Project

General

Profile

Activity

From 05/16/2012 to 06/14/2012

06/14/2012

03:06 PM Bug #2534: librbd: make sure watch is established on same header version as initial read was
Okay, this is blocked by #2592. Greg Farnum
02:37 PM Feature #1480 (Fix Under Review): librbd: image locking
wip-rbd-locking Greg Farnum
12:59 PM Bug #2535: rbd: random data corruption in vm
Sage Weil wrote:
> Just a bit of context: rbd without caching does a 'sparse-read' operation, which uses FIEMAP to d...
Sage Weil
12:52 PM Bug #2535: rbd: random data corruption in vm
Just a bit of context: rbd without caching does a 'sparse-read' operation, which uses FIEMAP to determine which parts... Sage Weil
12:50 PM Bug #2535: rbd: random data corruption in vm
Let's try a different tack: I pushed a osd-verify-sparse-read-holes branch to ceph.git (based on 0.47.2) that reads ... Sage Weil
09:09 AM Bug #2535: rbd: random data corruption in vm
Status update:
I tried modifying the iotester so that it would work directly on the block device, in the hopes I c...
Guido Winkelmann
09:09 AM Bug #2410: hung xfstest #68
disabled 68 in qa for the time being. Sage Weil
09:03 AM Bug #2522: xfstest #219
Sigh.. took a quick look and it's non-obvious why the repquota output doesn't match. Disabling this for now, but lea... Sage Weil

06/13/2012

02:28 PM Feature #2586 (Rejected): rbd: check/take locks on --lock
if you pass --lock to rbd, take an exclusive lock, do whatever, unlock Sage Weil
02:20 PM Feature #2585 (Resolved): rbd: clone command
A command for the rbd tool to create a child image from a parent. Example:
rbd clone --parent pool/image@snap pool...
Sage Weil
01:56 PM Feature #2467 (Rejected): qemu: implement bdrv_invalidate_cache
I've tested migration with caching, and read the code, and it looks like this is unnecessary. qemu is doing a flush b... Josh Durgin
01:20 PM Bug #2522: xfstest #219
The problem here appears to be that the output of the repquota
command is not what's expected. I think the group qu...
Alex Elder
11:20 AM Cleanup #2347 (Resolved): The rbd help text is misleading on required arguments
commit:67710a65c7cd1173c73c40241572d615dd7da1f3 Sage Weil
10:30 AM Bug #2572 (Resolved): krbd: writeback errors?
While trying to reproduce a null pointer messenger problem,
I kept hitting messages like this after some (fairly ran...
Alex Elder
10:27 AM Bug #2535: rbd: random data corruption in vm
Sage Weil wrote:
> Guido Winkelmann wrote:
> > Sage Weil wrote:
> > > Are there multiple partitions or is LVM on t...
Guido Winkelmann
09:23 AM Feature #2568 (Resolved): qa: run xfstests on qemu+rbd
This will build on #2566:
* stage xfstests on vdb, like a regular workunit, and:
* map additional rbd images to r...
Sage Weil
09:21 AM Feature #2567 (Resolved): qa: add qemu+rbd jobs to qa suite
Add a bunch of workunits to the qa suite that will run on top of rbd inside a vm. Sage Weil
09:20 AM Feature #2566 (Duplicate): teuthology: task to run rbd workunits in a vm
teuthology task that will:
* download workunit vm
* create and format rbd image
* mount, stage a workunit in rbd...
Sage Weil

06/12/2012

02:40 PM Feature #2561: rbd: copyup command
What? How does a class function of any kind provide atomicity in cross-OSD data copies? Greg Farnum
02:37 PM Feature #2561: rbd: copyup command
'rbd copyup pool/image' command to copy any missing objects up from the parent. simple O(n) operation that leverages... Sage Weil
02:11 PM Feature #2561 (Resolved): rbd: copyup command
'rbd copyup pool/image' command to copy any missing objects up from the parent. simple O(n) operation that leverages ... Sage Weil
02:39 PM Feature #2562 (Resolved): librbd: open parent images, read path, write path
- when we open an image, open the parent image too.
- make reads fall through to parent
- guard writes beyond paren...
Sage Weil
02:05 PM Feature #2560 (Resolved): rbd: safe parent deletion
- maintain map of parent/child pairs in each child pool... Sage Weil
02:04 PM Feature #2531: rbd: fencing broken clients
As I see it, we have two options that we need to choose between.
1) We can add fencing to librbd and let anybody do ...
Greg Farnum
01:55 PM Feature #2559 (Resolved): cls_rbd: copyup method
- client provides object content
- if object exists, fail with EEXIST (or 0, or something)
- if object does not exi...
Sage Weil
01:54 PM Feature #2558 (Resolved): cls_rbd: child/parent methods
On the new image header:
- set_parent(poolid, image (maybe id, maybe name), snapid)
On the per-pool child list:
...
Sage Weil
01:52 PM Feature #2557 (Rejected): QEMU support for image locking
We should convert QEMU to make use of rbd cooperative locking, once it's done (#1480).
And any other appropriate c...
Greg Farnum
01:50 PM Feature #2556 (Resolved): rbd tool: break image locks
Once #1480 is done, expose lock breaking via the rbd tool. Greg Farnum
10:10 AM Bug #2535: rbd: random data corruption in vm
Guido Winkelmann wrote:
> Sage Weil wrote:
> > Are there multiple partitions or is LVM on the disk, or is the file ...
Sage Weil
10:07 AM Bug #2535: rbd: random data corruption in vm
Sage Weil wrote:
> Are there multiple partitions or is LVM on the disk, or is the file system on the raw device?
...
Guido Winkelmann
09:29 AM Bug #2535: rbd: random data corruption in vm
Are there multiple partitions or is LVM on the disk, or is the file system on the raw device? Sage Weil
05:32 AM Bug #2535: rbd: random data corruption in vm
Am Montag, 11. Juni 2012, 09:30:42 schrieb Sage Weil:
> If you can reproduce it with 'debug filestore = 20' too, tha...
Guido Winkelmann
05:29 AM Bug #2535: rbd: random data corruption in vm
The bug also does not seem to have any effect with the setting "filestore fiemap = false" in ceph.conf. Guido Winkelmann

