-
Bug
-
Resolution: Done
-
Normal
-
None
-
4.13.0
-
None
-
Low
-
No
-
Storage Sprint 232
-
1
-
False
-
Description of problem:
This e2e test has a few flakes: [sig-storage][Feature:DisableStorageClass][Serial] should not reconcile the StorageClass when StorageClassState is Unmanaged [Suite:openshift/conformance/serial] Example: https://prow.ci.openshift.org/view/gs/origin-ci-test/pr-logs/pull/openshift_release/36373/rehearse-36373-periodic-ci-openshift-release-master-nightly-4.13-e2e-vsphere-ovn-upi-serial/1625549967886127104 The test was introduced in https://github.com/openshift/origin/pull/27704 From sippy, it seems only the Unmanaged test is flaking (on vSphere and AWS). This test seems particularly racy too. We set SC state to Unmanaged, immediately set AllowVolumeExpansion, and then periodically check the value of AllowVolumeExpansion until the test gives up. This model works for the Managed and Removed tests, where we expect the operator to reconcile the state eventually. But for unmanaged, we're changing something and expecting the operator NOT to reconcile it. This means we could set SC state to Unmanaged, immediately change AllowVolumeExpansion, and the operator could still revert that change before noticing the new StorageClassState value. But we never attempt to set AllowVolumeExpansion in this test again after the first try.
Version-Release number of selected component (if applicable):
4.13
How reproducible:
rarely (98% pass rate)
Steps to Reproduce:
1. 2. 3.
Actual results:
the Unmanaged tests occasionally fail
Expected results:
all DisableStorageClass tests pass consistently
Additional info: