Hi Greg,
Firstly, many thanks for your advice.
I'm perplexed as to why the crush map is upset; the host names look the
same, each node has a fixed IP on a single bond0 interface.
Perhaps the problems were an artefact of having "nodown" set?
As you suggested, I've unset "osd nodown" and am
On Thu, May 31, 2018 at 5:07 AM Jake Grimmett wrote:
> Dear All,
>
> I recently upgraded our Ceph cluster from 12.2.4 to 12.2.5
> & simultaneously upgraded the OS from Scientific Linux 7.4 to 7.5
>
> After reboot, 0.7% objects were misplaced and many pgs degraded.
>
> the cluster had no client
Dear All,
I recently upgraded our Ceph cluster from 12.2.4 to 12.2.5
& simultaneously upgraded the OS from Scientific Linux 7.4 to 7.5
After reboot, 0.7% objects were misplaced and many pgs degraded.
the cluster had no client connections, so I speeded up recovery with:
ceph tell 'osd.*'