Investigate freezing all filesystems in a pool before the thin pool fills up

[jbaublitz]

Follow up on discussion stratis-storage/project#652

We have had extensive conversations with the filesystem and block storage teams. The general consensus appears to be that approaches that will result in writes to XFS on thin pools returning ENOSPC are unlikely to be implemented due to performance concerns among other things.

We currently would like to consider remounting the filesystems on the pool as read-only once a certain fullness threshold is passed to avoid confusion from users if their data is not persisted and they do not call fsync.

[jbaublitz]

Unlikely to be workable in kernel or userspace.

[jbaublitz]

After discussing with the XFS team, XFS freeze may be an option.