Skip to content

changes.md

Latest commit

 

History

History
165 lines (126 loc) · 5.27 KB

changes.md

File metadata and controls

165 lines (126 loc) · 5.27 KB

Major Changes

0.4.13 (15 Sep 2018)

  • Added HTTP Request logging middleware. #22 #23

0.4.12 (7 Sep 2018)

  • Added .NumAttempts() method as a faster way to do len(w.Attempts()) #21

0.4.11 (24 Jan 2018)

  • Added String method to WorkUnitStatus. (#19)

0.4.10 (2 Jan 2018)

  • Fix CircleCI build. (#18)

0.4.9 (2 Jan 2018)

  • Switch to CircleCI 2.0. (#17)
  • Switch to dep dependency management tool. (#17)

0.4.8 (21 Dec 2017)

  • Fix a data race. (#16)

0.4.7 (29 Aug 2017)

  • Switch docker image repository to Docker Hub. (#15)

0.4.6 (7 July 2017)

  • Bug fix for panic on nil logger. (#14)

0.4.5 (2 Jun 2017)

  • Performance fix for the PostgreSQL backend when max_retries is used. (#12)
  • If max_retries causes some work units to get preemptively failed, RequestAttempts will return fewer than the max_getwork number of attempts, rather than attempting to get more work, on all backends.

0.4.4 (30 May 2017)

  • In the published Prometheus metrics, don't double-double-quote the "status" value. (#11)

0.4.3 (30 May 2017)

  • Add a max_retries work spec parameter to limit work unit retries (#8)
  • Port the generic test suite to use github.com/stretchr/testify/suite, rather than the home-grown cptest program in this package. (#9)
  • Publish coordinate summary style metrics to Promotheus. (#10)

0.4.2 (4 May 2017)

  • Requires Go 1.7 or later, and its standard context package (#4)
  • Add CircleCI integration, Quay reference Docker image (#1, #2)
  • PostgreSQL backend performance improvement for long-running systems with many old workers (#3)
  • The Go worker framework can take a list of runtime strings as parameters, it is not limited to "go" (#5)

0.4.1 (15 Aug 2016)

  • Bug fix in Go decoding of CBOR-RPC responses.
  • Bug fix adding work units in the PostgreSQL backend, where adding a work unit that already existed would affect all work specs with performance implications

0.4.0 (29 Jun 2016)

  • Bug fix relating to multiple concurrent Go workers and concurrent map access
  • Bug fixes for continuous work units on the PostgreSQL backend, including fixes for intervals longer than 3600 seconds, intervals longer than 86400 seconds, and concurrent attempts to execute continuous work units
  • Don't ignore a requested work spec name from Python coordinate run_one --from-work-spec

The github.com/satori/go.uuid library made its UUID parsing stricter in this timeframe as well, so UUID values passed in work unit data or elsewhere may be rejected if key bits do not conform to RFC 4122.

0.3.0 (8 Apr 2016)

  • Adds the ability for work units to declare an earliest time they can run. Changes WorkSpec.AddWorkUnit() to take a new WorkUnitMeta structure instead of the existing priority parameter, and adds a time.Duration parameter to Attempt.Retry(). Generated output work units in chained work units can also declare a delay before running the next unit. Makes corresponding (incompatible) changes to the REST API.
  • Adds a worker framework so that Coordinate work specs can run task implementations written in Go.
  • Add a coordbench tool to measure the performance of the Coordinate system.
  • Several performance improvements for the PostgreSQL backend, aimed at the "millions of work units" scale.
  • Switch tests from gocheck to testify.

0.2.0 (4 Jan 2016)

  • Adds a REST server interface: start coordinated -http :5980.
  • Adds a REST client backend: start coordinated -backend http://localhost:5980/.
  • Adds a runtime key to work specs to allow workers for multiple language runtimes to coexist.
  • Turns off detailed logging of every CBOR-RPC request and response.

0.1.2 (1 Dec 2015)

  • Build changes to support getting a working static Docker container out.

0.1.1 (20 Nov 2015)

  • Work around a code change in the codec library.

0.1.0 (18 Nov 2015)

  • Initial release.