Activity
From 06/13/2011 to 07/12/2011
07/10/2011
- 02:22 PM Bug #1295 (Resolved): mds: handle rmdir of subtree root
- If we rmdir an empty directory whose dirfrag(s) are exports/subtree roots,
- it gets journaled locally, replay is...
07/08/2011
- 02:26 PM Bug #1167: Fsstress: bad rstat assert
- merged into master in commit:73d98a057bfeaf6f9a87c38f85b199c6fdbbbf04
- 02:21 PM Bug #1167 (Resolved): Fsstress: bad rstat assert
07/07/2011
- 01:11 PM Bug #1289 (Resolved): misdirected mds request results in improper ENOENT during mds rename workunits
- Resolved by commit:3ef1e656e95eacb1d0d8acd530e2cd9cbe08d586.
We had some problems with false ENOENTs on lookups th... - 11:24 AM Bug #1289 (Resolved): misdirected mds request results in improper ENOENT during mds rename workunits
- Detected as #1288, the kclient is doing a lookup on mds0 for "srci2" immediately following a create "srci2" request. ...
07/06/2011
- 04:19 PM Bug #1284 (Can't reproduce): client: leaked inode ref with cfuse_workunit_snaps.yaml
- Seems to be pretty rare. ...
07/05/2011
- 02:40 PM Feature #1276 (Resolved): client: expose mds partition via virtual xattrs
- our options are basically and ioctl or virtual xattrs. virtual xattrs are easier to use and portable across fuse and...
- 02:36 PM Bug #1252 (Resolved): regression: cfuse fails snaptest-2 with 2-MDS system
- commit:e9e3883d0db4cc03183c0432773825f84170c42b
07/01/2011
- 10:15 AM Bug #1252 (Resolved): regression: cfuse fails snaptest-2 with 2-MDS system
- On head of master, it asserts: ...
- 09:00 AM Bug #1239 (Resolved): client: hang on umount after snaps workunit
06/29/2011
- 03:32 PM Bug #1243 (Resolved): inest lock blocks dir create for a long time
- commit:310032ee8128f6417ac302c0f5ecd27c691cbcc7
I haven't been able to figure out why this doesn't impact cfuse, t... - 01:43 PM Bug #1243 (Resolved): inest lock blocks dir create for a long time
- From the mailing list:
steps are:
we mount ceph on /mnt/test/
then create dir "/mnt/test/a/b/"
1) in dir "b" , us... - 09:35 AM Bug #1232 (Closed): client: inode ref leak?
- 09:34 AM Bug #1239 (Resolved): client: hang on umount after snaps workunit
- two inodes still have pins. looks like a ref count leak somewhere....
- 09:11 AM Bug #1238 (Resolved): mds: ~elist crash on untar_snap_rm
06/28/2011
- 03:09 PM Bug #1193 (Resolved): cmds crashes with stacktrace when using --reset-journal and an invalid rank
- Should be fixed by a pair of commits ending in commit:d1fcffad6c038c6712f1d83735e629017829a2fb
- 02:11 PM Bug #1238 (Resolved): mds: ~elist crash on untar_snap_rm
- INFO:teuthology.task.workunit.client.0.out:removed `k/coreutils-8.5/lib/stripslash.c'
INFO:teuthology.task.workunit.... - 12:28 PM Feature #1237 (Resolved): mds caps limit mount to some subdir
- About a year ago the ceph.conf supported something like:...
- 10:51 AM Feature #1236 (Resolved): libceph: set layout via virtual xattrs (libceph/cfuse)
- Apparently there actually is support in cfuse for ioctls, although it's hairy. Given that, we should support them in ...
06/27/2011
06/26/2011
- 10:33 PM Bug #1232 (Closed): client: inode ref leak?
- see ginormous cfuse log on vapre
06/24/2011
- 11:16 AM Bug #1225 (Resolved): client: recallstate removing caps when it shouldn't
- - can't have an inode with no caps
- cap lru needs to align with dentry lru to get reasonable results
06/22/2011
- 10:00 PM Bug #1206: NFS reexport file creation lags 1-3 seconds
- 0.3 seconds for kernel mount... 0.01s for cfuse.. I'm going to work on tracking this down..
- 09:05 PM Bug #1206: NFS reexport file creation lags 1-3 seconds
- ok, but are you saying that a native mount is also .5s for touch? that's no good either!
- 04:59 PM Bug #1206: NFS reexport file creation lags 1-3 seconds
- Ahh I don't seem to have access to close it...
- 04:58 PM Bug #1206: NFS reexport file creation lags 1-3 seconds
- Ahh okay... file create isn't really any slower than on native ceph kernel mount (rather than re-export)... so this i...
- 03:42 PM Bug #1206: NFS reexport file creation lags 1-3 seconds
- That's faster... 0.5s for a touch create.
