[sheepdog-users] vdi problem

[Liu Yuan]

On 06/17/2013 03:22 PM, Valerio Pachera wrote:
> First of all, everything got fixed!
> I suspect troubles where related to a disk that got full on node 'sheepdog001'.
> I didn't notice it before satturday (15/6/13).
> I added a node (3T) and unplugged the full disk.

>From your 'node info' we can see that there are plenty much enough room
left, but the problem some is disk was full and lead the cluster status
to be full.

I am trying to find out if our hash function for object distribution
work as expected,

$ ./collie/collie node md info --all

Id	Size	Used	Avail	Use%	Path
Node 0:
 0	2.0 GB	1.6 GB	419 MB	 79%	/tmp/sheepdog/0/obj
 1	482 MB	424 MB	58 MB	 87%	/tmp/sheepdog/3
Node 1:
 0	2.0 GB	2.0 GB	0.3 MB	 99%	/tmp/sheepdog/1/obj
Node 2:
 0	2.0 GB	2.0 GB	0.3 MB	 99%	/tmp/sheepdog/2/ob

With 500M and 2G disks in a node, we can see that objects are indeed
evenly distributed.

I don't have big enough disk to simulate the real workload. But I think
we should make sure different sized disks won't cause cluster-full problem.

Anybody help try different sized disks combination in one node and post
the result?

Thanks,
Yuan