Activity
From 07/20/2010 to 08/18/2010
08/18/2010
- 10:17 PM Revision 824de0d6 (ceph): mds: fix crash when REQSCATTER races with an actual scatter/MIX
- 08:49 PM Revision c8701f29 (ceph): mds: fix null snapflush logic
- We only want to do a null snapflush if we _know_ there isn't another one
coming: that is, there aren't any outstandin... - 08:45 PM Revision f3b323f6 (ceph): mds: don't send null ack in do_snap_update
- (In this case, it's a null flushsnap.)
- 08:44 PM Revision 466c002f (ceph): qa: add snaptest-snap-rm-cmp
- This (usually) reproduced a bug where:
- we write a big file
- snap it
- remove it. this makes the mds cow it.
... - 08:16 PM Revision 0bcdd0b4 (ceph): mds: remove forward-on-nonauth-rdlock behavior
- The problem is that we may be rdlocking items with a different auth than
the main item we are modifying, so forwardin... - 07:57 PM Revision e8f56d32 (ceph): auth: fix skipping of ~/ if $HOME subst fails
- We were breaking out of the loop entirely, and failing to check the rest
of the items in the path list. Doh! - 05:38 PM Revision 28d7a400 (ceph): mds: oops, fix null snapflush fix
- 05:27 PM Revision 625ffd3e (ceph): mds: fix null snapflush inode lookup
- Don't use pick_inode_snap is totally wrong (it depends on the current set
of snaps, etc.).. look up the inode directl... - 05:27 PM Revision 9908c0ea (ceph): mds: clean up {have,get}_inode() definitions
- 05:27 PM Revision 9092d65b (ceph): rgw: fix commit warning.
- size_t is a long unsigned, not a long long unsigned.
- 03:23 PM CephFS Bug #362 (Rejected): mds: rejoin crashes on snaptest-2 workload
- saw two crashes, running commit:c8701f29f0a6f3777c41f8952c054ba4dd41b9d4...
- 01:16 PM Bug #355: snapshots fail to maintain data integrity
- reproduced with commit:5a89b8b2c9e871ece9e01390eac317df8e4b7d86 and added snaptest-snap-rm-cmp.sh
- 01:14 PM rgw Bug #361 (Resolved): keyring must be specified in ceph.conf?
- 12:58 PM rgw Bug #361: keyring must be specified in ceph.conf?
- fixed by commit:e8f56d3294b552ddd1459914f1d3ce4ad9bf2c61 in testing, will go into 0.21.2
- 11:59 AM rgw Bug #361: keyring must be specified in ceph.conf?
- on ballgate0, you can ...
- 11:58 AM rgw Bug #361 (Resolved): keyring must be specified in ceph.conf?
- on ballgate0, i had to add keyring = /etc/ceph/keyring.bin to /etc/ceph/ceph.conf for it to start. otherwise i would...
- 10:26 AM Feature #352 (Resolved): unit tests to check snap parents
- Wrote snapmove.sh. Would like it to be a little more capable than it is but it will catch basic issues.
- 12:01 AM Revision 5a89b8b2 (ceph): qa: add simple snaprealm test script.
08/17/2010
- 11:38 PM Revision 6d0ccd9c (ceph): mds: handle no flushsnap
- We won't get a flushsnap when the client has EXCL/WR caps but no dirty
data. The MDS needs to release the snapped in... - 09:29 PM Revision 25516510 (ceph): qa: add capwb check for case where client doesn't send flushsnap
- 09:29 PM Revision d2205a94 (ceph): qa: snaptest-upchildreams.sh should clean up after itself
- 09:29 PM Revision 9dd2ffd7 (ceph): qa: snaptest-capwb check no snapflush case first
- 09:29 PM Revision 5f9cc443 (ceph): mds: fix rdlock referral on snapped inode to head
- Note: this still needs some fixing, as the head and snapped inodes may be
on different MDSs! - 07:15 PM Revision 60596c75 (ceph): mds: drop x/wrlocks before, rdlocks after sending reply
- This lets us issue the most leases/caps possible. It also ensure we can
issue caps in the snapped namespace when we ... - 07:00 PM Revision caec5d2b (ceph): client: direct requests in snapped namespace based on nonsnap parent
- We fixed snapdir before.. but we need to do the same for the entire snapped
portion of the namespace. - 07:00 PM Revision a4930d9e (ceph): qa: add snaptest-estale.sh
- Verify requests withing snapped namespace are directed to the proper
MDS. We should never get ESTALE, only ENOENT. - 02:30 PM CephFS Bug #356 (Resolved): mds: cap writeback/release race with cow_inode
- fixed by commit:78c759a8bc94870e303aaa73ee3504cbcc9de17c
- 12:48 PM CephFS Bug #356 (In Progress): mds: cap writeback/release race with cow_inode
- This also happens when the client has WR/EXCL caps but no dirty data.. in that case it sends no FLUSHSNAP.
Basical... - 02:26 PM CephFS Bug #360 (Resolved): mds: head/snapped snap_cap linkage may cross mdss
- The snap cap tracking and rdlock referral business may break due to the snapped inode and head inode being on differe...
- 11:13 AM CephFS Bug #29 (New): mds: rfiles underflow
- hit this again on commit:4b6deda44192db6e5a65488ceefa8c0cef2d82df, saved output to fatty:~sage/ceph/src/out.rstat.sna...
- 11:00 AM Feature #352 (In Progress): unit tests to check snap parents
- Going to write these to test my snaprealm changes that I'm doing to go along with root directory snapshots.
- 10:57 AM CephFS Bug #318 (Resolved): Multi-MDS snaptest-2 fails
- passes with cfuse too.
- 10:21 AM CephFS Bug #318: Multi-MDS snaptest-2 fails
- kclient passes as of commit:4b6deda44192db6e5a65488ceefa8c0cef2d82df
- 10:27 AM Feature #359 (Resolved): osd: use new btrfs snapshot ioctls
- This is awaiting the upstream merge so that the ioctl numbers and interface are fixed.
- 10:14 AM Bug #357 (Resolved): write error: Cannot allocate memory
- fixed by commit:fc88b070efbf911a31d825ecc497bd2cf5cf4a39 (rbd: fix client setup error path), pushed to rbd branch.
t... - 12:28 AM CephFS Feature #358 (Rejected): mds: efficient revert to snapshot
- Use case:
Not so lucky user has deleted <think of many GB/TB data> of valueable data. Copying back from the .snap di...
08/16/2010
- 11:45 PM Revision 4b6deda4 (ceph): qa: make snaptest-2 output less ugly
- 11:45 PM Revision 945c62c6 (ceph): mds: make inode first track dn first on rename
- This mirrors the logic in cc8f5ac47c77d1e336e16d8deb024d507e0e8c59. Make
the renamed inode first match the destdn to... - 11:01 PM Revision cc8f5ac4 (ceph): mds: make new inodes follow dn->first
- We can get a dn->first that is greater than the parent dir's seq. Notably,
when we do something like
mkdir foo
mk... - 10:41 PM Revision 7beeecc8 (ceph): mds: Use project_past_parent and pop_projected_snaprealm instead of
- add_past_parent hack.
Adjusted pop_projected to invalidate caches if needed, and removed
project_past_parent's delet... - 10:41 PM Revision ca1d1ede (ceph): mds: create a new sr_t struct to hold versioned data of a SnapRealm.
- 10:41 PM Revision b0a8f6be (ceph): mds: switch all SnapRealm users to make use of snaprealm.
- 10:41 PM Revision 6289b3e4 (ceph): mds: rename SnapRealm::snaprealm to SnapRealm::srnode
- 10:41 PM Revision bb0dcd2c (ceph): mds: add SnapRealm projection methods
- 10:41 PM Revision e4ea064e (ceph): mds: move snaprealm projection into CInode
- 10:41 PM Revision 830b62a9 (ceph): mds: Move SnapRealm::project_past_parent to the CInode and fix users.
- For now, delete the projected version of the inode so it matches (or fixes?)
old behavior. Once users have been fixed... - 10:41 PM Revision d7ac8eee (ceph): mds: use snaprealm projection methods for mksnap
- 10:41 PM Revision b590b4bc (ceph): mds: use snaprealm projection in rmsnap
- 10:41 PM Bug #357: write error: Cannot allocate memory
- It seems to be due to the missing 'secret' option....
- 10:33 PM Bug #357 (Resolved): write error: Cannot allocate memory
- When trying to add a RBD, i got a kernel panic....
- 09:51 PM Revision 064c2486 (ceph): client: correctly direct snapdir requests
- Use real dir (and its caps), not the virtual snapdir.
- 09:51 PM Revision 4e9ae5d9 (ceph): qa: add test that client updates realm children on snap update
- 09:51 PM Revision eb5213ee (ceph): qa: show rm/cp progress on snaptest-2
- 09:51 PM Revision 18cf92ac (ceph): mds: show cdentry authpin debug state
- 09:51 PM Revision 2921e135 (ceph): mds: flush log on cap writeback if !dirty and unstable locks
- The problem is if we revoke caps, nothing is dirty, but we do writeback
because we are adjusting max_size. Then we h... - 06:48 PM Revision ab202695 (ceph): qa: add snap-rm-diff.sh to look for issues with snapshot integrity.
- Currently passes the script, although running these steps manually
(especially with smaller files) fails a fair perce... - 03:29 PM Revision 1a0ab553 (ceph): mds: only kick head on snap rdlock if in SNAP_SYNC state
- We may, e.g., be in the SCAN state, which is totally unrelated to the
head! - 01:56 PM CephFS Bug #351: mds: fix cow on xattrs
- also verify cow_old_inode() doesn't need the same fix
- 12:58 PM CephFS Bug #356 (Resolved): mds: cap writeback/release race with cow_inode
- We cow and set up the snapped inode locks based on whether WR/EXCL caps are issued. If that races with an in-flight ...
- 11:45 AM Bug #355 (Resolved): snapshots fail to maintain data integrity
- I've been noticing snapshots that have corrupted data if you quickly delete the original contents after snapshotting....
- 10:14 AM Bug #354 (Resolved): Detect errors during transactions
- Right now portions of a transaction can fail and the code doesn't check for this, so you can go through and have some...
- 02:49 AM Revision cdb8a986 (ceph): monclient: gracefully error out when given an invalid monmap path
08/15/2010
- 08:13 PM Bug #350 (Resolved): cosd crash when trying a mkfs with a osdmap as "monmap"
- commit:cdb8a98601ca85ddc345eae519c8e8fc25de253f errors out gracefully with 'unable to decode monmap from file xxx' in...
08/14/2010
- 02:23 PM rgw Bug #353 (Resolved): Handle non-ascii filenames
- Yesterday i made a script which uploaded some random files to my S3 gateway just to test it.
