[linux-lvm] 15 snapshot limit?

Larry Dickson ldickson at cuttedge.com
Tue Sep 2 14:31:32 UTC 2008


Sorry, I did not make that last question clear:

What (if any) is the formula by which I can predict the number of snapshots
that are possible, given a certain size of the parent volume and/or
snapshot and a certain amount of available memory? Milan implied there was
such a limit, and I seem to remember it, but I can't seem to find it spelled
out anywhere.

Larry

On 8/29/08, Larry Dickson <ldickson at cuttedge.com> wrote:

> You said:
>
> > (But technically I tested >50 snapshot from one LV - should be limited
> > only by available memory in recent kernels)
>
> What is that limit? I suspect that is what I have run up against.
>
> Larry
>
>  On 8/29/08, Milan Broz <mbroz at redhat.com> wrote:
>>
>> Larry Dickson wrote:
>> > With a 768 GB logical volume in a 2235 GB volume group, and 4 GB of
>> > memory, I try to set up lots of snapshots (of 45 or 50 GB each) and
>> > can't get past 15. Sometimes I'll succeed in generating more than 15
>> > COWs in /dev/mapper, but lvdisplay never shows more than 15 snapshots in
>> > the "LV snapshot status source of" listing under the parent volume. I
>> > searched and found 15 snapshots mentioned in a couple places like
>> > wiki.samba.org/index.php/Shadow_Copies_with_Snapshots
>> > <http://wiki.samba.org/index.php/Shadow_Copies_with_Snapshots> - but
>> > nothing really authoritative.
>> >
>> > Is there really such a limit? Or is it a memory requirement - I seem to
>> > remember so many MB memory per GB of storage, but can't find that
>> > either. Versions are (variants of) lvm2-2.02.38, device-mapper-1.02.26,
>> > kernel-smp-2.6.20 (64 bit).
>>
>> Hi,
>> no, there is no hard limit for snapshots.
>>
>> But IIRC patches which solves some problems related with this (using
>> private dm-io
>> mempool) were commited in 2.6.22 kernel.
>> But anyway, many snapshot from one origin causes serious performance
>> degradation.
>> (Current snapsot implementation is not ideal in this situation - every
>> write to origin
>> is replicated N-times to COW snapshot area - iow: every snapshot is
>> independent and
>> need copy changed area separately.)
>>
>> (But technically I tested >50 snapshot from one LV - should be limited
>> only by available
>> memory in recent kernels).
>>
>> Milan
>> --
>> mbroz at redhat.com
>>
>>
>> _______________________________________________
>> 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/20080902/72474acc/attachment.htm>


More information about the linux-lvm mailing list