Project

General

Profile

Activity

From 01/10/2013 to 02/08/2013

02/08/2013

08:36 AM Cleanup #4053: ceph: cleanup ceph page vector functions
Apparently for cleanup there is no "need review" so I'm
marking this "Feedback". I've posted a series of patches
t...
Alex Elder
08:30 AM Cleanup #4053 (Resolved): ceph: cleanup ceph page vector functions
This is just documenting some cleanup activity I've done
that I'm about to post for review.
- delete bogus (re)decl...
Alex Elder
08:21 AM Subtask #4007 (Fix Under Review): libceph: support STAT osd operation
A patch implementing this has been posted to the
ceph-devel mailing list for review.
[PATCH] libceph: allow STAT ...
Alex Elder

02/07/2013

05:08 PM Bug #4047 (Resolved): removing a non-existing rbd image logs error in osd logs
when removing a non-existing rbd image floods osd logs even when the debug is turned off. This can be avoided.
ubu...
Tamilarasi muthamizhan
04:22 PM Bug #4045 (Resolved): snap unprotect on a snapshot that is already unprotected throws inappropria...
ceph version 0.56.2-7-gc3468f7 (c3468f76a5e68a6426f03e508d8ecf26950fca2a)
Trying to unprotect a snapshot, that is ...
Tamilarasi muthamizhan
01:49 PM Bug #4003 (Resolved): rbd: EBUSY errors from rbd unmap
closing this. phew! Sage Weil

02/06/2013

11:49 AM Subtask #4007: libceph: support STAT osd operation
This has turned out to be simple change. It was needed in
rbd as well, and I'll just add support to both under this...
Alex Elder
09:19 AM Subtask #4007: libceph: support STAT osd operation
It wasn't really possible to know this up front but
it looks like this is trivial. I've basically
completed it but...
Alex Elder
10:56 AM Bug #3697 (In Progress): rbd copy.sh test failing in nightly
Tamilarasi muthamizhan
08:56 AM Bug #3958: rbd fsx fails with EBUSY
this is causing several failures on master runs.. something has changed.
latest:
ubuntu@teuthology:/a/sage-2013-...
Sage Weil
08:31 AM Bug #4003: rbd: EBUSY errors from rbd unmap
The testing I've been doing now has shown no problems
now that teuthology has been updated.
The two other issues ...
Alex Elder
06:16 AM Bug #4003: rbd: EBUSY errors from rbd unmap
Seems to have done the trick! The kernel_untar_build.sh
task just finished for me without error, and it failed
rel...
Alex Elder
05:06 AM Bug #4003: rbd: EBUSY errors from rbd unmap
That sounds promising, I hope it works!
This was actually the last thing I was looking at last
night while waitin...
Alex Elder
07:20 AM Bug #4033 (Fix Under Review): krbd: add barriers near done flag operations
A fix for this has been posted for review.
[PATCH] rbd: add barriers near done flag operations
Alex Elder
06:15 AM Bug #4033 (Resolved): krbd: add barriers near done flag operations
I fixed this problem while investigating the rbd hangs
in http://tracker.ceph.com/issues/4003.
Somehow, I missed ...
Alex Elder

