Re: OSD won't go up after node reboot

[Евгений Д. <ineu.main@xxxxxxxxx>]

Data lives in another container attached to OSD container as Docker volume. According to `deis ps -a`, this volume was created two weeks ago, though all files in `current` are very recent. I suspect that something removed files in the data volume after reboot. As reboot was caused by CoreOS update, it might be newer version of Docker (1.6 -> 1.7) that introduced the problem. Or maybe it was container initialization process that somehow removed and recreated files.

I don't have this data volume anymore, so can only guess.

2015-08-31 18:28 GMT+03:00 Jan Schermer <jan@xxxxxxxxxxx>:

Is it possible that something else was mounted there?

Or is it possible nothing was mounted there?

That would explain such behaviour...

Jan

On 31 Aug 2015, at 17:07, Евгений Д. <ineu.main@xxxxxxxxx> wrote:

No, it really was in the cluster. Before reboot cluster had HEALTH_OK.

Though now I've checked `current` directory and it doesn't contain any data:

        root@staging-coreos-1:/var/lib/ceph/osd/ceph-0# ls current

        commit_op_seq  meta  nosnap  omap

while other OSDs do. It really looks like something was broken on reboot, probably during container start, so it's not really related to Ceph. I'll go with OSD recreation.

Thank you.

2015-08-31 11:50 GMT+03:00 Gregory Farnum <gfarnum@xxxxxxxxxx>:

On Sat, Aug 29, 2015 at 3:32 PM, Евгений Д. <ineu.main@xxxxxxxxx> wrote:
> I'm running 3-node cluster with Ceph (it's Deis cluster, so Ceph daemons are
> containerized). There are 3 OSDs and 3 mons. After rebooting all nodes one
> by one all monitors are up, but only two OSDs of three are up. 'Down' OSD is
> really running but is never marked up/in.
> All three mons are reachable from inside the OSD container.
> I've run `log dump` for this OSD and found this line:
>
> Aug 29 06:19:39 staging-coreos-1 sh[7393]: -99> 2015-08-29 06:18:51.855432
> 7f5902009700  3 osd.0 0 handle_osd_map epochs [1,90], i have 0, src has
> [1,90]
>
> Is it the reason why OSD cannot connect to the cluster? If yes, why could it
> happen? I haven't removed any data from /var/lib/ceph/osd.
> Is it possible to bring this OSD back to cluster without completely
> recreating it?
>
> Ceph version is:
>
> root@staging-coreos-1:/# ceph -v
> ceph version 0.94.2 (5fb85614ca8f354284c713a2f9c610860720bbf3)

It's pretty unlikely. I presume (since the OSD has no maps) that it's
never actually been up and in the cluster? Or else its data store has
been pretty badly corrupted since it doesn't have any of the requisite
metadata. In which case you'll probably be best off recreating it
(with 3 OSDs I assume all your PGs are still active).
-Greg

_______________________________________________
ceph-users mailing list
ceph-users@xxxxxxxxxxxxxx
http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com

_______________________________________________
ceph-users mailing list
ceph-users@xxxxxxxxxxxxxx
http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com