Project

General

Profile

Activity

From 12/19/2019 to 01/17/2020

01/17/2020

10:51 PM Tasks #4492 (New): mds: Define kill points involved in clustered migration and recovery
Patrick Donnelly
10:50 PM Feature #39129 (Fix Under Review): create mechanism to delegate ranges of inode numbers to client
Patrick Donnelly
09:49 PM Backport #42943: nautilus: mds: free heap memory may grow too large for some workloads
Patrick Donnelly wrote:
> https://github.com/ceph/ceph/pull/31802
merged
Yuri Weinstein
09:48 PM Backport #42631: nautilus: client: FAILED assert(cap == in->auth_cap)
Nathan Cutler wrote:
> https://github.com/ceph/ceph/pull/32065
merged
Yuri Weinstein
07:33 PM Bug #43596: mds: crash when enable msgr v2 due to lost contact
That's indeed very odd. I looked through the code but didn't find a good reason why this would happen. It is interest... Patrick Donnelly
01:13 PM Bug #43596: mds: crash when enable msgr v2 due to lost contact
Yes the MDS was upgraded to 14.2.6 also.
Below the mon log when I changed its addr. (Full file that day is at ceph...
Dan van der Ster
06:49 PM Bug #43644 (Triaged): mds: Empty directory check is done on the importer side (at import finish) ...
Patrick Donnelly
06:48 PM Bug #43644: mds: Empty directory check is done on the importer side (at import finish) during mig...
Zheng Yan wrote:
> you are right. we can do the check a export_dir and export_frozen. If directory is empty, abort. ...
Patrick Donnelly
12:13 PM Bug #43644: mds: Empty directory check is done on the importer side (at import finish) during mig...
you are right. we can do the check a export_dir and export_frozen. If directory is empty, abort. But we still need to... Zheng Yan
09:16 AM Bug #43644: mds: Empty directory check is done on the importer side (at import finish) during mig...
Sidharth Anupkrishnan wrote:
> In the current MDS code, the migration of empty directories is prohibited but it is ...
Sidharth Anupkrishnan
09:13 AM Bug #43644 (Rejected): mds: Empty directory check is done on the importer side (at import finish)...
In the current MDS code, the migration of empty directories is prohibited but it is actually exported during the migr... Sidharth Anupkrishnan
03:59 PM Bug #43649: mount.ceph fails with ERANGE if name= option is longer than 37 characters
It turns out that name= options can pretty much be arbitrarily long, so I reworked the code to remove the need for an... Jeff Layton
03:57 PM Bug #43649 (In Progress): mount.ceph fails with ERANGE if name= option is longer than 37 characters
Jeff Layton
03:13 PM Bug #43649 (Resolved): mount.ceph fails with ERANGE if name= option is longer than 37 characters
Aaron reported on the cephfs mailing list that some mount attempts were failing with ERANGE. For example:... Jeff Layton
10:01 AM Bug #43645 (Fix Under Review): mgr/volumes: subvolumes with snapshots can be deleted
Venky Shankar
09:28 AM Bug #43645 (Resolved): mgr/volumes: subvolumes with snapshots can be deleted
... Venky Shankar
07:50 AM Feature #24880 (Fix Under Review): pybind/mgr/volumes: restore from snapshot
clone from a snap: https://github.com/ceph/ceph/pull/32030
Most of this work will be required for restoring a subv...
Venky Shankar
05:46 AM Bug #42835 (Fix Under Review): qa: test_scrub_abort fails during check_task_status("idle")
Venky Shankar

01/16/2020

11:49 PM Bug #43640: nautilus: qa: test_async_subvolume_rm failure
Just the lines from the teuthology log for the mgr connection:... Patrick Donnelly
09:00 PM Bug #43640 (Need More Info): nautilus: qa: test_async_subvolume_rm failure
... Ramana Raja
08:17 PM Bug #43638 (Duplicate): nautilus qa: tasks/cfuse_workunit_suites_ffsb.yaml failure
Patrick Donnelly
08:05 PM Bug #43638 (Duplicate): nautilus qa: tasks/cfuse_workunit_suites_ffsb.yaml failure
... Ramana Raja
05:45 PM Bug #43637 (Triaged): nautilus: qa: Health check failed: Reduced data availability: 16 pgs inacti...
... Ramana Raja
02:46 PM Backport #43629 (Resolved): nautilus: mgr/volumes: provision subvolumes with config metadata stor...
https://github.com/ceph/ceph/pull/33122/ Nathan Cutler
02:46 PM Backport #43628 (Resolved): nautilus: client: disallow changing fuse_default_permissions option a...
https://github.com/ceph/ceph/pull/32915 Nathan Cutler
02:46 PM Backport #43627 (Rejected): mimic: client: disallow changing fuse_default_permissions option at r...
Nathan Cutler
02:45 PM Backport #43624 (Resolved): nautilus: mds: note features client has when rejecting client due to ...
https://github.com/ceph/ceph/pull/32914 Nathan Cutler
09:50 AM Bug #43601 (Fix Under Review): qa: ERROR: test_object_deletion (tasks.cephfs.test_damage.TestDamage)
Zheng Yan
12:25 AM Bug #43601 (Triaged): qa: ERROR: test_object_deletion (tasks.cephfs.test_damage.TestDamage)
Looks like it's just that the MDS is responding to a getattr request on the root inode with EROFS:... Patrick Donnelly
12:26 AM Bug #43125 (Can't reproduce): qa: ceph_volume_client not available "ModuleNotFoundError: No modul...
Patrick Donnelly
12:13 AM Documentation #43155 (Closed): CephFS Documentation Sprint 4
Patrick Donnelly
12:12 AM Bug #42637 (Fix Under Review): qa: ffsb suite causes SLOW_OPS warnings
Patrick Donnelly
12:00 AM Bug #16881 (Fix Under Review): RuntimeError: Files in flight high water is unexpectedly low (0 / 6)
Patrick Donnelly