02/05/2013

11:43 PM Bug #4003: rbd: EBUSY errors from rbd unmap
This was backing up qa stuff because the rbd.py qa task wasn't unmounting during cleanup. That bit is now fixed. I ... Sage Weil
10:54 PM Bug #4003: rbd: EBUSY errors from rbd unmap
new theory:
the reason umount hangs is because nuke is killing the client and osds at the same time. the umount i...
Sage Weil
10:41 PM Bug #4003: rbd: EBUSY errors from rbd unmap
I found that unmount was hanging too. I think somehow the
completion of the I/O is not getting propagated up when
...
Alex Elder
10:33 PM Bug #4003: rbd: EBUSY errors from rbd unmap
aha:... Sage Weil
10:15 PM Bug #4003: rbd: EBUSY errors from rbd unmap
btw i am able to reproduce the EBUSY with just... Sage Weil
08:28 PM Bug #4003: rbd: EBUSY errors from rbd unmap
I've added some instrumentation and find that the rbd
client is not dropping its watch at the end of the
kernel_unt...
Alex Elder
12:51 PM Bug #4003: rbd: EBUSY errors from rbd unmap
The interrupt issue has been fixed, but the other issue
(rbd device can't be unmapped because EBUSY) remains.
I h...
Alex Elder
11:35 AM Bug #4003: rbd: EBUSY errors from rbd unmap
I ran the kernel_untar_build.sh workunit using the
ceph "master" branch and the ceph-client "testing"
branch and go...
Alex Elder
11:13 AM Bug #4003: rbd: EBUSY errors from rbd unmap
I think I found *a* problem, possibly not *the* problem.
This commit:
bc7a62ee5 rbd: prevent open for image ...
Alex Elder
11:04 AM Bug #4003: rbd: EBUSY errors from rbd unmap
I am able to reproduce this problem by running
the kernel_untar_build.sh workunit.
I ran the test using the ceph ...
Alex Elder
08:53 AM Bug #4003: rbd: EBUSY errors from rbd unmap
> Alex, unless there is another high priority regression, can you
> look at this first?
Yes I will.
Alex Elder
08:52 AM Bug #4003: rbd: EBUSY errors from rbd unmap
Sam Lang wrote:
> I was able to verify that this happens with an older version of teuthology, one without the change...
Sage Weil
08:41 AM Bug #4003: rbd: EBUSY errors from rbd unmap
I was able to verify that this happens with an older version of teuthology, one without the changes I've made recentl... Sam Lang
05:18 AM Bug #4003: rbd: EBUSY errors from rbd unmap
I had the impression this might be a problem that
is holding up completion of the nightly test suite.
But I'm not...
Alex Elder
01:47 PM Feature #4021 (Resolved): rbd: openstack: add ability to copy volume to image for rbd
Ian Colle
01:46 PM Subtask #4020 (Resolved): rbd: openstack: simplify volume booting with new api: make image boot b...
Ian Colle
01:44 PM Subtask #4019 (Resolved): rbd: openstack: simplify volume booting with new api: add boot option t...
Ian Colle
01:44 PM Subtask #4018 (Resolved): rbd: openstack: simplify volume booting with new api: modify boot panel...
Ian Colle
01:42 PM Feature #4017 (Resolved): rbd: openstack: simplify volume booting with new api
Ian Colle
01:42 PM Feature #4013 (In Progress): rbd: openstack: extend nova boot api to support going from image to ...
Ian Colle
01:24 PM Feature #4013 (Resolved): rbd: openstack: extend nova boot api to support going from image to volume
Ian Colle
01:41 PM Subtask #4016 (Resolved): rbd: openstack: extend nova boot api: modify libvirt driver to support ...
Ian Colle
01:40 PM Subtask #4015 (Resolved): rbd: openstack: extend nova boot api: add block_dev_mapping_v2 to nova-...
Ian Colle
01:40 PM Subtask #4014 (Resolved): rbd: openstack: extend nova boot api: add block_dev_mapping_v2 to nova-api
Ian Colle
01:13 PM Bug #4012 (Won't Fix): rbd: image creation behaviour has to be uniform across bobtail and argonau...
rbd allows images to be created with size 0 in bobtail, but it fails in argonaut.
similarly,while in bobtail it do...
Tamilarasi muthamizhan
12:52 PM Bug #4010 (Fix Under Review): krbd: turn off interrupts for open/remove locking
Posted for review.
[PATCH] rbd: turn off interrupts for open/remove locking
Alex Elder
12:49 PM Bug #4010 (Resolved): krbd: turn off interrupts for open/remove locking
This fix is done. The problem was discovered while
investigating http://tracker.ceph.com/issues/4003.
This commi...
Alex Elder
10:12 AM Bug #3697: rbd copy.sh test failing in nightly
recent log : ubuntu@teuthology:/a/teuthology-2013-02-04_20:00:03-regression-bobtail-master-basic/15773 Tamilarasi muthamizhan
07:42 AM Subtask #4007 (Resolved): libceph: support STAT osd operation
In order to do layered writes we need to check whether
an object to be written exists before issuing the write.
Thi...
Alex Elder

02/04/2013

05:53 PM Bug #4003: rbd: EBUSY errors from rbd unmap
I really can't tell who's got a watch on the header
object. It should be getting removed when the object
gets unma...
Alex Elder
05:02 PM Bug #4003: rbd: EBUSY errors from rbd unmap
There is clearly something that is keeping the rbd image
from getting removed. I reproduced this with just running
...
Alex Elder
04:12 PM Bug #4003 (In Progress): rbd: EBUSY errors from rbd unmap
This sounds familiar, but I'm going to look a little
more closely to see if I can learn why it's happening.
Alex Elder
04:03 PM Bug #4003 (Resolved): rbd: EBUSY errors from rbd unmap
From the teuthology kernel untar task on rbd, we get EBUSY trying to unmap. I'm investigating that this isn't someho... Sam Lang

