operator run failed in outofcluster mod

• Which image of the operator are you using?
  Run postgres operator binary cmd with outofcluster flag directly as follows

./cmd --kubeconfig=/home/xxx/.kube/kubeconfig --outofcluster

• Where do you run it - cloud or metal? Kubernetes or OpenShift?
  Bare Metal K8s
• Are you running Postgres Operator in production?
  no
• Type of issue?
  question, may be a bug report

When operator run in outofcluster mod, it start failed with message

INFO[0000] Spilo operator
INFO[0000] no ConfigMap specified. Loading default values  pkg=controller
2021/02/18 17:17:50 Unable to detect operator namespace from within its pod due to: open /var/run/secrets/kubernetes.io/serviceaccount/namespace: no such file or directory

I found the help message about the --outofcluster is Whether the operator runs in- our outside of the Kubernetes cluster. In my opinion, outofcluster means for run operator binay command out of k8s cluster directly, and communicate with apiserver with kubeconfig, not need a pod in cluster.

Buf in fact, in outofcluster mod the operator still try to parse operator namespace use /var/run/secrets/kubernetes.io/serviceaccount/namespace which mounted and exist in a pod. Is my understanging wrong or a bug when operator run out of cluster ?

Reproduce the failure using ./cmd --kubeconfig=/home/xxx/.kube/kubeconfig --outofcluster outside Kubernetes and inspect the operator's namespace detection path. Check whether the service-account namespace lookup is expected in this mode or whether the kubeconfig should suffice. Done means the intended out-of-cluster behavior is established and covered by an appropriate regression test or clarified documentation.