01/15/2020

08:49 PM Bug #43599 (Fix Under Review): kclient: corrupt message failure on RHEL8 distribution kernel
Patrick Donnelly
08:25 PM Bug #43599 (In Progress): kclient: corrupt message failure on RHEL8 distribution kernel
Patrick Donnelly
08:20 PM Bug #43599: kclient: corrupt message failure on RHEL8 distribution kernel
This is just one of those places where the kernel client did not ever expect to see a struct be extended. I suspect t... Jeff Layton
06:43 PM Bug #43599: kclient: corrupt message failure on RHEL8 distribution kernel
Jeff Layton wrote:
> What kernel is this?...
Patrick Donnelly
05:50 PM Bug #43599: kclient: corrupt message failure on RHEL8 distribution kernel
What kernel is this? Jeff Layton
08:48 PM Bug #43600: qa: workunits/suites/iozone.sh: line 5: iozone: command not found
Unfortunately, CentOS 8 / RHEL 8 don't have this package. We'll need to filter out these distributions somehow.
Mo...
Patrick Donnelly
07:41 PM Bug #36507 (Duplicate): client: connection failure during reconnect causes client to hang
Thanks huanwen! Patrick Donnelly
07:39 PM Bug #42467: mds: daemon crashes while updating blacklist
Zheng, I think you may have inadvertently fixed this in... Patrick Donnelly
07:28 PM Bug #43216 (New): MDSMonitor: removes MDS coming out of quorum election
Patrick Donnelly
07:22 PM Bug #40608 (Duplicate): mds: assert after `delete gather` in C_Drop_Cache::recall_client_state
Fixed by: https://tracker.ceph.com/issues/38445 Patrick Donnelly
07:16 PM Bug #42941 (Rejected): mds: stuck "waiting for osdmap 273 (which blacklists prior instance)"
Cause was reverted. Patrick Donnelly
05:19 AM Feature #43349: mgr/volumes: provision subvolumes with config metadata storage in cephfs
backport note: additionally, include https://github.com/ceph/ceph/pull/32645 Venky Shankar
04:33 AM Feature #43349 (Pending Backport): mgr/volumes: provision subvolumes with config metadata storage...
Patrick Donnelly
04:33 AM Feature #43349 (Resolved): mgr/volumes: provision subvolumes with config metadata storage in cephfs
Patrick Donnelly
02:16 AM Bug #43362 (Pending Backport): client: disallow changing fuse_default_permissions option at runtime
Patrick Donnelly
02:15 AM Cleanup #43367 (Resolved): mds: reorg SimpleLock header
Patrick Donnelly
02:14 AM Cleanup #43386 (Resolved): mds: reorg SnapRealm header
Patrick Donnelly
02:13 AM Cleanup #43418 (Resolved): mds: reorg flock header
Patrick Donnelly
02:13 AM Cleanup #43424 (Resolved): mds: reorg inode_backtrace header
Patrick Donnelly
01:55 AM Bug #42986 (Fix Under Review): qa: Test failure: test_drop_cache_command_dead (tasks.cephfs.test_...
Patrick Donnelly
12:32 AM Bug #43513: qa: filelock_interrupt.py hang
Zheng Yan wrote:
> Looks like flock syscall was restarted after handling signal alarm. The script does not work with...
Patrick Donnelly
12:14 AM Bug #43554 (Fix Under Review): qa: test_full racy check: AssertionError: 29 not greater than or e...
Patrick Donnelly

01/14/2020

11:52 PM Bug #43601 (Resolved): qa: ERROR: test_object_deletion (tasks.cephfs.test_damage.TestDamage)
... Patrick Donnelly
11:15 PM Bug #43600 (Resolved): qa: workunits/suites/iozone.sh: line 5: iozone: command not found
... Patrick Donnelly
10:37 PM Bug #43542 (Resolved): mds/FSMap.cc: 1063: FAILED ceph_assert(count)
Sage Weil
10:32 PM Bug #43484 (Pending Backport): mds: note features client has when rejecting client due to feature...
Patrick Donnelly
10:10 PM Bug #43599 (Resolved): kclient: corrupt message failure on RHEL8 distribution kernel
... Patrick Donnelly
08:44 PM Bug #43541 (Resolved): qa/cephfs: don't test client on latest RHEL
Patrick Donnelly
08:44 PM Bug #43539 (Resolved): qa/cephfs: don't test kclient RHEL 7
Patrick Donnelly
08:34 PM Backport #42279: nautilus: qa: logrotate should tolerate connection resets
Nathan Cutler wrote:
> https://github.com/ceph/ceph/pull/31082
merged
Yuri Weinstein
08:33 PM Backport #42129: nautilus: doc/ceph-fuse: -k missing in man page
Nathan Cutler wrote:
> https://github.com/ceph/ceph/pull/30765
merged
Yuri Weinstein
05:39 PM Bug #43598 (Resolved): mds: PurgeQueue does not handle objecter errors
Here: https://github.com/ceph/ceph/blob/6ea89e01971462432e0bc8b128b950acec4d85fe/src/mds/PurgeQueue.cc#L555
The fi...
Patrick Donnelly
05:30 PM Bug #43596 (Need More Info): mds: crash when enable msgr v2 due to lost contact
> It seems to be stIt seems to be stable now after enabling v2 on all mons and restarting all mds's.able now after en... Patrick Donnelly
12:41 PM Bug #43596 (New): mds: crash when enable msgr v2 due to lost contact
We just upgraded from mimic v13.2.7 to v14.2.6 and when we enable msgr v2 on the mon which an MDS is connected to, th... Dan van der Ster
04:15 PM Bug #43440 (Fix Under Review): client: chdir does not raise error if a file is passed
Varsha Rao

