[PVE-User] Ceph noout doesn't work properly!

Ronny Aasen ronny+pve-user at aasen.cx
Tue Sep 4 22:16:52 CEST 2018

On 04.09.2018 21:02, Gilberto Nunes wrote:
> Hi
> I try to use noout in Ceph Luminous Cluster, but is seems doesn't work!
> I follow this:
> https://pve.proxmox.com/wiki/Ceph_Server#noout
> Try either via web gui and shell.
> I get some disk with I/O failure so try to stop OSD related that disk, and
> the Ceph start 86542/2196939 objects misplaced (3.939%)....
> The noout option is not supposed to prevent this behavior?
> The ceph cluster was builded with 6 servers of Proxmox 5.2-7
> Thanks a lot.

noout should prevent the mon's from automagically removing the osd from 
the crushmap,  if recovery did start, perhaps you edited the crushmap 
manually ? perhaps by removing the osd or setting the disk down/out 
manually ?

if you want to hard block recovery you can use nobackfill, norecovery.  
but the whole purpose of ceph is to keep your data safe, so letting it 
recover is a good thing imho.

kind regards
Ronny Aasen

More information about the pve-user mailing list