Activity
From 08/23/2020 to 09/21/2020
09/21/2020
- 09:53 PM Bug #47294: client: thread hang in Client::_setxattr_maybe_wait_for_osdmap
- All right, I'm going to shove some more debug information in Objecter and Monitor.
- 12:40 AM Bug #47294: client: thread hang in Client::_setxattr_maybe_wait_for_osdmap
- Xiubo Li wrote:
> Patrick Donnelly wrote:
> > Xiubo Li wrote:
> > > Hi Patrick,
> > >
> > > For this let's add ... - 09:04 PM Bug #47526 (Resolved): qa: RuntimeError: FSCID 2 not in map
- 09:02 PM Bug #36389: untar encounters unexpected EPERM on kclient/multimds cluster with thrashing
- ...
- 08:32 PM Bug #47565 (Resolved): qa: "client.4606 isn't responding to mclientcaps(revoke), ino 0x200000007d...
- ...
- 07:47 PM Bug #47563 (Resolved): qa: kernel client closes session improperly causing eviction due to timeout
- ...
- 05:08 PM Bug #45835 (Pending Backport): mds: OpenFileTable::prefetch_inodes during rejoin can cause out-of...
- 03:47 PM Bug #45835: mds: OpenFileTable::prefetch_inodes during rejoin can cause out-of-memory
- The fix was merged. Something needed to start the backports process?
- 03:21 PM Backport #47152: nautilus: pybind/mgr/volumes: add debugging for global lock
- Patrick Donnelly wrote:
> https://github.com/ceph/ceph/pull/36828
merged - 03:20 PM Backport #46948: nautilus: qa: Fs cleanup fails with a traceback
- Nathan Cutler wrote:
> https://github.com/ceph/ceph/pull/36714
merged - 03:20 PM Backport #46592: nautilus: ceph-fuse: ceph-fuse process is terminated by the logratote task and w...
- Nathan Cutler wrote:
> https://github.com/ceph/ceph/pull/36181
merged - 01:51 PM Feature #47162: mds: handle encrypted filenames in the MDS for fscrypt
- Jeff Layton wrote:
> Xiubo Li wrote:
> > Ceph has its own base64 encode/decode logic already in src/common/armor.c,... - 01:42 PM Feature #47162: mds: handle encrypted filenames in the MDS for fscrypt
- Xiubo Li wrote:
> Ceph has its own base64 encode/decode logic already in src/common/armor.c, which is the same with ... - 01:39 PM Feature #47162: mds: handle encrypted filenames in the MDS for fscrypt
I am planing to append a `fscrypt.alternate_name : ${raw_ciphertext}` pair to the xattr map when doing the create d...- 04:07 AM Feature #47162: mds: handle encrypted filenames in the MDS for fscrypt
- Ceph has its own base64 encode/decode logic already in src/common/armor.c, which is the same with the kernel does.
09/20/2020
- 11:01 AM Feature #47162: mds: handle encrypted filenames in the MDS for fscrypt
- Jeff Layton wrote:
> Xiubo Li wrote:
> >
> > Yeah, this looks good.
> >
> > BTW, what the alternat_name will s...
09/19/2020
- 05:56 PM Bug #47389: ceph fs volume create fails to create pool
- Hi Joshua,
I tried on the master and it works for me. The HEAD was at 240c46a75a44cb9363cf994cb264e9d7048c98a1 dat... - 12:29 AM Bug #47512 (Pending Backport): mgr/nfs: Cluster creation throws 'NoneType' object has no attribut...
- 12:27 AM Bug #47423 (Resolved): volume rm throws Permissioned denied error
- 12:24 AM Bug #47353 (Pending Backport): mds: purge_queue's _calculate_ops is inaccurate
09/18/2020
- 11:26 PM Bug #47518 (Resolved): qa: spawn MDS daemons before creating file system
- 11:22 PM Backport #47249 (In Progress): octopus: mon: deleting a CephFS and its pools causes MONs to crash
- 11:19 PM Backport #47248 (In Progress): nautilus: mon: deleting a CephFS and its pools causes MONs to crash
- 04:43 PM Bug #47499: Simultaneous MDS and OSD crashes when answering to client
- It just happened again on a different MDS with a different client and I found something in common. In all the crashes...
- 04:12 PM Bug #47526 (Fix Under Review): qa: RuntimeError: FSCID 2 not in map
- 02:02 AM Bug #47526 (Resolved): qa: RuntimeError: FSCID 2 not in map
- ...
- 03:42 PM Backport #46786 (In Progress): octopus: client: in _open() the open ref maybe decreased twice, bu...
- 03:41 PM Backport #46783 (In Progress): octopus: mds/CInode: Optimize only pinned by subtrees check
- 03:29 PM Backport #46637 (In Progress): octopus: mds: optimize ephemeral rand pin
- 02:56 PM Backport #46636 (In Progress): octopus: mds: null pointer dereference in MDCache::finish_rollback
- 02:53 PM Backport #46634 (In Progress): octopus: mds forwarding request 'no_available_op_found'
- 11:55 AM Feature #47162: mds: handle encrypted filenames in the MDS for fscrypt
- Xiubo Li wrote:
>
> Yeah, this looks good.
>
> BTW, what the alternat_name will store ? The full ciphertext bin... - 11:04 AM Feature #47162: mds: handle encrypted filenames in the MDS for fscrypt
- Zheng Yan wrote:
> Jeff Layton wrote:
[...]
> > I think that approach will give us the most flexibility going forw... - 10:39 AM Feature #47162: mds: handle encrypted filenames in the MDS for fscrypt
- Jeff Layton wrote:
> Xiubo Li wrote:
> >
> > Yeah, right.
> >
> > If the master key is absent, for the ->looku... - 11:04 AM Backport #47259 (In Progress): nautilus: client: FAILED assert(dir->readdir_cache[dirp->cache_ind...
- 10:59 AM Backport #47254 (In Progress): nautilus: client: Client::open() pass wrong cap mask to path_walk
- 10:52 AM Backport #47252 (In Progress): nautilus: mds: fix possible crash when the MDS is stopping
- 10:49 AM Backport #47246 (In Progress): nautilus: qa: Replacing daemon mds.a as rank 0 with standby daemon...
