Activity
From 06/22/2016 to 07/21/2016
07/21/2016
- 09:11 PM Bug #16771 (New): mon crash in MDSMonitor::prepare_beacon on ARM
- ceph 10.2.2
ubuntu 16.10
in Docker version 1.11.1, build 5604cbe
on arch armhf (rapsberry pi running hypriot)
<... - 08:11 PM Bug #16768 (Resolved): multimds: check_rstat assertion failure
- ...
- 04:31 PM Bug #16042 (Pending Backport): MDS Deadlock on shutdown active rank while busy with metadata IO
- 02:42 PM Bug #16668: client: nlink count is not maintained correctly
- Pull request with the fix is up here:
https://github.com/ceph/ceph/pull/10386 - 01:15 PM Bug #16764 (Resolved): ceph-fuse crash on force unmount with file open
Reproducing this in a vstart environment:
1. Mount a client
2. in python, do "f = open('mnt/foo.bin', 'w')"
3....- 11:08 AM Bug #16397: nfsd selinux denials causing knfs tests to fail
- The other thing of note is the logs seem to indicate that these hosts are running pretty bleeding-edge kernels -- 4.7...
- 11:06 AM Bug #16397: nfsd selinux denials causing knfs tests to fail
- As Scott Mayhew pointed out, the version of nfs-utils that ships in RHEL7.2 uses fopen to open the channel file, and ...
- 12:11 AM Bug #4212: mds: open_snap_parents isn't called all the times it needs to be
- I had a misunderstanding about what data a SnapRealm/sr_t has directly.
So, yes, right now we need *all* past_pare...
07/20/2016
- 08:19 PM Feature #16757 (New): enable MDS replacement migration
- Right now, without multi-mds the only way we have to switch MDSes is to do a failover from the current active to some...
- 02:02 PM Feature #16468: kclient: Exclude ceph.* xattr namespace in listxattr
- I gave a shot at fixing this today (kclient only) as per the email thread.
listxattr() does not return internal xa... - 02:00 PM Bug #16668: client: nlink count is not maintained correctly
- Ok, I have a couple of small patches that fix the testcase. One is a client-side patch to fix the ctime handling in f...
- 01:37 PM Support #16526: cephfs client side quotas - nfs-ganesha
- Oh, we just recently flipped the bit so quotas are enforced by default. This should work if you set "client quota = t...
- 09:21 AM Support #16526: cephfs client side quotas - nfs-ganesha
- For this test I was using the below versions:
ceph version 10.2.1 (3a66dd4f30852819c1bdaa8ec23c795d4ad77269)
nfs-... - 12:15 PM Bug #16754 (Can't reproduce): mounting cephfs root and sub-directory on the same node makes the s...
- Steps to reproduce:
*********************************************************************
ems@host1: sudo mount -t ... - 08:06 AM Bug #16737: Mounting ceph fs on client leads to kernel crash
- attaching full kernel log
- 06:15 AM Bug #3718: multi-client dbench gets stuck over NFS exported cephfs
- Probably not a bug any more?
- 06:14 AM Bug #1787 (Closed): mds: laggy oneshot replays pollute mdsmap
- One shot replay got zapped.
- 06:08 AM Bug #5864 (Closed): cfuse_workunit_suites_ffsb suite on Centos hangs with *** Got Signal Interrup...
- 06:08 AM Bug #4732 (Closed): uclient: client/Inode.cc: 126: FAILED assert(cap_refs[c] > 0)
- Things have changed.
- 05:59 AM Bug #4738 (Closed): libceph: unlink vs. readdir (and other dir orders)
- We have file locking and redid the listing code.
- 05:57 AM Bug #8432 (Closed): ceph-fuse process not dying
- These are definitely out of date, whatever the bug was.
- 05:52 AM Bug #9276: Client::get_file_extent_osds asserts in object_locator_to_pg if osd map is out of date
- This might be fixed now?
- 05:46 AM Bug #3845 (Closed): mds: standby_for_rank not getting cleared on takeover
- A bunch of this got rejiggered in John's multi-fs and follow-on work; it's probably gone.
- 05:42 AM Bug #9884 (Closed): too many files in /usr for multiple_rsync.sh
- Pretty sure we reduced the size and this isn't a problem any more.
- 05:41 AM Bug #10061: uclient: MDS: output cap data in messages
- This should also be exposed via the admin socket.
- 05:35 AM Bug #10542: ceph-fuse cap trimming fails with: mount: only root can use "--options" option
- I think this got resolved into one of the many fuse cache invalidate PRs, but I'm not sure.
- 05:28 AM Cleanup #11 (Resolved): mds: replace ALLOW_MESSAGES_FROM macro
- This got fixed up in the security stuff last summer.
- 12:29 AM Feature #16745 (Pending Backport): mon: prevent allocating snapids allocated for CephFS
- The MDS allocates its own snapids. In general, the monitor allocates self-managed snapids for librados users.
We n...
07/19/2016
- 11:29 PM Bug #11789 (Can't reproduce): knfs mount fails with "getfh failed: Function not implemented"
- 11:28 PM Bug #12209 (Won't Fix): CephFS should have a complete timeout mechanism to avoid endless waiting ...
- There's been no movement here and we didn't seem to like the idea.
- 11:26 PM Bug #13689 (Won't Fix): ceph-mds not build with libjemalloc
- We're switching to cmake so hopefully this is fixed.
- 11:23 PM Support #15268 (Resolved): CephFS mount blocks VM
- 11:22 PM Bug #15783: client: enable acls by default
- Zheng?
- 11:20 PM Documentation #3113 (Rejected): Ceph FS Options Could Use Some Additional Information
- The cephfs tool got zapped.
- 11:19 PM Fix #4286 (Rejected): SLES 11 - cfuse: disable 'big_writes'and 'atomic_o_trunc
- I think/hope we can ditch this now. There have been several SLES11 service packs and SLES12 is out now.