02/01/2013

02:18 PM Subtask #3741 (Resolved): krbd: rework request tracking code
commit 9ac90ea3d8dd6ab82f3665a132ca29e6ada56ad8
Author: Alex Elder <elder@inktank.com>
Date: Thu Nov 22 00:00:08 ...
Alex Elder
02:17 PM Feature #3754 (Closed): krbd: use new request tracking code for notify ack
commit 1c8c3c5c571607a188203142020d80aa58e5e280
Author: Alex Elder <elder@inktank.com>
Date: Fri Nov 30 17:53:04 ...
Alex Elder
02:16 PM Tasks #3755: krbd: use new request tracking code for sync object operations
commit 5d08568324f53368f927cc10927b1b105533c044
Author: Alex Elder <elder@inktank.com>
Date: Thu Jan 17 12:25:27 ...
Alex Elder
01:44 PM Tasks #3755 (Resolved): krbd: use new request tracking code for sync object operations
commit 304819b1a49937753ee01aa7ccf8d66547a0be36
Author: Alex Elder <elder@inktank.com>
Date: Sat Jan 19 00:30:28 ...
Alex Elder
02:11 PM Feature #3877 (Closed): krbd: don't wait for notify ack to complete
commit a8a34efcac7a33e7631fe8bf25530bd4be0417f8
Author: Alex Elder <elder@inktank.com>
Date: Thu Jan 17 12:18:46 ...
Alex Elder
01:38 PM Bug #3940 (Resolved): krbd: decrement obj request count when deleting
commit 150fde1984ec8454c163e4f89a50416cd68edbc4
Author: Alex Elder <elder@inktank.com>
Date: Fri Jan 25 17:08:55 ...
Alex Elder
01:38 PM Bug #3937 (Resolved): krbd: crash in rbd_assert(osd_req == obj_request->osd_req)
commit 8d93192992301f8c3a288c8cf4dc8598ac4b8427
Author: Alex Elder <elder@inktank.com>
Date: Fri Jan 25 17:08:55 ...
Alex Elder
01:37 PM Bug #3427 (Resolved): krbd: unmap does not remove block device properly
commit bc7a62ee52cffc735cb8383b6d26648883f1a01e
Author: Alex Elder <elder@inktank.com>
Date: Mon Jan 14 12:43:31 ...
Alex Elder
12:48 PM Bug #1740 (Resolved): krbd: don't return head data when reading from a non-existent snapshot
This was fixed a while ago. Josh Durgin
11:58 AM Bug #3980 (Won't Fix): rbd image created with size zero on a mixed cluster crashes rbd
creating a rbd image with size 0 is allowed in bobtail but not on argonaut.
on a mixed cluster running argonaut[bu...
Tamilarasi muthamizhan

01/31/2013

08:51 PM Bug #3978 (Resolved): krbd qa: concurrent.sh test leaves something read-only
I don't know what exactly is happening here, but it appears
that after running the "rbd/concurrent.sh" workunit, if
...
Alex Elder
02:21 PM Bug #3975 (Rejected): librbd: xfstests 008 failed inside qemu
This one's not a problem. This test pokes random holes in a
file (or maybe fills random spots). And when done it s...
Alex Elder
02:05 PM Bug #3975 (Rejected): librbd: xfstests 008 failed inside qemu
From xfstests output in ubuntu@teuthology:/a/teuthology-2013-01-29_20:00:04-regression-bobtail-master-basic/7794/remo... Josh Durgin

