have a shared secondary index with HALDB. Shared indexes are
defined with a CONST parameter on the XDFLD statement of the indexed
database. You must delete this parameter when you migrate to the database.
8.5.2 Estimating the sizes of secondary index partitions
HALDB secondary index records are often much larger than non-HALDB
secondary index records. There are two reasons for this. First, the pointers are
different. Each HALDB secondary index record includes the 28-byte extended
pointer set (EPS). Non-HALDB secondary index records have either a 4-byte
direct pointer or a symbolic pointer. Symbolic pointers are the length of the
concatenated key. Second, the root key of the target segment is stored in the
HALDB secondary index record. This is not done with non-HALDB secondary
index records. When allocating your HALDB secondary index partitions, you
must account for this increased size.
You must define partition boundaries for secondary indexes. The HALDB
Migration Aid (DFSMAID0) utility may be used to report on secondary indexes.
The total size of the secondary index is usually easy to estimate. All secondary
index entries are fixed length. The DBDGEN reports this length. It is the
RECORDSIZE in the RECOMMENDED VSAM DEFINE CLUSTER
PARAMETERS of the HALDB DBDGEN output listing. The number of entries
does not change during the migration. You can find the current number of entries
from the output of several utilities. When you recreate a secondary index as part
of a reorganization process, the HISAM Reload utility or any tool that you use
reports this information. Alternatively, you can use the REC-TOTAL value from
LISTCAT to find the number of records in the existing secondary index data set.
Finally, you may use the HALDB Migration Aid utility to find the number of
records. The number of records, their length, and free space requirements may
be used to estimate the size of an entire HALDB secondary index. For example,
if you had 1,000,000 entries, a record length of 48 bytes, and wanted an
additional 25 percent for free space, the index would require 60,000,000 bytes.
The size requirements of each partition will depend on the number of index
entries in the partition. You may estimate this from the output of the HALDB
Migration Aid utility (DFSMAID0) or from other means. If you already know the
number of secondary index entries in a key range you do not need to use the
Migration Aid utility. Otherwise, you should use the utility.
The Migration Aid utility provides accurate information about the number of
records in a key range and key range boundaries. It does not provide accurate
information about the number of bytes required for a segment or partition. For
this reason, you should not use the MAX control statement when reading a
secondary index. If you know the number of partitions you want, use the NBR
control statement. If you know the high keys that you want, use KR control
statements. When an NBR statement is used, the high keys that are reported for
each partition are accurate. When using KR control statements, the number of
segments reported for each partition is accurate.

Example 8-19 on page 108 shows some of the output of the Migration Aid utility
for a secondary index. This was generated with a NBR=4 control statement. We
have omitted the reports for partitions 1, 2, and 3. Only the report for partition 4
and the total database are shown. The numbers of segments reported in the
segments column are correct. Partition 4 will contain 158,518 secondary index
segments and the entire index will contain 634,078 segments. The information in
the bytes and prefix-incr columns is wrong. Nevertheless, all we need is the
number of segments in a partition to calculate its space requirements.
Example 8-19 DFSMAID0 output for a secondary index

partition 4 :
minimum key =
+0000 f0f0f1f6 f0f8f1f1 f5f0f0f0 f0f0f1f3 |0016081150000013|

+0010 f1f7 |17 |
maximum key =
+0000 f0f0f2f1 f1f0f3f0 f0f0f0f0 f0f0f0f5 |0021103000000005|

+0010 f4f6 |46 |

1) 'ORDRCUST' 158518 3804432 1268144 0
SUM) 158518 3804432 1268144 0
--------------------------------------------------------------------------------
sum of partitions:
1) 'ORDRCUST' 634078 15217872 5072624 0
SUM) 634078 15217872 5072624 0
-------------------------------------------------------------------------------
You may use the record size reported by DBDGEN and the number of segments
for a partition to estimate the size requirement for the partition. Of course, you
can add free space and room for expansion of the partition.
Restriction: DFSMAID0 reports of bytes and prefix-incr for secondary

indexes should never be used. They are inaccurate. DFSMAID0 may be used
to find high keys and the number of secondary index entries for each partition.
These are accurate.
8.5.3 Unloading databases with secondary indexes
When you unload a database with secondary indexes for migration to HALDB,
you can create unload files for its secondary indexes. This is done when you
specify the MIGRATX=YES control statement with HD Unload. This technique
does not read the secondary indexes. DD statements for the secondary index
data sets are not required. The information to create secondary index entries is
generated from the source segments. The unload files for secondary indexes
must be sorted before they are used to load the secondary indexes. The output
of HD Unload includes sort control statements to be used for the sorts.
Figure 8-1 illustrates a use of MIGRATX=YES to migrate a database and its two
secondary indexes.
Non-HALDB HALDB
Secondary
Index 1
Secondary
Sort HD Reload Index 1
Report
Sort
Ctl.
Indexed
Database
Indexed
HD Unload HD Reload Database
Sort
Secondary Ctl.
Index 2
Secondary
Sort HD Reload Index 2
Figure 8-1 Using MIGRATX=YES to migrate a database with secondary indexes
OSAM sequential buffering should be used with OSAM database data sets. This
will typically improve the performance of the unload.
The MIGRATX=YES technique does not preserve any user data in the secondary
indexes. This is rarely a problem because most installations do not have user
data. They avoid user data because it is lost when the reorganization of a
non-HALDB database requires the rebuilding of its secondary indexes. These
rebuilds do not preserve user data.
The SYSPRINT listing of HD Unload includes a Work File Statistics report. It tells
you which output files are to be used for each secondary index. This is not based
on secondary index names. You must use the report to understand this
relationship. The utility assigns DFSWRK01 to the first secondary index

relationship it encounters in the unload. The report also lists the number of
segments in the secondary index files and the sort field sizes and offsets. A
sample report is shown in Example 8-20. In it DFSWRK01 contains the records
for secondary index STUNUM, and DFSWRK02 contains the records for
secondary index STUCRT.
Example 8-20 Work file statistics report with MIGRATX=YES

W O R K F I L E S T A T I S T I C S
SINAME WFNAME SFNAME RCDTOTAL OFFSET LENGTH
STUNUM DFSWRK01 DFSSRT01 000087012 0124 0006
STUCRT DFSWRK02 DFSSRT02 000010702 0124 0010
You must sort each secondary index output file. Use the sort control statements
in the DFSSRTnn files for these sorts. Example 8-21 shows sample JCL for the
sorting of the STUNUM secondary index. The SORTIN DD specifies the
DFSWRK01 output from HD Unload. The SYSIN DD specifies the DFSSRT01
output from HD Unload. The output of this sort is input to HD Reload for the
STUNUM secondary index.
Example 8-21 Sample JCL to sort a secondary index output file

//*******************************************************************
//* SORT THE STUNUM SEC. INDEX RECORDS FROM MIGRATX=YES
//*******************************************************************
//*
//SORT01 EXEC PGM=SORT,REGION=2048K,PARM='CORE=MAX'
//SORTLIB DD DSN=SYS1.SORTLIB,DISP=SHR
//SYSOUT DD SYSOUT=*
//SORTWK01 DD UNIT=SYSDA,SPACE=(CYL,(50,5),,CONTIG)
//SYSIN DD DSN=JOUKO3.STUDENT.MIGR.SRT01,DISP=OLD
//SORTIN DD DSN=JOUKO3.STUDENT.MIGR.WRK01,DISP=OLD
//SORTOUT DD DSN=JOUKO3.STUDENT.MIGR.WRK01.SORTED,DISP=(NEW,CATLG),
// UNIT=3390,VOL=SER=TOTIMN,
// SPACE=(CYL,(50,10),RLSE)
The IMS Index Builder alternative

If you use the IMS Index Builder, you have the option of not migrating your
secondary indexes. Instead, you can migrate just your indexed database and use
Index Builder to create the secondary indexes. Since you do not need the unload
files for the secondary indexes, do not use MIGRATX=YES. Instead, use
MIGRATE=YES. This is the technique that is used when there are no secondary
indexes. It is explained in 8.3.1, “Unloading the existing database” on page 96.
8.5.4 Allocating secondary index data sets

DBDGEN output is useful when allocating secondary index data sets. The output
of DBDGEN for the secondary index reports required parameters for the
IDCAMS define. This is shown following the label RECOMMENDED VSAM
DEFINE CLUSTER PARAMETERS. The report includes misleading text that
states *NOTE1 - THIS IS THE VSAM DEFINITION OF A HALDB PRIMARY INDEX.
Actually, it is the VSAM definition of a HALDB secondary index. The required
parameters are the KEY values, RECORDSIZE values, INDEXED, and REUSE.
Example 8-22 shows IDCAMS DEFINE statements to allocate a secondary index
data set.
Example 8-22 Sample DEFINE for a secondary index data set

DEFINE CLUSTER( -
NAME(JOUKO3.HALDB.DB.PEOSKSI.A00001) -
INDEXED -
RECORDSIZE(86 86) -
CYL(20 5) -
SHAREOPTIONS(3 3) -
REUSE -
KEY(29,50) -
FREESPACE(10,10) -
VOLUMES(TOTIMN) -
)
8.5.5 Loading databases and their secondary indexes

As with other databases, HD Reload is used to load databases with secondary
indexes and their secondary indexes. When a segment that is the target of a
secondary index is loaded, an entry is created in the ILDS for its partition.
Otherwise, there are no special considerations for loading indexed databases.
The sorted output files from HD Unload are used as input to HD Reload when
creating a secondary index as part of the migration. Before loading a secondary
index, you must modify its DBD, define its partitions, allocate its data sets, and
initialize its partitions. These are the same steps required for migrating any other
databases.
HD Reload has options for creating ILDS entries. We explain these options in
8.4, “ILDS creation options” on page 100.

The IMS Index Builder alternative
If you use the IMS Index Builder, you do not use HD Reload to create the
secondary indexes. Instead, Index Builder reads the migrated indexed database,
and creates the secondary indexes.
Converting to /SX fields for uniqueness

If you add a /SX field to create uniqueness you need to add three steps before
you reload the secondary index. These steps are required to place the unload
records in the correct subsequence order. The steps are executed in place of the
sort explained in 8.5.3, “Unloading databases with secondary indexes” on
page 109. The steps are executed before the execution of HD Reload for the
secondary index. These steps are:
1. Split the file into three separate files containing the header record, the unload
records, and the trailer record. Example 8-23 shows sort JCL and control
statements for this step.
Example 8-23 Step to split the unload records

//SORTIN DD DSN=UNLOAD.OUTPUT,DISP=SHR
//HEADER DD DSN=HEADER.FILE,DISP=(NEW,PASS)
//TRAILER DD DSN=TRAILER.FILE,DISP=(NEW,PASS)
//ULCOPY DD DSN=UNLOAD.COPY,DISP=(NEW,PASS)
//SYSIN DD *
OPTION COPY
OUTFIL INCLUDE=(5,2,CH,EQ,X'0080'),FNAMES=HEADER
OUTFIL INCLUDE=(5,2,CH,EQ,X'0090'),FNAMES=TRAILER
OUTFIL SAVE,FNAMES=ULCOPY
RECORD TYPE=V
2. Sort the file containing unload records in the /SX subsequence order. You
must calculate the offset and the size of the sort field. The offset is 63 bytes
plus the size of the root key of the indexed database. The sort field size is the
size of the secondary index search field plus eight bytes. Example 8-24
shows sort JCL and control statements for this step. In this example, the
offset is 73 bytes and the sort field size is 18 bytes.
Example 8-24 Step to sort the unload records

//SORTIN DD DSN=UNLOAD.COPY,DISP=SHR
//SORTOUT DD DSN=UNLOAD.SORTED1,DISP=(,CATLG),
// UNIT=SYSDA,VOL=SER=000000,SPACE=(CYL,(1,5))
//SYSIN DD *
SORT FIELDS=(73,18,CH,A),FILSZ=E1000
RECORD TYPE=V
END
3. Merge the records into one file. Example 8-25 shows sort JCL and control
statements for this step.
Example 8-25 Step to merge the unload records

//SORTIN DD DSN=UNLOAD.HEADER,DISP=(OLD,DELETE),
// DD DSN=UNLOAD.SORTED1,DISP=(OLD,DELETE),
// DD DSN=UNLOAD.TRAILER,DISP=(OLD,DELETE),
//SORTOUT DD DSN=UNLOAD.SORTED2,DISP=(NEW,KEEP),
//SYSOUT DD SYSOUT=A
//SYSIN DD *
OPTION COPY
END
8.5.6 Secondary index pointers after the migration

After the migration reload the secondary index pointers are not healed. That is,
the direct pointers cannot be used. When the pointers are needed, the ILDS of
the target partition must be used to find the target segment.
8.5.7 Using MIGRATE=YES with secondary indexes

We highly recommend that you migrate databases with secondary indexes by
using the MIGRATX=YES control statement with HD Unload. The
MIGRATE=YES control statement is valid, but should not be used.
When an HD Unload MIGRATE=YES control statement is used to unload a

database with secondary indexes, output files are not created for the secondary
indexes. The secondary indexes must be unloaded with separate executions of
HD Unload. When a secondary index is unloaded, the indexed database must
also be read. Each secondary index record requires a read of the indexed
database. These reads are random. This can elongate the unload process
significantly. Figure 8-2 on page 114 illustrates the use of MIGRATE=YES to
migrate a database and its two secondary indexes. There are three executions of
HD Unload. Each execution must read the indexed database.

Non-HALDB HALDB
Secondary
Index 1
Secondary
HD Unload HD Reload Index 1
Indexed
Database
Indexed
HD Unload HD Reload Database
Secondary
Index 2
Secondary
HD Unload HD Reload Index 2
Figure 8-2 Using MIGRATE=YES to migrate a database with secondary indexes
8.6 Migrating databases with logical relationships

IMS does not support logical relationships between HALDB and non-HALDB
databases. Databases that are logically related to each other must be migrated
together. This section describes the special considerations for migrating logically
related databases. If a database also has secondary indexes, the considerations
discussed in “Migrating databases with secondary indexes” on page 101 must
also be observed.
8.6.1 Changing DBDs with logical relationships

There are three types of logical relationships for non-HALDB databases. They
are:
򐂰 Unidirectional
򐂰 Bidirectional with physical pairing
򐂰 Bidirectional with virtual pairing
HALDB supports unidirectional and bidirectional with physical pairing. It does not
support virtual pairing. The following types of migrations of logical relationships
are supported:
򐂰 Unidirectional to unidirectional
򐂰 Bidirectional with physical pairing to bidirectional with physical pairing
򐂰 Bidirectional with virtual pairing to bidirectional with physical pairing
8.6.2 Changing DBDs with logical relationships
The DBDs for databases with logical relationships may have to be modified when
converting to HALDB. There are two reasons for this. First, some pointer options
are not supported with HALDB. Second, HALDB does not support virtual paring
with bidirectional logical relationships.
The DBDs for logical databases are not changed when the physical databases
they reference are migrated. DBDs for logical databases have
ACCESS=LOGICAL in their DBD statements.
Changing pointer options for logical relationships

Logically related databases may require changes to pointers in addition to those
required for databases that do not have logical relationships.
HALDB does not use symbolic pointers. Any logical relationship implemented
with symbolic pointing must be changed. This is done by adding the LPARNT
keyword to the PTR parameter on the SEGM statement for the logical child. Even
though symbolic pointing is not used, the concatenated key of the logical parent
is stored in the logical child. This is done no matter what keywords you use in the
PARENT parameter. That is, PHYSICAL or P will be used even if you specify
VIRTUAL or V. To avoid the warning message that is issued when VIRTUAL or V
is specified, you should specify the PHYSICAL or P keyword.
Recognizing virtual pairing

You may determine if you have virtual pairing by examining your DBDs. If your
HDAM or HIDAM database contains a SEGM statement with a SOURCE
parameter, this segment is a virtual logical child. The first subparameter of
SOURCE identifies the real logical child. The third subparameter identifies the
database in which the real logical child resides. In Example 8-26 the NAMESKIL
segment in the DBD contains a SOURCE parameter. It is a virtual logical child.
The source parameter references the SKILNAME segment in the SKILLINV
database in Example 8-27 on page 116. SKILNAME is a real logical child.
Example 8-26 HIDAM DBD with a logical relationship

DBD NAME=PAYROLDB,ACCESS=HIDAM
DATASET DD1=PAYHIDAM,BLOCK=1648,SCAN=3
SEGM NAME=NAMEMAST,PTR=TWINBWD,RULES=(VVV), X
BYTES=150
LCHILD NAME=(INDEX,INDEXDB),PTR=INDX
LCHILD NAME=(SKILNAME,SKILLINV),PAIR=NAMESKIL,PTR=DBLE
FIELD NAME=(EMPLOYEE,SEQ,U),BYTES=60,START=1,TYPE=C
SEGM NAME=NAMESKIL,PARENT=NAMEMAST,PTR=PAIRED, X
SOURCE=((SKILNAME,DATA,SKILLINV))
FIELD NAME=(TYPE,SEQ,U),BYTES=21,START=1,TYPE=C

DBDGEN
FINISH
END
Example 8-27 HDAM DBD with a logical relationship

DBD NAME=SKILLINV,ACCESS=HDAM,RMNAME=(DFSHDC40,1,500,824)
DATASET DD1=SKILHDAM,BLOCK=1648,SCAN=0
SEGM NAME=SKILMAST,BYTES=31,PTR=TWINBWD
SEGM NAME=SKILNAME, X
PARENT=((SKILMAST,DBLE),(NAMEMAST,P,PAYROLDB)), X
BYTES=80,PTR=(LPARNT,LTWINBWD,TWIN)
FIELD NAME=(EMPLOYEE,SEQ,U),START=1,BYTES=60,TYPE=C
DBDGEN
FINISH
END
Making changes to the DBDs for physical pairing

To convert a virtually paired logical relationship to physical pairing, the following
must be done:
򐂰 Add an LCHILD statement under the parent of the segment that was the real
logical child. In Example 8-26 on page 115 this is the SKILMAST segment. In
this LCHILD statement do the following:
– The new LCHILD statement must have a NAME parameter that specifies
the segment that had the SOURCE parameter. In the example this is
NAMESKIL.
– The PAIR parameter must specify the segment that was specified in the
SOURCE parameter. This was the real logical child. In the example this is
SKILNAME.
򐂰 Find the LCHILD statement that references the segment that was the real
logical child. In Example 8-26 on page 115 this is the LCHILD statement in
the PAYROLDB database that includes NAME=(SKILNAME,SKILLINV). In
this LCHILD statement do the following:
a. Delete the PTR parameter.
b. If there is a RULES parameter, delete it.
򐂰 Find the SEGM statement that defined the virtual logical child. In
Example 8-26 on page 115 this is the NAMESKIL segment. In this SEGM
statement make the following changes:
a. Add a BYTES parameter. The lengths of the fixed intersection data in the
paired logical children must be the same. The BYTES parameter includes
both the fixed intersection data and the logical parent's concatenated key
(LPCK). The BYTES value for this segment may be calculated by taking
the BYTES value from the paired logical child's SEGM statement, adding
the LPCK size for this segment, and subtracting the LPCK size for the
paired logical child. In the example, the paired logical child for the
NAMESKIL segment is the SKILNAME segment. Its size is 80 bytes. The
LPCK for the NAMESKIL segment is the TYPE field in the SKILMAST
segment. Its size is 21 bytes. The LPCK for the SKILNAME segment is the
EMPLOYEE field in the NAMEMAST segment. Its size is 60 bytes. The
BYTES parameter for the NAMESKIL segment should be 41, which is 80 +
21 - 60.
b. Delete the SOURCE parameter.
c. Change the PARENT parameter so that it also references its logical
parent. In the example this is the SKILMAST segment. The PARENT
parameter should also include the PHYSICAL or P keyword.
d. Change the PTR parameter specifications. Include PAIRED. Include
TWIN, T, TWINBWD, TB, NOTWIN, or NT. Do not use NOTWIN or NT
unless you will have a maximum of only one instance of this segment type
under a parent.
򐂰 Find the SEGM statement for segment that was the real logical child. In
Example 8-26 on page 115 this is the SKILNAME segment. In this SEGM
statement make the following changes:
– Change the PARENT parameter so that neither SNGL nor DBLE is
specified. If VIRTUAL or V was specified, change this to PHYSICAL or P.
– Change the PTR parameter specifications. Delete LTWIN, LT, LTWINBWD,
or LTB keywords if they exist. Specify both the LPARNT and PAIRED
keywords.
Example 8-28 and Example 8-29 on page 118 show the DBDs from
Example 8-26 on page 115 and Example 8-27 on page 116 after they have been
converted to HALDBs with physical pairing. In addition to the changes made for
logical relationships, other changes for the conversion to HALDB have been
made. These are changes in the ACCESS parameter on the DBD statements
and deletion of the DATASET statements.
Example 8-28 PHIDAM DBD with a logical relationship

DBD NAME=PAYROLDB,ACCESS=PHIDAM
SEGM NAME=NAMEMAST,PTR=TWINBWD,RULES=(VVV), X
BYTES=150
LCHILD NAME=(SKILNAME,SKILLINV),PAIR=NAMESKIL
SEGM NAME=NAMESKIL, X
PARENT=((NAMEMAST),(SKILMAST,P,SKILLINV)), X
BYTES=41,PTR=(TWIN,PAIRED)

DBDGEN
FINISH
END
Example 8-29 PHDAM DBD with a logical relationship

DBD NAME=SKILLINV,ACCESS=PHDAM,RMNAME=(DFSHDC40,1,500,824)
SEGM NAME=SKILMAST,BYTES=31,PTR=TWINBWD
LCHILD NAME=(NAMESKIL,PAYROLDB),PAIR=SKILNAME
SEGM NAME=SKILNAME, X
PARENT=((SKILMAST),(NAMEMAST,P,PAYROLDB)), X
BYTES=80,PTR=(TWIN,PAIRED)
FIELD NAME=(EMPLOYEE,SEQ,U),START=1,BYTES=60,TYPE=C
DBDGEN
FINISH
END
8.6.3 Unloading logically related databases for migration

When a migration unload of a database with a logical child is executed, its
logically related database may also be read. This is done when any of the
following conditions exists:
򐂰 Physical pairing is used.
򐂰 The VIRTUAL option is specified (the logical parent’s concatenated key is not
stored in the logical child).
򐂰 Symbolic pointing is used.
򐂰 The database being unloaded contains the virtual logical child.
Stated another way, there are only two cases where the logically related
database is not read. They are:
򐂰 The database being unloaded contains the real logical child of a virtually
paired relationship that uses direct pointing, and the logical parent’s
concatenated key is stored in this logical child.
򐂰 The logical relationship is unidirectional using direct pointing, and the logical
parent’s concatenated key is stored in the logical child.
When reading the logically related database, either the logical parent or the
paired logical child of each logical child is read. That is, each instance of a logical
relationship requires a read of the related data. These reads are random. Each
read may require a physical I/O. This can greatly increase the elapsed time of the
unload.
Buffer pools must be provided for any logically related database that is read by
the unload. The DFSVSAMP DD statement for HD Database Unload
(DFSURGU0) must include definitions for these buffers.
Figure 8-3 illustrates a migration of two logically related databases to HALDBs.
HALDB
Non-HALDB
Database 1
Database 1
HD Unload HD Reload
Logically
Related
DBs Database 2
HD Unload HD Reload
Database 2
Figure 8-3 Migrating databases with logical relationships
8.6.4 Loading logically related databases for migration

As with other databases, HD Reload is used to load databases with logical
relationships during the migration. When a logical parent of a unidirectional
relationship is loaded or a logical child of a bidirectional relationship is loaded, an
entry is created in the ILDS for its partition. Otherwise, there are no special
considerations for loading logically related databases during a migration.
HD Reload has options for creating ILDS entries. We explain these options in
8.4, “ILDS creation options” on page 100.
8.6.5 Logical relationship pointers after the migration

After the migration reload the logical relationship pointers are not healed. That is,
the direct pointers cannot be used. When the pointers are needed, the ILDS of
the target partition must be used to find the logical parent or paired logical child.

8.7 Migrating from PDB or PDF
Some installations have used the IMS/ESA Partition Support Product (PDB) or
similar capabilities in the Partitioned Database Facility (PDF) product from Neon
Enterprise Software, Inc. to partition databases before HALDB was available.
These databases may be migrated to HALDB. The same migration techniques
may be used with either PDB or PDF.
8.7.1 Partition selection

PDB partition selection depends on the type of database organization. HIDAM,
HISAM, and secondary indexes are partitioned by key range. The HDAM
randomization routine selects the partition for HDAM. The randomizer can use
either of two techniques. First, it can assign a range of RAPs to a partition.
Second, it can have two stages. In the first stage a partition is selected. In the
second stage a RAP within the partition is selected.
When migrating from PDB to HALDB you may want to maintain the same
partition boundaries. If so, a migration of a HIDAM database or secondary index
would use key range partitioning with HALDB. A migration of an HDAM database
would use a HALDB partition selection exit routine which assigns a root to a
partition just as the randomization routine assigned a root to a PDB RAP range.
This means that the logic used in the randomization routine would be moved to
the partition selection exit routine.
When migrating from PDB to HALDB you may want to change partition
boundaries. If so, you can use any scheme acceptable to HALDB. Most users
choose key range partitioning, so a typical migration from HDAM with PDB to
PHDAM would change partitioning to key range partitioning.
8.7.2 DBD changes

PDB databases are defined by DBDs. These DBDs do not contain DATASET
statements, but contain PART statements. There is a PART statement for each
partition. When migrating from PDB to HALDB you should delete these PART
statements from the DBD. Other changes are the same that you would make for
other IMS databases as described in 8.1.1, “DBD changes” on page 90. The
PART statements for HIDAM, HISAM, and secondary index PDB databases
contain the high key value. If you want use the same partition boundaries with
HALDB, you should use these values in the Partition Definition utility or your
DBRC INIT.PART statements.
If HISAM is used with PDB, it must be converted to PHIDAM.
8.7.3 Using standard IMS utilities for the migration
You may migrate a PDB HDAM or HIDAM database as you would migrate other
HDAM or HIDAM databases. This uses the IMS HD Unload and IMS HD Reload
utilities. HD Unload will read an entire PDB database. It cannot be used to unload
a single partition. Using this technique you cannot migrate partitions in parallel. If
you have secondary indexes you must use the MIGRATX=YES statement with
HD Unload.
You cannot use this technique to migrate a PDB HISAM database to HALDB.
The MIGRATE=YES and MIGRATX=YES control statements are not valid when
unloading a HISAM database.
8.7.4 Using application programs for the migration

If you maintain the same partition boundaries when migrating to HALDB, you
may want to migrate the partitions of your PDB database in parallel. One way of
doing this is by using application programs. Each instance of an application
program would read a different partition and create an output file of its segments.
Another program reads the file and inserts in the segments in the HALDB
database. Different instances of this program would read the different files.
Figure 8-4 illustrates this technique.
PDB HALDB
Application Application
Program Program
Program Program
Program Program
Figure 8-4 Migration from PDB using application programs with PROCOPT=L
If you modify your partition boundaries, the load programs cannot use
PROCOPT=L. If you are migrating from PDB HDAM to PHDAM, you are unlikely
to maintain the same partition boundaries. If so, you cannot use PROCOPT=L for
each program. Only one program can load segments (PROCOPT=L) in a
partition. Multiple programs may insert segments in the same partition when
using PROCOPT=I. Figure 8-5 on page 122 illustrates this technique.

HALDB
PDB
Program Program
Program Program
Program Program
Figure 8-5 Migration from PDB using application programs with PROCOPT=I
These techniques are valid for migrating PDB HISAM databases to PHIDAM.
8.7.5 Using PSU Unload and application programs for the migration
PDB has its own unload and reload utilities. They are PSU Unload and PSU
Reload. They are used to reorganize individual partitions or reorganize partitions
in parallel. The PSU Unload utility does not have the capability to create an
unload data set with HALDB headers. In other words, it does not have the
capabilities the HD Unload has when MIGRATE=YES or MIGRATX=YES is used.
There are no such control statements for the PSU Unload utility. Nevertheless,
you may be able to use PSU Unload to avoid writing the application unload
programs shown in Figure 8-4 on page 121 and Figure 8-5. This is shown in
Figure 8-6 on page 123. Each instance of PSU Unload creates an unload file.
This file is used as input to the application program that inserts the segments in
the HALDB database. The file contains a header record, one record for each
segment, and a trailer record. The segment record includes the segment name
and the segment data. The output of PSU Unload may be mapped by using the
DSECT in the IMS DFSURGUF macro in SDFSMAC. If the same partition
boundaries are maintained, the load programs may use PROCOPT=L. If not,
PROCOPT=I must be used to avoid loading a partition multiple times.
PDB HALDB
PSU Application
Unload Program
PSU Application
Unload Program
PSU Application
Unload Program
Figure 8-6 Migration from PDB using PSU Unload
This technique is valid for migrating PDB HISAM databases to PHIDAM.
You cannot use HD Reload or PSU Reload to load the HALDB when using PSU
Unload to unload the PDB database. Migration with HD Reload requires that the
input file contain HALDB prefixes. PSU Unload does not create these prefixes.
PSU Reload cannot load a HALDB database.
8.8 Migrating from user partitioning

Some installations have done their own partitioning. Typically, they did this to
address database size limitations before PDB or HALDB was available. This
technique is called user partitioning. With this technique the installation creates
multiple IMS databases. Each database contains a portion of the records that the
application wants to view as one database.
If you have user partitioning, you will probably want to migrate these databases
to a single HALDB database. The migration may be done with user-written
application programs or a combination of the user programs and the HD Unload
utility.
8.8.1 Using application programs for the migration

You may write one or more application programs to migrate a user-partitioned
database to HALDB. Figure 8-7 on page 124 illustrates the use of one
application program. The program reads the segments from the existing
databases and inserts them in the HALDB database. PROCOPT=L may be used
for the HALDB database. Since the application must have PCBs for each of these
databases, the HALDB database must have a different DBD name from any of
the non-HALDB databases. Partition boundaries may be different in the two
implementations.

DBs HALDB
Application
Program
Figure 8-7 Migration from user partitioning with a single program
Alternatively, you may use multiple application program executions to migrate

from user partitioning. Figure 8-8 illustrates an implementation of this technique.
Separate instances of the same application program are used to read the
segments in existing databases. Each instance creates an output file of its
segments. These files are read by separate instances of another application
program, which inserts the segments in the HALDB using PROCOPT=L.
Segments loaded by different instances of the insert programs cannot be placed
in the same partition. You cannot load a partition more than once.
DBs HALDB
Program Program
Program Program
Program Program
Figure 8-8 Migration from user partitioning with multiple load programs
If segments from two different user-partitioned databases might be inserted in

the same HALDB partition, PROCOPT=I must be used. PROCOPT=I avoids the
restriction that a partition cannot be loaded more than one time. This technique is
illustrated in Figure 8-9 on page 125.
HALDB
DBs
Program Program
Program Program
Program Program
Figure 8-9 Migration from user partitioning with multiple insert programs
8.8.2 Using HD Unload and application programs for the migration

The techniques in Figure 8-8 on page 124 and Figure 8-9 may be modified to
use the HD Unload utility. This eliminates the need to write an application
program to read the databases. This is shown in Figure 8-10. Each instance of
HD Unload creates an unload file. This file is used as input to the application
program that inserts the segments in the HALDB database. The file contains a
header record, one record for each segment, and a trailer record. The segment
record includes the segment name and the segment data. The output of HD
Unload may be mapped by using the DSECT in the IMS DFSURGUF macro in
SDFSMAC. Do not include a MIGRATE=YES or MIGRATX=YES control
statement for the unloads. This would add extra overhead to their executions.
DBs HALDB
Application
HD Unload
Program
Application
HD Unload
Program
Application
HD Unload
Program
Figure 8-10 Migration from user partitioning using HD Unload

8.8.3 Warning on the use of HD Reload
HD Reload should not be substituted for the application programs shown in
Figure 8-10 on page 125. When you use HD Reload in conjunction with HD
Unload with a MIGRATE=YES of MIGRATX=YES control statement, the ILKs
created for inserted segments are based on their locations in the non-HALDB
databases. These ILKs use the relative byte address (RBA). These RBAs would
not be unique with user partitioning since the same RBA could exist in multiple
databases. If two segments with the same ILK were ever placed in the same
partition, the HALDB self-healing pointer scheme would fail. This is a database
integrity exposure.
8.9 Fallback to non-HALDB

IMS provides a capability to convert a HALDB database to its previous state as a
non-HALDB database.
If you have not updated the database after a migration to HALDB, you would
restore the non-HALDB database from the last image copy and any necessary
logs. Of course you would also have to change the DBD to its former
specifications and delete and reregister the database to DBRC.
If you have updated the database after a migration to HALDB, you may not be
able to fall back. If the data will no longer fit in the space limitations of a
non-HALDB database, fallback would not be possible. If the fallback is possible,
the steps for this fallback are the following:
1. Unload the database using HD Reload with a FALLBACK=YES control
statement.
2. Perform DBDGEN.
3. Delete the HALDB information for the database from DBRC.
4. Register the non-HALDB database to DBRC.
5. Execute Database Prereorganization using the DBR control statement.
6. Reload the database using HD Reload.
The database data sets must be image copied after the fallback.
If the database has secondary indexes, its secondary indexes must be rebuilt.
If the database has logically related databases they must be converted at the
same time. See 8.9.2, “Fallback with logical relationships” on page 127, for
additional considerations with logically related databases.
8.9.1 Fallback with secondary indexes
Secondary indexes are rebuilt as part of the fallback process for the indexed
database. Fallback uses the same process that you use when reorganizing the
non-HALDB database. This can be the execution of Prefix Resolution, HISAM
Unload, and HISAM Reload, or the execution of IMS Index Builder, or the use of
any tool that builds secondary indexes from the indexed database.
8.9.2 Fallback with logical relationships

The fallback process for logically related databases includes the normal utilities
for resolving logical relationships as part of a reorganization of a non-HALDB
database. This is the execution of Prereorganization, Prefix Resolution, and
Prefix Update.
If your non-HALDB databases used virtual pairing, the fallback process does not
allow you to fall back to your previous definitions. The migration to HALDB
required a conversion to physical pairing. Fallback maintains physical pairing.
Your new non-HALDB DBDs must include physical pairing definitions. The
additional space required for the physically paired segments could prevent a
fallback.

