[linux-lvm] Is it possible to invalidate snapshots other than by overfilling?

[Rob West]

Jonathan,

Thanks for responding.

I think you're right about why we're getting the message on boot up. We
definitely don't have any dm modules in our initrd.

However, I have verified that just the act of rebooting with a snapshot in
place doesn't cause the snapshot to become invalid.

One thing that I failed to mention before is that so far the cases where
we've seen this problem are also resyncing md mirrors on boot up. I have
verified that the resyncing alone doesn't cause the invalidation. A fellow
engineer is working on replicating the scenario even closer.

Cheers,
Rob

On Wed, Jul 2, 2008 at 11:42 AM, Jonathan Brassow <jbrassow at redhat.com>
wrote:

> It sounds like the module needed by device-mapper to support snapshots is
> not loaded when it is needed.  (Perhaps the module - dm-snapshot - is not in
> your initrd image?)
>
> After step 3, but before step 4, try doing a 'dmsetup table' to see if
> things are properly mapped.  A 'dmsetup status' wouldn't hurt either...
>
> Certainly, there should never be a kernel oops...  It's tough for me to say
> whether this bug is already fixed - given the age of your kernel.
>
>  brassow
>
>
> On Jul 1, 2008, at 4:10 PM, Rob West wrote:
>
>  Does anyone know of ways that snapshots can become invalid other than by
>> writing more to the origin than the COW device can hold?
>>
>> The reason I ask is that we're having cases where a snapshot is becoming
>> invalid, but we are very skeptical that it is due to overfilling. The size
>> of the origin is 83G, and the size of the snapshot is 8G.
>>
>> Here's what little we can tell from the logs:
>> 1. Backup script creates a snapshot and starts copying data from it.
>> 2. Before backup is finished (and thus, snapshot removed), the user
>> reboots.
>> 3. On reboot, we get the following message which seems to be normal for
>> this kernel:
>>    kernel: device-mapper: table: 253:2: snapshot-origin: unknown target
>> type
>> 4. About 23 hours later, the backup script tries to create the snapshot
>> again but fails b/c it's still there.
>> 5. The backup script tries to clean up by removing the snapshot, but
>> lvremove causes a kernel oops because the snapshot is invalid. (We have
>> identified a potential patch for this, but are trying to figure out why the
>> snapshot was invalid in the first place.)
>>
>> The kernel we're using is based off of Fedora 6 (2.6.18-1.2849).
>>
>> Not sure whether it matters, but device-mapper is version 1.02.07 release
>> 4.0.RHEL4 and lvm2 is version 2.02.06 release 6.0.RHEL4.
>>
>>
>> Thanks for any help,
>> Rob
>> _______________________________________________
>> linux-lvm mailing list
>> linux-lvm at redhat.com
>> https://www.redhat.com/mailman/listinfo/linux-lvm
>> read the LVM HOW-TO at http://tldp.org/HOWTO/LVM-HOWTO/
>>
>
> _______________________________________________
> linux-lvm mailing list
> linux-lvm at redhat.com
> https://www.redhat.com/mailman/listinfo/linux-lvm
> read the LVM HOW-TO at http://tldp.org/HOWTO/LVM-HOWTO/
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://listman.redhat.com/archives/linux-lvm/attachments/20080702/417fe160/attachment.htm>