Activity
From 07/31/2014 to 08/29/2014
08/29/2014
- 10:08 AM Bug #9260 (Resolved): hadoop fs gets EINVAL
- ...
- 09:59 AM Bug #9252 (Fix Under Review): Worker thread to advance MDS in absence of messages
- 09:59 AM Bug #9151 (Fix Under Review): mds should log/error/warn when segments are NOT getting trimmed
- 07:18 AM Feature #9287 (Rejected): qa: hadoop: add big top tests to suite
- 07:18 AM Feature #9286 (Rejected): qa: hadoop: test 2.x with teuthology
- 07:02 AM Feature #9284 (Resolved): mds: warn when clients are not responding to cache pressure
- 07:00 AM Feature #9283 (New): mds: limit inodes with caps to <100% of cache
- 07:00 AM Feature #9282 (Resolved): mds: warn (and kill?) sessions to clients which aren't revoking caps
- We need better ways of dealing with clients who aren't following our instructions. The most obvious of them is to sim...
- 06:37 AM Bug #9280 (Resolved): valgrind failures in ceph-fuse
/a/john-2014-08-29_03:49:04-fs-wip-jcsp-test-testing-basic-multi
Valgrind: client (Leak_DefinitelyLost, Leak_Ind...- 12:43 AM Bug #9123: kceph: had 130k+ inodes with write caps
- 12:43 AM Bug #9123: kceph: had 130k+ inodes with write caps
- I saw 10.214.137.25 in the mds log, 10.214.137.25 is gitbuilder-archive if IP hasn't changed. Maybe the issue and #89...
- 12:15 AM Bug #8962: kcephfs: client does not release revoked cap
- ...
08/28/2014
- 05:46 PM Bug #9266 (Resolved): ceph_test_libcephfs pool name failures
- ...
- 11:59 AM Bug #9266 (In Progress): ceph_test_libcephfs pool name failures
- 09:20 AM Bug #9266 (Resolved): ceph_test_libcephfs pool name failures
http://pulpito.front.sepia.ceph.com/teuthology-2014-08-25_23:04:01-fs-master-testing-basic-multi/451157/
Also fa...- 05:36 PM Bug #9276 (New): Client::get_file_extent_osds asserts in object_locator_to_pg if osd map is out o...
This would happen if files in the filesystem had layouts referring to pools that were no in the OSD map, which can ...- 04:21 PM Bug #9264 (Duplicate): mds: occasionally log segments can't trim
- 04:20 PM Bug #9264: mds: occasionally log segments can't trim
- ...
- 04:16 PM Bug #9264: mds: occasionally log segments can't trim
- ...
- 07:00 AM Bug #9264 (Duplicate): mds: occasionally log segments can't trim
- it happened with latest lab mds restart yesterday; we have the logs (for another 6 days or so)...
- 11:59 AM Bug #9260: hadoop fs gets EINVAL
- Could be related to #9266, a recurrence of something trying to look up pool names before osdmap is loaded in client. ...
- 10:08 AM Bug #9260: hadoop fs gets EINVAL
- Duh, that last exception was just libcephfs-java not being installed.
- 09:46 AM Bug #9260: hadoop fs gets EINVAL
Hmm, apparently there's more than one way this can fail:...- 11:31 AM Bug #9178: samba: ENOTEMPTY on "rm -rf"
- It might not need to if there's a client bug somewhere. (Or some other issue?)
- 06:40 AM Bug #9178: samba: ENOTEMPTY on "rm -rf"
- The strange thing is that the MDS never reply ENOTEMPTY.
- 09:17 AM Bug #9173 (Resolved): Crash in Server::_session_logged
- 08:09 AM Bug #8962: kcephfs: client does not release revoked cap
- ...
