The kubernetesDeleteAllForClustersubroutine is trying to delete all config maps (without any selector), which also includes system-generated configmaps like kube-root-ca.crt and openshift-service-ca.crt on OpenShift clusters. These configmaps are managed by cluster operators and are recreated shortly after each modification or deletion, so when weathervane is trying to check whether all resources were deleted it still finds those configmaps and eventually fails after several iterations.
Reproduction steps
Deploy and run weathervane 2.1.1 on Red Hat OpenShift cluster (tested on OCP 4.12)
...
Expected behavior
Only resources with properly set label (by default app=auction) should be deleted and checked.
I found in the KubernetesCluster.pm source code there are several calls to kubectl with the --selector=app=auction argument, but for some reason this is missing in the kubernetesDeleteAllForCluster subroutine.
Describe the bug
The
kubernetesDeleteAllForCluster
subroutine is trying to delete all config maps (without any selector), which also includes system-generated configmaps likekube-root-ca.crt
andopenshift-service-ca.crt
on OpenShift clusters. These configmaps are managed by cluster operators and are recreated shortly after each modification or deletion, so when weathervane is trying to check whether all resources were deleted it still finds those configmaps and eventually fails after several iterations.Reproduction steps
...
Expected behavior
Only resources with properly set label (by default
app=auction
) should be deleted and checked.I found in the KubernetesCluster.pm source code there are several calls to
kubectl
with the--selector=app=auction
argument, but for some reason this is missing in thekubernetesDeleteAllForCluster
subroutine.Additional context
No response