Open jan-g opened 1 month ago
The flaky failure condition seems related the cleanup order of the storageclass.
It's just passing the output of kustomize into kubectl delete. Every kustomize call is (re)installing the storageclass helper.
This PR explicitly tags the storageclass stuff so that kubectl delete has a label to only ditch the pod & pvc before it waits on the cleanup completing.
If we're unlucky, the local-path-provisioner can be killed before it has a chance to clean up orphened PVs. These may then hang around while subsequent tests run. Depending on the lexical sort order of those PVs, the
typeCheckCmd
inrunTest
may select a moribund PV, causing the test to randomly fail.