As a reference local filesystem and another cfs exported over nfs runs sub... - 03:14 PM Bug #1206: NFS reexport file creation lags 1-3 seconds
- It looks like after the mknod nfsd is calling write_inode via commit_metadata() in fs/nfsd/vfs.c. This triggers cap ...
- 04:39 PM Bug #1199 (Resolved): 4-MDS fsstress: remote ino lookup assert
- Pushed to master in commit:bc273957af2224d028626a961fdb1bf3bc153f40 and commit:e8b45732a1e77bf4811335ae7b8c97c72cd342ce.
- 04:37 PM Bug #1178 (Resolved): cfuse doesn't shut down -- Fw cap doesn't flush on one inode
- Pushed to master in commit:787f03cbd7489cb31f5feeec08f84d0b68aa5c0a.
06/21/2011
- 03:53 PM Bug #1195: ceph mds crash on version upgrade
- pushed fix commit:cc644b842261dbeefde804ed999061b8733a9190 to stable branch
- 09:21 AM Bug #1110: mds: ls -l hangs on concurrent writer
> - when the client sends caps back to the MDS, some of them are dirty. The MDS, as a rule, will always make sure...
06/20/2011
- 05:09 PM Bug #1206: NFS reexport file creation lags 1-3 seconds
- If I reexport the ceph filesystem via fuse/nfs, there is *no* lag in creating files.
- 05:02 PM Bug #1206 (Closed): NFS reexport file creation lags 1-3 seconds
- I'm attaching the kernel logs and mds logs for the creation of a file called scale-product/testfoo3 via a touch runni...
- 04:29 PM Bug #1199: 4-MDS fsstress: remote ino lookup assert
- The problem is that if the backtrace for a given inode changed, the AnchorServer didn't overwrite the old Anchor that...
- 03:31 PM Bug #1199: 4-MDS fsstress: remote ino lookup assert
- Okay, this seems to be an anchortable bug. It's getting the last trace that got committed for the inode in question, ...
- 04:28 PM Bug #1200 (Duplicate): 4-MDS fsstress remote ino lookup cycle
- Yep, this is the same as #1199.
- 03:15 PM Bug #1110 (Resolved): mds: ls -l hangs on concurrent writer
- Created a separate task for the request prioritiziation on the osd, #1203
- 09:22 AM Bug #1110: mds: ls -l hangs on concurrent writer
- Okay, so I found one other issue while testing this patch. So there are two reasons this is slow:
- when we revo... - 08:18 AM Bug #1110: mds: ls -l hangs on concurrent writer
- > Indeed: It was at epoch 5 which has a "modified" timestamp of 2011-06-15 22:37:48. This matches exactly the time wh...
06/17/2011
- 03:30 PM Bug #1200: 4-MDS fsstress remote ino lookup cycle
- Looks like this is the same root cause as #1199.
- 03:19 PM Bug #1200 (Duplicate): 4-MDS fsstress remote ino lookup cycle
- Notice how it's missing an expected ino in a dir that's marked complete, and then it just tries to do the lookup agan...
- 03:12 PM Bug #1199 (Resolved): 4-MDS fsstress: remote ino lookup assert
- ...
- 11:04 AM Bug #1178: cfuse doesn't shut down -- Fw cap doesn't flush on one inode
- Will be testing locally, push later. We had to make sure to check the mseq when setting a new auth_cap based on the M...
- 09:41 AM Bug #1195: ceph mds crash on version upgrade
- Oh, you have multiple MDSs.. can you dump the same object for whichever one(s) crashed in
13: (SessionMap::decode... - 09:38 AM Bug #1195: ceph mds crash on version upgrade
- Attached log from mds crash with suggested debugging enabled.
- 09:30 AM Bug #1195: ceph mds crash on version upgrade
- Attached result of above command.
- 09:25 AM Bug #1195: ceph mds crash on version upgrade
- Can you 'rados -p metadata get mds0_sessionmap /tmp/mds0_sessionmap' and attach? I'm curious what is in the object t...
- 09:14 AM Bug #1195: ceph mds crash on version upgrade
- Hmm, Ceph *should* be upgradable to newer versions. These backtraces don't look familiar though and i don't see anyth...
- 08:03 AM Bug #1195 (Resolved): ceph mds crash on version upgrade
- This may just not be something that's handled in Ceph yet, but while trying to upgrade from stable v0.28 to v0.29, I ...
- 08:59 AM Bug #1110: mds: ls -l hangs on concurrent writer
- Sage Weil wrote:
> Can you confirm that it hung when the fs was full? Because that's normal (and unrelated to this ...
06/16/2011
- 03:27 PM Bug #1178: cfuse doesn't shut down -- Fw cap doesn't flush on one inode
- I'm pretty sure the client is getting the right import/export messages, there's just something going wrong in the twi...
