How do you remove the last remaining OSD from a Proxmox/Ceph cluster?

[victorhooi]

I have a 3-node Ceph/Proxmox cluster.

I added some OSDs recently, using a separate volume as the DB/WAL device.

However, it turns out I didn't budget enough for the DB/WAL volumes, and I now need to re-create the OSDs from scratch.

Via the Web UI, I am able to select each OSD, and then go to "Stop", then "Out", then the "More -> Destroy".

However, for the last remaining osd (osd.0) - when I hit Stop, nothing happens.

Is there something special you need to do to remove the last remaining OSD?

Thanks,
Victor

 

[Alwin]

Is there something special you need to do to remove the last remaining OSD?

No. Try to stop the OSD on the CLI, to catch the output. It may just be old data in the GUI.

 

[victorhooi]

Just to update this - I called systemctl stop ceph-osd@0 multiple times, and it still appeared to be up (according to the Proxmox GUI).

However, what did work was simply calling ceph down on the OSD manually:

root@example-vm01:~# ceph osd tree
ID CLASS WEIGHT  TYPE NAME              STATUS REWEIGHT PRI-AFF
-1       1.88820 root default
-3       1.88820     host example-vm01
 0   ssd 1.88820         osd.0              up        0 1.00000
root@example-vm01:~# ceph osd down osd.0
marked down osd.0.

https://forum.proxmox.com/threads/ceph-osd-will-not-delete.58579/

 

[comfreak]

Yes, same issue and soloution for me.

A bug in Proxmox?!

 

[Alwin]

The last OSD is sometimes not marked as down. And has to be set manually.