- 01:29 AM Bug #47444 (Resolved): crash in FSMap::parse_role
09/17/2020
- 04:27 PM Bug #47518 (Fix Under Review): qa: spawn MDS daemons before creating file system
- 04:01 PM Bug #47518 (Resolved): qa: spawn MDS daemons before creating file system
- ...
- 03:52 PM Feature #47162: mds: handle encrypted filenames in the MDS for fscrypt
- Jeff Layton wrote:
> Probably something like the last one. I think we're best off avoiding any logic that requires t... - 03:36 PM Feature #47162: mds: handle encrypted filenames in the MDS for fscrypt
- Xiubo Li wrote:
>
> Yeah, right.
>
> If the master key is absent, for the ->lookup() the client will tell MDS t... - 05:19 AM Feature #47162: mds: handle encrypted filenames in the MDS for fscrypt
- Xiubo Li wrote:
> Jeff Layton wrote:
> > Xiubo Li wrote:
[...]
> Yeah, since the long name case is rare, and ... - 05:09 AM Feature #47162: mds: handle encrypted filenames in the MDS for fscrypt
- Jeff Layton wrote:
> Xiubo Li wrote:
> > Hi Jeff,
> >
> > One question:
> >
> > Currently the ext4 will just ... - 12:44 PM Bug #47515: pybind/snap_schedule: deactivating a schedule is ineffective
- (formatting fix)...
- 12:43 PM Bug #47515 (Resolved): pybind/snap_schedule: deactivating a schedule is ineffective
- Deactivating a snap schedule does not have any effect on the schedule. Schedules snapshots still get created by the s...
- 10:51 AM Bug #47512 (Fix Under Review): mgr/nfs: Cluster creation throws 'NoneType' object has no attribut...
- 10:45 AM Bug #47512 (Resolved): mgr/nfs: Cluster creation throws 'NoneType' object has no attribute 'repla...
- ...
09/16/2020
- 02:02 PM Bug #47499 (New): Simultaneous MDS and OSD crashes when answering to client
- We observed 4 MDSes and 2 OSDs segfaulting simultaneously when answering to one client. All the six tracebacks report...
- 12:27 PM Feature #47162: mds: handle encrypted filenames in the MDS for fscrypt
- Xiubo Li wrote:...
- 11:25 AM Feature #47162: mds: handle encrypted filenames in the MDS for fscrypt
- Xiubo Li wrote:
>
> With this approach there seems no need to covert the ciphertext to base64-encode text when s... - 11:16 AM Feature #47162: mds: handle encrypted filenames in the MDS for fscrypt
- Xiubo Li wrote:
> Hi Jeff,
>
> One question:
>
> Currently the ext4 will just store the ciphertext as the fina... - 08:10 AM Feature #47162: mds: handle encrypted filenames in the MDS for fscrypt
- For the 2nd approach, suggeset by Zheng, more detail in my mind is:
If we will store both the "based64-encoded-pla... - 06:13 AM Feature #47162: mds: handle encrypted filenames in the MDS for fscrypt
- Hi Jeff,
One question:
Currently the ext4 will just store the ciphertext as the final filename to the disk, and... - 02:22 AM Feature #47162: mds: handle encrypted filenames in the MDS for fscrypt
- From the source code, the encoded filename length will be roughly increased to 4/3 of the original filename....
- 12:05 PM Bug #47423 (Fix Under Review): volume rm throws Permissioned denied error
- 07:05 AM Feature #47490 (Pending Backport): Integration of dashboard with volume/nfs module
- Currently, there are two ways to create exports with mgr/volume/nfs module and
dashboard. Both use the same code[1]... - 03:33 AM Backport #47090 (In Progress): nautilus: After restarting an mds, its standy-replay mds remained ...
- 03:30 AM Backport #47088 (In Progress): nautilus: mds: recover files after normal session close
- 03:27 AM Backport #47084 (Need More Info): nautilus: mds: 'forward loop' when forward_all_requests_to_auth...
- Zheng, the backport for this is non-trivial. Can you take a look?
- 03:25 AM Backport #47017 (In Progress): nautilus: mds: kcephfs parse dirfrag's ndist is always 0
- 02:46 AM Bug #47488: Apparent deadlock in tasks.mgr.dashboard.test_cephfs.CephfsTest.test_snapshots
- To progress this further we really need more/better logs. Created https://github.com/ceph/ceph/pull/37176 to assist i...
- 02:40 AM Bug #47488 (New): Apparent deadlock in tasks.mgr.dashboard.test_cephfs.CephfsTest.test_snapshots
- /a/yuriw-2020-09-02_17:33:04-rados-wip-yuri-master-baseline-9.2.2020-distro-basic-smithi/5400010...
- 01:49 AM Bug #47294: client: thread hang in Client::_setxattr_maybe_wait_for_osdmap
- Patrick Donnelly wrote:
> Xiubo Li wrote:
> > Hi Patrick,
> >
> > For this let's add more debug logs to check wh...
09/15/2020
- 07:32 PM Bug #47294: client: thread hang in Client::_setxattr_maybe_wait_for_osdmap
- Xiubo Li wrote:
> Hi Patrick,
>
> For this let's add more debug logs to check where it is stucked in ?
>
> I w... - 01:09 PM Bug #47294: client: thread hang in Client::_setxattr_maybe_wait_for_osdmap
- Hi Patrick,
For this let's add more debug logs to check where it is stucked in ?
I went through the client_loc... - 12:54 PM Bug #47423 (In Progress): volume rm throws Permissioned denied error
- 11:22 AM Feature #47277: implement new mount "device" syntax for kcephfs
- Patrick Donnelly wrote:
> Venky Shankar wrote:
> > Patrick Donnelly wrote:
> > > Venky Shankar wrote:
> > > > Pat...