9
Chapter 9. Migration examples

This chapter provides the following migration examples:
򐂰 Migration of a HIDAM database without any secondary indexes to PHIDAM
򐂰 Migration of a HIDAM database with a secondary index to PHIDAM with
PSINDEX
򐂰 Migration of a PDB HIDAM database to PHIDAM
Note: These examples do not include provisions for fallback. You should
include them. As a minimum, you should be prepared to restore the database
to its state before the migration. This may be accomplished by making image
copies of the non-HALDB database data sets, copying the RECONs, and
saving copies of the non-HALDB DBDs. These copies could be used to
restore the database. See 8.9, “Fallback to non-HALDB” on page 126, for
information on falling back and including updates made while the database
was HALDB.

9.1 Migration of HIDAM database to PHIDAM
In this example, we migrate the New Order (NORDDB) database from HIDAM to
PHIDAM. It has no secondary indexes. The database has over 190,000
segments. The migration includes the following steps.
1. Image copy HIDAM database data sets.
2. Run the Migration Aid utility.
3. Unload the HIDAM database.
4. Save database information.
5. Delete the database from the RECONs.
6. DBDGEN for HALDB.
7. Define partitions using PDU.
8. Allocate HALDB data sets.
9. Initialize HALDB partitions.
10.Load HALDB database.
11.Image copy HALDB database data sets.
12.ACBGEN.
9.1.1 Image copy of non-HALDB database data sets

We take image copies of the original non-HALDB database data sets before we
start the migration to HALDB.
9.1.2 Running the Migration Aid utility

Before we start the actual migration, we run the HALDB Migration Aid utility
(DFSMAID0) to analyze the existing database. We run it several times varying
the input parameters. Example 9-1 shows the JCL we use for DFSMAID0. (Note
that you need to specify the DD cards for the databases in the JCL if you do not
have the fix for APAR PQ63751 for IMS Version 7 or PQ68573 for IMS Version 8.)
Example 9-1 JCL for HALDB Migration Aid (DFSMIAD0) utility

//JOUKO1H JOB (999,POK),JOUKO1,CLASS=A,NOTIFY=&SYSUID,
// MSGLEVEL=(1,1),MSGCLASS=X
/*JOBPARM SYSAFF=SC42
//*
//MAID0 EXEC PGM=DFSRRC00,PARM='ULU,DFSMAID0,NORDDB,,0,,,,,,,,,N,N'
//STEPLIB DD DSN=IMSPSA.IM0A.MDALIB,DISP=SHR
// DD DSN=IMSPSA.IMS0.SDFSRESL,DISP=SHR
//NORDDD DD DSN=IMSPSA.IM0A.NORDDB,DISP=SHR
//NOIXDD DD DSN=IMSPSA.IM0A.NORDPI,DISP=SHR
//IMS DD DSN=IMSPSA.IMS0.DBDLIB,DISP=SHR
//DFSVSAMP DD DSN=IMSPSA.IM0A.PROCLIB(DFSVSM0N),DISP=SHR
//SYSIN DD *
DBNAME=NORDDB NBR=4
We run our first analysis using a control statement with NBR=4. This gives the
key ranges and statistics for four partitions of equal size. Example 9-2 shows the
output of this execution of DFSMAID0.
The main purpose of the first run is to get a rough idea of the possible partition
characteristics.
Example 9-2 The output of DFSMAID0 with NBR=4 specified

-------------------------------------------------------------------------------
-
DFSMAID0 1.00 partition report for HIDAM database 'NORDDB ',
Fri Feb 14 02:28:08 2003
root key length = 14 bytes:
DFSMAID0 SYSIN:
" DBNAME=NORDDB NBR=04
"
nbr = 4 ;
-------------------------------------------------------------------------------
-
segment types in database 'NORDDB ':
1) 'NORDER ' ( dmbvldfg=00x sdbtflg=04x sdbptds=60x

)
-------------------------------------------------------------------------------
-
partition 1 :
minimum key =
+0000 f0f0f0f1 f0f1f0f0 f0f0f2f1 f0f1 |00010100002101 |
maximum key =
Chapter 9. Migration examples 131

+0000 f0f0f0f6 f0f3f0f0 f0f0f2f4 f4f3 |00060300002443 |

1) 'NORDER ' 48270 1158480 386160 0
SUM) 48270 1158480 386160 0
-------------------------------------------------------------------------------
-
partition 2 :
minimum key =
+0000 f0f0f0f6 f0f3f0f0 f0f0f2f4 f4f4 |00060300002444 |
maximum key =
+0000 f0f0f1f1 f0f5f0f0 f0f0f2f9 f0f2 |00110500002902 |

1) 'NORDER ' 48270 1158480 386160 0
SUM) 48270 1158480 386160 0
-------------------------------------------------------------------------------
-
partition 3 :
minimum key =
+0000 f0f0f1f1 f0f5f0f0 f0f0f2f9 f0f3 |00110500002903 |
maximum key =
+0000 f0f0f1f6 f0f8f0f0 f0f0f2f2 f0f8 |00160800002208 |

1) 'NORDER ' 48270 1158480 386160 0
SUM) 48270 1158480 386160 0
-------------------------------------------------------------------------------
-
partition 4 :
minimum key =
+0000 f0f0f1f6 f0f8f0f0 f0f0f2f2 f0f9 |00160800002209 |
maximum key =
+0000 f0f0f2f1 f1f0f0f0 f0f0f3f0 f0f0 |00211000003000 |

1) 'NORDER ' 48268 1158432 386144 0
SUM) 48268 1158432 386144 0
-------------------------------------------------------------------------------
-
sum of partitions:
1) 'NORDER ' 193078 4633872 1544624 0
SUM) 193078 4633872 1544624 0
-------------------------------------------------------------------------------
The report provides the minimum and maximum key values, the number of the
segments, and the number of the bytes needed for each partition. It also provides
a summary for all the partitions.
Another way to run the Migration Aid utility is to specify the maximum size in
bytes for each partition. Then the utility gives the number of partitions needed
and the key ranges. This is done by using the control statement MAX=nnnn. If
you specify a value that is larger than 2147483647, then 2147483647 (2G - 1)
will be used. This method could be useful, for example, when you want a partition
to fit on a single volume. In our next test, when we specify MAX=1000000. This
produces seven partitions for our database. We could have guessed this by
summing the bytes and prefix-incr values from the first report and dividing that by
one million.
For the purposes of this example, we think that four is the suitable number of
partitions, but we want to have high keys to be set based on warehouse
numbers. Our database has the warehouse number in the first four bytes of the
key. These values range from 0001 to 0021. We want the partitions to be about
the same size. We take the maximum key values for the partitions from the first
run and round them to the closest warehouse limit. We run the Migration Aid
utility one more time.

The key ranges that we use for this job are the following:
KR=C'00059999999999'
KR=C'00109999999999'
KR=C'00159999999999'
This job produces a report that shows the number of segments and the number
of bytes needed for each of the four partitions. Note that we only specify three
values for the high keys. The last partition has the maximum value for its high key
(16 bytes of X’FF’). Example 9-3 shows the output of the job.
Example 9-3 The output of DFSMAID0 with key ranges specified

-------------------------------------------------------------------------------
DFSMAID0 1.00 partition report for HIDAM database 'NORDDB ',
Fri Feb 14 02:36:14 2003
root key length = 14 bytes:
DFSMAID0 SYSIN:
"KR=C'00059999999999'
"
"KR=C'00109999999999'
"
"KR=C'00159999999999'
"
partitioning keys in effect:
+0000 f0f0f0f5 f9f9f9f9 f9f9f9f9 f9f9 |00059999999999 |
+0000 f0f0f1f0 f9f9f9f9 f9f9f9f9 f9f9 |00109999999999 |
+0000 f0f0f1f5 f9f9f9f9 f9f9f9f9 f9f9 |00159999999999 |
-------------------------------------------------------------------------------
segment types in database 'NORDDB ':
1) 'NORDER ' ( dmbvldfg=00x sdbtflg=04x sdbptds=60x )
------------------------------------------------------------------------------
partition 1 :
minimum key =
+0000 f0f0f0f1 f0f1f0f0 f0f0f2f1 f0f1 |00010100002101 |
maximum key =
+0000 f0f0f0f5 f1f0f0f0 f0f0f3f0 f1f7 |00051000003017 |

1) 'NORDER ' 46107 1106568 368856 0
SUM) 46107 1106568 368856 0
-------------------------------------------------------------------------------
partition 2 :
minimum key =
+0000 f0f0f0f6 f0f1f0f0 f0f0f2f1 f0f1 |00060100002101 |
maximum key =
+0000 f0f0f1f0 f1f0f0f0 f0f0f3f0 f2f6 |00101000003026 |

1) 'NORDER ' 45930 1102320 367440 0
SUM) 45930 1102320 367440 0
-------------------------------------------------------------------------------
-
partition 3 :
minimum key =
+0000 f0f0f1f1 f0f1f0f0 f0f0f2f1 f0f1 |00110100002101 |
maximum key =
+0000 f0f0f1f5 f1f0f0f0 f0f0f3f0 f2f5 |00151000003025 |

1) 'NORDER ' 46205 1108920 369640 0
SUM) 46205 1108920 369640 0
-------------------------------------------------------------------------------
partition 4 :
minimum key =

+0000 f0f0f1f6 f0f1f0f0 f0f0f2f1 f0f1 |00160100002101 |
maximum key =
+0000 f0f0f2f1 f1f0f0f0 f0f0f3f0 f0f0 |00211000003000 |

1) 'NORDER ' 54836 1316064 438688 0
SUM) 54836 1316064 438688 0
-------------------------------------------------------------------------------
sum of partitions:
1) 'NORDER ' 193078 4633872 1544624 0
SUM) 193078 4633872 1544624 0
-------------------------------------------------------------------------------
9.1.3 Unloading the HIDAM database

The next step in the conversion is to unload the database. For this example we
use the standard IMS HD Reorganization Unload utility (DFSURGU0) and
specify MIGRATE=YES in our SYSIN data. Example 9-4 shows the JCL used for
this job.
Example 9-4 JCL for unload of HIDAM database

//NORDUNLJ JOB CLASS=A,MSGCLASS=X,REGION=0M
//*
//ULU EXEC PGM=DFSRRC00,PARM='ULU,DFSURGU0,NORDDB,,,,,,,,,,,N'
//STEPLIB DD DSN=IMSPSA.IM0A.MDALIB,DISP=SHR
// DD DSN=IMSPSA.IMS0.SDFSRESL,DISP=SHR
//IMS DD DSN=IMSPSA.IMS0.DBDLIB.OLD,DISP=SHR
// DD DSN=IMSPSA.IMS0.PSBLIB,DISP=SHR
//DFSURGU1 DD DSNAME=IMSPSA.NORDDB.UNLOAD,DISP=OLD
//NORDDD DD DSNAME=IMSPSA.IM0A.NORDDB,DISP=SHR
//NOIXDD DD DSNAME=IMSPSA.IM0A.NORDPI,DISP=SHR
//DFSVSAMP DD DSN=IMSPSA.IM0A.PROCLIB(DFSVSM0S),DISP=SHR
//SYSIN DD *
MIGRATE=YES
/*
//DFSCTL DD *
SBPARM ACTIV=COND
/*
From the SYSPRINT, we check the number of unloaded segments and the return
code. These are shown in Example 9-5.
Example 9-5 Output of unload job for HIDAM database

TOTAL SEGMENTS IN DATA BASE = 193,078
DFS339I FUNCTION DU HAS COMPLETED NORMALLY RC=0
9.1.4 Saving database information

We take a backup of the RECON data sets and the DBD source. We would need
this information if we need to fall back to our original HIDAM database.
9.1.5 Deleting the database from the RECONs

Because we are converting the database from HIDAM to PHIDAM using the
same database name, we need to delete the database information from the
RECONs. We must do this before we can register the database as HALDB.
Example 9-6 shows the DBRC utility job to delete the HIDAM database,
NORDDB, and its primary index, NOIX.
Example 9-6 JCL for deleting database from the RECON

//DBRC EXEC PGM=DSPURX00
//STEPLIB DD DISP=SHR,DSN=IMSPSA.IMS0.SDFSRESL
// DD DISP=SHR,DSN=IMSPSA.IM0A.MDALIB
//IMS DD DISP=SHR,DSN=IMSPSA.IMS0.DBDLIB
//JCLPDS DD DISP=SHR,DSN=IMSPSA.IM0A.PROCLIB
//JCLOUT DD SYSOUT=(A,INTRDR)
//JCLOUTS DD SYSOUT=*
//SYSIN DD *
DELETE.DB DBD(NOIX)
/*
Example 9-7 on page 138 shows the SYSPRINT from the successful deletion of
the database information from the RECONs.

Example 9-7 The output of DBRC database deletion
IMS VERSION 8 RELEASE 1 DATA BASE RECOVERY CONTROL
DELETE.DB DBD(NOIX)
DSP0203I COMMAND COMPLETED WITH CONDITION CODE 00
DSP0220I COMMAND COMPLETION TIME 2003.045 19:04:12.4 -05:00
DSP0203I COMMAND COMPLETED WITH CONDITION CODE 00
DSP0220I COMMAND COMPLETION TIME 2003.045 19:04:12.6 -05:00
DSP0211I COMMAND PROCESSING COMPLETE
DSP0211I HIGHEST CONDITION CODE = 00
******************************* BOTTOM OF DATA ***************
9.1.6 Changing the DBD and running the DBDGEN

We make the following changes to convert our DBD from HIDAM to PHIDAM:
򐂰 Change ACCESS=HIDAM to ACCESS=PHIDAM in the DBD statement.
򐂰 Remove the DATASET statement. The Partition Definition utility is used to
define the database data sets.
򐂰 Remove the LCHILD statement for the primary index. The primary index
definition for the PHIDAM database is automatically generated from the
PHIDAM definition.
򐂰 We do not have data set groups, so we do not have to code a DSGROUP
parameter on our SEGM statements.
Note: We discard the DBD for the primary index. The primary index definition
is generated from the information in the PHIDAM DBD.
Example 9-8 shows the DBD source for the database before it is converted to
HALDB.
Example 9-8 DBD source for NORDDB HIDAM database

DBD NAME=NORDDB,ACCESS=(HIDAM,OSAM)
*
DATASET DD1=NORDDD,DEVICE=3390,SIZE=26624,FRSPC=(2,19)
*
SEGM NAME=NORDER,BYTES=14,PARENT=0,PTR=TB
FIELD NAME=(NOKEY,SEQ,U),BYTES=14,START=1,TYPE=C
* WWWWDDOOOOOOOO
* 4 2 8
FIELD NAME=NOWID,BYTES=4,START=1,TYPE=C
FIELD NAME=NODID,BYTES=2,START=5,TYPE=C
FIELD NAME=NOOID,BYTES=8,START=7,TYPE=C
*
LCHILD NAME=(INDXSEG,NOIX),PTR=INDX
*
DBDGEN
FINISH
END
Example 9-9 shows the DBD source for the database after it is converted to
HALDB.
Example 9-9 DBD source for NORDDB PHIDAM database

DBD NAME=NORDDB,ACCESS=(PHIDAM,OSAM)
*
* WWWWDDOOOOOOOO
* 4 2 8
*
DBDGEN
FINISH
END
9.1.7 Defining partitions with PDU

We use the HALDB Partition Definition utility (PDU) to define the partitions of the
HALDB database. We use the high key values that we found by running the
HALDB Migration Aid utility (DFSMAID0), as explained in 9.1.2, “Running the
Migration Aid utility” on page 130. For more details on how to use the Partition
Definition utility, see Chapter 4, “Partition Definition utility” on page 45.
9.1.8 Allocate HALDB data sets

We use IDCAMS to allocate our database data sets.
򐂰 We use IDCAMS ALLOCATE for our OSAM data sets. We specify the block
size that we define in PDU.

Note: IDCAMS cannot be used to allocate multi-volume OSAM data sets.
Multi-volume OSAM data sets should be allocated in the reload step. One
of the benefits of HALDB is that you can define partition sizes so that data
sets always fit on one volume. This can eliminate potential multi-volume
OSAM problems.
򐂰 We use IDCAMS DEFINE for our primary index clusters. We use the
information from the RECOMMENDED VSAM DEFINE CLUSTER
PARAMETERS in the SYSPRINT output of the PHIDAM DBDGEN. We
specify our desired CI sizes in the DEFINEs.
򐂰 We use IDCAMS DEFINE for our ILDSs. Since the NORDDB database has
no secondary indexes or logical relationships, the ILDS space allocations can
be very small. Example 9-10 shows our DEFINE statements for an ILDS.
Example 9-10 VSAM definition for defining ILDS

DEFINE CLUSTER (NAME(IMSPSA.IM0A.NORDDB.L00001) -
VOLUMES(TOTIML) -
CYLINDERS (1 1) -
RECORDSIZE(50 50) -
FREESPACE(25 10) -
SHR(3 3) -
SPEED -
KEYS(9 0)) -
INDEX(NAME(IMSPSA.IM0A.NORDDB.L00001.INDEX)) -
DATA(NAME(IMSPSA.IM0A.NORDDB.L00001.DATA) -
CISZ(4096) )
9.1.9 Initializing HALDB partitions

We must initialize the partitions. When we defined them, the PARTITION INIT
NEEDED flags were turned on. We use the Database Prereorganization utility
(DFSURPR0) to initialize the partitions. Example 9-11 shows the JCL for this job.
Example 9-11 JCL for partition initialization using DFSURPR0

//PRERSTEP EXEC PGM=DFSRRC00,
// PARM=(ULU,DFSURPR0,,,,,,,,,,,,Y)
//STEPLIB DD DISP=SHR,DSN=IMSPSA.IM0A.MDALIB
// DD DISP=SHR,DSN=IMSPSA.IMS0.SDFSRESL
//IMS DD DISP=SHR,DSN=JOUKO4.DBDLIB
//RECON1 DD DSNAME=IMSPSA.IM0A.RECON1,DISP=SHR
//PRINTDD DD SYSOUT=*
//DFSURCDS DD DISP=OLD,DSN=IMSPSA.IMS0.DFSURCD5
//SYSIN DD *
DBIL=NORDDB
/*
9.1.10 Loading HALDB database

We use the HD Reorganization Reload utility (DFSURGL0) with DBRC=Y to load
the database. The database data sets are allocated dynamically using
information in the RECONs. Example 9-12 shows the JCL for loading the
database.
Example 9-12 JCL for loading the HALDB database

//LOAD EXEC PGM=DFSRRC00,PARM='ULU,DFSURGL0,NORDDB,,,,,,,,,,,Y'
//DFSUINPT DD DISP=SHR,DSN=IMSPSA.NORDDB.UNLOAD
//DFSURWF1 DD DUMMY
//DFSURCDS DD DUMMY
//DFSCTL DD *
SBPARM ACTIV=COND
//
Example 9-13 shows output from the reload.
Example 9-13 Output of HD Reload

SEGMENT LEVEL STATISTICS
TOTAL SEGMENTS BY SEGMENT TYPE

SEGMENT SEGMENT
NAME LEVEL RELOADED DIFFERENCE
NORDER 1 193,078
TOTAL SEGMENTS IN DATA BASE

UNLOADED RELOADED DIFFERENCE

193,078 193,078
DATABASE NORDDB HAS BEEN SUCCESSFULLY RELOADED BY FUNCTION DR

FUNCTION DR HAS COMPLETED NORMALLY RC=00
******************************** BOTTOM OF DATA *************************
9.1.11 Image copy HALDB data sets

At the completion of the reload, the image copy needed flags are set for all of the
database data sets. We run an image copy job for each partition data set.
We do not image copy the ILDS or primary index data sets. These data sets do
not use image copies for recoveries. Instead, if they are lost they are rebuilt using
the HALDB Index/ILDS Rebuild utility (DFSPREC0). There are no image copy
needed flags for these data sets.
9.1.12 ACBGEN
Finally, we run an ACBGEN. The ACBGEN process for HALDB is the same as for
9.2 Migration of a database with a secondary index

In this example, we migrate the customer (CUSTDB) database from HIDAM to
PHIDAM. The database has approximately 630000 segments. It has a secondary
index. The migration is done using the MIGRATX=YES control statement with
unload.
The procedure is similar to the previous migration example in 9.1, “Migration of

HIDAM database to PHIDAM” on page 130.
1. Image copy the HIDAM database and secondary index data sets.
We run it against both the HIDAM database and the secondary index. See
8.5.2, “Estimating the sizes of secondary index partitions” on page 107, for
information on using the output of the utility with secondary indexes.
3. Unload the HIDAM database.
Only the HIDAM database is unloaded. We do not unload the secondary
index.
We execute HD Unload with a SYSIN control statement specifying
MIGRATX=YES. We include DD statements for DFSWRK01 and DFSSRT01.
These are output files. The DFSWRK01 data set contains the records used to
load the HALDB secondary index. The DFSSRT01 data set contains the sort
control statements used to sort the DFSWRK01 data set before the reload of
the HALDB secondary index.
Example 9-14 shows the JCL used for the unload step.
Example 9-14 JCL for unload with MIGRATX=YES

//UNLOAD EXEC PGM=DFSRRC00,PARM='ULU,DFSURGU0,CUSTDB,,0,,,,,,,,,N,N'
//CUSTDD DD DISP=SHR,DSN=IMSPSA.IM0A.CUSTDB
//CUSTSIDD DD DISP=SHR,DSN=IMSPSA.IM0A.CUSTSI
//DFSURGU1 DD DISP=(NEW,CATLG),DSN=JOUKO4.CUSTDB.UNLOAD,UNIT=SYSDA,
//DFSWRK01 DD DISP=(,CATLG,DELETE),DSN=IMSPSA.CUSTDB.DFSWRK01,
// UNIT=SYSDA,SPACE=(CYL,(15,5))
//DFSSRT01 DD DISP=(,CATLG,DELETE),DSN=IMSPSA.CUSTDB.DFSSRT01,
// UNIT=SYSDA,SPACE=(TRK,(1,1))
//SYSIN DD *
MIGRATX=YES
/*
//DFSCTL DD *
SBPARM ACTIV=COND
If there were more than one secondary index, we would include more
DFSWRKnn and DFSSRTnn DD statements.
4. Sort the secondary indexes files.
We sort the DFSWRK01 output file from the previous step. The records in the
DFSSRT01 file are used as sort control statements in SYSIN. The sorted
output file will be used as an input to the load of the secondary index.
Example 9-15 JCL for sorting secondary index file

//SORT EXEC PGM=SORT
//SORTWK01 DD UNIT=3390,SPACE=(CYL,(50,10),,CONTIG)
//SORTIN DD DSNAME=IMSPSA.CUSTDB.DFSWRK01,DISP=(OLD,PASS)
//SORTOUT DD DSNAME=JOUKO4.CUSTDB.SORTED,DISP=(,CATLG),
// UNIT=SYSDA,SPACE=(CYL,(15,5))
//SYSIN DD DSNAME=IMSPSA.CUSTDB.DFSSRT01,DISP=OLD

5. Save database information.
We back up the RECONs and save the source for our HIDAM and secondary
index DBDs.
6. Delete the HIDAM database and the secondary index from RECONs.
7. DBDGEN of PHIDAM and PSINDEX databases.
Example 9-16 shows the DBD source for the customer database before
conversion to HALDB.
Example 9-16 DBD source for CUSTDB HDAM database

***********************************************************************
* "CUSTDB" DATA BASE: CUSTOMER
***********************************************************************
DBD NAME=CUSTDB,ACCESS=(HDAM,OSAM),RMNAME=(RANDCUST,11,54900)
*
DATASET DD1=CUSTDD,DEVICE=3390,SIZE=8192
*
SEGM NAME=CUSTOMER,BYTES=679,PARENT=0
FIELD NAME=(CKEY,SEQ,U),BYTES=10,START=1,TYPE=C
* WWWWDDCCCC
* 4 2 4
FIELD NAME=CWID,BYTES=4,START=1,TYPE=C
FIELD NAME=CDID,BYTES=2,START=5,TYPE=C
FIELD NAME=CID,BYTES=4,START=7,TYPE=C
FIELD NAME=/CKCID,BYTES=4,START=7,TYPE=C
FIELD NAME=CFIRST,BYTES=16,START=11,TYPE=C
FIELD NAME=CMIDDLE,BYTES=2,START=27,TYPE=C
FIELD NAME=CLAST,BYTES=16,START=29,TYPE=C
FIELD NAME=CSTREET1,BYTES=20,START=45,TYPE=C
FIELD NAME=CCITY,BYTES=20,START=85,TYPE=C
FIELD NAME=CSTATE,BYTES=2,START=105,TYPE=C
FIELD NAME=CZIP,BYTES=9,START=107,TYPE=C
*
LCHILD NAME=(CUSTNAME,CUSTSI),PTR=INDX
XDFLD NAME=XCNAME,SRCH=(CWID,CDID,CLAST,CFIRST), X
SUBSEQ=(/CKCID)
*
DBDGEN
FINISH
END
Example 9-17 on page 145 shows the DBD source for the secondary index
before conversion to HALDB.
Example 9-17 DBD source for CUSTSI non-HALDB secondary index
***********************************************************************
* SECONDARY INDEX DATA BASE, INDEXED ON CWID, CDID, CLAST, CFIRST
***********************************************************************
DBD NAME=CUSTSI,ACCESS=INDEX
*
DATASET DD1=CUSTSIDD,DEVICE=3390,SIZE=20480
*
SEGM NAME=CUSTNAME,BYTES=42,PARENT=0
FIELD NAME=(CNAMEKEY,SEQ,U),START=1,BYTES=42
* WWWWDDLLLLLLLLLLLLLLLLFFFFFFFFFFFFFFFFCCCC
* 4 2 16 16 4
*
LCHILD NAME=(CUSTOMER,CUSTDB),INDEX=XCNAME,PTR=SNGL
*
DBDGEN
FINISH
END
Example 9-18 shows the DBD source for the customer database after
conversion to HALDB. The original versions of changed statements have
been converted to comments. Our original database has a specialized
randomizer that is sensitive to the number of RAPs in the database. Since we
do not need a specialized randomizer, we change to DFSHDC40.
Example 9-18 DBD source for CUSTDB PHDAM database

***********************************************************************
* "CUSTDB" DATA BASE: CUSTOMER
***********************************************************************
* DBD NAME=CUSTDB,ACCESS=(HDAM,OSAM),RMNAME=(RANDCUST,11,54900)
DBD NAME=CUSTDB,ACCESS=(PHDAM,OSAM),RMNAME=(DFSHDC40,11,15000)
*
* DATASET DD1=CUSTDD,DEVICE=3390,SIZE=8192
*
SEGM NAME=CUSTOMER,BYTES=679,PARENT=0
FIELD NAME=(CKEY,SEQ,U),BYTES=10,START=1,TYPE=C
* WWWWDDCCCC
* 4 2 4
FIELD NAME=CWID,BYTES=4,START=1,TYPE=C
FIELD NAME=CDID,BYTES=2,START=5,TYPE=C
FIELD NAME=CID,BYTES=4,START=7,TYPE=C
FIELD NAME=/CKCID,BYTES=4,START=7,TYPE=C
FIELD NAME=CFIRST,BYTES=16,START=11,TYPE=C
FIELD NAME=CMIDDLE,BYTES=2,START=27,TYPE=C
FIELD NAME=CLAST,BYTES=16,START=29,TYPE=C

FIELD NAME=CCITY,BYTES=20,START=85,TYPE=C
FIELD NAME=CSTATE,BYTES=2,START=105,TYPE=C
FIELD NAME=CZIP,BYTES=9,START=107,TYPE=C
*
LCHILD NAME=(CUSTNAME,CUSTSI),PTR=INDX
XDFLD NAME=XCNAME,SRCH=(CWID,CDID,CLAST,CFIRST), X
SUBSEQ=(/CKCID)
*
DBDGEN
FINISH
END
Example 9-19 shows the DBD source for the secondary index after
conversion to HALDB. The original versions of changed statements have
been converted to comments.
Example 9-19 DBD source for CUSTSI PSINDEX

***********************************************************************
* SECONDARY INDEX DATA BASE, INDEXED ON CWID, CDID, CLAST, CFIRST
***********************************************************************
DBD NAME=CUSTSI,ACCESS=PSINDEX
* DBD NAME=CUSTSI,ACCESS=INDEX
*
* DATASET DD1=CUSTSIDD,DEVICE=3390,SIZE=20480
*
SEGM NAME=CUSTNAME,BYTES=42,PARENT=0
FIELD NAME=(CNAMEKEY,SEQ,U),START=1,BYTES=42
* WWWWDDLLLLLLLLLLLLLLLLFFFFFFFFFFFFFFFFCCCC
* 4 2 16 16 4
*
LCHILD NAME=(CUSTOMER,CUSTDB),INDEX=XCNAME,RKSIZE=10
* LCHILD NAME=(CUSTOMER,CUSTDB),INDEX=XCNAME,PTR=SNGL
*
DBDGEN
FINISH
END
8. Define partitions for the PHIDAM database and PSINDEX using PDU.
9. Allocate PHIDAM and PSINDEX data sets.
10.Initialize PHIDAM and PSINDEX partitions.
11.Load PHIDAM and PSINDEX databases.
We can run the HD Reloads for the PHIDAM database and the PSINDEX in
parallel. The input to the reload for the PSINDEX is the sorted file produced in
step 4.
12.Image copy the PHIDAM and PSINDEX data sets.
13.ACBGEN.
9.3 Migrating a PDB database

In this example, we migrate an IMS/ESA Partition Support Product (PDB)
database. The database is a PDB version of the New Order (NORDDB)
database from our example in 9.1, “Migration of HIDAM database to PHIDAM” on
page 130. We migrate the PDB HIDAM database to PHIDAM.
The procedure is similar to the migration example in 9.1, “Migration of HIDAM

database to PHIDAM” on page 130.
1. Image copy the database data sets.
3. Unload the PDB HIDAM database.
We use the MIGRATE=YES control statement in the HD Reorganization
Unload job. We cannot use the PDB Unload utility. It does not support the
creation of an output file with HALDB headers for the records.
Each PDB partition has it own DD statement in the unload JCL. Example 9-20
shows the JCL for unloading the PDB HIDAM database.
Example 9-20 JCL for unloading PDB HIDAM database

//UNLOAD EXEC PGM=DFSRRC00,PARM='ULU,DFSURGU0,NORDDB,,0,,,,,,,,,N,N'
//NOIXDD DD DISP=SHR,DSN=IMSPSA.IM0A.NORDPI
//NORDD1 DD DISP=SHR,DSN=IMSPSA.IM0A.NORDD1
//DFSURGU1 DD DISP=(NEW,CATLG),DSN=JOUKO4.NORDDB.UNLOAD,UNIT=SYSDA,
//SYSIN DD *
MIGRATE=YES
4. Save the database information.

5. Delete the database from the RECONs.
6. DBDGEN for HALDB.

The PART statements are deleted from the DBD source. In PDB these are
used to describe partitions.
Example 9-21 shows the DBD definition for the PDB HIDAM database.
Example 9-21 DBD source for NORDDB PDB HIDAM database

DBD NAME=NORDDB,ACCESS=(HIDAM,OSAM)
*
PART DD1=NORDD1,SIZE=26624,FRSPC=(2,19),HIGHKEY=00059999999999
*
* WWWWDDOOOOOOOO
* 4 2 8
*
LCHILD NAME=(INDXSEG,NOIX),PTR=INDX
*
DBDGEN
FINISH
END
Example 9-22 shows the DBD definition for the HALDB PHIDAM database.
Example 9-22 DBD source for NORDDB HALDB PHIDAM database

DBD NAME=NORDDB,ACCESS=(PHIDAM,OSAM)
*
* WWWWDDOOOOOOOO
* 4 2 8
*
DBDGEN
FINISH
END
7. Define partitions using PDU.

