[PVE-User] After update Ceph monitor shows wrong version in UI and is down and out of quorum

Frank Thommen f.thommen at dkfz-heidelberg.de
Tue Jan 5 20:08:09 CET 2021


On 05.01.21 20:01, Frank Thommen wrote:
> 
> On 04.01.21 12:44, Frank Thommen wrote:
>>
>> Dear all,
>>
>> one of our three PVE hypervisors in the cluster crashed (it was fenced 
>> successfully) and rebooted automatically.  I took the chance to do a 
>> complete dist-upgrade and rebooted again.
>>
>> The PVE Ceph dashboard now reports, that
>>
>>    * the monitor on the host is down (out of quorum), and
>>    * "A newer version was installed but old version still running, 
>> please restart"
>>
>> The Ceph UI reports monitor version 14.2.11 while in fact 14.2.16 is 
>> installed. The hypervisor has been rebooted twice since the upgrade, 
>> so it should be basically impossible that the old version is still 
>> running.
>>
>> `systemctl restart ceph.target` and restarting the monitor through the 
>> PVE Ceph UI didn't help. The hypervisor is running PVE 6.3-3 (the 
>> other two are running 6.3-2 with monitor 14.2.15)
>>
>> What to do in this situation?
>>
>> I am happy with either UI or commandline instructions, but I have no 
>> Ceph experience besides setting up it up following the PVE instructions.
>>
>> Any help or hint is appreciated.
>> Cheers, Frank
> 
> In an attempt to fix the issue I destroyed the monitor through the UI 
> and recreated it.  Unfortunately it can still not be started.  A popup 
> tells me that the monitor has been started, but the overview still shows 
> "stopped" and there is no version number any more.
> 
> Then I stopped and started Ceph on the node (`pveceph stop; pveceph 
> start`) which resulted in a degraded cluster (1 host down, 7 of 21 OSDs 
> down). OSDs cannot be started through the UI either.
> 
> I feel extremely uncomfortable with this situation and would appreciate 
> any hint as to how I should proceed with the problem.
> 
> Cheers, Frank

OSDs and MDSs just took a bit to start, so from this side it looks ok 
now.  But the monitor still refuses to start.

Frank



More information about the pve-user mailing list