09/14/2020
- 10:00 PM Feature #47277: implement new mount "device" syntax for kcephfs
- There are other alternates too, fwiw (e.g.):
name@fs#/path
...or maybe just omit the ':' or anything to rep... - 08:52 PM Feature #47277: implement new mount "device" syntax for kcephfs
- Venky Shankar wrote:
> Patrick Donnelly wrote:
> > Venky Shankar wrote:
> > > Patrick Donnelly wrote:
> > > > Jef... - 12:33 PM Feature #47277: implement new mount "device" syntax for kcephfs
- Patrick Donnelly wrote:
> Venky Shankar wrote:
> > Patrick Donnelly wrote:
> > > Jeff Layton wrote:
> > > > Propo... - 09:00 PM Bug #47423: volume rm throws Permissioned denied error
- Rishabh Dave wrote:
> Unlike @volume rm@, @fs fail@ does not fail -
>
> [...]
>
> @volume rm@ too runs @fs fai... - 03:12 PM Bug #47423: volume rm throws Permissioned denied error
- The issue with ticket assignee was because my page wasn't refreshed before hitting submit button.
- 03:11 PM Bug #47423: volume rm throws Permissioned denied error
- Unlike @volume rm@, @fs fail@ does not fail -...
- 02:39 PM Bug #47423: volume rm throws Permissioned denied error
- ...
- 12:41 PM Bug #47423: volume rm throws Permissioned denied error
- From what I see on master in my local repo, this issue (getting @Permissioned denied@ on @volume rm@) is not just lim...
- 08:51 AM Bug #47423: volume rm throws Permissioned denied error
- Kefu Chai wrote:
> i suspect that it is https://github.com/ceph/ceph/pull/32581 which broke `test_cluster_set_reset_... - 12:42 AM Bug #47423: volume rm throws Permissioned denied error
- i suspect that it is https://github.com/ceph/ceph/pull/32581 which broke `test_cluster_set_reset_user_config` in `tas...
- 08:30 PM Documentation #47449 (New): doc: complete ec pool configuration section with an example
- https://docs.ceph.com/docs/master/cephfs/createfs/#using-erasure-coded-pools-with-cephfs
The section should provid... - 07:41 PM Bug #47444 (Fix Under Review): crash in FSMap::parse_role
- 07:19 PM Bug #47444 (In Progress): crash in FSMap::parse_role
- 05:19 PM Bug #47444 (Resolved): crash in FSMap::parse_role
- ...
- 01:34 PM Backport #47200 (In Progress): octopus: scheduled cephfs snapshots (via ceph manager)
09/13/2020
- 06:02 PM Bug #47423 (Triaged): volume rm throws Permissioned denied error
- 05:59 PM Bug #47389 (Triaged): ceph fs volume create fails to create pool
- 05:55 PM Feature #47277: implement new mount "device" syntax for kcephfs
- Venky Shankar wrote:
> Patrick Donnelly wrote:
> > Jeff Layton wrote:
> > > Proposed syntax looks wrong in the des... - 05:48 PM Bug #47379 (Rejected): mds: mark no warn on killed request
- PR was rejected
- 05:47 PM Bug #47353 (Fix Under Review): mds: purge_queue's _calculate_ops is inaccurate
09/12/2020
- 07:22 PM Backport #47248 (Need More Info): nautilus: mon: deleting a CephFS and its pools causes MONs to c...
- non-trivial because it depends on a series of changes to qa/tasks/cephfs/mount.py that have not been backported
- 07:22 PM Backport #47249 (Need More Info): octopus: mon: deleting a CephFS and its pools causes MONs to crash
- non-trivial because it depends on a series of changes to qa/tasks/cephfs/mount.py that have not been backported
- 06:55 PM Backport #47086 (Need More Info): nautilus: common: validate type CephBool cause 'invalid command...
- must be backported together with the fix for #47179
- 06:55 PM Backport #47085 (Need More Info): octopus: common: validate type CephBool cause 'invalid command ...
- must be backported together with the fix for #47179
- 11:54 AM Bug #47423 (Resolved): volume rm throws Permissioned denied error
- ...
09/11/2020
- 10:39 PM Bug #46985: common: validate type CephBool cause 'invalid command json'
- https://github.com/ceph/ceph/pull/37098 fixes a bug in https://github.com/ceph/ceph/pull/36459 and needs backport too.
- 03:03 AM Bug #46985: common: validate type CephBool cause 'invalid command json'
- This change causes the failure seen in #47179. Could we either revert it or modify it so it reinstates the old behavi...
09/10/2020
09/09/2020
- 05:59 PM Feature #47277: implement new mount "device" syntax for kcephfs
- One idea might be to just get rid of the ':' ?
name@fsname[.fscid]/path
...but that fsname/path looks like ... - 01:40 PM Feature #47277: implement new mount "device" syntax for kcephfs
- Venky Shankar wrote:
>
> The "=" is a bit offputting. FWIW, mount helper tries to resolve (parse host/IP:port + ge... - 12:26 PM Feature #47162 (In Progress): mds: handle encrypted filenames in the MDS for fscrypt
- 10:57 AM Bug #47379 (Rejected): mds: mark no warn on killed request
- It is unnecessary to report slow request on killed ones, otherwise cause continous false alarms.
09/08/2020
- 09:30 PM Bug #47367 (New): mgr/volumes: volumes plugin does not ensure passed in subvolume name does not h...
- The volumes plugin does not check and ensure that a subvolume name is passed as the parameter to calls that require t...
- 07:10 PM Feature #40401 (Fix Under Review): mgr/volumes: allow/deny r/rw access of auth IDs to subvolume a...
- 01:53 PM Feature #47161 (Rejected): mds: add dedicated field to inode for fscrypt context
- Fair enough then. I'll keep working with this as an xattr for now. Let's go ahead and close this out then, and I'll r...
- 06:19 AM Bug #47353 (Resolved): mds: purge_queue's _calculate_ops is inaccurate
- ...
- 04:45 AM Bug #47268 (Resolved): pybind/snap_schedule: scheduled snapshots get pruned just after creation
09/07/2020
- 08:31 PM Backport #47317 (In Progress): nautilus: mds: CDir::_omap_commit(int): Assertion `committed_versi...
