Activity
From 01/04/2011 to 02/02/2011
02/02/2011
- 10:55 PM Revision e3fe2dbc (ceph): RGW: RGW::GetObj - signed/unsigned bug
- get_obj returns an int, len is unsigned
Signed-off-by: Samuel Just <samuel.just@dreamhost.com> - 10:38 PM Revision d47789d6 (ceph): RBD.cc: pread returns a signed value
- The check for len < 0 here always returned true since len is uint64_t.
Signed-off-by: Samuel Just <samuel.just@dream... - 10:38 PM Revision 763fb136 (ceph): RGW: get_service_secret returns a bool, not an int
- Signed-off-by: Samuel Just <samuel.just@dreamhost.com>
- 10:38 PM Revision 10ceea99 (ceph): RGW: fix shadowed variable error
- int ret and bool ret shadow this->ret, setting ret before the goto
therefore had no effect in these cases.
Signed-of... - 10:38 PM Revision 960cd6d0 (ceph): RGW: fix ret in RGWCopyObj::init_common
- int ret shadows RGWCopyObj::ret.
Signed-off-by: Samuel Just <samuel.just@dreamhost.com> - 10:33 PM Revision 4587f1fe (ceph): mount.ceph: option parsing fix
- Passing -o secretfile would cause a segfault since searching for = would
result in a null pointer. New version check... - 09:09 PM Bug #759 (Resolved): osd: pgs spend a long time peering when marking osds out
- On the playground (with lots of data), I see that some PGs spend a long time in peering state after marking an OSD as...
- 07:02 PM Revision 2b9cb24e (ceph): Fix base64-decoding when input ends in newline.
- It used to return -EINVAL because it thought the end was not aligned
to 4 bytes.
Clean up superfluous src < end test... - 06:04 PM Revision acf40be9 (ceph): Fix inner loop index variable. i and j are too close together ;)
- Before this fix, j would keep increasing well beyond in_len,
and lead to segfaults. - 05:57 PM Revision 0b166c31 (ceph): mds: Remove redundant cap suppression code.
- Signed-off-by: Greg Farnum <gregory.farnum@dreamhost.com>
- 05:57 PM Revision 0b7174f6 (ceph): rados: Make rados tool "put" objects in chunks.
- Defaults to 4MB, but can be configured.
Signed-off-by: Greg Farnum <gregory.farnum@dreamhost.com> - 05:57 PM Revision cef3ce82 (ceph): mds: Remove unused variable.
- Fixes compile complaint. Was made useless in commit
6d29fc66326ff40443ac465c84d6f7bddf3cee4f.
Signed-off-by: Greg Fa... - 05:57 PM Revision c7ac6f92 (ceph): osd: Fix compile-time warning.
- store is properly initialized inside a try block, but the
compiler doesn't notice that and so thinks it may be used
u... - 05:10 PM Revision 0972202f (ceph): Speling fixes
- Trivial spelling fixes...
Signed-off-by: Sage Weil <sage.weil@dreamhost.com> - 05:10 PM Revision eb3f65a6 (ceph): Improve mount.ceph.8
- Seeing as I was wondering what "rbytes" might be, I guess others might
too...
Signed-off-by: Sage Weil <sage.weil@dr... - 03:50 PM Revision cb897a5b (ceph): confutils: check return values
- Signed-off-by: Sage Weil <sage.weil@dreamhost.com>
- 03:50 PM Revision 8ebcc80b (ceph): mds: check return values in dumper
- Signed-off-by: Sage Weil <sage.weil@dreamhost.com>
- 03:50 PM Revision 471b1897 (ceph): monitorstore: check return values
- And use open(2) instead of fopen.
Signed-off-by: Sage Weil <sage.weil@dreamhost.com> - 12:59 PM Bug #644: rsync can be sloooow
- Discovered that the kernel client is clearing its I_COMPLETE flag on every newly-created directory due to how the cap...
- 12:25 PM Revision e7144e14 (ceph): Merge branch 'objecter_balance_reads'
- 12:00 PM Revision 05f26581 (ceph): os: use scoped variables for fds
- Signed-off-by: Colin McCabe <colin.mccabe@dreamhost.com>
- 11:56 AM Revision e6030151 (ceph): osd: change unfound handling when using localize
- Change how cosd handles unfound objects when doing operations with
localize_reads. Specifically, don't wait for unfou... - 11:56 AM Revision 2faf6a09 (ceph): csyn: implement localize_reads cmdline option
- Signed-off-by: Colin McCabe <colin.mccabe@dreamhost.com>
- 11:56 AM Revision a4781df7 (ceph): test: add test_csyn
- Signed-off-by: Colin McCabe <colin.mccabe@dreamhost.com>
- 11:56 AM Revision a28dbc93 (ceph): objecter: send read to random replica if BALANCE_READS flag is set
- The flag is just an existing one; we may want to rename.
This assumes the OSD will discard any request sent prior to... - 11:56 AM Revision e165f8ea (ceph): objecter: implement FLAG_LOCALIZE_READS
- Read from a local replica, if any.
Signed-off-by: Sage Weil <sage@newdream.net> - 11:56 AM Revision 087d4159 (ceph): libceph: iface for CEPH_OSD_FLAG_LOCALIZE_READS
- Signed-off-by: Colin McCabe <colin.mccabe@dreamhost.com>
- 11:56 AM Revision 7912f5c7 (ceph): cfuse: add --localize-reads argument
- Signed-off-by: Colin McCabe <colin.mccabe@dreamhost.com>
- 11:56 AM Revision a4c9bdf7 (ceph): Hadoop: set localize_reads
- Signed-off-by: Colin McCabe <colin.mccabe@dreamhost.com>
- 11:29 AM Bug #758 (Resolved): dd-truncate elsewhere-dd again is slow
- Jim Schutt has reported some troubling behavior apparently involved with client caps surrounding distributed (but not...
- 11:05 AM Bug #757 (Resolved): creating multiple snapshots through librbd can lead to inconsistent state
- Fixed in commit:a23903c5d8122d330cc159fb0f6ae0a7d72dd349 by rereading the header after creating or removing a snapshot.
- 11:03 AM Bug #757 (Resolved): creating multiple snapshots through librbd can lead to inconsistent state
- With the following sequence of events:
create image foo
create snap of foo
resize image foo
create snap of foo
... - 10:54 AM Bug #756 (Resolved): OSD should have max write size
- To prevent crashes if an incoming write is larger than the journal, we need an OSD max write size in the config. OSD ...
- 09:57 AM Cleanup #755 (Resolved): Compartmentalize rados.cc
- The rados tool code is pretty disgusting at this point. It consists entirely of a main function, with only usage() an...
- 09:55 AM Feature #624 (Resolved): radostool: make 'put' write large objects in chunks
- commit:0b7174f6aa072b6c0d8291a6c87a3991f8c401cd
Set chunk size with -b, defaults to 4MB.
02/01/2011
- 10:39 PM Revision ece4f61a (ceph): FileStore: fix double close
- curr_fd is already closed if cp == cur_seq. This second close
occasionally ended up closing another thread's fd. Th... - 07:02 PM Revision 06fe1320 (ceph): FileStore: fix double close
- curr_fd is already closed if cp == cur_seq. This second close
occasionally ended up closing another thread's fd. Th... - 06:54 PM Revision 16ffcf77 (ceph): common: config.cc: use "admin" as the default id
- Signed-off-by: Colin McCabe <colin.mccabe@dreamhost.com>
- 05:28 PM Revision 79ed51fd (ceph): mds: make --dump-journal preserve offset
- Suggest user use tar -S to preserve sparseness.
Signed-off-by: Sage Weil <sage.weil@dreamhost.com> - 05:25 PM Revision ffd549ba (ceph): common: move init_g_conf into md_config_t ctor
- Make sure that g_conf is initialized with default values before anything
else happens.
Signed-off-by: Colin McCabe <... - 05:22 PM Revision 196b817b (ceph): common: config.cc: whitespace cleanup
- Signed-off-by: Colin McCabe <colin.mccabe@dreamhost.com>
- 05:09 PM Bug #754 (Resolved): mkfs script should clean up its entries in /tmp, and use TMPDIR
- Our mkfs shell script creates stuff in /tmp and doesn't clean it up. Should use "trap" to ensure that this stuff gets...
- 04:33 PM Revision a4375329 (ceph): common: config.cc: de-globalize g_fake_kill_after
- Signed-off-by: Colin McCabe <colin.mccabe@dreamhost.com>
- 04:13 PM Revision 6a03ffbe (ceph): common: config.cc: de-globalize show_config
- Signed-off-by: Colin McCabe <colin.mccabe@dreamhost.com>
- 04:06 PM Revision 68f929a6 (ceph): common: clean up g_conf.id initialization a bit
- Signed-off-by: Colin McCabe <colin.mccabe@dreamhost.com>
- 02:23 PM Revision 4380920b (ceph): common: remove ceph_set_default_id
- ceph_set_default_id was only ever used to set the default ID to "admin",
which it already was.
Signed-off-by: Colin ... - 12:37 PM Bug #752: High I/O wait when writing data
- While running rsync I sometimes see these messages appearing inside the VM:...
- 12:13 PM Feature #753 (Resolved): Setting identity for syslog
- Right now all syslog messages get their process name as their identity in syslog, for example:...
- 12:03 PM Revision 49e238b5 (ceph): FileStore: fix error handling for mkfs, umount
- In FileStore::umount: check if FDs are valid before closing them. Make
them invalid after closing them. Shut down Fil... - 11:08 AM Linux kernel client Bug #750: run "dd", printk " libceph: osd1 172.16.10.68:6805 socket closed"
- Jeff Wu wrote:
> ceph client driver is from ceph-client.git unstable
There are some error handling fixes recentl... - 09:38 AM Feature #719 (Resolved): mds journal dump should be simpler
- commit:79ed51fd5bc30756f6a4f2e682c90a4e085f43d0
- 05:10 AM Revision 2373972f (ceph): gitignore: ignore eclipse metadata
- Signed-off-by: Sage Weil <sage.weil@dreamhost.com>
- 05:09 AM Revision a45e8f2d (ceph): remove ancient active/ stuff
- 04:58 AM Revision f2713332 (ceph): osd: don't leak fd on error
- Signed-off-by: Sage Weil <sage.weil@dreamhost.com>
- 04:55 AM Revision 4d50af76 (ceph): crypto: don't clobber errno
- Signed-off-by: Sage Weil <sage.weil@dreamhost.com>
- 12:24 AM Revision 6dc8994b (ceph): Merge remote branch 'origin/ostimeo'
01/31/2011
- 05:32 PM Bug #644: rsync can be sloooow
- Started taking a look at this again.
- 04:23 PM Bug #735: Manual drive pull testing hangs filesystem
- merged by commit:6dc8994b750631c15e88553fd4fabdd9e4907989
- 03:29 PM Bug #735: Manual drive pull testing hangs filesystem
- The ioctl timeout is now implemented in the ostimeo branch ( commit:2a266bd09d0db3b8d8c4f33a101229de1a4301a3 )
- 03:37 PM Linux kernel client Cleanup #168 (Closed): new truncate sequence
- we don't use the inode .truncate op.
- 12:23 PM Linux kernel client Bug #742: ceph-client.git unstable , tiotest ,printk "timeout, ceph: mds0 caps stale",then oops
- Were the osds running on top of btrfs? Also, could you post the ceph.conf?
- 11:22 AM Bug #752: High I/O wait when writing data
- Oh, I forgot my bonnie++ results.
- 11:22 AM Bug #752: High I/O wait when writing data
- I've done some benchmarks today in the VM and noticed something weird.
In the VM I ran:... - 10:40 AM CephFS Feature #630: release caps on inodes unlinked by other clients
- Putting this back in the queue since we've pushed it back past 1.0.
- 10:39 AM CephFS Bug #551 (Can't reproduce): cfuse crash on quick mds restart
- When I created this Sage told me that this wasn't actually the cause, and it was a different bug. I can't seem to rep...
01/30/2011
- 05:34 AM Revision 548f329f (ceph): Merge branch 'mds_reset'
- Fixes: #602
- 05:17 AM Revision 6137c789 (ceph): Merge remote branch 'origin/stable'
- Conflicts:
src/osd/OSD.cc
01/29/2011
- 09:41 PM CephFS Feature #602 (Resolved): mds: handle corrupt/missing journals
- no repair flag, but journal discard and cluster reset are in place. not especially useful yet without the actual rep...
- 09:25 PM Bug #563 (Closed): osd: btrfs, warning at inode.c ( btrfs_orphan_commit_root )
- 08:33 PM Bug #740 (Can't reproduce): osd's not sharing osdmaps with osd_ping
- 02:24 AM Revision 0f3198e8 (ceph): OSD: update_osd_stat take heartbeat_lock
- Previously update_osd_stat had a race with code modifying heartbeat_from
causing the iterator increment to occasional... - 12:56 AM Revision cf59cdf7 (ceph): mds: skip a few more inodes during journal reset
- To be safe...
Signed-off-by: Sage Weil <sage.weil@dreamhost.com> - 12:55 AM Revision f8f2cae0 (ceph): mds: open mydir (along w/ root) inode from boot_start()
- Signed-off-by: Sage Weil <sage.weil@dreamhost.com>
- 12:55 AM Revision e405d4a4 (ceph): mds: implement journal reset
- This basically works. Remaining issues:
- mydir and root inodes are recreated from scratch but need to be
reconc... - 12:48 AM Revision 14c669c3 (ceph): Locker: Drop loner correctly!
- Our previous check for if we want to drop the loner was incorrect.
Now, it's fixed. Resolves a serious bug with inode... - 12:47 AM Revision 04728b85 (ceph): librados: fix C interface const, too.
- See 561224e95d6c66661d1bd6dce0e3d9da6f4a7e13
Signed-off-by: Greg Farnum <gregory.farnum@dreamhost.com> - 12:47 AM Revision 9b42e8eb (ceph): Locker: Drop loner correctly!
- Our previous check for if we want to drop the loner was incorrect.
Now, it's fixed. Resolves a serious bug with inode...
01/28/2011
- 11:24 PM Revision 02408ecf (ceph): config: remove dead stringtable cruft
- Signed-off-by: Sage Weil <sage@newdream.net>
- 08:57 PM Revision 9e4325b2 (ceph): mds: defer sending resolves until mdsmap.failed.empty()
- There is no point sending resolves while there are still failed nodes,
since we can't complete. We also trigger an a... - 08:35 PM Revision f711508b (ceph): mds: defer sending resolves until mdsmap.failed.empty()
- There is no point sending resolves while there are still failed nodes,
since we can't complete. We also trigger an a... - 05:45 PM Revision aed862b4 (ceph): mds: standardize option parsing
- - Use the standard macros.
- Simply --hot-standby and --journal-check options (always specify rank).