01/30/2013

02:10 PM Bug #3937: krbd: crash in rbd_assert(osd_req == obj_request->osd_req)
The patch is reviewed and ready to push to the testing
branch, and I will do that in a day or so.
I'm going to le...
Alex Elder
02:08 PM Bug #3940: krbd: decrement obj request count when deleting
Reviewed and ready to push to master. Will do that in a day or so. Alex Elder
02:07 PM Bug #3427: krbd: unmap does not remove block device properly
Reviewed and ready to push to the ceph-client "testing" branch.
I'm going to wait a day or two before pushing this...
Alex Elder
11:31 AM Bug #3964: krbd: 32-bit, kernel 3.2.0 system can't do O_DIRECT writes to mapped rbd image with sn...
...and to answer your other question Alex, there's now a workunit test Sage just added
in c782d2ac531cbb7650968e62f0...
Dan Mick
11:00 AM Bug #3964: krbd: 32-bit, kernel 3.2.0 system can't do O_DIRECT writes to mapped rbd image with sn...
Josh thinks 32-bitness probably doesn't matter, and remembers problems with snapshots that were fixed long ago; I gue... Dan Mick
10:55 AM Bug #3964: krbd: 32-bit, kernel 3.2.0 system can't do O_DIRECT writes to mapped rbd image with sn...
I don't know if Sage tested 32-bit, or if it matters, and no, that script was just a reproduction scenario; as far as... Dan Mick
06:25 AM Bug #3964: krbd: 32-bit, kernel 3.2.0 system can't do O_DIRECT writes to mapped rbd image with sn...
So is this then a request to port whatever it was that
fixed the problem back to 3.2?
If so, how do we prioritize...
Alex Elder
01:10 AM Bug #3964: krbd: 32-bit, kernel 3.2.0 system can't do O_DIRECT writes to mapped rbd image with sn...
added test to suite, commit:c782d2ac531cbb7650968e62f0b24e6136a64359 Sage Weil
12:15 AM Bug #3964: krbd: 32-bit, kernel 3.2.0 system can't do O_DIRECT writes to mapped rbd image with sn...
This works fine on current testing 3.6.0-00210-g8cc17ca Sage Weil
11:16 AM Bug #3961 (Resolved): 32-bit cls_rbd tries cls_log with %d for 64-bit int, segfaults
commit:e253830abac76af03c63239302691f7fac1af381 on next
Dan Mick
09:37 AM Subtask #3741: krbd: rework request tracking code
My testing on this code is nearly complete. However, I'm going
to hold off on pushing this (along with the changes ...
Alex Elder
06:34 AM Subtask #3741: krbd: rework request tracking code
Alex Elder
06:47 AM Bug #3927 (Closed): krbd: I/O errors (ENXIO) during rbd/kernel.sh workunit
It turns out this new behavior is a good thing, we're just
reporting errors now where we apparently did not previous...
Alex Elder
06:47 AM Bug #3745 (Rejected): krbd: individual response errors are ignored
I no longer believe this is a problem. Although there is no
aggregate result value for a collection of osd requests...
Alex Elder
06:35 AM Feature #3877: krbd: don't wait for notify ack to complete
Alex Elder
06:35 AM Tasks #3755: krbd: use new request tracking code for sync object operations
Alex Elder
06:35 AM Feature #3754: krbd: use new request tracking code for notify ack
Alex Elder

01/29/2013

11:40 PM Bug #3964: krbd: 32-bit, kernel 3.2.0 system can't do O_DIRECT writes to mapped rbd image with sn...
... Dan Mick
11:28 PM Bug #3964 (Won't Fix): krbd: 32-bit, kernel 3.2.0 system can't do O_DIRECT writes to mapped rbd i...
fghaas reported, I reproduced on a precise 32-bit system:
create an image, map, writes work fine, even with dd ofl...
Dan Mick
10:46 PM Bug #3961 (Resolved): 32-bit cls_rbd tries cls_log with %d for 64-bit int, segfaults
32-bit system: rbd create i -s 1; rbd rm i causes death of osd in cls_log();
presumably this is because of cls_log(%...
Dan Mick
08:33 PM Bug #3566 (Resolved): log max new = 1 can cause hang on process exit
fixed a few weeks ago, commit:813787af3dbb99e42f481af670c4bb0e254e4432 and a few prior commits Sage Weil
08:26 PM Bug #3958 (Resolved): rbd fsx fails with EBUSY
... Sage Weil
03:11 PM Bug #3952 (Resolved): krbd: no need for object header version
The header object watch operation had a sort of half implemented
use of the version of the object. It apparently is...
Alex Elder
03:08 PM Bug #3946 (Resolved): rbd fsx failing in nightly
Just an extra delete in a code path in flush_set that wasn't exercised before. Fixed by commit:3bc21143552b35698c9916... Josh Durgin
02:44 PM Bug #3946: rbd fsx failing in nightly
Reproducing locally seems to confirm this, since there was a recent change to replace commit_set() with flush_set():
...
Josh Durgin
12:06 PM Bug #3946: rbd fsx failing in nightly
I'm guessing these are related to recent objectcacher changes, since they didn't affect runs without caching. The cor... Josh Durgin
02:48 PM Feature #3949 (Resolved): krbd: create test script that exercises concurrent operations
I just committed the test script to the ceph master branch.
The script is located here: qa/workunits/rbd/concurrent...
Alex Elder
09:16 AM Feature #3949: krbd: create test script that exercises concurrent operations
Well the script is really nice. And I just got a new
crash while running it on a real machine (rather than
my UML ...
Alex Elder
08:22 AM Feature #3949 (Resolved): krbd: create test script that exercises concurrent operations
I suggested doing this in http://tracker.ceph.com/issues/3427.
That issue is about a bug where an image unmapping ca...
Alex Elder
11:10 AM Bug #3950: krbd: new assertion failure running concurrent rbd test
OK, I do have the osd request pointer now. It was available
in register R14. And with a little work I can determin...
Alex Elder
10:35 AM Bug #3950: krbd: new assertion failure running concurrent rbd test
The object being operated on is the rbd header image, in
this case named "image.5X5ZNB.rbd". The object request typ...
Alex Elder
10:06 AM Bug #3950: krbd: new assertion failure running concurrent rbd test
Weird. It looks to me like the object request that's
just completing is already done, meaning we got
a callback fr...
Alex Elder
09:19 AM Bug #3950 (Can't reproduce): krbd: new assertion failure running concurrent rbd test
(I think this is a new issue, I haven't investigated it yet.)
I hit an assertion failure while running my new test...
Alex Elder
10:34 AM Bug #3937: krbd: crash in rbd_assert(osd_req == obj_request->osd_req)
I've opened a new issue that has symptoms similar to this
but not identical:
http://tracker.ceph.com/issues/395...
Alex Elder
09:15 AM Bug #3427: krbd: unmap does not remove block device properly
I have opened this to cover developing that test script
http://tracker.ceph.com/issues/3949
Alex Elder