- 08:25 PM Backport #47316 (In Progress): octopus: mds: CDir::_omap_commit(int): Assertion `committed_versio...
- 08:25 PM Backport #46520 (In Progress): octopus: mds: deleting a large number of files in a directory caus...
- 08:28 AM Backport #46520: octopus: mds: deleting a large number of files in a directory causes the file sy...
- sorry, I made a mistake.
reset state to need more info. - 08:27 AM Backport #46520 (Need More Info): octopus: mds: deleting a large number of files in a directory c...
- 08:22 AM Backport #46520 (In Progress): octopus: mds: deleting a large number of files in a directory caus...
- 10:20 AM Feature #47277: implement new mount "device" syntax for kcephfs
- Patrick Donnelly wrote:
> Jeff Layton wrote:
> > Proposed syntax looks wrong in the description. I meant this:
> >... - 10:01 AM Backport #46524 (In Progress): octopus: non-head batch requests may hold authpins and locks
- 08:31 AM Backport #46522 (In Progress): octopus: mds: fix hang issue when accessing a file under a lost pa...
- 08:12 AM Bug #47294: client: thread hang in Client::_setxattr_maybe_wait_for_osdmap
- It was neither blocked by the "client_lock", nor by the RWRef's lock, because both kept working well:
For the ti... - 08:02 AM Backport #46516 (In Progress): octopus: client: directory inode can not call release_callback
- 03:28 AM Cleanup #47160 (In Progress): qa/tasks/cephfs: Break up test_volumes.py
09/06/2020
- 10:21 AM Backport #47157: nautilus: mgr/volumes: Mark subvolumes with ceph.dir.subvolume vxattr, to improv...
- This update was made using the script "backport-resolve-issue".
backport PR https://github.com/ceph/ceph/pull/36833
m... - 10:21 AM Backport #46796: nautilus: mds: Subvolume snapshot directory does not save attribute "ceph.quota....
- This update was made using the script "backport-resolve-issue".
backport PR https://github.com/ceph/ceph/pull/36404
m... - 09:54 AM Cleanup #47325 (Fix Under Review): client: remove unneccessary client_lock for objector->write()
- 09:35 AM Cleanup #47325 (Resolved): client: remove unneccessary client_lock for objector->write()
- 09:23 AM Bug #47294: client: thread hang in Client::_setxattr_maybe_wait_for_osdmap
- Patrick Donnelly wrote:
> Xiubo Li wrote:
> > From /ceph/teuthology-archive/pdonnell-2020-09-03_02:04:14-fs-wip-pdo...
09/05/2020
- 09:10 PM Backport #47317 (Resolved): nautilus: mds: CDir::_omap_commit(int): Assertion `committed_version ...
- https://github.com/ceph/ceph/pull/37035
- 09:10 PM Backport #47316 (Resolved): octopus: mds: CDir::_omap_commit(int): Assertion `committed_version =...
- https://github.com/ceph/ceph/pull/37034
09/04/2020
- 09:08 PM Feature #40401 (In Progress): mgr/volumes: allow/deny r/rw access of auth IDs to subvolume and su...
- 06:59 PM Bug #47293 (Resolved): client: osdmap wait not protected by mounted mutex
- 02:54 AM Bug #47293 (Fix Under Review): client: osdmap wait not protected by mounted mutex
- 06:54 PM Bug #47307 (Triaged): mds: throttle workloads which acquire caps faster than the client can release
- 06:28 PM Bug #47307 (Resolved): mds: throttle workloads which acquire caps faster than the client can release
- A trivial "find" command on a large directory hierarchy will cause the client to receive caps significantly faster th...
- 05:53 PM Bug #47294: client: thread hang in Client::_setxattr_maybe_wait_for_osdmap
- Xiubo Li wrote:
> From /ceph/teuthology-archive/pdonnell-2020-09-03_02:04:14-fs-wip-pdonnell-testing-20200903.000442... - 10:13 AM Bug #47294: client: thread hang in Client::_setxattr_maybe_wait_for_osdmap
- From /ceph/teuthology-archive/pdonnell-2020-09-03_02:04:14-fs-wip-pdonnell-testing-20200903.000442-distro-basic-smith...
- 08:07 AM Bug #47294: client: thread hang in Client::_setxattr_maybe_wait_for_osdmap
- Set the "ceph.dir.subvolume" won't fetch the osdmap, only for the pool related xattrs....
- 03:23 AM Bug #47294: client: thread hang in Client::_setxattr_maybe_wait_for_osdmap
- Patrick Donnelly wrote:
> Actually, I think it's more likely the hang is in
>
> https://github.com/ceph/ceph/blob... - 05:43 PM Bug #46882 (Resolved): client: mount abort hangs: [volumes INFO mgr_util] aborting connection fro...
- I don't think this issue exists in Octopus or Nautilus? I think this is fallout from Xiubo's work on breaking the cli...
- 05:41 PM Bug #46905 (Resolved): client: cluster [WRN] evicting unresponsive client smithi122:0 (34373), af...
- 05:29 PM Feature #47102 (Resolved): mds: add perf counter for cap messages
- 01:15 PM Feature #47162: mds: handle encrypted filenames in the MDS for fscrypt
- Will start it next week.
- 10:26 AM Feature #47277: implement new mount "device" syntax for kcephfs
- I will start taking a look next week
- 06:14 AM Feature #47266: add a subcommand to change caps in a simpler and clear way
- Closing this ticket based on conversation with Patrick.
09/03/2020
- 11:18 PM Bug #47293 (In Progress): client: osdmap wait not protected by mounted mutex
- 06:12 PM Bug #47293 (Resolved): client: osdmap wait not protected by mounted mutex
- https://github.com/ceph/ceph/blob/master/src/client/Client.cc#L11619
Accessing the client members before acquiring... - 06:37 PM Bug #47201 (Pending Backport): mds: CDir::_omap_commit(int): Assertion `committed_version == 0' f...
