Skip to content
This repository has been archived by the owner on Sep 27, 2018. It is now read-only.

Latest commit

 

History

History
45 lines (39 loc) · 2.51 KB

CHANGELOG.md

File metadata and controls

45 lines (39 loc) · 2.51 KB
Raw

[1.2.1] - 2017-08-03

  • Added targets in Makefile to verify Docker images (see README.md)
  • Upgraded Nokogiri dependency to 1.8.0
  • Upgraded default Ruby interpreter to JRuby 9.1.12.0
  • Dropped support for MRI in Gemfile.lock.

[1.2.0] - 2017-06-21

Note: this version requires Apicast-XC >= v1.3.0.

  • The priority auth renewer now only publishes each response in the redis pubsub channel once, unless there is an error while publishing. We were publishing each response several times, but that was because of a race condition caused by Apicast-XC that has been fixed. This change improves performance a bit because it lowers the number of commands issued to Redis.
  • Introduced minor changes in the Dockerfile that decrease the image size.

[1.1.0] - 2017-05-02

  • Switched from MRI to JRuby. our tests show that JRuby performs better than MRI when there is a high number of calls to be made to 3scale's backend in a flushing cycle. One of the reasons is that JRuby allows us to make several calls to 3scale's backend in parallel as well as parse their XML responses.
  • Updated dependencies: 3scale_client (2.11.0), gli (2.16.0), redis(3.3.3), concurrent-ruby(1.0.5), simplecov (0.14.1), rubocop (0.48.1).
  • Changed the timezone in the Dockerfile to UTC. This way, timestamps can be easily translated from log files or other sources to whatever is needed.
  • Increased the number of messages that the priority auth renewer can process at the same time. We accomplished that by reducing the waiting times between publishing attempts. Those waiting times were unnecessarily long.
  • Fixed a bug that affected the way thread pools were being used. We were using concurrent-ruby's ThreadPoolExecutor without specifying a max size for the queue, and that type of pool only spawns a new thread when the queue is full. That means that in practice, we specified a 'min:max' number of threads, but only min were spawned. Now we are using FixedThreadPools and always take the max number of threads for the input, ignoring the min.
  • Added logging that shows the run time of each of the 5 phases of a flushing cycle: 1- getting the reports from Redis, 2- reporting to 3scale, 3- waiting to leave some time to the reports to take effect, 4- getting the auths from 3scale, 5- renewing the auths in Redis. This is useful for debugging purposes.
  • Minor changes in the Dockerfile to make it run on CentOS / RHEL.
  • Improved performance of the flushing phase that takes care of renewing authorizations in Redis by using the hmset command.