- 11:12 PM Bug #16322 (Need More Info): ceph mds getting killed for no reason
- 11:08 PM Bug #15502 (Resolved): files read or written with cephfs (fuse or kernel) on client drop all thei...
- I think this is all cleaned up now.
- 10:54 PM Bug #4212: mds: open_snap_parents isn't called all the times it needs to be
- See the email thread at http://www.spinics.net/lists/ceph-devel/msg12818.html
Unfortunately it doesn't include any... - 08:39 PM Documentation #16743 (Resolved): client: config settings missing in documentation
- These include at least:
* client_cache_mid
* client_oc_size
* client_oc_max_objects
* client_oc_max_dirty
* cl... - 08:05 PM Bug #16668: client: nlink count is not maintained correctly
- I think the actual bug here is that, as you note, ll_lookup calls fill_stat without checking that it has As (and what...
- 05:01 PM Bug #16668: client: nlink count is not maintained correctly
- Actually we could probably just always return the updated inode attrs on unlink. There's always the possibility that ...
- 04:46 PM Bug #16668: client: nlink count is not maintained correctly
- Ok, I think I sort of get it now. Here's my reproducer:...
- 04:13 PM Bug #16668: client: nlink count is not maintained correctly
- Successful test -- the lookup after the unlink calls into _do_lookup:...
- 03:41 PM Bug #16668: client: nlink count is not maintained correctly
- Tracked down the problem with the ctime and it appears to be a fairly simple bug in fill_stat(). It was only looking ...
- 01:56 PM Bug #16737: Mounting ceph fs on client leads to kernel crash
- the screenshot does not contain full backtrace. please setup netconsole to get full kernel message
- 10:25 AM Bug #16737 (Resolved): Mounting ceph fs on client leads to kernel crash
- Mounting the cephfs on client side with IO running leads to the client crashing sometimes.
Client version:-
uname... - 01:50 PM Bug #16739 (Resolved): Client::setxattr always sends setxattr request to MDS
- If client has CEPH_CAP_AUTH_EXCL, it can updates xattr locally and marks CEPH_CAP_AUTH_EXCL dirty
- 01:43 PM Support #16738 (Closed): mount.ceph: unknown mount options: rbytes and norbytes
- Ceph: @v10.2.2 (45107e21c568dd033c2f0a3107dec8f0b0e58374)@
Linux Kernel: @4.6.3-300.fc24.x86_64@
Hello,
When t... - 01:32 PM Bug #16610: Jewel: segfault in ObjectCacher::FlusherThread
- https://github.com/ceph/ceph/pull/10304
- 01:32 PM Bug #16610 (Fix Under Review): Jewel: segfault in ObjectCacher::FlusherThread
- 05:07 AM Bug #16610: Jewel: segfault in ObjectCacher::FlusherThread
- Just to keep the full history in this issue, we have understood that the segfault only appears in VM with AMD62xx pro...
- 01:09 PM Cleanup #15923 (Fix Under Review): MDS: remove TMAP2OMAP check and move Objecter into MDSRank
- https://github.com/ceph/ceph/pull/10243
- 01:09 PM Cleanup #16035: Remove "cephfs" CLI
- https://github.com/ceph/ceph/pull/10243
- 11:33 AM Bug #16397: nfsd selinux denials causing knfs tests to fail
- I'm trying to get some clarification of what the application was doing when it got these AVC denials. In the meantime...
- 11:07 AM Bug #16397: nfsd selinux denials causing knfs tests to fail
- Just some notes. It looks like the machine has already been torn down and rebuilt, but the new machine is using the s...
- 05:08 AM Bug #16709: No output for "ceph mds rmfailed 0 --yes-i-really-mean-it" command
- Yes that would be ideal. as of now we cannot be sure if it has been actually removed or not.
- 04:46 AM Bug #16730 (Won't Fix): mds'dump display incomplete
- This is deliberate. "mds dump" dumps a specific filesystem (it defaults to the first one, but a client which is set u...
- 02:34 AM Bug #16730: mds'dump display incomplete
- mds dump" and "fs dump" are repeated,and "mds dump" display incomplete.
so delete "mds dump" I think is the best cho... - 01:32 AM Bug #16730 (Won't Fix): mds'dump display incomplete
- create "cephfs&&leadorfs2" fs when run "create fs flag set enable_multiple"...
07/18/2016
- 10:29 PM Bug #16397 (New): nfsd selinux denials causing knfs tests to fail
- Oh dear, this is happening again:
http://pulpito.ceph.com/teuthology-2016-07-13_02:25:02-knfs-jewel-testing-basic-... - 08:20 PM Cleanup #16035: Remove "cephfs" CLI
- For additional info, quoting Sage from an internal RH bug (sorry this is restricted, not sure why. https://bugzilla.r...
- 03:18 PM Cleanup #16035: Remove "cephfs" CLI
- (Agreed with merging "ceph-fs-common" into "ceph-common". I've never found an explanation for why that was its own pa...
- 02:25 PM Cleanup #16035: Remove "cephfs" CLI
- After the cephfs tool is dropped, mount.ceph will be the only thing remaining in the (deb-only) "ceph-fs-common" pack...
- 08:19 PM Bug #16691: sepia LRC lost directories
- Well, I checked the code again and the tmap2omap path looks appropriately durable.
I did notice one thing that hel... - 01:53 PM Bug #16691: sepia LRC lost directories
- Plan is for greg to look into the TMAP2OMAP OSD code to look for what might have causd that.
Afterwards John+Doug ... - 07:39 PM Bug #16709: No output for "ceph mds rmfailed 0 --yes-i-really-mean-it" command
- Without looking at the code, I would imagine that you're seeing EINVAL for rank 1 because there is no such rank (so i...
- 12:28 PM Bug #16709 (Resolved): No output for "ceph mds rmfailed 0 --yes-i-really-mean-it" command
- there is no output for the command ceph mds rmfailed 0 --yes-i-really-mean-it. The command is successful how many eve...
- 04:43 PM Bug #16592: Jewel: monitor asserts on "mon/MDSMonitor.cc: 2796: FAILED assert(info.state == MDSMa...