01/28/2013

09:18 PM Bug #3947 (Resolved): krbd: read zeroing freed bio?
This happened to me once before but I wasn't sure what
I did. Now I think I do know. This is with the new
request...
Alex Elder
05:02 PM Bug #3946 (Resolved): rbd fsx failing in nightly
... Sage Weil
02:11 PM Bug #3427 (Fix Under Review): krbd: unmap does not remove block device properly
I have posted two patches for review, the second of which
should fix this problem. I have not actually reproduced
...
Alex Elder

01/26/2013

12:44 PM Bug #3937 (Fix Under Review): krbd: crash in rbd_assert(osd_req == obj_request->osd_req)
A patch resolving this has been posted for review.
[PATCH 4/4] rbd: don't drop watch requests on completion
Alex Elder
12:43 PM Bug #3940 (Fix Under Review): krbd: decrement obj request count when deleting
A patch resolving this has been posted for review. Alex Elder
08:05 AM Bug #3940 (Resolved): krbd: decrement obj request count when deleting
The obj_request_count value keeps track of how many object requests
are associated with an image request. It is inc...
Alex Elder
07:57 AM Bug #3939 (Duplicate): krbd: circular locking report in sysfs code
I intended to write this up before but don't think I did.
I'm getting a "possible circular locking dependency detect...
Alex Elder

01/25/2013

03:25 PM Bug #3937: krbd: crash in rbd_assert(osd_req == obj_request->osd_req)
OK, with Josh's help I finally managed to reproduce the
problem intentionally to check my fix.
I'm building it no...
Alex Elder
11:11 AM Bug #3937: krbd: crash in rbd_assert(osd_req == obj_request->osd_req)
I have confirmed that every time a request registered to linger
is re-submitted the osd client will call the callbac...
Alex Elder
08:07 AM Bug #3937: krbd: crash in rbd_assert(osd_req == obj_request->osd_req)
I've decoded the osd request that's been provided to
rbd_osd_req_callback(). Its contents look completely
legitima...
Alex Elder
06:54 AM Bug #3937: krbd: crash in rbd_assert(osd_req == obj_request->osd_req)
Adding two things:
- this occurred during test 190 of the third consecutive pass
of xfstests with this in the teuth...
Alex Elder
05:04 AM Bug #3937 (Resolved): krbd: crash in rbd_assert(osd_req == obj_request->osd_req)
Looking at a crash this morning in the new request code due
to this failed assertion in rbd_osd_req_callback():
...
Alex Elder
09:54 AM Bug #3936: rbd: Strange dd speed behaviour (server side issue?)
It's pretty likely that this is a server-side behavior rather than a client-side one. Keep that in mind when reproduc... Greg Farnum
12:00 AM Bug #3936: rbd: Strange dd speed behaviour (server side issue?)
rados -p rbd bench 120 write -t 16
shows about 90-110 MB/sec.
Ivan Kudryavtsev
09:52 AM Bug #3654 (Resolved): libvirt: colons in ipv6 monitor addresses are not escaped when sent to qemu
Upstream commit c1509ab47edf61e9f20d11922526b9fca518d238 Josh Durgin
09:34 AM Bug #3927: krbd: I/O errors (ENXIO) during rbd/kernel.sh workunit
Yes, the ENXIO is expected. Assuming it's being propagated out to dd, and the test passes (outputs OK at the end of k... Josh Durgin
05:55 AM Bug #3427: krbd: unmap does not remove block device properly
We had some discussion about the whether an atomic bit
operation for this was sufficient, or whether a memory
barri...
Alex Elder