In this batch there w... - 03:40 AM Revision 27331997 (ceph): osd: update missing_loc earlier in push
- This just avoids l=-1 from showing up in the logs, makes the logic a bit
cleaner (keeps missing and missing_loc in sy...
08/13/2010
- 10:12 PM Revision e5c09126 (ceph): debian: fix tcmalloc dependency
- Depend on libgoogle-perftools0, not libtcmalloc-minimal0, since we link
against libtcmalloc, not libtcmalloc-minimal.... - 10:02 PM Revision 165c5f27 (ceph): client: fix xattr cap writeback
- 10:02 PM Revision 0be7dcf5 (ceph): client: fix capsnap writeback of uid/gid/mode/xattrs
- Some metadata wasn't making it into CapSnap and was subsequently getting
written back zeroed to the MDS.
Signed-off-... - 04:59 PM Feature #352 (Resolved): unit tests to check snap parents
- Make tests that create snapshots in directory trees and then move directories around while checking that snapshots sh...
- 04:29 PM CephFS Bug #351 (Resolved): mds: fix cow on xattrs
- in MDCache::cow_inode(),...
- 03:02 PM CephFS Bug #349 (Resolved): cfuse fails untar-snap-rm
- This was a problem with non-file metadata not getting stored in the CapSnap on the uclient. Fixed up by commit:0be7d...
- 10:25 AM Bug #350 (Resolved): cosd crash when trying a mkfs with a osdmap as "monmap"
- ...
08/12/2010
- 11:32 PM Revision a5e582df (ceph): Merge branch 'tcmalloc' into unstable
- 11:32 PM Revision 462d3174 (ceph): ceph.spec.in: require, build with tcmalloc
- 11:32 PM Revision 49f3ac46 (ceph): debian: require tcmalloc
- 10:06 PM CephFS Bug #344: cfuse should pass all qa tests
- Passes:
snaptest-1
trivial_sync
dbench
snaptest-2
bonnie++ (maybe, basic logging doesn't have much output there,... - 10:05 PM CephFS Bug #349 (Resolved): cfuse fails untar-snap-rm
- "p: cannot stat `.snap/k/coreutils-8.5/.x-sc_useless_cpp_parens': Input/output error" on a number of files. Don't thi...
- 10:04 PM CephFS Bug #348 (Resolved): cfuse fails fsx test
- + echo '------ running test fsx.sh ------'
------ running test fsx.sh ------
+ pwd
/more/gregf/testing/ceph/qa/tes... - 08:28 PM Revision 9069b8e5 (ceph): mds: only kick head on snap rdlock if auth
- - If we are non-auth, stick with the snap, and the auth will do the
inference.
- If we are auth, the head had better ... - 08:10 PM Revision 755aa4fd (ceph): Merge branch 'testing' into unstable
- 08:10 PM Revision c5e5237f (ceph): filestore: fix throttling on journal, op_queue for parallel and writeahead
- 08:10 PM Revision c3e0915b (ceph): filestore: fix race in op enqueuing to ensure throttle limit is honors
- 08:10 PM Revision bfee1fa3 (ceph): objectstore: fix transaction get_num_bytes
- The separate bytes field isn't used; use encoded bytes count for now.
- 07:05 PM Revision c4b7c97e (ceph): mon: mon_clock_drift_* for consistency
- 06:57 PM Revision 3345c3c5 (ceph): Merge branch 'testing' into unstable
- 06:57 PM Revision 42802008 (ceph): buffer: fix/improve debug
- 06:57 PM Revision dcedfb85 (ceph): osd: improve l_osd_buf accuracy
- 06:57 PM Revision 998d97fc (ceph): osd: whitespace
- 06:57 PM Revision c1da827d (ceph): msg: use RefCountedObject; overload get() only to cast return type
- Signed-off-by: Sage Weil <sage@newdream.net>
- 06:55 PM Revision a3eefd57 (ceph): msgr: don't leak message when sending to a closed connection
- 06:55 PM Revision a406cfab (ceph): osd: log push ops as push, not pull
- 06:55 PM Revision d00f43aa (ceph): logger: fix logger counter reset
- Reset inc keys after each line; reset all keys on full reset.
Signed-off-by: Sage Weil <sage@newdream.net> - 04:44 PM CephFS Bug #30 (In Progress): multimds: slave_request on getattr
- 04:43 PM Feature #138 (Resolved): Try out tcmalloc
- merged by commit:a5e582df85354d57b178697b7ad13a75078edb1e
- 12:48 PM Feature #138: Try out tcmalloc
- I guess Sage is going to handle the packaging once the kernel rbd split is accomplished.
- 12:49 PM CephFS Bug #319: allows snaps in root directory
- I'm reworking SnapRealms a little bit in order to make this work smoothly. Switching them to use the projected_versio...
08/11/2010
- 11:19 PM Revision 9341ddb7 (ceph): Merge branch 'testing' into unstable
- Conflicts:
configure.ac - 09:59 PM Revision 7aa332cd (ceph): ceph v0.21.1
- 08:30 PM Revision 255bf71d (ceph): mon: use elector's epoch
- This fixes a race with successive elections: we may see a new election
(X+1), then get a victory (X). The victory is... - 07:29 PM Revision acd17a59 (ceph): Makefile: include logrotate.conf in tarball
- 05:46 PM Revision a234df0d (ceph): debian: fix update_pbuilder.sh
- 05:37 PM Revision 13defe0a (ceph): mon: rename mon_lease_wiggle_room -> mon_allowed_clock_drift
- 05:36 PM Revision 0e7cfc6b (ceph): mon: backoff clock drift warnings
- 05:05 PM Revision fd080d53 (ceph): osd: write (empty) log, bounds on remove_pg start
- This zeros the log, and the bounds, when we start pg removal. Previously
we just removed the log and didn't write th... - 10:50 AM Tasks #347 (Rejected): set up man2html or similar to get man pages on web site
- 10:37 AM Bug #346 (Resolved): back-off on clock drift errors
- Done in commit:0e7cfc6bebe1d4b1b3a6c2c85fc169c64a75f1e9
- 09:10 AM Bug #346 (Resolved): back-off on clock drift errors
- Avoid filling the logs up with these errors.. some sort of exponential backoff perhaps?
- 10:07 AM rbd Feature #341: libvirt bindings
- The qemu-rbd stuff probably needs to go upstream before the libvirt patches do. That said, we have some patches we'r...
- 10:06 AM Bug #345 (Resolved): OSD crash: PG::read_state
- 10:06 AM Bug #345: OSD crash: PG::read_state
- fixed by commit:fd080d538e9594ed6203b20e2c65a91f5aaae2d4
for any of these that aren't starting, just do 'rmdir /sr... - 01:57 AM Bug #345: OSD crash: PG::read_state
- Checked out the code, it seems you have to specify the full OSD name or * to scrub:...
- 01:48 AM Bug #345: OSD crash: PG::read_state
- I just had the same crash on another osd. This OSD had some troubles with cephx, so i restarted it, then it crashed w...
- 12:24 AM Bug #345 (Resolved): OSD crash: PG::read_state
- This might be a duplicate of #279 but i'm not sure.
This morning i saw that 4 of my 12 OSD's were down (most of th... - 07:12 AM rgw Bug #342: Can't create files larger than 4GB
- Confirmed, tested a 4.4GB and 8.8GB file, both worked fine.
08/10/2010
- 10:02 PM Revision d708a746 (ceph): rgw: fix content length encoding
- 03:58 PM Bug #339 (Can't reproduce): OSD crash: ReplicatedPG::sub_op_modify
- The missing map on the replica apparently showed the object missing.
I audited the primary code and it should reco... - 03:55 PM CephFS Bug #319 (In Progress): allows snaps in root directory
- Been working on this today; I haven't run across any big architectural issues so far, just a lot of asserts and bad a...
- 02:43 PM CephFS Bug #344 (Resolved): cfuse should pass all qa tests
- Right now, it doesn't. Figure out which tests it fails and make them work.
Should probably create a new bug for each... - 02:38 PM rgw Feature #174 (Resolved): Support large files better
- Closing this one, last bug was fixed with commit:d708a746ffd4d75d7502127d2c43d11105f1e484. Was able to upload and dow...
- 11:15 AM rgw Feature #174: Support large files better
- Seems to be an issue with the reported Content-Length when downloading the file:...
- 09:16 AM rgw Feature #174: Support large files better
- I can confirm that, the FastCGI module works fine under Apache.
I uploaded a file of 1.1G and 4.4G (larger then RA... - 02:37 PM rgw Bug #342 (Resolved): Can't create files larger than 4GB
- Fixed with commit:d708a746ffd4d75d7502127d2c43d11105f1e484
- 12:42 PM rgw Bug #342: Can't create files larger than 4GB
- Just tried a 3.5GB file, which went wrong too.
In a "ListBucketResult" the ETag matches the original md5sum and al... - 10:27 AM rgw Bug #342 (Resolved): Can't create files larger than 4GB
- According to wido, when trying to write a 4.4 GB file, he ended up with a 400 MB file.
- 02:10 PM Linux kernel client Feature #42: Resize of rbd image
- resizing hte underlying block device has been possible with ext3 since RHEL 4 (~5 years?), so this is something we sh...
- 10:41 AM rgw Bug #343 (Resolved): Bad 100-Continue handling
- When uploading a file onboth fcgid and fastcgi, the PUT request with the 'Expect: 100-Continue' triggers an automatic...
- 09:06 AM Feature #138: Try out tcmalloc
- I can confirm that tcmalloc() works fine, seeing about 70% memory reduction on my OSD and MDS, great!
Tested it on... - 07:28 AM rbd Feature #341 (Resolved): libvirt bindings
- It might be nice to make some libvirt bindings for qemu-kvm/rbd
libvirt: http://libvirt.org/
We should/could su...
08/09/2010
- 09:17 PM Revision a8786f8a (ceph): monclient: fix memory leak
- 09:17 PM Revision 5e042490 (ceph): conf: fix memory leak
- 09:17 PM Revision 8dd4b46d (ceph): msgr: Fix leak of authorizer
- 09:17 PM Revision 9c18972d (ceph): auth: delete Exception once it's caught; fixing minor memory leak
- 09:17 PM Revision a43897b0 (ceph): mon: delete Subscriptions, don't just clear them. Fixes minor memory leak.
- 09:17 PM Revision c81be3ba (ceph): osd: rename shadowing variable for clarity
- 09:17 PM Revision 6573a1a3 (ceph): osd: put message before shutdown.
- Prevents (admittedly minor) memory leak.
- 09:17 PM Revision f1eb9a87 (ceph): osd: clean up ReplicatedPG leaks and comments
- 09:17 PM Revision 195c3827 (ceph): mds: remove dead code
- 09:17 PM Revision 3ebc768f (ceph): mds: MDCache documentation, and a few missing puts.
