Since openenshift/cluster-ingress-operator#817 merged, the e2e-aws-operator CI job has been failing for multiple PRs in the cluster-ingress-operator repository. In particular, the TestScopeChange test has been consistently failing. Example failures:
The operator is repeatedly logging errors like the following in those failing CI jobs:
The scope-wildcard dnsrecord is created for the TestScopeChange test.
Using search.ci, it seems that the failures occurred many times on #817 before it merged and then started occurring for the other PRs after #817 merged.
I filed a PR, openshift/cluster-ingress-operator#838, that reverts #817. I have run the e2e-aws-operator CI job on this PR twice. While the job has failed both times, the TestScopeChange test did not fail either time.
At this point, we have strong evidence that #817 is causing TestScopeChange to fail.
email@example.com did some testing and determined that there is some interaction between TestAllowedSourceRangesStatus and TestScopeChange. It may suffice to serialize some tests (TestScopeChanged is currently a parallel test, as is TestAllowedSourceRangesStatus and two other tests that #817 adds).
If the problem cannot be resolved by serializing tests, it may be necessary to revert #817 to unblock CI.
1. Run CI on a PR against the master branch of cluster-ingress-operator.
The TestScopeChange test fails as described.
TestScopeChange should not fail.