- Update usage()... - 05:05 PM Revision 334a7564 (ceph): common: _dout_lock: initialize _dout_lock first
- Signed-off-by: Colin McCabe <colin.mccabe@dreamhost.com>
- 04:15 PM Feature #739 (Resolved): 'ceph health' should be warning when pg's are not all active
- Fixed by commit:50976297fd227664f3b451d77a11863259a18f48
- 04:08 PM Bug #751 (Resolved): _dout_lock is used after been disctucted.
- Fixed by commit:334a7564a21017e78f66ee0c73dcec4aacbb7244
- 01:56 PM Bug #751: _dout_lock is used after been disctucted.
- I have confirmed that _dout_lock is *not* the last thing to be destroyed.
Jan 28 05:57:42 flab cconf: creating mu... - 01:29 PM Revision e373d95e (ceph): common: Put debug.cc last in linking order
- debug.cc needs to be last in the linking order so that its global
constructors will be initialized first and destroye... - 01:17 PM Bug #752 (Resolved): High I/O wait when writing data
- Like I said on IRC, I'm seeing a high load on my machine "noisy".
The setup is:
* Intel(R) Xeon(R) CPU 5110 1.6G... - 01:07 PM Bug #715 (Closed): btrfs: extent-tree.c ( btrfs_free_block_group & release_global_block_rsv)
- Closing this one, seems after all this was a real btrfs bug. Fixed in the latest tree.
- 12:19 PM Revision 2a266bd0 (ceph): os: FileStore: Add commit timeout
- Signed-off-by: Colin McCabe <colin.mccabe@dreamhost.com>
- 12:05 PM Bug #746: core dump on radostool failure
- Although it wasn't apparent from my bug report, 027335afe30127f841a5ea875e173ffc4cd7cf91 was a parent revision of 5a1...
- 10:53 AM Bug #746 (Resolved): core dump on radostool failure
- this looks like it is fixed by commit:027335afe30127f841a5ea875e173ffc4cd7cf91.
- 12:01 PM Revision eda48faf (ceph): rbd: Rados::init: clean up after failure
- Signed-off-by: Colin McCabe <colinm@hq.newdream.net>
- 11:27 AM Bug #714 (Won't Fix): ceph -s ,show osd number error
- It still says 6 because the osd existed at one point in time. There isn't a command to mark it as nonexistent curren...
- 10:52 AM Linux kernel client Bug #730: connection resets from kclient
- I'm hoping this is caused by the bad error handling in try_read() and try_write(). Need to do some more testing befo...
- 10:51 AM CephFS Bug #663 (Can't reproduce): cmds segfaults
- Hopefully this will turn up later. The new asserts will let us catch it sooner and get a useful stack dump.
- 10:51 AM Bug #652: ReplicatedPG _scrub missing clone
- Explanation: _scrub in the case that it hits a missing clone currently fails an assert and dies. It might be worth h...
- 10:46 AM CephFS Bug #733 (Can't reproduce): cmds crash: mds/LogEvent.cc:88: FAILED assert(p.end())
- Hmm not enough to go on I'm afraid. If you see this again please let us know!
- 10:42 AM Revision 1ea8f2b3 (ceph): os:FileStore:use std::string rather than huge bufs
- Signed-off-by: Colin McCabe <colin.mccabe@dreamhost.com>
- 10:42 AM Revision 184048cb (ceph): os: FileStore: remove default param
- Signed-off-by: Colin McCabe <colin.mccabe@dreamhost.com>
- 10:42 AM Revision ba5cbb36 (ceph): os: FileStore: ctor should init all class vars
- Signed-off-by: Colin McCabe <colin.mccabe@dreamhost.com>
- 09:24 AM Revision 35442744 (ceph): osd: fix mutual exclusion for _dispatch
- We want only one thread dispatching messages (either new or requeued), so
that we can preserve ordering. Previously ... - 05:33 AM Revision 50976297 (ceph): Merge remote branch 'origin/health2' into unstable
- 01:12 AM Linux kernel client Bug #742: ceph-client.git unstable , tiotest ,printk "timeout, ceph: mds0 caps stale",then oops
- ceph-client unstable:
make menuconfig:
kernel hacking:
i selected :
[*] RCU debugging: sparse-based checks ... - 12:27 AM Linux kernel client Bug #742: ceph-client.git unstable , tiotest ,printk "timeout, ceph: mds0 caps stale",then oops
- git ceph-client-standalone.git master-backport:
if printk "caps stale" ,then cosd blocked.
mon osd mds is at the ... - 12:53 AM Linux kernel client Bug #750: run "dd", printk " libceph: osd1 172.16.10.68:6805 socket closed"
- ceph client driver is from ceph-client.git unstable
- 12:46 AM Linux kernel client Bug #750: run "dd", printk " libceph: osd1 172.16.10.68:6805 socket closed"
- I try to reproduce this issue.
but ceph-client.git unstable + ceph 0.24.2 ,
one mon ,one mds ,two osd at the same h...
01/27/2011
- 09:56 PM Bug #749: test pipe queue if is empty??V0.24.2
- Sage Weil wrote:
> Is there a stack dump in the log (/var/log/ceph/whatever.log)? That will include which revision ... - 09:36 PM Bug #749: test pipe queue if is empty??V0.24.2
- Is there a stack dump in the log (/var/log/ceph/whatever.log)? That will include which revision of the code you're r...
- 09:39 PM Bug #744: osd class changes take too long to propagate
- I think we should use the 'subscribe' mechanism to pass class changes (just the names of updated classes?) to osds. ...
- 09:34 PM Linux kernel client Bug #750: run "dd", printk " libceph: osd1 172.16.10.68:6805 socket closed"
- From the error is sounds like the OSDs are down. Can you include 'ceph -s' output?
- 09:31 PM Bug #751 (Resolved): _dout_lock is used after been disctucted.
- cconf -c /etc/ceph/ceph.conf -l mon
------------gdb-------------------------
common/Mutex.h: In function 'void Mute... - 06:48 PM Revision adffe53e (ceph): units: add signals unit test
- Signed-off-by: Colin McCabe <colin.mccabe@dreamhost.com>
- 04:47 PM Revision bb72d95d (ceph): mds: cluster_fail instead of reset_cluster
- Mark all cluster members as failed, and blacklist. Do not force up/failed
ranks to stopped, as that requires the adm... - 04:34 PM Revision e59007c5 (ceph): mon: add mdsmap DOWN flag to prevent mdsmap updates
- This is intended to be set while doing critical cluster manipulation to
avoid cmds instances from starting up and get... - 04:34 PM Revision 9ebd7735 (ceph): mon: add 'mds reset_cluster' command
- Reset an MDS cluster back to a single node. The idea is:
- wipe out mds journals
- maybe set recovery flag
- mds... - 04:34 PM Revision c3d4014f (ceph): mdsmap: add flags
- Convert unused client_epoch field to flags to avoid a protocol change. It
is always 0 on current clusters. Lucky us!... - 03:54 PM Revision 73e76723 (ceph): .gitignore: vstart generated files
- Signed-off-by: Sage Weil <sage@newdream.net>
- 03:53 PM Revision 0c2ec2d0 (ceph): vstart: put tmp files in /tmp
- Signed-off-by: Sage Weil <sage@newdream.net>
- 10:23 AM Bug #735: Manual drive pull testing hangs filesystem
- We need to be ready to handle unresponsive FileStores in general. Even if the underlying filesystem is 100% perfect (...
- 08:52 AM CephFS Feature #602: mds: handle corrupt/missing journals
- see mds_reset branch.
Still need a process to reset the individual mds instances. This should:
- advance the jou...
01/26/2011
- 10:05 PM Revision 561224e9 (ceph): librados: Remove rados_pool_t& usage, and pointless consts.
- For some reason when I wrote this I passed rados_pool_t by reference
in some functions instead of by value. It's just... - 08:00 PM Linux kernel client Bug #750 (Won't Fix): run "dd", printk " libceph: osd1 172.16.10.68:6805 socket closed"
- Hi
i git ceph-client-standalone.git master-backport
build and insmod it.
ceph server: ceph 0.24.2
OS:linux-2.6... - 06:08 PM Revision 7d65f6ea (ceph): osd: restart if the osdmap client, heartbeat, OR cluster addrs don't match
- If we somehow get ourselves into a situation where the OSDMap addresses do
not match our actual addresses, restart an... - 06:08 PM Revision fbcf6690 (ceph): osd: preserve ordering when ops are requeued
- Requeue ops under osd_lock to preserve ordering wrt incoming messages.
Also drain the waiter queue when ms_dispatch t... - 06:08 PM Revision 47dc27a6 (ceph): osd: avoid extraneous send_boot() calls
- Only send_boot() on osdmap update if we are restarting. Otherwise we can
end up with too many MOSDBoot messages in f... - 05:51 PM Bug #749 (Can't reproduce): test pipe queue if is empty??V0.24.2
- void SimpleMessenger::dispatch_entry()
line 299
#0 0x00007fd99adc96c5 in raise () from /lib64/libc.so.6
Missing s... - 05:29 PM Revision fce3c8d8 (ceph): test_unfound.sh: kill cosds rather than mark out
- For this test, we need to kill cosds rather than mark them as out.
Otherwise, we cannot force objects to become unfou... - 04:04 PM Revision 6685baeb (ceph): disable scrubs during test_unfound
- Signed-off-by: Colin McCabe <colin.mccabe@dreamhost.com>
- 03:45 PM Revision 8aff0dd1 (ceph): mon: implement PGMonitor::get_health
- Signed-off-by: Colin McCabe <colin.mccabe@dreamhost.com>
- 01:54 PM Revision 5c5408bf (ceph): mon: OSDMonitor::get_health: const cleanup
- Signed-off-by: Colin McCabe <colin.mccabe@dreamhost.com>
- 01:47 PM Revision 23a7e88d (ceph): mon: MonitorStore::mkfs: use run_cmd
- Signed-off-by: Colin McCabe <colin.mccabe@dreamhost.com>
- 01:25 PM Feature #748 (Rejected): Design/implement a RADOS copy
- It would be nice if the OSDs could copy objects between themselves, rather than requiring the client to do a read and...
- 01:09 PM Revision 9ecd5fe7 (ceph): os: FileStore: use run_cmd instead of system
- Signed-off-by: Colin McCabe <colinm@hq.newdream.net>
- 01:08 PM Revision b68e1218 (ceph): common: Add run_cmd
- Signed-off-by: Colin McCabe <colinm@hq.newdream.net>
- 12:53 PM Bug #737 (Resolved): don't use system() in daemons
- fixed by commit:b68e121815fa1e551b074899517bf745795a2e75, commit:9ecd5fe7d001455c840e818e4e1f7b1f74bc1017, commit:23a...
- 12:51 PM Feature #618: osd: allow reads from replicas
- todo: add libceph interface to this
- 11:41 AM Linux kernel client Bug #747 (Can't reproduce): ladder0 crash
- ...
- 11:34 AM Bug #746 (Resolved): core dump on radostool failure
- radostool failed because of a problem with some daemons. However, instead of failing gracefully, it dumped core.
O... - 10:53 AM Feature #745 (Resolved): qa: radosgw tests
- creating, removing buckets. creating, fetching, removing objects. listing objects.
- 10:42 AM Bug #741 (Resolved): Removing rbd snapshots causes weird things to happen
- That was fixed at commit:da2d50180dfdc0e30b4348f2acceb2be650f20b7.
- 10:40 AM Bug #744 (Won't Fix): osd class changes take too long to propagate
- The class timeout is too long, we might want to consider some notification mechanism to let osds find out about new c...
- 10:18 AM Bug #743 (Resolved): osd: broken ordering when pg ops are requeued
- commit:fbcf66906e67adbe6769ba7b1853dd0161e977c6
- 10:03 AM Bug #743 (Resolved): osd: broken ordering when pg ops are requeued
- Incoming messages race with requeued ops and get out of order. This is problematic for osd_sub_op's in partricular.
... - 01:02 AM Linux kernel client Bug #742: ceph-client.git unstable , tiotest ,printk "timeout, ceph: mds0 caps stale",then oops
- if printk " ceph: mds0 caps renewed"
,then ,system hang.
[ 3495.283726] libceph: tid 139576 timed out on osd0,...
01/25/2011
- 11:49 PM Linux kernel client Bug #742: ceph-client.git unstable , tiotest ,printk "timeout, ceph: mds0 caps stale",then oops
- run ceph server and ceph client at the same machine.
- 11:48 PM Linux kernel client Bug #742: ceph-client.git unstable , tiotest ,printk "timeout, ceph: mds0 caps stale",then oops
- ceph server : ceph-0.24.2
- 11:47 PM Linux kernel client Bug #742 (Won't Fix): ceph-client.git unstable , tiotest ,printk "timeout, ceph: mds0 caps stale"...
- hi ,
i git ceph-client.git ,checkout to unstable.
build it at ubuntu 10.04 ,make install,reboot,
then run as linux... - 10:36 PM Revision ba998f05 (ceph): ReplicatedPG: _rollback_to fix the just cloned condition
- _rollback_to in the case that head was just cloned and that clone
includes snapid does not need to do anything. Prev... - 10:12 PM Revision bc98a0fb (ceph): MDSMonitor: fix bugs with standby-replay assignment.
- We were accidentally passing gid instead of rank into find_standby_for!
Also, if we got an MDS with rank -1 we went a... - 10:12 PM Revision f19ef8eb (ceph): dumper: rework slightly to prevent incorrect usage of g_conf.id.
- Signed-off-by: Greg Farnum <gregory.farnum@dreamhost.com>
- 05:40 PM Revision 81c0ec1a (ceph): mon: remove PGMap::pg_set
- We don't need an additional data structure to hold the keys to pg_stat.
We can just look at the keys of pg_stat.
Sig... - 05:14 PM Bug #741: Removing rbd snapshots causes weird things to happen
- You're right. Thanks.
- 05:03 PM Bug #741: Removing rbd snapshots causes weird things to happen
- This might be something that we already fixed. The older rbd class (version 1.2) had this issue which was fixed in ve...
- 04:56 PM Bug #741 (Resolved): Removing rbd snapshots causes weird things to happen
- A quick transcript should explain:...
- 05:09 PM Revision 212289f6 (ceph): vstart: Add --standby_mds setting, for auto-creating standby-replays.
- 05:08 PM Revision 5ce82c94 (ceph): Merge branch 'standby_replay' into unstable
- 04:58 PM Revision b17de794 (ceph): mon: PGMap::apply_incremental must maintain pg_set
- Signed-off-by: Colin McCabe <colin.mccabe@dreamhost.com>
- 04:50 PM Revision dba7c1fd (ceph): Merge branch 'testing' into unstable
- Conflicts:
configure.ac
src/Makefile.am
src/common/common_init.cc
src/common/debug.h
src/common/signal.cc
src/c... - 04:38 PM Revision 58637acc (ceph): debian: fix publish.sh for ubuntu
- 03:24 PM Revision 5beec0b0 (ceph): os: readdir_r: read into PATH_MAX-sized buf
- Fix the readdir_r uses in FileStore.cc
Signed-off-by: Colin McCabe <colin.mccabe@dreamhost.com> - 03:24 PM rbd Feature #699 (In Progress): support snapshot notify
- changes done in commit:9c01177349b435186025a088f612a6f5ce2f3de9, still needs to get it upstream.