- 09:17 PM Revision 744e3722 (ceph): MDS: Migrator leak fixes and documentation.
- 09:17 PM Revision 5cbbf103 (ceph): mds: Server documentation and some leak fixes.
- 09:17 PM Revision 43af88e2 (ceph): MDBalancer: fix a few leaks, add documentation.
- 09:17 PM Revision 0c6c48c0 (ceph): mds: add documentation to Locker, MDS, TableServer, Objecter
- 09:17 PM Revision 062f5312 (ceph): mds: add documentation to MDCache, Server, Locker
- 09:17 PM Revision e5923ac2 (ceph): mds: open_remote_dirfrag: don't lose an op if a remote mds is down, que...
- 09:17 PM Revision c8e7e1a7 (ceph): mds: remove an extra m->put().
- This was previously necessary but handle_client_session() now takes care
of the reference, and it's more uniform with... - 09:17 PM Revision 768eb41e (ceph): osd: remove some very dead code from ReplicatedPG.
- It looks like this used to be part of preprocess_op? It's not enclosed in a function now.
- 09:17 PM Revision 98932693 (ceph): osd: Remove PG::preprocess_op. It hasn't been used for a while now.
- If you bring it back, make sure you fix it for memory leaks!
- 09:17 PM Revision 4916cc13 (ceph): auth: no need to duplicate character arrays, just turn to string implic...
- Fixes small leak.
- 09:17 PM Revision ae6a530b (ceph): auth: delete char array, fixing memory leak.
- 09:17 PM Revision 4cbe2b40 (ceph): filejournal: fix small memory leak
- 09:17 PM Revision 6a3baec3 (ceph): auth: fix memory leak
- 03:36 PM Feature #138: Try out tcmalloc
- Dunno how to set up the packaging stuff, but the configure.ac/Makefile stuff wasn't too complicated and is pushed in ...
- 02:07 PM rgw Feature #174: Support large files better
- Replacing the apache fcgid module with fastcgi seems to solve the problem, as it doesn't buffer the entire uploaded d...
- 09:48 AM rgw Feature #174: Support large files better
- Actually, the ETag calculation is being done in chunks and is ok. The real problem is that fcgid apache module first ...
- 12:46 PM Feature #334: Version in librados
- A ''librados-config'' binary might be usefull here. This is also used by libcurl, libxml2 and PHP.
For example:
... - 08:34 AM Bug #262 (Resolved): osd: make sync work on ext4
- the flush issue is resolved by using a big sync(2) hammer. sucks, but that's currently what it takes. hopefully we'...
08/08/2010
- 03:59 PM Revision 6b93dc97 (ceph): filestore: flush using sync(2) hammer
- Since we can't easily detect ext3 (let alone whether we have data=journal),
by default use sync(2) as an overly large... - 09:51 AM CephFS Feature #340 (Closed): large directories, directory fragmenting
- directory fragments need to be reenabled, and tested.
fragment merging implementation needs to be finished, tested. - 08:48 AM Bug #262: osd: make sync work on ext4
- we could do a quick FREEZE/THAW sequence, but only if we're root.
a simple SYNCFS ioctl would be ideal...
08/07/2010
- 06:22 AM Bug #339 (Can't reproduce): OSD crash: ReplicatedPG::sub_op_modify
- Two OSD's got killed by the OOM killer, after restarting both (osd4 and osd5), one crash with the following message:
...
08/06/2010
- 11:10 PM Revision 9ac243eb (ceph): mds: fix snap lookups in CInode encoding
- 11:10 PM Revision 8778bacf (ceph): qa: remove random ordering from runallonce, for predictable resource ma...
- 10:02 PM Revision fc375915 (ceph): tcmalloc: add support for tcmalloc.
- Right now it's only used on the MDS and OSD, but it can be added to mons too.
- 09:58 PM Revision f5487fd1 (ceph): mds: don't try to recover filelocks if the client didn't encode any
- 09:49 PM Revision fd437244 (ceph): osd: simplify class waiter list
- No reason to break it down by pg_t?
- 09:13 PM Revision 840530e4 (ceph): rgw: make gateway dns name configurable
- 08:42 PM Revision 92665f8b (ceph): client: fill in num_fwd in messages to mds
- 07:49 PM Revision 2cac1668 (ceph): osd: class load shouldn't hang when failed to load
- Fixes bug #332. This happened if either the class existed, and it was either
corrupted, or failed to load from any ot... - 06:35 PM Revision 81008419 (ceph): qa: echo output
- 06:34 PM Revision 89b0f577 (ceph): Merge branch 'snap_cap' into unstable
- 06:34 PM Revision 43638640 (ceph): Merge branch 'testing' into unstable
- 06:00 PM Revision af96ba53 (ceph): fiemap: fix include
- 05:55 PM Revision 185f4e5d (ceph): fiemap: move read_fiemap from rbd.cc to fiemap.cc
- 05:08 PM Revision bce3a630 (ceph): mds: fix send_message_client_counted helper
- Don't use the outgoing message connection (usually there is none!).. use
the connection argument. Der. Broken since... - 04:56 PM Revision ec91ea1d (ceph): mon: add 'mds compat rm_* id' commands to adjust mdsmap compat set
- This'll be helpful if someone inadvertantly starts up a new cmds, the
compat set updates, and then they can't start o... - 02:30 PM Feature #336 (Closed): osd: metadata on tmpfs
- 02:14 PM Feature #336: osd: metadata on tmpfs
- use
journal dio = false
Thanks. Please close. - 10:44 AM Feature #336: osd: metadata on tmpfs
- Jan Smets wrote:
> Having the journal (or even data) on a memory filesystem would be usefull for testing.
>
> Fil... - 12:25 AM Feature #336 (Closed): osd: metadata on tmpfs
- Having the journal (or even data) on a memory filesystem would be usefull for testing.
Filesystem Size ... - 01:54 PM rgw Feature #313 (Resolved): Support loglevel
- both patches were merged in, commit:801150a723aaaba2f53995f1646abb9fabc37df5 and commit:840530e42a7adeb8045777d6cc4db...
- 01:49 PM rgw Feature #337: support bucketname hostname prefix for wildcard dns
- Configuring is being done via the apache configuration and requires mod_env installed, e.g., the following:
SetE... - 01:47 PM rgw Feature #337 (Resolved): support bucketname hostname prefix for wildcard dns
- Done, commit:840530e42a7adeb8045777d6cc4dbfb29344815e.
- 11:07 AM rgw Feature #337 (Resolved): support bucketname hostname prefix for wildcard dns
- ala *.amazonaws.com, or whatever it is.
- 12:20 PM Bug #332 (Resolved): class operations hangs when out of disk space
- Fixed with commit:2cac166825ac080e091789e659f4cb6ae250cc7a.
- 11:08 AM rgw Tasks #338 (Resolved): set up playground apache(s)
- 10:49 AM Bug #248 (Resolved): rbdtool import should use fiemap
- Added new osd ops: mapext, sparse-read. Implemented rbd import, export, copy using fiemap. Latest commit in the serie...
- 10:43 AM rbd Bug #322 (Closed): Can't create an image via rbd or qemu-kvm
- This was a case of class not loaded. AFAIR this was solved, just it took longer than usual for the osds to get the ne...
- 09:57 AM CephFS Bug #316 (Resolved): mds: track dirty snapped metadata without faking a cap
- see commit:ec91ea1db61290cd02a565be1d2076ceb74ba0c1
- 12:19 AM Revision 823c36c2 (ceph): autoconf: detect existence of linux/fiemap.h
08/05/2010
- 11:49 PM Revision d97e42e1 (ceph): mon: automatically incorporate new incompat features into mdsmap
- 11:48 PM Revision 846e979c (ceph): mds: include compatset in beacon message
- 11:48 PM Revision 9cdd4a64 (ceph): mds: add incompat clientranges bit for the format change
- 11:16 PM Revision 7e002127 (ceph): qa: add snap test for cap writeback
- This used to fail/hang before the cap cloning business was fixed up.
- 11:14 PM Revision afb8cdf0 (ceph): rbd: implement copy operation using sparse_read
- 11:05 PM Revision 4a4d6873 (ceph): mds: non-head requests always rdlock as_anon
- This ensures we can't do a snapped GETATTR and rdlock with EXCL on a
directory and miss out on cap metadata still wit... - 10:59 PM Revision af922877 (ceph): rbd: use sparse_read op for export
- 10:58 PM Revision 6f38165d (ceph): osd: implement sparse_read
- 09:52 PM Revision 1fba876f (ceph): mds: write flushsnap cap metadata into old_inode if appropriate
- Don't have a test case for this yet, unfortunately, but it looks
straightforward at least. :) - 09:09 PM Revision acfcfa02 (ceph): mds: ensure that rdlock on snap filelock flushes data
- Do the head filelock sync 'anonymously', so that we can_rdlock(-1) (no
particular client) and thus can't rdlock in th... - 09:01 PM Revision f8333deb (ceph): mds: rdlock head when snapped inode is snap->sync
- 08:08 PM Revision 81ab0a02 (ceph): mds: do not clone caps to snapped inodes
- Instead, explicitly track which locks need to be flushed (via a FLUSHSNAP)
with a LOCK_SNAP_SYNC lock state.
Restruc... - 07:23 PM Revision 2f03347d (ceph): mds: group finishers after eval_cap_gather does all locks
- This will reduce some churn, since we can update all the lock states before
worrying about anyone trying to take more... - 06:36 PM Revision 77ee6dc1 (ceph): config: set osd_min_down_reports to 3.
- 06:09 PM Revision d979e48d (ceph): osd: Adjust failure reporting.
- When a failure report is sent to the mon, the failed OSD is added to
pending_failures. If the OSD gets a heartbeat fr... - 06:09 PM Revision 5a8a2187 (ceph): osd: only access failure_queue while under heartbeat_lock.
- 06:08 PM Revision 83e58257 (ceph): osdmon: Adjust failure reporting.
- MOSDFailure can now be a failure or a not-failure report. If it is a
failure, OSDMon will add it to a map of failure ... - 04:24 PM CephFS Bug #316: mds: track dirty snapped metadata without faking a cap
- incompat bit?
need a way to remove bad (in)compat bits from mdsmap compatset (via teh montior) - 01:11 PM CephFS Bug #316: mds: track dirty snapped metadata without faking a cap
- ok, the cap tracking is done.
remaining issues:
- how to make an OPEN or STAT on the snapped inode trigger the head... - 02:47 PM Revision 7f9cef01 (ceph): makefile: define docdir for old automake versions
- Notably centos5.
Reported-by: Thomas Mueller <thomas@chaschperli.ch>
Signed-off-by: Sage Weil <sage@newdream.net> - 02:44 PM Revision 773dd825 (ceph): monmaptool: man page typo fix
- Signed-off-by: Sage Weil <sage@newdream.net>
- 02:35 PM Linux kernel client Bug #335 (Resolved): xattrs not coherent wrt snaps
- the ceph_queue_cap_snap doesn't include xattrs in the cap_snap struct, so xattrs prior to snap creation are not inclu...