8. Allocate the HALDB data sets.
9. Initialize the HALDB partitions.
10.Load the HALDB database.
11.Image copy the HALDB database data sets.
12.ACBGEN.

10
Chapter 10. Using the HALDB

Conversion and
Maintenance Aid
IMS High Availability Large Database Conversion and Maintenance Aid for
z/OS™ (product number 5655-K47) is a product in the IBM Data Management
Tools for IMS portfolio. IMS HALDB Conversion and Maintenance Aid is an ISPF
application that allows you to analyze, model, and convert existing IMS database
structures to the HALDB format.
IMS HALDB Conversion and Maintenance Aid creates DBD and DBRC
definitions, allocates HALDB partition data sets, and creates the conversion JCL.
Batch functions support the conversion process. This chapter provides an
overview of the product and examples of its use.
Note: This chapter is based on a pre-GA version of IMS High Availability

Large Database Conversion and Maintenance Aid for z/OS Version 2 product
and may not apply as such when the product becomes generally available. We
recommend that you consult the product documentation for more current
information. For example, the title line on all ISPF panels has changed to
reflect the final product name.

10.1 HALDB Conversion and Maintenance Aid product
The IMS HALDB Conversion and Maintenance Aid is designed to ease the
conversion process from existing full function to HALDB databases. All
necessary steps are created and then presented to you through an ISPF
interface. You submit the steps and check the output for zero return codes. The
IMS HALDB Conversion and Maintenance Aid then creates the next action. The
process continues until the conversion is complete. The HALDB Conversion and
Maintenance Aid Version 2 provides enhancements to the HALDB Conversion
Aid Version 1 (product number 5655-I01). Both products include the following:
򐂰 Support for conversion of HDAM, HIDAM, HISAM, and secondary indexes to
HALDB. HISAM is converted to PHIDAM.
򐂰 Support for migration from PDB or PDF.
򐂰 Conversion of DBDs.
򐂰 Allocation of database data sets.
򐂰 Allows user to choose the level of control for migration: Cruise control (largely
automatic) or manual (step-by-step) or in between.
򐂰 Inclusion of a generic partition selection exit IHCPSEL0.
– Based on subset of the root key
– Specification of offset and length
– Uses data string in DBRC definition
The HALDB Conversion and Maintenance Aid Version 2 provides the following
additional functions:
򐂰 It can create DBD source from the DBD load library. This allows the DBD
source libraries to be discarded.
򐂰 It provides a stand-alone EPS pointer healer utility. This allows you to heal all
secondary index pointers after a reorganization.
򐂰 It has a maintenance function that helps in maintaining (splitting or combining
partitions) your HALDB database once it has been converted from full
function.
򐂰 It can improve the performance when initially loading databases with the
secondary indexes. When a user application program creates a HALDB
database, the product provides a capability not to insert secondary indexes
immediately. They are saved in temporary files during the load process. Once
the database is loaded, all indexes are sorted and loaded to the index KSDS
file(s).
򐂰 It provides a capability to allow initial load programs to insert logical child
segments. That is, it removes the normal HALDB restriction that does not
allow logical child segments to be inserted when a user application program is
initially loading a database with logical relationships.
򐂰 It provides a function that can be used to enable BA status codes for
applications that do not issue the INIT STATUS call. This will prevent those
applications from abending with a U3303 abend. The second feature of this
function allows you to define a retry mechanism to retry the call after receiving
a BA status code. The third feature limits the BA handling to specified
databases only.
Section 10.4, “Additional functions” on page 176, describes these new functions.
10.1.1 The conversion process overview

The conversion process for a database with IMS HALDB Conversion and
Maintenance Aid consists of the following phases:
1. Phase 1 collects the necessary data and requests the DBDs that are to be
converted.
The user selects the environment, that is, a set of RECONs, DBDLIBs, and
other IMS data sets, and sets up a project for the database(s) to be converted
by entering the answers to the questions presented in ISPF panels. For many
questions, one option is to select Ask during the conversion. If selected, the
user is prompted during the implementation phase to answer the question.
If the database has relationships, such as logical relationships or secondary
indexes, those DBDs are automatically included. A collect job is created for
every primary DBD, in order to gather the information.
2. Phase 2 prepares the conversion process.
The partition boundaries are set and the file allocations are prepared. During
this phase IMS HALDB Conversion and Maintenance Aid reads old
databases for modeling and calculates partition sizes or partition high keys
based on the modeling information. IMS HALDB Conversion and
Maintenance Aid provides an efficient alternative to the HALDB Migration
Utility (DFSMAID0).
3. Phase 3 unloads the “old” databases.
4. Phase 4 implements the IMS HALDB Conversion and Maintenance Aid.
During the implementation phase, IMS HALDB Conversion and Maintenance
Aid performs the following functions:
– It defines partitions in DBRC. It uses data from reading the old database to
establish boundaries. It allows the user to adjust boundaries. All DBRC
information for the old database is removed for DBRC.
Chapter 10. Using the HALDB Conversion and Maintenance Aid 153
– It generates the new DBD for the database(s). When converting the DBD,
IMS HALDB Conversion and Maintenance Aid is able to handle virtual
pairing to physical pairing conversion and multiple data set groups. The
user has the choice to keep groups or combine them. IMS HALDB
Conversion and Maintenance Aid also saves the old definitions.
– It allocates the database data sets. It uses the data from reading the old
database for sizing. It allows the user to adjust sizes and other
parameters.
– It initializes the partitions.
– It creates the jobs for the reload and is able to select between MIGRATE
and MIGRATX based on whether the converted database(s) has
secondary indexes or not.
– It creates image copies.
– It creates an optional EPS pointer heal job that you can run.
For all the jobs, the JCL is presented to the user. The user submits the jobs.
The jobs maybe modified before they are submitted. An ISPF panel informs
the user of the end of each phase.
5. Phase 5 is a reminder that after databases are converted, ACBGENs and
online changes are required.
IMS HALDB Conversion and Maintenance Aid does not perform those
actions.
The following full-function DBDs cannot be converted by using IMS HALDB

Conversion and Maintenance Aid:
򐂰 HSAM
򐂰 Non-unique secondary indexes with five SUBSEQ parameters
򐂰 Logical related databases type 4; that is, those that appear to be virtually
paired without a virtual segment
10.2 Activating the tool

After the SMP/E installation of HALDB Conversion and Maintenance Aid, there
are only a few things you need to do before you can start using the product:
򐂰 Modify the startup CLIST for ISPF invocation and make it available to the
users of the product.
򐂰 Allocate the environment description data set.
10.2.1 Modifying the startup CLIST
A sample CLIST for invoking the HALDB Conversion and Maintenance Aid in an
ISPF environment is provided. You can find the sample CLIST in the product
sample library hlq.SIHCSAMP where hlq is the high-level qualifier specified
during the product installation. The CLIST has the name IHCXHAL in the sample
library.
You may want to copy this CLIST to a CLIST library that is already included in
your TSO allocation and modify it there. In the sample CLIST, the only things you
need to modify are the ISPF library names for the product target libraries unless
you are using the product default names. The names in the CLIST must match
the target library names.
By copying the CLIST to the general CLIST library, the start command is
available to all users who have the library defined in their TSO allocation. Then
you do not have to specify the library name where the CLIST resides when using
the start command. In our case we copied the CLIST into library
SYS1.OS390.CLIST and gave the name $IHCXHAL for it. The other way is to
modify the CLIST in the SIHCSAMP library and specify the library name in the
start command when starting the tool:
ex ‘hlq.SIHCSAMP(IHCXHAL)’
10.2.2 Allocating the environment description data set

Before using the HALDB Conversion and Maintenance Aid, you must allocate the
environment description data set. It is a VSAM key sequenced data set (KSDS)
that the conversion process uses to define and maintain environment-related
information. You can find the sample member IHCXKSDS in the product sample
library (hlq.SIHCSAMP). Change the library names, the data set name of the
KSDS to be allocated, and the volume information to meet your requirements,
and run the modified JCL.
Note: The HALDB Conversion and Maintenance Aid load library must be APF
authorized before running this allocation job.
10.3 Migration example with HALDB Conversion tool

This example explains how to use the IMS HALDB Conversion and Maintenance
Aid tool to convert an existing HDAM database, CUSTDB, and its secondary
index, CUSTSI, to PHDAM and PSINDEX. You have to perform the following
tasks:
򐂰 Define an IMS environment.
򐂰 Define a conversion project.
򐂰 Process a conversion project.
To start the IMS HALDB Conversion and Maintenance Aid, use the CLIST that
you had modified as described in the 10.2.1, “Modifying the startup CLIST” on
page 155. In our case we can start the tool by using the command $IHCXHAL
from the ISPF command option or from any TSO command line by entering the
command:
TSO $IHCXHAL
10.3.1 Define an IMS environment

Use the main menu (Figure 10-1) to select all major IMS HALDB Conversion and
Maintenance Aid actions. When you use the tool the first time, an environment
has not yet been defined and you must select option 1.
Figure 10-1 Main menu
After selecting option 1 you get the panel about environment maintenance
(Figure 10-2 on page 157), where you choose option 2 to add a new
environment.
Figure 10-2 Environment maintenance
Now you see a panel where you have to enter a name and a description for your
environment. The next panel (Figure 10-3) shows an example.
Figure 10-3 Definition of the new environment
In the next step you set definitions for your project (see Figure 10-4 on
page 158).
Figure 10-4 Project definition
To set up an environment, the tool must know which data sets it should use
(Figure 10-5). Specify your RESLIB, and perhaps a second data set for user
program library (user exits, randomizer, etc.). For MACLIB enter the name of your
MACLIB, and for LOADLIB the name of the IHC load library. Enter the data set
name and member name that contains your DFSVSAMP member. Specify your
MDALIB that is used to extract data set names of the databases and where the
RECON MDA are located if you want them dynamically allocated. Otherwise
specify the RECONs you want to use in the next panel. After that you see a panel
where you define your DBDLIBs (up to nine).
Figure 10-5 Defining data sets
The HALDB Conversion and Maintenance Aid is designed to run with preset
defaults. However, you can override defaults to accommodate the needs of your
environment. The following panels describe available options. You can get help
by pressing PF1. Default conversion rules are shown in Figure 10-6.
Figure 10-6 Conversion defaults
This panel contains the following fields:

򐂰 Multi DSG: Change to 1 if you want to convert multiple data set groups to a
single one.
򐂰 Convert VSAM to OSAM: You may specify whether you want to convert
VSAM to OSAM data sets. OSAM advantages are typically reduced CPU time
and the availability of OSAM sequential buffering.
򐂰 Change DBD name: If you change the name you must also change the PSBs
and logical DBDs (this tool does not perform these functions). If names are
not changed, DBRC information for old databases is deleted.
򐂰 Selection exit: Specify if you want to add one.
򐂰 Heal index pointer: Specify whether to heal the index pointers. The migration
process leaves the PSINDEX pointers in a healing needed status. If you want
them healed specify option 1, which will create the necessary JCL to run the
pointer healing utility for every PSINDEX database in the project.
If you choose option 3 for any field, the value for that option will be Ask during
Conversion. These rules apply to all of your projects in the selected environment
if you do not change them in your project definition.
The Save Source Statements panel is used to enable the saving of source
statements (Figure 10-7). This function saves the source for new DBDs and the
IDCAMS used for allocating the data sets into the specified data sets. These data
sets must be allocated before making the selection.
Figure 10-7 Save source statements
You can also specify the member name for the IDCAMS statements. With option
1 (DBDNAME) all IDCAMS statements for that DBD become part of a member
with this name. With option 2 (partition name) all statements of that partition are
included in a member with the partition name. With option 3 (DD name) the
IDCAMS statement per DD name is saved. Option 3 provides maximum flexibility.
We recommend that you save the IDCAMS statements. This allows you to delete
and allocate the data sets very easily. It is not necessary to save the DBD source
because you have the capability to see and reassemble it by choosing the
appropriate panel option.
The Partition Definition Rules (Figure 10-8 on page 161) panel sets partition
defaults at data collection time.
Figure 10-8 Partition Definition Rules
It contains the following fields:

򐂰 Combine Database records
Here you define how many database records should be combined for one
data point. These data points are used to analyze potential partition
boundaries and sizes. This parameter influences the performance of the
creation of new partitioning layouts. The smaller the specified parameter, the
greater the number of records stored in the project data set during the
collection phase. More records allow for a finer granularity in the analyses.
򐂰 Main DB Partition
Specify how a partitioning layout is to be created during data collection. You
can choose between number of partitions, size of partitions, specifying the
high keys, or delaying the selection until the creation of the JCL.
򐂰 Index Partition
Specify the number of PSINDEX partitions to be created or their size.
򐂰 PDB Conversion
You have to decide whether the existing ranges should be used or new
partition boundaries are to be created.
򐂰 Additional Partitioning Layouts to be created?
This function is available for the main database only.
Next you have to define the partition naming rules (Figure 10-9). A unique
partition name must be created for each partition in your installation. The tool
does not check the validity.
With Partition DSN specify the left-most portion of the partition data set name (for
naming conventions see 1.9, “Naming conventions” on page 17), which applies
to all data sets in your environment.
Figure 10-9 Naming
Now some space allocation panels follow. The first one defines the space
allocation rules for PHDAM and PHIDAM databases (Figure 10-10 on page 163).
In these panels the term cloning means that the old definitions should be taken.
Similar panels for PSINDEX data sets and ILDS follow (not shown).
Figure 10-10 Rules for PHDAM and PHIDAM
JCL defaults for the JOB statement you submit for all conversion JCLs must be
entered.
If you have the tools High Performance Unload or High Performance Reload you
may use them instead of the standard HD Unload and Reload utilities. One panel
offers the selection of the Unload utility or HP Unload (Figure 10-11). A similar
one offers the selection of the Reload utility or HP Reload.
Figure 10-11 Select utility for unload
The following panel is for the utility set up for backout (Figure 10-12). It provides
some defaults for the standard Image Copy utility (DFSUDMP0). These are the
number of copies you want and the naming scheme.
Figure 10-12 Utility set up for backup
Now the environment is set and you can start defining your project.
10.3.2 Define a project

We want to convert a full-function database to a new full-function HALDB, so we
have to enter option 4 of the main menu (Figure 10-1 on page 156). From the
Conversion panel we choose option 1 (Figure 10-13 on page 165) to add a new
project.
Figure 10-13 Conversion Full Function to HALDB
Next we follow the guided panel tour to set some other defaults before returning
to the Conversion panel. Basically you see the panels you already know from
defining the environment. Now we use them for determining the values for our
individual project. We may change the shown defaults or not. For example, if for
this particular database we need a partition selection exit routine, we change the
conversion default to Selection Exit: 1 (according to Figure 10-6 on page 159; 1
means YES). Then the following panel is shown (Figure 10-14 on page 166). We
may specify which part of the root key the selection exit should use (offset and
length) and how the partitions are to be defined.
Figure 10-14 Specify a Partition Selection Exit
If we want to be prompted, a panel will be presented where we may define our

key strings. In a summary panel we get all strings we have entered and we may
change them (Figure 10-15). In our migration example, however, we do not use
the partition selection exit.
Figure 10-15 Key strings defined for CUSTDB partition selection exit
One other default we might want to change for a specific database is additional
layouts. Perhaps we do not know the partition ranges and want to check which
would be the best choice. We select the appropriate row in Figure 10-8 on
page 161 and specify 1 to view the predefined partition layout or create
additional ones. We see a selection panel (Figure 10-16) and we must then
decide which layout to use.
Figure 10-16 Additional Layout Selection
Our database CUSTDB should have four partitions, so the layout stored looks
like Figure 10-17.
Figure 10-17 Stored layout for CUSTDB
We will use the stored layout. We select a new project, give it a name and a
description and specify a tracking and staging KSDS (DSN, space, volser), which
we allocate by generated JCL. Back in the Conversion panel we select function 3
to display our projects and select the appropriate one (Figure 10-18).
Figure 10-18 Panel displaying added project
10.3.3 Process a conversion project

Now we may start our project by selecting option 4. We get the Project Work
Panel (Figure 10-19).
Figure 10-19 Project Work Panel
With option 0 we may check all settings and eventually change them. However,
we want to continue our project and select option 1, which gives us a list of DBDs
to be selected. CUSTDB is the database we wish to convert, so we select it. We
can select one DBD after the other. If we select the corresponding secondary
index we get the message that it has already been selected because this tool
automatically selects all related databases.
After the selection we will be presented with a panel that shows the DBDs
indicated (Figure 10-20).
Figure 10-20 DBDs in project
We must now press the END key to display the continuation, which is the
confirmation panel of our DBD selection. We specify Y for selection complete (N
if we want to add more DBDs). When we have confirmed our selection, the
following panel gives us the name of the unload data set (Figure 10-21).
Figure 10-21 Panel naming unload file
Next we select whether we want to specify the high keys, a certain number of
partitions, or a certain size for our partitions. In our example we simply say we
want four partitions (Figure 10-22 on page 170).
Figure 10-22 Select partition layout
Now the setting is done and JCL for the unload is generated. We check and
submit it (Example 10-1).
Example 10-1 Unload JCL

//DEL EXEC PGM=IDCAMS
//SYSIN DD *
DELETE IMSPSA.IM0A.CUSTDB.UNLOAD.CUSTDB
SET MAXCC=0
/*
//UNLOAD EXEC PGM=IHCHALDB,COND=(4,LE),
// REGION=60M
//STEPLIB DD DISP=SHR,DSN=IHC.V2PTF2.SIHCLOAD
//DFSRESLB DD DISP=SHR,DSN=IMSPSA.IMS0.SDFSRESL
//RECON1 DD DISP=SHR,DSN=JOUKO1.RECON1
//CUSTDD DD DISP=SHR,DSN=IMSPSA.IM0A.CUSTDB
//CUSTSIDD DD DISP=SHR,DSN=IMSPSA.IM0A.CUSTSI
//IHCPROJ DD DISP=SHR,DSN=IMSPSA.PROJECT.HDO2
//DFSURGU1 DD DISP=(,CATLG),SPACE=(CYL,(156,50)),UNIT=SYSDA SHR,
// DSN=IMSPSA.IM0A.CUSTDB.UNLOAD.CUSTDB
//MSGPRINT DD SYSOUT=*
//IHCSYSIN DD *
UNLOAD DBD(CUSTDB) -
PARTNUM(4) -
SELTYP(1) -
KEYLIST(2) -
INDSEL(2) INDSIZE(1024) -
INDPART(1) INDMIN(1024) -
DBREC(5000)
/*
Most generated JCLs also contain an additional step (not shown in our
examples) that keeps track of the status of the project in the KSDS we defined
(10.3.2, “Define a project” on page 164).
Pressing the End key leads to the Project Checkpoint panel (Figure 10-23),
which shows the status of the project. We are asked if we want to start the next
phase (called implementation). We enter Y for yes.
Figure 10-23 Project Checkpoint panel
Now a series of panels with generated JCLs is shown to us, which we have to
submit. The sequence is:
1. JCL for DELETE.DB
2. JCL for DBDGEN
3. JCL for saving the source (if we had specified SAVE SOURCE as default)
4. JCL for INIT.DB and INIT.PART
5. JCL for deletion and allocation of the data sets
6. JCL for saving the source (if we had specified SAVE SOURCE as default)
7. JCL for initialization
8. JCL for reload
9. JCL for image copy
Not all of the jobs are shown here. The first does a DELETE.DB of CUSTDB,
which is still non-HALDB in the RECONs, so all related information is removed
from the RECONs. Then the new HALDB DBD is created by DBDGEN. You may
see the source in your save file or by choosing option 7 (other utilities) of the
main panel and then selecting option 1 (show DBD source). The next step
registers the HALDB in the RECONs (Example 10-2).
Example 10-2 JCL for INIT commands

//DBRC EXEC PGM=IHCZUTIL,COND=(4,LE),REGION=50M
//IHCSYSIN DD *
RUN PGM(IHCZDBRC)
//SYSIN DD *
INIT.DB DBD(CUSTDB) TYPHALDB
INIT.PART DBD(CUSTDB) PART(CUSTDB0) -
ANCHOR(11) HIBLOCK(13725) -
FBFF(0) FSPF(0) -
BLOCKSZE(8192) -
DSNPREFX(IMSPSA.IM0A.CUSTDB.CUSTDB) -
KEYSTRNG(X'F0F0F0F6F0F3F1F5F0F0')
FBFF(0) FSPF(0) -
BLOCKSZE(8192) -
FBFF(0) FSPF(0) -
BLOCKSZE(8192) -
FBFF(0) FSPF(0) -
BLOCKSZE(8192) -
KEYSTRNG(X'FF')
INIT.DB DBD(CUSTSI) TYPHALDB
INIT.PART DBD(CUSTSI) PART(CUSTSI0) -
DSNPREFX(IMSPSA.IM0A.CUSTDB.CUSTSI) -
KEYSTRNG(X'FF')
After registering the database and partitions, the corresponding data sets must
be allocated. Example 10.3 on page 155 shows the statements for the data data
set and the ILDS data set of partition 1, and the allocation JCL for the secondary
index data set. The allocation statements for the other partitions are similar.
Example 10-3 JCL for allocation (shows only partition 1)

DEFINE CLUSTER( -
NAME(IMSPSA.IM0A.CUSTDB.CUSTDB.L00001) -
CYLINDERS(1,1) -
INDEXED KEYS(9,0) -
RECSZ(50 50) -
CISZ(4096) -
FREESPACE(0,0) -
VOLUMES(TOTIML) -
REUSE )
ALLOCATE -
DATASET('IMSPSA.IM0A.CUSTDB.CUSTDB.A00001') -
NEW CATALOG DSORG(PS) -
SPACE(160,16) -
CYLINDERS
.......
.......
DEFINE CLUSTER( -
NAME(IMSPSA.IM0A.CUSTDB.CUSTSI.A00001) -
CYLINDERS(79,8) -
INDEXED KEYS(42,39) -
RECSZ(82 82) -
CISZ(4096) -
FREESPACE(0,0) -
VOLUMES(TOTIML) -
REUSE )
The next step sets the PINIT flags with a CHANGE command and initializes all
partitions (Example 10-4) with the DFSUPNT0 utility.
Example 10-4 Initialization

//JOBHCA JOB (999,POK),'LISTINGS',NOTIFY=JOUKO1,
// CLASS=A,MSGCLASS=T,TIME=1439,
// REGION=0M,MSGLEVEL=(1,1)
//*
//DBRC EXEC PGM=IHCZUTIL,
// COND=(4,LE),
// REGION=50M
//IHCSYSIN DD *
RUN PGM(IHCZDBRC)
//SYSIN DD *
CHANGE.DB DBD(CUSTDB0) PINIT
CHANGE.DB DBD(CUSTSI0) PINIT
//INIT EXEC PGM=DFSUPNT0,
// COND=(4,LE),
// REGION=50M
//DFSVSAMP DD DISP=SHR,
// DSN=IMSPSA.IM0A.PROCLIB(DFSVSM0R)
//SYSIN DD *
CUSTDB
CUSTSI
After this has completed successfully we may process the reload

(Example 10-5), which includes CHANGE commands to turn off the image copy
needed flags set by the initialization.
Example 10-5 Reload
//*=====================================================
//*
//* RELOAD PRIMARY DATABASE CUSTDB
//*
//*=====================================================
//RELD EXEC PGM=IHCHALDB,
// COND=(4,LE),
// REGION=60M
//DFSURGU1 DD DISP=OLD,
// DSN=IMSPSA.IM0A.CUSTDB.UNLOAD.CUSTDB
//IHCSYSIN DD *
RELOAD DBD(CUSTDB) UNLFILE(DFSURGU1)
The job output says Image copy needed for..., so the next job must be an
image copy (DFSUDMP0). We check and submit it. After the image copies of the
main database have been performed, we are asked whether we want to do an
image copy of our secondary index. If we choose NO, we must have a tool to
build the secondary index. We submit the generated image copy JCL and then
again we are presented with a checkpoint panel, which asks whether we want to
start the next phase: Post Processing.
After entering YES we see the following panel (Figure 10-24). This is just a
reminder and lists the tasks that are to be performed by the user.
Figure 10-24 Post-processing
If all shown tasks are performed, the HALDB database may be used and this
conversion project is finished. A last panel asks if the project data set should be
kept, reused, or deleted.
10.4 Additional functions
The HALDB Conversion and Maintenance Aid Version 2 for z/OS is a successor
product to the IMS HALDB Conversion Aid Version 1. As the name implies, the
new functions are primarily for the maintenance of databases after they have
been converted to HALDB. They include the capability to split or consolidate
HALDB partitions (choose option 5 of the main panel Figure 10-1 on page 156).
Other functions include back up, recovery, unload, and reload for individual
partitions, as well as INDEX pointer healing (choose option 6 of the main panel).
For example, our current HALDB NORDDB with a partition exit routine has four
partitions. Now we think that perhaps six partitions would be better, but which
keystrings should we use? We choose option 5, make our definitions, enter the
new keystrings, and submit the JCL generated. Example 10-6 shows the
analysis based on the new partitions. We can see that we managed to select
quite a good keystring, since all the partitions are almost equal in size. Only the
last partition has less segments, but this change may be done for the preparation
for the new data to be added in the last partition. You are asked if you want to
continue with an unload and if you enter YES, the unload JCL is generated.
Example 10-6 Output for testing new boundaries

Partitioning for DBD NORDDB
Part Roots % All Segments % Prefix Bytes % Data Bytes % Total Bytes %
Total 193,078 100.00 193,078 100.00 3,475,404 100.00 2,703,092 100.00 6,178,496 100.00
1 36,960 19.14 36,960 19.14 665,280 19.14 517,440 19.14 1,182,720 19.14
Key: 0004
F0F0F0F4
2 36,582 18.94 36,582 18.94 658,476 18.94 512,148 18.94 1,170,624 18.94
Key: 0008
F0F0F0F8
3 37,021 19.17 37,021 19.17 666,378 19.17 518,294 19.17 1,184,672 19.17
Key: 0012
F0F0F1F2
4 36,901 19.11 36,901 19.11 664,218 19.11 516,614 19.11 1,180,832 19.11
Key: 0016
F0F0F1F6
5 36,614 18.96 36,614 18.96 659,052 18.96 512,596 18.96 1,171,648 18.96
Key: 0020
F0F0F2F0
6 9,000 4.66 9,000 4.66 162,000 4.66 126,000 4.66 288,000 4.66
Key: ÿÿÿÿ
FFFFFFFF
Then a job with CHANGE commands for the old partitions and INIT commands
for the new partitions is created (Example 10-7 on page 177).
Example 10-7 Defining the new partitions
//DBRC EXEC PGM=IHCZUTIL,
// COND=(4,LE),
// REGION=50M
//IMS DD DISP=SHR,DSN=JOUKO1.DBDLIB1
//IHCSYSIN DD *
RUN PGM(IHCZDBRC)
//SYSIN DD *
CHANGE.PART DBD(NORDDB) PART(NORDDB0) -
KEYSTRNG(X'F0F0F15EF0F0F45EF0F0F0F4') -
DSNPREFX(IMSPSA.IM0A.NORDDB.NORDDB)
KEYSTRNG(X'F0F0F15EF0F0F45EFF') -
As already shown in the conversion example, jobs follow that allocate the data
sets, initialize the partitions, perform the reload, and all image copies needed.
Option 7, other utilities, provides such functions as the Partition Selection Exit
Test utility, OSAM Multi-volume, and the Check "BA" Status Code Table. Most are
self-explanatory, so only a few words about the last function are added here. You
may enable BA status codes for applications that do not issue the INIT STATUS
call. The tool provides an exit. You determine for which programs this call should
be created and you specify the databases for which the programs should wait if
they get a BA status code. For the mechanism to retry the call after receiving a
BA, two parameters may be specified, wait time and retry count. The retry is
made after wait time has exceeded, and the count tells how many times a retry
should be attempted. This function only applies to BMP and MPP regions and for
GU or GHU calls.
Option 8 (Figure 10-25) allows you to perform the following DBRC functions,
where the first three are LIST commands.
Figure 10-25 DBRC Support
If you want the definitions of your HALDB partitions copied into another RECON
environment, choose Clone Partition Definition. This does not transfer the
content of your database. Specifying your DBDLIB and the new RECONs on the
given panels leads you to the following panel (Figure 10-26).
Figure 10-26 Clone partition definitions
Pressing Enter will generate the needed JCL (Example 10-8 on page 179).
Example 10-8 Clone partition definitions
//*-----------------------------------------------------------------
//* CREATE DBRC COMMANDS
//*-----------------------------------------------------------------
//CRE EXEC PGM=IHCHALDB,
// REGION=60M
//IHCSYSIN DD *
COPYDBRC DBD(NORDDB) -
DSNPREF(IMSPSA.TEST) -
DSNDBD(YES) -
INCLIND(YES) -
TODD(DBRCOUT)
/*
//DBRCOUT DD DISP=(,PASS),SPACE=(TRK,(1,1)),UNIT=SYSALLDA
//*---------------------------------------------------------
//* APPLY TO OTHER DBRC
//*---------------------------------------------------------
//DBRC EXEC PGM=DSPURX00,REGION=50M,COND=(4,LE)
//RECON1 DD DISP=SHR,DSN=JOUKO1.NEW.RECON1
//SYSIN DD DISP=(OLD,DELETE),DSN=*.CRE.DBRCOUT
The HALDB Conversion and Maintenance Aid helps you implement the database
in another RECON environment if you choose Copy Partitions to a different
RECON. This will create cloned partition definitions in another RECON, and also
transport full image copies to these RECON data sets. It will resolve the
problems with partition IDs as they are described in 11.7.2, “Using image copies”
on page 207. Then you have to perform a timestamp recovery with
GENJCL.RECOV and test your database.
Part 3
Part 3 Application
considerations
In this part we discuss potential application changes and opportunities with
HALDB. We describe some application programming changes that you may need
to make, and other changes that you may want to implement.
We discuss the differences that HALDB creates for test environments and the
copying of databases.

11
Chapter 11. Application considerations

You may not need to change any of your application programs when you migrate
databases to HALDB, but there may be exceptions. In this chapter we explain the
exceptions and how to handle them. Exceptions include the initial load of logically
related databases and the processing of secondary indexes as databases.
We also explain why you might want to change applications to take advantage of
some HALDB capabilities. These capabilities include processing partitions in
parallel, processing individual partitions, and handling unavailable partitions.
We explain how to create test environments for HALDB and how to copy HALDB
databases between IMS environments.