- 03:22 PM Linux kernel client Feature #738 (Resolved): ino32 mount option
- done, commit:9c01177349b435186025a088f612a6f5ce2f3de9.
- 08:03 AM Linux kernel client Feature #738 (Resolved): ino32 mount option
- expose 32-bit inos via readdir and stat.
- 03:22 PM Linux kernel client Feature #700 (Resolved): rbd: support snapshot notify
- done, commit:fb021e0010a209bca2e4b206f7bfc47379f704e8
- 03:20 PM Feature #298 (Resolved): osd: "watch" operation
- done as of osd commit:7e206dfec07033d49770bafcc2c86f6dd736d7c2, ceph-client commit:fb021e0010a209bca2e4b206f7bfc47379...
- 03:06 PM Bug #727 (Rejected): sysfs: cannot create duplicate filename '/devices/rbd/0'
- Yeah, probably a duplicate of #693, closing it (should be Rejected-Duplicate status but there isn't one yet).
- 03:00 PM Revision 0fbbbad8 (ceph): os: FileStore::mkfs error handling fixes
- Clean up all resources on every exit path. Don't allocate multiple
PATH_MAX buffers on the stack when one will do. Fi... - 02:30 PM Bug #736 (Resolved): Buffer overflow in FileStore::collection_list_partial
- Fixed by commit:5beec0b00ccf1dce3358154367558d9a4e320e93
- 02:28 PM Bug #734 (Resolved): incorrect filestore errno checks
- Addressed in commit:0fbbbad8dd53603f06d226dca2c309236d3d85b8 and commit:6b4ee9ed3ed0f19966269357cac31ea594c0bcd3
- 02:04 PM Bug #722 (Resolved): Fix use of g_conf.id
- Fixed with standby-replay stuff and commit:f19ef8eb03521507cbb4f0071484d1419eafc03d
- 01:36 PM Bug #740: osd's not sharing osdmaps with osd_ping
- this is contention on map_lock. something is holding it for too long.
- 10:56 AM Bug #740 (Can't reproduce): osd's not sharing osdmaps with osd_ping
- on playground, latest testing
- 12:57 PM Bug #735: Manual drive pull testing hangs filesystem
- I have a Quarch box in the lab that I was just pointed to. It has an ssh interface to power cycle drives for failure...
- 08:56 AM Bug #735: Manual drive pull testing hangs filesystem
- Shouldn't btrfs be able to detect that the disk is gone and return appropriate error codes itself, rather than hanging?
- 08:29 AM Bug #735: Manual drive pull testing hangs filesystem
- Yep, this is a problem. The errors are causing btrfs operations to hang instead of return error codes.
What sho... - 10:37 AM Feature #739 (Resolved): 'ceph health' should be warning when pg's are not all active
- 10:32 AM CephFS Bug #733: cmds crash: mds/LogEvent.cc:88: FAILED assert(p.end())
- Odd, I can't repro this anymore. It was either fixed by some change between 785bf0fcbfb69efa8dd97340c8ee0079bb5ad55e ...
- 08:31 AM CephFS Bug #733: cmds crash: mds/LogEvent.cc:88: FAILED assert(p.end())
- Can you restart the mds with 'debug mds = 20' so we can see what events are getting replayed and which decode is fail...
- 09:06 AM Feature #687 (Resolved): Let the monitor auto-set idle MDSes as standbys
- Merged to unstable in commit:5ce82c94a1df22accefba18d04eab9c7672e7c02.
Just set "mds standby replay = true" in the... - 09:05 AM Feature #686 (Resolved): Specify standby behavior in configuration file
- Merged to unstable in commit:5ce82c94a1df22accefba18d04eab9c7672e7c02.
01/24/2011
- 08:53 PM Revision f7572de5 (ceph): v0.24.2
- 07:06 PM Revision bf749eee (ceph): Merge branch 'unstable' into standby_replay
- 06:59 PM Revision 943fd14f (ceph): filestore: don't wait min sync interval on explicit sync()
- Also, if we do wait longer, wait on the same cond.
Signed-off-by: Sage Weil <sage@newdream.net> - 06:59 PM Revision 3a30eb75 (ceph): msgr: include con in debug output
- Signed-off-by: Sage Weil <sage@newdream.net>
- 06:59 PM Revision 4a49a87d (ceph): msgr: make connection pipe reset atomic
- Close a small and unlikely race.
Signed-off-by: Sage Weil <sage@newdream.net> - 06:57 PM Revision b7bc4711 (ceph): MDSMap: split up find_standby_for into multiple functions.
- Usage of this function is rapidly diverging, in terms of what
is desired.
We now have "find_standby_for", which selec... - 06:57 PM Revision 95ed2e4a (ceph): MDSMonitor: Try to assign unassigned standby-replay MDSes during tick()
- We can now specify an MDS as standby-replay and let the monitor
assign it to any MDS. The monitor will only assign it... - 06:57 PM Revision 74ce6e1b (ceph): MDSMonitor: On restarting MDSes; set to standby-replay if appropriate.
- This way, if the primary MDS crashes and is replaced, but is supposed
to standby-replay its secondary on recovery, it... - 06:57 PM Revision 76fbdf17 (ceph): MDSMap: Update/fix print function.
- It previously didn't look at standby_for_name unless standby_for_rank
was set!
Also, we now let it print out standby... - 06:57 PM Revision 69447f79 (ceph): man: Update cmds manual.
- Signed-off-by: Greg Farnum <gregory.farnum@dreamhost.com>
- 06:57 PM Revision a9824b04 (ceph): MDS: MDSMonitor: Make MDS set standby-replay preferences, not MDSMonitor.
- The MDS has more information about its configuration than the MDSMonitor
does. Therefore, encode that information int... - 06:57 PM Revision 1b45088e (ceph): MDSMonitor: be more conservative with use of pending_mdsmap.
- Use the current mdsmap when looking for MDSes to standby-replay for,
as that way we know the other MDS is already up.... - 06:57 PM Revision 785c4987 (ceph): MDSMonitor: remove unused code.
- 06:57 PM Revision f4a07df9 (ceph): MDSMonitor: Don't create new map for standby-replay spam.
- If an MDS is unable to get into the standby-replay state for some
reason (MDS it should be following doesn't exist ye... - 06:57 PM Revision cc207ca5 (ceph): MDSMonitor: Adjust handling of MDSes asking for standby-replay.
- 1) If the MDS does not specify an MDS to follow, we mark them as
standing-by for -2. MDSMap::find_standby_for() has b... - 06:57 PM Revision 54bdac70 (ceph): mds: Adjust replay state changes and options parsing.
- The MDS used to interpret g_conf.id as a rank. It no longer does
so and requires that standby ranks/names be set via ... - 05:45 PM Revision ea2266d0 (ceph): os: fix minor typo in function defs
- Signed-off-by: Colin McCabe <colin.mccabe@dreamhost.com>
- 05:44 PM Revision 6b4ee9ed (ceph): os: fix some obvious error handling problems
- Fix some errors like checking errno when it may not have been set, doing
other operations which may change the value ... - 05:43 PM Bug #737 (Resolved): don't use system() in daemons
- We should replace system() with a simple wrapper function that just does fork + exec.
The use of system() is wrong... - 05:02 PM Bug #376: File corruption after cluster crashes
- if the bug is resolved?
- 04:16 PM Bug #736 (Resolved): Buffer overflow in FileStore::collection_list_partial
- FileStore::collection_list_partial calls readdir_r without actually allocating space for the dirent. A properly craft...
- 03:44 PM Revision 017495db (ceph): Makefile: use CXXFLAGS more consistently
- Signed-off-by: Colin McCabe <colin.mccabe@dreamhost.com>
- 03:44 PM Revision 76a1e09a (ceph): Makefile: remove unecessary header check
- We already check for libcrypto++ using PKG_CHECK_MODULES; we don't need
to fish for header files.
Signed-off-by: Col... - 03:44 PM Revision ffb208e3 (ceph): Makefile: use new Spirit headers where available
- Use new boost::spirit header files where available, to eliminate the
annoying compiler warning on newer systems.
Sig... - 03:43 PM Revision a2a7d039 (ceph): test: Add test_rw
- Test reading and writing lots of objects from the object store.
Signed-off-by: Colin McCabe <colin.mccabe@dreamhost.... - 02:42 PM Feature #619: objecter: optionally read from replicas
- Sigh. Reposting patch for clarity:...
- 02:41 PM Feature #619 (Resolved): objecter: optionally read from replicas
- I ran test_rw.sh with the following patch:
diff --git a/src/osdc/Objecter.h b/src/osdc/Objecter.h
index 37a23fb..... - 12:31 PM Bug #735 (Resolved): Manual drive pull testing hangs filesystem
- It appears that drive failure problems in my configuration are not making their way up through the stack to kill off ...
- 11:27 AM Bug #734 (Resolved): incorrect filestore errno checks
01/23/2011
- 10:56 PM CephFS Bug #733 (Can't reproduce): cmds crash: mds/LogEvent.cc:88: FAILED assert(p.end())
- When I start cmds, I get this crash:
mds/LogEvent.cc: In function 'static LogEvent* LogEvent::decode(ceph::buffer...
01/22/2011
- 10:40 AM Bug #563: osd: btrfs, warning at inode.c ( btrfs_orphan_commit_root )
- Maybe it's "just" a btrfs issue, not a ceph-related one.
Btrfs was known to lack some consistency in metadata alloca... - 02:01 AM Revision 28d08f2d (ceph): config: add new mds_standby options.
- Signed-off-by: Greg Farnum <gregory.farnum@dreamhost.com>
- 02:01 AM Revision a139b24c (ceph): messages: Let MMDSBeacon set_standby_for_name from a c-string.
- Signed-off-by: Greg Farnum <gregory.farnum@dreamhost.com>
- 02:01 AM Revision 4b60223a (ceph): mds: Keep journaler in readonly mode until replay completes.
- Previously we were switching it off for the final non-standby replay
when a standby-replay got activated. This caused...
01/21/2011
- 10:30 PM Revision 785bf0fc (ceph): ReplicatedPG: fix snap_trimmer log version bug
- Previously, ctx->at_version would be the same as ctx->obs->oi.version
leading to the log entry having prior_version =... - 10:20 PM Revision 3e4a82e5 (ceph): FileJournal: don't overflow the journal size.
- Previously we were casting it to a uint64_t, but the left shift
occurs before the cast, so we were overflowing in som... - 06:08 PM Revision aaed6eb3 (ceph): msgr: always start reaper
- If we didn't explicitly bind (i.e. are a client), then we don't start
the accepter. That's fine. But the reaper thre... - 05:35 PM Revision 027335af (ceph): monclient: fix locking
- Hold lock in handle_* methods; assert lock held in all _* methods.
Fixes: #731
Signed-off-by: Sage Weil <sage@newdre... - 03:57 PM Linux kernel client Bug #730: connection resets from kclient
- Sage Weil wrote:
> on the osd side, i see:
> [...]
> the thing is the client side isn't noticing the disconnect. ... - 03:55 PM Linux kernel client Bug #730: connection resets from kclient
- on the osd side, i see:...
- 03:03 PM Revision de013c26 (ceph): msgr: don't need to reinstall signals after daemon
- Signed-off-by: Colin McCabe <colin.mccabe@dreamhost.com>
- 02:53 PM Revision 444e930a (ceph): mds: respawn must unblock signals before exec
- Signed-off-by: Colin McCabe <colin.mccabe@dreamhost.com>
- 02:53 PM Revision 59e8e165 (ceph): common: move signal blocking into signal.cc
- Signed-off-by: Colin McCabe <colin.mccabe@dreamhost.com>
- 02:47 PM Revision ba000d9c (ceph): common: add signal_mask_to_str
- Signed-off-by: Colin McCabe <colin.mccabe@dreamhost.com>
- 02:45 PM Revision 78e31f1d (ceph): mds: respawn must unblock signals before exec
- Signed-off-by: Colin McCabe <colin.mccabe@dreamhost.com>
- 02:41 PM Revision f2ce966b (ceph): common: move signal blocking into signal.cc
- Signed-off-by: Colin McCabe <colin.mccabe@dreamhost.com>
- 02:05 PM Revision a01bc976 (ceph): common: add signal_mask_to_str
- Signed-off-by: Colin McCabe <colin.mccabe@dreamhost.com>
- 12:18 PM Revision 3b6ce334 (ceph): units: RandomRoundTrips: fix endptr for buffer
- Signed-off-by: Colin McCabe <colin.mccabe@dreamhost.com>
- 10:51 AM Revision 942eb0e0 (ceph): workunits: direct_io_test: fix error message
- Signed-off-by: Colin McCabe <colin.mccabe@dreamhost.com>
- 10:08 AM Messengers Bug #732 (Resolved): msgr: leaking sockets when connection retrying?
- commit:aaed6eb3d066144c1e661cc004767afc356f6b37
- 09:44 AM Messengers Bug #732 (Resolved): msgr: leaking sockets when connection retrying?
- ...
- 09:45 AM Bug #731 (Resolved): msgr assert from monclient
- fixed by commit:027335afe30127f841a5ea875e173ffc4cd7cf91
01/20/2011
- 09:52 PM Bug #731 (Resolved): msgr assert from monclient
- ...
- 08:38 PM Revision 715bbcab (ceph): init-ceph: more verbose status
- Signed-off-by: Colin McCabe <colinm@hq.newdream.net>
- 07:18 PM Revision 6a20b83a (ceph): workunits: direct_io_test: add cmdline options
- Signed-off-by: Colin McCabe <colin.mccabe@dreamhost.com>
- 07:07 PM Revision 42709d1b (ceph): workunits: direct_io_test: test more iterations
- Test seeking around in an O_DIRECT file to see if we read what we
should.
Signed-off-by: Colin McCabe <colin.mccabe@... - 06:46 PM Revision f1d7af97 (ceph): workunits: add direct_io_test
- Signed-off-by: Colin McCabe <colin.mccabe@dreamhost.com>
- 06:22 PM Feature #659 (Resolved): direct io unit test
- I added direct_io_test for testing this feature. commit:f1d7af97f3a30d2ecdd0e48f804d8347d319f3eb, commit:42709d1bff91...
- 06:07 PM Revision f33a33e9 (ceph): mds: store migrator lock set over course of export
- We were recalculating the migrator locks at the beginning and end of
export, but we were getting inconsistent sets at... - 04:28 PM Linux kernel client Bug #730 (Closed): connection resets from kclient
- on ladder0 i see lots of...
