Details
-
Task
-
Resolution: Fixed
-
L3 - Default
-
None
-
None
Description
Situation:
Currently when we do a release on a maintenance branch, the Jenkins responsible for CI of the same branch is overloaded during/after the release because various jobs start to because they detected a SCM change because of newly committed version poms.
This also leads to lots of failed jobs because they aren't triggered in any order so the jobs are missing some artifacts they need which are produced by an upstream build.
AT:
- The Jenkins for the maintenance branch is triggered in a correct way, so there are no failed jobs because of run ordering / race conditions.
Possible solutions:
- The release job of the maintenance build triggers the top-most 'distro-pipeline'-job when it has committed all changes. There is still the possibility of a race condition but it is quite low because the Jenkins jobs have a quiet period of 5 seconds in which they don't start and also the SCM polling must occur right in the same time window.