[Sheepdog] sheepdog question

[MORITA Kazutaka]

At Mon, 10 Oct 2011 15:34:18 +0200,
Dennis Jacobfeuerborn wrote:
> 
> On 10/10/2011 06:07 AM, MORITA Kazutaka wrote:
> > At Sun, 9 Oct 2011 00:53:48 +0200,
> > Marin Marušić wrote:
> >> Hi,
> >>
> >> I have a question, if I have uneven storage space available across various
> >> servers in the cluster, will I still be able to use Sheepdog? Or must I have
> >> the same storage configuration in each and every server? FE I have 3 servers
> >> with 6x500GB disks in raid 10, and have 3 servers with only 2 500GB drives
> >> RAID1. And I can't change that config for the time being :)
> >
> > Run multiple sheep daemons proportional to its capacity (3 daemons on
> > servers with 6x500 GB and one daemon on the other), and Sheepdog would
> > work as you expect, I think.  But the best way is to stop using RAID
> > and run a sheep daemon on for each disk.
> 
> What happens if one of the disks dies? Can the daemon responsible for this 
> disk deal with that and can I hot-plug a new disk an easily tell the daemon 
> to treat it as such?

Yes, the daemon automatically leaves Sheepdog if the disk dies, and
you can easily add a new disk if you start a new sheep daemon.

> This would make such a "one daemon per disk" setup 
> superior to a raid-0 setup as only a smaller chunk of data would disappear 
> when a disk dies.

Yes, in terms of reliability, the above approach I described is better
than RAID0.


Thanks,

Kazutaka