-
Notifications
You must be signed in to change notification settings - Fork 21
Description
What did you do to encounter the bug?
I was doing helm upgrade
to set operator.watchNamespace=*
, and some RBAC resource content are seem immutable, so i uninstall MCK and install again.
During uninstall, seem all RBAC resources are removed, including mongodb-kubernetes-appdb
which is referenced by the StatefulSet of the existing MongoDBCommunity instance.
It makes :-
- the
mongodb-agent
fails on readiness probe - the pod endpoints will be not ready (I have extra exposing services which will not function at time)
- the StatefulSet warns of the missing
mongodb-kubernetes-appdb
serviceaccount (if pods are restarted/deleted during this status, the new ones will not be scheduled)
p.s. although I install the MCK again very soon, above status still long for 30-60 minutes (I have no idea about this)
Also, if I apply a new MongoDBCommunity in another namespace, the mongodb-kubernetes-appdb
resources will not be created in that namespace automatically so the pods will not be scheduled (same reason to point 3 above).
Operator Information
- Operator Version: 1.2.0
- MongoDB Image used: mongodb-community-server:8.0.12-ubi9
- MongoDB Agent Image used: mongodb-agent-ubi:108.0.2.8729-1
Kubernetes Cluster Information
- Distribution: Native K8s (on-premise, AlmaLinux 9.6)
- Version: 1.32.7
- Image Registry location: Quay