Ceph » rbd

12/27/2012

04:52 PM Bug #3688 (Won't Fix): rbd allows image of size 0 to be created

ceph version : 0.55.1-360-g6356739 (635673928a6b4dae6d4712cacad81cbac6412dc3)
rbd allows images created with zero ... Tamilarasi muthamizhan

12:32 PM Bug #3427: krbd: unmap does not remove block device properly

I am going to assume that the racing open is the cause of
the problem reported by Nikola Kotur.
To fix it, I will... Alex Elder

12:17 PM Bug #3427: krbd: unmap does not remove block device properly

> For RBD, wasn't the use_count something we just added? Would it cover this situation?
No.
The first warning i... Alex Elder

08:53 AM Bug #3427: krbd: unmap does not remove block device properly

For cephfs, the vfs normally handles that.
For RBD, wasn't the use_count something we *just* added? Would it cove... Sage Weil

08:37 AM Bug #3427: krbd: unmap does not remove block device properly

I also note, having taken a little closer look at Nikola Kotur's
kernel log that both an open and a close appear to ... Alex Elder

08:31 AM Bug #3427: krbd: unmap does not remove block device properly

It looks to me like the osd client code has nothing in place
to protect itself from one of its users (ceph client, m... Alex Elder

11:28 AM Bug #2689: qemu iozone test hangs

Testing again since some possible causes were fixed. Josh Durgin

10:54 AM Bug #3685 (Closed): xfs test 186 fails in the nightlies

logs: ubuntu@teuthology:/a/teuthology-2012-12-26_19:00:03-regression-next-testing-basic/28039
... Tamilarasi muthamizhan

12/26/2012

02:38 PM Bug #3427: krbd: unmap does not remove block device properly

I haven't spent time on this in almost a month so wanted to just
provide an update. We have been looking at and try... Alex Elder

12/24/2012

02:56 PM Feature #3580 (Resolved): rbd import from stdin could try harder to sparsify images

Sage Weil

09:22 AM Bug #3654 (Fix Under Review): libvirt: colons in ipv6 monitor addresses are not escaped when sent...

Sage Weil

08:45 AM Fix #3665: librbd: deadlock during flatten

the problem is that we are holding the snap_lock and then waiting for io. but we mostly use snap_lock as a tight inne... Sage Weil

12/21/2012

10:03 AM Fix #3665 (Resolved): librbd: deadlock during flatten

Ran into this trying to reproduce #3631.
The test_librbd_fsx process is still running on plana34 for debugging.
... Josh Durgin

08:32 AM Bug #3664 (Resolved): osdc/ObjectCacher.cc: 517: FAILED assert(!i->size())

... Sage Weil

12/20/2012

03:49 PM Bug #3524: test_librbd_fsx: crash after flatten

Sam saw this come up again in: ubuntu@teutholog:/a/sam-ooo3/19022
It's a different cause of the same symptom. In t... Josh Durgin

12/19/2012

04:38 PM Bug #3654 (Resolved): libvirt: colons in ipv6 monitor addresses are not escaped when sent to qemu

Given xml like:... Josh Durgin

11:47 AM Bug #3611 (Resolved): rbd.py: segfault with many snapshots

This was caused by c3107009f66bc06b5e14c465142e14120f9a4412. Reverting it fixes the problem. There is a corrected imp... Josh Durgin

12/18/2012

01:26 PM Bug #3642 (Resolved): librbd: watch is sent with assert version, which fails on resends

Instead of using an assert version op, establish the watch before reading the header. This hasn't actually caused any... Josh Durgin

08:54 AM Bug #3611: rbd.py: segfault with many snapshots

This survived overnight testing (with the python librbd tests) with 56 passes. Josh Durgin

12/17/2012

10:59 PM Bug #3611 (Fix Under Review): rbd.py: segfault with many snapshots

wip-3611 contains a respin of the bad commit. It's passing test_stress_watch with failure injection and the python te... Josh Durgin

11:09 AM Bug #3611: rbd.py: segfault with many snapshots

also, ubuntu@teuthology:/a/teuthology-2012-12-15_19:00:04-regression-next-testing-basic/16289 Tamilarasi muthamizhan