- 06:34 PM Bug #47294: client: thread hang in Client::_setxattr_maybe_wait_for_osdmap
- Xiubo, I'd also suggest adding debugging entry/exit points for these methods. (If you're feeling motivated, debugging...
- 06:32 PM Bug #47294 (Triaged): client: thread hang in Client::_setxattr_maybe_wait_for_osdmap
- Actually, I think it's more likely the hang is in
https://github.com/ceph/ceph/blob/e4a37f6338cf39e76228492897c1f2... - 06:29 PM Bug #47294 (Resolved): client: thread hang in Client::_setxattr_maybe_wait_for_osdmap
- ...
- 05:48 PM Bug #47292 (In Progress): cephfs-shell: test_df_for_valid_file failure
- ...
- 05:35 PM Feature #47277: implement new mount "device" syntax for kcephfs
- Jeff Layton wrote:
> Proposed syntax looks wrong in the description. I meant this:
>
> [...]
>
> Note that if ... - 04:45 PM Bug #42688 (Triaged): Standard CephFS caps do not allow certain dot files to be written
- 04:35 PM Cleanup #46802 (In Progress): mds: do not use asserts for RADOS failures
- 11:31 AM Bug #47268 (Fix Under Review): pybind/snap_schedule: scheduled snapshots get pruned just after cr...
- 09:31 AM Backport #46473 (In Progress): octopus: mds: make threshold for MDS_TRIM warning configurable
- 07:49 AM Backport #46943 (In Progress): nautilus: mds: segv in MDCache::wait_for_uncommitted_fragments
- 07:45 AM Backport #46941 (In Progress): nautilus: mds: memory leak during cache drop
- 07:38 AM Backport #46787 (In Progress): nautilus: client: in _open() the open ref maybe decreased twice, b...
- 07:35 AM Backport #46784 (In Progress): nautilus: mds/CInode: Optimize only pinned by subtrees check
- 07:26 AM Backport #46633 (In Progress): nautilus: mds forwarding request 'no_available_op_found'
09/02/2020
- 05:04 PM Feature #47277: implement new mount "device" syntax for kcephfs
- Proposed syntax looks wrong in the description. I meant this:...
- 05:02 PM Feature #47277 (Resolved): implement new mount "device" syntax for kcephfs
- Currently, a mount has to pass in a device string like this:
mon_addr1,mon_addr2:/path
It's problematic for... - 04:41 PM Bug #47276: MDSMonitor: add command to rename file systems
- I think we should also rethink allowing "." in file system names. Jeff is about to open a ticket to change the mount ...
- 04:40 PM Bug #47276 (Resolved): MDSMonitor: add command to rename file systems
- We've added character restrictions on file system names but there's no mechanism for fixing a legacy file system name...
- 10:43 AM Bug #47268 (Resolved): pybind/snap_schedule: scheduled snapshots get pruned just after creation
- Sample run link with PR https://github.com/ceph/ceph/pull/34552: https://pulpito.ceph.com/vshankar-2020-09-02_08:40:2...
- 09:20 AM Feature #47266 (Closed): add a subcommand to change caps in a simpler and clear way
- I am not sure if there's a better way to do it but AFAIS changing permission flag or path within the cap isn't very c...
- 08:50 AM Feature #47264 (Resolved): "fs authorize" subcommand should work for multiple FSs too
- Currently assigning caps for a second FS to an already existing client (which holds caps for a different FS already) ...
- 05:57 AM Feature #47148: mds: get rid of the mds_lock when storing the inode backtrace to meta pool
- Currently will queue some encoding excepting the encodings which need to access the CDir/CInode members in the finish...
- 05:54 AM Feature #47148 (Fix Under Review): mds: get rid of the mds_lock when storing the inode backtrace ...
- 05:09 AM Backport #47260 (Resolved): octopus: client: FAILED assert(dir->readdir_cache[dirp->cache_index] ...
- https://github.com/ceph/ceph/pull/37370
- 05:09 AM Backport #47259 (Resolved): nautilus: client: FAILED assert(dir->readdir_cache[dirp->cache_index]...
- https://github.com/ceph/ceph/pull/37232
- 05:05 AM Backport #47255 (Resolved): octopus: client: Client::open() pass wrong cap mask to path_walk
- https://github.com/ceph/ceph/pull/37369
- 05:05 AM Backport #47254 (Resolved): nautilus: client: Client::open() pass wrong cap mask to path_walk
- https://github.com/ceph/ceph/pull/37231
- 05:05 AM Backport #47253 (Resolved): octopus: mds: fix possible crash when the MDS is stopping
- https://github.com/ceph/ceph/pull/37368
- 05:05 AM Backport #47252 (Resolved): nautilus: mds: fix possible crash when the MDS is stopping
- https://github.com/ceph/ceph/pull/37229
- 05:04 AM Backport #47249 (Resolved): octopus: mon: deleting a CephFS and its pools causes MONs to crash
- https://github.com/ceph/ceph/pull/37256
- 05:04 AM Backport #47248 (Rejected): nautilus: mon: deleting a CephFS and its pools causes MONs to crash
- https://github.com/ceph/ceph/pull/37255
- 05:04 AM Backport #47247 (Resolved): octopus: qa: Replacing daemon mds.a as rank 0 with standby daemon mds...
- https://github.com/ceph/ceph/pull/37367
- 05:04 AM Backport #47246 (Resolved): nautilus: qa: Replacing daemon mds.a as rank 0 with standby daemon md...
- https://github.com/ceph/ceph/pull/37228
- 03:58 AM Backport #47158: octopus: mgr/volumes: Mark subvolumes with ceph.dir.subvolume vxattr, to improve...
- Shyamsundar Ranganathan wrote:
> Conflicts (and also depends) with backports in https://github.com/ceph/ceph/pull/36... - 03:56 AM Backport #47158 (Need More Info): octopus: mgr/volumes: Mark subvolumes with ceph.dir.subvolume v...
- Changing status to reflect that issue is waiting for an external event.
09/01/2020
- 01:53 PM Bug #47236 (New): Getting "Cannot send after transport endpoint shutdown" after changing subvolum...
