Xen users - Nov 2005 - RE: RE: Tutorial : Debian, Xen and CLUSTER / GFS Supp ort

>> >In theory.  In practice, the snapshot target is horribly unstable
and
>> >will nuke your data sooner rather than later :-).
>> is
>> much more stable.
>
>It''s being actively developed?  Sounds good!
Actively is not the word I think... bug fixes so far.
>I thought it was a dead thing, haven''t seen a single thread about
it
>on dm-devel for at least 6 months.
>
>Which ML did you see activity on ?
I can see quite a few threads on https://www.redhat.com/archives/dm-devel/
But, unless I am wrong DM and LVM are included in the kernel tree, so kernel
maintainers are working on these bugfixes too I think.
>> I''am trying to use snapshot features under heavy IO stress
>> and so far no problem. I will test and dig a bit more in the ML to be
>>sure
>> that is working as intended.
>How are you testing?
>Which patches to dm-snapshot are you using?
>I''ve got an HP ProLiant (x86_64) here that''s not going
into production
>for a month.  I''d like to chip in with some dm-snapshot (vs. Xen)
>testing.
For my LVM/DM snapshot tests, I have created 4 block devices of 8GB in LVM
group of 50GB. I am using DM and LVM2 libs and daemon from the Debian Etch
packages. I wrote a program to write randomly in the 4 block devices (it
looks like it''s write about 20MB/s, I didn''t really optimize
it.). While
this program is active, I do a snapshot of the first volume, and copy it on
a tape device, I destroy snapshot, then do the same with second, third and
fourth volume. I launched the loop of snapshot/backups many times in day
without a single problem.
My test were made on a Dell 1750 (32bit) with a Perc 3D/I SCSI card.

Did you experience any problem?


Regards,

	Christophe Painchaud



Christophe Painchaud


_______________________________________________
Xen-users mailing list
Xen-users@lists.xensource.com
http://lists.xensource.com/xen-users

PAINCHAUD Christophe wrote:
> For my LVM/DM snapshot tests, I have created 4 block devices of 8GB in LVM
> group of 50GB. I am using DM and LVM2 libs and daemon from the Debian Etch
> packages. I wrote a program to write randomly in the 4 block devices (it
> looks like it''s write about 20MB/s, I didn''t really
optimize it.). While
> this program is active, I do a snapshot of the first volume, and copy it on
> a tape device, I destroy snapshot, then do the same with second, third and
> fourth volume. I launched the loop of snapshot/backups many times in day
> without a single problem.
I''m very impressed!
> > I''ve got an HP ProLiant (x86_64)
> My test were made on a Dell 1750 (32bit) with a Perc 3D/I SCSI card.
>
> Did you experience any problem?
Yes.  I took one snapshot of a 3GB full 20GB ext3 volume onto a 10GB
snapshot target volume.  After executing the LVM2 snapshot command,
the original 20GB volume was trashed.  Seems like all disk accesses to
the 20GB volume were misaligned by a few kB''s or so.  I could see
directories and files, but only because that''s cached in RAM. 
Contents of all files were swapped randomly.  Totally trashed.

I wrote a bug report to dm-devel and got no replies.
I wrote a bug report to RedHat (Fedora Core 4, which I was using at
the time) and got no replies.

:-/.

_______________________________________________
Xen-users mailing list
Xen-users@lists.xensource.com
http://lists.xensource.com/xen-users