11:08 AM Bug #3611: rbd.py: segfault with many snapshots

recent log: ubuntu@teuthology:/a/teuthology-2012-12-15_19:00:04-regression-next-testing-basic/16281 Tamilarasi muthamizhan

10:53 PM Bug #3585: Image import via QEMU-IMG results in a corrupt rbd

Thanks for the logs. All the differences there are zeroes where actual data should be, but the librbd debug log shows... Josh Durgin

12:40 PM Feature #3635: rbd cli: call "udevadm settle" after use of add/remove kernel interface

Trivial change. Biggest decision is which libc routine to use to spawn the command... Dan Mick

11:42 AM Feature #3635 (Resolved): rbd cli: call "udevadm settle" after use of add/remove kernel interface

The rbd command line interface creates mappings by sending
output to the /sys/bus/rbd/add file system entry, and rem... Alex Elder

10:53 AM Bug #3600: rbd: assert in objectcacher destructor after flatten

Tried to reproduce this behavior to no avail.
There are operations on the test that do hang for a long time, but a... Joao Eduardo Luis

12/16/2012

12:44 PM Bug #2872 (Resolved): RBD resize command allows image size -1

Sage Weil

11:02 AM Bug #2689: qemu iozone test hangs

let's retest this with all of the recent caching fixes? Sage Weil

12/15/2012

09:45 AM Fix #3588 (In Progress): rbd.py's clone should take stripe parms, call rbd_clone2

Sage Weil

09:45 AM Feature #2601 (Resolved): rbd: Show image size with an "ls"

Sage Weil

09:44 AM Feature #2634 (Resolved): teuthology: add networking to qemu task

Sage Weil

09:43 AM Bug #3619 (In Progress): librbd: read_iterate sparse behavior broken

Sage Weil

09:42 AM Bug #2689: qemu iozone test hangs

Sage Weil

12/14/2012

07:15 PM Bug #3611: rbd.py: segfault with many snapshots

It looks like the op in the objecter has been corrupted, similar to #3613. In this case, op->objver ends up pointing ... Josh Durgin

04:07 PM Bug #3589 (Resolved): rbd.py should check for method existence before calling new methods

Josh Durgin

07:28 AM Bug #2410: hung xfstest #68

This appears to be an XFS problem, where the file
system is having trouble getting space in its
journal. I inquire... Alex Elder

07:13 AM Bug #2608: rbd: hung xfstest 270

We should re-evaluate this with XFS found in newer kernels.
Maybe this should just be closed and re-opened (or open
... Alex Elder

06:08 AM Feature #3418: krbd: write path (layering)

I did a little research on this before starting on the write
path. This work will require the kernel rbd client, th... Alex Elder

06:00 AM Feature #3417: krbd: read path (layering)

Work on this really started in November 2012.
In October there were a number of cleanup tasks we agreed
should ge... Alex Elder

12/13/2012

11:48 PM Bug #3585: Image import via QEMU-IMG results in a corrupt rbd

Attached files as requested.
Compare was stopped early to save on file size. Matt Anderson

05:03 PM Bug #3611: rbd.py: segfault with many snapshots

Finally got a backtrace. It seems something is overwriting a Mutex::Locker on the stack:... Josh Durgin

01:31 PM Bug #3619 (Resolved): librbd: read_iterate sparse behavior broken

Instead of getting a NULL for a hole, we get a zeroed buffer.
Reported on the ML Sage Weil

12/12/2012

04:13 PM Bug #3611: rbd.py: segfault with many snapshots

Downgrading priority since this isn't an actual bug. Josh Durgin

03:41 PM Bug #3611: rbd.py: segfault with many snapshots

Without lockdep, I could not reproduce a crash.
Running with lockdep enabled results in this backtrace:... Josh Durgin

11:52 AM Bug #3611 (Resolved): rbd.py: segfault with many snapshots

From the nightly python api tests, test_many_snaps failed with a segfault in all runs.
Logs are in:... Josh Durgin

