Had a fun time trying to figure out why when I mounted a PVC hosted on an NFS server with a Dell EMC Isilon storage array, and deployed a job which mounted this PVC, the container would not start and remain in an “Unavailable” state with no logs and no hint of what’s going on.
Turns out I had the “All Dirs” option for the specific export on the NFS server (Isilon) set to “No”, which means mounting subdirectories inside that share will fail. The share was defined at a higher level than the folder I was giving my Rancher persistent volume definition, e.g.:
Isilon NFS Export was defined for “/ifs/erictest”
Rancher PV definition was for “/ifs/erictest/k8s”
After trying a Linux NFS server (worked fine), and a second Isilon cluster of ours (worked fine, mounting the export-level folder), I spotted the “All Dirs” option on my original Isilon export and modified it, then everything worked right.