Something I had suggested off-list (repeated here if anyone else finds themselves in a similar scenario):
since only one PG is dead and the OSD now seems to be alive enough to start/mount: consider taking a backup of the affected PG with
ceph-objectstore-tool --op export --pgid X.YY
(That might also take a loong time)
That export can later be imported into any other OSD if these three dead OSDs turn out to be a lost cause.
(Risk: importing the PG somewhere else might kill that OSD as well, depending on the nature of the problem; I suggested new OSDs as import target)
Paul
On Thu, Jun 13, 2019 at 3:52 PM Sage Weil <sage@xxxxxxxxxxxx> wrote:
On Thu, 13 Jun 2019, Harald Staub wrote:
> Idea received from Wido den Hollander:
> bluestore rocksdb options = "compaction_readahead_size=0"
>
> With this option, I just tried to start 1 of the 3 crashing OSDs, and it came
> up! I did with "ceph osd set noin" for now.
Yay!
> Later it aborted:
>
> 2019-06-13 13:11:11.862 7f2a19f5f700 1 heartbeat_map reset_timeout
> 'OSD::osd_op_tp thread 0x7f2a19f5f700' had timed out after 15
> 2019-06-13 13:11:11.862 7f2a19f5f700 1 heartbeat_map reset_timeout
> 'OSD::osd_op_tp thread 0x7f2a19f5f700' had suicide timed out after 150
> 2019-06-13 13:11:11.862 7f2a37982700 0 --1-
> v1:[2001:620:5ca1:201::119]:6809/3426631 >>
> v1:[2001:620:5ca1:201::144]:6821/3627456 conn(0x564f65c0c000 0x564f26d6d800
> :-1 s=CONNECTING_SEND_CONNECT_MSG pgs=18075 cs=1 l=0).handle_connect_reply_2
> connect got RESETSESSION
> 2019-06-13 13:11:11.862 7f2a19f5f700 -1 *** Caught signal (Aborted) **
> in thread 7f2a19f5f700 thread_name:tp_osd_tp
>
> ceph version 14.2.1 (d555a9489eb35f84f2e1ef49b77e19da9d113972) nautilus
> (stable)
> 1: (()+0x12890) [0x7f2a3a818890]
> 2: (pthread_kill()+0x31) [0x7f2a3a8152d1]
> 3: (ceph::HeartbeatMap::_check(ceph::heartbeat_handle_d const*, char const*,
> unsigned long)+0x24b) [0x564d732ca2bb]
> 4: (ceph::HeartbeatMap::reset_timeout(ceph::heartbeat_handle_d*, unsigned
> long, unsigned long)+0x255) [0x564d732ca895]
> 5: (ShardedThreadPool::shardedthreadpool_worker(unsigned int)+0x5a0)
> [0x564d732eb560]
> 6: (ShardedThreadPool::WorkThreadSharded::entry()+0x10) [0x564d732ed5d0]
> 7: (()+0x76db) [0x7f2a3a80d6db]
> 8: (clone()+0x3f) [0x7f2a395ad88f]
>
> I guess that this is because of pending backfilling and the noin flag?
> Afterwards it restarted by itself and came up. I stopped it again for now.
I think that increasing the various suicide timeout options will allow
it to stay up long enough to clean up the ginormous objects:
ceph config set osd.NNN osd_op_thread_suicide_timeout 2h
> It looks healthy so far:
> ceph-bluestore-tool --path /var/lib/ceph/osd/ceph-266 fsck
> fsck success
>
> Now we have to choose how to continue, trying to reduce the risk of losing
> data (most bucket indexes are intact currently). My guess would be to let this
> OSD (which was not the primary) go in and hope that it recovers. In case of a
> problem, maybe we could still use the other OSDs "somehow"? In case of
> success, we would bring back the other OSDs as well?
>
> OTOH we could try to continue with the key dump from earlier today.
I would start all three osds the same way, with 'noout' set on the
cluster. You should try to avoid triggering recovery because it will have
a hard time getting through the big index object on that bucket (i.e., it
will take a long time, and might trigger some blocked ios and so forth).
(Side note that since you started the OSD read-write using the internal
copy of rocksdb, don't forget that the external copy you extracted
(/mnt/ceph/db?) is now stale!)
sage
>
> Any opinions?
>
> Thanks!
> Harry
>
> On 13.06.19 09:32, Harald Staub wrote:
> > On 13.06.19 00:33, Sage Weil wrote:
> > [...]
> > > One other thing to try before taking any drastic steps (as described
> > > below):
> > >
> > > ceph-bluestore-tool --path /var/lib/ceph/osd/ceph-NNN fsck
> >
> > This gives: fsck success
> >
> > and the large alloc warnings:
> >
> > tcmalloc: large alloc 2145263616 bytes == 0x562412e10000 @ 0x7fed890d6887
> > 0x562385370229 0x5623853703a3 0x5623856c51ec 0x56238566dce2 0x56238566fa05
> > 0x562385681d41 0x562385476201 0x5623853d5737 0x5623853ef418 0x562385420ae1
> > 0x5623852901c2 0x7fed7ddddb97 0x56238536977a
> > tcmalloc: large alloc 4290519040 bytes == 0x562492bf2000 @ 0x7fed890d6887
> > 0x562385370229 0x5623853703a3 0x5623856c51ec 0x56238566dce2 0x56238566fa05
> > 0x562385681d41 0x562385476201 0x5623853d5737 0x5623853ef418 0x562385420ae1
> > 0x5623852901c2 0x7fed7ddddb97 0x56238536977a
> > tcmalloc: large alloc 8581029888 bytes == 0x562593068000 @ 0x7fed890d6887
> > 0x562385370229 0x5623853703a3 0x5623856c51ec 0x56238566dce2 0x56238566fa05
> > 0x562385681d41 0x562385476201 0x5623853d5737 0x5623853ef418 0x562385420ae1
> > 0x5623852901c2 0x7fed7ddddb97 0x56238536977a
> > tcmalloc: large alloc 17162051584 bytes == 0x562792fea000 @ 0x7fed890d6887
> > 0x562385370229 0x5623853703a3 0x5623856c51ec 0x56238566dce2 0x56238566fa05
> > 0x562385681d41 0x562385476201 0x5623853d5737 0x5623853ef418 0x562385420ae1
> > 0x5623852901c2 0x7fed7ddddb97 0x56238536977a
> > tcmalloc: large alloc 13559291904 bytes == 0x562b92eec000 @ 0x7fed890d6887
> > 0x562385370229 0x56238537181b 0x562385723a99 0x56238566dd25 0x56238566fa05
> > 0x562385681d41 0x562385476201 0x5623853d5737 0x5623853ef418 0x562385420ae1
> > 0x5623852901c2 0x7fed7ddddb97 0x56238536977a
> >
> > Thanks!
> > Harry
> >
> > [...]
> > _______________________________________________
> > 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
_______________________________________________ ceph-users mailing list ceph-users@xxxxxxxxxxxxxx http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com
