ubuntu-repository-cache units fail at rebooting on azure

[Junien F]

Hi,

Because the Azure local storage can be wiped in a lot of situations (machine restarted, shut down, resized, etc...), and because the label of said disk is used to mount it in /etc/fstab, booting with a wiped disk will put the server in emergency mode, which means it won't fully boot.

The charm shouldn't use /etc/fstab at all (or with the "noauto" option) and format/partition/create a filesystem/mount the ephemeral disk in its "start" hook.

Thanks

[Junien F (axino)]

See also https://bugs.launchpad.net/ubuntu-repository-cache/+bug/1835319

[Haw Loeung (hloeung)]

Closing this one off. For Azure, we're now using persistent storage volumes and fstab also mounts the devices using 'nofail'. e.g.:

| # /srv/ubuntu-repository-cache was on /dev/sdc1 during installation
| UUID=37e15f62-7bf3-47a0-9112-3d567392fbbc /srv/ubuntu-repository-cache ext4 nofail,relatime,rw 0 0

The change to using 'nofail' landed in Juju 2.8.2 per LP:1877380.