- 01:53 PM Revision 6c85b10b (ceph): common: pthread_cond_signal: don't need lock
- from the man page:
The pthread_cond_broadcast() or pthread_cond_signal() functions may be
called by a thread whether... - 11:44 AM Revision 189cf33f (ceph): common: re-install sighandlers after daemon()
- Signed-off-by: Colin McCabe <colin.mccabe@dreamhost.com>
- 11:44 AM Revision ad8951ae (ceph): signals: signal.cc: trim includes
- Signed-off-by: Colin McCabe <colin.mccabe@dreamhost.com>
- 11:44 AM Revision 6041302e (ceph): common: move signal handler stuff into signal.cc
- Signed-off-by: Colin McCabe <colin.mccabe@dreamhost.com>
- 11:34 AM Revision ba57520a (ceph): signals: signal.cc: trim includes
- Signed-off-by: Colin McCabe <colin.mccabe@dreamhost.com>
- 10:09 AM CephFS Bug #721 (Resolved): mds crash with multiple mds
- commit:f33a33e9473295f0e6cd60aee17a007542994b1e. was broken by commit:e6ee6c4fbcaadcb6ac0bce99f6955d696459c651
- 06:14 AM Bug #727: sysfs: cannot create duplicate filename '/devices/rbd/0'
- I think there is something else wrong, my "alpha" image seems corrupted somehow:
My virtual machine "alpha" uses t... - 06:05 AM Bug #727: sysfs: cannot create duplicate filename '/devices/rbd/0'
- I'm not sure, but it looks like it was a duplicate indeed.
I compiled the latest RBD driver (master branch: 9aae8f... - 01:44 AM Revision e6b9731d (ceph): ReplicatedPG.cc: fix use of potentially invalid pointer
- rollback_to may not be initialized if ret != 0.
Signed-off-by: Samuel Just <samuel.just@dreamhost.com> - 01:44 AM Revision d1b85e06 (ceph): ReplicatedPG.cc: update coi version and prior_version to match log
- Caused error where oi on clone would not get updated version when snaps
was updated. oi.version would lag behind the... - 01:44 AM Revision 48ebab6d (ceph): ReplicatedPG.cc: fix snap_trimmer object context error
- Previously, snap_trimmer would get the clone object information from the
object store rather than using find_object_c... - 01:44 AM Revision 4e3a4e28 (ceph): ReplicatedPG,PG,OSD: snap_trimmer should run only when the PG is clean
- Signed-off-by: Samuel Just <samuel.just@dreamhost.com>
- 12:28 AM Bug #674: tiobench stress test , OSD timeout
- Hi ,i tried to modify some of codes which can decrease the timeout issue;
For instance:
mount.ceph monip:6789/ ...
01/19/2011
- 10:35 PM Bug #727: sysfs: cannot create duplicate filename '/devices/rbd/0'
- This looks like #693. Would happen when first trying to add unexisting image (the sysfs cleanup was broken), and then...
- 01:19 PM Bug #727 (Rejected): sysfs: cannot create duplicate filename '/devices/rbd/0'
- I just tried to load a RBD image with the RBD module in 2.6.37 (final), I added the device with:...
- 09:15 PM Feature #609: osd: query pool/pg for objects with given xattr
- 09:14 PM CephFS Feature #630: release caps on inodes unlinked by other clients
- I think sending a client_caps to (other) clients with caps notifying them of nlink==0 is the way to do this without c...
- 09:10 PM Bug #729 (Closed): weird kernel BUG on metropolis
- This is a known problem with ext3 and xattrs on 2.6.32. Either upgrading to a newer kernel (2.6.34 or later IIRC?) o...
- 05:58 PM Bug #729 (Closed): weird kernel BUG on metropolis
- I'm not sure if this is our issue, or an ext3 issue. Anyway, it came up on metropolis, and I need write down the info...
- 07:20 PM Revision 79c5dc99 (ceph): mds: use direct replay test when deciding whether to rebalance.
- The previous use of standby_for_rank testing was prone to errors
and I think would have ended up causing bugs if it w... - 06:48 PM CephFS Bug #723 (Resolved): mds: cmds ignores signals (like SIGTERM) after respawn
- fixed by commit:afc9ae068caeeab1e58f96a95e91eb7cb0384ab3
- 06:46 PM Revision 8e77b356 (ceph): test: add new unit test "gather" for C_Gather.
- Signed-off-by: Greg Farnum <gregory.farnum@dreamhost.com>
- 05:50 PM Revision 457e3e09 (ceph): mds: fix journaling of root default_file_layout
- We need to include the default_file_layout (if any) on root inodes, too.
Fixes: #725
Signed-off-by: Sage Weil <sage@... - 05:25 PM Revision 820c54fe (ceph): common: re-install sighandlers after daemon()
- Signed-off-by: Colin McCabe <colin.mccabe@dreamhost.com>
- 05:22 PM Revision 2f72156e (ceph): common: move signal handler stuff into signal.cc
- Signed-off-by: Colin McCabe <colin.mccabe@dreamhost.com>
- 01:44 PM Feature #728 (Closed): automate package builds
- current process for .debs is, in root of ceph.git,
checkout proper branch
$ ./release.sh <stable|testing|unstable... - 01:14 PM Revision 35ef7bc9 (ceph): signals: handle_fatal_signal: use SA_NODEFER
- SA_RESETHAND | SA_NODEFER allows the "re-trigger default signal handler"
trick to work for signals other than SIGSEGV... - 01:14 PM Revision 3326b753 (ceph): signals: backtrace some more exotic fatal signals
- We're not likely to see these, but if we do, we want it in the logs!
Signed-off-by: Colin McCabe <colinm@hq.newdream... - 01:14 PM Revision 22e20614 (ceph): signals: Handle SIGILL, SIGBUS, SIGFPE.
- Print out a backtrace when we get SIGILL, SIGBUS, or SIGFPE. Fix a bug
where we failed to install a SIGABRT handler.
... - 12:16 PM Revision 047ed152 (ceph): cephtool: handle init failures correctly
- Signed-off-by: Colin McCabe <colin.mccabe@dreamhost.com>
- 12:08 PM Revision b13efe2c (ceph): cephtool: fix usage a bit
- Signed-off-by: Colin McCabe <colin.mccabe@dreamhost.com>
- 12:08 PM Revision 1b46b0d8 (ceph): common: fix common_usage a bit
- Signed-off-by: Colin McCabe <colin.mccabe@dreamhost.com>
- 11:21 AM Bug #711 (Resolved): safetimer teardown crash if auth fails
- fixed by commit:047ed152a3112e47c6e1adf2c526aa46245b77a3
- 10:28 AM Cleanup #726 (Closed): Make libcommon self-sufficient
- While writing a unit test for C_Gather I noticed that libcommon doesn't seem to be self-sufficient. I tried linking m...
- 09:51 AM CephFS Bug #725 (Resolved): mds: set_layout on root inode isn't persistent
- commit:457e3e09bc78c297f83f0e85757a4d238a1da968
- 09:32 AM CephFS Bug #725 (Resolved): mds: set_layout on root inode isn't persistent
- ...
01/18/2011
- 11:13 PM Revision 6152f522 (ceph): mon: remove rank from failed when taking over for failed node
- Leaving it there leaves a broken MDSMap, and prevents rejoin because
MDSMap::is_rejoining() is always false.
Signed-... - 11:09 PM Revision e45d8590 (ceph): mds: kick discovers when peers enter active|clientreplay|rejoin
- We process discovers when active, clientreplay, or later stages of rejoin.
Wait until then to resend pending discover... - 09:27 PM Revision e276683d (ceph): mon: fix 'ceph mds fail <N>' command
- We need to remove the mds_info from the map for cmds to take notice.
Fixes: #720
Signed-off-by: Sage Weil <sage@newd... - 09:16 PM Revision 7e99efc4 (ceph): PG: fix adjust_local_snaps bug
- current must be removed from to_remove in the loop for the loop to
terminate (and not cause a double erasure from sna... - 06:22 PM Revision d69e5f51 (ceph): units: base64 should test more round-trips
- Test 1000 round trips.
Signed-off-by: Colin McCabe <colin.mccabe@dreamhost.com> - 06:10 PM Revision 88530e13 (ceph): units: better names for the base64 test
- Signed-off-by: Colin McCabe <colin.mccabe@dreamhost.com>
- 05:39 PM Revision 219a1c1f (ceph): units: don't need -lcrypto
- Signed-off-by: Colin McCabe <colin.mccabe@dreamhost.com>
- 03:57 PM Revision 08110f10 (ceph): mon: Allow ceph mds fail <name>
- Formerly we only allowed "ceph mds fail <rank>", and also failed to
check if rank was correctly parsed.
Signed-off-b... - 03:56 PM Revision e5b78695 (ceph): health: fix health display for MDSes
- When there are laggy MDSes, we should set health to WARN. Failed MDSes
should set health to FAIL. Stopped MDSes shoul... - 03:56 PM Revision 93a729f2 (ceph): mds: add comment explaining MDSMap::max_mds
- Signed-off-by: Colin McCabe <colin.mccabe@dreamhost.com>
- 03:37 PM Revision 9e7a2b63 (ceph): Merge branch 'purged_snaps' into testing
- 03:27 PM Linux kernel client Bug #227 (Resolved): caps flush races with caps export
- 03:24 PM Linux kernel client Bug #471 (Can't reproduce): NULL pointer dereference __list_add+0x42/0x89 kick_requests+0x24/0x9e
- 03:24 PM Linux kernel client Bug #477 (Can't reproduce): kernel BUG at fs/inode.c:295
- 03:24 PM Linux kernel client Bug #651 (Resolved): osd_client: need to recalculate request mapping for every osdmap incremental
- fix in usntable, commit:05d7263709810c998d396a47dcf09ccb2f643c90 'libceph: fix osd request queuing on osdmap updates'
- 03:23 PM Linux kernel client Bug #718 (Resolved): sync hangs
- fixed in unstable branch, commit:55ee8fe37598475ed363d078cb50d19e0524c69f 'ceph: fix flushing of caps vs cap import'
- 03:19 PM CephFS Bug #724 (Rejected): mds: cmds ignores signals (like SIGTERM) after respawn
- Duplicate of #723! :)
- 02:04 PM CephFS Bug #724 (Rejected): mds: cmds ignores signals (like SIGTERM) after respawn
- To reproduce:
- vstart
- ceph mds fail 0
- wait for mds.a to respawn and come back up (replay, reconnect, etc.)
-... - 02:49 PM Bug #720 (Resolved): marking an MDS that is operational as failed causes an assert
- 01:29 PM Bug #720: marking an MDS that is operational as failed causes an assert
- commit:e276683d196c725cdd317342f8b8962a83d12bad
- 12:08 PM Bug #720 (Resolved): marking an MDS that is operational as failed causes an assert
- How to reproduce:
1. start all cluster nodes
2. ./ceph mds fail a
Backtrace:... - 01:36 PM CephFS Bug #723 (Resolved): mds: cmds ignores signals (like SIGTERM) after respawn
- To reproduce:
- vstart
- ceph mds fail 0
- wait for mds.a to respawn and come back up (replay, reconnect, etc.)
-... - 01:33 PM Bug #722 (Resolved): Fix use of g_conf.id
- There are a few things I've added that incorrectly interpret g_conf.id as a rank: Dumper and cmds standbys and maybe ...
- 12:47 PM CephFS Bug #721 (Resolved): mds crash with multiple mds
- Multiple mds (3). Running commit:4385aa567de31d0ebe46e25eac802f329cf0851d.
Using the kernel client, doint a simple... - 12:17 PM Bug #713 (Resolved): ceph health doesn't notice downed mds
- fixed by commit:cf25cd9e0b5eccb69cdb77433116de5c8cda7dd2
- 11:22 AM Bug #702 (Resolved): osd: failed interval_set assertr in snap_trimmer FAILED assert(p->first > st...
- 10:51 AM Revision 9077fa5a (ceph): automake: check for libedit
- Signed-off-by: Colin McCabe <colin.mccabe@dreamhost.com>
- 09:56 AM Bug #688 (Resolved): should add dependency on libedit to rpm, deb, configure.ac
- resolved by commit:9077fa5aeaba53975909bacc6a30a5d683ed6cd9
(the rpm and deb already depend on libedit) - 09:14 AM Feature #719 (Resolved): mds journal dump should be simpler
- Right now the dumper just creates a file which contains the contents of the mds journal from read_pos to the probed e...
- 04:07 AM Revision c8d1029d (ceph): osd: avoid polluting peer_info with acting[0]
- This can confuse code that iterates over peer_info and finds the primary
(with bad info).
Signed-off-by: Sage Weil <... - 12:50 AM Bug #704: it hang both in client and osd
- it is strange, previous osd0 hang,this time osd1 hang.....
- 12:43 AM Bug #704: it hang both in client and osd
- this time ,i mkcephfs again.
at first mount.ceph is ok,i cp fine /mnt/ceph ,this is ok,but sync hang,you can ls now....
01/17/2011
- 11:47 PM Revision 546d65f1 (ceph): osd: make misdirected request warnings specify osdmap epoch
- Signed-off-by: Sage Weil <sage@newdream.net>
- 11:47 PM Revision b9b8d7c8 (ceph): vstart: set pgp_bits to a sane value
- We need to specify pgp_bits as well or else we get the default (which makes
no sense when it's way bigger than pg_bit... - 11:46 PM Revision 0a3eec1f (ceph): osd: disallow pgp_num > pg_num on osdmap creation
- Signed-off-by: Sage Weil <sage@newdream.net>
- 11:44 PM Revision fcab5e81 (ceph): Revert "objecter: use raw_pg_to_pg when needed"
- This reverts commit cfd87ceefb46358adaa1751975c8d3a6b063bdf9.
The client side behavior here is correct: we should fe... - 11:36 PM Revision 7e3d2ae0 (ceph): mon: make 'pg map <pgid>' output more explicit
- Show effective pgid in parens.
Signed-off-by: Sage Weil <sage@newdream.net> - 09:38 PM Revision 6f946c86 (ceph): Merge branch 'testing' into unstable
- Conflicts:
src/mds/journal.cc - 09:23 PM Revision afefd737 (ceph): osd: rebind heartbeat_messenger (with cluster one) when wrongly marked ...
- This keeps things clean.
Signed-off-by: Sage Weil <sage@newdream.net> - 09:22 PM Revision d27a0aaf (ceph): messenger: let rebind() avoid multiple ports
- We need to rebind two messengers, which means avoiding both old ports.
Signed-off-by: Sage Weil <sage@newdream.net> - 09:20 PM Linux kernel client Bug #718: sync hangs
- on v2.6.37,...
- 08:40 PM Linux kernel client Bug #718 (Resolved): sync hangs
- There were a few reports, and I'm also seeing it on current unstable. Haven't checked other branches yet.
