A major release, brings incompatible changes, please read the CHANGELOG carefully and adapt your configs.

• Fix high load on a single thread instead of many
• Fix non-cut buffer passed from async udp listener when multimessage is off

This major release introduces more flexibility for choosing metric aggregates, a big internal async/await rewrite and some more features.

See CHANGELOG for details.

See CHANGELOG.md in project root for details

This is a bugfix release

See CHANGELOG.md for this release details

Cheers to everyone on Highload++ conference!

See changelog for more information

• moved to new tokio, excluding consul(hyper is not ready yet)
• retry strategy has beed added to carbon backend
• carbon backend now has it's own section in config.toml. To use new config:
  • move backend parameter to [carbon] section as address
  • move backend-interval parameter to [carbon] section as interval
  • also note that new parameters has been added for retrying send. See config.toml in this repo for more information
• ingress(internal meric) now counts packets instead of packet chunks
• egress(internal metric) counting is now more precise
• new version of bincode is used for peer protocol. Showed not to work with previous one, so restart will most probably be required

Release 0.3.2

This is mostly a bugfix release.

Changes:

• Fix to allow low volume of metrics to be sent
• Fix for snapshot distribution bug for number of nodes reater than 2
• Consul key name is now customizable
• ms metric's leaf value is now suffixed as ".last", which is less confusing, more similar to statsd/brubeck and less random

Some statistics about metrics added to this release.

One can be notified now if too much metrics come in aggregation period (thanks @errx for the patch).
See config.toml for customizing such parameters.
Also some minor performance optimizatons done(timestamp string replaced with Cow, which should reduce allocations during aggregation).