CephFS NodeGetVolumeStats procedure should accept standard staging path

[nixpanic]

The NodeGetVolumeStats procedure in CephFS does not handle requests for using the inital staging path correctly.

This isn't a problem with Kubernetes, as that uses the publish path by default:

/var/lib/kubelet/pods/d88426aa-2d97-41dd-b754-2794f97f8ba0/volumes/kubernetes.io~csi/pvc-78770630-74dd-4b6a-bef3-3e7307e5f52d/mount

The CSI-Addons Volume Condition Reporter uses the staging path by default, which appends /globalmount to the requested target directory:

/var/lib/kubelet/plugins/kubernetes.io/csi/openshift-storage.cephfs.csi.ceph.com/b38dba31761526eebc9b397a08986ee4dab5fa2515f8dc85bd47059ee9c8709a/globalmount

The /globalmount is not part of the CSI specification, and callers of NodeGetVolumeStats need to know about this Ceph-CSI specific thing.

Ideally Ceph-CSI handles NodeGetVolumeStats requests in such a way, that it is checked if the path was used for staging, or was used for publishing (but without the known kubelet format of the paths). A difficulty is that there should not be any checks on the path itself, as that can cause hangs (csi-addons/kubernetes-csi-addons#850).

This issue is handled in RBD with getStagingPath(). However, it uses Mounter.GetMountRefs which calls stat on the path. The stat syscall can potentially hang, and should not be called in functions that return a Volume Condition.

[github-actions[bot]]

This issue has been automatically marked as stale because it has not had recent activity. It will be closed in a week if no further activity occurs. Thank you for your contributions.

[github-actions[bot]]

This issue has been automatically closed due to inactivity. Please re-open if this still requires investigation.