We need to document how to add more volume nodes that guest VMS can access.
On diablo, you can indeed deploy several nova-volume. Under Diablo, every service would report himself to the scheduler - that last electing one when there is a volume to spawn.
On Essex, you set --iscsi_ip_address=$-ip" per volume node.
On trunk, set the option "--iscsi_ip_prefix=nnn.nnn.nnn.yyy" in the nova.conf - apparently this makes every node registered into the database.
Vish describes in this post: http://osdir.com/ml/openstack-cloud-computing/2012-05/msg00860.html:
> FYI iscsi_ip_prefix doesn't exist in essex. The ip is passed back to the
> compute node based on what it has stored in the database, so the compute
> node no longer finds it through discovery and matching to the prefix. You
> should only need iscsi_ip_address on the volume node to make sure that the
> db entry is created properly.
>
> Vish
Moving to high as I believe we should try to revise the volume information in the install/deploy guide as it is not production level currently. See comments on http:// docs.openstack. org/essex/ openstack- compute/ install/ apt/content/ installing- the-cloud- controller. html.