01/13/2020

09:44 PM Bug #43493 (Fix Under Review): osdc: fix null pointer caused program crash
Patrick Donnelly
04:29 PM Backport #42790: nautilus: mgr/volumes: add `fs subvolume resize infinite` command
Jos Collin wrote:
> https://github.com/ceph/ceph/pull/31332
merged
Yuri Weinstein
04:28 PM Backport #42615: nautilus: mgr/volumes: add `fs subvolume extend/shrink` commands
Nathan Cutler wrote:
> https://github.com/ceph/ceph/pull/31332
merged
Yuri Weinstein
04:28 PM Backport #42142: nautilus: mds:split the dir if the op makes it oversized, because some ops maybe...
Nathan Cutler wrote:
> https://github.com/ceph/ceph/pull/31302
merged
Yuri Weinstein
04:27 PM Backport #42424: nautilus: qa: "cluster [ERR] Error recovering journal 0x200: (2) No such file ...
Nathan Cutler wrote:
> https://github.com/ceph/ceph/pull/31084
merged
Yuri Weinstein
04:27 PM Backport #42422: nautilus: test_reconnect_eviction fails with "RuntimeError: MDS in reject state ...
Nathan Cutler wrote:
> https://github.com/ceph/ceph/pull/31083
merged
Yuri Weinstein
04:26 PM Backport #42158: nautilus: osdc: objecter ops output does not have useful time information
Nathan Cutler wrote:
> https://github.com/ceph/ceph/pull/31081
merged
Yuri Weinstein
02:37 PM Bug #43567: qa: UnicodeDecodeError in TestGetAndPut.test_put_and_get_without_target_directory
... Rishabh Dave
12:12 PM Bug #43567 (Fix Under Review): qa: UnicodeDecodeError in TestGetAndPut.test_put_and_get_without_t...
Rishabh Dave
11:41 AM Bug #43567 (Resolved): qa: UnicodeDecodeError in TestGetAndPut.test_put_and_get_without_target_di...
decode() is run on a type @str@ -... Rishabh Dave
12:31 PM Feature #22446 (Resolved): mds: ask idle client to trim more caps
While running with --resolve-parent, the script "backport-create-issue" noticed that all backports of this issue are ... Nathan Cutler
12:28 PM Bug #40283 (Resolved): qa: add testing for lazyio
While running with --resolve-parent, the script "backport-create-issue" noticed that all backports of this issue are ... Nathan Cutler
12:28 PM Cleanup #40694 (Resolved): mds: move MDSDaemon conf change handling to MDSRank finisher
While running with --resolve-parent, the script "backport-create-issue" noticed that all backports of this issue are ... Nathan Cutler
12:27 PM Bug #41148 (Resolved): client: _readdir_cache_cb() may use the readdir_cache already clear
While running with --resolve-parent, the script "backport-create-issue" noticed that all backports of this issue are ... Nathan Cutler
12:27 PM Bug #41310 (Resolved): client: lazyio synchronize does not get file size
While running with --resolve-parent, the script "backport-create-issue" noticed that all backports of this issue are ... Nathan Cutler
12:26 PM Bug #41835 (Resolved): mds: cache drop command does not drive cap recall
While running with --resolve-parent, the script "backport-create-issue" noticed that all backports of this issue are ... Nathan Cutler
12:26 PM Bug #41837 (Resolved): client: lseek function does not return the correct value.
While running with --resolve-parent, the script "backport-create-issue" noticed that all backports of this issue are ... Nathan Cutler
12:22 PM Backport #42161 (Resolved): nautilus: qa: add testing for lazyio
This update was made using the script "backport-resolve-issue".
backport PR https://github.com/ceph/ceph/pull/30769
m...
Nathan Cutler
12:22 PM Backport #41888 (Resolved): nautilus: client: lazyio synchronize does not get file size
This update was made using the script "backport-resolve-issue".
backport PR https://github.com/ceph/ceph/pull/30769
m...
Nathan Cutler
12:22 PM Backport #42147 (Resolved): nautilus: mds: mds returns -5 error when the deleted file does not exist
This update was made using the script "backport-resolve-issue".
backport PR https://github.com/ceph/ceph/pull/30767
m...
Nathan Cutler
12:22 PM Backport #42145 (Resolved): nautilus: client: return error when someone passes bad whence value t...
This update was made using the script "backport-resolve-issue".
backport PR https://github.com/ceph/ceph/pull/30766
m...
Nathan Cutler
12:21 PM Backport #42121 (Resolved): nautilus: client: no method to handle SEEK_HOLE and SEEK_DATA in lseek
This update was made using the script "backport-resolve-issue".
backport PR https://github.com/ceph/ceph/pull/30764
m...
Nathan Cutler
12:21 PM Backport #42040 (Resolved): nautilus: client: _readdir_cache_cb() may use the readdir_cache alrea...
This update was made using the script "backport-resolve-issue".
backport PR https://github.com/ceph/ceph/pull/30763
m...
Nathan Cutler
12:21 PM Backport #42035 (Resolved): nautilus: client: lseek function does not return the correct value.
This update was made using the script "backport-resolve-issue".
backport PR https://github.com/ceph/ceph/pull/30762
m...
Nathan Cutler
12:21 PM Backport #42339 (Resolved): nautilus: mds: move MDSDaemon conf change handling to MDSRank finisher
This update was made using the script "backport-resolve-issue".
backport PR https://github.com/ceph/ceph/pull/30761
m...
Nathan Cutler
12:21 PM Backport #41899 (Resolved): nautilus: mds: cache drop command does not drive cap recall
This update was made using the script "backport-resolve-issue".
backport PR https://github.com/ceph/ceph/pull/30761
m...
Nathan Cutler
12:21 PM Backport #41865 (Resolved): nautilus: mds: ask idle client to trim more caps
This update was made using the script "backport-resolve-issue".
backport PR https://github.com/ceph/ceph/pull/30761
m...
Nathan Cutler
11:48 AM Backport #43573 (Resolved): nautilus: cephfs-journal-tool: will crash without any extra argument
https://github.com/ceph/ceph/pull/32913 Nathan Cutler
11:48 AM Backport #43572 (Rejected): mimic: cephfs-journal-tool: will crash without any extra argument
Nathan Cutler
11:48 AM Backport #43568 (Resolved): nautilus: qa: test setUp may cause spurious MDS_INSUFFICIENT_STANDBY
https://github.com/ceph/ceph/pull/32912 Nathan Cutler
03:03 AM Bug #43218 (Rejected): kclient: when looking up the snap dirs sometime will hit WARN_ON
This is not a bug and will close it.... Xiubo Li
01:43 AM Feature #9477 (Closed): Handle kclient shutdown with dead network more gracefully
this can be handled by 'umount -f' Zheng Yan
01:43 AM Feature #8368 (Resolved): kernel: Notify users of mds disconnect and allow them to react to it
Zheng Yan
01:30 AM Feature #8368: kernel: Notify users of mds disconnect and allow them to react to it
resolved by https://tracker.ceph.com/issues/39967 Zheng Yan