In one ... - 09:18 PM Revision 818fa33a (ceph): osd: drop messages from before we moved back to boot state
- We want to make sure we ignore any messages sent to us before we moved
back to the boot state (after being wrongly ma... - 04:03 PM Feature #717 (Resolved): cephtool: make -s/-w use subscribe instead of paxos watch; deprecate pax...
- We need to transition the ceph tool to the subscribe framework and away from the raw paxos watch framework. This wil...
- 01:48 PM Bug #716 (Resolved): incorrect RESETSESSION hangs osd peering after wrongly marked down
- fixed by commit:818fa33a6610ba85903e647addaf84ec71ae9653
- 01:38 PM Bug #716 (Resolved): incorrect RESETSESSION hangs osd peering after wrongly marked down
- Saw this on playground during repeering. Cause seems to be that we _share_map_incoming on the rebound messenger whil...
- 01:24 PM Bug #710 (Resolved): cauthtool --create-keyring generates bad keyring
- 03:04 AM Bug #715 (Closed): btrfs: extent-tree.c ( btrfs_free_block_group & release_global_block_rsv)
- On 'noisy' I'm trying to run bonnie++ inside a Qemu-RBD virtual machine. After a while this gets stuck, the VM become...
01/16/2011
- 09:57 PM Bug #714 (Won't Fix): ceph -s ,show osd number error
- i have 6 osds,i down osd4,out osd4,but ceph -s show still 6 osds, 5 up, 5 in.it should be 5 osds,5 up ,5 in
how to ... - 09:35 PM Support #675: hot to adjust replicate level,disk not the same size
- and at highest level debug
- 09:34 PM Support #675: hot to adjust replicate level,disk not the same size
- wo do not know what you mean,can you take above as example, 50G,and 3T,how to adjust device weight????
i try several... - 09:29 PM Bug #704: it hang both in client and osd
- at the first time ,i mkcephfs -a --mkbtrfs.client can mount.ceph,
but if you ls /data/osd0 at osd0-node,it will hang... - 11:28 AM CephFS Bug #663: cmds segfaults
- The ceph test-cluster has since been replaced by MooseFS. I'm sorry that I can't provide any further useful feedback ...
01/15/2011
- 11:52 PM Revision 8aa4a957 (ceph): Merge remote branch 'origin/overload2' into unstable
- 04:07 PM Feature #95 (Resolved): mon: adjust overload based on osd disk utilization
- merged in commit:8aa4a9576857ce796fc9de1b0b9ae2144a830cb1
- 04:07 PM RADOS Feature #433 (New): improve osd reweighting
- Sage Weil wrote:
> merged in commit:8aa4a9576857ce796fc9de1b0b9ae2144a830cb1
whoops, wrong bug, ignore that.
- 04:06 PM RADOS Feature #433 (Resolved): improve osd reweighting
- merged in commit:8aa4a9576857ce796fc9de1b0b9ae2144a830cb1
- 04:04 PM Feature #619: objecter: optionally read from replicas
- see objecter_balance_reads branch for untested client-side implementation
- 01:36 AM Revision ace8ef90 (ceph): Make non-runnable source files not have execute bit set.
- 01:27 AM Revision 5a0bc6b7 (ceph): Sanitize environment before running clitests.
- This avoids CEPH_KEYRING etc from slipping in.
- 01:27 AM Revision 886c7d35 (ceph): Command line -c should still override CEPH_CONF from env.
- 01:27 AM Revision ba4de2e1 (ceph): Add TODO note for cauthtool -c conflict.
- This triggered in unrelated ways while testing CEPH_CONF
support. - 01:27 AM Revision fa7e469c (ceph): Make clitests avoid existing ceph.conf files.
- 01:27 AM Revision 3c656ad5 (ceph): Make run-cli-tests not abort on first failure.
- 12:57 AM Revision 0c6fb520 (ceph): cfuse: fix crash when bad address provided
- 12:34 AM Revision 7d7af85c (ceph): auth: new rotating secret ttl should depend on now() + ttl
- Before it only depended on the previous rotating secret (which was
always bigger than g_clock.now()). Since the ticke... - 12:33 AM Revision ae47f3c3 (ceph): Merge remote branch 'newdream/unstable' into unstable
- 12:27 AM Revision 58f6d09e (ceph): Also distribute pre-downloaded tarballs.
- Found by "make distcheck".
- 12:27 AM Revision a549955e (ceph): Include the cli tests in a dist tarball.
- Found by "make distcheck".
- 12:27 AM Revision 95d7d6e2 (ceph): Clean up test virtualenv on "make clean".
- Found by "make distcheck".
- 12:27 AM Revision dab99b04 (ceph): Patch cram so it is able to write *.err files outside the source tree.
- https://bitbucket.org/brodie/cram/issue/9/allow-read-only-directories-for-t
Found by "make distcheck". - 12:27 AM Revision bdc1e260 (ceph): Make git ignore more generated files in src/gtest.
- 12:27 AM Revision 0259e57b (ceph): Make git ignore *.t.err files generated by clitests.
- 12:27 AM Revision a90329ae (ceph): Revert "unit tests: do standard ceph init before tests"
- This reverts commit 77f90978dc38d025e05300f7650a550ae15704fa.
Unit tests should not parse the normal "-c ceph.conf" ... - 12:27 AM Revision 5045695c (ceph): Teach run-cli-tests about build dirs.
- An optional argument points can tell it where to put
generated files (in this case, virtualenv). Provide
the argument... - 12:22 AM Revision 4385aa56 (ceph): MDS: Use new C_Gather::get_num_remaining() in MDCache.
- It was using get_num(), which now reports the number created.
This probably wouldn't have worked previously except th... - 12:12 AM Revision a93b970a (ceph): C_Gather: Set debug #ifdefs to remove set.
- This way when we're confident it works right, we can
remove the set<Context*> and just rely on ref counting.
Further... - 12:11 AM Revision 55cf6bad (ceph): C_Gather: Rewrite for thread safety.
- Previously, C_Gather wasn't thread safe at all,
and there was an issue with creating subs while some
subs were being ...
01/14/2011
- 11:24 PM Revision d8073c8b (ceph): Include run-cli-tests in release tarball.
- Found by "make distcheck".
- 11:24 PM Revision f599b0b6 (ceph): run-cli-tests is in srcdir not in build dir.
- Found by "make distcheck".
- 11:24 PM Revision b0fc2ca0 (ceph): Rename variable in run-cli-tests.
- Emphasize the fact that the path is the source dir, not the build dir.
- 11:22 PM Revision 52e1edb0 (ceph): Merge branch 'tests-broken' into unstable
- 11:21 PM Revision 5f239fa6 (ceph): Fix clitests for cconf usage change.
- 11:08 PM Revision 29825c75 (ceph): mds: call MonClient::shutdown when doing a journal dump.
- Previously we got a failed assert since nothing was calling this.
Signed-off-by: Greg Farnum <gregf@hq.newdream.net> - 11:07 PM Revision 986e5c19 (ceph): Merge branch 'tests-broken' into unstable
- 11:06 PM Revision 62d9de47 (ceph): Fix clitests for cauthtool usage change.
- 11:04 PM Revision 73dd32c2 (ceph): Merge commit 'cfae10b8f8b0d91f37dc6eb72f3b3f8285bb15e7' into tests-brok...
- 11:01 PM Revision 1d7460a9 (ceph): Plaintext keyring format is supposed to be user-friendly, so test it.
- 11:00 PM Revision 0d0e03e0 (ceph): Now that cauthtool has two kinds of keyrings, test them both.
- 10:33 PM Revision 67bec9c4 (ceph): Fix a bug where "cauthtool --create-keyring" (no --bin) wrote garbage.
- This only triggered when running without --gen-key or --add-key.
- 06:08 PM Revision 1bae352e (ceph): os: don't crash on no-journal case
- JournalingObjectStore::commit_start should handle the case where journal is
null. This will occur if the user doesn't... - 05:22 PM Bug #713 (Resolved): ceph health doesn't notice downed mds
- The output of 'ceph health' doesn't seem to reflect the status of the mds nodes. I stopped first one and then both m...
- 04:54 PM Bug #674: tiobench stress test , OSD timeout
- added tiotest to qa/workunits
- 04:50 PM CephFS Bug #712: cfuse crashes when Ceph daemons started with vstart.sh
- ahmm.. that is:
./cfuse -m 127.0.0.1:6789 /mnt/ceph - 04:47 PM CephFS Bug #712 (Resolved): cfuse crashes when Ceph daemons started with vstart.sh
- should be:
./cfuse -m 127.0.0.1:6789:/ /mnt/ceph
Anyway, commit:0c6fb520dc0c88bf39a1859ff9aa6d8ab1db70d5 fixes th... - 04:34 PM CephFS Bug #712 (Resolved): cfuse crashes when Ceph daemons started with vstart.sh
- There appears to be a regression in cfuse as of v0.24.1, which was not present in v0.23. Starting a local "cluster" ...
- 04:28 PM Bug #462 (In Progress): cephx: verify_authorizer_reply exception in decode_decrypt
- Shouldn't happen any more with commit:7d7af85c3a8b90ef93a152131a4b8fd66ceb6996. We always keep 3 ("rotating") keys fo...
- 04:10 PM Revision c85d5ead (ceph): common: Add CEPH_CONF override
- Signed-off-by: Colin McCabe <colin.mccabe@dreamhost.com>
- 03:56 PM Bug #711 (Resolved): safetimer teardown crash if auth fails
- ...
- 02:53 PM Bug #710: cauthtool --create-keyring generates bad keyring
- Fixed in 67bec9c42373dd70d4d6fbc8364b8e811256c0bb.
- 02:26 PM Bug #710 (Resolved): cauthtool --create-keyring generates bad keyring
- When running cauthtool --create-keyring and without specifying --add-key, the resulted keyring is corrupt (instead of...
- 01:58 PM Revision 77f90978 (ceph): unit tests: do standard ceph init before tests
- Signed-off-by: Colin McCabe <colin.mccabe@dreamhost.com>
- 12:38 PM Revision 512cc974 (ceph): qa: Add tiobench test
- Signed-off-by: Colin McCabe <colinm@hq.newdream.net>
- 11:35 AM Revision 9af160c5 (ceph): mon: Implement reweight-by-utilization
- Signed-off-by: Colin McCabe <colinm@hq.newdream.net>
- 11:14 AM Revision 6f36e9fe (ceph): osd: adjust_osd_weights: small const cleanup
- Signed-off-by: Colin McCabe <colinm@hq.newdream.net>
- 11:14 AM Revision 97b8eebc (ceph): mon: some PGMap cleanups
- Const cleanup. Don't do multiple lookups for the same item in a
container.
Signed-off-by: Colin McCabe <colinm@hq.ne... - 11:13 AM Revision 08caffe7 (ceph): vstart.sh: don't depend on /usr/bin/host
- Signed-off-by: Colin McCabe <colinm@hq.newdream.net>
- 10:14 AM Feature #709 (Resolved): nagios plugin
- Make a generic nagios plugin (for widespread use) that uses 'ceph health' to check status.
It should let you con... - 09:58 AM Bug #665 (Closed): mds(uo:replay),clent can not mount,return error 5 input/output
- To diagnose this sort of problem we need to see 'ceph -s' (to know if osds are up, etc.). If they are, then the mds ...
- 09:54 AM Bug #664 (Closed): osd auto dead,zombie state.
- Not sure what this would be. Usually I see things like this when a process had a large unlinked file open, and the d...
- 09:52 AM Support #675 (Closed): hot to adjust replicate level,disk not the same size
- The crush weight should be proportional to the disk size, or node throughput, whichever you prefer. It depends on if...
- 09:50 AM Bug #704: it hang both in client and osd
- can you include 'ceph -s' output?
- 09:49 AM Bug #703 (Resolved): 0.23.1update to 0.24.1,old ceph can not run
- The no journal bug was recnetly fixed in unstable branch, see commit:1846355176ecd5b60e93da8cf20cea0ef6ed4326, cherry...
- 09:34 AM CephFS Bug #663: cmds segfaults
- 09:24 AM Bug #694 (Resolved): no osd journal causes cosd segfault
- 06:08 AM Revision 3f4434f7 (ceph): mds: use common helper to journal a client session close
- We saw a bug where an ESession close was followed by an EMetaBlob on that
session (see 6d0dc4bf64b2792d6fc007268c5a42... - 06:08 AM Revision 6d0dc4bf (ceph): mds: tolerate (with warning) replayed op with bad prealloc_inos
- This comes up when an ESesssion close is followed by an EMetaBlob that
uses a prealloc_ino. That isn't supposed to h... - 05:51 AM Revision 86337127 (ceph): mds: improve debug output on ESession journal replay
- Signed-off-by: Sage Weil <sage@newdream.net>
- 12:20 AM Revision b60ef3a7 (ceph): OSD,ReplicatedPG: Do not run snap_trimmer while the pg is degraded
- snap_trimmer causes replica crashes if the replica is missing
objects.
Signed-off-by: Samuel Just <samuelj@hq.newdre... - 12:11 AM Revision 2b2d52cf (ceph): config: keyring uses a search path again
- 12:11 AM Revision 58a0f384 (ceph): cauthtool: default keyring format is plaintext, add --bin
- 12:11 AM Revision cfae10b8 (ceph): keyring: default keyring file name is 'keyring'
- update accordingly the tools, scripts, man page
- 12:11 AM Revision ef8a60fd (ceph): cconf: fix usage parsing, add --resolve search
- --resolve-search is used to resolve a search path result
01/13/2011
- 11:47 PM Revision 62024d04 (ceph): qa: add xattr check
- Signed-off-by: Sage Weil <sage@newdream.net>
- 10:04 PM CephFS Bug #708 (Resolved): mds replay crash on slide0
- fixed by commit:3f4434f7b9e02ea092f277b448bcb9b8edd5b143 and commit:6d0dc4bf64b2792d6fc007268c5a42ae4e2e583c
- 05:22 PM CephFS Bug #708 (Resolved): mds replay crash on slide0
- The mds on slide0 won't start on slide0 (or slide1, it seems) and this is the end of the log file before it exits. T...
- 09:43 PM Revision cd3b35a3 (ceph): PG: added adjust_local_snaps, activate now checks local collections
- adjust_local_snaps handles removing local collections contained in
to_check. On activate, pg will now remove local c... - 09:43 PM Revision 6e610a90 (ceph): OSD: _pg_process_info refactor to use adjust_local_snaps
- Changes _pg_process_info to use adjust_local_snaps. Also accounts for
the incoming info not being a superset of the ... - 09:43 PM Revision 1b3ed033 (ceph): ReplicatedPG: snap_trimmer skip removed snaps without collections
- If no writes are made between two snapshots, the first won't get a snap
collection. Subsequently removing that snap ... - 09:43 PM Revision cf39db24 (ceph): PG: change snap_collections to an interval_set
- Previously, the set of local snap collections was represented using a
set, which complicates set operations with inte... - 09:24 PM Revision f650e3da (ceph): Merge branch 'testing' into unstable
- Conflicts:
configure.ac - 09:16 PM Revision f2755a53 (ceph): PG: activate should not enqueue snap_trimmer on a replica
- Previously, activate would queue_snap_trim() for replicas if snap_trimq
ended up non-empty, guaranteeing a crash for ... - 09:14 PM Revision e060d7a1 (ceph): filejournal: rewrite completion handling, fix ordering on full->notfull
- Rewriting the completion handling to be simpler, clearer, so that it is
easier to maintain a strict completion orderi... - 07:32 PM Revision 1869ffc0 (ceph): Fix confusing comment about gtest automake hookup.