- 11:33 AM Feature #82 (Resolved): mon: osd failure smarts
- Pushed in commit:77ee6dc1cc8e34d0d0be02c90c976058603f78b2.
The OSDMonitor will only mark an OSD down after it gets a... - 10:49 AM Feature #334 (Resolved): Version in librados
- When writing a application which uses librados it could be usefull if you could get the version number out of it some...
- 10:07 AM Linux kernel client Bug #333 (Resolved): dup flushsnap messages
- client appears to be sending dup flushsnap messages
this...
08/04/2010
- 11:45 PM Revision aa5eced1 (ceph): osd: add a mapext operation for mapping extents (fiemap)
- 07:19 PM Revision e2bb2e79 (ceph): rbd: import using fiemap
- still requires some cleanup, but basically works
- 06:40 PM Revision 5b5c0066 (ceph): osdmap: only decode internal addresses if map is new enough version
- 05:37 PM Revision 8c4d2689 (ceph): Merge branch 'testing' into unstable
- 05:01 PM Revision 6f7523a0 (ceph): rados: -i and -o are still there for other commands
- Signed-off-by: Sage Weil <sage@newdream.net>
- 04:56 PM Revision a7038da7 (ceph): ceph.spec.in: fix docdir paths
- _docdir is /usr/share/doc, not /usr/share/doc/ceph (as it is with debian).
- 04:51 PM Revision 1ecda511 (ceph): rados: fix usage
- 04:51 PM Revision 1cf45017 (ceph): ceph.spec.in: fix sample.ceph.conf chown
- Reported-by: Xiaoguang Liu <syslxg@gmail.com>
- 04:51 PM Revision ef711e2e (ceph): osd: fix heartbeat to/from map updates
- - always make _inst match _to and _from (don't adjust in note_down_osd)
- only mark down old hb connection in update_... - 01:03 PM Bug #331 (Closed): OSD crash: OSDMap::Incremental::decode
- Fixed by commit:5b5c0066f1bbfdc8c03cfacffab8969e23377f90. Only applied if you upgraded your machines with an in-place...
- 11:58 AM Bug #331: OSD crash: OSDMap::Incremental::decode
- Fixed by 5b5c0066f1bbfdc8c03cfacffab8969e23377f90
- 11:31 AM Bug #331 (Closed): OSD crash: OSDMap::Incremental::decode
- After upgrading to the latest unstable i'm seeing a OSD crash in my whole cluster (30 OSD's)....
- 12:12 PM Bug #332 (Resolved): class operations hangs when out of disk space
- osds can't load class in if there's not enough space to write down the temporary library file. It should return some ...
- 10:36 AM Bug #330: Crash on OSD::_share_map_outgoing(const entity_inst_t&)
- Fixed (more) by commit:ef711e2eead039b9819b8380f7b1ea6ebd84160d
- 12:59 AM Bug #330: Crash on OSD::_share_map_outgoing(const entity_inst_t&)
- The commit did not work, my OSD's kept crashing.
I place three new coredumps (preserved the timestamp) in the same... - 10:25 AM Bug #247 (Resolved): osd: allow separate interface for internal cluster traffic
- Merged this into unstable.
08/03/2010
- 11:13 PM Revision fa27b31e (ceph): osd: cover map_in_progress by osd_lock instead of an individual lock.
- The separate lock is no longer needed and removing it reduces lock contention
and makes it easier to avoid races else... - 10:21 PM Revision 207a1bdc (ceph): osd: Unlock osd_lock in _dispatch before locking handle_map_lock
- 09:56 PM Revision 595543f8 (ceph): osd: when checking OSD identities via require_same_or_newer_map, use cl...
- 09:56 PM Revision 24989feb (ceph): OSD: use get_cluster_inst instead of get_inst when checking other OSD i...
- 09:56 PM Revision 42c00850 (ceph): osdmap: remove unused versions of get_[cluster_]inst
- 09:56 PM Revision d7d615bc (ceph): osd: use the client_messenger for the logclient.
- 09:56 PM Revision b588a56e (ceph): config: Error out when you can't parse an addr from the config file.
- Also, fix up a few other addr parsing cases to break properly!
- 09:56 PM Revision c26cb727 (ceph): cosd: adjust client/cluster messenger setup to be more consistent
- and poison better
- 09:56 PM Revision dfb85d23 (ceph): osd: adjust map_in_progress guarding
- 09:56 PM Revision c7d23529 (ceph): osd: guard _dispatch to prevent execution of new messages while the map...
- 09:56 PM Revision 2fbf1e9f (ceph): osd: switching dispatch_running to be a counter
- Now it works with multiple messengers dispatching.
- 09:56 PM Revision bc9cb931 (ceph): osdmap: make room in incremental for separate cluster/client OSD addrs.
- 09:56 PM Revision ded7b8e2 (ceph): osd: make new cluster_messenger, client_messenger pointers.
- Don't remove old messenger, just alias it to client_messenger.
- 09:56 PM Revision 7f701120 (ceph): osdmap: add osd_cluster_addr to full map
- Include it in lookups and add get_cluster_inst et al to match get_inst et al
- 09:56 PM Revision abff0990 (ceph): osd: use cluster/client_messenger as appropriate.
- Modify send_incremental_map to choose based on the type of the entity.
- 09:56 PM Revision ef9a4eeb (ceph): osd: remove messenger pointer, fix calls to it in PG and ReplicatedPG
- 09:56 PM Revision ed28d5e6 (ceph): osdmap: Make sure osd_cluster_addr matches osd_addr size on decode
- 09:56 PM Revision 26ff6cb5 (ceph): osdmap: move osd_cluster_addr decoding, and make sure that size matches...
- 09:56 PM Revision 6abf48da (ceph): config: rename g_my_addr to g_public_addr, add g_cluster_addr
- 09:56 PM Revision eb708208 (ceph): msgr: set up bind and Accepter::bind to take in addresses and bind to them
- 09:56 PM Revision 7eeca9b8 (ceph): cosd: create separate cluster/client messengers if separate addresses a...
- 09:56 PM Revision d290a046 (ceph): cosd: move messenger existence checks to where they're useful
- 09:56 PM Revision 382290ef (ceph): messages: add cluster_addr to MOSDBoot, use it in OSD and OSDMonitor
- 09:56 PM Revision d0f7412e (ceph): config: add OPT_ADDR type, move "public_addr"/"cluster_addr" processing...
- Necessary side effect: move public_addr/cluster_addr into g_conf.
- 09:56 PM Revision d6c8f8c8 (ceph): config: correctly notify when OPT_ADDR config options are found.
- 09:56 PM Revision 582f9285 (ceph): osd: register the OSD on cluster_messenger if needed
- 09:56 PM Revision 28b54973 (ceph): osd: be more careful with cluster/client messenger use.
- 09:56 PM Revision cec513ff (ceph): osd: add Mutex and flag to cover execution of handle_map_lock.
- Fixes a crash when the cluster and client messenger both dispatch an
OSDMap and their handling gets intertwined. - 06:07 PM Revision 4ea24d78 (ceph): osd: fix replies to dup/already-committed requests
- We can safely reply if oldv is _older_ than what's on disk. Otherwise the
request ends up on the list and never gets... - 04:43 PM Revision 930ef4e5 (ceph): Merge branch 'testing' into unstable
- 04:42 PM Revision 22a7d42b (ceph): assert: only include acconfig.h ifdef __CEPH__
- Signed-off-by: Sage Weil <sage@newdream.net>
- 04:40 PM Revision 0ffb08c8 (ceph): librados.hpp fix
- [ The following text is in the "UTF-8" character set. ]
[ Your display is set for the "iso-8859-1" character set.... - 04:39 PM Revision 4bcba121 (ceph): osd: fix map sharing with old heartbeat peers
- The addr is in old_inst, since it's an old peer. Fixes a crash like
osd/OSD.cc: In function 'void OSD::_share_map_o... - 04:39 PM Revision be7d84b2 (ceph): mon: fix signed/unsigned compile warning
- 04:17 PM Bug #247: osd: allow separate interface for internal cluster traffic
- Am running the osd_msgrs branch in qa today, unless that finds problems that weren't exposed by a bunch of system sta...
- 03:45 PM CephFS Bug #316: mds: track dirty snapped metadata without faking a cap
- this illustrates the problem:...
- 10:57 AM Feature #138: Try out tcmalloc
- let's turn this on for cmds and cosd.
and update configure.ac to detect it.
and set debian/control and ceph.spec... - 10:56 AM CephFS Feature #321: allow dns names for monitor switch
- see MonClient::create_initial_monmap, and mount/mount.ceph.c for an example of using getaddrinfo()
- 10:52 AM Feature #82: mon: osd failure smarts
- A simple approach would be to index the osd peers/heartbeat graph in the PGMonitor and only mark an osd down when som...
- 10:10 AM Linux kernel client Feature #23 (Resolved): fcntl/flock advisory lock support
- This stuff got pushed into unstable yesterday.
- 09:47 AM Bug #330 (Resolved): Crash on OSD::_share_map_outgoing(const entity_inst_t&)
- fixed by commit:4bcba121ac165939b3a825564d67fbc1f2e74cf1
- 01:46 AM Bug #330: Crash on OSD::_share_map_outgoing(const entity_inst_t&)
- I got the cluster working again, by starting it in the following order:
* Kill monitor and mds
* Start all the OS... - 01:22 AM Bug #330 (Resolved): Crash on OSD::_share_map_outgoing(const entity_inst_t&)
- When upgrading to the latest unstable all my OSD's (30 in total) crashed with the following message:...
- 12:07 AM CephFS Bug #312: MDS crash: LogSegment::try_to_expire(MDS*)
- I just saw this crash again, uploaded the log (mds.0.log.1.gz) and corefile (core.cmds.node13.18006) to *logger.ceph....
08/02/2010
- 11:27 PM Revision 6573635b (ceph): mds: fix cap encoding skipping when snaprealm differs
- We were avoiding adding a new cap, but _any_ issue is problematic, not just
adds (which we can't really be sure about... - 11:01 PM Revision ccdfcd39 (ceph): osdmap: show flags
- Signed-off-by: Sage Weil <sage@newdream.net>
- 10:36 PM Revision d9a97ac4 (ceph): Merge branch 'locks' into unstable
- 10:35 PM Revision aed240ca (ceph): ceph_fs: use __le64 instead of __u64 for over-the-wire data
- 10:35 PM Revision 0b964c3d (ceph): msgr: Add CEPH_FEATURE_FLOCK to default features set, remove extraneous...
