I found a misconfiguration in my ceph config dump: mgr advanced mgr/cephadm/migration_current 5 and changing it to 3 solved the issue and the orchestrator is back to working properly. That's something to do with the previous failed upgrade to Quincy, which updated automatically. If I understand correctly migration_current is somehow a safety feature in the upgrade. If you have more info, please let me know. Regards, Reza On Mon, 29 Aug 2022 at 10:50, Reza Bakhshayeshi <reza.b2008@xxxxxxxxx> wrote: > Hi > > I'm using the pacific version with cephadm. After a failed upgrade from > 16.2.7 to 17.2.2, 2/3 MGR nodes stopped working (this is a known bug of > upgrade) and the orchestrator also didn't respond to rollback services, so > I had to remove the daemons and add the correct one manually by running > this command: > > ceph orch daemon add mgr --placement=<my_host> > > As it was mentioned in some bugs I tried removing admin label and > reapplying them as well. > > Now, that the cluster status is healthy, but the orchestrator still > doesn't work properly when I'm going to add RGW node. Also I cannot upgrade > to a newer version: > > ceph orch host add <name> <IP> rgw-swift > ceph orch apply rgw swift --realm=<realm-name> --zone=<zone-name> > --placement="label:rgw-swift" --port=<port-num> > > I can't see any error logs. It seems like just not responding anymore. > I Also tried these commands: > > ceph orch pause/cancel/resume > ceph orch module enable/disable > > Do you have any idea? > > Best, > Reza > _______________________________________________ ceph-users mailing list -- ceph-users@xxxxxxx To unsubscribe send an email to ceph-users-leave@xxxxxxx
 |