[SOLVED] Removing a ZFS Pool from UI after disk deletion

[xantonin]

I removed a ZFS Pool before I could have a chance to remove it from the UI, and now the UI shows the pool still under the ZFS Tab.

I did the following:

• When I try to delete it from the UI, it says
  • command 'zpool list -vHPL ssd-raid10' failed: not a valid block device
  • (Note: Removal from "Datacenter" level is OK, it's still at PVE level)
• Tried to remove it with "pvesm remove ssd-raid10"
  • It says "delete storage failed: storage 'ssd-raid10' does not exist"
• Tried to re-add it to /etc/pve/storage.cfg and re-run the above - same thing.
• Disabled the systemctl service
• Ran "systemctl clean zfs-import@ssd\x2draid10.service"
• Checked to delete any related service files: (they're gone though)
  • find /etc/systemd/ -name "*x2draid10.service"
  • find /user/lib/ -name "*x2draid10.service"
• Ran systemctl daemon-reload
• Ran "service pveproxy restart"

I'm about running out of ideas.

 

[xantonin]

I just noticed the device shows up under "zpool status" still.

root@pve:~# zpool status
  pool: ssd-raid10
 state: SUSPENDED
status: One or more devices are faulted in response to IO failures.
action: Make sure the affected devices are connected, then run 'zpool clear'.
   see: https://openzfs.github.io/openzfs-docs/msg/ZFS-8000-HC
config:

        NAME                                               STATE     READ WRITE CKSUM
        ssd-raid10                                         DEGRADED     0     0     0
          mirror-0                                         DEGRADED     0     6     0
            ata-Samsung_SSD_850_EVO_250GB_S21NNSAFC47172A  REMOVED      0     0     0
            ata-Samsung_SSD_850_EVO_250GB_S21NNSAFC67440K  ONLINE       0     6     0
          mirror-1                                         DEGRADED     0    18     0
            ata-Samsung_SSD_850_EVO_250GB_S21NNSAFC46507N  REMOVED      0     0     0
            ata-Samsung_SSD_850_EVO_250GB_S21NNSAFC21972A  ONLINE       3    18     0

What really happened here is that I kept the disks in PVE, but moved them into the HBA for a VM and then booted the VM which had full control of the HBA through PCI Passthrough. This was all for my lab system and the disks were empty, so I wasn't worried about data loss - you probably don't want to do this without removing the zpool in production first (without destroying it) - which is why we're here. I guess I thought something would unlink the drives/pools once the VM took over. NOPE. My bad.

Anyway, since the pool still shows up, I'll try to nuke it from there. Figures I find a new thing the moment I start a thread for help. I'll log things to help others.

root@pve:~# zpool destroy -f ssd-raid10
cannot unmount '/ssd-raid10': pool or dataset is busy
could not destroy 'ssd-raid10': could not unmount datasets

root@pve:~# zpool export -f ssd-raid10
cannot unmount '/ssd-raid10': pool or dataset is busy

root@pve:~# mount
ssd-raid10 on /ssd-raid10 type zfs (rw,xattr,noacl)

Oh dear. I definitely messed up something. I'll probably have to shut down the VM or force unmount it. I'm not sure how/why it thinks 2 of the 4 drives are online though. They should all be removed.

 

[xantonin]

Ooops I put this in the wrong forum, I meant Proxmox VE: Installation and configuration, can someone move this?

I'll try shutting down my VM once it finishes it's task and update this if that works to remove the pool.

 

[xantonin]

Restarting the VM removed the dead zpool.