01/12/2020

12:52 AM Bug #36635: mds: purge queue corruption from wrong backport
I think we can just add an upgrade note to Octopus to not upgrade from 13.2.2. Patrick Donnelly

01/11/2020

01:49 PM Feature #24461: cephfs: improve file create performance buffering file unlink/create operations
OK, I move the wait to client side. See commit "client: wait for async creating before sending request or cap message... Zheng Yan
01:13 AM Bug #43543 (Triaged): mds: scrub on directory with recently created files may fail to load backtr...
Patrick Donnelly
12:25 AM Backport #43558 (In Progress): nautilus: mds: reject forward scrubs when cluster has multiple act...
Patrick Donnelly
12:19 AM Backport #43558 (Resolved): nautilus: mds: reject forward scrubs when cluster has multiple active...
https://github.com/ceph/ceph/pull/32602 Patrick Donnelly
12:19 AM Backport #43559 (Rejected): mimic: mds: reject forward scrubs when cluster has multiple active MD...
Patrick Donnelly
12:18 AM Bug #43483 (Pending Backport): mds: reject forward scrubs when cluster has multiple active MDS (m...
Patrick Donnelly
12:16 AM Bug #43249 (Resolved): cephfs-shell: exit failure when non-interactive command fails
Patrick Donnelly

01/10/2020

10:24 PM Bug #43251 (Resolved): mds: track client provided metric flags in session
Patrick Donnelly
10:22 PM Cleanup #43366 (Resolved): mds: reorg SessionMap header
Patrick Donnelly
10:15 PM Bug #43554 (Resolved): qa: test_full racy check: AssertionError: 29 not greater than or equal to 30
... Patrick Donnelly
09:24 PM Backport #43506 (In Progress): nautilus: MDSMonitor: warn if a new file system is being created w...
Patrick Donnelly
08:19 PM Backport #42161: nautilus: qa: add testing for lazyio
Nathan Cutler wrote:
> https://github.com/ceph/ceph/pull/30769
merged
Yuri Weinstein
08:19 PM Backport #41888: nautilus: client: lazyio synchronize does not get file size
Nathan Cutler wrote:
> https://github.com/ceph/ceph/pull/30769
merged
Yuri Weinstein
08:18 PM Backport #42147: nautilus: mds: mds returns -5 error when the deleted file does not exist
Nathan Cutler wrote:
> https://github.com/ceph/ceph/pull/30767
merged
Yuri Weinstein
08:18 PM Backport #42145: nautilus: client: return error when someone passes bad whence value to llseek
Nathan Cutler wrote:
> https://github.com/ceph/ceph/pull/30766
merged
Yuri Weinstein
08:17 PM Backport #42121: nautilus: client: no method to handle SEEK_HOLE and SEEK_DATA in lseek
Nathan Cutler wrote:
> https://github.com/ceph/ceph/pull/30764
merged
Yuri Weinstein
08:17 PM Backport #42040: nautilus: client: _readdir_cache_cb() may use the readdir_cache already clear
Nathan Cutler wrote:
> https://github.com/ceph/ceph/pull/30763
merged
Yuri Weinstein
08:16 PM Backport #42035: nautilus: client: lseek function does not return the correct value.
Nathan Cutler wrote:
> https://github.com/ceph/ceph/pull/30762
merged
Yuri Weinstein
08:16 PM Backport #42339: nautilus: mds: move MDSDaemon conf change handling to MDSRank finisher
Nathan Cutler wrote:
> https://github.com/ceph/ceph/pull/30761
merged
Yuri Weinstein
08:15 PM Backport #41899: nautilus: mds: cache drop command does not drive cap recall
Nathan Cutler wrote:
> https://github.com/ceph/ceph/pull/30761
merged
Yuri Weinstein
08:15 PM Backport #41865: nautilus: mds: ask idle client to trim more caps
Nathan Cutler wrote:
> https://github.com/ceph/ceph/pull/30761
merged
Yuri Weinstein
07:42 PM Bug #43540 (Duplicate): qa: test_export_pin (tasks.cephfs.test_exports.TestExports) failure
Real error is here:... Patrick Donnelly
12:57 PM Feature #24461: cephfs: improve file create performance buffering file unlink/create operations
I still don't understand what value this flag adds. Why not just always have requests involving an inode wait on the ... Jeff Layton
03:09 AM Feature #24461: cephfs: improve file create performance buffering file unlink/create operations
Jeff Layton wrote:
> Great! I'll still plan to add in a sanity check for this in the client too.
Patrick is right...
Zheng Yan
03:01 AM Feature #24461: cephfs: improve file create performance buffering file unlink/create operations
Jeff Layton wrote:
> Zheng Yan wrote:
> > mainly for wait_for_create_inode() function in MDS. Also make mds print e...
Zheng Yan
11:02 AM Bug #43440: client: chdir does not raise error if a file is passed
Not a cephfs shell bug. The error should be raised by ceph_chdir(). Varsha Rao
07:44 AM Bug #43543: mds: scrub on directory with recently created files may fail to load backtraces and r...
This issue exists since scrub is first implemented. Should be easy to fix, just ignore checking backtrace if dirty_pa... Zheng Yan