- In an Ubuntu 20.04 environment with Ceph Nautilus (ceph version 14.2.11-99-gaf0268dc91 (af0268dc910f84b47655e83a83ca5...
08/31/2020
- 08:44 PM Bug #47224 (Resolved): various quota failures
- https://pulpito.ceph.com/pdonnell-2020-08-31_20:09:39-fs-master-distro-basic-smithi/
https://pulpito.ceph.com/pdon... - 08:23 PM Bug #47202 (Pending Backport): qa: Replacing daemon mds.a as rank 0 with standby daemon mds.b" in...
- 08:22 PM Feature #47168 (Resolved): client: support getting ceph.dir.rsnaps vxattr
- 08:21 PM Bug #47125 (Pending Backport): mds: fix possible crash when the MDS is stopping
- 08:19 PM Bug #42365 (Pending Backport): client: FAILED assert(dir->readdir_cache[dirp->cache_index] == dn)
- 08:06 PM Bug #47182 (Pending Backport): mon: deleting a CephFS and its pools causes MONs to crash
- 11:58 AM Backport #47158: octopus: mgr/volumes: Mark subvolumes with ceph.dir.subvolume vxattr, to improve...
- Conflicts (and also depends) with backports in https://github.com/ceph/ceph/pull/36803
Awaiting merge of the above... - 11:30 AM Feature #3244: qa: integrate Ganesha into teuthology testing to regularly exercise Ganesha CephFS...
- closed already via https://github.com/ceph/ceph/blob/master/qa/tasks/cephfs/test_nfs.py ?
- 09:15 AM Bug #47201 (Fix Under Review): mds: CDir::_omap_commit(int): Assertion `committed_version == 0' f...
08/30/2020
- 06:48 AM Bug #47201: mds: CDir::_omap_commit(int): Assertion `committed_version == 0' failed.
- Backporting note:
* octopus backport should be done together with #46273
* nautilus backport has no dependency an... - 06:46 AM Backport #46520 (Need More Info): octopus: mds: deleting a large number of files in a directory c...
- setting "Need More Info" to prevent blind automated backport
- 05:45 AM Feature #47161: mds: add dedicated field to inode for fscrypt context
- Jeff Layton wrote:
> The prototype implementation uses an xattr, so I'm aware how that works, but there is more to t...
08/29/2020
- 12:49 AM Bug #47011 (Pending Backport): client: Client::open() pass wrong cap mask to path_walk
- 12:45 AM Bug #47202 (Fix Under Review): qa: Replacing daemon mds.a as rank 0 with standby daemon mds.b" in...
- 12:44 AM Bug #47202 (Resolved): qa: Replacing daemon mds.a as rank 0 with standby daemon mds.b" in cluster...
- ...
- 12:21 AM Backport #46520: octopus: mds: deleting a large number of files in a directory causes the file sy...
- PR introduced a bug: https://tracker.ceph.com/issues/47201
- 12:20 AM Bug #47201 (Resolved): mds: CDir::_omap_commit(int): Assertion `committed_version == 0' failed.
- ...
08/28/2020
- 05:05 PM Bug #47182 (Fix Under Review): mon: deleting a CephFS and its pools causes MONs to crash
- 04:48 AM Bug #47182 (Resolved): mon: deleting a CephFS and its pools causes MONs to crash
- This is a clone of #41228. The bug is back in Octopus:
https://tracker.ceph.com/issues/41228#note-11
and in Nau... - 02:42 PM Backport #47200 (Rejected): octopus: scheduled cephfs snapshots (via ceph manager)
- https://github.com/ceph/ceph/pull/37142
- 02:39 PM Bug #46278 (Resolved): mds: Subvolume snapshot directory does not save attribute "ceph.quota.max_...
- While running with --resolve-parent, the script "backport-create-issue" noticed that all backports of this issue are ...
- 01:04 PM Backport #47058 (Resolved): nautilus: mgr/volumes: Clone operation uses source subvolume root dir...
- 01:04 PM Backport #47058: nautilus: mgr/volumes: Clone operation uses source subvolume root directory mode...
- Nathan Cutler wrote:
> https://github.com/ceph/ceph/pull/36744
Merged as part of PR https://github.com/ceph/ceph/... - 10:38 AM Feature #47161: mds: add dedicated field to inode for fscrypt context
- The prototype implementation uses an xattr, so I'm aware how that works, but there is more to this than just setting ...
- 08:26 AM Feature #47161: mds: add dedicated field to inode for fscrypt context
- xattr is the most suitable place, because client can create file and set xattr at the same time. This is similar to s...
- 08:01 AM Bug #21777: src/mds/MDCache.cc: 4332: FAILED assert(mds->is_rejoin())
- mds.node188-2(rank 2) receive MMDSCacheRejoin message from mds.node185-0(rank 1), but mds.node188-2 is in resolve sta...
- 07:42 AM Bug #41072 (Pending Backport): scheduled cephfs snapshots (via ceph manager)
- 01:30 AM Feature #47168: client: support getting ceph.dir.rsnaps vxattr
- v2: https://patchwork.kernel.org/patch/11742015/
08/27/2020
- 09:01 PM Backport #47178 (In Progress): nautilus: qa: after the cephfs qa test case quit the mountpoints s...
- 08:57 PM Backport #47178 (Resolved): nautilus: qa: after the cephfs qa test case quit the mountpoints stil...
- https://github.com/ceph/ceph/pull/36863
- 08:56 PM Bug #44408 (Pending Backport): qa: after the cephfs qa test case quit the mountpoints still exist
- 08:43 PM Backport #47157 (Resolved): nautilus: mgr/volumes: Mark subvolumes with ceph.dir.subvolume vxattr...
- 07:32 PM Backport #46796 (Resolved): nautilus: mds: Subvolume snapshot directory does not save attribute "...