- 10:35 PM Revision f30efb28 (ceph): mds: new low-level debug output, and turn down previous debugs
- 08:27 PM Revision 3e42191b (ceph): debian: add ceph-client-tools dep to ceph
- Signed-off-by: Thomas Mueller <thomas@chaschperli.ch>
Signed-off-by: Sage Weil <sage@newdream.net> - 08:26 PM Revision f725f203 (ceph): debian: move utils rbd,rados,csyn to ceph-client-tools
- Signed-off-by: Thomas Mueller <thomas@chaschperli.ch>
Signed-off-by: Sage Weil <sage@newdream.net> - 08:26 PM Revision c61a6c32 (ceph): debian: move mount.ceph to ceph-client-tools
- Signed-off-by: Thomas Mueller <thomas@chaschperli.ch>
Signed-off-by: Sage Weil <sage@newdream.net> - 08:26 PM Revision 3fc47c2e (ceph): Merge branch 'testing' into unstable
- 08:26 PM Revision 9ccf9cb6 (ceph): debian: move fuse-utils dep to ceph-fuse
- Reported-by: Thomas Mueller <thomas@chaschperli.ch>
Signed-off-by: Sage Weil <sage@newdream.net> - 08:23 PM Revision 4a7c3553 (ceph): osd: avoid most of handle_osd_map if there are no new maps in message
- We don't want to take the time to flush things to disk and shut down all
the worker threads if there is nothing new i... - 08:23 PM Revision 9bfb8da9 (ceph): osd: send latest osdmap when we stop sending heartbeats to peer
- Otherwise, they may still expect them, and may mark us down.
- 08:23 PM Revision baa69ad7 (ceph): cclass: fix .so loading file
- The only thing I've found, is that the *.so files are included in the -devel
rpm. This leads to the fact that the 'fi... - 08:14 PM Revision 025bce64 (ceph): osd: send heartbeat request from update_heartbeat_peers
- This keeps our from_stamps in a consistent state, and sends the hb request
sooner rather than later.
Signed-off-by: ... - 08:13 PM Revision 90a2f201 (ceph): mon: make backward-compat mon id's start at '0', not 'a'
- Signed-off-by: Sage Weil <sage@newdream.net>
- 08:13 PM Revision 30cdc3dd (ceph): mon: drop redundant pg_temp mappings
- 07:32 PM Revision 88c17f42 (ceph): osd: add asserts when indexing pg log by reqid
- This should help catch callers that submit dup entries with the same
request id, causing other problems down the line. - 05:39 PM Revision df813369 (ceph): mds: locking. remove zero-valued client held/waiting lock counts from t...
- In the unlikely event that you have a very large cluster doing a lot of
locking of different files amongst the whole ... - 05:39 PM Revision b443f130 (ceph): mds: locking. Keep trap of neighbors and coalesce them
- 05:39 PM Revision 8dd1960b (ceph): mds: locks. Fix possible naughty dereference, and fix existence checks
- 05:39 PM Revision 2cfbddc2 (ceph): mds: Fix ceph_lock_state_t to use uint64_t
- 05:39 PM Revision 0d4bfea5 (ceph): mds: locking. remove ceph_filelock* lists and replace with iter lists
- This allows for more efficient deletion of elements.
- 05:39 PM Revision fa058c41 (ceph): mds: handle file locking in MClientCaps on a reconnect.
- 05:39 PM Revision fe0c5119 (ceph): mds: need to size values read from a bufferlist
- 05:39 PM Revision 5850363d (ceph): mds: Determine if there are file locks based on flockbl size, not dirty...
- 05:39 PM Revision b4e9e4c0 (ceph): mds: replace accidnetal __u64 with uint64_t
- 05:39 PM Revision 40bb86ab (ceph): mds: better printing on lock messages
- 05:39 PM Revision 8d85f344 (ceph): mds: drop distributed locks on waiting filelock messages, so they don't...
- 05:39 PM Revision c0eccba3 (ceph): mds: locking. Fix waiting lock count map deletion
- 05:39 PM Revision eb669ad2 (ceph): mds: locking. Compare iterator to right list!
- 05:39 PM Revision e501da4b (ceph): mds: debug output and some comments.
- 05:39 PM Revision e555ac67 (ceph): mds: recover filelocks properly on client reconnect
- 05:39 PM Revision eb3d9b44 (ceph): msg: rename MClientReply::dir_bl to extra_bl. Fix uses
- 05:39 PM Revision aab26ff6 (ceph): mds: add messenging support for advisory file locking
- 05:39 PM Revision f60e20e5 (ceph): mds: fcntl. add skeleton [add, remove]_lock functions.
- Implement helper functions to be used later.
- 05:39 PM Revision b75fe3b8 (ceph): mds: fcntl. implement add_lock function
- Still need to write some of the helpers
- 05:39 PM Revision 27ff76bd (ceph): mds: fnctl. implement adjust_locks method, with helpers
- 05:39 PM Revision 1fe8ab34 (ceph): mds: fnctl. implement half of remove_lock.
- Still need to handle all the waiting_locks and set them if proper.
- 05:39 PM Revision e41056ac (ceph): mds: fnctl. add get_waiting_overlaps function; fix get_overlapped locks
- so that it won't do bad things if there aren't any.
- 05:39 PM Revision 18b1d772 (ceph): mds: fnctl. finish remove_locks so it'll activate whichever locks it can.
- This is lazy but it should work.
- 05:39 PM Revision b7b349e2 (ceph): mds: fcntl. add proper encoding to ceph_filelock and ceph_lock_state_t
- 05:39 PM Revision 4520fdfa (ceph): mds: add distributed lock support for fcntl/flock support
- 05:39 PM Revision 4ea76e8d (ceph): mds: Implement file lock message handling.
- 05:39 PM Revision 68e592a8 (ceph): mds: cleanup
- 05:39 PM Revision 5d375ea7 (ceph): mds: ceph_filelock and filelock_change need to be explicitly sized
- 05:39 PM Revision 0dbcdd84 (ceph): mds: implement look_for_lock, for fcntl F_GETLK-style things
- 05:39 PM Revision cc652fe7 (ceph): mds: fcntl. Keep track of how many locks/waiting locks each client has.
- 05:39 PM Revision 71cf9d6a (ceph): mds: fcntl locks can efficiently remove all locks from one client.
- 05:39 PM Revision 80412537 (ceph): mds: Now actually requeues blocked locks whenever locks are removed.
- 05:39 PM Revision 6a3600ae (ceph): mds: Remove unnecessary if check.
- We already know that head_in exists, or we would have entered the previous
if-block and returned from the function. - 05:39 PM Revision c14fa1d5 (ceph): ceph_fs: Add cap bits for advisory file locks
- 05:39 PM Revision cc845462 (ceph): mds: use the right size mask for file cap bits
- 05:39 PM Revision 367ff978 (ceph): ceph_fs: CEPH_LOCK_UNLOCK should have its own bit, for safety
- 05:39 PM Revision 8315b5db (ceph): mds: Move lock print function; add debugging; fix fencepost errors.
- 05:39 PM Revision a11a3197 (ceph): mds: conditionally include flock metadata in MClientCaps
- 05:39 PM Revision fd61e1b4 (ceph): mds: conditionally include flock metadata in MClientReconnect
- mds: move flockbl into each reconnect record
Simpler for the client side.
mds: keep old reconnect struct
mds: put ... - 05:21 PM Revision 6be6c48e (ceph): mds: remove size, mtime, atime from cap reconnect
- 04:57 PM Revision 2136ee76 (ceph): mds: make mislinked dentry during replay workaround tolerate directorie...
- This should probably all go away when the original bug is fixed!
- 04:28 PM Linux kernel client Bug #283 (Resolved): ceph_add_cap: couldn't find snap realm, NULL ptr deref
- This is a server-side problem with CInode::encode_inodestat, fixed by commit:6573635ba48a9b6c4f364e8f8b7132c90ea2e8e9...
- 04:15 PM Revision 67236ae7 (ceph): move logrotate conf out of debian/ to fix ceph.spec
- 02:12 PM Bug #306 (Resolved): osd: adjusting replication should be fast, keep data available
- This was a problem with handle_osd_map doing all the slow steps even when there was no new information in the osdmap....
- 02:11 PM Bug #282 (Resolved): osd: heartbeat can't keep up with large cluster changes
- 02:11 PM Bug #282: osd: heartbeat can't keep up with large cluster changes
- fixed what i think is the last issue here in commit:9bfb8da9f925642bca46528a999124cd8b28ba2a
- 01:47 PM Cleanup #324 (Resolved): standalone ceph-client-tools package
- 12:15 PM Bug #326: OSD crash PG::IndexedLog::unindex
- Saw this crash again, just added a new core file (core.node10.2629) to the logger machine. Also uploaded the log from...
- 10:04 AM CephFS Bug #328 (Resolved): MDS crash: MDCache::remove_inode(CInode*)
- Fixed the replay workaround in commit:2136ee763659e84f5715974450b89e8dea31a717
The original source of the problem,... - 09:43 AM CephFS Bug #328 (Resolved): MDS crash: MDCache::remove_inode(CInode*)
- Today i tried to sync kernel.org again, this went fine, until my log parition filled up and mds0 got stuck.
Someho... - 10:03 AM CephFS Bug #329 (Resolved): mds: mislinked dentry found during journal replay
- There is a FIXME error that is logged during replay when we encounter a dentry that is already linked and a journal e...
07/31/2010
- 04:14 PM Revision 1d221964 (ceph): auth: fix cast
- 12:14 AM Revision e3721638 (ceph): rgw: Support a loglevel of 0
07/30/2010
- 11:35 PM Revision 309f0652 (ceph): cauthtool: correct man page sample of minimal fs client caps
- 11:12 PM Revision 71532447 (ceph): rgw: fix setting acls on the bucket
- 07:06 PM Revision b502be7a (ceph): move mkcephfs to /sbin
- 07:05 PM Revision 7465b5e3 (ceph): debian: more cleanups
- https://code.launchpad.net/~clint-fewbar/review-new-branches/ceph-new-pkg/+merge/31303
- 05:24 PM Revision 8b8fd7e9 (ceph): auth: define AES iv in common header
- Signed-off-by: Sage Weil <sage@newdream.net>
- 04:34 PM Bug #327 (Closed): OSD crash PG::IndexedLog::print
- I'm not sure if this is the same issue as #326, that's why i open a new issue.
At the same moment #326 took place,... - 04:27 PM Bug #326 (Resolved): OSD crash PG::IndexedLog::unindex
- I've just seen this crash on one of my OSD's running with the latest unstable.
Have no idea what went wrong (Was j... - 04:19 PM Revision 325b41a3 (ceph): add files to .gitignore
- add generated files to gitignore.
