Description of problem: Sometimes builds are run out of order. > $ oc get builds --namespace=03d3u > NAME TYPE FROM STATUS STARTED DURATION > ruby-hello-world-1 Source Git@e79d887 Cancelled 6 minutes ago 23s > ruby-hello-world-2 Source Git@e79d887 Cancelled 4 minutes ago 12s > ruby-hello-world-3 Source Git New > ruby-hello-world-4 Source Git@e79d887 Running About a minute ago 1m56s > ruby-hello-world-5 Source Git@e79d887 Running About a minute ago 1m53s > ruby-hello-world-6 Source Git@e79d887 Running About a minute ago 1m57s Version-Release number of selected component (if applicable): STG 3.4.1.2 How reproducible: often Steps to Reproduce: 1. oc new-build openshift/ruby~https://github.com/openshift/ruby-hello-world 2. oc start-build bc/ruby-hello-world # 3 times 3. oc patch bc/ruby-hello-world -p '{"spec":{"runPolicy":"Parallel"}}' 4. oc start-build bc/ruby-hello-world # 3 times 5. oc get build -w Actual results: see above, parallel builds may kick in before a serial build Expected results: parallel builds wait for serial builds to complete
Actually I could not reproduce when builds execute and complete normally. If after step #4 though, I issue the command > oc cancel-build ruby-hello-world-1 then builds could end up like: > NAME TYPE FROM STATUS STARTED DURATI ON > ruby-hello-world-1 Source Git@e79d887 Cancelled 2 minutes ago 36s > ruby-hello-world-2 Source Git New > ruby-hello-world-3 Source Git New > ruby-hello-world-4 Source Git Pending > ruby-hello-world-5 Source Git Pending > ruby-hello-world-6 Source Git@e79d887 Running 10 seconds ago 9s So it appears to be related to canceling build. Or that just makes it easier to reproduce. If the issue does not manifest after canceling first build, then it can manifest after second build is canceled.
This was fixed by https://github.com/openshift/origin/pull/12699
Can't reproduce anymore with v3.6.135