[Linux-cluster] Replication for iSCSI target?

[Fajar A. Nugraha]

On Tue, Apr 5, 2011 at 3:08 AM, Michael McGlothlin
<michaelm at plumbersstock.com> wrote:
> All the storage appliances I've looked at were either severely limited
> (2TB limit, no replication) or expensive ($4000 per node)

Nexentastor Community Edition has 18 TB usage capacity limit, and
should have active-passive replication support.

> Speed isn't nearly the issue for me that
> reliability is. DRBD sounds good in that it works at the block level

Give drbd a try then. Just make sure you do enough tests first. I had
a test setup with no fencing (yes, I know it's bad, that's why it's
only a test) with dbrd+ocfs2, and sometimes a node failure cause
unpredicatable cluster-wide error (both nodes restarting, split brain,
etc). In production environment timeout for fencing should be lower
than other subsystems (e.g. drbd).

-- 
Fajar