11.1 Initial loads
Application programs that initially load databases generally do not have to be
changed when the databases are migrated to HALDB. On the other hand, you
may find situations in which it is required or advantageous to modify existing
initial load programs.
The only required change is for application programs that initially load logical
children in logically related databases. This is not allowed with HALDB. If you
have an initial load application program that inserts logical child segments, it
must be modified. We describe these modifications in “Initial loads of databases
with logical relationships” on page 188.
11.1.1 PHDAM considerations

Initially loads of PHDAM databases may insert root segments in any order. This
is also true for HDAM databases. Any program that was used for initial loads of
HDAM databases may be used without modification to load PHDAM databases.
Many installations load HDAM databases in RAP sequence. This loads the
records in physical sequence in the database. This is done by sorting the input
records into the order that the randomization routine creates. To do the same
thing with PHDAM, one would have to sort the records within each partition. With
PHDAM, randomization is done within a partition, not across partitions. That is,
the partition is selected first and then randomization is done. Each partition could
have different randomization parameters. In some cases, the techniques used
with HDAM might not work for PHDAM. For example, the use of different
randomization routines for different partitions would prevent a general sort for the
entire database. The IBM High Performance Load (HP Load) product includes a
Physical Sequential Sort for Reload (PSSR) program. PSSR may be used to sort
records in RAP sequence. It includes HALDB support to sort records for a
partition, a set of partitions, or all partitions in a database.
11.1.2 PHIDAM considerations

Initial loads of PHIDAM databases must insert root segments in key sequence
within a partition. They do not have to be in key sequence when crossing
partition boundaries. Figure 11-1 on page 185 illustrates this. The figure shows a
PHIDAM database that uses a partition selection exit routine. Roots are in key
sequence within the partitions. They are not in key sequence when crossing
partition boundaries. Two possible load sequences are shown. The first shows
loading the records in key sequence within each partition. The second shows
loading the records in key sequence across the entire database. Other
sequences would also be valid as long as they load records in a partition in key
sequence.
Partition Partition Partition Partition

A B C Z
Keys: Keys: Keys: Keys:

20A12 24B12 08C74 04Z88
45A16 47B02 17C12 65Z49
62A37 78B23 34C32 77Z20
77A18 89B00 75C89 97Z91
Valid database load sequences:
1. 20A12, 45A16, 62A37, 77A18, 24B12, 47B02, 78B23, 89B00,

08C74, 17C12, 34C32, 75C89, 04Z88, 65Z49, 77Z20, 97Z91
2. 04Z88, 08C74, 17C12, 20A12, 24B12, 34C32, 45A16, 47B02,

62A37, 65Z49, 75C89, 77A18, 77Z20, 78B23, 89B00, 97Z91
Figure 11-1 Loading sequences for a PHIDAM database
11.1.3 Loading partitions in parallel

You may want to load partitions in parallel. This should reduce the time to load a
HALDB. You may use separate concurrently executing jobs to do this.
You may be able to load partitions in parallel by making only minor changes to
the load process you use with a non-HALDB database. Figure 11-2 on page 186
illustrates a situation where the original load program does not have to be
modified. Its input file is split into multiple files. There is a separate file for each
partition’s key range. The original program is executed multiple times. Each
execution reads the input file for its partition’s key range.
Chapter 11. Application considerations 185

Non-HALDB process HALDB parallel load process
First step creates file First step creates a file for each partition
read by load program
Load program is unchanged
Create
Create load multiple
input file load input
files
Load Load Load Load

program program program program
Partition 1 Partition 2 Partition 3
Non-HALDB HALDB
Figure 11-2 Converting a load process to parallel loads
Each partition may be loaded only once. When you load partitions in parallel, two
jobs cannot load records in the same partition. An attempt to load more records
in a partitions that has already been loaded will result in the original records
being lost.
11.1.4 Initial loads of databases with secondary indexes

The initial load of a database with secondary indexes also creates the secondary
indexes. We explain an exception to this in the next section, “BLDSNDX
parameter” on page 187. A load of a non-HALDB database with secondary
indexes requires utility steps to create the indexes. These steps are not required
with HALDB databases.
When source segments are loaded in HALDB databases, their secondary index
entries are inserted in the secondary index partitions. The buffer pools used by
the load program must include buffers for the secondary indexes. The secondary
index partitions must have been initialized previously.
Restriction: Partition initialization turns the image copy needed flags on for
data sets in the partitions.
When we wrote this book, initial loads of indexed databases failed if the image
copy needed flags were on for secondary index partition data sets. IBM plans
to change this with maintenance to IMS Version 8. The maintenance will allow
authorizations for initial loads when the image copy needed flags are on for
secondary index data sets.
Initial loads do not create entries in the ILDS. They are not needed because any
secondary index entries will have accurate RBA pointers in their extended pointer
sets (EPSs).
BLDSNDX parameter
APARs PQ55840 for IMS Version 7 and PQ56463 for IMS Version 8 added a
capability to eliminate the building of secondary indexes by initial load programs.
This is invoked by including a BLDSNDX parameter on the OPTIONS statement
in the DFSVSAMP data set. Example shows a use of this statement.
Example 11-1 BLDSNDX statement in DFSVSAMP

DFSVSAMP DD *
OPTIONS,BLDSNDX=NO
VSRBF=4096,500
IOBF=(8192,200)
/*
If BLDSNDX=YES is specified or if a BLDSNDX parameter is not included,

secondary index entries are created by the initial load.
If you include a BLDSNDX=NO parameter, you must build your secondary index
by other means. You can use the IMS Index Builder tool to do this.
When you use BLDSNDX=NO, the secondary index partitions are not authorized
and their data sets are not allocated.
The use of BLDSNDX=NO and the IMS Index Builder may shorten the load and
index creation process. When BLDSNDX=NO is not used, inserts in secondary
indexes are random. This could lengthen the load process considerably. IMS
Index Builder reads the loaded database, creates index entries without writing
them to the secondary indexes, sorts the entries in secondary index key
sequence, and then writes them to the secondary indexes in a sequential
process. When you have many secondary index entries this process may require
less elapsed time.

11.1.5 Initial loads of databases with logical relationships
The process for initially loading databases with logical relationships differs with
HALDB. When you load logical children in a non-HALDB database, your load
programs create work files. These work files are processed by the Prefix
Resolution and Prefix Update utilities. They create the appropriate pointers and
counters in your logically related databases. HALDB does not use work files or
the Prefix Resolution and Prefix Update utilities. Logical relationships cannot be
created as part of the load process. This means that load programs cannot insert
logical child segments.
If you have a database that you initially load as part of your normal operations
and it contains logical children, you will have to modify your load process. The
logical children must be inserted in the database in a different job or step. This
program must use PROCOPT=I or A, not PROCOPT=L.
Figure 11-3 illustrates a logical relationship. Logical child LC2 in database DBX
points to logical parent LP2 in database DBY. Logical child LC1 in database DBY
points to logical parent LP1 in database DBX. LC2 and LC1 are paired logical
children.
DBX DBY
LP2
LP1 LC1
LC2
LPn - logical parent LCn - logical child

Figure 11-3 Logical children and logical parents
Load steps for non-HALDB logically related databases

The steps for loading the databases in Figure 11-3, if they are non-HALDB, are:
1. Execute the Prereorganization utility.
2. Execute the initial load application program for DBX. This includes the loading
of logical children LC2. The program creates a work file.
3. Execute the initial load application program for DBY. This includes the loading
of logical children LC1. The program creates a work file.
4. Execute the Prefix Resolution utility. This utility sorts the work file records.
5. Execute the Prefix Update utility. This utility reads the output of the Prefix
Resolution utility and updates the pointers and counters for the logical
relationships.
Load steps for HALDB logically related databases

The steps for loading the databases in Figure 11-3 on page 188 (if they are
HALDBs) do not include the Prefix Resolution and Prefix Update utility steps.
The Prereorganization utility is used, but it has a different purpose. It is used to
initialize the partitions in the databases. Since logical children cannot be loaded,
they must be added by an update (PROCOPT=I or A) program. The steps are:
1. Execute the Prereorganization utility to initialize the partitions in DBX.
2. Execute the Prereorganization utility to initialize the partitions in DBY.
3. Execute the initial load application program for DBX. Do not load logical
children LC2. The program does not create a work file.
4. Execute the initial load application program for DBY. Do not load logical
children LC1. The program does not create a work file.
5. Execute an update application program for a logical database. Insert either
logical children LC1 or LC2, but not both. The insertion of either of the paired
segments will create its paired logical child.
Steps 3 and 4 can use modified versions of the application programs used in
steps 2 and 3 for the non-HALDB process. The modifications would eliminate the
inserts of the logical children. The modifications for one of these programs
should also save the information about the logical children. This information is
needed by the update program used in step 5. For example, the modification of
the program that loads DBX could write a file with a record for each LC2
segment. The record would include the concatenated key and the data for the
segment. Then the program in step 5 could read this file and insert the LC2
segments using a PCB with PROCOPT=I.
11.2 Processing partitions in parallel

You may find it advantageous to have different instances of your programs
process the partitions of a database in parallel. They could reduce the elapsed
time required for the processing. The following sections discuss how this may be
done with HALDB.
11.2.1 DBRC authorization of partitions

DBRC authorization applies to HALDB partitions, not the entire database.
Different partitions of the same database may be authorized to different

subsystems concurrently. For example, two batch jobs may be updating different
partitions at the same time without the use of the IRLM. This is not possible with
non-HALDB full-function databases. Authorization for non-HALDB full function
applies to the entire database. HALDB authorization of partitions is similar to the
authorization of areas for Fast Path DEDBs. Both HALDB partitions and DEDB
areas are authorized independently.
11.2.2 Parallel processing within an online system

You can easily process different partitions in parallel within an online system.
Multiple dependent regions may process records in the same or in different
partitions. Data sharing is not required.
11.2.3 Parallel processing with block level data sharing

If you use block level data sharing, you can easily process different partitions in
parallel with multiple subsystems. The subsystems may be online systems or
batch jobs. Block level data sharing requires the IRLM and the registration of
databases at SHARELVL=2 or 3.
11.2.4 Parallel processing without block level data sharing

You may be able to process different partitions in parallel from different
subsystems without using block level data sharing, but there are some
restrictions. A partition may be authorized concurrently to multiple subsystems,
such as multiple batch jobs, if the authorizations are only for reads. If any
subsystem is authorized to update a partition, no other subsystem can have
update or read with integrity authorization. Other subsystems could have read
without integrity (ACCESS=RO or PROCOPT=GO) authorizations. This
restriction may not be serious if each program updates only one partition.
Different programs could be processing different partitions simultaneously. On
the other hand, if each program must also update another database, it is less
likely that you can restrict each program to a different partition of the other
database.
Figure 11-4 on page 191 illustrates a situation where parallel processing cannot
be done. Programs that update different partitions of the target database update
all of the partitions of the secondary index. For example, an insert of a target
database record with a key of CML might create a secondary index with a key of
632. An insert of a record with a key of RZL might create a secondary index with
a key of 980. The target database records are in different partitions but their
secondary index entries are in the same partition of the secondary index. Similar
situations exist with logical relationships.
Secondary Index
000-499 500-999
AAA-JZZ KAA-PZZ QAA-ZZZ
Target Database
Figure 11-4 Database partitions and secondary index partitions
In some cases the use of partition selection exit routines can make parallel
processing of partitions with secondary indexes or logical relationships feasible.
Figure 11-5 illustrates as example of this. The keys of the target database and
the secondary index have some common data. The third byte of the target
database key contains a one-byte code that also appears in the second byte of
the secondary index key. The valid values for code are A, B, and C. You can write
partition selection exit routines that assign records to partitions based only on
this code. This will cause all secondary index entries pointing to the partition of
the target database to reside in the corresponding partition of the secondary
index.
Secondary Index
nAnnn nBnnn nCnnn
xxAxx xxBxx xxCxx
Target Database
Figure 11-5 Alignment of secondary index and target database partitions

11.2.5 Program modifications for parallel processing
If you want your application programs to process partitions in parallel, you will
probably need to modify the existing programs. The following changes may be
required:
򐂰 Restriction of the program to one or a set of partitions
򐂰 Splitting of the inputs to feed multiple programs
򐂰 Consolidation of the outputs of multiple programs
򐂰 Handling the inability to access a partition
You will need to restrict your program to one partition or a set of partitions. This
can be done easily with the HALDB control statement. This statement limits a
PCB to one partition of a database. The specification of this statement and its
use are described in 11.3, “Restricting a PCB to one partition” on page 193. If
you wish to use a PCB to access multiple partitions, you cannot use the HALDB
control statement. Instead, your program must understand which records reside
in these partitions and access only those records. Information about which
records are in which partitions is not available from control blocks or calls that the
application may use. You must devise a scheme to provide this information to
your programs. We will not discuss this further. It is much easier to design
systems so that application programs either have access to all partitions or are
limited to only one partition.
Your current program may read an input file. If you wish to modify it so that
multiple instances of the program process different partitions, you need to split
the input file. Each instance should read a different file. Similarly, if your current
program reads a database to get its input, different instances of the program
need to read different parts of the database.
Your current program may produce output files or reports. If you create multiple
instances of the program, you will need to consolidate output files or reports from
the separate programs.
You may need to modify your program so that it can handle situations when a
partition is not available. An example of this requirement is a program that
typically reads records from a partition, but occasionally must access a record
out of this range. This is explained in Appendix 11.4, “Handling unavailable
partitions” on page 197.
Application programs that process a database sequentially are typically the

easiest to restrict to one partition. With key range partitioning these programs
can begin with the lowest key in the partition and terminate when they reach the
end of the partition. This is easily done with the HALDB control statement, which
is explained in 11.3, “Restricting a PCB to one partition” on page 193. If the
HALDB control statement is not used, the program must know the lowest key in
the partition so that it may begin with it. It must also know the highest possible
key in the partition so that it can determine when to terminate. If the highest
possible key does not exist, the program will attempt to cross to the next partition.
The program may not have authorization to process this partition. This could
occur with a batch program not using block level data sharing. Another program
could have authorization for the next partition. This problem is avoided when the
HALDB control statement is used.
Application programs that process a database randomly may be more difficult to

restrict to one partition. Often these programs decide which records to access
based on data they read from other sources. You may need to modify how these
programs get their data from the other sources. For example, you may need to
read the other sources first and then create separate input files. Each file would
contain data for one partition and would be read by one application program.
11.3 Restricting a PCB to one partition

The HALDB control statement may be used to restrict a batch program (DLI,
DBB, BMP, or JBP) to one partition of a HALDB database. This capability was
introduced by APARs PQ57313 for IMS Version 7 and PQ58600 for IMS Version
8. The control statement may be used with DLI, DBB, BMP, and JBP regions. It
may not be used with MPP, IFP, or JMP regions, or with DBCTL or ODBA
threads.
The control statement is used with the DFSHALDB DD statement. It restricts a

PCB to one partition. It specifies the PCB and the partition.
//DFSHALDB DD *
HALDB PCB=(nnn|dddddddd,ppppppp)
nnn - DBPCB number

dddddddd - DBPCB label or name
ppppppp - partition name
Figure 11-6 HALDB control statement
The format of the statement is shown in Figure 11-6. You may specify either the
PCB label, name, or number. The capability to specify the PCB label or name
was added by APARs PQ65486 for IMS Version 7 and PQ65489 for IMS Version

8. The PCB label is the label on the PCB statement in the PSB. The PCB name is
the name of the database specified on the PCB statement. The PCB number is
the relative PCB number among the database PCBs in the PSB. Our examples
show the use of the PCB number, not the PCB label or name. You may have
multiple HALDB control statements. Each statement would restrict a different
PCB. Up to ten statements may be specified.
The HALDB control statement may be used with PHDAM, PHIDAM, and
PSINDEX databases. It restricts the PCB to the partition of the database or
secondary index specified on the PCB statement in the PSB. If a PROCSEQ
parameter is specified on the PCB, the statement restricts the PCB to a partition
of the secondary index. Otherwise, it restricts the PCB to a partition of the
database specified on the DBD parameter.
When you include a DFSHALDB DD statement you should also include a

SYSHALDB DD statement. SYSHALDB is used for information and error
messages about the HALDB control statement. You may use the SYSOUT
parameter on the SYSHALDB DD statement.
11.3.1 Processing a partition sequentially

You can easily process a partition sequentially by specifying the HALDB control
statement. When the control statement is used, a request for the first segment in
a database returns the first segment in the partition. For example, an unqualified
GN with no previous position would return the first segment in the partition. A get
next call that reaches the end of the partition returns a GB status code. Without
the control statement, the GB status code indicates that the end of the database
has been reached. These characteristics make it easy to use a program that
otherwise processes the entire database to process only a partition. Other
modifications may be required. For example, the program may produce a report
or output file. Multiple instances of the program that processed different partitions
would produce multiple reports or files. You would need to consolidate them.
11.3.2 Requesting segments outside of the partition

A call that requests a segment in another partition returns an FM status code.
For example, a GU call for key=2,546,773 with a control statement restricting the
PCB to a partition with keys between 1,000,000 and 1,999,999 would return the
FM status code. Your existing programs probably do not expect an FM status
code. If a program might request a record outside of the partition, it should be
modified to handle the FM status code before you use the HALDB control
statement.
11.3.3 Secondary index and logical relationship considerations
You should be careful when using the HALDB control statement with secondary
indexes and logical relationships. The control statement specifies a partition of
the database or secondary index that is used to enter the database. When you
specify the PROCSEQ parameter on the PCB, you must specify a partition of the
secondary index. When you do not specify PROCSEQ, you must specify a
partition of the indexed database.
Figure 11-7 shows the use of the control statement with PROCSEQ on the PCB.
The control statement restricts the PCB to one partition of the database specified
in the control statement. It does not restrict the use of partitions in the indexed
database. In Figure 11-7 the statement restricts the third PCB to partition AAA in
the secondary index. It does not restrict the partitions accessed in the indexed
database. Partition AAA has index entries pointing to each of the partitions in the
indexed database.
Control Statement PSB

HALDB PCB(3,AAA) PCB ...
PCB ...
PCB DBD=DB00A,PROCSEQ=SI001
Secondary Index: SI001

Partition AAA Partition BBB Partition CCC
Partition W Partition X Partition Y Partition Z
Indexed Database: DB00A
Figure 11-7 Using a HALDB control statement with PROCSEQ on the PCB
Figure 11-8 on page 196 shows the use of the control statement without a
PROCSEQ on the PCB. The control statement restricts the PCB to one partition
of the indexed database. It does not restrict the use of partitions in the secondary

index. In the figure the statement restricts the second PCB to partition X of the
indexed database. Partition X has index entries in all of the partitions of the
secondary index. An update of a source segment in partition X of database
DB00A could cause an update in any partition of the secondary index. Similarly,
a call qualified on a secondary index field could use any of the partitions of the
secondary index.

HALDB PCB(2,X) PCB ...
PCB DBD=DB00A,...
PCB ...
Indexed Database: DB00A

Partition W Partition X Partition Y Partition Z
Partition AAA Partition BBB Partition CCC
Secondary Index: SI001
Figure 11-8 Using a HALDB control statement without PROCSEQ on the PCB
Figure 11-9 on page 197 shows the use of the control statement with logical
relationships. The control statement restricts the PCB to one partition of the
physical database containing the root segment in logical database LDBABC. This
physical database is DB0L1. It does not restrict the use of partitions in the
physical database DB0L2. Partition XAA in database DB0L1 has logical
relationships with records in all of the partitions of database DB0L2.
Physical Databases HALDB PCB(1,XAA) PCB DBD=LDBABC,...
DBOL1 PCB ...
DBOL2
PCB ...
D
A
LP
C
B
LC
E Database: DB0L1
Partition XAA Partition XBB Partition XCC
Logical Database LDBABC
B C D
Partition L01 Partition L02 Partition L03 Partition L04

E
Database: DB0L2
Figure 11-9 Using a HALDB control statement with logical relationships
11.4 Handling unavailable partitions

A partition may not be available to an application program. This is likely to
happen if you deallocate individual partitions with /DBR commands or if you are
using multiple batch jobs without block level data sharing. A partition is
unavailable if a /DBR, /STOP, or /LOCK command has been issued for it. DBRC
flags such as recovery needed or prohibit further authorization make a partition
unavailable. A partition is unavailable for update if a /DBD command has been
issued for it or if a /STA DB command with ACCESS=RD or RO has been issued
for it. The DBRC image copy needed flag also makes a partition unavailable for
update. The use of batch jobs without block level data sharing may make
partitions unavailable. For example, your program may attempt to process
records in another partition. As explained in 11.2.4, “Parallel processing without
block level data sharing” on page 190, DBRC authorization processing may
prevent access to the other partition. If a partition may be unavailable or
unavailable for update and your program might attempt to access the partition,
your program must handle the unavailable partition condition.
Attempts to access a partition that is not available result in an unavailable data

condition. If your program has issued an INIT STATUS GROUPx call or EXEC
DLI ACCEPT STATUSGROUP command, an attempt to access an unavailable
partition returns a BA status code. If you have not issued this INIT call or this
ACCEPT command, your program will abend. A U3303 abend code is produced.
Obviously, you should include the call or command in your program if it might
attempt to access an unavailable partition.

If your program receives a BA status code due to an unavailable partition, a GN
call using the same PCB will move to the next available partition. If there are no
more available partitions, the GN call will return a GB status code indicating that
the end of the database has been reached. Programs that process databases
sequentially and that may encounter unavailable partitions, may need to issue
the GN call after receiving a BA. They also should be designed to understand
that they have not processed all of the data in the database.
There are two reasons for the unavailable data condition. One is an unavailable
partition. The other is a lock reject. A lock reject can occur when a database
record is not available due the failure of a data-sharing subsystem holding a lock
on the record. It can also occur due to the failure of a CICS® system or ODBA
thread with an in-doubt unit of work holding a lock on the database record. ODBA
threads include DB2-stored procedures. If your program receives a BA status
code, it could be due to either an unavailable partition or a lock reject. It does not
receive information indicating which is the cause.
11.4.1 Database PCB status code priming

When a program is scheduled the database PCB status code field indicates the
status of the database. This is the same for HALDB and non-HALDB databases.
The PCB does not include information about the status of partitions in the
database. A “blank blank” status code indicates that the database is available.
NA indicates that the database is not available. NU indicates that the database is
available, but not available for update. The INIT DBQUERY call may be used to
get the same information in the status code fields. That is, it restores these fields
to the settings that they had when the application was scheduled.
Since there is no information about the status of individual partitions, a “blank

blank” could indicate that the database is available, but some or all of its
partitions might not be available. The program is not given a way to determine the
status of partitions.
11.5 Processing secondary indexes as databases

If you have an application program that processes a secondary index as a
database, you may have to adjust the program and its PSB. Adjustments are
required when the secondary index uses a /SX field to create unique keys.
Adjustments may be required when symbolic pointing is used.
11.5.1 Secondary indexes with /SX fields
The /SX field is four bytes for non-HALDB databases but eight bytes with
HALDB. This increases the size of the secondary index's key field and changes
the offset to any following fields. The three changes that may be required are:
򐂰 The KEYLEN parameter on the PCB statement for the secondary index in the
PSB by must be at least the size of the VSAM key of the secondary index. It
should be increased by four bytes.
򐂰 The I/O area used by the secondary index segment must be increased by
four bytes.
򐂰 The offset to any duplicate data fields or user fields that the application
program processes must be adjusted by four bytes.
Figure 11-10 shows the change to the program I/O area when a secondary index
with a /SX field is migrated to HALDB.
Program I/O Area for Non-HALDB Secondary Index Segment

Subsequence Duplicate User
Search Field
Field with /SX Data Field Data
Program I/O Area for HALDB Secondary Index Segment

Subsequence Duplicate User
Search Field
Field with /SX Data Field Data
Figure 11-10 Program I/O area for secondary index with /SX field
Determining which programs are affected by /SX fields

You should examine your PSBs to determine if you need to make adjustments.
You may use the following steps.
1. See if you have a PSB with a PCB statement on which the DBDNAME field
specifies a secondary index database name. This indicates that the
secondary index is being processed as a database.
2. Find the INDEX parameter on the LCHILD statement in the DBD for the
secondary index. This parameter specifies the name of an indexed field in the
indexed database.
3. Find the XDFLD statement in the DBD for the indexed database whose
NAME parameter has the same value as the INDEX parameter found in step
2.

4. Determine if there is a SUBSEQ parameter on this XDFLD statement with a
value beginning with /SX. If not, you do not have a /SX field and you do not
need to make any adjustments. If there is a SUBSEQ value that begins with
/SX, a /SX field is being used. The KEYLEN parameter on the PCB statement
and the program's I/O area length should be incremented by four.
5. Determine if there is also a DDATA parameter on the XDFLD statement. If
there is, you have one or more duplicate data fields following the /SX field in
your secondary index segment. The offsets in your application program must
be adjusted by four bytes.
6. Determine if there is also user data. These are fields that are not maintained
by IMS. User application programs are required to create, delete, or replace
user data fields. When a non-HALDB database is reorganized, user data is
lost. It is not recreated when the secondary index is recreated. For these
reasons most installations do not include user data in their secondary
indexes. User data fields are created by making the segment larger than
required to hold the sequence, subsequence, and duplicate data fields.
Compare the value of the BYTES parameter in the SEGM statement of the
secondary index DBD with the combined sizes of the SRCH, SUBSEQ, and
DDATA fields defined in the XDFLD statement in the DBD for the indexed
database. If the BYTES value is larger, you have space for user data.
These conditions are illustrated by Example 11-2, which shows the PSB;
Example 11-3, which shows the secondary index DBD; Example 11-4 on
page 201, which shows the indexed database DBD; and Figure 11-11 on
page 201, which shows the secondary index segment.
Example 11-2 PSB using secondary index as a database

PCB TYPE=DB,DBDNAME=PEOSKSI,PROCOPT=G,KEYLEN=25
SENSEG NAME=SKILLCD,PARENT=0
PSBGEN LANG=COBOL,PSBNAME=SKLQRY
END
Example 11-3 DBD for secondary index with /SX and duplicate data fields
DBD NAME=PEOSKSI,ACCESS=INDEX
DATASET DD1=PESKSI,DEVICE=3390,SIZE=8192
SEGM NAME=SKILLCD,BYTES=31,PARENT=0
FIELD NAME=(SKTYPE,SEQ,U),BYTES=25,START=1,TYPE=C
LCHILD NAME=(SKILL,PEOPLE),INDEX=SKILIDX
DBDGEN
FINISH
END
Example 11-4 DBD for indexed database with /SX and duplicate data fields
DBD NAME=PEOPLE,ACCESS=(HDAM,OSAM), X
RMNAME=(DFSHDC40,1,500,824)
DATASET DD1=SKILHDAM,BLOCK=1648,SCAN=0
SEGM NAME=NAMEMAST,BYTES=150,PTR=TWIN
SEGM NAME=SKILL,PARENT=NAMEMAST,BYTES=80,PTR=TWIN
FIELD NAME=(SKILLTYP,SEQ,U),BYTES=21,START=1,TYPE=C
FIELD NAME=RATING,BYTES=6,START=22,TYPE=C
FIELD NAME=/SXA
LCHILD NAME=(SKILLCD,PEOSKSI),PTR=INDX
XDFLD NAME=SKILIDX,SRCH=SKILLTYP,SUBSEQ=/SXA,DDATA=RATING
DBDGEN
FINISH
END
Segment SKILLCD from Secondary Index PEOSKSI

SKTYPE /SXA RATING
(RBA) duplicate data
21 bytes 4 bytes 6 bytes
0 21 25
Figure 11-11 Secondary index segment with /SX and duplicate data fields
Adjusting programs and PSBs that are affected by /SX fields

The changes required when migrating to HALDB are in the examples and figures
that follow. Example 11-5 shows the PSB with the KEYLEN parameter increased
from 25 to 29 bytes. Example 11-6 shows the DBD for the secondary index with
the BYTES parameter on the SEGM statement increased from 31 to 35 bytes
and the ACCESS parameter on the DBD statement changed to PSINDEX.
Example 11-7 on page 202 shows the DBD for the indexed database with the
ACCESS parameter on the DBD statement changed to PHDAM. Figure 11-12 on
page 202 shows the secondary index segment with the larger /SX field and
resulting change in the offset to the duplicate data field.
Example 11-5 PSB using HALDB secondary index as a database

PCB TYPE=DB,DBDNAME=PEOSKSI,PROCOPT=G,KEYLEN=29
SENSEG NAME=SKILLCD,PARENT=0
PSBGEN LANG=COBOL,PSBNAME=SKLQRY
END
Example 11-6 DBD for HALDB secondary index with /SX and duplicate data field
DBD NAME=PEOSKSI,ACCESS=PSINDEX
SEGM NAME=SKILLCD,BYTES=35,PARENT=0

FIELD NAME=(SKTYPE,SEQ,U),BYTES=29,START=1,TYPE=C
LCHILD NAME=(SKILL,PEOPLE),INDEX=SKILIDX,RKSIZE=21
DBDGEN
FINISH
END
Example 11-7 DBD for indexed HALDB database with /SX and duplicate data fields
DBD NAME=PEOPLE,ACCESS=(PHDAM,OSAM), X
RMNAME=(DFSHDC40,1,100,824)
SEGM NAME=NAMEMAST,BYTES=150,PTR=TWIN
SEGM NAME=SKILL,PARENT=NAMEMAST,BYTES=80,PTR=TWIN
FIELD NAME=(SKILLTYP,SEQ,U),BYTES=21,START=1,TYPE=C
FIELD NAME=RATING,BYTES=6,START=22,TYPE=C
FIELD NAME=/SXA
LCHILD NAME=(SKILLCD,PEOSKSI),PTR=INDX
XDFLD NAME=SKILIDX,SRCH=SKILLTYP,SUBSEQ=/SXA,DDATA=RATING
DBDGEN
FINISH
END
Segment SKILLCD from HALDB Secondary Index PEOSKSI

SKTYPE /SXA RATING
(ILK) duplicate data
21 bytes 8 bytes 6 bytes
0 21 29
Figure 11-12 HALDB Secondary index segment with /SX and duplicate data fields
11.5.2 Secondary indexes with symbolic pointers

If you have an application program that processes a secondary index as a
database and the secondary index uses symbolic pointing, you may need to
make adjustments to your program. These adjustments can be avoided if the
HALDB secondary index definition accounts for the differences in pointing with
HALDB.
HALDB does not use symbolic pointers. All HALDB secondary indexes use the
extended pointer set. Symbolic pointing for non-HALDB secondary indexes
places the concatenated key of the target segment in the data area of the
secondary index segment. Figure 11-13 on page 203 illustrates its place.
Program I/O Area for Non-HALDB Sec. Index Segment using Symbolic Pointing
Subsequence Duplicate Concatenated

Search Field User Data
Field Data Field Key Field
Figure 11-13 Non-HALDB secondary index segment with symbolic pointing
When you convert the secondary index to HALDB, you can place the
concatenated key field in the segment at the same place by including it as a
duplicate data field. If it is the only or last duplicate data field, it will occupy the
same place in the data portion of the segment.
Determining if programs are affected by symbolic pointers

You should examine your PSBs to determine if you need to make adjustments.
You may use the following steps.
1. See if you have a PSB with a PCB statement on which the DBDNAME field
specifies a secondary index database name. This indicates that the
secondary index is being processed as a database.
2. Find the LCHILD statement in the DBD for this secondary index. If it includes
a PTR=SYMB parameter, this secondary index is using symbolic pointing.
Figure 11-8, Figure 11-9, and Figure 11-10 on page 204 show the PSB and
DBDs for a non-HALDB database and its secondary index, which uses symbolic
pointing.
Example 11-8 PSB using secondary index with symbolic pointing as a database
PCB TYPE=DB,DBDNAME=CONTRSI,PROCOPT=G,KEYLEN=8
SENSEG NAME=CONTR,PARENT=0
PSBGEN LANG=COBOL,PSBNAME=CONQRY
END
Example 11-9 DBD for secondary index with symbolic pointing

DBD NAME=CONTRSI,ACCESS=INDEX
DATASET DD1=CONTSI,DEVICE=3390,SIZE=8192
LCHILD NAME=(CONTRACT,ENGAGEM),INDEX=CONTRIDX,PTR=SYMB
DBDGEN
FINISH
END

Example 11-10 DBD for indexed database with symbolic pointing
DBD NAME=ENGAGEM,ACCESS=HDAM,RMNAME=(DFSHDC40,1,500,824)
DATASET DD1=ENGAHDAM,BLOCK=1648,SCAN=0
LCHILD NAME=(CONTR,CONTRSI),PTR=SYMB
DBDGEN
FINISH
END
Adjusting DBDs that are affected by symbolic pointing

You can include the symbolic field in the DBD for the secondary index by adding
a /CK field as the last or only specification in the DDATA parameter of the XDFLD
statement in the DBD for the indexed database. A FIELD statement for the /CK
field also must be added. When converting the DBDs for the secondary index
and the indexed database, you must also eliminate the specifications of symbolic
pointing. This is done by deleting the PTR=SYMB parameter on the LCHILD
statement in the DBD for the secondary index and replacing PTR=SYMB with
PTR=INDX on the LCHILD statement in the DBD for the indexed database.
Example 11-11 and Example 11-12 show the DBDs for the secondary index and
the indexed database after they have been converted to HALDB. The secondary
index includes a concatenated key field as duplicate data. The PSB is not
changed when making this conversion to HALDB.
Example 11-11 DBD for PSINDEX with concatenated key as duplicate data field
DBD NAME=CONTRSI,ACCESS=PSINDEX
LCHILD NAME=(CONTRACT,ENGAGEM),INDEX=CONTRIDX,RKSIZE=10
DBDGEN
FINISH
END
Example 11-12 DBD for indexed database with concat. key field as duplicate data
DBD NAME=ENGAGEM,ACCESS=PHDAM,RMNAME=(DFSHDC40,1,500,824)
FIELD NAME=/CK1,BYTES=18,START=1
LCHILD NAME=(CONTR,CONTRSI),PTR=INDX
XDFLD NAME=CONTRIDX,SRCH=CONTRNO,DDATA=/CK1
DBDGEN
FINISH
END
11.6 Handling test environments

In general, application testing with HALDB databases is similar to testing with
non-HALDB databases. There may be some changes required. We discuss
these considerations in the following sections.
11.6.1 DBRC registration

You must register all HALDB databases with DBRC. If you do not register
non-HALDB test database, you will have to change your test processes when
migrating to HALDB.
DBRC only allows one instance of a database to be registered in a set of

RECONs. Multiple databases with the same database name are not allowed.
This implies that you must register multiple test versions of the same HALDB
database in different sets of RECONs. Since a single batch job or online system
can use only one set of RECONs, you must define all of its HALDB databases
and registered non-HALDB databases in the same set of RECONs. This could
force you to change some of your test procedures. If your batch test jobs shared
databases that were not updated but had separate instances of databases that
were updated, you are affected. You cannot use this test configuration with
HALDB. Instead, you must have separate instances of all databases.
Figure 11-14 on page 206 shows a non-HALDB test environment. Batch jobs A
and B update separate copies of database X. They share databases Y and Z,
which they only read. The databases are not registered.