06/09/2012

08:06 PM Bug #2535: rbd: random data corruption in vm
The information that *should* let us fully diagnose:
* set
debug osd = 20
debug filestore = 20
debug ms = ...
Sage Weil
08:04 PM Bug #2535 (Resolved): rbd: random data corruption in vm
From ML:... Sage Weil

06/08/2012

04:42 PM Bug #2534 (Resolved): librbd: make sure watch is established on same header version as initial re...
Right now there's a race where it doesn't. Greg Farnum
10:12 AM Feature #2531: rbd: fencing broken clients
I talked to Sam about the combination of blacklisting, bad client writes, and changing primaries that we discussed an... Greg Farnum
10:06 AM Bug #2478 (New): krbd: unmap on 3.4.0: scheduling while atomic...
Sage Weil
09:51 AM Bug #2532: rbd command allows passing in -K </path/to/secret>, but long version of (--secret) doe...
That's probably best. It is always easier though when all subcommands under the main command, rbd in this case used o... Sam Zaydel
09:00 AM Bug #2532: rbd command allows passing in -K </path/to/secret>, but long version of (--secret) doe...
Oh, i see.
I think the right fix is to make '--secret' and synonym for '--keyfile', and fix up rbd to use the conf...
Sage Weil
08:20 AM Bug #2532: rbd command allows passing in -K </path/to/secret>, but long version of (--secret) doe...
When I try to use --keyfile=<file> with map, it seemingly fails, but using --secret=<file> succeeds. ... Sam Zaydel
08:13 AM Bug #2532: rbd command allows passing in -K </path/to/secret>, but long version of (--secret) doe...
This is part of the rbd cmd helper message. It seems that for the map command one uses --secret.... Sam Zaydel
07:00 AM Bug #2532: rbd command allows passing in -K </path/to/secret>, but long version of (--secret) doe...
the option is --keyfile <file>... where did you see --secret <file> documented? Sage Weil
05:49 AM Bug #2532 (Resolved): rbd command allows passing in -K </path/to/secret>, but long version of (--...
While rolling back a snapshot I succeed when I pass in `-K with location of key file, but it looks like I fail when I... Sam Zaydel

06/07/2012

09:38 PM Feature #2531 (Resolved): rbd: fencing broken clients
Sage Weil
04:25 PM Documentation #2530 (Closed): Doc: rbd manpage doesn't mention watch; usage does, and it works
Dan Mick
03:10 PM Bug #2528 (Resolved): Mounted RBD image appears to go read-only after a snapshot is created
I have been able to repeat this a number of times. Essentially, I create a small rbd device, using the map command in... Sam Zaydel

06/06/2012

01:22 PM Bug #2522: xfstest #219
ubuntu@teuthology:/a/nightly_coverage_2012-06-05-b Sage Weil
01:21 PM Bug #2522 (Closed): xfstest #219
... Sage Weil

06/05/2012

09:56 AM Feature #1480: librbd: image locking
lock(entity)
unlock(entity)
new code should lock before open, unlock on close.
the rbd map tool have 'lock lis...
Sage Weil

05/30/2012

02:28 PM Bug #2488 (Can't reproduce): rbd: EIO from rbd map
... Sage Weil
08:35 AM Bug #2478: krbd: unmap on 3.4.0: scheduling while atomic...
with patch above, this trace appears after ~15 sec from unmap operation
[ 2453.203365] ------------[ cut here ]---...
Andrey Korolyov

05/26/2012

01:30 PM Bug #2478 (In Progress): krbd: unmap on 3.4.0: scheduling while atomic...
commit:cd9d9f5df6098c50726200d4185e9e8da32785b3 already fixes this; we just neglected to send it upstream for 3.4. I... Sage Weil
06:05 AM Bug #2478 (Can't reproduce): krbd: unmap on 3.4.0: scheduling while atomic...
On 3.4.0 on the 'rbd unmap' call kernel rbd driver produces following trace(same for vanilla):
[ 8660.786269] BUG...
Andrey Korolyov

05/23/2012

12:52 PM Feature #2468 (Resolved): librbd: provide a way for a user to flush and invalidate the cache
This could be an admin socket command. This would make live migration work with older qemu when rbd caching is enabled. Josh Durgin
12:37 PM Feature #2467 (Resolved): qemu: implement bdrv_invalidate_cache
This is used during live migration to clear librbd's cache on the destination host before starting the guest there. I... Josh Durgin
12:35 PM Feature #2466 (Resolved): librbd: add invalidate_cache function to interface
Qemu requires this to make live migration work when caching is enabled. Josh Durgin

05/21/2012

02:48 PM Bug #2457 (Resolved): libvirt: migration fails with rbd in 0.9.11 and 0.9.12
As reported on libvirt-users:
https://www.redhat.com/archives/libvirt-users/2012-May/msg00088.html
Looks like a...
Josh Durgin
 

Also available in: Atom