https://github.com/twitter/heron
Revision 5267285d84d2f1d74ae6e6fb15cf7b93cf542dfc authored by Maosong Fu on 16 November 2017, 01:49:29 UTC, committed by GitHub on 16 November 2017, 01:49:29 UTC
A lot of Twitter internal topology owners complained heron failing to kill an aurora topology because
the Aurora service can be unavailable or unstable for a while, or it can sometimes provide a false negative response.

This pull request adds retry mechanism to kill an Aurora topology with more confidence. It can increase the robustness of AuroraScheduler regardless of the Aurora Service People can also specify the retry attempts and interval via config.

This pull request also re-format some coding-styles for RuntimeManagerRunner.java.

Tested with actual Twitter distributed env.
1 parent 57e910f
History
Tip revision: 5267285d84d2f1d74ae6e6fb15cf7b93cf542dfc authored by Maosong Fu on 16 November 2017, 01:49:29 UTC
Retry when failing to kill an Aurora topology (#2554)
Tip revision: 5267285

README.md

back to top