- 06:34 PM Revision 4b2ffba9 (ceph): unit: add IncorrectBase64Encoding test
- Signed-off-by: Colin McCabe <colinm@hq.newdream.net>
- 06:23 PM Revision e7153f38 (ceph): unit: Add test/base64.cc
- Signed-off-by: Colin McCabe <colinm@hq.newdream.net>
- 05:50 PM Revision 983e62c8 (ceph): Add a target to build but not run unittests.
- Use with "make -C src unittests".
- 04:58 PM Revision 34adf360 (ceph): common: fix buffer::list::decode_base64
- buffer::list::decode_base64 needs to check for decode failures.
Signed-off-by: Colin McCabe <colinm@hq.newdream.net> - 03:58 PM Feature #705 (Resolved): plaintext keyring files
- Done, concluded by commit:cfae10b8f8b0d91f37dc6eb72f3b3f8285bb15e7.
- 03:57 PM Bug #707 (Resolved): cconf crashes when no args supplied
- 03:54 PM Bug #707: cconf crashes when no args supplied
- Fixed, commit:ef8a60fd8577091352f4d9d0893d659ca7e28e58.
- 02:14 PM Bug #707 (Resolved): cconf crashes when no args supplied
- probably missing a simple check after the args parsing
- 01:08 PM Bug #666 (Resolved): OSD: ReplicatedPG::eval_repop
- fixed by commit:e060d7a115ff1477de0024d24fdc0397dc4c97b1
- 08:03 AM Bug #703: 0.23.1update to 0.24.1,old ceph can not run
- it is because no osd journal.
01/12/2011
- 11:13 PM Revision 1cdb01b4 (ceph): ReplicatedPG: Fix oi.size bug in _rollback_to
- _rollback_to calls _delete_head before cloning the clone into place.
_delete_head sets the object info size to 0. _r... - 09:51 PM Revision 9c80239b (ceph): ReplicatedPG: register_object_context and register_snapset_context cleanup
- Previously, get_object_context and get_snapset_context did not register
the resulting objects. In some cases, these ... - 09:28 PM Revision 688bdca2 (ceph): Fix src/test/run-cli-tests to work with any cwd.
- 08:32 PM Bug #704: it hang both in client and osd
- at mds0
#cat /proc/469/statck
futex_wait)queue)me+0xc5/0xe4
futex_wait+0x143/0x2f9
do_futex+0x9c/0x852
sys_futex... - 08:25 PM Bug #704: it hang both in client and osd
- #ls /data/osd0
cat /proc/31361/statck
__mutex_fastpath_lock_retval+0x18/0x1a
vfs_readdir+0x59/0xb2
sys_getdents+0... - 08:07 PM Revision 8f327d11 (ceph): ReplicatedPG: snap_trimmer work around
- Currently, an OSD bug is causing snap_trimq to contain some snaps
already in purged_snaps. This work around should l... - 08:04 PM Revision 7ab1f75e (ceph): Merge branch 'clitests-3' into unstable
- 07:10 PM Revision 903ed4af (ceph): Adjust clitests after cauthtool changes.
- 06:58 PM Revision 96424a09 (ceph): Merge commit '735eb400dc617c599f8cb42af91bab00931eeaff' into clitests-z
- 06:52 PM Revision f3bfbda1 (ceph): Adjust clitests after cauthtool changes.
- 06:49 PM Revision de3b9e74 (ceph): Merge commit 'e9a70f15029d397ebf0414e5f16fda321af5f55b' into clitests-4
- 06:24 PM Revision d3505a38 (ceph): Fix osdmaptool error reporting.
- 06:21 PM Revision cf54bd8c (ceph): osd: prevent PG objects from being copied
- Signed-off-by: Colin McCabe <colinm@hq.newdream.net>
- 06:21 PM Revision 095574b0 (ceph): osd: call prior_set_affected only if we have one
- Don't call prior_set_affected if the prior set hasn't been built. This
will be the case unless we're a primary doing ... - 06:21 PM Revision 62a5c206 (ceph): osd: remove unused PG global
- Signed-off-by: Colin McCabe <colinm@hq.newdream.net>
- 06:21 PM Revision 039aa1e7 (ceph): osd: clean up loop in proc_replica_log
- We don't need to update lu on (almost) every iteration, only on the
final one. Use a const iterator.
Signed-off-by: ... - 06:21 PM Revision 6fedbf8b (ceph): osd: de-globalize PG::oldest_update
- Making oldest_update a class variable complicates log merging and wastes
space in the PG struct. Even though memory i... - 06:21 PM Revision 3a822d6d (ceph): osd: avoid creating some temporary coll_t objects
- PG::coll caches the value of coll_t(this->info.pgid). So use PG::coll
when appropriate rather than constructing a new... - 06:21 PM Revision 70f3c2c2 (ceph): conf: ConfFile can parse bufferlists
- 06:21 PM Revision 0568bee8 (ceph): Revert "client: Remove the I_COMPLETE flag from the parent directory in...
- This reverts commit c43455cee4b7b45de6bd04454a40bc7016f2d6d1. We don't
need this fix any more since we now handle har... - 06:21 PM Revision 735eb400 (ceph): auth: change the plaintext keyring format
- 06:21 PM Revision 838ebddb (ceph): uclient: Switch how inodes link to dentries a bit.
- Inodes now have a set of parent dentries, rather than a single
pointer. This allows the cache to accurately represent... - 06:21 PM Revision 540f42ce (ceph): osd: Put all prior_set fields into a struct
- Keep all the prior set stuff together.
Signed-off-by: Colin McCabe <colinm@hq.newdream.net> - 05:18 PM Revision fad09722 (ceph): Remove outdated TODO note.
- The test originally used cat <<EOF, which made
the mon$id in the config file get expanded at
that time. - 04:46 PM Linux kernel client Bug #706 (Resolved): large dir size even with norbytes on 32-bit
- Fixed, commit:db5f8e20aa9294bd6db2d1635c45c98cf42e282a.
- 04:06 PM Linux kernel client Bug #706 (Resolved): large dir size even with norbytes on 32-bit
- see Jan's email
- 11:50 AM Feature #705 (Resolved): plaintext keyring files
- Keyring files should be easier to modify and not require the cumbersome cauthtool for that. This should ease installa...
- 09:24 AM CephFS Cleanup #588 (Resolved): Allow Inodes to have multiple parent Dentries
- Merged to unstable in commit:b6271764cda3652d7fb0e3472732ae31c1adbbe9.
- 12:43 AM Revision f4a0c2f8 (ceph): Add CLI tests for osdmaptool and friends.
- Uses a python package "cram" as test runner.
Requires PIP (python-pip.deb) installed on the
build machine, to actuall...
01/11/2011
- 11:16 PM Revision 9a5a7fe4 (ceph): Git ignored files cleanup.
- Make gitignore entries not match recursively.
I wanted to introduce a directory "osdmaptool" to contain cli tests
fo... - 10:49 PM Bug #704 (Closed): it hang both in client and osd
- 1.# sync ;in client,it hang ......never return
2.in osd0,
# ls /date/osd0 ;hang there,never return. - 09:21 PM Bug #703: 0.23.1update to 0.24.1,old ceph can not run
- Core was generated by `/usr/bin/cosd -i 0 -c /tmp/ceph.conf.7906'.
Program terminated with signal 11, Segmentation f... - 09:06 PM Bug #703: 0.23.1update to 0.24.1,old ceph can not run
- Core was generated by `/usr/bin/cosd -i 0 -c /tmp/ceph.conf.7906'.
Program terminated with signal 11, Segmentation f... - 08:54 PM Bug #703: 0.23.1update to 0.24.1,old ceph can not run
- meanwhile,osd dead,but ceph -s show 2 osds is up and in.
- 08:50 PM Bug #703 (Resolved): 0.23.1update to 0.24.1,old ceph can not run
- so after i mkcephfs
1.mds {0=up:creating},
2.osd can not run,even i run cosd -i -c ceph.conf.
osd.log:
journal ... - 06:29 PM Revision 61bd155f (ceph): osd: OSD::queue_pg_for_deletion: avoid double del
- Signed-off-by: Colin McCabe <colinm@hq.newdream.net>
- 06:22 PM Revision 27f172f4 (ceph): osd: OSD::queue_pg_for_deletion: avoid double del
- Signed-off-by: Colin McCabe <colinm@hq.newdream.net>
- 05:50 PM Revision 6e6c22ea (ceph): mds: avoid double-pinning stray inodes
- We make multiple iterations through populate_mydir(). Only pin each stray
once. Fixes #689 and crashes like
mds/CI... - 05:22 PM Revision 83d3fb2c (ceph): debian: make update_pbuilder.sh a bit smarter
- Signed-off-by: Sage Weil <sage@newdream.net>
- 03:32 PM Bug #701 (Rejected): osd: scrub crash, failed clone_size assert
- was running old code, problem went away with latest testing branch
- 03:25 PM Bug #701: osd: scrub crash, failed clone_size assert
- on kvmtest-osd1 and 2
- 03:24 PM Bug #701 (Rejected): osd: scrub crash, failed clone_size assert
- ...
- 03:31 PM Bug #702 (Resolved): osd: failed interval_set assertr in snap_trimmer FAILED assert(p->first > st...
- ...
- 12:48 PM Bug #696 (Resolved): osd: _put_pool, assert(p->num_pg > 0)
- 11:29 AM Bug #696: osd: _put_pool, assert(p->num_pg > 0)
- Tried your fix, seems to be working, but like I said on IRC, I can't be 100% sure. I couldn't reproduce it today, so ...
- 10:05 AM Bug #696: osd: _put_pool, assert(p->num_pg > 0)
- See if this fixes it up: commit:27f172f4c7d1afafeda8c9ec1d530f5e02ea1b7a
in the latest unstable.
- 07:32 AM Bug #696: osd: _put_pool, assert(p->num_pg > 0)
- Colin McCabe wrote:
> Also, Sage, is there a reason why PG::~PG() is doing a pool->put(), but OSD::_remove_pg() also ... - 05:21 AM Bug #696: osd: _put_pool, assert(p->num_pg > 0)
- Hi Colin,
The logging was indeed low. But I was able to reproduce the crash yesterday with a higher logging enable... - 10:57 AM Linux kernel client Feature #700 (Resolved): rbd: support snapshot notify
- 10:55 AM rbd Feature #699 (Resolved): support snapshot notify
- 10:49 AM Bug #698 (Resolved): cosd memory usage with large number of pools
- I reported this on the mailing list a week ago but never filed it here. Still present in 0.24.1.
I've got a 3 nod... - 10:00 AM CephFS Cleanup #588 (In Progress): Allow Inodes to have multiple parent Dentries
- Okay, going to swap this with an implementation using sets rather than lists, then done!
- 09:36 AM CephFS Bug #689 (Resolved): cmds crash: mds/CInode.h:1088: FAILED assert(ref_set.count(by) == 0)
- Thanks, Ravi. There's no corruption. Fixed by commit:6e6c22ea23918e72211eb9cbdac64aec4abf7e7c.
- 12:30 AM Bug #590: osd/PG.cc:1645: FAILED assert(info.last_complete >= log.tail || log.backlog)
- Okay, everything is fine now... got rid of ENOSPC and had a scrub error that would not repair. I upgraded to 0.24.1 a...
01/10/2011
- 11:51 PM Revision e9a70f15 (ceph): keyring: can parse plain text keyring files
- 11:41 PM Revision e189222f (ceph): ReplicatedPG: Fix bug in rollback
- Previously, _rollback_to assumed that the rollback was a noop if
ctx->clone_obc was set and it's prior version matche... - 11:20 PM Revision 1229af51 (ceph): Port encoding tests over to gtest.
- 11:19 PM Revision 9b0d5776 (ceph): Use Google Test framework for unit tests.
- Use ``make check`` to run the tests.
The src/gtest directory comes from ``svn export
http://googletest.googlecode.co... - 11:07 PM Bug #696: osd: _put_pool, assert(p->num_pg > 0)
- Hi Wido,
I took a quick glance at the logs in osd_3_crash_rgw_pool_remove.tar.gz, but it looks like those are the ... - 07:41 PM Bug #696: osd: _put_pool, assert(p->num_pg > 0)
- It looks like we already have "_get_pool" and "_put_pool" printed out in the logs when the relevant functions are cal...
- 12:47 AM Bug #696: osd: _put_pool, assert(p->num_pg > 0)
- I'm able to reproduce the crash on my 'noisy' machine.
After setting the logging to 20 I got:... - 10:07 PM Revision 1a9ef3fa (ceph): Make git ignore generated files.
- 11:30 AM Feature #697: create rbd library
- We've discussed it on IRC, might be usefull as reference: http://irclogs.ceph.widodh.nl/index.php?date=2011-01-10
- 11:25 AM Feature #697 (Resolved): create rbd library
- As suggested by Wido: we can create a new library -- librbd (similar to librados) that'll allow modifying rbd images....
- 10:32 AM Bug #694: no osd journal causes cosd segfault
- fixed by commit:1846355176ecd5b60e93da8cf20cea0ef6ed4326
01/09/2011
- 10:23 PM CephFS Bug #689: cmds crash: mds/CInode.h:1088: FAILED assert(ref_set.count(by) == 0)
- Log sent in email due to paranoia about security ;)
Also, I've verified that I get the same crash with 0.24.1. The... - 09:34 PM Revision 18463551 (ceph): os: don't crash on no-journal case
- JournalingObjectStore::commit_start should handle the case where journal is
null. This will occur if the user doesn't... - 08:28 PM Bug #696: osd: _put_pool, assert(p->num_pg > 0)
- We may want to add get/put pool debug prints (if there aren't any already) so we can nail this one down...
- 01:36 AM Bug #696 (Resolved): osd: _put_pool, assert(p->num_pg > 0)
- Yesterday I tried to remove some pools from my Ceph system on 'noisy', running unstable commit eace4398cb163a670ff6bb...
01/08/2011
- 08:02 PM Support #695 (Closed): two questions about osd and kclient,important!!
- This is a bug tracker, not a question-answer forum or a wiki. Please send your questions to the developer list and we...
- 07:03 PM Support #695 (Closed): two questions about osd and kclient,important!!
- 1.ceph-client-standalone.git ,ceph-client.git
what is the different of the client.git???i know ceph-client-standal... - 02:11 PM CephFS Bug #689: cmds crash: mds/CInode.h:1088: FAILED assert(ref_set.count(by) == 0)
- Can you attach the whole log?
