Troubleshooting

Troubleshooting

PersistenceReachable=False

Check the datastore host/port and that the password Secret exists with the referenced key.
The operator pings the datastore from its own pod (in temporal-system), so a Postgres host must be a namespace-qualified FQDN (e.g. temporal-pg-rw.temporal.svc.cluster.local), not a bare service name.
For Postgres, confirm the database and visibility database exist. The operator does not create databases — both temporal and temporal_visibility must exist before schema setup.
remaining connection slots are reserved for ... SUPERUSER (SQLSTATE 53300) means Postgres ran out of connections. Each service pod pools to both stores; raise max_connections (200 is a good starting point for a small cluster).

Inspect the schema Jobs: kubectl get jobs -l app.kubernetes.io/component=schema.
View logs of a failed schema Job for temporal-sql-tool / temporal-cassandra-tool errors.

Ensure cert-manager is installed and the referenced issuerRef exists.
Check the Certificate resources: kubectl describe certificate <cluster>-internode.

kubectl get deploy -l app.kubernetes.io/instance=<cluster> and inspect pod events (image pulls, probe failures, membership join issues).
On mTLS clusters the request-serving pods use TCP probes rather than gRPC ones (the native gRPC prober cannot present a client certificate); see the “mTLS health probes” section under Operations for details.