Signed-off-by: Thomas Mueller <thomas@chaschperli.ch>
Signed-off-by: Sage Weil <sa... - 04:18 PM Revision bcbad88d (ceph): redirect stdout to /dev/null when reloading
- the reload command prints to stdout. example:
# service ceph reload
=== mon.0 ===
Reloading Ceph mon0 on hoe001...ki... - 09:25 AM Bug #325 (Can't reproduce): writes starve reads
- From: Henry C Chang <henry_c_chang@tcloudcomputing.com>
To: ceph-devel@vger.kernel.org
Subject: writes-starving-rea... - 09:25 AM Cleanup #324 (Resolved): standalone ceph-client-tools package
07/29/2010
- 10:49 PM Revision b72c1bb6 (ceph): monclient: use default port if not specified via -m
- Reported-by: Thomas Mueller <thomas@chaschperli.ch>
Signed-off-by: Sage Weil <sage@newdream.net> - 08:32 PM Revision 728c7669 (ceph): mon: we support MONNAMES feature
- 08:24 PM Revision 9052986f (ceph): Merge branch 'unstable' into mon-remove
- Conflicts:
src/cmon.cc
src/include/ceph_fs.h
src/mon/MonClient.cc
src/mon/Monitor.cc - 06:37 PM Revision 11dc1af1 (ceph): mon: fix signed/unsigned compile warning
- 06:32 PM Revision 719aa186 (ceph): debian: add update_pbuilder.sh script; don't do it during build
- 04:38 PM Revision f3ac22ab (ceph): bump unstable version to v0.22~rc
- 04:37 PM Revision ea2ce9d4 (ceph): cclass.in: fix libdir definition
- 04:37 PM Revision 090436f5 (ceph): v0.21
- 04:29 PM Revision fd4029cb (ceph): debian/ceph.logrotate: enhance postrotate for "service" util
- [ The following text is in the "UTF-8" character set. ]
[ Your display is set for the "iso-8859-1" character set.... - 04:17 PM Revision 6db89756 (ceph): mon: trim old logm states
- Signed-off-by: Sage Weil <sage@newdream.net>
- 04:14 PM Revision 183984e1 (ceph): ceph.spec.in: create /var/log/ceph/stat dir
- the logrotate config expects a var/log/ceph/stat dir - create it.
Signed-off-by: Sage Weil <sage@newdream.net> - 04:14 PM Revision e1a97685 (ceph): debian: ceph-cfuse: move ceph from Recommends to Suggests
- on a standalone fuse client ceph serverside is not needed. if it's
"Recommended" the default behaviour is to install ... - 03:50 PM CephFS Feature #321: allow dns names for monitor switch
- the port part of htis is fixed by commit:b72c1bb6e9b77e1ab6c21bfd761f7fe3cc493647
- 09:27 AM CephFS Feature #321 (Resolved): allow dns names for monitor switch
- if the -m switch is set to a dns name cfuse fails:
# cfuse -m node001:6879 /mnt
unable to parse addrs in 'node001... - 02:45 PM Bug #323 (Duplicate): Get the wireshark dissector up to date
- 02:16 PM rbd Bug #322: Can't create an image via rbd or qemu-kvm
- My bad, this is a duplicate of #263, but the solution of #263 doesn't apply here.
- 02:12 PM rbd Bug #322 (Closed): Can't create an image via rbd or qemu-kvm
- ...
- 01:28 PM Feature #195 (Resolved): monitor naming and removal support
- merged in commit:9052986fd3fb5fc065fd9781bbfdb131194e845a
- 04:57 AM rgw Feature #313: Support loglevel
- One small glitch in the patch, setting a loglevel of 0 was not supported due to a wrong if-condition.
Attached a p... - 04:51 AM rgw Cleanup #320 (Resolved): Remove some duplicates and unused declarations
- I've done some small cleanups in the gateway, mostly unused declarations, double things or just unneeded comment line...
- 04:04 AM Revision ce07f280 (ceph): ceph.spec.in: rbdtool -> rbd
- Reported-by: markn1
07/28/2010
- 11:28 PM Revision b0616f85 (ceph): debian: drop gcc, g++ build deps
- 11:25 PM Revision 47688a43 (ceph): debian: fix weak-library-dev-dependency
- E: ceph source: weak-library-dev-dependency librados1-dev on librados1
E: ceph source: weak-library-dev-dependency li... - 11:25 PM Revision 07d880ea (ceph): debian: update to latest standards version
- 11:25 PM Revision 39a3db6b (ceph): debian: fix debhelper depend version
- dh_lintian was introduced in 6.0.7
- 11:25 PM Revision 58181fe5 (ceph): debian: add trailing newline
- 11:25 PM Revision 487f0e85 (ceph): debian: remove gcc/g++ build-depends
- (as per Clint Byrum)
- 11:25 PM Revision 9c6ae54c (ceph): release: drop wildcard
- 10:41 PM Revision 3ed08a33 (ceph): check_version: don't globber .git_version if we're not in a git repo
- namely, when you extract the release tarball!
- 10:00 PM Revision 8b0bf8e1 (ceph): filestore: add option to disable use of btrfs clone range ioctl
- 09:57 PM Revision eb1f5874 (ceph): filestore: fix fallback clone_range implementation on short files
- If we read less data from the source, don't loop forever... return -ERANGE
instead. - 08:57 PM Revision 43d28dbc (ceph): release: put .git_version in release tarball
- 08:37 PM Revision 9b8e2b4f (ceph): pgmon: Remove creating_pgs which are from deleted pools.
- 07:47 PM Revision 425e052e (ceph): osd: drop broken (and useless) PG::merge_log assert
- The assert should be olog.backlog, not log.backlog. But it's the same
condition as the if guard that's 3 lines up, m... - 06:55 PM Revision 05a18e20 (ceph): mds: handle this log replay error better:
- Log it to the central log so it gets noticed, and
delete the "wrong inode" instead of just unlinking it, to prevent l... - 06:41 PM Revision 801150a7 (ceph): rgw: Support loglevel settings
- The gateway now supports changing the verbosity of the logging, this can be achieved by changing the
RGW_LOG_LEVEL en... - 06:41 PM Revision 593937d6 (ceph): librados: quiet down some log messages
- 06:41 PM Revision b01cc38f (ceph): rgw: set default log level to 20
- 11:56 AM CephFS Bug #312 (Can't reproduce): MDS crash: LogSegment::try_to_expire(MDS*)
- Unfortunately we can't solve the root cause of this because we don't have detailed logging when the inodes in questio...
- 10:12 AM CephFS Bug #312 (In Progress): MDS crash: LogSegment::try_to_expire(MDS*)
- 09:54 AM rgw Feature #313: Support loglevel
- I gave this some work today and created a patch, see the attachment.
The gateway will read the evirionment variabl...
07/27/2010
- 08:22 PM Revision 6bd40ac9 (ceph): qa: consistent snaptest-%d.sh naming
- 07:53 PM Revision 9127cd9d (ceph): mds: fix uninitialized LeaseStat for null lease
- ./include/encoding.h: In member function void
Locker::issue_client_lease(CDentry*, client_t, ceph::bufferlist&, utime... - 07:49 PM Revision 3de9c8d2 (ceph): client: some whitespace cleanup
- 07:49 PM Revision 7fbe1655 (ceph): debug: no name symlink when explicit --log-file
- 07:49 PM Revision 5c714bfb (ceph): osd: log when we get marked down but aren't
- 07:13 PM Revision 8195899a (ceph): qa: add localized version of Thomas Mueller's snaptest-2.sh
- http://github.com/vinzent/ceph-testsuite/blob/master/tests/snaptest-2
- 06:18 PM Revision 2d35d247 (ceph): rgw: exit after 5 seconds from SIGUSR1 anyway
- 05:28 PM Revision a2ba0d35 (ceph): rgw: make sure error structure is initialized
- 05:28 PM Revision 58cf95bb (ceph): rgw: Make sure all times are converted to GMT, so we can compare them c...
- 05:28 PM Revision 3d1f3072 (ceph): rgw: fix returned error code in If-None-Match
- 05:28 PM Revision 50a89fcb (ceph): rgw: handle sigusr1, get notified for shutdown requests
- 05:28 PM Revision a48b1504 (ceph): rgw: fix that we return data when partial content is requested
- 05:28 PM Revision 0acfd014 (ceph): rgw: support partial content (206) response
- Couldn't get this one to work yet, as apache (or probably one of
the apache filters) rewrites the content-length valu... - 05:28 PM Revision 234b6916 (ceph): We shoud always return the Accept-Ranges header, with a GET or HEAD req...
- This is to let the client know we accept requests for partial content.
On HEAD requests we should also return the Co... - 05:28 PM Revision 017588ed (ceph): We should sent back the Last-Modified header in GMT as specified by RFC...
- 05:28 PM Revision 577095b3 (ceph): rgw: fix error status code
- 05:28 PM Revision d9ff7e09 (ceph): rgw: fix return code for prepare_get_obj() on the fs storage
- 05:28 PM Revision b628ba19 (ceph): rgw: fix date/time format string
- 02:14 PM CephFS Bug #319 (Closed): allows snaps in root directory
- 12:57 PM Linux kernel client Bug #317 (Resolved): sync hangs when doing concurrent syncs
- Fixed by commit:e33b35cf88cd06d69332bd399d5720edc85c17fa.
Problem was that we didn't use complete_all and wake_up_... - 11:38 AM Linux kernel client Bug #317 (Resolved): sync hangs when doing concurrent syncs
- For example, running iozone and doing a sync concurrently, sync will hang (as iozone does sync internally).
- 12:01 PM CephFS Bug #288 (Closed): cmds disappears under snapshot load
- All right, it works on one MDS. Opened #318 to track issues with the multi-mds cluster.
- 12:01 PM CephFS Bug #318 (Resolved): Multi-MDS snaptest-2 fails
- See #288. Looks like a problem with flushing (or other handling) of snapshotted inodes.
- 11:11 AM CephFS Bug #316 (Resolved): mds: track dirty snapped metadata without faking a cap
- 10:57 AM rgw Feature #174: Support large files better
- Probably the ETag calculation is not being done right. Should be done in chunks too.
- 10:56 AM rgw Feature #302 (Resolved): Support If-Modified-Since
- 10:54 AM rgw Bug #310 (Resolved): rgw fcgi processes don't go down
- Should be fixed now with commit:50a89fcb5206e5afde88c01cd5370518b86c470d and commit:2d35d247844d030fe628b227c836ff657...
- 10:47 AM Feature #315 (Resolved): mds disaster recovery tools
- things like
- extract mds journal
- edit/wipe mds journal
- 10:47 AM Feature #314 (Closed): osd disaster recovery tools
- things like:
- edit or wipe pg logs
- "hide" problematic pgs to get the system online
- throttle osd cluster recov... - 10:14 AM Linux kernel client Feature #206: make a 'soft' mode
- Make 'hard' and 'soft' mount options, ala nfs. Default is 'hard'....
