Prometheus was not collecting metrics due to disk quota being exceeded. The PV still existed but was not bound to a PVC. The solution was to edit the PV, create a new PVC referencing the existing PV, copy the data from the old PV location to the new, and restart Prometheus. Lessons learned include properly sizing storage, verifying requirements, and having disaster recovery processes for monitoring infrastructure.
34. $ kubectl get pv
NAME CAPACITY ACCESS MODES
RECLAIM POLICY STATUS CLAIM
STORAGECLASS REASON AGE
pvc-dfb6cde7-e3d0-4e6a-b46d-4ac5dd3d97fb 20Gi RWX
Retain Released monitoring/prometheus-pvc
csi-cephfs 8d