01/09/2020

11:14 PM Bug #43543: mds: scrub on directory with recently created files may fail to load backtraces and r...
If you flush the journal:... Patrick Donnelly
11:08 PM Bug #43543 (Resolved): mds: scrub on directory with recently created files may fail to load backt...
On a vstart cluster, copy a directory tree into CephFS and do a recursive scrub concurrently:... Patrick Donnelly
08:25 PM Bug #43514 (Pending Backport): qa: test setUp may cause spurious MDS_INSUFFICIENT_STANDBY
Patrick Donnelly
07:57 PM Bug #43542 (Fix Under Review): mds/FSMap.cc: 1063: FAILED ceph_assert(count)
Patrick Donnelly
07:48 PM Bug #43542 (Resolved): mds/FSMap.cc: 1063: FAILED ceph_assert(count)
... Patrick Donnelly
07:54 PM Feature #24461: cephfs: improve file create performance buffering file unlink/create operations
Great! I'll still plan to add in a sanity check for this in the client too. Jeff Layton
07:25 PM Feature #24461 (In Progress): cephfs: improve file create performance buffering file unlink/creat...
Jeff Layton wrote:
> There's a potential problem I spotted today with copying the layouts from the first synchronous...
Patrick Donnelly
07:21 PM Feature #24461: cephfs: improve file create performance buffering file unlink/create operations
There's a potential problem I spotted today with copying the layouts from the first synchronous create.
Suppose we...
Jeff Layton
07:10 PM Feature #24461: cephfs: improve file create performance buffering file unlink/create operations
Zheng Yan wrote:
> mainly for wait_for_create_inode() function in MDS. Also make mds print error if it failed to han...
Jeff Layton
02:05 PM Feature #24461: cephfs: improve file create performance buffering file unlink/create operations
mainly for wait_for_create_inode() function in MDS. Also make mds print error if it failed to handle async request. Zheng Yan
11:42 AM Feature #24461: cephfs: improve file create performance buffering file unlink/create operations
Zheng Yan wrote:
> please add a flag that tell if a request is async.
> https://github.com/ukernel/ceph/commit/54f...
Jeff Layton
07:13 PM Bug #43515 (Resolved): qa: SyntaxError: invalid token
Patrick Donnelly
07:12 PM Bug #43487 (Resolved): qa: test_acls does not detect rhel8
Patrick Donnelly
06:22 PM Feature #118 (Rejected): kclient: clean pages when throwing out dirty metadata on session teardown
Excellent. In that case, let's go ahead and close this out.
Jeff Layton
06:43 AM Feature #118: kclient: clean pages when throwing out dirty metadata on session teardown
Case1:
In the case when unmounting, the vfs will do this for us.
Case2:
In the case when the session is reconne...
Xiubo Li
06:09 PM Bug #43541 (Fix Under Review): qa/cephfs: don't test client on latest RHEL
Rishabh Dave
06:08 PM Bug #43541 (Resolved): qa/cephfs: don't test client on latest RHEL
Rishabh Dave
06:04 PM Bug #43540 (Duplicate): qa: test_export_pin (tasks.cephfs.test_exports.TestExports) failure
... Ramana Raja
06:02 PM Bug #43539 (Fix Under Review): qa/cephfs: don't test kclient RHEL 7
Rishabh Dave
05:42 PM Bug #43539 (Resolved): qa/cephfs: don't test kclient RHEL 7
Just fix the symlin qa/cephfs/mount/kclient/overrides/distro/rhel/rhel_7.yaml. Rishabh Dave
11:34 AM Feature #42530 (Fix Under Review): cephfs-shell: add setxattr and getxattr
Varsha Rao
06:48 AM Feature #43435 (Fix Under Review): kclient:send client provided metric flags in client metadata
Under review in V2's "[Patch v2 8/8] ceph: send client provided metric flags in client metadata"
https://patchwork...
Xiubo Li
06:45 AM Feature #43423 (Fix Under Review): mds: collect and show the dentry lease metric
Xiubo Li
06:44 AM Bug #37617: CephFS did not recover re-plugging network cable
I also enconter the same problem. when I use ls or other operations on the mountpoint,it will failed. Even if the net... geng jichao
05:50 AM Feature #4386 (Resolved): kclient: Mount error message when no MDS present
Xiubo Li
05:49 AM Feature #4386: kclient: Mount error message when no MDS present
Fixed in:
https://github.com/ceph/ceph/pull/32164
https://patchwork.kernel.org/patch/11283665
Xiubo Li

01/08/2020

03:27 PM Bug #43516 (Resolved): qa: verify sub-suite does not define os_version
Patrick Donnelly
02:38 PM Feature #24461: cephfs: improve file create performance buffering file unlink/create operations
please add a flag that tell if a request is async.
https://github.com/ukernel/ceph/commit/54f6bbdc85505ddea21583e9c...
Zheng Yan
01:54 PM Bug #43513: qa: filelock_interrupt.py hang
Looks like flock syscall was restarted after handling signal alarm. The script does not work with python3, but work w... Zheng Yan
08:58 AM Bug #43522 (Fix Under Review): qa: update xfstests_dev to install python2 instead of python on ub...
Rishabh Dave
08:47 AM Bug #43522 (Resolved): qa: update xfstests_dev to install python2 instead of python on ubuntu 19
Rishabh Dave
08:42 AM Bug #43393 (In Progress): qa: add support/qa for cephfs-shell on CentOS 9 / RHEL9
Rishabh Dave

