Unbuntu 20 and OpenZFS basics

1) Check the backup setup?  It suppose to be doing hourly snapshots...

This will depend on how your backups are tooled.

a) First check local machine's CRON file to see if you're running a push backup, where data is pushed to an archive machine.

If you can't find any CRON jobs running, then...

b) You'll check your backup archive machine's CRON file to local the pull backup script, where data is pulled to the archive machine.

c) Every backup setup is unique. If you get stumped, you'll likely hire someone to help you... decipher how backups are generated...

2) How to increase the size of a drive that is almost running out of space.

You can only do this if you have free space or by adding additional physical disks.

Rarely will someone... add physical disks... then leave this space unused...

So your likely first step will be to add physical disks, then go through the process of adding these disks as a separate ZFS pool or integrating your new disk(s) into an existing ZFS pool.

ZFS is complex... One minor wrong command can leave you in an impossible to reverse travesty... so before you do any ZFS work, ensure your full backups are working correctly + you've made a hot backup before doing any ZFS pool work.

3) What type of issues would cause me to have to reboot the hypervisor?

None.

Rebooting your Hypervisor or any VM running ZFS should just work.

Random Aside: Working with ZFS... has worked very poorly each time I've spun up ZFS.

1) ZFS, at least last time I tested it on Ubuntu Focal a year ago... was horribly slow all the time...

Painfully slow, compared with EXT4.

2) ZFS glitched periodically, where all I/O would just hang... sometimes for minutes... then resume...

No rhyme or reason to this. No errors logged. Just I/O pausing + resuming.

3) Since I run 1000s of production sites, both slow I/O + glitchy I/O are unacceptable.

So I fell back to EXT4 in all cases.

4) For the 1x case where tail packing or sub block allocation is useful (1,000,000s of IMAP files), I've been testing.

Dovecot + mdbox + ZSTD compression + mysqlfs + MariaDB-10.6.X as backing store.

This means only 1x inode/table, rather than 1x inode/file, so in this specific use case massive volume mail servers inode usage can drop from 95%+ to <1%.

5) If you find your daylight hours eaten up by ZFS maintenance or slow/glitchy I/O problems, switching to EXT4 might be useful.

Awesome... But what are the commands to do so?