- Dzianis reported that he upgraded to 10.2.2 without ever upgrading to 10.2.0 (and downgrading after, if that's even p...
- 04:42 PM Bug #16592: Jewel: monitor asserts on "mon/MDSMonitor.cc: 2796: FAILED assert(info.state == MDSMa...
- I make (maybe wrong, but no way back) one-shot upgrade: stop all client, stop all ceph daemons (mds,osd,mon) and run ...
07/16/2016
- 05:08 PM Bug #16592: Jewel: monitor asserts on "mon/MDSMonitor.cc: 2796: FAILED assert(info.state == MDSMa...
- PR for added assertions: https://github.com/ceph/ceph/pull/10316
07/15/2016
- 07:18 PM Bug #16668: client: nlink count is not maintained correctly
- I set up a ganesha + ceph test rig today and was able to reproduce the problem. Interestingly, it does not reproduce ...
- 04:24 PM Bug #16592: Jewel: monitor asserts on "mon/MDSMonitor.cc: 2796: FAILED assert(info.state == MDSMa...
- So, rambling brain dump of my current thoughts on this:
I haven't been able to reproduce this problem. There are t... - 03:19 PM Backport #16697 (Fix Under Review): jewel: ceph-fuse is not linked to libtcmalloc
- PR for jewel is https://github.com/ceph/ceph/pull/10303
- 09:36 AM Backport #16697 (Resolved): jewel: ceph-fuse is not linked to libtcmalloc
- https://github.com/ceph/ceph/pull/10303
- 03:28 AM Bug #16655: ceph-fuse is not linked to libtcmalloc
- https://github.com/ceph/ceph/pull/10303
- 02:25 AM Bug #16691: sepia LRC lost directories
- what do you mean they are old? what does 'rados stat xxxx' show?
07/14/2016
- 11:33 PM Documentation #16664 (Resolved): Standby Replay configuration doc is wrong
- 04:12 PM Documentation #16664: Standby Replay configuration doc is wrong
- Backport: https://github.com/ceph/ceph/pull/10298
I can't mark this issue as Resolved for some reason. - 11:18 PM Bug #16655: ceph-fuse is not linked to libtcmalloc
- tcmalloc is also missing from @ldd /usr/bin/ceph-fuse@ in ceph-fuse-0.94.7-0.el7, FYI, so this has gone on for quite ...
- 01:39 PM Bug #16655 (Pending Backport): ceph-fuse is not linked to libtcmalloc
- 09:48 PM Bug #16691 (Resolved): sepia LRC lost directories
- If you log in to the sepia long-running cluster, it has 37 directories whose objects it lost.
I spot-checked one o... - 05:39 PM Bug #16640 (New): libcephfs: Java bindings failing to load on CentOS
- Let's leave this open to work out if there is a change to the build we can make to avoid the java bindings requiring ...
- 04:01 PM Bug #16640: libcephfs: Java bindings failing to load on CentOS
- Noah, John, I'm guessing the Java bindings ought to link to the versioned libcephfs_jni.so.1.0.0 instead of the unver...
- 05:38 PM Feature #4139 (Resolved): MDS: forward scrub: add scrub_stamp infrastructure and a function to sc...
- I think Greg meant to mark this Resolved.
- 01:11 AM Feature #4139: MDS: forward scrub: add scrub_stamp infrastructure and a function to scrub a singl...
- This bit has been done forever: we have admin socket interfaces to scrub a dentry or recursive folder.
- 01:44 PM Bug #16610: Jewel: segfault in ObjectCacher::FlusherThread
- looks like ObjectCacher::bh_write_adjacencies() passed an empty list to ObjectCacher::bh_write_scattered(). Maybe the...
- 01:25 PM Bug #16668: client: nlink count is not maintained correctly
- It also occurred to me yesterday that I was using the path-based calls, whereas ganesha would likely be using the ll ...
- 10:39 AM Bug #8255 (Resolved): mds: directory with missing object cannot be removed
- This kind of issue should be handled cleanly (MDS will raise 'damaged' health alert, specifics in "damage ls") as of ...
- 01:14 AM Feature #12275 (Duplicate): Handle metadata migration during forward scrub
- #4143 and #4144
- 01:03 AM Feature #12141: cephfs-data-scan: File size correction from backward scan
- This was discussed elsewhere, but we need to be able to disable file size correction as well – via a config option at...
07/13/2016
- 11:37 PM Bug #13271 (Resolved): Missing dentry in cache when doing readdirs under cache pressure (?????s i...
- Zheng fixed this.
- 11:28 PM Feature #14427: qa: run snapshot tests under thrashing
- https://github.com/ceph/ceph/pull/9955 improves snapshots and https://github.com/ceph/ceph-qa-suite/pull/1073 enables...
- 11:25 PM Bug #10834 (Closed): SAMBA VFS module: Timestamps revert back to 01-01-1970
- Closing in favor of #16679, since this is really about birthtime and we're adding a real one.
- 11:25 PM Bug #16679 (New): Samba: hook up to birthtime correctly
- https://github.com/ceph/ceph/pull/9965 is adding birthtime to Ceph internally. Once done, we need to plug samba in to...
- 11:20 PM Feature #12671: Enforce cache limit during dirfrag load during open_ino (during rejoin)
- If we do #13688, we probably won't need this one or can put it off.
- 11:17 PM Fix #5268 (New): mds: fix/clean up file size/mtime recovery code
- 11:15 PM Bug #15379 (Closed): ceph mds continiously crashes and going into laggy state (stray purging prob...
- We have open tickets about improving purge, and the specific issue here seems to have been addressed.
- 11:09 PM Feature #3314: client: client interfaces should take a set of group ids
- This is a natural part of what I'm already doing for #16367.
- 11:07 PM Bug #8090 (New): multimds: mds crash in check_rstats
- 05:28 AM Bug #8090: multimds: mds crash in check_rstats
- There may no longer be an issue now that #8094 is resolved?