- 10:07 AM rgw Bug #309 (Resolved): Error status is not correctly initialized
- 07:29 AM rgw Bug #309: Error status is not correctly initialized
- I can confirm it has been fixed. States now work, no more "caching" like i saw.
- 10:03 AM rgw Feature #301 (Resolved): Support partial content response
- 10:03 AM rgw Feature #301: Support partial content response
- Great! I modified your fix a bit. We should only return the data if there's no actual error, and pushed it.
- 12:49 AM rgw Feature #301: Support partial content response
- There was still a problem that with partial content no data was returned. This was due to a error in rgw_rest.cc.
... - 08:06 AM rgw Feature #313 (Resolved): Support loglevel
- Like spoken about yesterday, it would be usefull if we could set a loglevel for the RADOS Gateway.
We would use th... - 04:33 AM CephFS Bug #312 (Can't reproduce): MDS crash: LogSegment::try_to_expire(MDS*)
- This morning i upgraded my cluster to the latest unstable, afterwards i tried to mount the cluster, which failed.
...
07/26/2010
- 11:43 PM Revision e2b1a4ee (ceph): mds: Use get_oldest_snap() (not first) in handle_client_lssnap.
- 11:13 PM Revision 49c21837 (ceph): rgw: remove unused label 'done' (compiler warning)
- 10:58 PM Revision 0ae06c81 (ceph): msgr: fix direct send to loopback connection
- If we send directly to the local/loopback connection, we need to
queue it for local delivery, not go through the usua... - 10:03 PM Revision 997d67e5 (ceph): mon: revamp subscribe protocol [backward compatible protocol change]
- Before, we would provide "have" and a bool "onetime" flag. The struct was
also screwed up with an extra __le64. The... - 06:44 PM Revision fa48a1a8 (ceph): mds: restore CInode::get_oldest_snap()
- It's either 'first', for the first for the oldest old_inode.
- 05:27 PM Revision bc9a7955 (ceph): mds: defer anchortable updates before CLIENTREPLAY, not ACTIVE
- Otherwise we hang trying to replay an op that interacts with the
anchortable. - 05:24 PM Revision ad6f92c0 (ceph): Merge remote branch 'origin/large_objects' into unstable
- 05:24 PM Revision 0775c30b (ceph): mds: get rid of old lease mask cruft
- Signed-off-by: Sage Weil <sage@newdream.net>
- 05:24 PM Revision ac9305f8 (ceph): mds: fix lease revocation mask vs type
- 04:43 PM CephFS Bug #288: cmds disappears under snapshot load
- Switched back to using get_oldest_snap, works on a single-MDS install as of commit: e2b1a4ee119a68b403582ae3bc15b54e9...
- 03:23 PM RADOS Cleanup #311 (New): osd: remove read(len=0) full object behavior
- ..,after the objecter doesn't need it (see #303)
- 03:23 PM Feature #175 (Resolved): Make the system large-object safe
- the recovery portion of this is done.
Remaining bits include #303, and then removing the '0 == full object' logic ... - 03:22 PM Bug #308 (Resolved): osd can't get full map history
- resolved by commit:997d67e5b1ce6df1b467a9c8b284f1289fbee22a
- 12:10 PM Bug #308 (Resolved): osd can't get full map history
- 10.07.26_20:55:17.810053 7f424cf8d710 osd10 0 handle_osd_map epochs [78,78], i have 0
10.07.26_20:55:17.810093 7f424... - 03:05 PM rgw Feature #302: Support If-Modified-Since
- The patch seems fine, just a small issue is that you changed the return code in the unmodified-since case. This would...
- 02:01 PM rgw Feature #302: Support If-Modified-Since
- See the attached patch, this will convert all times to GMT, so they can be compared correctly.
- 03:20 AM rgw Feature #302: Support If-Modified-Since
- There is some more to this and it might need some more work to do so.
Right now, all the times stored by the RADOS... - 02:58 PM rgw Bug #310 (Resolved): rgw fcgi processes don't go down
- Need to implement a signal handler for the SIGUSR1 signal, to receive apache requests to go down.
- 02:57 PM rgw Bug #309: Error status is not correctly initialized
- should be fixed by commit:21c8144c0d631530704acb1f08f6dddfa6c83eb1
- 02:55 PM rgw Bug #309 (Resolved): Error status is not correctly initialized
- Reported by wido: once getting a "304" response, we keep on getting that error, even when operating on different obje...
07/24/2010
- 04:50 PM Revision 9932e4b4 (ceph): osd: on dups, reply if committed, or wait until ondisk
- 04:49 PM Revision 47fd7fe2 (ceph): osd: helper should reply with ACK on error, op flags on success
- 04:37 PM Revision b0332590 (ceph): osd: maintain pg last_update_ondisk
- Only defined when is_active(). Only printed when it lags (i.e. we are
in the process of replicating and/or writing a... - 04:37 PM Revision fb359c8e (ceph): osd: map indexed caller_ops back to relevant entry
- This let's us look up a version from the osd_reqid_t.
- 10:11 AM Bug #307 (Resolved): osd: fix wrnoop on dup op
- commit:9932e4b40713df2e33f53583274d9ff35adca03b
07/23/2010
- 11:05 PM Revision 3733b177 (ceph): initscripts: fix type.id check in ceph.conf
- 10:53 PM Revision 8f9f57a2 (ceph): Merge branch 'unstable' into large_objects
- 10:52 PM Revision b2194357 (ceph): osd: make assemble_backlog more tolerant of races
- The assemble_backlog is merging data generated while racing with online
updates. It needs to tolerate races with thi... - 10:51 PM Revision dd684835 (ceph): osd: clean up size learning during pull
- Signed-off-by: Sage Weil <sage@newdream.net>
- 10:51 PM Revision 1d0c91d5 (ceph): osd: populate in-memory snapset context with just-pulled SnapSet
- This can come from the SNAPDIR or NOSNAP (head) object.
Signed-off-by: Sage Weil <sage@newdream.net> - 10:51 PM Revision c6111924 (ceph): osd: create missing temp collection/dir
- 10:51 PM Revision b59cb10d (ceph): osd: skip degraded objects we're already recoverying in recover_replicas
- Otherwise we'll start_recovery_op, dup the push, etc.
Signed-off-by: Sage Weil <sage@newdream.net> - 10:50 PM Revision 9520917d (ceph): mds: drop remove_client_lease return value
- Unused
- 10:50 PM Revision 74fcb6bf (ceph): mds: fix dentry lease mask/lockid disconnect
- There is only one lock here, no need to be tricky with the mask.
- 10:50 PM Revision c33171c0 (ceph): mds: no return value for issue_client_lease()
- Return value is unused. Also drop obsolete inode lease method declaration.
Signed-off-by: Sage Weil <sage@newdream.... - 10:50 PM Revision 34957ff3 (ceph): mds: fix lease issue mask
- We define 1 to be the only lease "mask" we currently support (for dentry)
and divorce ourselves from the CEPH_LOCK na... - 10:50 PM Revision 15abbfbe (ceph): mds: create leases with non-zero seq
- We want to avoid the situation:
- mds issues lease, seq 0
- mds times out lease, seq 0
- client sends seq 0 release
... - 10:50 PM Revision 92273f70 (ceph): osd: log/debug hb set changes
- 10:50 PM Revision b33b53b1 (ceph): osdmap: recalc num_osds after applying incremental
- 10:50 PM Revision 4ab66fa4 (ceph): osd: clean_up_local should ignore snapped, snapdir objects
- The snap_trimmer is responsible for removing any/all snap objects.
snapdir only exists if snapped objects exist and ... - 10:50 PM Revision fdd13cb2 (ceph): filestore: make stat print correct errno
- 10:50 PM Revision 622e9b47 (ceph): osd: simplify heartbeat checks
- - Only check heartbeats when we have heartbeat_lock and osdmap rdlocked,
and thus _know_ heartbeat info and map are i... - 10:49 PM Revision 9e3650be (ceph): osd: red is good enough; don't blink
- 09:37 PM Revision f731d4d6 (ceph): mds: fix snap lookups in CInode encoding
- 09:37 PM Revision 7bba8447 (ceph): qa: runoncfuse shouldn't start cfuse in background any more.
- 09:37 PM Revision 9c1ceb1e (ceph): mds: only try and lssnaps on a dir for as long as it's been around
- 08:21 PM Revision ff53e9b3 (ceph): mon: don't crash on 'osd pool set' command
- 06:46 PM Revision e986721b (ceph): osd: zero ondisklog pointers when starting pg deletion
- This fixes a problem where the osd stops part way through pg cleanup. It
seens the old ondisklog bounds, but then fa... - 06:43 PM Revision 16c0fea2 (ceph): librados: remove use of snapid_t
- 04:42 PM Revision 6b7693bf (ceph): objecter: default debug level 0 (no 'laggy' messages)
- 04:31 PM Bug #307 (In Progress): osd: fix wrnoop on dup op
- 03:27 PM Bug #307 (Resolved): osd: fix wrnoop on dup op
- - need to return success, but not until the request has definitely committed.
- shouldn't do the old noop nonsense..... - 02:35 PM rgw Feature #302: Support If-Modified-Since
- Seems that the times were not parsed correctly and converted to GMT.
Attached patch will fix this and make the if-... - 01:20 PM rgw Feature #302: Support If-Modified-Since
- Yes, Amazon does that. The RFC specifies that when you do so, you have to return a 304 Not Modified header.
Like y... - 09:52 AM rgw Feature #302: Support If-Modified-Since
- Actually we do implement the If-Modified-Since, if it's not working that's a bug. One difference is that we don't ret...
- 08:24 AM rgw Feature #302 (Resolved): Support If-Modified-Since
- "The If-Modified-Since request-header field is used with a method to make it conditional: if the requested variant ha...
- 02:13 PM Bug #306 (Resolved): osd: adjusting replication should be fast, keep data available
- it's currently not... something isn't working as intended.
- 01:24 PM Bug #305 (Resolved): Monitor crash on "ceph osd pool set"
- fixed in commit:ff53e9b33d1911d047d8be0cdaa65e1e6c62d7f1
- 01:18 PM Bug #305 (Resolved): Monitor crash on "ceph osd pool set"
- When running "ceph osd pool set" with no arguments, the monitor who receives this command crashes....
- 01:13 PM Linux kernel client Bug #304: GPF in writepages_finish
- the first crash is addr.c:534,
WARN_ON(!PageUptodate(page));
bad page pointer page=5b5b5b.., i=0. - 01:13 PM Linux kernel client Bug #304: GPF in writepages_finish
- another node got this, probably the same bug?...
- 01:02 PM Linux kernel client Bug #304 (Can't reproduce): GPF in writepages_finish
- untar_snap_rm.sh...