Batch Job A Batch Job B
Update Update
Read Read
DB X DB X
Version A Version B
DB Y DB Z
Figure 11-14 Non-HALDB test environment
Figure 11-15 shows a corresponding HALDB test environment. Batch jobs A and
B each have their own copies of all databases and their own set of RECONs. All
of the databases are registered in the corresponding RECONs. The registration
requirement with HALDB may require you to make similar changes.
Batch Job A Batch Job B
Update Update
DB X Read Read DB X
Version A RECONA RECONB Version B
DB Y DB Z DB Y DB Z
Copy A Copy A Copy B Copy B
Figure 11-15 HALDB test environment
11.6.2 Testing with non-HALDB databases

You may want to do some testing with non-HALDB databases. Application
functional testing may be done with non-HALDB databases. The incompatibilities
between HALDB and non-HALDB databases are minor. They will not affect most
programs. The incompatibilities apply to initial loads of logical children,
processing some secondary indexes as databases, and taking advantage of
HALDB exclusive capabilities. These capabilities include limiting a batch job, or a
BMP or JBP region, to a single partition; processing partitions in parallel; and
making some partitions unavailable to a subsystem. Most processing will not be
affected by these incompatibilities. This allows you to use non-HALDB versions
of the databases for testing. Of course, you would want to do system and
regression testing with HALDB databases before using them in production.
11.7 Copying databases to different environments

You may want to copy a database from one system to another. Some users do
this to build a test database from production data. Others copy a database from
an online system for processing by batch jobs such as report generators. You
could have other reasons. HALDB adds additional considerations for these types
of copies.
11.7.1 The partition ID

The partition ID is a number assigned to a partition when it is defined. Each
newly defined partition is assigned a new number. The last partition ID number
assigned is stored in the master database record in the RECONs. It is
incremented when a new partition is defined. You cannot change this number.
Partition IDs are stored in database data sets and in the RECONs. A partition
cannot be opened if the ID stored in the data set does not match the ID stored in
the RECONs. This limits the kinds of copies that may be done between systems.
IBM plans to address this limitation. See 11.7.7, “Planned enhancement for
copying partition definitions” on page 209, for a discussion of the planned
enhancements.
11.7.2 Using image copies

Image Copy utilities copy the partition ID that is stored in the database data set.
When the image copy is restored by a Recovery utility, the partition ID is
restored. If the image copy is restored to a different system, the data set is
unusable if the partition ID does not match the value in the RECONs. It is unlikely
that they would match. Over the life of a HALDB database, partitions are likely to
be added and deleted. This is done as the space requirements and root key
distributions change. You should expect the partition IDs to be non-contiguous.
For example, a database with five partitions could have partition IDs of 1, 4, 5, 8,
and 12. It is cumbersome to define partitions for a copy of the database with
matching numbers. For this reason, you may not be able to use image copies to
create copies of a database for use in another system. IBM plans to address this
limitation. See 11.7.7, “Planned enhancement for copying partition definitions” on
page 209, for a discussion of the planned enhancements.

11.7.3 Using unloads
Unload and Reload utilities do not copy partition IDs. Reload uses the partition ID
defined in its RECONs. This makes it easy to unload data in a database and
insert it into another instance of the database using a different set of RECONs.
The new copy will have new partition IDs. You can have a different number of
partitions and you can have different partition boundaries when you use unload
and reload to create the new instance of the database.
11.7.4 Copying part of the data

You may want to copy only a part of data. This is often true when creating test
databases. You need only a subset of the data for testing. You may be able to
unload one partition and reload only its data in the test database.
Alternatively, you may use the FABHEXTR exit in the IMS High Performance
Unload tool (program number 5655-E06) to extract a subset of the database
records in a database. This tool includes a PARTEXTR control statement, which
causes the extract to unload a subset of records from each partition of a HALDB
database. This would typically provide a more representative subset of the entire
database. The output of IMS High Performance Unload may be used as input to
the HD Reload utility, which creates the test database.
11.7.5 Copying partition definitions

The Partition Definition utility (PDU) allows you to export and import partition
definitions. The export function writes partition definitions for a database to an
output data set. The import function reads this data set and creates these
partition definitions for the same database name in another set of RECONs. The
import function allows you to easily copy your partition definitions before you
move or copy a database to another set of RECONs. Of course, after the import
you may modify the partition definitions. The partition definitions created by
import do not maintain the partition IDs from the RECONs used for export.
Instead, it assigns new partition ID numbers. IBM plans to allow you to copy the
partition ID numbers with this process. These plans are explained in 11.7.7,
“Planned enhancement for copying partition definitions” on page 209.
11.7.6 Timestamp recoveries to a copy

Some installations do a timestamp recovery to create a special instance of a
database. The timestamp could be the end of a month, quarter, or year. Such
copies are useful for some kinds of report processing. With non-HALDB
databases the timestamp copy is created with image copies and, possibly,
change accumulation and log inputs. As we mentioned in 11.7.2, “Using image
copies” on page 207, the partition ID is unlikely to allow you to use this method
with HALDB databases. We explain IBM plans to eliminate this restriction in
11.7.7, “Planned enhancement for copying partition definitions” on page 209. In
the meantime, you must use another method for creating your timestamp copy.
An alternative for creating a timestamp copy of a database requires the copying

of your RECONs. The copy of the RECONs has the same partition IDs as the
original. This solves the partition ID matching problem. It allows you to use image
copies for a timestamp recovery, which creates a copy of the database. The data
set names will need to be changed if the copy will be used on the system with the
original data sets. These names may be changed by modifying the data set
name prefixes in the RECON copy. The RECON copy is used for the timestamp
recovery process. GENJCL.RECOV may be used to generate the timestamp
recovery job. This technique assumes that the database was not allocated to any
subsystem at the time to which you are recovering. The RECON copy must be
used by the job that reads the database copy.
11.7.7 Planned enhancement for copying partition definitions

IBM plans to enhance the export and import functions. The enhancement will
allow you to export and import partition IDs with the partition definitions. Import
will use the exported partition IDs for the partition definitions it creates. Then
image copies from an exporting system may be used as inputs to recoveries in
the system that imports the definitions. This will allow you to more easily copy
databases from one system to another.

Part 4
Part 4 Administration
In this part we discuss the administration of the HALDB. First we describe the
IMS commands that can be used with HALDB, and then we have a chapter for
the back up and recovery of HALDB. Different options of reorganization and the
self-healing pointers process are discussed also in this part of the book, as well
as considerations for when you need to change your existing HALDB databases.
We also provide a description and examples of how you can utilize the IMS Data
Management Tools for IMS administering the HALDB databases.

12
Chapter 12. HALDB online commands

In this chapter we discuss the use of IMS commands with HALDB. We explain
how to use the commands for entire databases and individual partitions. We
provide examples of the commands and their responses.

12.1 Online commands with HALDB
Commands and keywords used with HALDB are the same commands that are
used with other full-function databases. There are seven of these commands:
򐂰 /DBDump
򐂰 /DBRecovery
򐂰 /DISplay
򐂰 /LOCk
򐂰 /STArt
򐂰 /STOp
򐂰 /UNLock
In the command you may specify either a HALDB master database name or a
partition name.
IMS keeps track of partition statuses and database statuses separately. For
example, a partition may be stopped but its master database may be started.
Alternatively, the partition may be started but its master database may be
stopped. Before opening, authorizing, or scheduling a partition, IMS always
examines the status of the partition and the database. If either one has a status
that would prevent the action, IMS does not take the action.
Each partition has the access limitations of both itself and its master database.
For example, if the master database has an access intent of read (RD) and one
of its partitions has an access intent of update (UP), the partition cannot be
updated. Alternatively, if the master database has an access intent of update
(UP) and one of its partitions has an access intent of read (RD), the partition
cannot be updated. Similar considerations apply to other statuses that affect
access limitations, such as being stopped, locked, or DBRed.
Commands issued with a partition name act only on the partition and its
statuses. Commands issued against the master database act only on the master
database and its statuses. This means that a start of a master database does not
start any of its partitions. If they are stopped, they remain stopped.
The following sections show how to use the commands.
12.2 /DISPLAY command

The /DIS DB command may be used to display the status of a HALDB database
or partition.
In the TYPE column the master database is shown as PHDAM, PHIDAM, or
PSINDEX. Partitions are shown as PART. Non-HALDB full-function databases
are shown as DL/I.
12.2.1 Display of a HALDB master database

The display command for a HALDB master database shows the HALDB master
and, typically, each of its partitions. When using key range partitioning, the
partitions are shown in high key order. When using a partition selection exit
routine, the partitions are shown in partition name sequence.
Example 12-1 shows the output of a /DISPLAY command for a master database.
Example 12-1 /DIS DB command for master database

/DIS DB CUSTDB
DFS4444I DISPLAY FROM ID=IM1A

DATABASE TYPE TOTAL UNUSED TOTAL UNUSED ACC CONDITIONS
CUSTDB PHDAM UP
CUSTDB1 PART UP ALLOCS
*2003091/114910*
If a /DBR command has been issued for the master database, a /DISPLAY
command of the master database does not show the partitions. Figure 12-2
shows the output of the /DIS command in this situation. We explain more about
the effects of /DBR commands in 12.3, “/DBRECOVERY command” on
page 217.
Example 12-2 /DIS DB command for master database after /DBR command
/DIS DB CUSTDB

CUSTDB PHDAM UP STOPPED
*2003091/115232*
12.2.2 Display of a HALDB partition

The display command for a partition shows the partition and, typically, its master
database.
Chapter 12. HALDB online commands 215

Example 12-3 shows the output of a /DISPLAY command for a partition.
Example 12-3 /DIS DB command for a partition

/DIS DB CUSTDB1

CUSTDB PHDAM UP
*2003091/120027*
Example 12-4 shows an example of the use of the /DIS DB command to display
multiple partitions. For each partition, the master database and the partition are
displayed.
Example 12-4 /DIS DB command for multiple partitions

/DIS DB CUSTDB1 CUSTDB2 CUSTDB3 CUSTDB4
DFS4444I DISPLAY FROM ID=IM1A 284

CUSTDB PHDAM UP
CUSTDB PHDAM UP
CUSTDB PHDAM UP
CUSTDB PHDAM UP
*2003091/161000*
If a /DBR command has been issued for a master database, a /DISPLAY

command for one of its partitions does not show the master database.
Example 12-5 shows the output of the /DIS command in this situation.
Example 12-5 /DIS DB command for a partition after /DBR command for master
/DIS DB CUSTDB1

CUSTDB1 PART UP NOTOPEN
*2003091/121151*
12.3 /DBRECOVERY command
The /DBR DB command may be used to deallocate and stop a HALDB database
or partition.
12.3.1 DBR of a HALDB master database

The /DBR DB command for a HALDB master database closes and deallocates
all partition data sets and unauthorizes all partitions in the database. It also sets
flags in the control block for the master database. It does not set flags in the
control blocks for the partitions.
The /DBR DB command for a master database is used in the way that a /DBR DB
command is used for non-HALDB databases. It applies to the entire database.
Example 12-6 shows an example of the use of the /DBR command for a master
database.
Example 12-6 /DBR DB command for master database

/DBR DB CUSTDB
DFS058I 11:52:23 DBRECOVERY COMMAND IN PROGRESS IM1A

DFS0488I DBR COMMAND COMPLETED. DBN= CUSTDB RC= 0 IM1A
If you issue a /DBR command for the HALDB master database, the HALDB
partitions are not shown when a display command for the master database is
issued. A display command for the master database only shows the master
database with its STOPPED status.
12.3.2 DBR of a HALDB partition

The /DBR DB command for a partition closes and deallocates all data sets in the
partition and unauthorizes the partition. It also sets flags in the control block for
the partition. It does not set flags in the control block for the master database.
The /DBR DB command for a partition should only be used if you wish to DBR a
subset of the partitions in a HALDB database. If you wish to DBR the entire
database, issue the command for the master database, not for the partitions.
Example 12-7 shows an example of the use of the /DBR DB command for a
partition.
Example 12-7 /DBR DB command for a partition

/DBR DB CUSTDB1

DFS0488I DBR COMMAND COMPLETED. DBN= CUSTDB1 RC= 0 IM1A
Example 12-8 shows a /DIS DB command for the master database after the
/DBR command for a partition.
Example 12-8 /DIS of master database after a /DBR for a partition

/DIS DB CUSTDB

CUSTDB PHDAM UP
CUSTDB1 PART UP STOPPED, NOTOPEN
*2003091/123659*
If you issue a /DBR DB command for a partition, you must explicitly issue a /STA
DB command for the partition before it may be accessed. A /STA DB command
for the master database will not allow you to access the partition. When you issue
the /DBR DB command for the partition, flags are set in the partition’s control
block indicating its status. These flags are not affected by a /STA DB command
for the master database.
12.4 /DBDUMP command

The /DBD DB command may be used to prevent transactions or programs from
updating HALDB databases and partitions. If the data sets of the database or
partition are open for update, they are closed and reopened for read. These are
the same actions that occur with non-HALDB full-function databases when the
command is issued for them. The command may be used to ensure that image
copies taken while the command is in effect are not fuzzy.
12.4.1 DBD of a HALDB master database

The /DBD DB command for a HALDB master database closes and reopens all
partition data sets and reauthorizes all partitions in the database. It also sets
flags in the control block for the master database. It does not set flags in the
control blocks for the partitions.
The /DBD DB command for a master database is used in the way that a /DBD DB
command is used for non-HALDB databases. It applies to the entire database.
12.4.2 DBD of a HALDB partition
The /DBD DB command for a partition closes and reopens all data sets in the
partition and reauthorizes the partition. It also sets flags in the control block for
the partition. It does not set flags in the control block for the master database.
The /DBD DB command for a partition should only be used if you wish to DBD a
subset of the partitions in a HALDB database. If you wish to DBD the entire
database, issue the command for the master database, not for the partition.
12.5 /START command

The /START DB command may be used to allow transactions or programs to
update HALDB databases and partitions or to change their access intents. The
command has the same functions that it has with non-HALDB databases.
12.5.1 Start of a HALDB master database

The /START DB command for a HALDB master database resets flags for the
master database that were set by previous /DBR, /DBD, or /STOP commands
issued for the master database. It does not reset flags in the control blocks for the
partitions. If the ACCESS parameter is used, the access intent setting for the
master database is set, but the access intent setting for the partitions in the
database are not affected. The OPEN parameter is not supported when the
command is issued for a HALDB master database.
12.5.2 Start of a HALDB partition

The /START DB command for a partition resets flags for the partition that were
set by previous /DBR, /DBD, or /STOP commands issued for the partition. It does
not reset flags in the control block for the master database. The command may
also be used to set the access intent for a partition.
The /START DB command for a partition should only be used in three

circumstances. First, previous /DBR, /DBD, or /STOP commands have been
issued for the partition. Second, you wish to change the access intent for a single
partition or subset of partitions. Third, you wish to use the OPEN parameter to
open the partition data sets. In the third case, you may issue the command after
you have issued the /START DB command for the master database.

12.6 /STOP command
The /STOP DB command may be used to stop the use of HALDB databases or
partitions. The command has the same functions that it has with non-HALDB
databases.
12.6.1 Stop of a HALDB master database

The /STOP DB command for a HALDB master database closes the partition data
sets and unauthorizes the partitions in the database. It sets a flag in the control
block for the master database. It does not set flags in the control blocks for the
partitions.
12.6.2 Stop of a HALDB partition

The /STOP DB command for a HALDB partition closes the partition data sets
and unauthorizes the partition. It sets a flag in the control block for the partition. It
does not set flags in the control block of its master database.
The /STOP DB command for a partition should only be used if you wish to stop a
partition or a subset of the partitions in a HALDB database. If you wish to stop
the entire database, issue the command for the master database, not for the
partition.
12.7 /LOCK command

The /LOCK DB command may be used to prevent the use of HALDB master
databases or partitions. The command has the same function that it has with
non-HALDB databases. It prevents subsequently scheduled programs from
accessing the database or partition. It does not close the data sets or affect
currently scheduled programs.
12.7.1 Lock of a HALDB master database

The /LOCK DB command for a HALDB master database sets a flag in the control
block for the master database. It does not set flags in the control blocks for the
partitions. The flag for the master database is turned off by an /UNLOCK DB
command for the master database.
12.7.2 Lock of a HALDB partition
The /LOCK DB command for a HALDB partition sets a flag in the control block for
the partition. It does not set the flag in the control blocks for the master database.
The flag for the partition is turned off by an /UNLOCK DB command for the
partition.
12.8 /UNLOCK command

The /UNLOCK DB command may be used to undo the actions of a /LOCK DB
command. It allows the use of the HALDB master database or partition to be
resumed. The command has the same function that it has with non-HALDB
databases.
12.8.1 Unlock of a HALDB master database

The /UNLOCK DB command for a HALDB master database resets the lock flag
in the control block for the master database. It does not reset flags in the control
blocks for the partitions.
12.8.2 Unlock of a HALDB partition

The /UNLOCK DB command for a HALDB partition resets the lock flag in the
control block for the partition. It does not reset the flag in the control block for the
master database.
12.9 Using the ALL keyword with commands

The ALL keyword may be used with IMS database commands. The ALL keyword
causes the command to act on all HALDB master databases, but not on HALDB
partitions. For example, a /START DB ALL command would start all HALDB
master databases but would not start any partitions.
If you DBR, DBD, or STOP a partition, the /START DB ALL command will not
reset the condition. You must specify the partition explicitly in the /START DB
command.
12.10 Commands while databases are in use

/DBR, /DBD, and /START commands may be rejected. For non-HALDB
databases this occurs when a PSB for a currently scheduled BMP or JBP has a

PCB that references the database. The same is true with HALDB master
databases. If the PSB for a currently scheduled BMP or JBP references a
HALDB database, a /DBR, /DBD, or /START command for the master database
is rejected. Message DFS0565I is issued in these cases.
When no currently scheduled BMP or JBP references the database, but a

currently scheduled MPP, IFP, JMP, DBCTL thread, or ODBA thread references
the master database, the command waits for the program to terminate. Then the
command is processed.
Commands for HALDB partitions are treated differently. If the command

references a partition, it is rejected only if a currently scheduled BMP or JBP has
accessed the partition in its current scheduling. The DFS0565I message is
issued.
When no currently scheduled BMP or JBP has accessed the partition, but a
currently scheduled MPP, JMP, IFP, DBCTL thread, or ODBA thread has
accessed the partition, the command waits for the program to terminate.
12.11 Command examples

This section shows some combinations of commands that illustrate the use of
commands for master databases and partitions.
Example 12-9 shows the results of a /STOP command for a master database.
The partition attributes that are shown in the CONDITIONS column are not
changed by the /STOP command for the master database. The partitions cannot
be accessed because the master database, CUSTSI, is stopped.
Example 12-9 /STOP command for a master database

/DIS DB CUSTSI

CUSTSI PSINDEX UP
CUSTSI1 PART UP ALLOCS
*2003091/171903*
/STOP DB CUSTSI
DFS058I 17:24:27 STOP COMMAND IN PROGRESS IM1A

DFS0488I STO COMMAND COMPLETED. DBN= CUSTSI RC= 0 IM1A
/DIS DB CUSTSI

CUSTSI PSINDEX UP STOPPED
*2003091/172536*
Example 12-10 shows the results of a /DBR command for a partition followed by
a /START command for its master database. The partition remains unavailable
after the /START command because it does not reset the flags for the partition.
/DISPLAY commands are included to show the results of the /DBR and /START
commands.
Example 12-10 /DBR for a partition and /START for its master database
/DIS DB CUSTDB
DFS4444I DISPLAY FROM ID=IM1A 312

CUSTDB PHDAM UP
*2003091/164943*
/DBR DB CUSTDB1

DFS0488I DBR COMMAND COMPLETED. DBN= CUSTDB1 RC= 0 IM1A
/DIS DB CUSTDB

CUSTDB PHDAM UP
*2003091/165254*
/STA DB CUSTDB
DFS058I 16:54:33 START COMMAND IN PROGRESS IM1A

DFS0488I STA COMMAND COMPLETED. DBN= CUSTDB RC= 0 IM1A
/DIS DB CUSTDB

CUSTDB PHDAM UP
*2003091/165608*
Example 12-11 shows the results of a /STOP command for a partition followed by
a /START DB ALL command. After the /STOP command, the /DISPLAY
command response shows that the master database conditions have not
changed. After the /START command, the partition remains unavailable because
the ALL keyword does not cause the command to act on partitions. It only acts on
the master database. /DISPLAY commands are included to show the results of
the /STOP and /START commands.
Example 12-11 /STOP for a partition and /START DB ALL

/DIS DB CUSTDB

CUSTDB PHDAM UP
CUSTDB1 PART UP NOTOPEN, ALLOCS
*2003091/170647*
/STOP DB CUSTDB2
DFS058I 17:07:37 STOP COMMAND IN PROGRESS IM1A

DFS0488I STO COMMAND COMPLETED. DBN= CUSTDB2 RC= 0 IM1A
/DIS DB CUSTDB

CUSTDB PHDAM UP
+ CUSTDB2 PART UP STOPPED, NOTOPEN,
ALLOCS
*2003091/170838*
/START DB ALL
DFS058I 17:09:31 START COMMAND IN PROGRESS IM1A

DFS0488I STA COMMAND COMPLETED. KEYWORD ALL IM1A
/DIS DB CUSTDB

CUSTDB PHDAM UP
+ CUSTDB2 PART UP STOPPED, NOTOPEN,
ALLOCS
*2003091/171025*
Tip: If you have explicitly stopped, deallocated, locked, or reset the access
intent for a partition, you cannot reset the condition with a command for its
master database or by using the ALL keyword. You must issue the /START or
/UNLOCK command using the partition name.

13
Chapter 13. Backup and recovery

In this chapter we describe the backup and recovery processes for HALDB
databases. We explain the different techniques used with different types of data
sets. We provide examples of the JCL used for these utilities. We show DBRC
GENJCL commands to generate JCL.

13.1 Backup and recovery overview
The backup and recovery processes used with HALDB databases vary
depending on the data set types.
Most data sets are backed up with image copies and recovered with the
Database Recovery utility (DFSURDB0). These are the same processes that are
used for non-HALDB databases. These processes are used for PHDAM and
PHIDAM data data sets and PSINDEX data sets. These data sets have A–J
suffixes for their DD names.
PHIDAM primary index data sets and ILDSs do not use these processes. They
are not backed up with image copies. Database change log records are not
written for them. The Index/ILDS Rebuild utility (DFSPREC0) is used to rebuild
these data sets.
13.2 Image copies for HALDB data sets

Image copies are used to back up the HALDB database data sets that have A–J
suffixes. These are PHDAM and PHIDAM data data sets and PSINDEX data
sets. Image copies are not used for PHIDAM primary indexes or ILDSs. IMS
provides three image copy utilities:
򐂰 The Database Image Copy utility (DFSUDMP0)
򐂰 The Database Image Copy 2 utility (DFSUDMT0)
򐂰 The Online Database Image Copy utility (DFSUICP0)
All of these utilities may be used for HALDB data sets. As with all HALDB
processing, DBRC is required.
Both fuzzy and clean image copies may be made for HALDB database data sets.
The IMS Image Copy Extensions is an alternative to the use of the Image Copy
utilities provided by IMS. It has full support for HALDB. This tool is explained in
16.4, “IMS Image Copy Extensions” on page 276.
You can also use various utilities supplied by the operating system to make your
backup copies. These are called non-standard or user image copies. As with
non-HALDB databases, these copies do not interact with DBRC. You must notify
DBRC of these user image copies. Use the NOTIFY.UIC command to do this.
13.2.1 Database Image Copy utility (DFSUDMP0) example
Example 13-1 shows the Database Image Copy utility (DFSUDMP0) JCL for data
sets in database NORDDB. This database has one data set group and four
partitions, so each partition has one data set to copy. The job creates one image
copy data set for each partition. The control statements specify the master
database name and the DD name for each data set. Partition names are not
used on control statements. Dynamic allocation is used for the database data
sets.
Example 13-1 JCL for Image Copy utility (DFSUDMP0)

//IC EXEC PGM=DFSUDMP0,PARM='DBRC=Y'
//DBIMAGE1 DD DISP=(,CATLG),DSN=IMSPSA.NORDDB1.IMCOPY,
// UNIT=3390,VOL=SER=TOTIMK,SPACE=(CYL,(1,1))
//SYSIN DD *
D1 NORDDB NORDDB1A DBIMAGE1
/*
13.2.2 Database Image Copy 2 utility (DFSUDMT0) example

The Database Image Copy 2 utility (DFSUDMT0) is an alternative to the Image
Copy utility (DFSUDMP0). It allows you to create clean image copies with
minimum down time for the database. It also may be used to create fuzzy image
copies. It uses the DUMP function of DFSMSdss™ to create the copy of the data
sets. Copies are made in two phases, logical copy and physical copy. The logical
copy is very quick. Updates may be resumed while DFSMSdss is creating the
physical copy. These processes are the same with HALDB and non-HALDB
database data sets.
Example 13-2 on page 230 shows the Database Image Copy 2 utility JCL to copy
four database data sets to one output data set. This capability was added to
Database Image Copy 2 in IMS Version 8. The control statements specify the
Chapter 13. Backup and recovery 229

master database name and the DD names for the data sets. Partition names are
not used on control statements. Dynamic allocation is used for the database data
set.
Example 13-2 JCL for Image Copy 2 utility (DFSUDMT0)

//IC EXEC PGM=DFSRRC00,PARM=(ULU,DFSUDMT0,,,,,,,,,,,,Y)
//DBIMAGE1 DD DISP=(,CATLG),DSN=IMSPSA.IMCOPY2.NORDDB,
//SYSIN DD *
1 NORDDB NORDDB1A DBIMAGE1 XL
S NORDDB NORDDB2A XL
/*
13.2.3 Using GENJCL to create image copy JCL

The following GENJCL command may be used to create image copy JCL. The
dbname in the command is either a partition name or a master database name:
GENJCL.IC DBD(dbname)
GENJCL.IC may be used to create either the Database Image Copy

(DFSUDMP0) or the Database Image Copy 2 (DFSUDMT0) JCL. GENJCL.OIC
may be used to create the Online Database Image Copy (DFSUICP0) JCL.
If the master database name is specified on a GENJCL command, JCL is

generated for each partition in the database. JCL is generated for all A–J data
sets in the partitions.
If a partition name is specified on a GENJCL command, JCL is generated only

for that partition. JCL is generated for all A–J data sets in the partition.
13.3 Recoveries of HALDB data sets

HALDB database data sets that have A–J suffixes are recovered with the same
processes that are used for non-HALDB database data sets. These HALDB data
sets are PHDAM and PHIDAM data data sets and PSINDEX data sets. The
recovery process may use the Database Change Accumulation utility
(DFSUCUM0) and the Database Recovery utility (DFSURDB0). Alternatively,
tools such as the IMS Online Recovery Service (ORS) may be used.
13.3.1 Change Accumulation utility with HALDB

The Change Accumulation utility (DFSUCUM0) may be used for HALDB
database data sets that have A–J suffixes. These data sets are PHDAM and
PHIDAM data data sets and PSINDEX data sets. The utility has the same
function with HALDB data sets that it has with non-HALDB database data sets. It
accumulates database changes from log records and writes them to a change
accumulation data set.
To include a HALDB data set in Change Accumulation utility processing, specify

the master database name and the DD name on the change accumulation DB0
control statement.
HALDB database data sets may be included in change accumulation groups

defined to DBRC. Different data sets from the same partition or different
partitions in the same database may be in different CA groups. Data sets from
different partitions, databases, or areas may be in the same CA group. The only
rule is that a data set cannot be in two CA groups. GENJCL.CA will produce the
appropriate control statements when HALDB database data sets are included in
a change accumulation group.
13.3.2 Database Recovery utility and ORS with HALDB

The Database Recovery utility (DFSURDB0) may be used to recover HALDB
database data sets that have A–J suffixes. These data sets are PHDAM and
PHIDAM data data sets and PSINDEX data sets. The utility has the same
function with HALDB data sets that it has with non-HALDB database data sets.
Its inputs are a combination of an image copy, a change accumulation data set,
and logs. It recovers one database data set in one execution.
To recover a HALDB data set, specify the master database name in the PARM
field of the EXEC statement of the Database Recovery utility, and specify the
master database name and the DD name on the control statement.
Since all PHDAM and PHIDAM VSAM data data sets and all PSINDEX data sets
must be defined with the REUSE attribute, they may be reused. They do not have
to be deleted and redefined before they are recovered.
The IMS Online Recovery Service (ORS) tool (product number 5655-E50) may
be used to recover IMS database data sets. Like the Database Recovery utility, it
may be used to recover HALDB database data sets that have A–J suffixes.
These data sets are PHDAM and PHIDAM data data sets and PSINDEX data

sets. It does not recover ILDSs and PHIDAM primary index data sets. The
partitions being recovered cannot be allocated to any online system, including
the one doing the recovery. A /DBR command for the partitions should be used to
deallocate them. The master database must be started on the system doing the
recovery. Do not issue a /DBR command for the master database.
Since HALDB databases must be registered in RECONs, it is not possible to

recover a HALDB database data set without using DBRC. Techniques for
creating copies of HALDB database are explained in 11.7, “Copying databases
to different environments” on page 207.
Timestamp recoveries
Timestamp recoveries restore database data sets to their state at a previous
time. They are typically done to recover from an application or operational error
that incorrectly updated a database.
When a timestamp recovery is done, typically all data sets in the database are
recovered to the same time. This avoids potential data integrity problems.
Logically related databases and secondary indexes for the database are also
recovered to the same time. If there are databases that are related by application
logic, they are also recovered to this time. These considerations apply to HALDB
databases. All partitions and all related databases should be recovered to the
same time to avoid potential data integrity problems.
When a timestamp recovery is done for a PHIDAM database, the PHIDAM

indexes must be rebuilt. The Index/ILDS Rebuild utility (DFSPREC0) is used to
do this. We explain the use of this utility in 13.4, “Rebuilding ILDS and PHIDAM
index data sets” on page 236.
If a timestamp recovery restores a database to a time preceding the last

reorganization and the database has secondary indexes or logical relationships,
the ILDSs must be rebuilt. The Index/ILDS Rebuild utility (DFSPREC0) is used to
do this. We explain the use of this utility in 13.4, “Rebuilding ILDS and PHIDAM
index data sets” on page 236.
It is possible to do a timestamp recovery of a partition without recovering all

partitions in the database to the same time. This is a dangerous practice. If the
database has a secondary index or logically related database, all of the partitions
of the index or logically related database will probably have pointers to the
recovered partition and other partitions. This would force you to recover all
partitions of the database, its secondary indexes, and logically related databases
to the same time.
GENJCL.RECOV examples
Example 13-3 shows the GENJCL.RECOV step used to generate the recovery
JCL for all partitions of the NORDDB database.
Example 13-3 GENJCL.RECOV for partitions of NORDDB

//JCLOUT DD DISP=SHR,DSN=JOUKO4.HALDB.CNTL(RECOVOUT)
//SYSIN DD *
GENJCL.RECOV NOJOB DBD(NORDDB) MEMBER(RECOVJCL)
/*
Example 13-4 shows the GENJCL.RECOV step used to generate the recovery
JCL for only partition NORDDB1.
Example 13-4 GENJCL.RECOV for partition NORDDB1

//JCLOUT DD DISP=SHR,DSN=JOUKO4.HALDB.CNTL(RECOVOUT)
//SYSIN DD *
GENJCL.RECOV NOJOB DBD(NORDDB1) MEMBER(RECOVJCL)
/*
Database recovery example

Example 13-5 shows the JCL that was generated by the GENJCL.RECOV
command in Example 13-3. Since the database has four partitions and one data
set group, four recovery steps are generated.
Example 13-5 Generated recovery JCL for database NORDDB

//RCV1 EXEC PGM=DFSRRC00,
// PARM='UDR,DFSURDB0,NORDDB,,,,,,,,,,,Y,,,,,,,,'
//STEPLIB DD DISP=SHR,DSN=IMS810C.SDFSRESL

//IMS DD DISP=SHR,DSN=IMS810C.DBDLIB
//NORDDB1A DD DSN=IMSPSA.IM0A.NORDDB.A00001,
// DISP=OLD,
// DCB=BUFNO=10
//DFSUDUMP DD DSN=JOUKO3.IMCOPY2.NORDDB.C01,
// DISP=OLD,DCB=BUFNO=10
//DFSVDUMP DD DUMMY
//DFSUCUM DD DUMMY
//DFSULOG DD DUMMY
// DSN=IMS810C.PROCLIB(DFSVSMDB)
//SYSIN DD *
S NORDDB NORDDB1A
/*
// DISP=OLD,
// DCB=BUFNO=10
//DFSUDUMP DD DSN=IMSPSA.IMCOPY.NORDDB2,
//DFSVDUMP DD DUMMY
//DFSUCUM DD DUMMY
//DFSULOG DD DUMMY
//SYSIN DD *
S NORDDB NORDDB2A
/*
// DISP=OLD,
// DCB=BUFNO=10
//DFSVDUMP DD DUMMY
//DFSUCUM DD DUMMY
//DFSULOG DD DUMMY
//SYSIN DD *
S NORDDB NORDDB3A
/*
// DISP=OLD,
// DCB=BUFNO=10
//DFSVDUMP DD DUMMY
//DFSUCUM DD DUMMY
//DFSULOG DD DUMMY
//SYSIN DD *
S NORDDB NORDDB4A
/*
Example 13-6 shows the JCL that was generated by the GENJCL.RECOV
command in Example 13-4 on page 233. The GENJCL.RECOV command
specified a partition name. The database has only one data set group, so only
one data set is recovered for the partition.
Example 13-6 Generated recovery JCL for partition NORDDB1

//*
// DISP=OLD,
// DCB=BUFNO=10
//DFSUDUMP DD DSN=IMSPSA.NORDDB1.IMCOPY,
//DFSVDUMP DD DUMMY
//DFSUCUM DD DUMMY
//DFSULOG DD DUMMY

//SYSIN DD *
S NORDDB NORDDB1A
/*
13.4 Rebuilding ILDS and PHIDAM index data sets

PHIDAM primary index data sets and ILDSs do not use the same backup and
recovery processes that are used for other HALDB data sets. They are not
backed up with image copies. Database change log records are not written for
them. They are not recovered with the Database Recovery utility. In a sense, they
are never recovered. They are rebuilt. The Index/ILDS Rebuild utility
(DFSPREC0) is used to rebuild these data sets.
The Index/ILDS Rebuild utility scans the HALDB partition and recreates the
PHIDAM primary index, the ILDS, or both. A single execution of the utility will
recover only one partition.
The master database name is specified in the PARM field of the EXEC
statement. The SYSIN statement specifies the partition name and the recovery
type. RECOVTYP can be either INDEX, ILE, or BOTH.
INDEX rebuilds only the PHIDAM index data set. ILE rebuilds only the ILDS.
BOTH rebuilds the PHIDAM index data set and the ILDS. When BOTH is used,
the utility rebuilds the index and then rebuilds the ILDS. The ILDS cannot be
rebuilt if the index does not exist.
Only one SYSIN control statement is allowed.
The DFSVSAMP data set provides buffer pool definitions for the data sets being
read and written in the process.
Example 13-7 shows the JCL to rebuild the ILDS and the PHIDAM index data set
for partition NORDDB1 in database NORDDB.
Example 13-7 JCL for rebuilding ILDS and index data set
// PARM='ULU,DFSPREC0,NORDDB,,,,,,,,,,,Y,N'
//*
//SYSIN DD *
PARTITION=NORDDB1,RECOVTYP=BOTH
/*
If the database has no secondary index or logical relationships, there will be no

records in ILDS. The rebuild job will complete with RC=04 when BOTH or ILDS is
specified for the RECOVTYP.
The PHIDAM index data set or ILDS must be deleted and redefined before it is
rebuilt. The VSAM REUSE attribute cannot be used for these data sets.
13.4.1 Performance of the Index/ILDS Rebuild utility

The Index/ILDS Rebuild utility scans the partition to rebuild the PHIDAM index or
ILDS. When rebuilding both data sets, it rebuilds the index before rebuilding the
ILDS.
When rebuilding the PHIDAM index, the utility reads all of the root segments in
the partition. It only needs to read the first data data set. This is the one with the
A prefix.
If the roots have twin pointers, the utility reads blocks until it finds a root segment.
When it finds the a root segment, it follows the twin backward pointers until it
finds the first root in the partition. It then follows twin forward pointers to find all of
the roots and rebuild their index entries.
If the roots do not have twin pointers, the utility reads all of the blocks in the data
set. It builds index entries for each root that it finds.
When rebuilding the ILDS, the utility does unqualified GN calls for all segments
that are targets of Extended Pointer Sets (EPSs). These are secondary index
targets, logical parents for unidirectional logical relationships, and paired logical
children for bidirectional logical relationships.
The utility uses database buffer pools. You must supply a DFSVSAMP statement
with buffer pools defined for the data sets being read and written. As with other
IMS database processing, performance is often optimized with a large number of
buffers.
If the database uses OSAM, you should use OSAM sequential buffering. The
read process is almost always a sequential process. In fact, when recovering the
PHIDAM index without twin pointers for the roots, the read is strictly sequential.
OSAM sequential buffering will eliminate waits for reads.

