otto/tools: cephadm upgrade script cephfs cluster with multiple fs

[JoshuaGabriel]

Upgrade the MDS daemons of a single CephFS filesystem without disturbing the
others, working around the cephadm limitation where
ceph orch upgrade ... --services mds.
prepares (fails / scales down) ALL filesystems instead of only the targeted one.

written by @FredNass

some features

• non-interactive mode by default (--interactive is available)
• setting mds_join_fs and refuse_standby_for_another_fs on some or all filesystems
• upgrading all --all or multiple filesystems -f cephfs,cephfs2 (one after the other) in a single run
• downgrading MDSs with --force-downgrade and appropriate warning.
• using a specific repo other than quay.io/ceph/ceph

[JoshuaGabriel]

uploading it for now, I have looked at it briefly and it looks good, but will give a second review

[trociny]

Lets name the file clyso-cephfs-upgrade (without .sh suffix), so it follows our convention we use for other files.

If we do not like clyso- prefix for these scripts or prefer ceph-filesystem instead of cephfs, lets change it for all scripts in tool directory.

[FredNass]

There you go @trociny:

• Added -I/-J/-R short options,
• fail_fs is now the default,
• and filesystem/method can be passed positionally.

I intentionally left --force-downgrade without a short option, to avoid a dangerous downgrade caused by confusing -F with -f (--filesystem).

I edited the script in your repo @JoshuaGabriel. Just noticed that its name is "ceph-filesystem-upgrade.sh" instead of "clyso-ceph-filesystem-upgrade.sh" in my repo. We might want to rename it to match other tools names.

[JoshuaGabriel]

should be renamed now

[trociny]

Otherwise LGTM. Note, I have not tested it though.