scylla_create_devices fails when var-lib-scylla.mount already exists

[ShlomiBalalis]

During a restart_then_repair nemesis in SCT, we've noticed that when we run scylla_create_devices on a restarted scylla instance the script fails when var-lib-scylla.mount already exists, a behavior we've never seen beforehand, and it is not a part of the known procedure.

As for SCT, at the moment we have a temporary fix for the issue: scylladb/scylla-cluster-tests#2839 which we will remove once this issue is fixed

[yaronkaikov]

@fruch ^

[fruch]

@fruch ^

We need to handle this inside create_devices scripts.
otherwise it would be break any user that would be trying to follow the procedure:
https://docs.scylladb.com/operating-scylla/procedures/cluster-management/rebuild-node/

if one would follow this procedure, he would fail can calling scylla_create_devices after restart of a node

[yaronkaikov]

@fruch almost 5 years :-), is this still relevant? i guess not really

[fruch]

the patch is still in SCT code, and I don't think anyone tried running this flow without that patch.

So I guess it's still relevant for the procedure mentioned about: https://docs.scylladb.com/operating-scylla/procedures/cluster-management/rebuild-node/

scylla_create_devices should be able to recreate /etc/systemd/system/var-lib-scylla.mount, as part of the procedure.
or the docs should mention that you first need to clear the mount file, and then call it.