01/07/2020

10:11 PM Bug #42238 (Resolved): cephfs-shell: setxattr() is passed extra length argument
Patrick Donnelly
10:07 PM Bug #43517 (Resolved): qa: random subvolumegroup collision
... Patrick Donnelly
09:59 PM Bug #43336 (Resolved): qa: test_unmount_for_evicted_client hangs
Patrick Donnelly
09:58 PM Cleanup #42563 (Resolved): mds: reorg MDSTableServer header
Patrick Donnelly
09:57 PM Cleanup #42690 (Resolved): mds: reorg Mutation header
Patrick Donnelly
09:56 PM Bug #43438 (Pending Backport): cephfs-journal-tool: will crash without any extra argument
Patrick Donnelly
09:21 PM Bug #43516 (Fix Under Review): qa: verify sub-suite does not define os_version
Patrick Donnelly
09:14 PM Bug #43516 (Resolved): qa: verify sub-suite does not define os_version
... Patrick Donnelly
09:12 PM Feature #24461: cephfs: improve file create performance buffering file unlink/create operations
Yes, setting a zero length i_xattrs buffer on the new inode seems to have corrected the problem. I believe what was h... Jeff Layton
05:24 PM Feature #24461: cephfs: improve file create performance buffering file unlink/create operations
At this point, I'm 90% sure the problem is in xattrs. Basically, after creating the file async we're leaving the i_xa... Jeff Layton
04:00 PM Feature #24461: cephfs: improve file create performance buffering file unlink/create operations
I threw in a hack to do this:... Jeff Layton
02:44 PM Feature #24461: cephfs: improve file create performance buffering file unlink/create operations
also see https://github.com/ceph/ceph/pull/30969 Zheng Yan
02:42 PM Feature #24461: cephfs: improve file create performance buffering file unlink/create operations
For current async create code. ceph_mds_reply_inode::max_size is 0. client can't write to the new file until it gets ... Zheng Yan
02:00 PM Feature #24461: cephfs: improve file create performance buffering file unlink/create operations
dynamic debugging from the client, with async dirops disabled. This is during the write calls:... Jeff Layton
01:32 PM Feature #24461: cephfs: improve file create performance buffering file unlink/create operations
I can use strace to get timing statistics on individual calls though. With async dirops disabled:... Jeff Layton
01:18 PM Feature #24461: cephfs: improve file create performance buffering file unlink/create operations
Now that I look closer, I don't think strace -c is measuring what we need. It's looking at CPU time in each syscall. ... Jeff Layton
01:16 AM Feature #24461: cephfs: improve file create performance buffering file unlink/create operations
Thanks for checking. I'll have to play around with this more myself. Patrick Donnelly
09:07 PM Bug #43515 (Fix Under Review): qa: SyntaxError: invalid token
Patrick Donnelly
09:06 PM Bug #43515 (Resolved): qa: SyntaxError: invalid token
... Patrick Donnelly
08:53 PM Bug #43514 (Fix Under Review): qa: test setUp may cause spurious MDS_INSUFFICIENT_STANDBY
Patrick Donnelly
08:48 PM Bug #43514 (Resolved): qa: test setUp may cause spurious MDS_INSUFFICIENT_STANDBY
... Patrick Donnelly
08:33 PM Bug #43513 (Resolved): qa: filelock_interrupt.py hang
... Patrick Donnelly
02:45 PM Backport #43509 (Resolved): nautilus: 'ceph -s' does not show standbys if there are no filesystems
https://github.com/ceph/ceph/pull/32912 Nathan Cutler
02:44 PM Backport #43506 (Resolved): nautilus: MDSMonitor: warn if a new file system is being created with...
https://github.com/ceph/ceph/pull/32600 Nathan Cutler
02:44 PM Backport #43505 (Rejected): mimic: MDSMonitor: warn if a new file system is being created with an...
Nathan Cutler
02:42 PM Backport #43503 (Resolved): nautilus: mount.ceph: give a hint message when no mds is up or cluste...
https://github.com/ceph/ceph/pull/32910 Nathan Cutler
02:42 PM Backport #43502 (Resolved): mimic: mount.ceph: give a hint message when no mds is up or cluster i...
https://github.com/ceph/ceph/pull/32911 Nathan Cutler
12:43 PM Documentation #43154 (In Progress): doc: migrate best practice recommendations to relevant docs
Rishabh Dave
12:34 PM Bug #43496: qa: xfstest_dev.py crashes while calling teuthology.misc.get_system_type
Oh, BTW, the crash happened locally, not on teuthology. Rishabh Dave
12:33 PM Bug #43496 (Fix Under Review): qa: xfstest_dev.py crashes while calling teuthology.misc.get_syste...
Rishabh Dave
12:30 PM Bug #43496 (Resolved): qa: xfstest_dev.py crashes while calling teuthology.misc.get_system_type
teuthology.misc.get_system_type calls teuthology.misc.sh. Fix: add a wrapper method of teuthology.misc.sh to vstart_r... Rishabh Dave
12:33 PM Bug #43486 (Fix Under Review): qa: test_acls: cannot find packages on centos 8
Rishabh Dave
08:00 AM Bug #43486: qa: test_acls: cannot find packages on centos 8
I am checking if xfstests-dev runs fine without btrfs-prog-devel. The reason for it's absence on CentOS 8 is (AFAIS) ... Rishabh Dave
07:14 AM Bug #43486 (In Progress): qa: test_acls: cannot find packages on centos 8
Rishabh Dave
11:07 AM Bug #43483 (In Progress): mds: reject forward scrubs when cluster has multiple active MDS (more t...
Milind Changire
08:42 AM Backport #43338 (New): nautilus: qa/tasks: add remaining tests for fs volume
Jos Collin
03:07 AM Bug #43493 (Can't reproduce): osdc: fix null pointer caused program crash
PurgeRange.oncommit NULL error wenpeng li

