[sheepdog-users] Did i get sheepdog wrong?

[Stefan Priebe]

Am 30.06.2012 um 03:23 schrieb icez network <icez at icez.net>:

> Sheepdog 'does' use the technique to divide the vdi ('disk image' as
> you might concern) into 4MB 'object' and use the consistency hashing
> to distribute these 'object' across the cluster. So if you're running
> a cluster consist of 2 nodes with 2 copies, the object will be stored
> on both node. By the way, you still can access your 'vdi' from every
> node within the cluster.

That's what I expected but it doesn't work for me. I've 4 nodes with 4 disks each. And disk usage rises only on ONE sheep while writing to disk.


> You may use command 'collie node info' to get the current usage of each node.

Yes usage stays 0 on ALL nodes and just free space lowers...

Stefan

> On Sat, Jun 30, 2012 at 4:23 AM, Stefan Priebe <s.priebe at profihost.ag> wrote:
>> Hello list,
>> 
>> i think i got sheepdog wrong. I thought sheepdog will spread the WHOLE disk
>> image across all nodes to archieve maximum performance. Instead it seems
>> sheepdog just holds X copies of whole images. And hen places different
>> images onto different nodes.
>> 
>> Is this correct? So it is working completely different than ceph / rbd-
>> 
>> Greets
>> Stefan
>> --
>> sheepdog-users mailing lists
>> sheepdog-users at lists.wpkg.org
>> http://lists.wpkg.org/mailman/listinfo/sheepdog-users
> 
> 
> 
> -- 
> Personal hosting by icez network
> http://www.thzhost.com