Hi all,
I just updated a Ceph cluster from Nautilus to Octopus and followed the documentation in
order to migrate from the original ceph-ansible setup to cephadm.
Overall, this worked great, but there's one part that I couldn't figure out yet
and that doesn't seem to be documented: How do I migrate the OSDs to the new managed
approach using service specifications?
Currently, "ceph orch ps" shows me each OSD and "ceph orch ls" lists
them as "osd.2", with "9/0" running with unmanaged placement (iirc
osd.2 was the first one I adopted so that's probably where the name comes from).
I tried writing a service specification that should match the current deployment and
applying that, but the new entries are just sitting there at 0/3 running.
For node-exporter, I solved this problem by just removing the old containers and services
manually and waiting for Ceph to recreate the new ones, but for OSDs that approach
doesn't really seem practical (unless it was a matter of just stopping/removing the
old container, but that doesn't seem to do the trick in my tests).
Is there a proper way to do this? Or is the cluster just stuck with unmanaged OSDs if it
was created without cephadm?
Thanks,
Lukas
Show replies by date