DNS Brokenness Can Break Storage-Clients Boot-time Mounting

[ferricoxide]

Problem Description:

When deploying to a VPC where the storage-clients' resolver-setup cannot resolve the Gluster storage-server(s) by name, attempts to mount Gluster shares via systemd will silently fail.

Expected Behavior:

Reboot GlusterFS storage-client and all GlusterFS shares defined in /etc/fstab mount as defined.

Actual Behavior:

Reboot GlusterFS storage-client and all GlusterFS shares defined in /etc/fstab (may) silently fail

(Detailed) Steps to reproduce:

1. Deploy a GlusterFS storage-cluster
2. Deploy a GlusterFS storage-client
3. Add IP-based mount-definition in client's /etc/fstab file
4. Use mount -a to verify that fstab-defined mount-definition works as expected
5. Reboot storage-client
6. Find that fstab-defined mount is missing from list of active mounts and that system logs (journalctl -b 0) indicated that the mount-action successfully occurred .

(Optional) Fix recommendation:

Verified option:

1. Ensure that all members of the storage-cluster are resolveable
2. If DNS doesn't provide adequate resolver-response, add storage servers' self-advertised hostnames to storage client's /etc/hosts table
3. Reboot and verify the defined GlusterFS mounts are present

Alternate (Unverified) option:

1. Remove GlusterFS definitions from storage client's /etc/fstab file
2. Create a systemd mount service-definition for each GlusterFS share to mount to the client
3. Verify correctness of service-definition by executing systemctl start <MOUNT_PATH>.mount` and ensure that targeted mount-point is active