- 03:46 PM Bug #47172 (Resolved): mgr/nfs: Add support for RGW export
- Current interface for CephFS:
https://docs.ceph.com/en/latest/cephfs/fs-nfs-exports/
The "ceph nfs cluster crea... - 03:42 PM Feature #45746: mgr/nfs: Add interface to update export
- Add an option to output export config in json format. Then user can use this json file to modify the existing export ...
- 03:09 PM Bug #45745 (Rejected): mgr/nfs: Move enable pool to cephadm
- Because we need to create/enable the nfs-ganesha pool for all orchestrator backends, I think this should stay in the ...
- 01:51 PM Feature #47102 (Fix Under Review): mds: add perf counter for cap messages
- 10:56 AM Backport #47096: nautilus: mds: provide altrenatives to increase the total cephfs subvolume snaps...
- This update was made using the script "backport-resolve-issue".
backport PR https://github.com/ceph/ceph/pull/36804
m... - 09:50 AM Feature #47168 (Resolved): client: support getting ceph.dir.rsnaps vxattr
- https://patchwork.kernel.org/patch/11740285/
08/26/2020
- 11:35 PM Backport #47096 (Resolved): nautilus: mds: provide altrenatives to increase the total cephfs subv...
- 02:38 AM Backport #47096 (In Progress): nautilus: mds: provide altrenatives to increase the total cephfs s...
- 09:23 PM Cleanup #47160: qa/tasks/cephfs: Break up test_volumes.py
- To be clear: this is as simple as breaking the volumes tests into separate classes in teh same file. Then the yaml fr...
- 07:16 PM Cleanup #47160 (Resolved): qa/tasks/cephfs: Break up test_volumes.py
- The test_volumes has become unwieldy with growing number of non-trivial tests to test growing set of features in mgr/...
- 07:49 PM Feature #47162 (Resolved): mds: handle encrypted filenames in the MDS for fscrypt
- Once you turn a filename into encrypted text then it can have non-legal and non-printable embedded characters. To mak...
- 07:25 PM Feature #47161 (Rejected): mds: add dedicated field to inode for fscrypt context
- fscrypt requires that each encrypted inode contain an encryption context:
https://www.kernel.org/doc/html/late... - 06:44 PM Backport #47157 (In Progress): nautilus: mgr/volumes: Mark subvolumes with ceph.dir.subvolume vxa...
- 05:48 PM Backport #47157 (Resolved): nautilus: mgr/volumes: Mark subvolumes with ceph.dir.subvolume vxattr...
- https://github.com/ceph/ceph/pull/36833
- 05:48 PM Backport #47158 (Resolved): octopus: mgr/volumes: Mark subvolumes with ceph.dir.subvolume vxattr,...
- https://github.com/ceph/ceph/pull/38612
- 05:47 PM Bug #47154 (Pending Backport): mgr/volumes: Mark subvolumes with ceph.dir.subvolume vxattr, to im...
- 05:30 PM Bug #47154 (Resolved): mgr/volumes: Mark subvolumes with ceph.dir.subvolume vxattr, to improve sn...
- Fix for tracker https://tracker.ceph.com/issues/46074 introduces the vxattr ceph.dir.subvolume that can be use to mar...
- 03:31 PM Backport #47152 (In Progress): nautilus: pybind/mgr/volumes: add debugging for global lock
- 03:28 PM Backport #47152 (Resolved): nautilus: pybind/mgr/volumes: add debugging for global lock
- https://github.com/ceph/ceph/pull/36828
- 03:28 PM Backport #47151 (Resolved): octopus: pybind/mgr/volumes: add debugging for global lock
- https://github.com/ceph/ceph/pull/37366
- 03:28 PM Fix #47149 (Pending Backport): pybind/mgr/volumes: add debugging for global lock
- 02:10 PM Fix #47149 (Resolved): pybind/mgr/volumes: add debugging for global lock
- To help diagnose deadlocks we believe to be happening.
- 12:02 PM Bug #47140 (Duplicate): mgr/volumes: unresponsive Client::abort_conn() when cleaning stale libcep...
- ACK. Thx
- 11:51 AM Bug #47140: mgr/volumes: unresponsive Client::abort_conn() when cleaning stale libcephfs handle
- @Venky looks like a duplicate of https://tracker.ceph.com/issues/46882
From the logs further down,
Job ID: 537667... - 07:25 AM Bug #47140: mgr/volumes: unresponsive Client::abort_conn() when cleaning stale libcephfs handle
- https://pulpito.ceph.com/vshankar-2020-08-26_05:34:12-fs-wip-pdonnell-testing-20200826.032941-distro-basic-smithi/537...
- 07:20 AM Bug #47140 (Duplicate): mgr/volumes: unresponsive Client::abort_conn() when cleaning stale libcep...
- Libcephfs connection pool in mgr (mgr_util) identifies stale filesystem handles and cleans them up by calling abort_c...
- 09:32 AM Feature #47148 (In Progress): mds: get rid of the mds_lock when storing the inode backtrace to me...
- 09:32 AM Feature #47148 (Resolved): mds: get rid of the mds_lock when storing the inode backtrace to meta ...
- The objector->mutate() may take a long time to finish. We can get rid of the mds_lock when doing this.
- 09:30 AM Backport #47081 (In Progress): nautilus: mds: decoding of enum types on big-endian systems broken
- 09:26 AM Backport #47080 (In Progress): octopus: mds: decoding of enum types on big-endian systems broken
- 09:20 AM Feature #20 (Resolved): client: recover from a killed session (w/ blacklist)
- While running with --resolve-parent, the script "backport-create-issue" noticed that all backports of this issue are ...
- 09:17 AM Bug #44276 (Resolved): pybind/mgr/volumes: cleanup stale connection hang
- While running with --resolve-parent, the script "backport-create-issue" noticed that all backports of this issue are ...
- 09:16 AM Feature #45371 (Resolved): mgr/volumes: `protect` and `clone` operation in a single transaction
- While running with --resolve-parent, the script "backport-create-issue" noticed that all backports of this issue are ...
- 09:14 AM Backport #47147 (Resolved): octopus: pybind/mgr/nfs: Test mounting of exports created with nfs ex...
