[cnpg] Multi-Attach error for volume

[jkleinlercher]

When cnpg updates and does a RollingUpdate on the replicaset the new pod cannot start because of

Multi-Attach error for volume "pvc-c2edf35f-ba31-48a3-946f-db9afb366aa3" Volume is already used by pod(s) sx-cnpg-pgadmin4-7798468bd6-ln42z

I found some good explanation in https://medium.com/@golusstyle/demystifying-the-multi-attach-error-for-volume-causes-and-solutions-595a19316a0c which is an exact match for our problem.

• we use a deployment with replicaset RollingUpdate strategy
• our PVC is ReadWriteOnce

So it is more a question why it doesn't occur more often.

[jkleinlercher]

this happened today when cnpg got updated

workaround: simply delete the old replicaset, then the new replicaset with its pod can start

still the question is why cnpg upstream chart is designed like this. For me it seems this can happen always

[jkleinlercher]

according to https://github.com/rowanruseler/helm-charts/tree/main/charts/pgadmin4 we can configure rollout strategy in deployment an access mode in pvc. however, need to evaluate which config is the best for use