01/24/2013

11:59 PM Bug #3936: rbd: Strange dd speed behaviour (server side issue?)
I also tried to do:
dd if=/dev/rbd/rbd/test of=/dev/null bs=4M - the same situation.
Ivan Kudryavtsev
11:57 PM Bug #3936 (Rejected): rbd: Strange dd speed behaviour (server side issue?)
I have 3 node/15 osds (5 on each), every on separate drive installation (with SSD cache), journal in RAMFS. XFS as ba... Ivan Kudryavtsev
08:15 AM Bug #3927: krbd: I/O errors (ENXIO) during rbd/kernel.sh workunit
Hey! I just looked at the test, and here's how it ends:
# remove snapshot and detect error from mapped snapshot
...
Alex Elder
08:15 AM Bug #3927: krbd: I/O errors (ENXIO) during rbd/kernel.sh workunit
This is the relevant portion of the yaml file:
- workunit:
clients:
all:
- rbd/map-unmap.sh
...
Alex Elder
08:09 AM Bug #3927 (Closed): krbd: I/O errors (ENXIO) during rbd/kernel.sh workunit
I'm seeing ENXIO errors at what I believe to the "rbd/kernel.sh
teuthology workunit while testing the new request co...
Alex Elder
05:49 AM Feature #3926 (Resolved): krbd: use slab allocation for common data structures
There are some common data structures--like image and object
requests--that are very frequently allocated and would ...
Alex Elder
05:29 AM Bug #3925 (Resolved): krbd: sysfs write lockdep warnings
... Alex Elder

01/23/2013

11:27 AM Tasks #2853 (In Progress): krbd: read path
With my patches for the basic new request code now
out for initial review, I've started working on this
feature. I...
Alex Elder
11:20 AM Subtask #2852 (In Progress): krbd: open parent on open
The many patches have now been posted for review.
Included in that is a small, temporary patch that enables
this ...
Alex Elder
05:21 AM Fix #3665: librbd: deadlock during flatten
possibly here: ... Sage Weil

01/22/2013

02:40 PM Feature #3877 (Fix Under Review): krbd: don't wait for notify ack to complete
I've posted this code for review. I continue to do testing. Alex Elder
02:39 PM Subtask #3741 (Fix Under Review): krbd: rework request tracking code
I've posted this code for review. I continue to do testing. Alex Elder
02:39 PM Tasks #3755 (Fix Under Review): krbd: use new request tracking code for sync object operations
I've posted this code for review. I continue to do testing. Alex Elder
02:39 PM Feature #3754 (Fix Under Review): krbd: use new request tracking code for notify ack
I've posted this code for review. I continue to do testing. Alex Elder
02:19 PM Feature #3893 (Rejected): krbd: document the new request code
There are bits and pieces of the new request code
documented for the kernel rbd client--in the comments
and in the ...
Alex Elder
10:27 AM Bug #3889 (Won't Fix): krbd: handle zero-length requests
I'm pretty sure there are some special zero-length
requests (like flush) that can come down from the
block layer. ...
Alex Elder

01/21/2013

09:35 AM Bug #3737: Higher ping-latency observed in qemu with rbd_cache=true during disk-write
Hi Josh,
according to our conversation I did some testing.
I started the dd if=/dev... of=/tmp/doof.dat bs=4k cou...
Oliver Francke

