kube-arangodb icon indicating copy to clipboard operation
kube-arangodb copied to clipboard
verified publisher icon arangodb

Operator setting the pods' RestartPolicy cause gitops tools to fail health checks

Open sarahhenkens opened this issue 4 years ago • 1 comments

The operator starts all its server pod with the RestartPolicy set to Never. This causes ArgoCD to fail to report it as running.

https://github.com/arangodb/kube-arangodb/blob/13f3e2a09b4c6c08f050efffc364d498b1293dcf/pkg/util/k8sutil/pods.go#L433

Can this be changed to Always and let the pod automatically restart itself on failure?

Related ticket with repro steps: https://github.com/argoproj/argo-cd/issues/7259

sarahhenkens avatar Sep 20 '21 02:09 sarahhenkens

Hello!

We do not want to allow Pod restarts, full lifecycle is managed by Operator (Operator recreate pod, takes care about shards).

It is also used in Graceful shutdowns, when we send request thru API. This logic will change (to send Delete + handle it in finalizers), but ETA is next year.

Best Regards, Adam.

ajanikow avatar Sep 20 '21 10:09 ajanikow