13.4.2 Using GENJCL.USER to generate Index/ILDS Rebuild JCL
You can use GENJCL.USER to generate the JCL for an Index/ILDS Rebuild job.
IMS supplies a sample DBRC skeletal JCL member for this purpose. The
member is DSPUPJCL. It is in the SDFSISRC data set. This is the data set that
contains IMS supplied DBRC skeletal JCL members as well as sample
application and miscellaneous source modules.
Example 13-8 shows the syntax of the GENJCL.USER command with member
DSPUPJCL. User key %MDBNAME specifies the master database name. User
key %DBNAME specifies the partition name. User key %RECTYP specifies the
type of recovery. It may have values of INDEX, ILE, or BOTH.
Example 13-8 Syntax of GENJCL.USER to generate DFSPREC0 JCL

GENJCL.USER MEMBER(DSPUPJCL)
USERKEYS((%MDBNAME,’master dbname’),
(%DBNAME,’partition name’),
(%RCVTYP,’recovery type’))
Example 13-9 shows the GENJCL.USER job used to generate the rebuild JCL
for both the ILDS and the PHIDAM index data set of partition NORDB1 in
database NORDDB.
Example 13-9 GENJCL.USER for DFSPREC0

//JOUKO3S JOB (999,POK),JOUKO3,CLASS=A,NOTIFY=&SYSUID,
// MSGLEVEL=(1,1),MSGCLASS=X,REGION=0M
/*JOBPARM SYSAFF=SC42
//JCLPDS DD DISP=SHR,DSN=IMSPSA.IMS0.SDFSISRC
//JCLOUT DD DISP=SHR,DSN=JOUKO3.HALDB.CNTL(PREC0)
//SYSIN DD *
GENJCL.USER NOJOB MEMBER(DSPUPJCL) -
USERKEYS((%MDBNAME,'NORDDB'),(%DBNAME,'NORDDB1'),(%RCVTYP,'BOTH'))
/*
Example 13-10 shows the JCL generated by the job in Example 13-9.
Example 13-10 Generated DFSPREC0 JCL

//UPREC1 EXEC PGM=DFSRRC00,
// PARM='ULU,DFSPREC0,NORDDB,,,,,,,,,,,Y,N'
//SYSPRINT DD SYSOUT=A
//IMS DD DSN=JOUKO4.DBDLIB,DISP=SHR
//DFSVSAMP DD DSN=IMS.VSAM.PARM(OPTIONS),DISP=SHR
//SYSIN DD *
PARTITION=NORDDB1,RECOVTYP=BOTH
/*
13.5 Recovering HALDB secondary index data sets

HALDB secondary index database data sets may be recovered with the same
processes used to recover PHDAM and PHIDAM database data sets. You may
use GENJCL.RECOV to generate the Database Recovery utility JCL for
PSINDEX data sets. The secondary index database name is specified in the
GENJCL.RECOV command. Example 13-11 shows a job used to generate the
recovery steps for the data sets of all the partitions of the CUSTSI secondary
index.
Example 13-11 Job to generate recovery of HALDB secondary index

//JCLOUT DD DISP=SHR,DSN=JOUKO4.HALDB.CNTL(RECOVOU3)
//SYSIN DD *
GENJCL.RECOV DBD(CUSTSI) ONEJOB LIST MEMBER(RECOVJCL)
Example 13-12 shows the JCL generated by the job in Example 13-11. The
secondary index has three partitions. Each is recovered.
Example 13-12 Generated recovery JCL for HALDB secondary index

//IVPGNJCL JOB (999,POK),
// 'JJ',
// CLASS=A,MSGCLASS=X,MSGLEVEL=(1,1),
// REGION=64M
/* JOBPARM SYSAFF=SC42
// PARM='UDR,DFSURDB0,CUSTSI,,,,,,,,,,,Y,,,,,,,,'

//CUSTSI1A DD DSN=IMSPSA.IM0A.CUSTSI.A00001,
// DISP=OLD
//DFSUDUMP DD DSN=IMSPSA.IC1.CUSTSI1.CUSTSI1A.D03073.T143654,
//DFSVDUMP DD DUMMY
//DFSUCUM DD DUMMY
//DFSULOG DD DUMMY
//SYSIN DD *
S CUSTSI CUSTSI1A
/*
// DISP=OLD
//DFSVDUMP DD DUMMY
//DFSUCUM DD DUMMY
//DFSULOG DD DUMMY
//SYSIN DD *
S CUSTSI CUSTSI2A
/*
// DISP=OLD
//DFSVDUMP DD DUMMY
//DFSUCUM DD DUMMY
//DFSULOG DD DUMMY
//SYSIN DD *
S CUSTSI CUSTSI3A
/*
13.5.1 Using IMS Index Builder for recoveries

The IMS Index Builder is an alternative to using the Database Recovery utility or
IMS Online Recovery Service (ORS) to recover secondary indexes. Instead of
recovering the secondary index, you can rebuild it from the data in the indexed
database. This tool is explained in 16.3, “IMS Index Builder” on page 275.
13.6 Batch Backout utility with HALDB

The Batch Backout utility (DFSBBO00) provides the same function with HALDB
that it provides for other IMS databases. It backs out updates to databases that
were made by a transaction or batch program. Batch Backout reads logs from
batch jobs or online systems. These logs include undo records for all HALDB
databases, including PHIDAM primary index updates. No changes are required
to the use of the Batch Backout utility when HALDB databases are updated.
Dynamic allocation of HALDB database data sets may be used.

14
Chapter 14. HALDB reorganization

One of the primary advantages of HALDB is its simplified and shortened
reorganization process. In this chapter we contrast the processes used with
non-HALDB and HALDB databases. We show the unload and reload jobs used
for reorganizing HALDB databases. We explain your options for the HALDB
process.
Reorganizations of databases with logical relationships and secondary indexes

do not require the execution of utilities to update these pointers. Instead, HALDB
uses a self-healing pointer process to correct these pointers when they are used.
We explain how this process works.

14.1 The reorganization process
The reorganization of an IMS full-function database is an unload and reload of
the database. Other steps may be required for non-HALDB databases.
14.1.1 Non-HALDB reorganization overview

The reorganization of a non-HALDB database requires an unload of the entire
database by one program execution followed by a reload of the entire database
by another.
If the database has secondary indexes or logical relationships, additional steps

are required. In these cases, the reload produces a work file. The work file is
processed by the Prefix Resolution utility.
For secondary indexes, the output of Prefix Resolution is processed by the

HISAM Unload utility. Its output it processed by the HISAM Reload utility.
Alternatively, you may use a tool such as IMS Index Builder to rebuild secondary
indexes. This tool recreates secondary indexes by reading and processing either
the work file from the reload or the source and target segments from the
reorganized database.
For logical relationships, the output of the Prefix Resolution utility is processed by
the Prefix Update utility. Prefix Update updates the logical pointers.
Figure 14-1 shows the processes used to reorganize a non-HALDB database

with logical relationships and secondary indexes. A lot of time is used to update
pointers in the logically related database and rebuild the secondary indexes.
Unload DB Reload DB Resolve Log. Rel. Rebuild Sec. Indexes
SI-3
SI-2
DB-B
DB-A DB-A
SI-1
Database Scan Prefix Resolution
HD Unload HD Reload Prefix Resolution HISAM Unload
Prefix Update HISAM Reload
or
Index Builder
Figure 14-1 Reorganization of non-HALDB database
14.1.2 HALDB reorganization overview
A reorganization of a HALDB database may be done with one or more parallel
processes. These processes unload one or more partitions and reload them. If
the database has secondary indexes or logical relationships, additional steps are
not required. The HALDB self-healing process makes updates of pointers during
the reorganization unnecessary. The amount of time required for a reorganization
depends on the sizes of the partitions. Smaller partitions reduce the time. You
may reduce your reorganization time by creating more partitions and
reorganizing them in parallel.
Figure 14-2 shows the processes used to reorganize a HALDB database with
logical relationships and secondary indexes. In this case, the partitions are
reorganized by parallel processes. Each partition can be unloaded and reloaded
in less time than unloading and reloading the entire database. This is much faster
than the process for a non-HALDB database. Additionally, no time is required for
updating pointers in the logically related database or rebuilding secondary
indexes. This further shortens the process.
DB-A
Related databases not processed by
Unload
Unload
DB Reload
ReloadDB the reorganization
Part1 Part2
Part1
Secondary Indexes
SI-1 SI-2 SI-3

Unload Reload
Part 1 Part 1 Part 1
Part2 Part2
Unload Reload Part 3 Part 3 Part 3
Part3 Part3
Logically Related Database
Unload Reload DB-B

Part4 Part4
Figure 14-2 Reorganization of a HALDB database
Chapter 14. HALDB reorganization 245

14.1.3 Options for reorganization
You have several options when reorganizing a HALDB database.
򐂰 You may reorganize all of your partitions or only a subset. If only one partition
is disorganized, you may unload and reload it without processing other
partitions.
򐂰 You may reorganize partitions in parallel or you may reorganize the database
with one process. The degree of parallelism is determined by the number of
reorganization jobs you run. Each job may process one or multiple partitions.
To increase the parallelism you may increase the number of reorganization
jobs and decrease the number of partitions each job processes.
򐂰 You may reuse existing database data sets or you may delete them after they
are unloaded and allocate new data sets for the reload.
򐂰 You may add partitions, delete partitions, or change partition boundaries. We
discuss these changes in 15.2, “Changing partition definitions” on page 259.
14.1.4 Unloading HALDB partitions and databases

HALDB partitions or databases may be unloaded with the HD Unload utility
(DFSURGU0). You may unload an entire HALDB database by not including a
SYSIN DD statement. To limit the unload to a partition or set of partitions, you
must include a control statement in your SYSIN data set. The control statement
identifies the name of the first partition to unload. If you wish to unload more than
one partition, the control statement must also include the number of partitions to
unload. Consecutive partitions are unloaded. For key range partitioning, this is
determined by the high keys. When using a partition selection exit routine, this is
determined by the order assigned by the exit routine.
Example 14-1 shows a control statement to unload one partition.
Example 14-1 HD Unload control statement to unload one partition

SYSIN DD *
PARTITION=PEO02
Example 14-2 shows a control statement to unload three partitions.
Example 14-2 HD Unload control statement to unload multiple partitions

SYSIN DD *
PARTITION=PEO04,NUMBER=3
Do not include DD statements for the HALDB database data sets. HD Unload
uses dynamic allocation for HALDB data sets. This is not true for non-HALDB
databases.
You should enable OSAM sequential buffering for databases using OSAM.
You must supply buffer pools for all data sets in the partitions that are unloaded.
This includes the ILDSs.
Example 14-3 shows a sample job that unloads a HALDB partition.
Example 14-3 Sample unload JCL

//JOUKO3C JOB (999,POK),JOUKO3,CLASS=A,NOTIFY=&SYSUID,
//JOBLIB DD DSN=IMSPSA.IMS0.SDFSRESL,DISP=SHR
// DD DSN=IMSPSA.IM0A.MDALIB,DISP=SHR
//*******************************************************************
//* HD UNLOAD FOR THE PARTITION PEO02 OF PEOPLE DATABASE
//*******************************************************************
//UNLOAD EXEC PGM=DFSRRC00,REGION=1024K,
// PARM='ULU,DFSURGU0,PEOPLE,,,,,,,,,,,Y,N'
//DFSURGU1 DD DSN=JOUKO3.UNLOAD.PEO02,UNIT=3390,VOL=SER=TOTIMN,
// SPACE=(CYL,(10,5),RLSE),DISP=(NEW,CATLG)
//DFSVSAMP DD *
IOBF=(4096,50)
VSRBF=8192,50
/*
//DFSCTL DD *
SBPARM ACTIV=COND
/*
//SYSIN DD *
PARTITION=PEO02
/*
The High Performance Unload tool is an alternative to the use of HD Unload. You
may use HP Unload to unload any number of partitions or the entire database.
14.1.5 Reallocating HALDB database data sets

You do not have to delete and redefine HALDB database data sets before you
reload them. This applies to both OSAM and VSAM data sets. VSAM data sets,
other than ILDSs, must be specified with the REUSE option. HALDB honors this
option.

If you delete and redefine database data sets, you do not have to initialize the
data sets for partitions that are reloaded. You must initialize any partition whose
data sets are redefined if the partition is not reloaded. The requirements for
partition initialization are explained in 6.1, “Partition initialization function” on
page 74.
If you delete and redefine VSAM data sets, you will receive an IEC161I 152-061
message when reloading a partition. This is not an error message. It indicates
that a VSAM data set was empty when it was opened. Example 14-4 shows the
message for an ILDS.
Example 14-4 IEC161I message during reload

IEC161I 152-061,JOUKO3D,RELOAD,PEO01L,,,
IEC161I JOUKO3.HALDB.DB.PEOPLE.L00001,
IEC161I JOUKO3.HALDB.DB.PEOPLE.L00001.DATA,CATALOG.TOTICF2.VTOTCAT
Concurrent partition initialization jobs

Partition initialization steps in concurrently executing reorganization jobs for the
same database may cause job failures. The failures occur because of a timing
problem. The following example illustrates the problem. A database has two
partitions, A and B. We have two reorganization jobs, one for each partition. Job
A reorganizes partition A. Job B reorganizes partition B. The steps of the jobs
are:
1. Unload the partition.
2. Delete the partition data sets.
3. Define the partition data sets.
4. Set the partition init needed (PINIT) flag for the partition.
5. Execute the Prereorganization utility.
6. Reload the partition.
Step 5 initializes all partitions for which the PINIT flag is set. Step 5 from the two
jobs may execute at approximately the same time. Step 5 of job A may attempt to
initialize both partitions. First, it gets DBRC authorization for partition A. Before it
completes, step 5 of job B executes. The PINIT flag is still on for both partitions.
Step 5 of job B attempts to initialize both partitions. It requests DBRC
authorization for partition A. This fails. Job B abends due to the authorization
failure. It does not reload partition B. The solution is not to set the PINIT flag. Do
not include steps 4 and 5. This technique assumes that there is data in both
partitions. If a partition does not have data and you delete and redefine its data
sets, the partition must be initialized. Empty partitions should be rare. HD Unload
will complete with RC=4 if it unloads an empty partition. You may use this to alert
yourself to this condition.
14.1.6 Reloading HALDB partitions and databases
HALDB partitions and databases may be reloaded with HD Reload
(DFSURGL0). HD Reload reads the output file from HD Unload. You do not
specify the partitions to be reloaded. They are determined by the records in the
input file to HD Reload.
Do not include DD statements for the HALDB database data sets. HD Reload
uses dynamic allocation for HALDB data sets. This is not true for non-HALDB
databases.
You must supply buffer pools for all data sets in the partitions that are reloaded.
This includes the ILDSs.
HD Reload sets the image copy needed flag for data sets in partitions that it
loads. You should image copy them as you would any database data sets after
they have been reloaded.
Example 14-5 shows a sample job that reloads HALDB partitions. The partitions
it reloads depend on the records in the input file.
Example 14-5 Sample reload JCL

//JOUKO3D JOB (999,POK),JOUKO3,CLASS=A,NOTIFY=&SYSUID,
//JOBLIB DD DSN=IMSPSA.IMS0.SDFSRESL,DISP=SHR
// DD DSN=IMSPSA.IM0A.MDALIB,DISP=SHR
//*******************************************************************
//* HD RELOAD FOR THE PEOPLE DATABASE
//*******************************************************************
//RELOAD EXEC PGM=DFSRRC00,REGION=1024K,
// PARM='ULU,DFSURGL0,PEOPLE,,,,,,,,,,,Y,N'
//DFSUINPT DD DSN=JOUKO3.UNLOAD.PEOPLE,DISP=OLD
//DFSVSAMP DD *
VSRBF=8192,50
IOBF=(4096,50)
/*
//DFSSTAT DD SYSOUT=*

ILDS updates
HD Reload updates the ILDS for partitions that contain targets of logical
relationships or secondary indexes. It has three options for updating ILDSs. They
are:
򐂰 No control statement
򐂰 NOILDS control statement
򐂰 ILDSMULTI control statement
No control statement
When you do not specify a control statement in the SYSIN data for HD Reload,
an ILDS entry is updated or created when a target of a secondary index or logical
relationship is inserted in the partition. An entry exists if a previous
reorganization loaded the target segment in the partition. The updates to the
ILDS are done in VSAM update mode. When a CI or CA is filled, it must be split
by VSAM. Free space in the ILDS may help avoid these splits. Updates may be
random or sequential. This depends on the order in which these segments are
inserted and their ILKs. The ILDS keys are based on the ILK that is based on the
location of the target segment when it was created.
You may create free space in an ILDS by copying it using REPRO. REPRO
honors the free space parameters in the VSAM DEFINE.
You may delete and redefine the ILDS before reloading. You may want to do this
to eliminate entries in the ILDS for target segments that are no longer in the
partition. HD Reload never deletes an entry. The only way to delete these entries
is to delete and redefine the ILDS. On the other hand, an empty ILDS will contain
no free space. A reload with a large number of target segments may require a
large number of CI and CA splits.
NOILDS control statement

The NOILDS control statement was added to IMS Version 7 by APAR PQ36991
and to IMS Version 8 by PQ54227. When you specify a NOILDS control
statement in the SYSIN data, HD Reload does not update or create entries in the
ILDSs. They must be created by a separate process. The HALDB Index/ILDS
Rebuild utility (DFSPREC0) is used to do this. Separate executions of
DFSPREC0 are required for each partition. These executions may be done in
parallel and on different machines. Obviously, the use of NOILDS improves the
performance of HD Reload since it does not have to write the ILDS entries. On
the other hand, DFSPREC0 must read the partitions to create the ILDSs. We do
not recommend the use of NOILDS unless you have specific needs for it.
ILDSMULTI control statement

The ILDSMULTI control statement was added to IMS Version 7 by APAR
PQ36991 and to IMS Version 8 by PQ54227. The ILDSMULTI option only applies
to migration reloads. We explain its use in 8.4.2, “HD Reload with an ILDSMULTI
control statement” on page 100.
14.1.7 Reorganizing HALDB secondary indexes

You may need to reorganize your HALDB secondary indexes. Many installations
never reorganize non-HALDB secondary indexes. Since they are recreated with
every reorganization of their indexed databases, they do not become highly
disorganized. Reorganizations of HALDB databases do not require the
recreation of their secondary indexes. As entries are added to these secondary
indexes they may become disorganized.
HD Unload and HD Reload may be used to reorganize HALDB secondary

indexes. The restrictions and recommendations for reorganizing other HALDB
databases also apply to secondary indexes with one exception. HALDB
secondary indexes have no ILDSs. The HD Reload control statements should not
be used with secondary indexes.
14.2 The self-healing pointer process

Reorganizations of databases with logical relationships and secondary indexes
do not require the execution of utilities to update pointers. Instead, HALDB uses
a self-healing pointer process to correct logical relationship and secondary index
pointers. This process is implemented by placing a target key and an extended
pointer set (EPS) in the secondary index or logically related database and by
using an indirect list data set (ILDS) in each partition of PHDAM and PHIDAM
databases.
14.2.1 The elements of the self-healing process

Figure 14-3 on page 252 shows the elements used to implement the self-healing
pointer process.

ILDS ILE
Secondary Index or
ILK
Logically related database ILE Segment Code
Partition ID
Secondary index entry
Current reorg # : 5
KSDS Current RBA
or logical child segment
...
EPS KEY
Partition DBDS
EPS
Partition ID Current reorg # :5
Reorg # : 5
RBA
ILK
ILK
... Target segment
Figure 14-3 HALDB pointer before a reorganization
Each secondary index entry and each logical child segment contains the key of
its target record. For secondary indexes the key of the target’s root segment is
included in the prefix. For logical child segments, the concatenated key of the
logical parent is included in the segment data.
Each segment in a PHDAM or PHIDAM database has an Indirect List Key (ILK).
The ILK is unique for the segment type across the entire database. It is
composed of the RBA, partition ID, and partition reorganization number of the
segment when it was first created. The ILK for a segment never changes. It is
maintained across reorganizations.
Each secondary index entry or logical child segment has an extended pointer set
(EPS). The extended pointer set includes the ILK of its target segment. It also
contains the RBA, partition ID, and partition reorganization number for the target
segment. These parts of the EPS may not be accurate. That is, they may not
reflect the current location of the target segment or the current reorganization
number of the target segment’s partition. In Figure 14-3 they are accurate.
The target segment has an Indirect List Entry (ILE) in the Indirect List Data Set
(ILDS) for a partition. The ILE contains accurate information about the target
segment. This includes its current RBA, the correct partition ID, and the current
reorganization number for the partition. The key of the ILE is composed of the
ILK and the segment code of the target segment.
The reorganization number for a partition is physically stored in the partition’s first
database data set. This number is initialized by partition initialization or load, and
incremented with each reorganization that reloads segments in the partition.
14.2.2 Finding target segments
When IMS accesses the target segment from the secondary index entry or
logical child segment, it must first determine the partition in which the target
resides. It uses the key in the secondary index or logical child to determine the
partition. Next it must determine the location in the target partition database data
set. It compares the partition ID and reorganization number of the target partition
with the partition ID and reorganization number stored in the EPS. If they match,
IMS uses the RBA in the EPS to locate the target segment. If they do not match,
the RBA in the EPS cannot be used.
When the RBA in the EPS cannot be used, IMS uses the information in the ILE to
locate the target segment. The ILE key is found by using the ILK from the EPS
and the target’s segment code. The ILE is read from the ILDS of the partition
determined from the target’s key.
Figure 14-4 illustrates a situation where the RBA in the EPS cannot be used. In
the figure, the target partition has been reorganized three times since the EPS
was accurate. This has moved the target segment and updated the
reorganization number in the partition data set. The EPS still contains a
reorganization number of 5, but the reorganization number in the partition data
set is now 8. The information in the ILE has been updated by HD Reload. IMS
uses the ILK from the EPS to find the ILE and uses the RBA in the ILE to find the
target segment.
ILDS ILE
Secondary Index or
ILK
Partition ID
Current reorg # : 8
KSDS Current RBA
...
EPS KEY
Partition DBDS
EPS
Reorg # : 5
ILK
RBA
ILK Target segment
...
Figure 14-4 HALDB pointer after a reorganization
We will see that even though the retrieval is indirect, often the CI containing the
ILE will already be in IMS’s buffer pool. Nevertheless, it would be good to avoid

the indirect process if possible. We would prefer to get the target segment
location from the EPS without reading the ILE. The self-healing process allows
IMS to limit the use of ILEs.
14.2.3 Healing pointers

The self-healing process updates or corrects the information in EPSs. When the
ILE is used, the information about the current location of the segment in the ILE
is moved to the EPS. This allows IMS to avoid the indirect process if the EPS is
used for a later retrieval. This correction to the EPS in the database buffer pool is
always done. Due to locking considerations, the update may not be written to the
database on DASD. The buffer containing the entry or segment with the updated
EPS is marked as altered if the application program is allowed to update the
database. The call must be done with a PCB allowing updates, and the
subsystem must have an access intent for the partition that allows updates. If
updates are not allowed, the buffer is not marked as altered. When the
application reaches a sync point, it does not write buffers to DASD if they are not
marked as altered. If the updated EPS is not written to DASD, the next time it is
retrieved from DASD and used to find its target, IMS must use the indirect
process. That is, IMS must read the ILE again.
Figure 14-5 shows the EPS after it has been healed. The RBA points to the
current location. The partition ID is correct. The partition reorganization number
matches the number stored in the partition database data set.
ILDS ILE
Secondary Index or
ILK
Partition ID
Current reorg # : 8
KSDS Current RBA
...
EPS KEY
Partition DBDS
EPS
Reorg # : 6
ILK
RBA
ILK Target segment
...
Figure 14-5 HALDB pointer after the self-healing process
Performance of the self-healing process
The performance of the self-healing process may be much more efficient than
you would anticipate.
Many pointers may be healed with a small number of ILDS reads. This is due to
the use of IMS database buffering. ILDSs are database data sets. They use
database buffer pools in the same way that other database data sets use them. If
a CI is already in its buffer pool, it does not have to be read from DASD. Each ILE
is 50 bytes. You specify the CI sizes for your ILDSs. An 8 K ILDS CI holds up to
163 ILEs. A 16 K CI holds up to 327 ILEs. So a single CI may hold many ILEs.
After a reorganization IMS may need to heal many pointers to the reorganized
partitions. When there are frequent uses of the CIs in an ILDS, they tend to
remain in their buffer pool. One read of an ILDS CI may be sufficient to heal
hundreds of pointers. As with most IMS database tuning, having a large number
of buffers for frequently used data sets may be highly beneficial.
See 2.1.7, “Buffer pools” on page 25, for an explanation of how to assign HALDB
data sets, including ILDSs, to buffer pools.
Another benefit of the self-healing process is that it does not waste resources
healing pointers that are not used. In many secondary indexes, only a small
number of entries are actually used. With a non-HALDB database, the entire
index is rebuilt every time the indexed database is reorganized. With HALDB, the
index is not rebuilt and only a small number of referenced index entries are
updated. HALDB does not use resources to update pointers that are never used.
As we stated in 14.2.3, “Healing pointers” on page 254, when an EPS is updated,

an application only marks buffers altered if the application is allowed to make
updates. If updates are allowed and a block level data sharing environment is
being used, a block lock is requested for the altered block. Block level data
sharing environments exist when the IRLM is used and the share level for the
database is either 2 or 3. The block locks are held until the application program
commits its unit of work. This could cause a performance problem.
Potential self-healing performance problem

Usually update programs commit frequently. This is good programming practice.
Occasionally, a program that is allowed to do updates does not actually do them.
For example, a program with a PCB specifying PROCOPT=A may only read. In
this case, it may not commit frequently. Since it only reads, it never holds many
locks. This could change with the implementation of HALDB. If the program runs
in a block level data sharing environment and invokes the healing process, it will
hold block locks until they are committed. This could cause two problems. First, it
might hold the locks for a long time and cause other programs to wait before they
may update the blocks. Second, it could hold many locks. This could cause a
storage shortage in the IRLM or a lock structure.

If you have this potential problem, you have four alternatives to address it. First, if
the program does not make updates, you could use PROCOPT=G. Second, your
program could commit frequently. Third, you could invoke the pointer healing
process before this application is run. Another program or utility could be run
before this program. The HALDB Conversion and Maintenance Aid tool supplies
a pointer healing utility. Fourth, you could rebuild a secondary index with IMS
Index Builder. Index Builder creates EPSs with accurate RBAs.
We believe this problem is rare. We expect that most users will be able to let the
pointer healing process occur without taking any special precautions.
Tip: Do not rebuild your secondary indexes after a reorganization. Let the
self-healing process of HALDB correct the pointers. This will shorten the
outage for reorganizations and will tend to minimize the use of resources.
15
Chapter 15. Changing existing HALDB

databases
After you have migrated a database to HALDB or created a new HALDB
database, you may want to change it. Changes include modifications to the
database definition and changes to the partitions. Partitions may be changed to
accommodate more records, balance the amount of data across partitions, or
delete records. In this chapter we explain how these changes are made.