- 12:58 PM CephFS Bug #288 (In Progress): cmds disappears under snapshot load
- Looks like there's more to it than this, I got another crash farther on. Continuing to study.
Also there might be ... - 11:57 AM Linux kernel client Bug #150 (Can't reproduce): order:1 page allocation failure
- we've fixed a bunch of memory leaks. haven't seen this recently.
- 11:53 AM Feature #303 (Resolved): objecter: limit in-flight ops and/or bytes written
- 11:51 AM Bug #279 (Resolved): OSD crash on empty pglog
- 11:51 AM Bug #279: OSD crash on empty pglog
- i think this is fixed by commit:e986721b06973a9379a37961e49ce334c480fa1a. not a disk full problem at all.
- 11:49 AM Bug #300: osd: make pg removal atomic
- actaully commit:e986721b06973a9379a37961e49ce334c480fa1a
- 11:48 AM Bug #300 (Resolved): osd: make pg removal atomic
- commit:9a3c990d3708b3ec67a6a980c93fa814aece1c2e
- 11:44 AM Feature #33 (Closed): O_LAZY or equivalent
- 10:20 AM Bug #282: osd: heartbeat can't keep up with large cluster changes
- Greg Farnum wrote:
> Do we still think this is an issue after commit:856999eda434fa9b7d93b152427cf7c82240f220 ("osd:... - 10:17 AM Bug #282: osd: heartbeat can't keep up with large cluster changes
- Do we still think this is an issue after commit:856999eda434fa9b7d93b152427cf7c82240f220 ("osd: clear failure_queue w...
- 10:06 AM Linux kernel client Bug #139: BUG ceph_dentry_info: Objects remaining on kmem_cache_close()
- actually, commit:252af5214682191e34e57204e1a31924fb82c207
- 07:43 AM rgw Feature #301: Support partial content response
- Uh, my bad, the Last-Modified header value was not a GMT timestamp. The attached patch fixes this by using "gmtime()"...
- 07:06 AM rgw Feature #301: Support partial content response
- I've done some work on this today, while doing so i found some more issues.
The attached patch fixes that with GET...
07/22/2010
- 09:27 PM Revision d665d494 (ceph): initscripts: use $type.$id, not $name, when possible
- 09:26 PM Revision de9262e9 (ceph): initscripts: accept osd0 or osd.0, regardless of what conf uses
- 09:17 PM Revision d9a7acbf (ceph): Merge remote branch 'origin/large_objects' into unstable
- 07:01 PM Revision ad1cd81f (ceph): mds: be careful obeying REQRDLOCK
- Only do a simple_sync() if we are stable, auth, and not already sync. The
client request can race with other state c... - 05:53 PM Revision b8c1bda8 (ceph): debian: fix typo in libcrush description
- Reported-by: Wido
- 04:30 PM rgw Feature #301 (Resolved): Support partial content response
- When playing rgw-hosted movie via vlc, can't fast forward/set movie position. This happens because we don't reply wit...
- 04:14 PM Bug #300 (Resolved): osd: make pg removal atomic
- currently we can start pg removal, delete pglog, then crash while deleting objects, then fail on restart in load_pgs ...
- 03:31 PM CephFS Bug #30: multimds: slave_request on getattr
- In order to change this, we need a way to wake up the replica waiter when we find out the auth changed, so we can res...
- 02:37 PM Bug #273 (Can't reproduce): OSD crashes during replication level change
- 02:30 PM Bug #290 (Resolved): take osd1 or osd.1 to initscript
- commit:de9262e995085e892df481bb5607b39510a616a7
- 02:19 PM CephFS Bug #297 (Closed): MDS crash on Objecter::handle_osd_op_reply
- this is just out of memory. opened up #299 to improve logging.
- 06:20 AM CephFS Bug #297: MDS crash on Objecter::handle_osd_op_reply
- I've tried restarting the MDS'es multiple times, all resulting in the same crash again from both MDS'es.
The core-... - 05:49 AM CephFS Bug #297: MDS crash on Objecter::handle_osd_op_reply
- Forgot to note my cluster state:...
- 05:49 AM CephFS Bug #297 (Closed): MDS crash on Objecter::handle_osd_op_reply
- While doing a rsync of kernel.org again, both my MDS'es crashed....
- 02:14 PM Linux kernel client Bug #291 (Resolved): leaked dentry (?) in ceph_readdir_prepopulate when ceph_init_dentry fails
- fixed by commit:68301be0b91242139ffa215c1163a78f3e18db38
- 02:11 PM Linux kernel client Bug #139 (Resolved): BUG ceph_dentry_info: Objects remaining on kmem_cache_close()
- ah, finally fixed by commit:59bb6d659605044c2c62e2f5c76b712db36f7da8
- 01:33 PM Feature #245: Logging to syslog
- Thought about this some. A few things:
the dout() macros just need to use a temp stringstream (not *_dout), then ... - 01:28 PM Cleanup #299 (New): catch std::bad_alloc and die with helpful error in log on ENOMEM
- 12:12 PM CephFS Bug #288: cmds disappears under snapshot load
- Sage should have got this in commit:1271fdd0e345d64493c386167e38e3bfea7c52e6. Will test and confirm.
- 12:04 PM CephFS Bug #296 (Resolved): Locker::simple_sync assert(0)
- fixed by commit:ad1cd81fd00067a2da88daf053b82683a2e64a0b
- 11:17 AM Feature #298 (Resolved): osd: "watch" operation
- - returns when object changes
- client needs to support long-running ops (w/o timing out or whatever)
- need to exp...
07/21/2010
- 08:30 PM Revision 1271fdd0 (ceph): mds: initialize snaprealm created, current_parent_since on creation
- Need to initialize created and current_parent_since on new snaprealms
when they are created, or else we get incorrect... - 08:30 PM Revision 6aa32f8c (ceph): Merge remote branch 'origin/debian' into unstable
- 06:03 PM Revision 93a2ba8e (ceph): rgw: fix calls to finish_get_obj() with uninitialized handle
- 04:31 PM Revision 0bd671f7 (ceph): debian: drop unused *.modules.in
- 04:27 PM Revision ef2ecdab (ceph): fuse.cc: make license GPL
- Originally based on Miklos' fusexmp.c, which is GPL. Keep that license.
LGPL makes no sense here anyway. - 04:27 PM Revision dd9d23b3 (ceph): debian: cleanup from Clint Byrum <clint@ubuntu.com>
- With a few changes by me (changelog, copyright clarification)
- 04:27 PM Revision 68fc3fc8 (ceph): debian: no more ceph-kclient-soruce
- 03:26 PM Linux kernel client Bug #271: dropping dirty+flushing AxFxw state for
- It does look like a multiple mds only bug. This message can occur when ci->i_caps is empty while the ci->i_flushing_i...
- 02:40 AM Linux kernel client Bug #271: dropping dirty+flushing AxFxw state for
- I just saw this again, saved the logs on *logger.ceph.widodh.nl* in */srv/ceph/issues/issue_271*...
- 01:35 AM rgw Feature #174: Support large files better
- I think i was a bit to early with reporting. A few minutes later the file showed up with the correct filesize and a E...
- 01:27 AM rgw Feature #174: Support large files better
- After that commit the Content-Length works fine.
Uploading large files still fails, for example:...
07/20/2010
- 10:07 PM Revision 24c75167 (ceph): clean up new pool crush rule values [backward compatible protocol change]
- Default to -1 (not 0!), which means 'server default'.
- 10:07 PM Revision 6cb3ea76 (ceph): mon: specify default new pool size/rule/pg_num in conf
- 08:24 PM Revision 856999ed (ceph): osd: clear failure_queue when marked down
- This prevents bleed through of failures (due to not getting hearbeats, due
to us being marked down) so they don't get... - 08:24 PM Revision e8eec86c (ceph): mon: set/clear pause{rd,wr} osdmap flags
- 08:24 PM Revision 9ddb6ab2 (ceph): osd: infer missing map is empty when last_complete==last_update
- 07:39 PM Revision 618f7f25 (ceph): osd: clean up pulling first/complete business
- 07:30 PM Revision 46247640 (ceph): osd: push in chunks
- Signed-off-by: Sage Weil <sage@newdream.net>
- 07:13 PM Revision f3eb9645 (ceph): rgw: fix content-length reporting
- and also some other various issues that came up while at it
- 05:38 PM Revision 190df1c6 (ceph): debug: fix log append
- ios::ate does not work as advertised!
Signed-off-by: Sage Weil <sage@newdream.net> - 05:36 PM Revision 538ceedc (ceph): debug: say 'append' or 'new' when opening log
- Signed-off-by: Sage Weil <sage@newdream.net>
- 05:25 PM Revision 2483096e (ceph): osd: wait for non-existant pools (w/o crashing); clean up handle_op()
- Make sure the pool exists before calling OSDMap::raw_pg_to_pg(). Fixes
crash on operations shortly after pg pools ar... - 04:37 PM CephFS Bug #296 (Resolved): Locker::simple_sync assert(0)
- Started seeing this today while working on #288. Once it strikes a run it sticks around, but it's intermittent about ...
- 04:33 PM CephFS Bug #295 (Can't reproduce): mds: can't rmdir due to dir size underflow
- [dir 10000181ec7 /ceph1/untar_snap_rm.sh/ [3b,head] auth v=5988 cv=5988/5988 state=1073741826|complete f(v0 m10.07.2...
- 11:55 AM rgw Feature #174: Support large files better
- commit:f3eb96457b193b1f5d79cf2b41a3cda690c0eab0 fixes the content length issue.
- 02:54 AM rgw Feature #174: Support large files better
- Removing the following lines in *rgw_rados.cc* the content-length is returned correctly:
*Line 621:**... - 02:08 AM rgw Feature #174: Support large files better
- The commit from last night seems to have broken the Content-Length header, see the attached screenshot.
There is a... - 11:21 AM Linux kernel client Feature #294 (Resolved): honor PAUSE* flags in osdmap
- 10:43 AM Bug #293 (Can't reproduce): cmon crash during paxos update
- hmm, i fixed this by fixing logm/last_committed to have the actual last committed state (3141 i think). I'm not sure...
- 08:50 AM Bug #293 (Can't reproduce): cmon crash during paxos update
- Today i experienced some crashes of my monitors and mds'es due to my disks filling up with logs.
I had to re-start... - 10:29 AM Bug #292 (Resolved): OSD crash raw_pg_to_pg
- fixed by commit:2483096e664c772478bb7dcf718297f7779fe5c7
- 02:06 AM Bug #292 (Resolved): OSD crash raw_pg_to_pg
- In my cluster osd28 just got marked as down, i assume the heartbeat problem again. ( I was playing with the S3 Gatewa...
- 12:20 AM Revision a0cd80cf (ceph): rgw: fix copy
Also available in: Atom