Hi,
I'm glad you could resolve the issue.
Zitat von Iztok Gregori via ceph-users <[email protected]>:
On 15/04/26 12:29, Iztok Gregori via ceph-users wrote:
On 15/04/26 12:08, Eugen Block via ceph-users wrote:
Yeah, looks like it. I just compared the output with a fresh
Tentacle install, I only enabled log-to-file during bootstrap, so
log_to_file is true in my cluster as well. There's no
leveldb_log_to_ceph_log option anymore since the default has been
rocksdb for a while now.
I put 'log_to_journald' to true and journalctl started to log. A
lot (with debug=20) :-)
Now I'm ready to debug the joining problem. I will keep you posted.
Just an update, the restart of the Octopus daemon didn't had an
effect, one the new monitors were started they still stayed in the
"electing" phase causing the 'ceph ...' commands to hang. Not a lot
in the logs, or nothing that I was able to point out.
After that we decided to stop the Octopus daemon and ran the cluster
with the remaining 2 Quincy ones. This did the job.
Once the Octopus monitor wasn't running any more the new Quincy
monitors rapidly joined the cluster. And now we are able to continue
our upgrade path to Squid/Tentacle and cephadm.
I'm not 100% sure of what was wrong, it seems that in an not
completely upgraded Quincy monitor cluster is not possible to add
new monitors. Once the monitors were all the same version, it
started to work again.
Thanks a lot for your help!
Cheers
Iztok
_______________________________________________
ceph-users mailing list -- [email protected]
To unsubscribe send an email to [email protected]
_______________________________________________
ceph-users mailing list -- [email protected]
To unsubscribe send an email to [email protected]