- 06:19 AM Bug #9252: Worker thread to advance MDS in absence of messages
- Testing on wip-jcsp-test
- 06:19 AM Bug #9151: mds should log/error/warn when segments are NOT getting trimmed
- Testing on wip-jcsp-test
08/27/2014
- 05:02 PM Bug #9260 (Resolved): hadoop fs gets EINVAL
- This will fail on hadoop fs -put with EINVAL. No apparent problems in the libcephfs log....
- 02:37 PM Feature #4583 (Resolved): libcephfs: add test that kills a client and verifies mds cleans it up
- ...
- 02:22 PM Feature #7810 (Resolved): libcephfs: add a test that freezes + unfreezes a client, and then verif...
- ...
- 02:21 PM Feature #4886 (Resolved): teuthology: add tests that use the MDS dumper
- 02:21 PM Feature #4886: teuthology: add tests that use the MDS dumper
- ...
- 10:03 AM Bug #8962: kcephfs: client does not release revoked cap
- ...
- 06:51 AM Bug #9252 (Resolved): Worker thread to advance MDS in absence of messages
As we move dispatchers outside of the MDS (first Objecter, now Beacon), there are some cases that don't progress pr...
08/26/2014
- 04:07 PM Bug #9105: ~ObjectCacher behaves poorly on EBLACKLISTED
- 04:04 PM Bug #9105: ~ObjectCacher behaves poorly on EBLACKLISTED
- NB the handing for this case in rbd landed with wip-objecter, keep this ticket open for general purpose ObjectCacher ...
- 04:05 PM Bug #9238: Floating point exception in Locker::calc_new_client_ranges
- Loic: if the original change was on firefly too then yes
- 11:34 AM Bug #9238: Floating point exception in Locker::calc_new_client_ranges
- Should this be backport to firefly ?
- 11:05 AM Bug #9238 (Resolved): Floating point exception in Locker::calc_new_client_ranges
- 09:50 AM Bug #9238 (Fix Under Review): Floating point exception in Locker::calc_new_client_ranges
- https://github.com/ceph/ceph/pull/2331
- 09:29 AM Bug #9238 (Resolved): Floating point exception in Locker::calc_new_client_ranges
On master, MDS starts fine first time and then crashes on second start.
Floating point error on Locker::calc_new...- 06:26 AM Bug #9212: mon election delays mds beacon
- mds sent beacon...
- 06:09 AM Bug #9152 (Fix Under Review): mds: beacon needs to not take mds_lock
08/25/2014
- 05:07 PM Bug #8878 (Resolved): mds lock cycle (wip-objecter)
- 03:56 PM Bug #9216: mds may regard active clients as stale due to slow pg recovery
- I haven't got that far yet, but if I had to guess I'd say it is not about caps, since when this happens, all existing...
- 10:26 AM Bug #9216: mds may regard active clients as stale due to slow pg recovery
- Interesting. Did you establish the mechanism by which the clients are being stale? Do they have a renew caps request ...
- 01:59 AM Bug #9216 (New): mds may regard active clients as stale due to slow pg recovery
- I occasionally get fuse and ceph.ko mounts into weird states, and I can generally track them down to the mds's decidi...
- 10:29 AM Bug #9212: mon election delays mds beacon
- 10:21 AM Bug #9212: mon election delays mds beacon
- Did we identify why it was blacklisted? I don't think we have any tests that should make it that slow or whatever.
- 09:30 AM Bug #9212 (Rejected): mon election delays mds beacon
- EBLACKLISTED
08/24/2014
- 09:25 AM Bug #9212 (Won't Fix): mon election delays mds beacon
- ubuntu@teuthology:/a/teuthology-2014-08-22_23:04:01-fs-master-testing-basic-multi/444359...
08/22/2014
- 02:57 AM Bug #4545: error creating empty object store. Invalid argument.
- i maybe found the problem.
before you mkcephfs,you should ensure the dir(/var/lib/ceph/osd/ceph-0) empty.
once i wr...