- 11:06 PM Feature #7321 (Duplicate): qa: multimds thrasher
- #10792
- 10:49 PM Bug #16668 (In Progress): client: nlink count is not maintained correctly
- Noted on irc that cap handling that involves the root directory (so, anything in root and frequently things in its im...
- 06:21 PM Bug #16668: client: nlink count is not maintained correctly
- I rolled up a testcase for this:...
- 12:44 PM Bug #16668: client: nlink count is not maintained correctly
- MDS revokes CEPH_CAP_LINK_EXCL when unlinking files. It's odd, but I can't see how does it cause problem
- 10:47 PM Feature #10498 (New): ObjectCacher: order wakeups when write calls block on throttling
- 10:38 PM Feature #15393 (Resolved): ceph-fuse: Request for logrotate for client side log files
- ceph-fuse was included in Jewel! commit:98744fdf9bda9d3b14bbf7f528f05ba50a923f97
- 10:34 PM Feature #10060: uclient: warn about stuck cap flushes
- This should be pretty simple by looking at each session->flushing_caps_tids!
- 10:30 PM Feature #6511 (Rejected): MDS: add special purging options for testing
- This is kind of vague now and will get caught up in our future purge fixes anyway.
- 10:23 PM Feature #15067 (Resolved): mon: client: multifs: enable clients to map a filesystem name to a FSCID
- 10:22 PM Feature #15068 (In Progress): fsck: multifs: enable repair tools to read from one filesystem and ...
- I think Doug is working on this as well as #15069?
(Reset if not.) - 10:15 PM Bug #16640 (Resolved): libcephfs: Java bindings failing to load on CentOS
- 12:01 PM Bug #16640: libcephfs: Java bindings failing to load on CentOS
- I suppose the convention of putting the unversioned libraries into -dev packages is based on the idea that built code...
- 09:57 PM Feature #6290 (Resolved): Journaler: warn and shut down if we hit end of journal too early
- Looks like this got fixed in our Journaler refactor.
- 09:50 PM Feature #16676: flush dirty data to journal on SIGTERM
- We sort of assume that there's a standby and the client who will replay the op, but if we've lost the client it's (al...
- 09:36 PM Feature #16676 (New): flush dirty data to journal on SIGTERM
- When it receives SIGTERM, the MDS should commit unsafe data to its journal before terminating.
- 03:36 PM Feature #15615 (Fix Under Review): CephFSVolumeClient: List authorized IDs by share
- https://github.com/ceph/ceph/pull/9864
https://github.com/ceph/ceph-qa-suite/pull/1080 - 03:31 PM Feature #15406 (Fix Under Review): Add versioning to CephFSVolumeClient interface
- https://github.com/ceph/ceph/pull/9864
https://github.com/ceph/ceph-qa-suite/pull/1080 - 06:13 AM Bug #16610: Jewel: segfault in ObjectCacher::FlusherThread
- Alas, the gdb log does not give us much more to go on.
Thread 1 (Thread 0x7f891cdfa700 (LWP 5467)):
#0 0x00007f8... - 05:48 AM Cleanup #13868 (Resolved): mds: MDCache::cap_import_paths is never used
- This member no longer turns up when grepping.
- 05:43 AM Bug #7206 (Can't reproduce): Ceph MDS Hang on hadoop workloads
- If this was a time issue, we fixed a bunch of weird stuff in the switch to solely client-directed mtime updates.
- 05:42 AM Bug #6458 (Can't reproduce): journaler: journal too short during replay
- The journal format is different now too; this is probably not useful any more.
- 05:39 AM Bug #8405: multimds: FAILED assert(dir->is_frozen_tree_root())
- I don't think we've run many multi-mds tests in a while so this is probably still an issue?
- 05:28 AM Fix #8094 (Resolved): MDS: be accurate about stats in check_rstats
- Zheng fixed this ages ago.
- 05:26 AM Bug #10996 (Can't reproduce): dumpling MDS: failed MDLog assert
- Dumpling is old and we don't seem to have seen the error again.
- 05:18 AM Bug #14641 (Duplicate): don't let users specify 0 on stripe count or object size
- 05:10 AM Bug #8255: mds: directory with missing object cannot be removed
- John, much of this is handled now with the metadata damaged flags. What's left?
- 01:12 AM Feature #13688: mds: performance: journal inodes with capabilities to limit rejoin time on failover
- This might have already been done...Zheng, maybe?
- 12:57 AM Cleanup #3677 (Closed): libcephfs, mds: test creation/addition of data pools, create policy
- Things have changed a lot and we definitely test adding multiple data pools now.
- 12:51 AM Bug #4023: kclient: d_revalidate is abusing d_parent
- Is this still an issue?
- 12:50 AM Bug #6770: ceph fscache: write file more than a page size to orignal file cause cachfiles bug on EOF
- fscache has been through a lot of changes; anybody know if this is still a problem?
- 12:41 AM Bug #5950 (Rejected): kcephfs: cephfs set_layout -p 4 gets EINVAL
- I think we actually got rid of the cephfs tool at last.
- 12:20 AM Bug #7685 (Can't reproduce): hung/failed teuthology test: cfuse_workunit_misc
- 12:19 AM Bug #11294: samba: DISCONNECTED inode warning
- This doesn't look anything like #11835 to me; I've not been tracking closely enough to know if we're still seeing han...
- 12:01 AM Bug #12895: Failure in TestClusterFull.test_barrier
- Is this still a problem? It looks to me like the code is still there but I don't think the test has been failing.
07/12/2016
- 11:52 PM Bug #5360 (Rejected): ceph-fuse: failing smbtorture tests
- We have other tickets about smbtorture but we also fixed a bunch; who knows which one this was.
- 11:51 PM Feature #4906 (Resolved): ceph-fuse: use the Preforker class
- See auto-associated revision 66f0704c; this got done years ago.
- 11:48 PM Bug #5731 (Can't reproduce): failed pjd link permissions check
- So much stuff has changed and we haven't linked any other failures to this ticket.
