I ended up manually cleaning up the OSD host, removing stale LVs and DM entries, and then purging the OSD with `ceph osd purge osd.19`. Looks like it's gone for good. /Z On Sat, 4 May 2024 at 08:29, Zakhar Kirpichenko <zakhar@gmail.com> wrote:

Hi!

An OSD failed in our 16.2.15 cluster. I prepared it for removal and ran `ceph orch daemon rm osd.19 --force`. Somehow that didn't work as expected, so now we still have osd.19 in the crush map:

-10 122.66965 host ceph02 19 1.00000 osd.19 down 0 1.00000

But OSD has been cleaned up on the host, although incompletely, as both block and block.db LVs still exist.

If I try to remove the OSD again, I get an error:

# ceph orch daemon rm osd.19 --force Error EINVAL: Unable to find daemon(s) ['osd.19']

How can I clean up this OSD and get rid of it completely, including the crush map? I would appreciate any suggestions or pointers.

Best regards, Zakhar