08/21/2014
- 04:44 PM Bug #5762 (Resolved): teuthology: Failed MPI runs lead to a hung test instead of a failure
- 09:55 AM Bug #9152 (In Progress): mds: beacon needs to not take mds_lock
- wip-9152
- 09:50 AM Bug #9177: ceph-fuse: failing MPI mdtest runs
- The compiler is spitting out a warning about getcwd -- no evidence that that's what it's actually hitting in this ins...
- 08:53 AM Bug #9177: ceph-fuse: failing MPI mdtest runs
- http://qa-proxy.ceph.com/teuthology/teuthology-2014-08-20_23:04:01-fs-next-testing-basic-multi/439228/
- 08:29 AM Bug #9177: ceph-fuse: failing MPI mdtest runs
- How did you track it down to getcwd? If that is the issue there are a bunch of avenues of attack here, and we should ...
- 06:31 AM Bug #9177: ceph-fuse: failing MPI mdtest runs
- mdtest has a getcwd call into an unzeroed buffer that it doesn't check the error of. If fuse is failing the getcwd f...
- 06:56 AM Bug #9151 (In Progress): mds should log/error/warn when segments are NOT getting trimmed
- 05:56 AM Feature #9189 (Resolved): Expose client identifying metadata to MDS, e.g. hostname
Currently, when doing e.g. a "session ls" on an MDS's admin socket, we get client IDs and IP addresses. It would b...- 05:35 AM Bug #9173 (Fix Under Review): Crash in Server::_session_logged
https://github.com/ceph/ceph/pull/2297
08/20/2014
- 10:47 AM Bug #9173: Crash in Server::_session_logged
- Better log.
- 06:30 AM Bug #9173 (Resolved): Crash in Server::_session_logged
Hit by mds_client_recovery task...- 10:33 AM Bug #9178: samba: ENOTEMPTY on "rm -rf"
- http://qa-proxy.ceph.com/teuthology/teuthology-2014-08-10_23:14:02-samba-next-testing-basic-plana/415869/
- 10:30 AM Bug #9178 (Resolved): samba: ENOTEMPTY on "rm -rf"
- ...
- 10:14 AM Bug #9177 (Resolved): ceph-fuse: failing MPI mdtest runs
- ...
08/19/2014
- 09:13 AM Bug #9152: mds: beacon needs to not take mds_lock
- Hmm, the beacon send code doesn't need to hold the lock on its own, but it's triggered by the SafeTimer, which is jus...
- 09:05 AM Bug #9151: mds should log/error/warn when segments are NOT getting trimmed
- What kind of logging do we want? I assume you mean journal segments, and this is a bog standard operation...
If it's... - 06:49 AM Fix #4286: SLES 11 - cfuse: disable 'big_writes'and 'atomic_o_trunc
08/17/2014
- 01:00 PM Bug #9152 (Resolved): mds: beacon needs to not take mds_lock
- any random task that holds the mds lock for a long time prevents beacons, which will trigger a failover
- 12:48 PM Bug #9151 (Resolved): mds should log/error/warn when segments are NOT getting trimmed
08/16/2014
- 03:42 PM Bug #8574 (Resolved): teuthology: NFS mounts on trusty are failing
- chef adds a dummy export and restarts nfs-kernel-server now
- 02:41 PM Bug #8574: teuthology: NFS mounts on trusty are failing
- root@mira055:~# service nfs-kernel-server restart
* Stopping NFS kernel daemon ...
08/15/2014
- 11:08 AM Feature #8869 (Resolved): MDS: support standby-replay on old-format journals
- This merged a couple of weeks ago in https://github.com/ceph/ceph/commit/440c820cce2c262570ab78e352bed8a630d41be5
- 04:45 AM Bug #9105: ~ObjectCacher behaves poorly on EBLACKLISTED
- Punting on a general purpose fix for ObjectCacher for the time being, and just fixing this in librbd teardown.