- 11:43 PM Bug #11499 (Can't reproduce): ceph-fuse: don't try and remount during shutdown
- We haven't seen this again.
- 11:42 PM Fix #13126 (Resolved): qa: ceph-fuse flushes very slowly in some workunits
- I spot-checked one wow that the ObjectCacher is coalescing IOs to a single object. It looks like things have gotten f...
- 11:32 PM Bug #14735 (Resolved): ceph-fuse does not mount at boot on Debian Jessie
- I don't think there are likely to be any more infernalis releases now that Jewel is out.
- 11:28 PM Feature #16467: ceph-fuse: Exclude ceph.* xattr namespace in listxattr
- This applies to the kernel client as well, right?
- 11:16 PM Feature #15634 (Resolved): Enable fuse_use_invalidate_cb by default
- This got merged beginning of June.
- 09:33 PM Bug #16668: client: nlink count is not maintained correctly
- I suspect the kclient has a similar problem. I'll test it out when I get a chance. I do agree that we probably ought ...
- 08:53 PM Bug #16668 (Resolved): client: nlink count is not maintained correctly
- Frank reported in #ceph-devel that we don't seem to update nlink correctly from the Client. Looking through the sourc...
- 08:31 PM Bug #16655: ceph-fuse is not linked to libtcmalloc
- Okay, I guess it was just introduced in some autotools refactor or update then. Thanks!
- 08:23 PM Bug #16655: ceph-fuse is not linked to libtcmalloc
- Greg Farnum wrote:
> But the fix is on top of master, which should already work?
It looks like the bug is in t... - 05:47 PM Bug #16655: ceph-fuse is not linked to libtcmalloc
- The link is definitely missing in v10.2.2....
- 05:35 PM Bug #16655: ceph-fuse is not linked to libtcmalloc
- I'm a little confused about the cause here. Ken says
>confirmed that /usr/bin/ceph-fuse is linked to libtcmalloc.so.... - 10:12 AM Bug #16655: ceph-fuse is not linked to libtcmalloc
- https://github.com/ceph/ceph/pull/10258
- 08:27 AM Bug #16655 (Fix Under Review): ceph-fuse is not linked to libtcmalloc
- 03:43 AM Bug #16655 (Resolved): ceph-fuse is not linked to libtcmalloc
- For ceph-fuse binary at http://download.ceph.com/rpm-jewel/el7/x86_64/ceph-fuse-10.2.2-0.el7.x86_64.rpm
[root@zh... - 08:29 PM Documentation #16664 (Fix Under Review): Standby Replay configuration doc is wrong
- 07:46 PM Documentation #16664: Standby Replay configuration doc is wrong
- PR: https://github.com/ceph/ceph/pull/10268
- 07:33 PM Documentation #16664 (Resolved): Standby Replay configuration doc is wrong
- The config settings here are wrong:
http://docs.ceph.com/docs/master/cephfs/standby/
The settings should be pre... - 04:09 PM Bug #16640: libcephfs: Java bindings failing to load on CentOS
- I saw this before with Debian. It looks like it's now showing with with rhelish stuff. The non-devel package includes...
- 11:25 AM Bug #16610: Jewel: segfault in ObjectCacher::FlusherThread
- Just another update after further investigation and discussion in the mailing list.
1. I have tried to run the app... - 10:42 AM Bug #16643 (Won't Fix): MDS memory leak in hammer integration testing
- MDS leaks are ignored by default in the valgrind task, so presumably you're only seeing this because something else f...
- 10:32 AM Feature #16656: mount.ceph: enable consumption of ceph keyring files
- I'll go ahead and grab this one. Not a high priority but definitely a nice-to-have from a usability perspective.
- 09:42 AM Feature #16656 (Resolved): mount.ceph: enable consumption of ceph keyring files
- Jeff pointed this out in doc review:
> we really ought to fix up the mount helper to use the same sort of keyring ... - 09:46 AM Feature #16570 (Fix Under Review): MDS health warning for failure to enforce cache size limit
- https://github.com/ceph/ceph/pull/10245
07/11/2016
- 06:46 PM Bug #16610: Jewel: segfault in ObjectCacher::FlusherThread
- Just as a quick update, we're waiting on some more information from Goncalo concerning the possibility of nodes runni...
- 01:54 PM Feature #15619 (In Progress): Repair InoTable during forward scrub
07/09/2016
- 07:45 AM Bug #16643 (Won't Fix): MDS memory leak in hammer integration testing
- Lots of Leak_PossiblyLost and Leak_DefinitelyLost in
smithfarm@teuthology:/a/smithfarm-2016-07-08_15:27:38-fs-ham...
07/08/2016
- 10:36 PM Feature #16419: add statx-like interface to libcephfs
- Possibly. The thing is that the btime should only ever change due to an deliberate setattr call. It's unlike the othe...
- 10:20 PM Feature #16419: add statx-like interface to libcephfs
- We need to be able to serve an accurate btime. I suppose we could break our rules and assume it won't get changed in ...
- 09:54 PM Feature #16419: add statx-like interface to libcephfs
- Aside from the stuff Greg noticed in his latest review pass, I noticed a number of flaws in the original patchset and...
- 08:02 PM Feature #16419: add statx-like interface to libcephfs
- Changing the description since this has ballooned a bit in scope. We want to add btime support and a change_attribute...
- 10:21 PM Bug #16640 (Won't Fix): libcephfs: Java bindings failing to load on CentOS
- http://qa-proxy.ceph.com/teuthology/jspray-2016-07-08_05:19:56-fs-master-distro-basic-mira/302088/teuthology.log
<... - 12:49 PM Feature #16631 (New): ObjectCacher cache size stats for ceph-fuse
- Currently the perf stats from ObjectCacher don't include the actual size of the cache (get_stat_clean, get_stat_dirty...
- 08:26 AM Bug #16588 (Fix Under Review): ceph mds dump show incorrect number of metadata pools.
- https://github.com/ceph/ceph/pull/10202
- 07:28 AM Backport #16625 (In Progress): jewel: Failing file operations on kernel based cephfs mount point ...
