Hi Eugen,
thanks for the information! On the test development I found out that the
best approach was to fire "ceph orch daemon rm mon. --force" to
remove the stopped daemon. After a few minutes, Ceph restarts the daemon
again on the host in a running state. This command works also on mon
There's an existing tracker issue [1] that hasn't been updated since a
year. The OP reported that restarting the other MONs did resolve it,
have you tried that?
[1] https://tracker.ceph.com/issues/52760
Zitat von Mevludin Blazevic :
Its very strange. The keyring of the ceph monitor is the
Its very strange. The keyring of the ceph monitor is the same as on one
of the working monitor hosts. The failed mon and the working mons also
have the same selinux policies and firewalld settings. The connection is
also present since, all osd deamons are up on the failed ceph monitor node.
Did you check the permissions? To me it reads like the permission
denied errors prevent the MONs from starting and then as a result they
are removed from the monmap:
ceph-8c774934-1535-11ec-973e-525400130e4f-mon-sparci-store1[786211]:
debug 2022-12-13T10:24:21.599+ 7f317ba4d700 -1
The keyring is the same, but I found the following log lines:
Dec 13 12:22:18 sparci-store1
ceph-8c774934-1535-11ec-973e-525400130e4f-mon-sparci-store1[813780]:
debug 2022-12-13T11:22:18.016+ 7f789e7f3700 0
mon.sparci-store1@1(probing) e18 removed from monmap, suicide.
Dec 13 12:22:18
So you get "Permission denied" errors, I'm guessing either the mon
keyring is not present (or wrong) or the mon directory doesn't belong
to the ceph user. Can you check
ls -l /var/lib/ceph/FSID/mon.sparci-store1/
Compare the keyring file with the ones on the working mon nodes.
Zitat von
Hi Eugen,
I assume the mon db is stored on the "OS disk". I could not find any
error related lines in cephadm.log, here is what journalctl -xe tells me:
Dec 13 11:24:21 sparci-store1
ceph-8c774934-1535-11ec-973e-525400130e4f-mon-sparci-store1[786211]:
debug 2022-12-13T10:24:21.392+
Hi,
do the MONs use the same SAS interface? They store the mon db on local
disk, so it might be related. But without any logs or more details
it's just guessing.
Regards,
Eugen
Zitat von Mevludin Blazevic :
Hi all,
I'm running Pacific with cephadm.
After installation, ceph