- 04:44 AM Bug #9105 (Fix Under Review): ~ObjectCacher behaves poorly on EBLACKLISTED
- https://github.com/ceph/ceph/pull/2263
08/14/2014
- 04:14 PM Bug #9101: multimds: unlinked file is not pruned from replica mds caches
- 03:20 PM Bug #9123 (Can't reproduce): kceph: had 130k+ inodes with write caps
- in #9121 the client had more than 130k inodes open for write, resulting in a huge file recovery queue. there definit...
- 02:37 PM Bug #9121 (In Progress): mds: inode stuck recovering after client restart
- recovery is working.. there are just a lot of inodes queued:
2014-08-14 14:40:06.695087 7fd45f757700 10 mds.0.cach... - 02:10 PM Bug #9121 (Resolved): mds: inode stuck recovering after client restart
- ...
- 01:51 PM Bug #9105: ~ObjectCacher behaves poorly on EBLACKLISTED
- John Spray wrote:
> This is happening when the librbd-using client is blacklisted, ObjectCacher fails to flush when ... - 10:16 AM Bug #9105: ~ObjectCacher behaves poorly on EBLACKLISTED
- This is happening when the librbd-using client is blacklisted, ObjectCacher fails to flush when requested, and ImageC...
- 09:44 AM Bug #9105: ~ObjectCacher behaves poorly on EBLACKLISTED
- Started failing in 061c8e93f76dc4fd6290d6d15723d76e73267444 where rbd_cache and rbd_cache_writethrough_until_flush we...
- 06:34 AM Bug #8725 (Resolved): mds crashed in upgrade:dumpling-x:stress-split-master-testing-basic-plana
08/13/2014
- 03:35 PM Bug #8964 (Resolved): kcephfs: client does not resend requests on mds restart
- 03:13 PM Bug #8725 (Fix Under Review): mds crashed in upgrade:dumpling-x:stress-split-master-testing-basic...
- https://github.com/ceph/ceph/pull/2254
- 02:15 PM Bug #9105 (New): ~ObjectCacher behaves poorly on EBLACKLISTED
In ceph master 78dc4df
http://qa-proxy.ceph.com/teuthology/teuthology-2014-08-11_23:00:01-rbd-master-testing-bas...- 01:44 PM Bug #8962: kcephfs: client does not release revoked cap
- ...
- 01:19 PM Bug #8962: kcephfs: client does not release revoked cap
- ...
- 01:39 PM Bug #9101: multimds: unlinked file is not pruned from replica mds caches
- looks like the problem is that another mds has the inode in its cache and isn't trimming it (or being asked to trim i...
- 01:13 PM Bug #9101 (Fix Under Review): multimds: unlinked file is not pruned from replica mds caches
- https://github.com/ceph/ceph/pull/2250
- 11:36 AM Bug #9101: multimds: unlinked file is not pruned from replica mds caches
- Here is the debug data when using a ceph-fuse client.
We did reproduce the problem - 11:15 AM Bug #9101 (New): multimds: unlinked file is not pruned from replica mds caches
- as a result, deleted files stay pinned for a long time and space does not get removed.
- 12:38 PM Feature #9029 (Resolved): min/max uid for snapshot creation
08/12/2014
- 07:31 AM Bug #9056: fuse kmod + ceph-fuse triggers "BUG: sleeping function called from invalid context"
- ...
- 06:51 AM Bug #9056 (Resolved): fuse kmod + ceph-fuse triggers "BUG: sleeping function called from invalid ...
- 05:10 AM Bug #9056: fuse kmod + ceph-fuse triggers "BUG: sleeping function called from invalid context"
- This is supposed to be fixed upstream in v3.16-rc6 by commit c55a01d360af, will close this when we've seen a clean fs...
- 06:56 AM Bug #8648: Standby MDS leaks memory over time
- Any change you can run one of these in standby under massif for a while? that will tell us what is leaking!