- 07:18 AM Backport #16625 (Resolved): jewel: Failing file operations on kernel based cephfs mount point lea...
- https://github.com/ceph/ceph/pull/10199
- 07:27 AM Backport #16626 (In Progress): hammer: Failing file operations on kernel based cephfs mount point...
- 07:18 AM Backport #16626 (Resolved): hammer: Failing file operations on kernel based cephfs mount point le...
- https://github.com/ceph/ceph/pull/10198
- 07:06 AM Bug #16013: Failing file operations on kernel based cephfs mount point leaves unaccessible file b...
- *master PR*: https://github.com/ceph/ceph/pull/8778
- 07:05 AM Bug #16013 (Pending Backport): Failing file operations on kernel based cephfs mount point leaves ...
07/07/2016
- 09:53 PM Backport #16621 (Resolved): jewel: mds: `session evict` tell command blocks forever with async me...
- https://github.com/ceph/ceph/pull/10501
- 09:53 PM Backport #16620 (Resolved): jewel: Fix shutting down mds timed-out due to deadlock
- https://github.com/ceph/ceph/pull/10500
- 08:58 PM Bug #16592: Jewel: monitor asserts on "mon/MDSMonitor.cc: 2796: FAILED assert(info.state == MDSMa...
- Should note that this is maybe related to: http://tracker.ceph.com/issues/15591
- 05:44 PM Bug #16610: Jewel: segfault in ObjectCacher::FlusherThread
- Log is now here: /ceph/post/i16610/client.log
- 02:04 PM Bug #16610 (Resolved): Jewel: segfault in ObjectCacher::FlusherThread
- ...
- 03:10 PM Feature #15942: MDS: use FULL_TRY Objecter flag instead of relying on an exemption from full chec...
- Related: https://github.com/ceph/ceph/pull/9087
- 03:09 PM Cleanup #16144 (Resolved): Remove cephfs-data-scan tmap_upgrade
- 03:08 PM Cleanup #16195 (In Progress): mds: Don't spam log with standby_replay_restart messages
- 03:05 PM Bug #16288 (Pending Backport): mds: `session evict` tell command blocks forever with async messen...
- 03:04 PM Bug #16396 (Pending Backport): Fix shutting down mds timed-out due to deadlock
- 01:05 PM Feature #16570 (In Progress): MDS health warning for failure to enforce cache size limit
- 01:04 PM Bug #15485 (Duplicate): drop /usr/bin/cephfs
- 11:28 AM Bug #16588: ceph mds dump show incorrect number of metadata pools.
- h3. original description
Ceph mds dump shows metadata pool count as 2, even though only one metadata pool is prese... - 08:54 AM Bug #16588: ceph mds dump show incorrect number of metadata pools.
- Hi Xiaoxi,
You are right about the bug. The metadata_pool field should be left blank. I have changed the descripti... - 08:48 AM Bug #16588: ceph mds dump show incorrect number of metadata pools.
- Rohith Radhakrishnan wrote:
> Ceph mds dump shows metadata_pool id as 0. When no FS is present, then metadata_pool ... - 08:34 AM Bug #16588: ceph mds dump show incorrect number of metadata pools.
- Hmm, yes, this is because metadata_pool is initialized to 0 , this seems worth to fix.
The bug is , when no FS pr... - 06:50 AM Bug #16588: ceph mds dump show incorrect number of metadata pools.
- ceph osd pool stats
*there are no pools!*
ems@rack2-client-3:~$ ceph mds dump
dumped fsmap epoch 3
fs_name ceph... - 06:42 AM Bug #16588: ceph mds dump show incorrect number of metadata pools.
- on what basis is the pool id generated? There are no existing pools. So shouldn't the count start with 0 or 1?
Als...
07/06/2016
- 03:33 PM Feature #15406 (In Progress): Add versioning to CephFSVolumeClient interface
- 06:29 AM Bug #16588 (Rejected): ceph mds dump show incorrect number of metadata pools.
- This is not a bug.
The numbers following "data_pools" and "metadata_pool" are not count, but the pool ids.
root... - 03:44 AM Bug #16588: ceph mds dump show incorrect number of metadata pools.
07/05/2016
- 09:00 PM Bug #16042 (Fix Under Review): MDS Deadlock on shutdown active rank while busy with metadata IO
- PR: https://github.com/ceph/ceph/pull/10142
- 05:44 PM Bug #16592 (Need More Info): Jewel: monitor asserts on "mon/MDSMonitor.cc: 2796: FAILED assert(in...
- We've seen a few reports on the ceph-user mailing lists of the latest jewel....
- 11:42 AM Bug #16588 (Resolved): ceph mds dump show incorrect number of metadata pools.
- Ceph mds dump shows metadata_pool id as 0. When no FS is present, then metadata_pool id should be left blank.
ceph...
07/02/2016
- 07:48 AM Backport #16320 (In Progress): jewel: fs: fuse mounted file systems fails SAMBA CTDB ping_pong rw...
- 07:35 AM Backport #16313 (In Progress): jewel: client: FAILED assert(root_ancestor->qtree == __null)
- 07:31 AM Backport #16215 (In Progress): jewel: client: crash in unmount when fuse_use_invalidate_cb is ena...
- 07:29 AM Backport #16515 (In Progress): jewel: Session::check_access() is buggy
- 07:26 AM Backport #16560 (In Progress): jewel: mds: enforce a dirfrag limit on entries
- 07:22 AM Backport #16037: jewel: MDSMonitor::check_subs() is very buggy
- QA suite backported in https://github.com/ceph/ceph-qa-suite/pull/1075
- 07:11 AM Backport #16037 (In Progress): jewel: MDSMonitor::check_subs() is very buggy
07/01/2016
- 08:28 PM Feature #15069 (In Progress): MDS: multifs: enable two filesystems to point to same pools if one ...
