We need to know exactly what data is stored on uffizi:/srv/storage/space as well as banco:/srv/storage/space, and make a decision about what we keep/move away/delete.
I have moved /srv/storage/space/annex on a dedicated zfs volume.
The thing I am note sure about is the order of operations at boot time, between mounts and nfs exports.
We need to ensure that /srv/storage/space is mounter before zfs start, then and only the nfs exports should occur.
What needs to be checked on next reboot is:
- will /srv/storage/space be mounted before zfs is started (probably not according to the content of /lib/systemd/system/zfs-mount.service)
- will nfs exports take place after zfs volumes are mounted (probably ok)
Also we may prefer to use zfs-share.service to declare nfs exports for the annex (and any other zfs-backed volumes that needs to be exported via nfs) instead of managing this 'by hand' in /etc/expots
See the sharenfs property of a zfs volume in 'man zfs'.
uffizi has definitely rebooted a bunch of times since we've done this. It's pretty clear the boot sequence is okay.
It definitely wouldn't hurt to recover the 100 TB /srv/storage/space partition from the old array (if anything, to make sure the data is not something we want to lose).
I've re-launched ncdu on the directory, this time within a tmux session (ncdu). It's definitely going to take a while!