15.1 Changing database definitions
You may change the database definition of an existing HALDB database. This is
similar to the changes that you could make to a non-HALDB database. Some
changes require that the database be unloaded and reloaded. Some do not. We
discuss these changes and techniques for adding secondary indexes in the
following sections. Partitions are not part of the database definition. We discuss
changes to them in 15.2, “Changing partition definitions” on page 259.
15.1.1 Changes not requiring unload and reload of the database

Some database changes do not require you to unload and reload the database.
These are the changes that can be made to non-HALDB databases without the
unload and reload. These changes include the following:
򐂰 Adding a field definition that does not increase the size of a segment
򐂰 Deleting a field definition that does not decrease the size of a segment and is
not used with a secondary index
򐂰 Adding, deleting, or changing a data capture exit routine
򐂰 Adding, deleting, or changing the data conversion user exit routine
(DFSDBUX1)
These changes are made by updating the DBD while the database is not in use.
15.1.2 Changes requiring unload and reload of the database

Some database changes require you to unload the database, change your
definitions, and reload the database. These are the same types of changes that
require an unload and reload with non-HALDB databases. These changes
include:
򐂰 Adding or deleting a segment type
򐂰 Changing the definition of a segment size
򐂰 Changing the definition of a sequence field
򐂰 Changing the location of a segment in the hierarchy
򐂰 Changing pointer options
򐂰 Adding, deleting, or changing a segment edit/compression exit routine
15.1.3 Using online change with HALDB

DBD changes for HALDB databases may be implemented in online systems with
online change. The database cannot be in use during the online change process.
You should issue a /DBR DB command for the database before making the
change. You may issue a /START DB command for the database after the
change is committed. DBD changes affect every partition, so the master
database name should be used for these commands. For more information on
using online commands with HALDB, see Chapter 12, “HALDB online
commands” on page 213.
15.1.4 Adding a secondary index

IMS does not provide a utility to add a secondary index to an existing HALDB
database. The easiest way to add one is with a tool, such as the IMS Index
Builder. This tool reads an existing HALDB database and creates one or more
secondary indexes for it. It requires new definitions in the indexed database DBD,
but does not require changes in the database data sets. It does not need to add
any entries to the ILDSs because the pointers in the newly created secondary
index are accurate. Later, when the database or any of its partitions are
reorganized, entries for target segments will be added to the ILDSs.
If you do not have a tool to add a secondary index, you may use the following
steps:
1. Create an unload file for the indexed database. You may use HD Unload or an
application program that you write.
2. Add the secondary index definitions to the indexed database DBD.
3. Create the DBD for the secondary index.
4. Define the partitions for the secondary index.
5. Allocate the data sets for the secondary index.
6. Initialize the secondary index partitions.
7. Load the indexed database. You must provide the program to do this. It reads
the file created in step 1. When the indexed database is loaded, secondary
index entries are created. This is a random process so it may significantly
elongate the load time.
If you use HD Unload in step 1, the output file contains a header record, one
record for each segment, and a trailer record. The segment record includes the
segment name and the segment data. Your load program for step 7 may map the
records in this file by using the DSECT in the IMS DFSURGUP macro in
SDFSMAC.
15.2 Changing partition definitions

You can make changes that only affect one partition. These include:
򐂰 Changing the data set name prefix
Chapter 15. Changing existing HALDB databases 259

򐂰 Changing the randomizing module or randomization parameters for a partition
򐂰 Changing the free space parameters for a database data set
򐂰 Changing the OSAM block size or VSAM CI size for a database data set
These changes cannot be made while the partition is authorized, but other
partitions in the database may be authorized. All of these parameters, other than
the VSAM CI size, are kept in the RECONs. They are specified in the Partition
Definition utility or in DBRC CHANGE commands. If you change them, the
partition init needed (PINIT) flag is set for the partition.
You can make changes that affect multiple partitions. These include adding
partitions, deleting partitions, changing partition boundaries, and changing a
partition selection exit routine. These changes cannot be made while the affected
partitions are authorized, but other partitions in the database may be authorized.
15.2.1 Changing high key definitions

You may change the distribution of records across partitions when using key
range partitioning. This is done by adding partitions, deleting partitions, or
changing the high key specification for partitions. These changes generally
require that you unload affected partitions, change your definitions, and reload
the unloaded segments. Affected partitions are those whose range of keys have
changed. The change either moves records into or out of affected partitions.
Using the HALDB Migration Aid utility (DFSMAID0)

You may use the HALDB Migration Aid utility (DFSMAID0) to analyze existing
HALDB databases before changing high key definitions. The use of DFSMAID0
is explained in 8.2, “HALDB Migration Aid utility (DFSMAID0)” on page 93. When
the input to the utility is a HALDB database, the prefix-incr and length-incr
column values will always be zeros since segment sizes will not change. An
entire HALDB database is analyzed by DFSMAID0. It cannot be restricted to a
single partition or subset of partitions.
Adding a partition
You may want to add a partition. This is typically done when a partition grows too
large. Adding a partition causes some of the records in an existing partition to be
moved to the new partition. Figure 15-1 on page 261 shows the addition of a
partition. In the figure partition D is added to the database. Its high key is 3M.
Previously defined partitions A, B, and C had high keys of 2M, 4M, and high
values. The addition of the new partition will require the movement of records
with keys above 2M and up to 3M from partition B to partition D. This means that
partition B is affected by the change. When partition D is defined with a high key
of 3M, IMS sets the PINIT flag for partitions B and D. Partition initialization will
initialize these two partitions. Partitions A and C are not affected by the change.
The steps in the process are:
1. Unload partition B.
2. Define the new partition with the Partition Definition utility or DBRC
commands and allocate the new partition’s data sets.
3. Initialize partitions B and D.
4. Reload using the output of step 1. This loads partitions B and D. The image
copy needed flag is set for the data sets in B and D.
A B C
2M 4M High Values
1. HD Unload
.
2. Add partition D
3. Initialize B and D
4. HD Reload
A D B C
2M 3M 4M High Values
Figure 15-1 Adding a partition with key range partitioning
Deleting a partition
You may want to delete a partition. This is typically done when a partition has
little data. Deleting a partition causes its records to be moved to another partition.
Figure 15-2 on page 262 shows the deletion of partition B. Its high key is 4M. The
deletion of partition B will require the movement of its records to partition C. This

means that partition C is affected. When the definition of partition B is deleted,
IMS sets the PINIT flag for partition C. Partition initialization will initialize partition
C. Partitions A and D are not affected by the change. The steps in the process
are:
1. Unload partitions B and C.
2. Delete the definition for partition B with the Partition Definition utility or DBRC
commands.
3. Initialize partition C.
4. Reload using the output of step 1. This loads partition C. The image copy
needed flag is set for the data sets in C.
A B C D
1. HD Unload
.
2. Delete partition B
3. Initialize C
4. HD Reload
A C D
2M 6M High Values
Figure 15-2 Deleting a partition with key range partitioning
Changing the high key of a partition

You may want to change the high key of a partition. This is typically done when
the amount of data in partitions is not balanced. Changing a high key causes
some records to be moved from one partition to another. Figure 15-3 on
page 263 shows the change of the high key for partition B. Its high key is
changed from 4M to 5M. The change will require the movement of records with
keys above 4M and up to 5M from partition C to partition B. This means partitions
B and C are affected. When the definition of partition B is changed, IMS sets the
PINIT flag for partitions B and C. Partition initialization will initialize these
partitions. Partitions A and D are not affected by the change. The steps in the
process are:
1. Unload partitions B and C.
2. Change the definition for partition B with the Partition Definition utility or
DBRC commands.
3. Initialize partitions B and C.
4. Reload using the output of step 1. This loads partitions B and C. The image
copy needed flag is set for the data sets in B and C.
A B C D
1. HD Unload
2. Change high key for partition B

3. Initialize B and C
4. HD Reload
A B C D
Figure 15-3 Changing a high key with key range partitioning
Adding a higher key partition

You may add a partition that has a higher key than any existing partition. This is
likely to happen if the keys are based on time or an ascending value. As the key

values increase, you can add new partitions to hold new records. Figure 15-4 on
page 264 illustrates this situation. The key for the database is based on time. The
high order part of the key contains the year. There is a partition for each year. A
new partition is required for each new year. There are no records with keys above
2003zzz. Before they are added to the database, partition D with a high key of
2004zzz is added. Since no records are moved from an existing partition, they
are not affected. The steps in the process are:
1. Define partition D.
2. Initialize partition D. The image copy needed flag is set for the data sets in D.
A B C
2001zzz 2002zzz 2003zzz
1. Add partition D
2. Initialize D
A B C D
2001zzz 2002zzz 2003zzz 2004zzz
Figure 15-4 Adding a higher key partition with key range partitioning
Deleting the lowest key partition

You may delete the partition that has the lowest key range. This is likely to
happen if the keys are based on time or an ascending value. As the key values
increase, you can delete a partition that has become empty or whose records are
no longer of interest. Figure 15-5 on page 265 illustrates this situation. The key
for the database is based on time. The high order part of the key contains the
year. There is a partition for each year. In this example, partition A with a high key
of 2001zzz is no longer needed. Either all of its records have been deleted or
they are no longer needed. In fact, this is an efficient way to delete all the records
in partition A. Since no records are moved from an existing partition, they are not
affected. No partitions need to be initialized. There is only one step in the
process. It is to delete partition A.
A B C D
2001zzz 2002zzz 2003zzz 2004zzz
1. Delete partition A
B C D
2002zzz 2003zzz 2004zzz
Figure 15-5 Deleting the lowest key partition with key range partitioning
If there is a secondary index, you must be careful when deleting a partition. If the
secondary index has entries pointing to the deleted partition, the secondary
index becomes invalid. You must rebuild it with a tool such as the IMS Index
Builder. If the secondary index does not have entries pointing to the deleted
partition, it is not affected and does not have to be rebuilt. If there are no records
in the deleted partition, the secondary index does not have to be rebuilt. It is
unaffected by the partition deletion.
15.2.2 Making changes with a partition selection exit routine

If you are using a partition selection exit routine, you must understand which
partitions are affected by any changes you make. When a change causes
records to be moved from one partition to another, HD Unload must be run
before the changes are made. HD Reload must be run after the changes are
made. You must unload any existing partition that is affected by the change. All
affected partitions must be initialized before they are loaded by HD Reload.
Typical changes are discussed in the following sections.
Adding a partition
You may want to add a partition. This is typically done when a partition grows too
large. Adding a partition causes some of the records in one or more existing
partitions to be moved to the new partition. This is similar to adding a partition to
a database that uses key range partitioning. We explained this in “Adding a
partition” on page 260. You must unload the partition or partitions from which
records will be moved. After the unload, define the new partition and make any
necessary changes to your exit routine. You may also need to change the

partition strings for some partitions. When you use a partition selection exit
routine, IMS cannot know which existing partitions are affected by the addition of
the partition. IMS does not set the PINIT flag for existing partitions. You must set
it for them. IMS sets the flag for the new partition. Once you have set the PINIT
flag in the appropriate partitions, you initialize them. Then you run HD Reload. It
loads data in partitions according to the assignments made by your exit routine.
The image copy needed flag is set for the data sets in the partitions that are
initialized or loaded by HD Reload.
Deleting a partition
You may want to delete a partition. This is typically done when a partition has
little data. Deleting a partition causes its records to be moved to one or more
other partitions. This is similar to deleting a partition when using key range
partitioning. We explain this in “Deleting a partition” on page 261. You must
unload the partition to be deleted and any other affected partitions. These are the
partitions into which the records from the deleted partition will be moved. After
the unload, delete the definition for the partition to be deleted and make any
necessary changes to your exit routine. You may also need to change the
partition strings for some partitions. When you use a partition selection exit
routine, IMS cannot know which partitions are affected by the deletion of the
partition. IMS does not set the PINIT flag for existing partitions unless their
definitions, such as their partition strings, are changed. You must set the PINIT
flag for partitions when they are affected but you make no changes to their
definitions. Once you have set the PINIT flag in the appropriate partitions, you
initialize them. Then you run HD Reload. It loads data in partitions according to
the assignments made by your exit routine. The image copy needed flag is set for
the data sets in the partitions that are initialized or loaded by HD Reload.
Changing the exit routine

You may want to change your exit routine so that it assigns records to different
existing partitions. This is typically done when the amount of data in the partitions
is not balanced. You can make this change by switching exit routines or modifying
the existing exit routine. As with other changes, you must unload any partitions
that are affected by the change. These are the partitions from which or into which
records will be moved. After the unload, you may need to set the PINIT flag in the
affected partitions or reset it in unaffected partitions. If you change the name of
your exit routine, the PINIT flag is set in all of the partitions. If some are not
affected by the change, you should turn off their PINIT flags. You may do this with
the DBRC CHANGE.DB command. If you do not change the name of your exit
routine or change any partition definitions, you must set the PINIT flag for all of
the affected partitions. You may need to change some partition definitions, such
as their partition strings or randomizing parameters. If you change the definition
of a partition, IMS will set the PINIT flags for it. Once the PINIT flags are set
correctly, you initialize the affected partitions. After making the changes to the
exit routine, you run HD Reload. It loads data in partitions according to the
assignments made by changes. The image copy needed flag is set for the data
sets in the partitions that are initialized or loaded by HD Reload.
15.2.3 Changing partition names

You cannot change the name of a partition without deleting it and redefining it.
This will cause the new partition to have a new partition ID. We explain the
implications of changing partition ID numbers in the next section.
To change a partition name you must unload the partition, delete its definition,
redefine the partition with the new name, initialize the partition, and reload it. The
deletion of the existing definition also deletes image copy and ALLOC record
information about the partition from the RECONs.
15.2.4 Partition ID numbers

A partition ID number is assigned to a partition when it is defined. The last
number assigned is stored in the RECONs. Each new partition in a HALDB
database is assigned a new higher number. These numbers are not reused. If
you delete the definition of a partition and redefine it, it will have a new higher
partition ID. Since partition IDs are physically stored in partition data sets, data
sets created for a partition cannot be used with other partitions or with the same
partition if it is deleted and redefined.
The assignment of a new partition ID has important implications for deleting and
redefining a partition definition and the possibility of backing out our change. This
is illustrated by the example in Figure 15-6 on page 268. In this example, we
have a database with three partitions. This is shown in step 1. We decide to
delete partition C. In step 2 we unload partitions B and C, delete partition C,
initialize B, and run reload. This changes the database to two partitions. If we
change our minds and try to back out our changes, we have a potential problem.
In step 3 we redefine partition C. It has a new partition ID. Even though we may
have image copied our partition C data sets before deleting the partition in step
2, these image copies are useless. They contain partition ID 3. Our new definition
requires partition ID 4. There is an alternative. We could use our HD Unload to
reload partitions B and C after redefining C. A reload in step 3 would be
successful. Reload does not attempt to overwrite the new partition ID. After the
reload, the image copy needed flags would be set for the data sets in partitions B
and C.

1. Partition C exists 2. Partition C deleted 3. Partition C recreated
Partition Partition Partition Partition Partition Partition

Name ID Name ID Name ID
A 0001 A 0001 A 0001
B 0002 B 0002 B 0002
C 0003 C 0004
Figure 15-6 Partition ID change when deleting and recreating a partition
IBM plans to provide another alternative for this situation. We describe this
enhancement to the export and import functions of the Partition Definition utility
in 11.7.7, “Planned enhancement for copying partition definitions” on page 209.
Using this capability you could export the partition definitions along with their
partition IDs before deleting partition C. If you decided to back out your change,
you could import the definitions in step 3. Since the definitions for partitions A
and B already exist, they would not be imported. Only the definition of partition C
would be imported. It would have a partition ID of 3. This would allow you to
restore the image copies of the partition C data sets that you made before
deleting the partition. Importing only partition C would require that your execution
of import specified “Try all partitions” for its processing option.
15.3 Changing secondary indexes

Changes to secondary indexes are like changes to other HALDB databases. You
may add, delete, or modify their partitions. Affected partitions must be unloaded,
initialized, and reloaded. The indexed database is unaffected by changes to
secondary index partitions. The HALDB Migration Aid utility (DFSMAID0) may be
used to analyze changes in the high keys for secondary index partitions.
If changes to the secondary index require changes to the definition of the

indexed database, you may have to unload and reload the indexed database.
The rules governing these changes for all databases apply. We explained them
in 15.1.1, “Changes not requiring unload and reload of the database” on
page 258, and 15.1.2, “Changes requiring unload and reload of the database” on
page 258. An example of a change to a secondary index is a change in a
subsequence field. If you wish to add or modify a subsequence field, the indexed
DBD would have to be changed. If the field in the indexed database already
exists or does not require other changes to the DBD, the indexed database does
not have to be unloaded and reloaded. Of course, you will need to recreate the
secondary index using the new definitions. Since reorganizations of the indexed
database do not cause its secondary indexes to be recreated, you must use
other means for this. You may use a tool such as the IMS Index Builder.
Alternatively, you may use a technique similar to one used to add a secondary
index. We describe this in 15.1.4, “Adding a secondary index” on page 259.
When you make changes to an indexed database that do not require changes to
secondary index definitions, you do not need to make any changes to the
secondary indexes. They do not need to be unloaded, reloaded, or rebuilt. The
self-healing pointer scheme of HALDB provides this capability. The reload
process for the indexed database updates the ILDSs. This is all that is required to
ensure that the secondary index pointers may be used to find the moved target
segments.
15.4 Changing logically related databases

You may change the partitions of a logically related database without making any
changes to other logically related databases. When you add, delete, or modify
partitions in a database, you do not need to unload and reload any logically
related database. The self-healing pointer scheme of HALDB provides this
capability.

16
Chapter 16. Using IMS tools for HALDB

This chapter provides information on the following IMS tools:
򐂰 IMS High Performance Unload
򐂰 IMS High Performance Load
򐂰 IMS Index Builder
򐂰 IMS Image Copy Extensions
This chapter also provides some examples of how to use the tools with HALDB
databases.

16.1 IMS High Performance Unload
IMS High Performance Unload provides two unloaded utilities, FABHURG1 and
FABHFSU, both of which run as HSR application programs. Both unload utilities
are usually run in the ULU region. In that ULU region, all database segment
types are unload automatically.
The unload data set produced by these unload utilities can be used as input to
the IMS HD Reorganization Reload utility and to other reload utilities that are
compatible with it.
The functions and benefits of both FABHURG1 and FABHFSU include:

򐂰 Fast segment retrieval from databases
High Performance Unload’s segment retrieval engine, HSSR engine, provides
the facility to retrieve segments much faster than DL/1. Also, batch DL/I
application programs that use GN calls to read a database can use the facility.
Programs that read large portions of a database sequentially may get
significant reductions in elapsed time and CPU use. You do not have to
recompile or relink-edit the application program to use HSSR engine.
򐂰 Unload of a database without decompressing compressed segments
Both FABHURG1 and FABHFSU can unload a database that uses the
Segment Edit/Compression Exit without decompressing the segments. This
can decrease the CPU time and elapsed time.
򐂰 Unload of a selected partition or a selected sequence of partitions of a
HALDB
Both FABHURG1 and FABHFSU provide a function for unloading a particular
partition or a sequence of partitions from a HALDB. Unloading partition data
sets in parallel, using the multiple FABHURG1 or FABHFSU jobs, reduces the
elapsed time required for unloading a HALDB.
򐂰 User exit facility to allow additional processing for retrieved segments
Each of the Unload utilities provides the User Exit facility, which provides
more control on unload process than the IMS HD Reorganization Unload
utility.
򐂰 Ability to continue processing after sequence errors
High Performance Unload optionally does sequence-key checks for twin
chains.
򐂰 Read of a corrupted database
High Performance Unload options give you greater control when you are
unloading a database that has an incorrect pointer.
– The SKERROR option can be used to bypass pointer errors.
– The BYINDEX option can be used to force the roots of the HIDAM or
PHIDAM database to be accessed from the primary index.
Example 16-1 shows the JCL to unload the entire customer (CUSTDB)
database. The unloaded data set is defined by the SYSUT2 DD statement.
The HALDB Partition Definition report is produced if PARTINFO DEF is specified.
The partition-wide Segment Statistics report is produced if SEGSTAT PART is

specified.
Example 16-1 JCL to unload entire database

//UNLOAD EXEC FABHULU,MBR=FABHURG1,DBD=CUSTDB,DBRC=Y
// DD DISP=SHR,DSN=HPS.SHPSLMD0
//HSSROPT DD *
/*
//SYSIN DD *
SEGSTAT PART
//SYSUT2 DD DISP=(NEW,CATLG),DSN=JOUKO4.CUSTDB.HPUNLOAD,UNIT=SYSDA,
For more information on IMS High Performance Unload, refer to the IMS High
Performance Unload for OS/390, V1R1, User’s Guide, SC27-0936.
16.2 IMS High Performance Load

IMS High Performance Load provides high performance reload processing for
databases. It is a performance replacement of the basic HD Reorganization
Reload utility (DFSURGL0). It analyzes database performance and prepares
reports containing statistics on space use and segment pointers.
The functions of this utility include:

򐂰 High Performance Load initializes an HDAM and HIDAM database. It also
performs as an performance replacement of the IMS HALDB partition data
Chapter 16. Using IMS tools for HALDB 273

set Initialization utility (DFSUPNT0). It initializes PHDAM and PHIDAM
database partitions for the DBRC RECON record whose partition init needed
flag is on.
򐂰 High performance Load reloads an HD unloaded database data set with
compressed segments in a compressed form without calling the segment
compression routine.
򐂰 High Performance supports logical relationships and secondary indexes. This
is also true for a HALDB.
򐂰 High Performance Load generates statistics reports on data sets, segments,
and segment pointers. The reports serve as a valuable aid in tuning the
database.
Using High Performance Load might benefit an installation by:

򐂰 Reducing time and costs for reorganization of IMS databases
򐂰 Having more flexible control over the reorganization of IMS databases
򐂰 Providing partition initialization of PHDAM or PHIDAM databases within its
reload processing
򐂰 Providing statistical data needed for database tuning and space analysis
Example 16-2 shows the JCL for High Performance Load. It is basically the same
as for the IMS HD Reorganization utility, with the addition of a DD statement.
Example 16-2 JCL to load the entire database

//LOAD EXEC PGM=DFSRRC00,PARM='ULU,DFSURGL0,CUSTDB,,,,,,,,,,,Y'
// DD DISP=SHR,DSN=HPS.SHPSLMD0
//DFSUINPT DD DISP=SHR,DSN=JOUKO4.CUSTDB.HPUNLOAD
//FRRIN DD *
DATASPACE=YES
For more information on IMS High Performance Load, refer to the IMS High
Performance Load for OS/390, V1R1, User’s Guide, SC27-0938.
16.3 IMS Index Builder
The IMS Index Builder provides the features to build or rebuild IMS secondary
indexes, and to rebuild HIDAM and/or PHIDAM primary indexes. Support is
provided for both IMS non-HALDB and HALDB.
The IMS Index Builder helps to simplify index recovery and maintenance,
reduces index maintenance time, and eliminates the need to image copy index
databases. It gives an easy-to-use one-step procedure for maintaining primary
and secondary indexes.
Some of benefits of using the IMS Index Builder are:

򐂰 It minimizes the elapsed time needed to build one or more index databases.
򐂰 It uses both parallel sort and parallel load whenever more than one index is
being built. This powerful feature significantly reduces the time needed to
build multiple indexes of a single physical database.
򐂰 Multiple indexes can be built or rebuilt in a single JOB step.
򐂰 Reorganization time can be reduced by using the Index Builder instead of the
unload/load process for each index.
򐂰 Image copy time and external storage (tape or DASD) may be reduced by
eliminating the need to copy secondary indexes.
򐂰 When used in a recovery scenario, the lengthy image copy restore and log
updates process can be replaced by a single Index Builder job.
Example 16-3 JCL to build all secondary indexes from HALDB

//IMSIB EXEC PGM=IIUSTART,REGION=0M
//STEPLIB DD DISP=SHR,DSN=IIU.V2R3M0.SIIULMOD
// DD DISP=SHR,DSN=IIU.V2R3M0.SIIULMOD
//IIUIN DD *
PROC BLD_SECONDARY,CUSTDB,SELECTED
INDEX CUSTSI
INPUT IBSCAN,DBRC=Y
SORTID A,MAINSIZE=MAX FILSZ=E3000000
//*
//IIUSTAT DD SYSOUT=*
//IIUSNAP DD SYSOUT=*

//IIUSORT DD SYSOUT=*
In 8.5, “Migrating databases with secondary indexes” on page 101, we explain

how you can use the IMS Index Builder in a migration process. We present more
information in “The IMS Index Builder alternative” on page 110.
In “BLDSNDX parameter” on page 187 we explain how you may use the IMS
Index Builder to build secondary indexes after an initial load of an indexed
database. This is an alternative to building the secondary index as part of the
initial load. This use of IMS Index Builder may supply substantial performance
benefits.
In 13.5.1, “Using IMS Index Builder for recoveries” on page 241, we explain how
IMS Index Builder may be used for recoveries of secondary indexes.
We explain how to use IMS Index Builder to add a secondary index to an existing
HALDB database in 15.3, “Changing secondary indexes” on page 268.
For more information on the IMS Index Builder, refer to the IMS Index Builder for
Z/OS, V2R3 User’s Guide, SC27-0930.
16.4 IMS Image Copy Extensions

Image Copy Extension (ICE) for OS/390 extends the capabilities of the IMS
Database Image Copy utility (DFSUDMP0). It includes all of its capabilities
including the CIC option for fuzzy copies. It also has an extension to the
capabilities of the IMS Database Recovery utility (DFSURDB0).
The functions of IMS Image Copy Extensions include:

򐂰 Image copies can be generated in compressed format, therefore reducing the
number of tapes or DASD space that is needed to create the image copies.
You can decompress the image copy records with the database recovery
function found in the IMS Image Copy Extensions product.
򐂰 A database pointer can be validated with a hash pointer check. This is done
while the image copies are taken. This reduces the elapsed time needed to
execute the image copies and pointer checker.
򐂰 Multiple image copies may be written to one tape volume without rewinding
the tape and remounting it for each image copy. This reduces the number of
mounts issued and the amount of operator intervention required for image
copies. This stacking support may be used with or without dynamic allocation.
򐂰 ICE eliminates steps by automatically allocating input and output data sets.
Input and output data sets do not require DD statements. They will be
dynamically allocated.
The tests uses the CUSTOMER database (PHIDAM/OSAM) that was converted
from HIDAM to PHIDAM in 9.2, “Migration of a database with a secondary index”
on page 142.
Example 16-4 shows the JCL of Image copy extensions for HALDB. This JCL will
make an image copy of all HLADB partitions.
Example 16-4 JCL of ICE for HALDB

//ICE EXEC PGM=FABJMAIN,REGION=0M
//STEPLIB DD DISP=SHR,DSN=HPS.V1R1M0.SHPSLMD0
//HISTORY DD DISP=SHR,DSN=IMSPSA.HPPC.HISTORY
//DFSPRINT DD SYSOUT=*
//ICEPRINT DD SYSOUT=*
//SYSUDUMP DD DUMMY
//MSGOUT DD SYSOUT=*
//REPORTS DD SYSOUT=*
//ICEIN DD *
GLOBAL DBRC=Y,COMP=Y,ICDALLOC=Y,ICNMRULE=Y,UNIT=SYSDA,
ICHLQ=IMSPSA,SPACE=(CYL,10,10)
IC DBD=CUSTDB,ICOUT=*
/*
If the parameter PART is omitted in the ICEIN DD statement, as in Example 16-4,

all partitions in the database are processed, one after the other.
Example 16-5 shows the ICEIN DD statement with PART parameters. To take
advantage of one of the benefits of HALDB, the partitions can be image copied
concurrently by running an image copy for each partition explicitly in different
jobs, thus reducing the total elapsed time.
Example 16-5 ICEIN DD statement with PART parameter

//ICEIN DD *
GLOBAL DBRC=Y,COMP=Y,ICDALLOC=Y,ICNMRULE=Y,UNIT=(SYSDA,3),
ICHLQ=IMSPSA,SPACE=(CYL,10,10)
IC DBD=CUSTDB,PART=CUSTDB1,ICOUT=*

/*
For more information on Image Copy Extensions, refer to the IMS Image Copy
Extensions for OS/390, V1R1, User’s Guide, SC27-0924.
Part 5
Part 5 Appendixes
This part contains the following appendixes:
򐂰 Appendix A, “Sample GENJCL.USER skeletons for HALDB allocation” on
page 281
򐂰 Appendix B, “HALDB functions added by APARs” on page 291
򐂰 Appendix C, “Summary of HALDB utilities” on page 293

A
Appendix A. Sample GENJCL.USER

skeletons for HALDB
allocation
This appendix explains how you may easily allocate IMS HALDB database data
sets with the use of DBRC GENJCL.USER commands. HALDB databases often
contain many data sets. The creation of jobs to allocate these data sets could be
time consuming for two reasons. First, there may be many data sets. Second, the
allocation jobs must use the data set names assigned by IMS in the partition
definition process. These challenges may be easily overcome by automating the
creation of allocation jobs. This appendix explains this automation and provides
examples of jobs that use it. These examples may be adapted for use in your
installation.
This appendix is a slightly modified version of the technical document that can be
found on Web at the following address:
http://www.ibm.com/support/techdocs/atsmastr.nsf/WebIndex/TD100491

Overview
This appendix explains how you may use DBRC GENJCL.USER commands to
create data set allocation jobs easily. There are four steps in the process.
1. Creation of skeletal JCL
Create skeletal JCL members for the various types of database data sets.
These types are indexes, indirect list data sets (ILDSs), ESDS data sets, and
OSAM data sets. The skeletal JCL members include the IDCAMS DEFINE
statements for the VSAM data sets or ALLOCATE statements for OSAM data
sets.
2. Definition of DBDS groups
Define DBDS groups for each HALDB database when you define the
database. There is one DBDS group for each of its data set types. For
example, a PHIDAM database would have a group for its indexes, a group for
its ILDSs, and a group for each of its data set groups.
3. Execution of GENJCL.USER commands
Use GENJCL.USER commands to generate the allocation jobs. These
commands specify the skeletal JCL members and DBDS groups defined in
the previous steps. The commands also may specify user variables, such as
primary and secondary allocation amounts.
4. Execution of allocation jobs
Execute the jobs generated by the previous step to allocate the data sets. If
necessary, you may modify the jobs to change the allocations for specific
partitions. A more complete explanation of these steps follows. The examples
shown are meant to be illustrations of how you may implement these
techniques. You should modify them for your environment.
Creation of skeletal JCL

A typical installation will want at least four skeletal JCL members. One member is
used for allocating PHIDAM prime index and secondary index data sets. These
are KSDSs. One member is used for allocating Indirect List Data Sets. These are
KSDSs. One member is used for allocating VSAM data data sets. These are
ESDSs. Data data sets are the data sets that contain segments in PHIDAM and
PHDAM databases. One member is used for allocating OSAM data data sets.
The following shows how these skeletal JCL members could be defined. You
should modify them to meet your installation's needs. For example, if
SMS-managed storage is used, the values for STORCLAS and MGMTCLAS
would have to meet your installation's standards. If you do not use
SMS-managed storage, you would replace STORCLAS and MGMTCLAS with
the appropriate VOLUMES parameter for your installation.
Index skeletal JCL member

You may use the following member to create PHIDAM index and secondary index
data sets. In the examples in this publication, its member name is DBDDDEFX. It
includes the following user variables:
%UPRIM Primary space allocation amount in cylinders
%USEC Secondary space allocation amount in cylinders
%UKSIZE Size of key
%UKOFFST Offset of key
%UCISZ Data component CI size
%URECSZ Record size
%UFREECI CI free space percentage
%UFREECA CA free space percentage
Example A-1 shows the sample member DBDDDEFX.
Example: A-1 Sample DBDDDEFX member

%DELETE (%STPNO NE '00000')
//S%STPNO EXEC PGM=IDCAMS
//SYSIN DD *
%ENDDEL
%SELECT DBDS((%DBNAME,%DDNAME))
DELETE %DBDSN CLUSTER
DEFINE CLUSTER (NAME(%DBDSN) -
STORCLAS(STANDARD) -
MGMTCLAS(STANDARD) -
CYLINDERS (%UPRIM %USEC) -
REUSE -
SHR(3 3) -
KEYS(%UKSIZE %UKOFFST) -
SPEED -
BWO(TYPEIMS) -
RECORDSIZE(%URECSZ %URECSZ) -
FREESPACE(%UFREECI %UFREECA) ) -
INDEX(NAME(%DBDSN.INDEX)) -
DATA(NAME(%DBDSN.DATA) -
CISZ(%UCISZ) )
%ENDSEL
Appendix A. Sample GENJCL.USER skeletons for HALDB allocation 283

ILDS skeletal JCL member
You may use the following member to create ILDS data sets. In the examples in
this section, its member name is DBDDDEFL. It includes the following user
variables:
%UCISZ Data component CI size
Example A-2 shows the sample member DBDDDEFL.
Example: A-2 Sample DBDDDEFL member

//SYSIN DD *
%ENDDEL
CYLINDERS (%UPRIM %USEC) -
RECORDSIZE(50 50) -
FREESPACE(25 10) -
SHR(3 3) -
SPEED -
KEYS(9 0)) -
INDEX(NAME(%DBDSN.INDEX)) -
DATA(NAME(%DBDSN.DATA) -
CISZ(%UCISZ) )
%ENDSEL
VSAM data skeletal JCL member

You may use the following member to create VSAM data data sets. In the
examples in this section, its member name is DBDDDEFV. It includes the
following user variables:
%UCISZ CI size
%URECSZ Record size
Example A-3 on page 285 shows the sample member DBDDDEFV.
Example: A-3 Sample DBDDDEFV member
//SYSIN DD *
%ENDDEL
CYLINDERS(%UPRIM %USEC) -
REUSE -
SHR(3 3) -
CISZ(%UCISZ) -
RECORDSIZE(%URECSZ %URECSZ) -
NONINDEXED SPEED ) -
DATA(NAME(%DBDSN.DATA))
%ENDSEL
OSAM data skeletal JCL member