- 08:19 PM Cleanup #16144 (Fix Under Review): Remove cephfs-data-scan tmap_upgrade
- https://github.com/ceph/ceph/pull/10100
- 07:39 PM Cleanup #16144 (In Progress): Remove cephfs-data-scan tmap_upgrade
- 12:07 PM Bug #16556: LibCephFS.InterProcessLocking failing on master and jewel
- Thanks Kefu, I guess the lockdep one is either a cephfs or msgr issue so we'll keep this ticket open to look into it.
- 03:39 AM Bug #16556: LibCephFS.InterProcessLocking failing on master and jewel
- LibCephFS.Fchown is fixed by https://github.com/ceph/ceph/pull/10081,
but we still have... - 11:11 AM Feature #15066: multifs: Allow filesystems to be assigned RADOS namespace as well as pool for met...
- Just in case I lose it, the draft code for splitting messengers was here: https://github.com/jcsp/ceph/tree/wip-15399...
- 10:48 AM Feature #16570 (Resolved): MDS health warning for failure to enforce cache size limit
This can have many causes, but it is a sign that something is not wrong, and a possible precursor to the MDS dying ...
06/30/2016
- 08:47 PM Bug #16042: MDS Deadlock on shutdown active rank while busy with metadata IO
- I'm able to reproduce this with vstart.sh and `cp -a /usr ...`. I'm seeing this every 10 seconds:...
- 04:35 PM Backport #16560 (Resolved): jewel: mds: enforce a dirfrag limit on entries
- https://github.com/ceph/ceph/pull/10104
- 02:21 PM Bug #16556: LibCephFS.InterProcessLocking failing on master and jewel
- Jeff points out that we can also get it to blow up with just a passing test like bin/ceph_test_libcephfs --gtest_fil...
- 01:53 PM Bug #16556: LibCephFS.InterProcessLocking failing on master and jewel
- ...
- 01:51 PM Bug #16556 (New): LibCephFS.InterProcessLocking failing on master and jewel
- Maybe related to https://github.com/ceph/ceph/pull/9995 ?
Failures on master here: http://pulpito.ceph.com/jspray-... - 02:19 PM Bug #16186: kclient: drops requests without poking system calls on reconnect
- Ahh, the reason I could reproduce this yesterday is because the client box was running a v4.5 kernel. With a v4.7-rc5...
- 11:47 AM Bug #16164 (Pending Backport): mds: enforce a dirfrag limit on entries
06/29/2016
- 06:26 PM Support #16528: Stuck with CephFS with 1M files in one dir
- Thank you!
Raised "mds cache size" to 3M and it took couple of minutes to list this dir. - 05:40 PM Support #16528 (Closed): Stuck with CephFS with 1M files in one dir
- Assuming your MDS server has enough memory (it probably does), turn up the "mds cache size" to a number larger than 1...
- 04:48 PM Support #16528 (Closed): Stuck with CephFS with 1M files in one dir
- I'm pretty much stukc with cephfs (jewel 10.2.2) with 1 million 0 byte files in one dir left behind from unsuccessful...
- 06:24 PM Bug #16186: kclient: drops requests without poking system calls on reconnect
- The fio threads at this point are all sitting in ceph_get_caps:...
- 05:38 PM Bug #16186: kclient: drops requests without poking system calls on reconnect
- Ok, the mds session evict command definitely did the trick. Once I issued that (while running a fio test in another s...
- 05:38 PM Support #16526: cephfs client side quotas - nfs-ganesha
- How are you evaluating that the quotas are ignored? There isn't any integration, certainly, but the Ceph client libra...
- 02:56 PM Support #16526 (Resolved): cephfs client side quotas - nfs-ganesha
- I am not sure If this is best logged on the nfs-ganesha project or here.
Ceph quotas are configured using virtual ... - 10:45 AM Feature #16523 (Resolved): Assert directory fragmentation is occuring during stress tests
- Currently we enable fragmentation and set a low (100) frag size limit, but nothing actually validates that there is a...
- 06:40 AM Backport #16515 (Resolved): jewel: Session::check_access() is buggy
- https://github.com/ceph/ceph/pull/10105
- 02:10 AM Bug #16358: Session::check_access() is buggy
- Yes, it could happen for normal case (newly created file). We should backport it
- 12:11 AM Bug #16358: Session::check_access() is buggy
- Whoops, yes. Luckily only for users of hard links, but that's good enough reason!
- 12:02 AM Bug #16358 (Pending Backport): Session::check_access() is buggy
- Seems like this could be serious enough to backport (Zheng: this could happen in normal use, right?)
- 01:11 AM Bug #16367 (In Progress): libcephfs: UID parsing breaks root squash (Ganesha FSAL)
- My basic approach here is to just stop automatically setting UID/GID within the Client class code base at all. It cur...
06/28/2016
- 09:53 PM Bug #16358 (Resolved): Session::check_access() is buggy
- 08:55 PM Bug #16407 (Rejected): LibCephFS.UseUnmounted failed
- 01:10 AM Bug #16407: LibCephFS.UseUnmounted failed
- @John Spray, This is my fault, please closed it.
Thanks. - 07:57 PM Feature #11171: Path filtering on "dump cache" asok
- For test, see https://github.com/ceph/ceph-qa-suite/pull/1066
- 02:27 PM Bug #16397 (Can't reproduce): nfsd selinux denials causing knfs tests to fail
- Ok, talked with Bruce (knfsd maintainer) and the SELinux folks and the consensus is that we have no clue as to why th...
- 01:11 PM Bug #16397: nfsd selinux denials causing knfs tests to fail
- Anyway, the first AVC denial is here:
avc: denied { add_name } for pid=22038 comm="rpc.mountd" name="channel" ... - 12:38 PM Bug #16397: nfsd selinux denials causing knfs tests to fail
- Ok, looking at the log, I do see the SELinux denials. I am new teuthology though...
So you have ubuntu boxes that ... - 01:21 PM Cleanup #15923 (In Progress): MDS: remove TMAP2OMAP check and move Objecter into MDSRank
- 01:21 PM Cleanup #16035 (In Progress): Remove "cephfs" CLI
06/27/2016
- 08:01 PM Bug #16288 (In Progress): mds: `session evict` tell command blocks forever with async messenger (...