- 02:08 PM CephFS Bug #689: cmds crash: mds/CInode.h:1088: FAILED assert(ref_set.count(by) == 0)
- Here's (a portion of) the mds log with debug mds = 20:
2011-01-08 14:16:35.724241 b68bdb70 mds0.cache.ino(100) pro... - 01:45 PM CephFS Tasks #366 (Resolved): test snaptests against clustered mds failures
- 01:35 PM Linux kernel client Bug #634 (Can't reproduce): Kernel client takes too long to recover after a MDS restart
- 01:34 PM Linux kernel client Cleanup #683 (Rejected): confused of struct ceph_xattr
- I'm not too concerned about the naming here; all xattrs are inode xattrs.
- 12:50 AM Revision 630565f3 (ceph): v0.24.1
01/07/2011
- 11:01 PM Revision eace4398 (ceph): test_split.sh: add many_pools test
- Signed-off-by: Colin McCabe <colinm@hq.newdream.net>
- 10:25 PM Revision a64ddbb6 (ceph): ReplicatedPG: get_object_context ssc refcount leak
- If obc->obs.ssc is non-null, the second get_snapset_context ends up
leaking a snapset reference.
Signed-off-by: Samu... - 10:17 PM Revision 7f357cf2 (ceph): mds: force dirfrag fragmention when replaying metablob
- We can have non-auth (and thus ambiguously fragmented) dirs in our cache.
When those get replayed, adjust our fragmen... - 10:17 PM Revision 63125adb (ceph): mds: make thrash_exports select random frags
- We were always picking the first frag.
Signed-off-by: Sage Weil <sage@newdream.net> - 10:17 PM Revision e6ee6c4f (ceph): mds: take rdlocks on bounding dftlocks; clean up migrator lock code
- We need to take an rdlock on bounding dirfrags during migration for a
rather irritating reason: when we export the bo... - 10:17 PM Revision fde4b97e (ceph): mds: fix _dout_lock recursion recursion
- The get_snaps() method also something to dout. We need to take care to
not do that as part of the ostream operator<<... - 09:45 PM Revision d2dbe9e9 (ceph): ReplicatedPG: register_object_context and register_snapset_context cleanup
- Previously, get_object_context and get_snapset_context did not register
the resulting objects. In some cases, these ... - 08:40 PM Revision 2da3ac14 (ceph): librados: check for initialization before doing certain operations
- 07:22 PM Revision 7b20c290 (ceph): librados: fix api declaration
- 06:50 PM Revision baae4bc7 (ceph): librados: add rados->version, include librados.h from .hpp
- 06:46 PM Revision a3424776 (ceph): librados-config: add man page
- 06:45 PM Revision 98105ad7 (ceph): librados-config: added a command line tool to dump librados version
- 03:50 PM Bug #694 (Resolved): no osd journal causes cosd segfault
- When ceph.conf doesn't include an "osd journal" configuration option, cosd will throw a segfault.
- 03:36 PM Bug #646 (Can't reproduce): [OOPS] mount.ceph monip:6789: /mnt/ceph ,ceph-client-standalone.git +...
- 02:45 PM Linux kernel client Bug #693: Adding a non-existent image creates a bad sysfs node
- commit:ee0450e4ede0dde6b4c721dfbd0a578b5e13053c in the unstable branch.
- 02:41 PM Linux kernel client Bug #693: Adding a non-existent image creates a bad sysfs node
- That's commit:64f2c53e25165d78808f482651a92685fd28ffe9.
- 02:41 PM Linux kernel client Bug #693 (Resolved): Adding a non-existent image creates a bad sysfs node
- Fixed in 64f2c53e25165d78808f482651a92685fd28ffe9. Had a bad cleanup sequence.
- 11:48 AM Linux kernel client Bug #693 (Resolved): Adding a non-existent image creates a bad sysfs node
- To reproduce:
echo "10.0.1.247 name=admin rbd doesnotexist" > /sys/bus/rbd/add
/sys/bus/rbd/devices/0 is create... - 02:25 PM Bug #666: OSD: ReplicatedPG::eval_repop
- FWIW, I seem to be able to reproduce this at will.
I've been trying to understand what I'm seeing in the logs - it... - 02:00 PM CephFS Bug #684 (Resolved): unstable fails some workunits
- fixed by commit:fde4b97e8c4fe5e7a657815ba59449a66e85ecba
- 12:00 PM CephFS Bug #684: unstable fails some workunits
- Attached the full-thread backtrace of the MDS after it hangs on snaptest-upchildrealms.
- 12:47 PM Bug #691 (Resolved): librados: Segfault when open_pool is run when not initialized
- 12:25 PM Bug #691: librados: Segfault when open_pool is run when not initialized
- Fixed with commit 2da3ac1499944e7aac358a47a4d5651f3a1257e0.
- 04:14 AM Bug #691 (Resolved): librados: Segfault when open_pool is run when not initialized
- While working with phprados I experienced a crash when I tried to open a pool when I did not initialize librados.
... - 11:13 AM Bug #673 (Resolved): cmon: SimpleMessenger::Pipe::discard_queue
- 11:03 AM Feature #334 (Resolved): Version in librados
- 11:03 AM Feature #334 (In Progress): Version in librados
- 10:43 AM Feature #334: Version in librados
- rados->version() implemented in commit:baae4bc7c5390681d4549bdfb9ec4204aa515eb1.
- 10:02 AM Feature #334: Version in librados
- Thanks, but this has only been done in the librados C library, not in the C++. Has this been done un purpose (just to...
- 04:30 AM phprados Feature #692 (Resolved): Throw an exception if we try to open a pool before we are initialized
- Fixed by commit b5d2a5871c8cf7da6f350060055a7d83da9f3330
We now throw an exception if we try to open a pool before... - 04:15 AM phprados Feature #692 (Resolved): Throw an exception if we try to open a pool before we are initialized
- Related to #691
- 04:25 AM phprados Bug #690 (Resolved): Segfault when a xattr has no value during listing
- Both fixed by bb3c344b69133a9307e076cca4523af38a233e21
- 04:17 AM phprados Bug #690 (In Progress): Segfault when a xattr has no value during listing
- Same goes for getxattr, example:...
- 04:06 AM phprados Bug #690 (Resolved): Segfault when a xattr has no value during listing
- Right now phprados segfaults during the return value array assignment when we list the xattr's of a object, where one...
01/06/2011
- 11:59 PM Revision 86653700 (ceph): ReplicatedPG: clone_overlap should contain one entry per clone
- Previously, writefull and _delete_head would remove the last
entry from snapset.clone_overlap. Now, the last entry b... - 11:44 PM Revision 3a4afd98 (ceph): mon: Introduce Monitor::leader_since
- Introduce Monitor::leader_since to keep track of when the current
monitor became the leader.
Signed-off-by: Colin Mc... - 11:44 PM Revision a9420716 (ceph): osd: Rename osd_mon_report_interval
- Rename osd_mon_report_interval to osd_mon_report_interval_min.
Signed-off-by: Colin McCabe <colinm@hq.newdream.net> - 11:44 PM Revision 668c0d96 (ceph): mon: remove obsolete comment
- Signed-off-by: Colin McCabe <colinm@hq.newdream.net>
- 11:44 PM Revision 5947769e (ceph): mon: don't allow Monitor to be copied
- Signed-off-by: Colin McCabe <colinm@hq.newdream.net>
- 11:44 PM Revision fc032afd (ceph): osd: Introduce osd_mon_report_interval_max
- After every g_conf.osd_mon_report_interval_max seconds, we send out a PG
stat update even if nothing has changed. Thi... - 11:44 PM Revision 396dce9e (ceph): mon: Always forward the PGStats to the leader
- Always forward the PGStats to the leader, even if they are the same as
the old PGStats. The leader will mark as down ... - 11:44 PM Revision 494b812c (ceph): mon: mark osds down for not sending MOSDPGStat
- PGMonitor::prepare_pg_stats should check to see if the stats in the
MOSDPgStats message are the same as the ones we a... - 11:44 PM Revision 2a7aad1b (ceph): osd: Create all_osds_die test
- Signed-off-by: Colin McCabe <colinm@hq.newdream.net>
- 11:39 PM Revision 0ef19e6b (ceph): Merge branch 'standby_replay' into unstable
- 11:37 PM Revision 3df23749 (ceph): mds: Add is_any_replay() method and fill it in as appropriate.
- This way we don't need to remember to call all three of is_replay(),
is_standby_replay(), is_oneshot_replay().
Signe... - 10:50 PM Revision b1b4c03e (ceph): Merge remote branch 'origin/unstable' into standby_replay
- Conflicts:
src/cmds.cc
src/mds/MDS.cc
src/mds/MDS.h - 10:43 PM Revision 8dbe529e (ceph): librados: add library api versioning
- 09:44 PM CephFS Bug #689: cmds crash: mds/CInode.h:1088: FAILED assert(ref_set.count(by) == 0)
- Apparently the objdump is a bit big to post as an attachment. I've uploaded it at http://p-static.net/cmds-objdump.gz
- 09:42 PM CephFS Bug #689 (Resolved): cmds crash: mds/CInode.h:1088: FAILED assert(ref_set.count(by) == 0)
- I just upgraded to Ceph 0.24, and when I started up the cluster I got
this crash.
2011-01-06 21:23:48.744864 b685... - 07:12 PM Revision 92824b64 (ceph): MDSMonitor: Do not set the rank of an MDS in standby-replay
- or oneshot-replay modes.
This was causing issues with identification in various circumstances,
and turns out to be u... - 07:12 PM Revision 7561f809 (ceph): MDSMonitor: Remove STATE_ONESHOT_REPLAY from takeover logic in tick().
- If something dies during a journal-check we shouldn't have anybody
doing standby for them, so assert out!
Signed-off... - 07:12 PM Revision f84a0909 (ceph): lru: change control flow and an assert to keep purpose clearer.
- Signed-off-by: Greg Farnum <gregf@hq.newdream.net>
- 07:12 PM Revision d14c7dd4 (ceph): elist: add a clear_list function.
- Signed-off-by: Greg Farnum <gregf@hq.newdream.net>
- 07:12 PM Revision ae4dda41 (ceph): mdcache: adjust trim() to handle running during standby-replay.
- This just means it needs to handle files on the open list and not
trim them. Add a check for that with an assert, and... - 07:12 PM Revision 8ca291c0 (ceph): mdlog: Add some helper functions for accessing segments map data.
- Signed-off-by: Greg Farnum <gregf@hq.newdream.net>
- 07:12 PM Revision b8b172d2 (ceph): MDS: make standby_trim_segments functional. Hurray, hot standbys work!
- Signed-off-by: Greg Farnum <gregf@hq.newdream.net>
- 07:12 PM Revision ff7340aa (ceph): mds: After probing the journal, reset if we've fallen behind.
- Previously, if the journal got trimmed and we missed log entries,
we failed out in the journaling step and stopped.
T... - 07:12 PM Revision bf0675fa (ceph): journaler: Add init_headers function, call when reading head off disk.
- Uninitialized headers were causing a failed assert during replay,
and there's no good reason to leave them set at the... - 07:12 PM Revision 8d88f3b2 (ceph): mdlog: return EAGAIN if replay falls off the tail of the journal.
- This can happen when we're following an active journal, and
would previously cause the MDS to shut down. Now we retur... - 07:12 PM Revision 7472806b (ceph): MDS: add gids to the logger file names.
- This is just to make differentiating between the standby's files
and stuff easier.
Signed-off-by: Greg Farnum <gregf... - 07:12 PM Revision cd4cd01a (ceph): MDS: rename replay Contexts -- they were ambiguous at best.
- Signed-off-by: Greg Farnum <gregf@hq.newdream.net>
- 07:12 PM Revision 60a4189d (ceph): mdcache: change replay trimming a bit.
- Previously we were re-inserting dentrys on the open list. But if
there weren't any other available dentrys to trim, t... - 07:12 PM Revision b0676cb9 (ceph): journaler: delete Contexts on finish() in new functions.
- Previously we weren't, and leaked memory.
Signed-off-by: Greg Farnum <gregf@hq.newdream.net> - 07:12 PM Revision 71f6a089 (ceph): MDS: miscellaneous standby-replay fixes and cleanups.
- Signed-off-by: Greg Farnum <gregf@hq.newdream.net>
- 07:12 PM Revision 3cb8c080 (ceph): Journaler: Remove the unused read_pos field.
- Rename it to unused_field, fill the in-memory read_pos
from header.expire_pos, and fill unused_field with the expire_... - 07:12 PM Revision b9aaa117 (ceph): MDLog: don't change expire_pos or read_pos on replay.
- These are unnecessary or rendered irrelevant by previous commit
removing read_pos from the on-disk Header.
Signed-of... - 07:12 PM Revision 3f5371cd (ceph): MDS: MDSMonitor: if MDS is in standby-replay and its leader goes down,
- take over as the MDS!
This means we can now exit standby-replay.
Signed-off-by: Greg Farnum <gregf@hq.newdream.net> - 07:12 PM Revision 7dcbfb81 (ceph): MDS: make use of the hooks to start standby-replay.
- This doesn't include trim, and there's no way to exit the replay!
Signed-off-by: Greg Farnum <gregf@hq.newdream.net> - 07:09 PM Revision d55a6579 (ceph): osd, rados: pgls filter cleanups
- 06:52 PM Revision cfd87cee (ceph): objecter: use raw_pg_to_pg when needed
- Signed-off-by: Colin McCabe <colinm@hq.newdream.net>
- 06:35 PM Revision e1675f32 (ceph): Journaler: set state to STATE_ACTIVE in _finish_probe_end.
- This was never actually getting set, although it doesn't matter
since STATE_ACTIVE and STATE_PROBING are defined to b... - 06:35 PM Revision 8de5681d (ceph): Journaler: remove unused vector<snapid_t> snaps from recover().
- Signed-off-by: Greg Farnum <gregf@hq.newdream.net>
- 06:35 PM Revision 25d6576a (ceph): Journaler: add new reread_head function and state.
- This is to facilitate the forthcoming up_shadow MDS state.
Signed-off-by: Greg Farnum <gregf@hq.newdream.net> - 06:35 PM Revision 93597910 (ceph): Journaler: Add a read-only setting, and asserts to make it fail on writ...
- Signed-off-by: Greg Farnum <gregf@hq.newdream.net>
- 06:35 PM Revision cfd839d2 (ceph): Journaler: Add reset() function, which returns it to the immediate post...
- Signed-off-by: Greg Farnum <gregf@hq.newdream.net>
- 06:35 PM Revision fb074f45 (ceph): Journaler: Add function reprobe, to search for the new end of log.
- Add new REPROBING state and split up new function probe() from _finish_read_head.
Signed-off-by: Greg Farnum <gregf@... - 06:35 PM Revision 14829c41 (ceph): Journaler: use uint64_6 instead of int64_t.