01/20/2013

08:26 PM Feature #3877: krbd: don't wait for notify ack to complete
I have implemented this in the new request code.
It will be posted for review along with the rest
of that new code ...
Alex Elder
08:14 PM Feature #3877 (In Progress): krbd: don't wait for notify ack to complete
Ian points out that "I've already implemented this change"
suggests that the status of this issue should at least
b...
Alex Elder
08:26 PM Subtask #3741 (In Progress): krbd: rework request tracking code
Considering this "is actually work that's mostly complete"
I'm (finally) marking it "In Progress."
This code is f...
Alex Elder
08:22 PM Feature #3754 (In Progress): krbd: use new request tracking code for notify ack
I have completed implementing sending synchronous acknowledgement
in response to a watch request notification. It i...
Alex Elder
08:19 PM Tasks #3755 (In Progress): krbd: use new request tracking code for sync object operations
I have completed implementing all of these in the new request
code:
- synchronous object read (for v1 header object...
Alex Elder

01/19/2013

08:40 AM Feature #3877 (Closed): krbd: don't wait for notify ack to complete
When we receive notification of a change to an rbd image's header
object we need to refresh our information about th...
Alex Elder

01/18/2013

02:01 PM Bug #3871 (Resolved): krbd: initial header read may be out of date
Currently krbd uses the version parameter of a watch operation to try to prevent this, but that was never implemented... Josh Durgin
10:15 AM Tasks #3755: krbd: use new request tracking code for sync object operations
The sync header read operation was another one that was needed.
That's basically done too.
All of this will be re...
Alex Elder
10:09 AM Tasks #3755: krbd: use new request tracking code for sync object operations
I have been looking in detail at how the watch requests are
implemented and in the process identified a few potentia...
Alex Elder
08:58 AM Feature #3861 (Resolved): rbd: consider splitting rbd_osd_req_op_create()
When it was out for review, Josh suggested that it might
be better to have separate (type-checking) functions for
b...
Alex Elder
08:07 AM Bug #3859 (Resolved): osd_client: define ceph_osdc_clear_request_linger()
There is a ceph_osdc_set_request_linger() function that
sets a flag on a request and takes an additional reference.
...
Alex Elder
08:04 AM Bug #3858 (Resolved): osd_client: ceph_osdc_wait_request() seems wrong
The only error wait_for_completion_interruptible() will
return is ERESTARTSYS. So if that gets returned inside
cep...
Alex Elder
07:14 AM Feature #1491: qemu: make qemu-img convert fast
This was rejected because feature is not relevant anymore. At the time, when I was looking at it there was some obvio... Yehuda Sadeh

01/17/2013

06:37 PM Bug #3413 (Resolved): rbd bench-write fails with assert when rbd caching turned on
commit:d81ac8418f9e6bbc9adcc69b2e7cb98dd4db6abb Josh Durgin
01:39 PM Bug #3413 (Fix Under Review): rbd bench-write fails with assert when rbd caching turned on
branch wip-rbd-bench-write Josh Durgin
02:24 PM Bug #3847 (Resolved): rbd: figure out correct byte order for watch version
In the process of refactoring rbd code that builds up osd
operations I noticed that for NOTIFY_ACK and WATCH operati...
Alex Elder
12:32 PM Feature #1491 (Rejected): qemu: make qemu-img convert fast
Yehuda Sadeh
09:50 AM Feature #3838 (New): krbd: use common functions for striping calculations
With the STRIPINGV2 feature bit, format 2 striping has the same parameters as cephfs striping. Re-work the rbd object... Josh Durgin
09:12 AM Feature #3754: krbd: use new request tracking code for notify ack
Yay! Sage Weil
04:52 AM Feature #3754: krbd: use new request tracking code for notify ack
Yeehah! All tests passed, including the previously-failing
blogbench.sh, fsstress, and two passes through xfstests.
Alex Elder

01/16/2013

09:23 PM Feature #3754: krbd: use new request tracking code for notify ack
OK, that quick fix wasn't enough.
I had a spinlock protecting the check for something being
complete. But that w...
Alex Elder
08:13 PM Feature #3754: krbd: use new request tracking code for notify ack
Well that's unfortunate. I hit the same problem. I'll
need to take a closer look I guess.
Alex Elder
07:39 PM Feature #3754: krbd: use new request tracking code for notify ack
Seems to be working better. It may end up being an
atomic rather than protecting with a spinlock, but
either way, ...
Alex Elder
03:15 PM Feature #3754: krbd: use new request tracking code for notify ack
I've pretty much implemented this feature but having done
this I'm looking at a crash that happened with this code
...
Alex Elder
11:38 AM Feature #3817 (Resolved): librbd: make cache write-through until a flush is encountered
Writeback caching is unsafe if higher layers don't send flushes. qemu can be accidentally misconfigured to not send f... Josh Durgin