- Still no reproducer, but
https://github.com/ceph/ceph/pull/9971
may help. - 01:44 PM Bug #16407: LibCephFS.UseUnmounted failed
- Can you update us? Where are you you seeing the issue and is there a new fix PR?
- 09:13 AM Bug #16042: MDS Deadlock on shutdown active rank while busy with metadata IO
- Could it be via following paths to call MDSDaemon::ms_handle_reset() like async msgr?
One mds thread: ... -> Simpl... - 03:44 AM Bug #16186: kclient: drops requests without poking system calls on reconnect
- there is a 'ceph daemon mds.xxx session evict' command, which makes mds close client session. (use 'ceph daemon mds.x...
06/25/2016
- 05:32 PM Bug #16186: kclient: drops requests without poking system calls on reconnect
- Ok, I tried reproducing this by issuing a stat() while outbound traffic from the client was blocked (on a v4.7-rc4 ke...
06/24/2016
- 08:21 PM Bug #16186: kclient: drops requests without poking system calls on reconnect
- I don't suppose we have a way to reproduce this, do we? Maybe drive a lot of MDS ops and continually stop and restart...
- 05:08 PM Feature #11171 (Fix Under Review): Path filtering on "dump cache" asok
- https://github.com/ceph/ceph/pull/9925
- 10:15 AM Bug #16042: MDS Deadlock on shutdown active rank while busy with metadata IO
- Interesting, #16396 is with async messenger (and is probably the issue we're seeing in current master testing), but w...
- 03:12 AM Bug #16042: MDS Deadlock on shutdown active rank while busy with metadata IO
- Hi guys,
Looks like this issue is very similar to this one here: http://tracker.ceph.com/issues/16396 - 10:07 AM Feature #16468 (Resolved): kclient: Exclude ceph.* xattr namespace in listxattr
- See this thread: http://www.spinics.net/lists/ceph-devel/msg30948.html
Some userspaces tools (notably rsync) try t... - 10:06 AM Feature #16467 (New): ceph-fuse: Exclude ceph.* xattr namespace in listxattr
- See this thread: http://www.spinics.net/lists/ceph-devel/msg30948.html
Some userspaces tools (notably rsync) try t...
06/23/2016
- 07:44 PM Bug #16186: kclient: drops requests without poking system calls on reconnect
- Well, if we have unsafe requests the MDS will in fact have committed them (assuming the MDS didn't crash or something...
- 01:53 PM Bug #16186: kclient: drops requests without poking system calls on reconnect
- If the mds has torn down the client's session, then I don't see what can reasonably be done other than to return an e...
- 06:33 PM Bug #16288: mds: `session evict` tell command blocks forever with async messenger (TestVolumeClie...
- Not to take away Doug's thunder, but I gather he's been unable to reproduce it. The AsyncMessenger may have already b...
- 05:44 PM Bug #15921: segfault in cephfs-journal-tool (TestJournalRepair failure)
- As far as I can tell, we don't even have the backtrace of the segfault in either of those logs, and the sha1 isn't av...
- 01:20 PM Bug #16013 (Resolved): Failing file operations on kernel based cephfs mount point leaves unaccess...
- 11:59 AM Bug #16367: libcephfs: UID parsing breaks root squash (Ganesha FSAL)
- I don't know if I should open a new issue for this, but it looks like even with another ID something is still wrong:
... - 04:51 AM Bug #16396: Fix shutting down mds timed-out due to deadlock
- https://github.com/ceph/ceph/pull/9884
06/22/2016
- 09:09 PM Bug #16186: kclient: drops requests without poking system calls on reconnect
- But if we restart requests from scratch, we're dramatically re-ordering them. We can seemingly send files back in tim...
- 09:01 PM Bug #16186: kclient: drops requests without poking system calls on reconnect
- I think it is working the way it is supposed to work.
We skip unsafe requests because the mds already got them and... - 08:59 PM Bug #16407: LibCephFS.UseUnmounted failed
- You appear to have closed your own PR. And generally speaking we pass around negative error numbers, so readdir() is ...
- 08:44 AM Bug #16407: LibCephFS.UseUnmounted failed
- https://github.com/ceph/ceph/pull/9860
- 07:36 AM Bug #16407 (Rejected): LibCephFS.UseUnmounted failed
- 2016-06-22T15:03:06.176 INFO:tasks.workunit.client.0.plana146.stdout:[ RUN ] LibCephFS.StripeUnitGran
2016-06-2... - 08:55 PM Support #16043 (Closed): MDS is crashed
- 07:40 PM Feature #16228: Create teuthology task for Samba ping_pong test
- (Copied from #16417) See Greg's draft https://github.com/gregsfortytwo/ceph-qa-suite/tree/wip-pingpong
- 07:40 PM Feature #16417 (Duplicate): test pingpong on ceph-fuse
- 05:10 PM Feature #16417 (Duplicate): test pingpong on ceph-fuse
- See #12653. We should integrate pingpong into our nightly test suite, to verify consistency on the kernel client and ...
- 06:10 PM Feature #16419: add statx-like interface to libcephfs
- Yeah, that's what I mean. We have ceph_ll_getattr now (afaict), so we need something like a ceph_ll_getattrx (that na...
- 06:01 PM Feature #16419: add statx-like interface to libcephfs
- Jeff Layton wrote:
> What I'm thinking is that we should add something along the lines of what David Howells has pro... - 05:39 PM Feature #16419: add statx-like interface to libcephfs
- What I'm thinking is that we should add something along the lines of what David Howells has proposed for the new stat...
- 05:35 PM Feature #16419 (Resolved): add statx-like interface to libcephfs
- samba, in particular, can make use of the birthtime for an inode. Have ceph track the btime in the inode and provide ...
- 01:01 PM Feature #15615: CephFSVolumeClient: List authorized IDs by share
- https://github.com/ceph/ceph/pull/9864
Also available in: Atom