- Since the values can never be negative, this is far more appropriate,
and it results in fewer casts than the other wa... - 06:35 PM Revision 18d89d13 (ceph): Journaler: Set the privacy of new functions correctly.
- Signed-off-by: Greg Farnum <gregf@hq.newdream.net>
- 06:35 PM Revision c9e1f992 (ceph): Journaler: redefine states to make them all unique.
- Apparently PROBING and ACTIVE being identical was a mistake.
Signed-off-by: Greg Farnum <gregf@hq.newdream.net> - 06:35 PM Revision 8da26fde (ceph): Journaler: make reread_head an asynchronous function.
- This better fits the spirit of the other functions, and the MDS itself.
Signed-off-by: Greg Farnum <gregf@hq.newdrea... - 06:35 PM Revision a34a0cf8 (ceph): Journaler: make reprobe() an asynchronous function.
- This better fits the spirit of the other functions, and the MDS itself.
Signed-off-by: Greg Farnum <gregf@hq.newdrea... - 06:35 PM Revision cc0eef4b (ceph): mds: Create new STATE_ONESHOT_REPLAY for the MDS.
- This takes over the previous behavior of STATE_STANDBY_REPLAY,
allowing standby-replay to be used for the upcoming co... - 06:35 PM Revision 02d46401 (ceph): mds: rename is_standby_replay() to is_oneshot_replay.
- This better represents its current purpose.
- 06:35 PM Revision e30cbaa9 (ceph): mds: extend the use of uint64_t instead of (signed) loff_t, et al.
- Signed-off-by: Greg Farnum <gregf@hq.newdream.net>
- 06:35 PM Revision 6337842c (ceph): mds: add expire_pos to the ESubtreeMap.
- This will allow more efficient trimming during standby_replay.
Signed-off-by: Greg Farnum <gregf@hq.newdream.net> - 06:35 PM Revision 87e40c0f (ceph): journaler: add reread_head_and_probe function.
- It does both so callers don't need to implement
intermediate bottom-half handlers.
Signed-off-by: Greg Farnum <gregf... - 06:35 PM Revision a8a7e730 (ceph): MDS: Implement the hooks for standby_replay.
- This commit adds the necessary state checks and machinery
for the MDS to go through a "looping" replay.
It does not y... - 06:31 PM Revision 4f15b8fa (ceph): objecter, librados: propagate extra pgls info to client
- 04:47 PM CephFS Bug #684: unstable fails some workunits
- Logs for each case are on kai:/home/gregf/logs
- 10:13 AM CephFS Bug #684: unstable fails some workunits
- Updated with snaptest-2 issue.
- 08:53 AM CephFS Bug #684 (Resolved): unstable fails some workunits
- This is on both cfuse and the kclient, so I assume it's an MDS issue.
v0.24 worked correctly; bisecting this is diff... - 04:40 PM Bug #688 (Resolved): should add dependency on libedit to rpm, deb, configure.ac
- should add dependency on libedit to rpm, deb, configure.ac(04:35:35 PM)
One user saw this error because his libed... - 03:02 PM Feature #609: osd: query pool/pg for objects with given xattr
- yehudasa@skinny:~/ceph/src$ ./rados ls -p metadata --filter="parent 0x100"
602.00000000
dirino=100 dname=stray2 v=... - 03:01 PM Feature #687 (Resolved): Let the monitor auto-set idle MDSes as standbys
- If we have more MDS daemons than configured MDSes, let the monitor set the idle ones as standbys automatically.
This... - 03:00 PM Feature #686 (Resolved): Specify standby behavior in configuration file
- We need ways to set certain MDSes as standbys of each other, in the config file.
- 02:54 PM Bug #629 (Resolved): cosd segfaults when deleting a pool containing degraded objects
- 02:52 PM Feature #334 (Resolved): Version in librados
- Fixed in commit:42c240d389cd2811f61d0f85aa5422d5473bb7dd. Will be useful for qemu rbd.
- 02:49 PM Bug #666: OSD: ReplicatedPG::eval_repop
- I have to use our managed transfer app for that; you should be getting an email about the transfer.
Let me know if... - 12:46 PM Bug #666 (In Progress): OSD: ReplicatedPG::eval_repop
- 12:46 PM Bug #666: OSD: ReplicatedPG::eval_repop
- Can you post logs?
Thanks! - 10:59 AM Bug #666: OSD: ReplicatedPG::eval_repop
- Hit another instance of this on current unstable (d386ea2ea0b).
I've got logs from debug osd = 20 + debug filestor... - 02:47 PM CephFS Feature #91 (Resolved): mds: up:shadow mode
- Well, this seems to be working as best I can tell.
There are some odd issues with virtual memory usage growing by ... - 02:15 PM Feature #685 (Duplicate): libcephmon: interact with ceph monitors via a library
- Hi,
Right now librados nor libceph give you a full internal look of how the Ceph system is behaving.
Some infor... - 10:36 AM CephFS Cleanup #588: Allow Inodes to have multiple parent Dentries
- When the client changes are merged into testing it also passes snaptest-upchildrealms and snaptest-2. I suspect this ...
- 09:19 AM CephFS Cleanup #588: Allow Inodes to have multiple parent Dentries
- Pushed the branch uclient_dentries; it's only a few commits.
There were actually surprisingly few things that need... - 10:35 AM Bug #682 (Resolved): higher pgp_bits setting leads to strange radostool errors
- Looks like we have a winner...
Submitted the patch as commit:cfd87ceefb46358adaa1751975c8d3a6b063bdf9
C. - 10:00 AM Feature #94 (Resolved): mon: monitor some osds
- Implemented in the osd_timeo branch.
C. - 02:16 AM Revision d386ea2e (ceph): common: dout_create_rank_symlink: init if needed
- Signed-off-by: Colin McCabe <colinm@hq.newdream.net>
- 12:50 AM Revision 2d781ff6 (ceph): osd, rados: pgls filter fixes
- 12:48 AM Revision 1f3b9c69 (ceph): mds: remove stray reference& in FragmentMarking context
- Led to confusing occasional(!) crashes on marking completion.
Signed-off-by: Sage Weil <sage@newdream.net> - 12:48 AM Revision 17bc2935 (ceph): mds: change refragment journaling/store strategy
- We had a serious problem before where we were updating the cache and
redivvying up the dentries among fragments, but ... - 12:48 AM Revision d1d26a53 (ceph): mds: make adjust_dir_fragments always adjust fragtree
- If we have the inode but no dirfrags, we still need to adjust the
inode dirfragtree.
Signed-off-by: Sage Weil <sage@... - 12:48 AM Revision cdf6af04 (ceph): Merge remote branch 'origin/testing' into unstable
- 12:48 AM Revision 21d49d91 (ceph): mds: fix can_authpin assert on post-fragment commit
- We wan to ignore the authpinnability check here; we already have the
(old) frag frozen, so no worries about starvatio... - 12:48 AM Revision 15c5e0c2 (ceph): mds: add mds_debug_frag option
- Verify dirfragtree matches any open dirfrags.
Signed-off-by: Sage Weil <sage@newdream.net>
01/05/2011
- 11:53 PM Revision 8adaa047 (ceph): common: make command-line programs log to stderr
- command-line programs (as opposed to daemons) should send their logs to
stderr rather than to a log file, syslog, etc... - 11:30 PM Linux kernel client Cleanup #683 (Rejected): confused of struct ceph_xattr
- file xattrs.c,line 371,376
struct ceph_xattr ,it should be struct ceph_inode_xattr
- 11:29 PM Revision bba356b8 (ceph): rgw_admin: call common_set_defaults as non-daemon
- Signed-off-by: Colin McCabe <colinm@hq.newdream.net>
- 10:10 PM Bug #682: higher pgp_bits setting leads to strange radostool errors
- try this?...
- 05:01 PM Bug #682: higher pgp_bits setting leads to strange radostool errors
- by the way, it should be 100% reproducible, just pull the latest unstable and run "./test/test_unfound.sh run"
- 04:47 PM Bug #682 (Resolved): higher pgp_bits setting leads to strange radostool errors
- git-bisect shows that commit:8f104243f823c3b5c827a4c7e59d637d38846e3f broke test_unfound.sh, test_lost.sh, and probab...
- 08:40 PM Revision f3771689 (ceph): debian: update scripts to build ubuntu (maverick, lucid) packages too
- Signed-off-by: Sage Weil <sage@newdream.net>
- 05:33 PM Revision 6dce7b0f (ceph): mds: move flock types into separate header
- Signed-off-by: Sage Weil <sage@newdream.net>
- 05:18 PM Revision 4b2ba49e (ceph): rados tool: Remove duplicate line in usage
- Signed-off-by: Wido den Hollander <wido@widodh.nl>
Signed-off-by: Sage Weil <sage@newdream.net> - 03:14 PM Feature #609 (In Progress): osd: query pool/pg for objects with given xattr
- 12:34 PM Feature #679: build .debs for ubuntu releases
- commit:f377168905b3d9530cca380b1c06ed34e7e343a1
- 12:34 PM Feature #679 (Resolved): build .debs for ubuntu releases
- 09:04 AM Cleanup #681 (Resolved): Incorrect usage radostool, duplicate line for 'getxattr'
- thanks. commit:4b2ba49e287b6afe50e99772173f0c1482fffc13
- 06:01 AM Cleanup #681 (Resolved): Incorrect usage radostool, duplicate line for 'getxattr'
- Just a small thing I noticed today:...
- 02:25 AM Revision 065e2acb (ceph): common: generic_dout needs to take the dout mutex
- Signed-off-by: Colin McCabe <colinm@hq.newdream.net>
- 01:18 AM Revision c46d8d5d (ceph): osd: add pgls filtering by parent ino
- 01:02 AM Revision edd5686e (ceph): common: handle_fatal_signal: print threadid in hex
- Signed-off-by: Colin McCabe <colinm@hq.newdream.net>
- 12:14 AM Revision 4a1c1b4a (ceph): mds: fix ancestor backtrace encoding
- Use explicit types to capture the encoding. Include object ino in the
inode_backtrace_t so that the xattr can stand ...
01/04/2011
- 11:27 PM Revision f66d4728 (ceph): frag: const cleanup fragset_t
- Signed-off-by: Sage Weil <sage@newdream.net>
- 11:27 PM Revision 4c53c726 (ceph): mds: make get_dirfrags_under behave when dirfragtree is not coherent wi...
- This is (currently) the case during replay/resolve, although it's not
clear that it should be.
Signed-off-by: Sage W... - 11:27 PM Revision a5f9a36c (ceph): mds: make resolve adjust dir fragmentation as needed
- During resolve, adjust dir fragmentation as needed based on the subtrees
the sender explicitly claims. The given fra... - 11:27 PM Revision 3d04d96b (ceph): mds: force fragmentation for ambiguous imports as well
- Handle needed refragmentation for processing ambiguous bounds. That means
forcing the peers' subtree root fragmentat... - 11:01 PM Revision 1b0bb7a1 (ceph): osd, objecter: pgls filtering option
- 10:38 PM Revision fab61391 (ceph): PG: Fixes bug in _scrub with checking clones
- I introduced this bug in
4a4a1e53c7d380cd0b582c1d0685fd0ef4ef1711.
curclone++ not curclone--.
Signed-off-by: Samuel ... - 07:29 PM Revision 8f104243 (ceph): osd: set default pg_bits higher; pgp_bits to old value
- Signed-off-by: Sage Weil <sage@newdream.net>
- 06:50 PM Revision aafb79c7 (ceph): assert: print thread id in hex
- Signed-off-by: Sage Weil <sage@newdream.net>
- 06:27 PM Revision 4a4a1e53 (ceph): PG: Fix bug in scrub when checking clone sizes
- Previosly, _scrub checked:
assert(p->second.size == snapset.clone_size[curclone])
curclone was, however, an index in... - 06:20 PM Revision 3d4e1d99 (ceph): client: fix frag selection code
- Calling fragtree_t::contains() on a non-frag_t is nonsense and will crash.
And a fragtree is a complete partition of ... - 06:18 PM Revision 062c4cc8 (ceph): vstart.sh: specify keyring in ceph.conf
- Signed-off-by: Sage Weil <sage@newdream.net>
- 05:16 PM Revision d2578107 (ceph): remove ancient uofs.h
- Signed-off-by: Sage Weil <sage@newdream.net>
- 05:09 PM Revision 9befd1ea (ceph): mkcephfs: Clarified numosd message
- Signed-off-by: Matthew Roy <matthew@royhousehold.net>
Signed-off-by: Sage Weil <sage@newdream.net> - 03:10 PM CephFS Bug #594: mds: frag split/merge vs replay
- commit:3d04d96b5cd0ab9d94904d035347c00fe19f687a fixes things for resolve. rejoin still needs work.
- 02:47 PM Feature #609 (Resolved): osd: query pool/pg for objects with given xattr
- implemented it as a new PGLS_FILTER operation, but reuses the PGLS operation code. There's no class extension now, bu...
- 09:01 AM Support #680 (Closed): what's device_open,uofs_*
- uofs is an ancient implementation OBFS, a userspace object file system from UCSC. It should probably be removed from...
- 05:39 AM Support #680 (Closed): what's device_open,uofs_*
- what is uofs? device_open
- 05:24 AM Revision 78f35a64 (ceph): osd: send sufficient log to compensate for replicas with last_complate ...
- If a replica has last_complete < log.tail and no backlog, send enough log
for them to get back into a consistent stat... - 05:24 AM Revision ff035ab3 (ceph): osd: generate backlog if needed to get last_complete >= log.tail || bac...
- If primary or a replica has a mistrimmed pg log, we need to generate the
backlog during peering. This sucks, because... - 05:24 AM Revision 462cb841 (ceph): osd: clean up backlog generation checks a bit
- Signed-off-by: Sage Weil <sage@newdream.net>
- 05:24 AM Revision 88c445b1 (ceph): mds: start new log segment resolve start, not replay finish
- Signed-off-by: Sage Weil <sage@newdream.net>
- 05:24 AM Revision 6c73da0a (ceph): mds: assert no submit_entry during replay state
- We should never submit items to the journal during replay.
Signed-off-by: Sage Weil <sage@newdream.net> - 01:44 AM Revision 16c603b2 (ceph): common: Implement max open files
- In init-ceph, call ulimit -n if the user has set a maximum number of
open files, and the current maximum number of fi... - 12:23 AM Bug #665: mds(uo:replay),clent can not mount,return error 5 input/output
- how to locate error in logs?
mds.log is attached.
Sage Weil wrote:
> You can't mount until mds replay completes. ... - 12:14 AM Revision 5dc66244 (ceph): osd: Make g_conf.osd_max_notify_timeout a uint32_t
- Make g_conf.osd_max_notify_timeout a uint32_t. Squashes an annoying
compiler warning and avoids the awkward issue of ...
Also available in: Atom