01/15/2013

07:51 PM Feature #3754: krbd: use new request tracking code for notify ack
I'm not sure yet whether the problem has to do with this
or whether it's in the existing "new request" code. But
I...
Alex Elder
02:03 PM Feature #3635: rbd cli: call "udevadm settle" after use of add/remove kernel interface
commit:15bb00cafc31305cacf3c4684a429c2c9ee6f804 in master
Dan Mick
02:03 PM Feature #3635 (Resolved): rbd cli: call "udevadm settle" after use of add/remove kernel interface
Dan Mick
02:02 PM Feature #3784: rbd: issue modprobe when rbd map is called
commit:e94b06a19218decaf7d2d7b009bd862040f20285 in master
Dan Mick
02:01 PM Feature #3784 (Resolved): rbd: issue modprobe when rbd map is called
Dan Mick
12:08 PM Subtask #3741: krbd: rework request tracking code
I found the source of my trouble, and in the process understood
a little more about some subtlety in bio reference c...
Alex Elder

01/14/2013

08:02 PM Subtask #3741: krbd: rework request tracking code
OK, I ran a test and got a crash. The bio built for
an object request gets handed off to an osd request.
I need to...
Alex Elder
07:32 PM Subtask #3741: krbd: rework request tracking code
I spent the day trying to find the memory leak and finally
found it. The structure being leaked was a bio. It was
...
Alex Elder
06:48 AM Subtask #3741: krbd: rework request tracking code
For some reason my tests started hanging on Friday when
I added memory debug code for catching leaks and reuses.
I ...
Alex Elder
11:55 AM Bug #2871 (Resolved): rbd export command hangs when trying to export an image of size 0 to a loca...
Not certain which recent fix resolved this, but it works now.
Dan Mick
11:32 AM Bug #3585 (Closed): Image import via QEMU-IMG results in a corrupt rbd
Great, glad to hear it's fixed. Josh Durgin
11:09 AM Bug #3427: krbd: unmap does not remove block device properly
Patch posted for review. I'm not sure I'll be able to test
the scenario very well but hopefully it can be seen by
...
Alex Elder
09:56 AM Bug #3427: krbd: unmap does not remove block device properly
Implementing the change I described now. Alex Elder

01/11/2013

02:17 PM Subtask #3741: krbd: rework request tracking code
Unfortunately my system crashed after an hour or so. The
crash was in the network driver, and a little analysis
su...
Alex Elder
10:45 AM Subtask #3741: krbd: rework request tracking code
My full test run isn't complete but I seem to have resolved
whatever problem I was hitting yesterday. I have not ye...
Alex Elder

01/10/2013

08:23 PM Bug #3585: Image import via QEMU-IMG results in a corrupt rbd
This seems to be fixed in QEMU 1.3.0 and Ceph 0.56.1
I've tried QED -> Raw -> Ceph -> Raw then QED -> Ceph -> Raw an...
Matt Anderson
07:10 PM Feature #3635 (In Progress): rbd cli: call "udevadm settle" after use of add/remove kernel interface
Dan Mick
07:09 PM Feature #3784 (In Progress): rbd: issue modprobe when rbd map is called
Dan Mick
06:04 PM Feature #3784 (Resolved): rbd: issue modprobe when rbd map is called
rbd map will not work unless the rbd kernel module is loaded, and this must be done manually. Add code to rbd to cau... Dan Mick
05:59 PM Bug #3518 (Can't reproduce): rbd import file --format 2 creates an image named '--format'
Dan Mick
05:59 PM Bug #3518: rbd import file --format 2 creates an image named '--format'
It seems that this no longer happens as of e6f284e945f45e39c57921149d4551d9e78557a5,
so closing non-reproducible.
Dan Mick
04:37 PM Bug #3688 (Won't Fix): rbd allows image of size 0 to be created
I claim that zero-sized images are legal, if not particularly useful in that size...but one might well want to create... Dan Mick
02:23 PM Subtask #3741: krbd: rework request tracking code
I am leaving shortly for a few hours. In reviewing this
new code I find a few things that make it a little hard
ma...
Alex Elder
01:00 PM Subtask #3741: krbd: rework request tracking code
I did some testing yesterday and found that I got I/O errors
while running xfstests. This was unexpected; I thought...
Alex Elder
 

Also available in: Atom