12/11/2012

05:04 PM Bug #3413: rbd bench-write fails with assert when rbd caching turned on

Josh Durgin

04:58 PM Bug #3589 (Fix Under Review): rbd.py should check for method existence before calling new methods

wip-rbdpy-compat Josh Durgin

04:53 PM Feature #2568 (Resolved): qa: run xfstests on qemu+rbd

Josh Durgin

03:34 PM Bug #3600: rbd: assert in objectcacher destructor after flatten

The hang on selfmanaged_snap_create seems like a monitor issue; bouncing the monitor, or hanging out in gdb long enou... Dan Mick

12/10/2012

07:57 PM Bug #3600: rbd: assert in objectcacher destructor after flatten

Hm #2. Both runs are deadlocked on the mutex IoCtxImpl::selfmanaged_snap_create::mylock. Will look for owner next. Dan Mick

07:23 PM Bug #3600: rbd: assert in objectcacher destructor after flatten

Looking at the code, I don't really see a clean "stop caching" mechanism. While I look, hacked fsx to do only write/... Dan Mick

11:40 AM Bug #3600 (Duplicate): rbd: assert in objectcacher destructor after flatten

From ubuntu@teuthology:/a/teuthology-2012-12-09_19:00:03-regression-master-testing-gcov/10977:... Josh Durgin

11:40 AM Bug #3524 (Resolved): test_librbd_fsx: crash after flatten

That's a different bug. Created #3600 to track it. Josh Durgin

11:36 AM Bug #3524 (In Progress): test_librbd_fsx: crash after flatten

recent log :
ubuntu@teuthology:/a/teuthology-2012-12-09_19:00:03-regression-master-testing-gcov/10977 Tamilarasi muthamizhan

11:11 AM Bug #3585: Image import via QEMU-IMG results in a corrupt rbd

Since the size isn't an issue, it'd be great if you could:

1) generate a log of qemu-img convert with 'rbd cache ... Josh Durgin

12/07/2012

06:55 PM Bug #3585: Image import via QEMU-IMG results in a corrupt rbd

I ran ... Matt Anderson

12:44 PM Bug #3585: Image import via QEMU-IMG results in a corrupt rbd

It looks like qemu-img info is also reporting the size after using integer division and multiplication by 512, so it ... Josh Durgin

12:36 PM Bug #3585: Image import via QEMU-IMG results in a corrupt rbd

Just reproduced a bad size (504 bytes less) when using qemu-img 1.1.2 to convert a 1024119288 byte file. It seems to ... Josh Durgin

01:18 AM Bug #3585: Image import via QEMU-IMG results in a corrupt rbd

All of the imported images are showing the exact same size. TS7 is a qemu-img import and TS6 is a rbd import. Using q... Matt Anderson

01:05 AM Bug #3585 (In Progress): Image import via QEMU-IMG results in a corrupt rbd

As a workaround you can use 'rbd import file pool/image' on a raw file. Does the corrupted image show the correct siz... Josh Durgin

12:21 AM Bug #3585 (Closed): Image import via QEMU-IMG results in a corrupt rbd

This is a follow on from the mailing list topic VM Corruption on "0.54 when 'client cache = false'". After upgrading ... Matt Anderson

05:25 PM Bug #3589 (In Progress): rbd.py should check for method existence before calling new methods

Josh Durgin

04:02 PM Bug #3589 (Resolved): rbd.py should check for method existence before calling new methods

If rbd.py is upgraded past librbd, it should not fail because e.g. rbd_create3 does not exist in the old version. Josh Durgin

03:59 PM Fix #3588: rbd.py's clone should take stripe parms, call rbd_clone2

In danger of forgetting about this unless I write it down somewhere:... Dan Mick

03:57 PM Fix #3588 (Resolved): rbd.py's clone should take stripe parms, call rbd_clone2

Dan Mick

03:58 PM Bug #2677 (Resolved): librbd: create does not clean up well

Josh Durgin

11:06 AM Bug #2677 (Fix Under Review): librbd: create does not clean up well

Josh Durgin

12/06/2012

11:28 AM Bug #3562 (Can't reproduce): incorrect progress shown for rbd resize image

Neither Josh nor I have a good theory for how this even could have happened, and I can't reproduce it, so...reopen if... Dan Mick