You may use the following member to create OSAM data data sets. In the
examples in this section, its member name is DBDDDEFO. It includes the
following user variables:
Example A-3 shows the sample member DBDDDEFO.
Example: A-4 Sample DBDDDEFO member

//S%STPNO EXEC PGM=IDCAMS,DYNAMNBR=100
//SYSIN DD *
%ENDDEL
ALLOCATE -
DSNAME('%DBDSN') -
FILE(%DDNAME) -
DSORG(PS) -
NEW CATALOG -
SPACE (%UPRIM %USEC) CYLINDERS

%ENDSEL
Defaults member
You may use the following member to contain default values for user variables. In
the examples in this section, its member name is DBDDFLTD. The
GENJCL.USER commands specify this member in their DEFAULTS parameter.
Example A-3 on page 285 shows the sample member DBDDFLTD.
Example: A-5 Sample DBDDFLTD

%UCISZ = '4096'
%UFREECI = '40'
%UFREECA = '10'
Definition of DBDS groups

The second step in the process is the definition of DBDS groups for each HALDB
database. Typically, you would create the groups for a database when you define
the database. These groups are used with GENJCL.USER commands to create
the allocation jobs. There is one DBDS group for each of its data set types. A
PHIDAM database would have a DBDS group for its indexes, a group for its
ILDSs, and a group for each of its data set groups. That is, you would create a
DBDS group for the A data sets, another for the B data sets, etc. A PHDAM
database would have a DBDS group for its ILDSs, and a group for each of its
data set groups. A PSINDEX database would have a DBDS group for its index
data sets.
Example A-6 on page 287 shows an example of a PHIDAM database with two
data set groups and a secondary index. The database name is RZLDBT. Its
partition names are DBTP01, DBTP02, DBTP03, DBTP04, and DBTP05. The
secondary index is RZLSIT1. Its partition names are SIT1P01, SIT1P02,
SIT1P03, and SIT1P04. The HALDB naming convention creates DD names from
these partition names. For example, the DD name of the ILDS data set for the
DBTP01 partition is DBTP01L. The DD name for the PHIDAM index of this
partition is DBTP01X. The DD names of the data data sets are DBTP01A and
DBTP01B. The following DBRC commands create the four DBDS groups for this
database and the DBDS group for the secondary index. The groups have been
named to indicate the database and the type of data sets in the group. For
example, DBDS group RZLDBTA contains the A data sets for the partitions in
database RZLDBT.
Example: A-6 JCL for defining the DBDS groups
//DBDSGRP JOB (999,XXX),
// 'INIT DBDSGRPS',
// CLASS=A,MSGCLASS=T,MSGLEVEL=(1,1),
// NOTIFY=LEWIS,
// REGION=4M
//D EXEC PGM=DSPURX00
//STEPLIB DD DISP=SHR,DSN=IMS.SDFSRESL
//IMS DD DISP=SHR,DSN=IMS.DBDLIB
//********************************************
//*
//SYSIN DD *
INIT.DBDSGRP GRPNAME(RZLDBTA) MEMBERS( -
(DBTP01 ,DBTP01A), -
(DBTP05 ,DBTP05A))
INIT.DBDSGRP GRPNAME(RZLDBTB) MEMBERS( -
(DBTP01 ,DBTP01B), -
(DBTP05 ,DBTP05B))
INIT.DBDSGRP GRPNAME(RZLDBTL) MEMBERS( -
(DBTP01 ,DBTP01L), -
(DBTP05 ,DBTP05L))
INIT.DBDSGRP GRPNAME(RZLDBTX) MEMBERS( -
(DBTP01 ,DBTP01X), -
(DBTP05 ,DBTP05X))
INIT.DBDSGRP GRPNAME(RZLSIT1A) MEMBERS( -
(SIT1P01 ,SIT1P01A), -
(SIT1P04 ,SIT1P04A))

Execution of GENJCL.USER commands
The third step in the process is the use of GENJCL.USER commands to
generate the allocation jobs. User variables, such as primary and secondary
allocation amounts, must be specified on the GENJCL.USER commands.
Each GENJCL.USER command opens, writes, and closes the output data set.
For this reason, each command uses a different member of data set
IMS.JCLOUT. If multiple commands wrote to the same member, later commands
would overwrite the output of earlier commands.
Example A-7 shows GENJCL.USER commands that use the DBDS groups and
skeletal JCL that were created in the first two steps. OSAM is used for the data
data sets. Skeletal JCL member DBDDDEFO is used for these data sets.
Skeletal JCL member DBDDDEFX is used for the PHIDAM index data sets and
the secondary index data sets. The keys of the PHIDAM index are 5 bytes long
at offset 5. The keys of the secondary index are 5 bytes long at offset 34. These
values are specified for the %UKSIZE and %UKOFFST user variables. The
PHIDAM index record size is 10 bytes. The secondary index record size is 40
bytes. These values are specified for the %URECSZ user variable. Similarly,
space and free space variables are also specified. Skeletal member DBDDDEFL
is used for the ILDSs.
Example: A-7 Example GENJCL.USER commands

//GENJCLT JOB (999,XXX),
// 'GENJCL.USER',
// CLASS=A,MSGCLASS=T,MSGLEVEL=(1,1),
// NOTIFY=LEWIS,
// REGION=4M
//D EXEC PGM=DSPURX00
//STEPLIB DD DISP=SHR,DSN=IMS.SDFSRESL
//IMS DD DISP=SHR,DSN=IMS.DBDLIB
//JCLPDS DD DISP=SHR,DSN=IMS.JCLPDS
//JCLOUTA DD DISP=SHR,DSN=IMS.JCLOUT(ALLOCTA)
//JCLOUTB DD DISP=SHR,DSN=IMS.JCLOUT(ALLOCTB)
//JCLOUTX DD DISP=SHR,DSN=IMS.JCLOUT(ALLOCTX)
//JCLOUTL DD DISP=SHR,DSN=IMS.JCLOUT(ALLOCTL)
//JCLOUTS DD DISP=SHR,DSN=IMS.JCLOUT(ALLOCTS)
//********************************************
//*
//SYSIN DD *
GENJCL.USER GROUP(RZLDBTA) MEMBER(DBDDDEFO) JCLOUT(JCLOUTA) LIST -
USERKEYS((%UPRIM,'2'),(%USEC,'1')) -
DEFAULTS(DBDDFLTD) ONEJOB
GENJCL.USER GROUP(RZLDBTB) MEMBER(DBDDDEFO) JCLOUT(JCLOUTB) LIST -
USERKEYS((%UPRIM,'2'),(%USEC,'1')) -
GENJCL.USER GROUP(RZLDBTX) MEMBER(DBDDDEFX) JCLOUT(JCLOUTX) LIST -
USERKEYS((%UPRIM,'2'),(%USEC,'1'),(%UKSIZE,'5'), -
(%UKOFFST,'5'),(%UCISZ,'4096'),(%URECSZ,'10')) -
GENJCL.USER GROUP(RZLDBTL) MEMBER(DBDDDEFL) JCLOUT(JCLOUTL) LIST -
USERKEYS((%UPRIM,'1'),(%USEC,'1'))
GENJCL.USER GROUP(RZLSIT1A) MEMBER(DBDDDEFX) JCLOUT(JCLOUTS) LIST -
USERKEYS((%UPRIM,'1'),(%USEC,'1'),(%UKSIZE,'5'), -
(%UKOFFST,'34'),(%UCISZ,'4096'),(%URECSZ,'40')) -
Execution of allocation jobs

The final step in the process is the execution of the allocation jobs generated in
the third step. In the example, the JCL for the jobs is written to members
ALLOCTA, ALLOCTB, ALLOCTX, ALLOCTL, and ALLOCTS in IMS.JCLOUT.
The process produces the same allocation parameters for the data sets in every
partition. For example, the A data sets for every partition will have the same
space specifications. If one or more of your partitions require different
parameters, you would have to modify the allocation specifications for such
partitions before you execute the jobs.

B
Appendix B. HALDB functions added by

APARs
Some capabilities have been added to HALDB by APARs. Other capabilities will
be delivered by APARs. This appendix provides a list of these APARs and the
functions they deliver.

HALDB functions delivered by APARs
Table B-1 lists APARs that deliver HALDB functions not in the original delivery of
IMS Version 7. Those that were included in the original delivery of IMS Version 8
are indicated by Included in base.
Table B-1 Functions added by APAR

IMS V7 APAR IMS V8 APAR Function
PQ35893 Included in base INIT.DB and INIT.PART DBRC commands.
PQ36991 PQ54227 Support for ILDSMULTI and NOILDS control

statements in HD Reload.
PQ38626 Included in base HALDB support in DBCTL IVP.
PQ48421 PQ73858 Support for DISABLE/ENABLE of partitions so

that DBRC recovery information will not be lost
when a partition is deleted, then redefined. A new
batch export capability to allow the defining of
HALDBs at different sites to be exactly the same.
For example, partition IDs can be preserved
across export/import.
PQ49638 PQ55002 Unconditional partition initialization by the

HALDB Partition Initialization utility.
PQ52858 PQ59888 CHANGE.DB, CHANGE.PART, DELETE.DB,

and DELETE.PART DBRC commands.
PQ55840 PQ56463 BLDSNDX parameter on the OPTIONS

statement to eliminate secondary index updates
by initial loads.
PQ57313 PQ58600 HALDB control statement to provide single

partition processing by batch jobs.
PQ61206 PQ72776 Eliminate allocation and open of ILDS when

database has no logical relationships or
secondary indexes.
PQ65486 PQ65489 Allow use of PCB name or PCB label in HALDB

control statement for single partition processing.
C
Appendix C. Summary of HALDB utilities

This appendix lists all of the database utilities that can be run against HALDBs
and refers to the chapters where we discuss more about them. Additional
information about the utilities can also be found in the manual IMS Version 8
Utilities Reference: Database and Transaction Manager, SC27-1308.

IMS database utilities for HALDBs
Table C-1 lists all of the databases utilities that can be run against HALDBs.
Table C-1 Utilities that can be run against HALDBs

Utility Description Comment
DFSMAID0 HALDB Migration Aid Refer to 8.2, “HALDB Migration Aid

utility (DFSMAID0)” on page 93.
DFSPREC0 HALDB Index/ILDS Rebuild Refer to “Generating Index/ILDS

Rebuild utility JCL” on page 41.
DFSUPNT0 HALDB Partition Data Set Refer to 6.4, “Database Partition Data
Initialization Set Initialization utility” on page 76.
%DFSHALDB HALDB Partition Definition Invocation of the utility by a CLIST.

%DFSHALDB is the TSO invocation
of module DSPXPDDU. Refer to
Chapter 4, “Partition Definition utility”
on page 45.
DFSURGU0 HD Reorganization Unload Applies to PHDAM, PHIDAM, and

PSINDEX. Refer to 8.3.1, “Unloading
the existing database” on page 96,
9.1.3, “Unloading the HIDAM
database” on page 136, and 14.1.4,
“Unloading HALDB partitions and
databases” on page 246.
DFSURGL0 HD Reorganization Reload Applies to PHDAM, PHIDAM, and

PSINDEX. Refer to 8.3.8, “Reloading
the HALDB database” on page 99,
9.1.10, “Loading HALDB database”
on page 141, and 14.1.6, “Reloading
HALDB partitions and databases” on
page 249.
DFSURPR0 Prereorganization Refer to 6.3, “Database

Prereorganization utility” on page 76,
and 9.1.9, “Initializing HALDB
partitions” on page 140.
Utility Description Comment
DFSUDMP0 Image Copy Refer to Chapter 13, “Backup and

recovery” on page 227.
DFSUICP0 Online Image Copy
DFSUDMT0 Database Image Copy 2
DFSUCUM0 Change Accumulation
DFSURDB0 Database Recovery
DFSBBO00 Batch Backout
Database Recovery Control (DBRC) is required for execution of any utility

operating on a HALDB. Each utility checks for the presence of DBRC. If DBRC is
not present, the utility issues an error message and terminates.
Image copy utilities reject any attempt to image copy HALDB ILDSs or PHIDAM
prime index data sets. Recovery utilities reject any attempt to recover HALDB
ILDSs or PHIDAM prime index data sets. Both Image Copy and Recovery utilities
can only run against a particular data set of a HALDB partition.
Appendix C. Summary of HALDB utilities 295

Abbreviations and acronyms
ACB Application Control Block HLQ High-level Qualifier
APAR Authorized Program Analysis HSAM Hierarchic Sequential Access
Report Method
APF Authorized Program Facility HTML Hyper Text Markup Language
BMP Batch Message Program HTTP Hyper Text Transfer Protocol
CDS Control Data Set IBM International Business
CI Control Interval Machines Corporation
CICS Customer Information Control IFP Ims Fast Path Program

System ILDS Indirect List Entry Data Set
DASD Direct Access Storage Device ILE Indirect List Entry
DB2 Database 2 ILK Indirect List Key
DBA Database Administrator IMS Information Management
DBCTL Database Control System
DBD Database Description IRLM Internal Resource Lock

Manager
DBDS Database Data Set
ISPF Interactive Systems
DBRC Data Base Recovery Control Productivity Facility
DEDB Data Entry Database ITSO International Technical
DL/I Data Language/i Support Organization
DLI/SAS Dl/i Separate Address Space IVP Installation Verification
Program
EBCDIC Extended Binary Coded
Decimal Interchange Code JBP Java Batch Processing
Region
EPS Extended Pointer Set
JCL Job Control Language
EQE Error Queue Element
JES Job Entry Subsystem (Jes2
ESDS Entry Sequenced Data Set
Or Jes3)
FSEAP Free Space Element Anchor
JMP Java Message Processing
Point
Region
HALDB High Availability Large
KR Key Range
Database
KSDS Key Sequenced Data Set
HDAM Hierarchical Direct Access
Method LPCK Logical Parent Concatenated
Key
HIDAM Hierarchical Indexed Direct
Access Method MPP Message Processing
Program
HISAM Hierarchical Indexed
Sequential Access Method ODBA Open Database Access

ORS Online Recovery Service
OSAM Overflow Sequential Access
Method
PCB Program Communication
Block
PDS Partitioned Data Set
PDU Partition Definition Utility
PHDAM Partitioned Hierarchical Direct
Access Method
PHIDAM Partitioned Hierarchical
Indexed Direct Access
Method
PID Partition Identification
PROCOPT Processing Option
PSB Program Specification Block
PSINDEX Partitioned Secondary Index
PTF Program Temporary Fix
RAP Root Anchor Point
RBA Relative Byte Address
RSR Remote Site Recovery
SMP/E System Modification
Program/extended
SVL Silicon Valley Laboratories
TB Twin Backward (Pointer)
TSO Time Sharing Option
VOLSER Volume Serial (Number)
VSAM Virtual Storage Access
Method
WWW World Wide Web
XRF Extended Recovery Facility
Related publications
The publications listed in this section are considered particularly suitable for a
more detailed discussion of the topics covered in this publication.
IBM Redbooks
For information on ordering these publications, see “How to get IBM Redbooks”
on page 300. Note that some of the documents referenced here may be available
in softcopy only.
򐂰 A DBA’s View of IMS Online Recovery Service, SG24-6112
򐂰 Ensuring IMS Database Integrity Using IMS Tools, SG24-6533
򐂰 IMS DataPropagator Implementation Guide, SG24-6838
򐂰 IMS e-business Connectors: A Guide to IMS Connectivity, SG24-6514
򐂰 IMS Installation and Maintenance Processes, SG24-6574
򐂰 IMS in the Parallel Sysplex, Volume I: Reviewing the IMSplex Technology,
SG24-6908
򐂰 IMS Primer, SG24-5352
򐂰 IMS Version 7 High Availability Large Database Guide, SG24-5751
򐂰 IMS Version 8 Implementation Guide - A Technical Overview of the New
Features, SG24-6594
򐂰 IMS Version 7 Java Update, SG24-6536
򐂰 IMS Version 7 Performance Monitoring and Tuning Update, SG24-6404
򐂰 IMS Version 7 Release Guide, SG24-5753
򐂰 Using IMS Data Management Tools for Fast Path Databases, SG24-6866
Other publications
These publications are also relevant as further information sources:
򐂰 IMS Image Copy Extensions for OS/390, V1R1, User’s Guide, SC27-0924
򐂰 IMS Index Builder for Z/OS, V2.R User’s Guide, SC27-0930
򐂰 IMS High Performance Load for OS/390, V1R1, User’s Guide, SC27-0938

򐂰 IMS High Performance Unload for OS/390, V1R1, User’s Guide, SC27-0936
򐂰 IMS Version 7 Customization Guide, SC26-9427
򐂰 IMS Version 8 Administration Guide: Database Manager, SC27-1283
򐂰 IMS Version 8 Command Reference, SC27-1291
򐂰 IMS Version 8 Customization Guide, SC27-1294
򐂰 IMS Version 8 Utilities Reference: Database and Transaction Manager,
SC27-1308
Online resources
These Web sites and URLs are also relevant as further information sources:
򐂰 IMS home page
http://www.ibm.com/ims
򐂰 IBM Redbooks home page
http://www.ibm.com/redbooks
򐂰 IBM Data Management Tools home page
http://www.ibm.com/software/data/db2imstools/
򐂰 Technical documents: Flashes, presentations, hints, tips, and Technotes
http://www.ibm.com/support/techdocs/
How to get IBM Redbooks

You can search for, view, or download Redbooks, Redpapers, Hints and Tips,
draft publications and Additional materials, as well as order hardcopy Redbooks
or CD-ROMs, at this Web site:
ibm.com/redbooks
Index
Symbols C
Change Accumulation 230–231
/DBD 214, 221
change accumulation group 42
/DBDUMP 218
change version numbe 33
/DBR 214, 221
CHANGE.DB 35–36, 68, 75, 78, 266, 292
/DIS 214
CHANGE.DBDS 36–37, 64, 68, 70
/LOCK 214, 220
CHANGE.PART 36–37, 68–69, 292
/START 214, 219, 221
CI size 260
/START DB ALL 221
CICS 198
/STOP 214, 220
command 213
/SX 92–93
CONST 92
/UNLOCK 214, 221
Numerics D
data capture exit 258
5655-B01 (IMS Version 7) ii
Data Capture Exit routine 24
5655-C56 (IMS Version 8) ii
data conversion user exit 258
5655-E50 (Online Recovery Service Version 1) 231
Data Conversion User Exit routine 24
5655-I01 (HALDB Conversion Aid Version 1) 152
data set name prefix 18, 33, 259
5655-K47 (HALDB Conversion and Maintenance
database
Aid Version 2) 151
PHIDAM 7
Database Change Accumulation 230
A database data set group 42
ACBGEN 130, 142, 147 Database Data Set Initialization 99
ACCESS 91, 138, 219 database group 42
access intent 219 Database Image Copy 228–229
ACCESS operand 26 Database Image Copy 2 228–229
ALL keyword 221 DATABASE macro 25
ALLOC 40, 267 Database Partition Data Set Initialization 76
ALLOCATE 139 Database Partition Data Set Initialization utility 74
allocation 35 Database Prefix Resolution 76
APAR 78, 100–101, 250, 291–292 Database Prereorganization 99
authorization 35 Database Prereorganization utility 74–76
Database Recovery 228, 231, 233
Database Scan 76
B
backout needed 37 Database structure 11
Batch Backout 241 DATASET 91, 138
Bidirectional logical relationships 10 DATASET statement 26
BLDSNDX 292 DB record 32
block size 260 DBCTL 292
BMP 10, 178, 221–222 DBCTL thread 222
Buffer subpools 25 DBD 91, 144
BYTES 92–93 DBDGEN 7–8, 12, 21–22, 24, 45, 49, 62, 90, 130,
144

DBDS record 32 error queue element 37
DBORG 40 extended pointer set 14
DBRC 9, 31 Extended Recovery Facility (XRF) 3
DBRC authorization for HALDB 35
DBRC CHANGE commands 260
DBRC command 35, 261, 263, 292
F
FIELD 93
DBRC command authorization 47
free space 91, 260
DBRC commands 35, 45
free space element anchor point 11, 94
DBRC group 42
FSEAP 11, 94
DDNAME 91
DEFINE 98–99
DELETE 38 G
DELETE.DB 38, 71, 292 GENJCL 40, 227
DELETE.DBDS 38 GENJCL.CA 231
DELETE.PART 38, 71, 292 GENJCL.IC 230
DFS0565I 222 GENJCL.OIC 230
DFSBBO00 241 GENJCL.RECOV 233, 235
DFSDBUX1 258 GENJCL.USER 41
DFSMAID0 93, 130, 139, 260, 268 GENMAX 37
DFSMDA 9, 35 Global DMB number 32
DFSMSdss 229 group 42
DFSPREC0 41, 142
DFSPSE00 84–85
H
DFSSRT01 142 H 91
DFSSRTnn 143 HALDB Conversion and Maintenance Aid 87
DFSUCUM0 231 HALDB Database Data Set Initialization 99
DFSUDMP0 228–229 HALDB master 9
DFSUDMT0 228–229 HALDB Migration Aid utility 260, 268
DFSUICP0 228 HALDB Migration Aid utility (DFSMAID0) 23
DFSURDB0 228, 231 HALDB Partition Initialization utility 292
DFSURGL0 99, 141 HB 91
DFSURGU0 136 HD Reload 90, 249, 265–267
DFSURGUP macro 259 HD Reorganization Reload 99, 141
DFSVSAMP 237 HD Reorganization Unload 136, 147
DFSWRK01 142 HD Unload 90, 249, 259, 265
DFSWRKnn 143 HDAM 6
DSGROUP 91–92, 138 HIDAM 6–7, 92, 130
DSGROUP parameter 26 HIER 91
DSPDBHRC macro 32–33 HIERBWD 91
DSPDSHRC macro 34 high-used-RBA 74
DSPPTNRC macro 34 HIKEY 63
DSPUPJCL 41
DUMP 229
dynamic allocation 26, 31, 35 I
IBM High Performance Load 74
IDCAMS 139
E IEC-161I 248
EPS 14 IEC161I 100, 248
EQE 37 IFP 193, 222
IHCXHAL 155 LIST.HISTORY 38
ILDS 11, 14–15, 34, 37, 98, 142, 259, 269, 292 logical relationship 6, 10
ILDSMULTI 100–101, 250, 292 logical relationships 5, 22
ILDSMULTI control statement 250
ILE 14
ILK 12, 14
M
master database 22, 32, 36, 43, 49, 62
image copy 130, 147, 228, 267
master database record 32
Image Copy 2 228–229
MAX 133
Image Copy Extensions 228
MAX control statement 95
image copy needed 34, 37, 75–76, 100, 142, 174,
MIGRATE=YES 90, 96, 101, 136, 147
187, 197, 261, 263–264, 266–267
Migration Aid 130, 139, 147
IMS HALDB Conversion and Maintenance Aid
Migration Aid utility 260, 268
151–152
MIGRATX=YES 101, 142
IMS Image Copy Extensions 228
MPP 178, 193, 222
IMS Index Builder 259, 265, 269
multi-volume OSAM 140
IMS monitor 3
IMS Online Recovery Service 231
IMS Performance Analyzer 3 N
IMS/ESA Partition Support Product 147 naming conventions 17
IMS/ESA Partition Support Product (PDB) 6 NBR 131
Index/ILDS Rebuild utility 41 NBR control statement 95
indirect list data set 11, 18, 282 NOILDS 101, 250, 292
indirect list entry 14 NOILDS control statement 250
indirect list key 12, 14 NOTWIN 92
INIT 36
INIT.DB 24, 32, 36, 62, 82, 292
O
INIT.DBDS 36 ODBA 193, 198
INIT.PART 24, 33–34, 36, 62, 64, 69, 292 ODBA thread 222
IOBF statement 25 Online Database Image Copy 228
ISPF 46 Online Recovery Service 231
IVP 292 OPEN 219
ORS 231
J OSAM 5, 22
JBP 10, 221–222 OSAM block size 91, 260
JMP 193, 222 OSAM sequential buffering 237
K P
key range 80 parallel processing 7
key ranges 10 PART record 32
KEYSTRNG 63, 65 partition database data set record 32
KR control statement 95, 107, 134 partition database record 32–33
partition DBDS record 34
partition definition process 7
L Partition Definition utility 8, 22, 24, 26, 33–35,
LCHILD 92, 138
45–46, 62, 82, 97, 120, 138–139, 208, 260–261,
LIST 38
263
LIST.DB 38
Partition Definition utility (PDU) 22
LIST.DBDS 38
partition high key 34
Index 303
partition ID 11, 14–15, 34, 267 Prefix Resolution 188, 244
partition init needed 33, 37, 67, 75, 248, 260, 274 Prefix Resolution utility 244
Partition init needed flag 34 Prefix Update 188
partition initialization 35, 73, 90 Prereorganization utility 248
Partition Initialization utility 75 primary index 26, 34, 142
partition name 18, 43, 267 prohibit further authorization 36
partition record 32–33 PSINDEX 3, 11, 15, 18, 80, 91–92
partition selection 10, 16, 79 PSNAME 91
partition selection exit 17, 36 PSNAME parameter 26
Partition selection exit routine 32 PTR 91
partition selection exit routine 11, 75, 79, 81 PTR=INDX 92
partition string 34, 37 PTR=SNGL 92
partition structure initialization 83 PTR=SYMB 92
partition structure modification 83
partition structure termination 83
partitioned HDAM 3, 11
R
randomization parameter 260
partitioned HIDAM 3
randomizing module 260
partitioned secondary index 3
RAP 94, 120, 145, 184
PARTSEL 63
RBA 12–15, 29, 102, 126, 187, 252–254, 256
PCB label 292
read only 37
PCB name 292
RECON 31, 46, 90
PDATA 42
recovery group 42
PDB 147
recovery needed 34, 37
PDU 46, 97, 139
RECOVPD 37
PHDAM 3, 11, 14–15, 18, 74, 80, 91
Redbooks Web site 300
PHIDAM 3, 7, 11, 14–15, 18, 74, 80, 91–92, 130
Contact us xv
PHIDAM primary index 37
Remote Site Recovery (RSR) 3
physical pairing 10, 15
REORG 40
PILDS 42
reorganization 5–7, 17
PINDEX 42
reorganization number 11, 14–15
PINIT 34, 37, 260, 263, 266
REUSE 37, 98–99
PQ35893 62, 292
RKSIZE 92
PQ36991 100–101, 250, 292
RMNAME 91
PQ38626 292
RMNAME parameter 26
PQ48421 292
root anchor point 94
PQ49638 78, 292
PQ52858 62, 292
PQ54227 100–101, 250, 292 S
PQ55002 78, 292 SAMPLE control statement 95
PQ55840 292 sample partition selection exit 84
PQ56463 292 SCAN 91
PQ57313 292 SDFSISRC 238
PQ58600 292 secondary index 5–6, 13, 16–17, 22, 87, 90, 251
PQ59888 62, 292 Secondary Index Database Maintenance Exit rou-
PQ61206 292 tine 24
PQ63751 130 secondary indexes 6
PQ65486 292 security 47
PQ65489 292 SEED 96
PQ68573 130 SEGM 91, 138
Segment Edit/Compression Exit routine 24
selection exit 10–11
self healing 245, 251
self healing pointer 269
share level 36
shared secondary index 92
single partition processing 292
subsequence 92–93
symbolic keyword 42
symbolic pointers 92
symbolic pointing 17
system definition 25
T
T 91
TB 91
Timestamp recovery 232
timestamp recovery 16
TSO 46
TSO %DFSHALDB 46
TWIN 91
TWINBWD 91
TYPE 215
TYPEFP 63
TYPEIMS 63
TYPHALDB 38, 63
U
Unconditional partition initialization 292
V
VSAM 5, 22
VSAM CI size 260
VSRBF statement 25
X
XDFLD 92
Index 305
The Complete IMS HALDB Guide All You Need to Know to Manage HALDBs
Back cover ®
The Complete IMS HALDB

Guide
Examine how to This IBM Redbook describes the High Availability

migrate databases to Large Database (HALDB) capability available with INTERNATIONAL
HALDB IMS. IMS HALDB was introduced with IMS Version TECHNICAL
7. It allows IMS databases to grow to almost SUPPORT
unlimited sizes while providing increased ORGANIZATION
Explore the benefits availability. This book updates IMS Version 7 High
of using HALDB Availability Large Database Guide, SG24-5751, as
well as adding topics that were not covered in the
Learn how to previous book. BUILDING TECHNICAL
administer and INFORMATION BASED ON
This publication provides a broad explanation of PRACTICAL EXPERIENCE
modify HALDBs HALDB and its uses. Specific areas covered
include:
IBM Redbooks are developed by
򐂰 HALDB overview, definition, and structure the IBM International Technical
򐂰 Migration from non-HALDB databases Support Organization. Experts
򐂰 Application considerations from IBM, Customers and
򐂰 HALDB database administration Partners from around the world
create timely technical
information based on realistic
This publication documents our hands-on scenarios. Specific
experience in a test environment. It includes recommendations are provided
migration and administration examples. Some IBM to help you implement IT
Data Management Tools for IMS are also discussed solutions more effectively in
in this publication. Special emphasis is given to the your environment.
IMS HALDB Conversion and Maintenance Aid
product. Examples of the use of these tools are
provided.
For more information:
ibm.com/redbooks
SG24-6945-00 ISBN 0738453439

The Complete IMS HALDB Guide

Uploaded by

Copyright:

Available Formats

The Complete IMS HALDB Guide

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

The Complete IMS HALDB Guide

Uploaded by

Copyright:

Available Formats

Front cover

The Complete IMS

Explore the benefits of using

Learn how to administer

The Complete IMS HALDB Guide

First Edition (June 2003)

© Copyright International Business Machines Corporation 2003. All rights reserved.

Part 1. HALDB overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1

Chapter 1. HALDB introduction and structure . . . . . . . . . . . . . . . . . . . . . . . 3

Chapter 2. Defining HALDB databases . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21

© Copyright IBM Corp. 2003. All rights reserved. iii

Chapter 3. HALDB and DBRC . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31

Chapter 4. Partition Definition utility . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45

Chapter 5. Batch definition of HALDB. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61

Chapter 6. Partition initialization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73

Chapter 7. Partition selection. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79

Chapter 8. Migration from non-HALDB to HALDB . . . . . . . . . . . . . . . . . . . 89

Chapter 9. Migration examples . . . . . . . . . . . . . . . . . . . ....... ...... . 129

Part 3. Application considerations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 181

Chapter 11. Application considerations . . . . . . . . . . . . . . . . . . . . . . . . . . 183

Part 4. Administration. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 211

Chapter 12. HALDB online commands . . . . . . . . . . . . . . . . . . . . . . . . . . . 213

Chapter 14. HALDB reorganization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 243

Chapter 15. Changing existing HALDB databases. . . . . . . . . . . . . . . . . . 257

Part 5. Appendixes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 279

Appendix A. Sample GENJCL.USER skeletons for HALDB allocation . 281

Appendix B. HALDB functions added by APARs . . . . . . . . . . . . . . . . . . . 291

Appendix C. Summary of HALDB utilities . . . . . . . . . . . . . . . . . . . . . . . . 293

Abbreviations and acronyms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 297

Related publications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 299

© Copyright IBM Corp. 2003. All rights reserved. xi

CICS® Language Environment® Sequent®

The following terms are trademarks of other companies:

This publication documents our hands-on experience in a test environment. It

The team that wrote this redbook

Cornelia Hallmen is an IMS instructor in Munich, Germany. She has over 11

© Copyright IBM Corp. 2003. All rights reserved. xiii

Raymond Keung is a Systems Analyst/IMS Database Administrator in Toronto,

Rich Lewis is a Senior Consulting IT Specialist with IBM Advanced Technical

Thanks to the following people for their contributions to this project:

Rich Conway, Bob Haimowitz, Julie Czubik

Become a published author

We want our Redbooks™ to be as helpful as possible. Send us your comments

© Copyright IBM Corp. 2003. All rights reserved. 1

Chapter 1. HALDB introduction and

HALDB support includes the following:

This chapter provides an overview of HALDB and its structure.

© Copyright IBM Corp. 2003. All rights reserved. 3

HALDB also has availability and manageability benefits. Partitions may be

HALDB has several benefits:

Each of these limitations is specifically addressed by HALDB while maintaining

HALDB increases the storage capacity of full-function databases by providing the

HALDB improves the manageability of databases. Large databases can be split

Availability of data can also be improved with HALDB. Allocation, authorization,

Chapter 1. HALDB introduction and structure 5

1.3 Candidates for HALDB

Future IMS database enhancements

1.4 HALDB definition process

Chapter 1. HALDB introduction and structure 7

Figure 1-1 Changes to the database definition with HALDB