openstack team mailing list archive

[Florian Haas <florian@xxxxxxxxxxx>]

On Thu, Feb 16, 2012 at 3:09 PM, Tom Ellis <tom.ellis@xxxxxxxxxxxxx> wrote:
> nova-volume service
>  * Current weakness in the HA setup, unless you are willing to use iscsi
> tgtd with DRBD. I believe this would still have some problems when
> failing over with the initiators that are logged in.

Nope, this is well understood and supported. It all amounts to getting
the Pacemaker configuration right, but DRBD-backed, tgt-based iSCSI
targets are perfectly capable of failover that is fully transparent to
initiators.

That being said, of course nova-volume with RBD (RADOS block device)
would be another approach that may provide better reliability (more
replicas, where DRBD is limited to 2 -- 4 with some trickery), better
scaleout (seamless expansion and contraction of the storage space),
and better self-healing capabilities (automatic rebalancing if
individual replicas fail).

Hope this is useful.

Cheers,
Florian

-- 
Need help with High Availability?
http://www.hastexo.com/now