01/06/2020

09:40 PM Bug #43329 (Resolved): cephfs-shell: AttributeError when undefined an conf opt is attemptted to read
Patrick Donnelly
08:42 PM Documentation #37746 (Resolved): doc: how to mount a subdir with ceph-fuse/kclient
Patrick Donnelly
08:35 PM Bug #43460 (Resolved): qa: loff_t type missing for fsync-tester
Patrick Donnelly
08:31 PM Fix #42450 (Pending Backport): MDSMonitor: warn if a new file system is being created with an EC ...
Patrick Donnelly
08:28 PM Bug #43326 (Resolved): mds: batch getattr/lookup bug
Patrick Donnelly
08:27 PM Bug #42088 (Pending Backport): 'ceph -s' does not show standbys if there are no filesystems
Patrick Donnelly
08:20 PM Feature #43294 (Pending Backport): mount.ceph: give a hint message when no mds is up or cluster i...
Patrick Donnelly
07:48 PM Bug #43487 (Fix Under Review): qa: test_acls does not detect rhel8
Patrick Donnelly
07:46 PM Bug #43487 (Resolved): qa: test_acls does not detect rhel8
... Patrick Donnelly
07:44 PM Bug #43486 (Resolved): qa: test_acls: cannot find packages on centos 8
... Patrick Donnelly
06:24 PM Bug #43484 (Fix Under Review): mds: note features client has when rejecting client due to feature...
Patrick Donnelly
06:15 PM Bug #43484 (Resolved): mds: note features client has when rejecting client due to feature incompat
Currently we get a message like:... Patrick Donnelly
03:17 PM Bug #43407: mds crash after update to v14.2.5
The first ESubtreeMap in the journal was wrong. It should also contains dir 0x1... Zheng Yan
02:52 PM Bug #43407 (Triaged): mds crash after update to v14.2.5
Patrick Donnelly
03:07 PM Bug #43483 (Resolved): mds: reject forward scrubs when cluster has multiple active MDS (more than...
Forward scrub may cause the MDS to hit various assertions if there is more than one rank. Have the MDS check if there... Patrick Donnelly
02:40 PM Bug #43440 (Triaged): client: chdir does not raise error if a file is passed
Patrick Donnelly

01/03/2020

11:49 PM Bug #43460 (Fix Under Review): qa: loff_t type missing for fsync-tester
Patrick Donnelly
11:44 PM Bug #43460 (Resolved): qa: loff_t type missing for fsync-tester
... Patrick Donnelly
11:35 PM Bug #43459 (Fix Under Review): qa: FATAL ERROR: libtool does not seem to be installed.
Patrick Donnelly
11:28 PM Bug #43459 (In Progress): qa: FATAL ERROR: libtool does not seem to be installed.
Patrick Donnelly
11:24 PM Bug #43459 (Resolved): qa: FATAL ERROR: libtool does not seem to be installed.
... Patrick Donnelly
07:34 PM Bug #43407: mds crash after update to v14.2.5
Status update:
I have tried
cephfs-journal-tool event recover_dentries summary
followed with
cephfs-journal-tool...
Marco Savoca
03:29 PM Bug #43407: mds crash after update to v14.2.5
> 2. recover journal events:
> cephfs-journal-tool journal export backup.bin
Do you mean
_cephfs-journal-tool ev...
Marco Savoca
02:23 PM Bug #43407: mds crash after update to v14.2.5
mds shows there are some ENoOp log events. This means some region of mds log was erased by cephfs-journal-tools. Why ... Zheng Yan
11:18 AM Feature #24461: cephfs: improve file create performance buffering file unlink/create operations
I built a tree based on 1e2fe722c41d4cc34094afb157b3eb06b4a50972, which is the commit just before the merge of Zheng'... Jeff Layton
03:14 AM Feature #24461: cephfs: improve file create performance buffering file unlink/create operations
Patrick Donnelly wrote:
> Zheng Yan wrote:
> > Patrick Donnelly wrote:
> > > The baseline performance is surprisin...
Zheng Yan
01:51 AM Feature #43423: mds: collect and show the dentry lease metric
Patches are ready and waiting for the depending PR [1] to be merged.
[1] https://github.com/ceph/ceph/pull/26004
Xiubo Li

01/02/2020

09:19 PM Bug #43407: mds crash after update to v14.2.5
Yes I had 3 filesystems (namespaces), one for every mds daemon, and the setup was working up to the update to v14.2.5... Marco Savoca
07:00 PM Bug #43407: mds crash after update to v14.2.5
Were you using multiple MDS before?
Can you increase MDS debugging:
ceph config set mds debug_mds 10
and res...
Patrick Donnelly
08:10 PM Feature #24461: cephfs: improve file create performance buffering file unlink/create operations
Looking at my home-grown testcase, the results look pretty good, but an untarring a random kernel tarball is consider... Jeff Layton

01/01/2020

09:38 AM Documentation #43154: doc: migrate best practice recommendations to relevant docs
https://docs.ceph.com/docs/master/cephfs/fuse/ - This is the location of the FUSE docs. Zac Dover

12/31/2019

12:46 PM Bug #43440 (Resolved): client: chdir does not raise error if a file is passed
... Varsha Rao
06:14 AM Feature #41566 (In Progress): mds: support rolling upgrades
Jos Collin
04:13 AM Feature #43435: kclient:send client provided metric flags in client metadata
Patch is ready and the test output is:... Xiubo Li
04:11 AM Bug #43438 (Fix Under Review): cephfs-journal-tool: will crash without any extra argument
Xiubo Li
04:10 AM Bug #43438: cephfs-journal-tool: will crash without any extra argument
The fixing PR: https://github.com/ceph/ceph/pull/32452 Xiubo Li
04:01 AM Bug #43438 (In Progress): cephfs-journal-tool: will crash without any extra argument
Xiubo Li
04:00 AM Bug #43438 (Resolved): cephfs-journal-tool: will crash without any extra argument
... Xiubo Li

12/30/2019

04:16 PM Bug #41565 (Fix Under Review): mds: detect MDS<->MDS messages that are not versioned
Jos Collin
05:42 AM Feature #43435 (In Progress): kclient:send client provided metric flags in client metadata
Xiubo Li
05:42 AM Feature #43435 (Resolved): kclient:send client provided metric flags in client metadata
This will send the kclient provided metric flags to the MDS server.
Xiubo Li

12/27/2019

11:09 AM Cleanup #43426 (Fix Under Review): mds: reorg mdstypes header
Varsha Rao