- 06:55 AM Bug #8651 (Won't Fix): crashing mds in an active-active mds setup
- this MDS got blacklisted. there is an open issues somewhere to make the shutdown more friendly, but the behavior is ...
- 06:50 AM Bug #8725: mds crashed in upgrade:dumpling-x:stress-split-master-testing-basic-plana
- we probably have to do a reencoding trick like we do in MOSDMap?
- 06:48 AM Bug #8876 (Resolved): kcephfs: hang on read of length 0
08/11/2014
- 04:09 AM Bug #8878 (In Progress): mds lock cycle (wip-objecter)
- I think all these are OK now in wip-mds-contexts: remaining failures on that branch are all outside MDS.
08/10/2014
- 02:16 AM Bug #8725: mds crashed in upgrade:dumpling-x:stress-split-master-testing-basic-plana
- "same error":http://pulpito.ceph.com/loic-2014-08-10_09:59:49-upgrade:firefly-x:stress-split-wip-9025-chunk-remapping...
- 12:53 AM Bug #8725: mds crashed in upgrade:dumpling-x:stress-split-master-testing-basic-plana
- Another "similar crash":http://pulpito.ceph.com/loic-2014-08-08_12:17:30-upgrade:firefly-x:stress-split-wip-9025-chun...
- 12:39 AM Bug #8725: mds crashed in upgrade:dumpling-x:stress-split-master-testing-basic-plana
- And the same trace at "upgrade:firefly-x:stress-split":http://pulpito.ceph.com/loic-2014-08-08_12:13:20-upgrade:firef...
- 12:33 AM Bug #8725: mds crashed in upgrade:dumpling-x:stress-split-master-testing-basic-plana
- Looks like a similar problem at "upgrade:firefly-x:stress-split":http://pulpito.ceph.com/loic-2014-08-08_12:13:20-upg...
08/09/2014
- 05:50 PM Bug #9056: fuse kmod + ceph-fuse triggers "BUG: sleeping function called from invalid context"
http://pulpito.front.sepia.ceph.com/john-2014-08-09_14:56:53-fs-wip-mds-contexts-testing-basic-plana/409236/
http:...- 05:48 PM Bug #9056 (Resolved): fuse kmod + ceph-fuse triggers "BUG: sleeping function called from invalid ...
kernel 5f740d7e1531099b888410e6bab13f68da9b1a4d
wip-mds-contexts (aka wip-objecter) 7be59771bff09e2b46b5467627cb...
08/07/2014
08/06/2014
- 12:25 PM Feature #9029 (Resolved): min/max uid for snapshot creation
- On shared systems like shared hosting it might be useful to prevent regular users from creating snapshots on CephFS.
... - 07:11 AM Feature #9026 (Resolved): client: vxattr support for rctime, rsize, etc.
08/03/2014
- 09:11 PM Bug #8962: kcephfs: client does not release revoked cap
- another similar hang:...
08/01/2014
- 03:51 PM Bug #8622 (Resolved): erasure-code: rados command does not enforce alignement constraints
- commit:7a58da53ebfcaaf385c21403b654d1d2f1508e1a
- 12:40 AM Bug #8962: kcephfs: client does not release revoked cap
- ...
07/31/2014
- 08:54 PM Bug #8962: kcephfs: client does not release revoked cap
- Zheng Yan wrote:
> Sage Weil wrote:
> > Zheng Yan wrote:
> > > no clue what happened. please dump the mds cache wh... - 07:32 PM Bug #8962: kcephfs: client does not release revoked cap
- Sage Weil wrote:
> Zheng Yan wrote:
> > no clue what happened. please dump the mds cache when it happens next time
... - 10:11 AM Bug #8962: kcephfs: client does not release revoked cap
- Zheng Yan wrote:
> no clue what happened. please dump the mds cache when it happens next time
We have a dump, act...
Also available in: Atom