- https://github.com/ceph/ceph/pull/37365
- 09:05 AM Backport #46957 (Resolved): octopus: pybind/mgr/nfs: add interface for adding user defined config...
- This update was made using the script "backport-resolve-issue".
backport PR https://github.com/ceph/ceph/pull/36635
m... - 09:05 AM Backport #46795 (Resolved): octopus: mds: Subvolume snapshot directory does not save attribute "c...
- This update was made using the script "backport-resolve-issue".
backport PR https://github.com/ceph/ceph/pull/36403
m... - 09:05 AM Backport #46591 (Resolved): octopus: ceph-fuse: ceph-fuse process is terminated by the logratote ...
- This update was made using the script "backport-resolve-issue".
backport PR https://github.com/ceph/ceph/pull/36195
m... - 09:04 AM Backport #46528 (Resolved): octopus: mgr/volumes: `protect` and `clone` operation in a single tra...
- This update was made using the script "backport-resolve-issue".
backport PR https://github.com/ceph/ceph/pull/36126
m... - 09:04 AM Backport #46402 (Resolved): octopus: client: recover from a killed session (w/ blacklist)
- This update was made using the script "backport-resolve-issue".
backport PR https://github.com/ceph/ceph/pull/35962
m... - 09:04 AM Backport #46389 (Resolved): octopus: pybind/mgr/volumes: cleanup stale connection hang
- This update was made using the script "backport-resolve-issue".
backport PR https://github.com/ceph/ceph/pull/35962
m... - 12:42 AM Backport #47059 (In Progress): octopus: mgr/volumes: Clone operation uses source subvolume root d...
- 12:35 AM Backport #46820 (In Progress): octopus: pybind/mgr/volumes: Add the ability to keep snapshots of ...
08/25/2020
- 10:10 PM Backport #46957: octopus: pybind/mgr/nfs: add interface for adding user defined configuration
- Nathan Cutler wrote:
> https://github.com/ceph/ceph/pull/36635
merged - 10:09 PM Backport #46795: octopus: mds: Subvolume snapshot directory does not save attribute "ceph.quota.m...
- Nathan Cutler wrote:
> https://github.com/ceph/ceph/pull/36403
merged - 10:08 PM Backport #46591: octopus: ceph-fuse: ceph-fuse process is terminated by the logratote task and wh...
- Nathan Cutler wrote:
> https://github.com/ceph/ceph/pull/36195
merged - 10:08 PM Backport #46528: octopus: mgr/volumes: `protect` and `clone` operation in a single transaction
- Nathan Cutler wrote:
> https://github.com/ceph/ceph/pull/36126
merged - 10:07 PM Backport #46402: octopus: client: recover from a killed session (w/ blacklist)
- Nathan Cutler wrote:
> https://github.com/ceph/ceph/pull/35962
merged - 10:07 PM Backport #46389: octopus: pybind/mgr/volumes: cleanup stale connection hang
- Nathan Cutler wrote:
> https://github.com/ceph/ceph/pull/35962
merged - 04:37 AM Bug #47125 (Fix Under Review): mds: fix possible crash when the MDS is stopping
- 04:32 AM Bug #47125 (In Progress): mds: fix possible crash when the MDS is stopping
- Before I have hit one crash without any useful logs months ago, just be possible caused by this.
- 04:30 AM Bug #47125 (Resolved): mds: fix possible crash when the MDS is stopping
- While the MDS daemon is stopping and if it call journaler->flush(), it may be crash dues to the onsafe parameter in J...
08/24/2020
- 08:20 PM Feature #46989 (Pending Backport): pybind/mgr/nfs: Test mounting of exports created with nfs expo...
- 07:37 PM Bug #47051 (Duplicate): fs/upgrade/volume_client: Command failed with status 124: 'sudo adjust-ul...
- 07:34 PM Bug #47051: fs/upgrade/volume_client: Command failed with status 124: 'sudo adjust-ulimits ceph-c...
- Older log with this occurrence: https://pulpito.ceph.com/pdonnell-2020-08-08_02:16:26-fs-wip-pdonnell-testing-2020080...
- 07:25 PM Bug #47051: fs/upgrade/volume_client: Command failed with status 124: 'sudo adjust-ulimits ceph-c...
- I thought this might be fixed by https://github.com/ceph/ceph/pull/36499 after earlier discussions with Neha but it s...
- 06:36 PM Bug #47015: mds: decoding of enum types on big-endian systems broken
- Ulrich Weigand wrote:
> Thanks for creating the backport requests!
>
> Would it make sense to also include this c... - 11:20 AM Bug #47015: mds: decoding of enum types on big-endian systems broken
- Thanks for creating the backport requests!
Would it make sense to also include this commit for backporting:
https... - 06:36 PM Backport #47081: nautilus: mds: decoding of enum types on big-endian systems broken
- Note to backporters: please include https://github.com/ceph/ceph/pull/35920
- 06:36 PM Backport #47080: octopus: mds: decoding of enum types on big-endian systems broken
- Note to backporters: please include https://github.com/ceph/ceph/pull/35920
- 01:51 PM Feature #47034: mds: readdir for snapshot diff
- cephfs-mirror could use this to read the changes to entire subtrees (not just a single directory) given a snapshot.
- 01:50 PM Bug #47009 (Resolved): TestNFS.test_cluster_set_reset_user_config: command failed with status 32:...
- backport PR: https://github.com/ceph/ceph/pull/36748
- 10:41 AM Bug #47009 (Pending Backport): TestNFS.test_cluster_set_reset_user_config: command failed with st...
- 01:48 PM Feature #47102: mds: add perf counter for cap messages
- separate perf counter for revokes and release
- 01:30 AM Feature #47102 (Resolved): mds: add perf counter for cap messages
- 01:47 PM Feature #45747 (Resolved): pybind/mgr/nfs: add interface for adding user defined configuration
- backport pr: https://github.com/ceph/ceph/pull/36748
Note that this is an exception. I don't plan to include volum... - 01:47 PM Bug #47006 (Triaged): mon: required client features adding/removing
Also available in: Atom