12/26/2019

05:26 PM Cleanup #43425 (Fix Under Review): mds: reorg snap header
Varsha Rao
03:01 PM Cleanup #43425 (Resolved): mds: reorg snap header
Varsha Rao
03:03 PM Cleanup #43426 (Resolved): mds: reorg mdstypes header
Varsha Rao
03:02 PM Cleanup #43424 (Fix Under Review): mds: reorg inode_backtrace header
Varsha Rao
01:58 PM Cleanup #43424 (Resolved): mds: reorg inode_backtrace header
Varsha Rao
06:15 AM Feature #43423: mds: collect and show the dentry lease metric
https://tracker.ceph.com/issues/24285 Xiubo Li
06:12 AM Feature #43423: mds: collect and show the dentry lease metric
Locally the patch is ready, but depend on https://github.com/ceph/ceph/pull/26004, which hasn't been merged yet.
<...
Xiubo Li
06:10 AM Feature #43423 (Resolved): mds: collect and show the dentry lease metric
Kclient will collect the dentry lease metric and send it to the MDS, currently this hasn't been shown in the perf stats. Xiubo Li

12/25/2019

11:20 AM Bug #43410 (Won't Fix): mds:When the directory level is above 3000, the following assertions will...
Zheng Yan

12/24/2019

09:02 AM Cleanup #43418 (Fix Under Review): mds: reorg flock header
Varsha Rao
07:38 AM Cleanup #43418 (Resolved): mds: reorg flock header
Varsha Rao
06:19 AM Bug #43410: mds:When the directory level is above 3000, the following assertions will appear
Zheng Yan wrote:
> mds call FOO::adjust_nested_auth_pins functions for each directory level, which caused stack over...
huanwen ren
06:14 AM Bug #43410: mds:When the directory level is above 3000, the following assertions will appear
Zheng Yan wrote:
> full calltrace ?
I use gdb for mounting, this information is complete
huanwen ren
02:31 AM Bug #43410: mds:When the directory level is above 3000, the following assertions will appear
mds call FOO::adjust_nested_auth_pins functions for each directory level, which caused stack overflow. mimic and late... Zheng Yan
01:53 AM Bug #43410: mds:When the directory level is above 3000, the following assertions will appear
full calltrace ? Zheng Yan

12/23/2019

11:14 AM Cleanup #43408 (Fix Under Review): mds: reorg StrayManager header
Varsha Rao
10:51 AM Cleanup #43408 (Resolved): mds: reorg StrayManager header
Varsha Rao
11:14 AM Bug #43410: mds:When the directory level is above 3000, the following assertions will appear
I don't see any exceptions in the log print
@Patrick Donnelly
@Zheng Yan
huanwen ren
11:10 AM Bug #43410 (Won't Fix): mds:When the directory level is above 3000, the following assertions will...
When I use the script to continuously create directories,
but the directory level is above 3000, the following asse...
huanwen ren
11:10 AM Bug #43409 (Closed): mds:When the directory level is above 3000, the following assertions will ap...
huanwen ren
11:09 AM Bug #43409 (Closed): mds:When the directory level is above 3000, the following assertions will ap...
When I use the script to continuously create directories,
but the directory level is above 3000, the following asse...
huanwen ren

12/22/2019

11:36 PM Bug #43407 (Triaged): mds crash after update to v14.2.5
All MDS crashed and not able to restart after update from v14.2.4 to v14.2.5
*systemctl status:*...
Marco Savoca

12/21/2019

03:57 AM Feature #24461: cephfs: improve file create performance buffering file unlink/create operations
Zheng Yan wrote:
> Patrick Donnelly wrote:
> > The baseline performance is surprising I think. That's with the same...
Patrick Donnelly

12/20/2019

12:29 AM Bug #42723 (In Progress): pybind/mgr/volumes: add upgrade testing
Patrick Donnelly

12/19/2019

08:31 PM Bug #43393 (Resolved): qa: add support/qa for cephfs-shell on CentOS 9 / RHEL9
Patrick Donnelly
08:00 PM Documentation #41688 (Resolved): doc: client config reference improvements
Patrick Donnelly
07:57 PM Bug #43250 (Resolved): qa/test_cephfs_shell: TestDu.test_du_works_for_hardlinks fails
Patrick Donnelly
07:46 PM Bug #43392 (Resolved): MDSMonitor: support automatic failover to standbys with stronger affinity
Initial work by Sage: https://github.com/ceph/ceph/pull/32015
The next step is to failover to a standby with stron...
Patrick Donnelly
02:38 PM Bug #43329 (Fix Under Review): cephfs-shell: AttributeError when undefined an conf opt is attempt...
Rishabh Dave
02:30 PM Cleanup #43387 (Fix Under Review): mds: reorg SnapServer header
Varsha Rao
02:23 PM Cleanup #43387 (Resolved): mds: reorg SnapServer header
Varsha Rao
02:16 PM Cleanup #43386 (Fix Under Review): mds: reorg SnapRealm header
Varsha Rao
02:08 PM Cleanup #43386 (Resolved): mds: reorg SnapRealm header
Varsha Rao
01:48 PM Feature #39129: create mechanism to delegate ranges of inode numbers to client
You're right. I just pushed a patch to be squashed in on top of the existing series. I'm testing it now with the clie... Jeff Layton
01:40 PM Feature #39129: create mechanism to delegate ranges of inode numbers to client
Jeff Layton wrote:
> That's not a bad idea. We'd have to keep track of a separate set of newly-added ino_t's to send...
Zheng Yan
12:09 PM Feature #39129: create mechanism to delegate ranges of inode numbers to client
That's not a bad idea. We'd have to keep track of a separate set of newly-added ino_t's to send in the reply, but tha... Jeff Layton
09:01 AM Feature #39129: create mechanism to delegate ranges of inode numbers to client
Jeff Layton wrote:
> I have patches for this for the MDS, and the kernel, but I keep hitting a race where the client...
Zheng Yan
 

Also available in: Atom