12/05/2012

06:10 PM Feature #3580 (Resolved): rbd import from stdin could try harder to sparsify images

rbd import from stdin currently seeks over full blocks of zeros that it gets from
stdin; it could try harder to accu... Dan Mick

05:56 PM Bug #3556 (Resolved): rbd rm command fails to remove the new format rbd image completely

This was fixed by reverting the tmap_rm-of-a-nonexistent-key behavior (the reverting commit is 5aecf2c092edcabc3b757a... Dan Mick

05:52 PM Bug #3133 (Duplicate): perfcounters: sigfpe when rbd caching is enabled

#3521 is the same problem Josh Durgin

05:47 PM Bug #3524 (Resolved): test_librbd_fsx: crash after flatten

commit:41e16a3b40efb80a5ed7a5587438569ca86c85a3
Dan Mick

12/03/2012

09:22 PM Bug #3524 (In Progress): test_librbd_fsx: crash after flatten

Dan Mick

06:33 PM Bug #3524: test_librbd_fsx: crash after flatten

I think I found the smoking gun. Two completions are racing to update partial
in add_partial_sparse_result(); there... Dan Mick

03:54 PM Bug #3566 (Resolved): log max new = 1 can cause hang on process exit

Investigating other things entirely, I had set log max err = 1 in [global]; I started
observing occasional hangs, an... Dan Mick

11/30/2012

11:21 PM Bug #3549: rbd: info should report striping

commit:7794de61b5f4b5c2e20c99e26c8578361d9b0247 Dan Mick

11:21 PM Bug #3549 (Resolved): rbd: info should report striping

Dan Mick

10:30 PM Bug #3549: rbd: info should report striping

Not yet resolved; feature_str() should return striping when enabled. Dan Mick

09:08 PM Bug #3549 (Resolved): rbd: info should report striping

Sage Weil

09:08 PM Bug #3503 (Resolved): rbd: import "-" no longer works

Sage Weil

09:08 PM Bug #3547 (Resolved): rbd: suites/iozone.sh failure

Sage Weil

09:05 PM Feature #3380 (Resolved): rbd: export and import functions should support stdout

Sage Weil

08:36 PM Feature #1908 (Resolved): rbd: test map/unmap more extensively

Sage Weil

08:36 PM Feature #1908: rbd: test map/unmap more extensively

Sage Weil

04:03 PM Bug #3562 (Can't reproduce): incorrect progress shown for rbd resize image

Logs:ubuntu@teuthology:/a/teuthology-2012-11-30_07:00:03-regression-testing-master-basic/6540... Tamilarasi muthamizhan

11/29/2012

11:03 AM Bug #3556 (Resolved): rbd rm command fails to remove the new format rbd image completely

when a rbd image is created using new format, removing the header followed by removing the rbd image using "rbd rm" d... Tamilarasi muthamizhan

11/28/2012

03:24 PM Bug #3524: test_librbd_fsx: crash after flatten

I've reproduced with logs, and am examining the ops. Working theory is that the flatten happened between the first p... Dan Mick

03:22 PM Bug #3478 (Closed): rbd: copy destination pool is assumed to be source pool

commit:81d383073865d221bb825a8990ae9b34bd491bef (next)
commit:3b0c36052806a4574ede6a9f0594dc83f7997bcc (master) Dan Mick

01:55 PM Bug #3549 (Resolved): rbd: info should report striping

feature_str() should have striping added for rbd info. Should rbd ls -l report striping
parms too?.. Dan Mick

12:57 PM Bug #3547 (Resolved): rbd: suites/iozone.sh failure

... Alex Elder

11:45 AM Bug #3427: krbd: unmap does not remove block device properly

Indeed, that's where this is from. So, there are running, or lingering, requests on the OSD. Shouldn't it try to kick... Nikola Kotur

08:30 AM Bug #3427 (In Progress): krbd: unmap does not remove block device properly

... Alex Elder

02:46 AM Bug #3427: krbd: unmap does not remove block device properly

The same bug hit me in kernel 3.6.7.
Ceph version: 0.54
Kernel log:... Nikola Kotur