- 02:56 PM Bug #1178: cfuse doesn't shut down -- Fw cap doesn't flush on one inode
- oh, interesting. in that case it won't get an IMPORT event because it wasn't exported.. and the client doesn't find ...
- 02:48 PM Bug #1178: cfuse doesn't shut down -- Fw cap doesn't flush on one inode
- Oh, so it's an inode where the auth got changed, and it only has the cap for the new auth but it doesn't think it is ...
- 01:48 PM Bug #1178 (In Progress): cfuse doesn't shut down -- Fw cap doesn't flush on one inode
- Looking at this now.
- 02:11 PM Bug #1114: NFS export extreme slowdown
- I think I've figured out the more general symptom.
File creation via NFS-ontop-of-ceph takes 1-3 seconds per file.
... - 11:29 AM Bug #1110: mds: ls -l hangs on concurrent writer
- Can you confirm that it hung when the fs was full? Because that's normal (and unrelated to this issue); if the FS is...
- 01:42 AM Bug #1110: mds: ls -l hangs on concurrent writer
- > > Otherwise, there's nothing there. If the situation is reproducible, the best would be an mds log generated with ...
- 09:39 AM Bug #1193 (Resolved): cmds crashes with stacktrace when using --reset-journal and an invalid rank
- Here are the two stack traces I noticed that are different....
- 07:56 AM Bug #1170: 2-MDS fsstress: SR fails subtree assert
- I'm seeing this with the 0.29 packages, using Ubuntu Natty and Linux 3.0rc3 built from git. Only some of the crashed ...
06/15/2011
- 08:06 PM Bug #1158: Unfinished freeze hangs fsstress
- Unfortunately, adjust_nested_auth_pins never sees the person who actually grabbed it. The others print out the grabbi...
- 06:51 PM Bug #1158: Unfinished freeze hangs fsstress
- If you can reproduce, you can enable the auth pin set define in mdstypes.h, which tracks who the pinners are.
//#d... - 06:04 PM Bug #1158: Unfinished freeze hangs fsstress
- Well, it's a nested auth pin....
- 04:32 PM Bug #1114: NFS export extreme slowdown
- Here's the time sequence with each x-axis increment being 20 seconds.
Plotted separately to avoid scaling axes and a... - 04:16 PM Bug #1114: NFS export extreme slowdown
- Here's a 20-second interval time evolution of /sys/kernel/debug/ceph/*/bdi/stats
I can go ahead and graph these.. - 06:54 AM Bug #1110: mds: ls -l hangs on concurrent writer
- Sage Weil wrote:
> I do see this:
> 2011-06-10 11:41:04.325925 7f931b987700 mds0.objecter FULL, paused modify 0x7f...
06/14/2011
- 12:25 PM Bug #1110: mds: ls -l hangs on concurrent writer
- Oh! Sorry, I didn't scroll down. :)
I do see this:
2011-06-10 11:41:04.325925 7f931b987700 mds0.objecter FULL,... - 10:08 AM Bug #1110: mds: ls -l hangs on concurrent writer
- Sage Weil wrote:
> Oh, that log is from the 8th, and doesn't include the fix I pushed on the 9th fixing that particu... - 09:39 AM Bug #1110: mds: ls -l hangs on concurrent writer
- Oh, that log is from the 8th, and doesn't include the fix I pushed on the 9th fixing that particular bug. Did you at...
- 01:41 AM Bug #1110: mds: ls -l hangs on concurrent writer
- Sage Weil wrote:
> Andre, any chance you can pass along more information about the hang you were seeing? Log, or th...
06/13/2011
- 10:21 PM Bug #1137 (Can't reproduce): MDS Crash
- If this turns up again, let us know! I suspect it may be related to the rename journaling changes; I'll be testing f...
- 10:19 PM Bug #1110: mds: ls -l hangs on concurrent writer
- Andre, any chance you can pass along more information about the hang you were seeing? Log, or the specific workload?
- 08:57 PM Bug #1084: blogbench won't finish: waiting for Fr cap forever
- It is working great. I haven't see this problem after applying the patches.
- 09:40 AM Bug #1084 (Resolved): blogbench won't finish: waiting for Fr cap forever
- I can't reproduce this on my mds_issue_caps, so I've merged it into master, commit:0f8fbd562cc58351541ac4f60a1dc98bdd...
- 12:00 PM Bug #1183 (Resolved): mds: snaptest-snap-rm-cmp.sh failing
- osd regression, fixed by commit:056a0c5a008fef0924b17c7e24c85cf73b3b390e
- 11:03 AM Bug #1183 (Resolved): mds: snaptest-snap-rm-cmp.sh failing
- this used to work :(
- 10:07 AM Bug #1181 (Can't reproduce): mds: old_inodes crash
- ...
Also available in: Atom