[Openstack-operators] Neutron in one cell, Nova network in another
rbryant at redhat.com
Mon Dec 9 17:30:45 UTC 2013
On 12/07/2013 11:45 AM, Tim Bell wrote:
> Anyone with experience running mixed network models between cells ?
> I wonder if it is possible as a way of gradually migrating to Neutron.
I may need to look closer but this strikes me as something that probably
won't work. There is tighter coupling betweens cells than I'd really like.
What I've been recommending as a way to gradually migrate is to set up
an independent Nova installation (region) that uses Neutron.
Two Nova deployments:
Region A (nova-network)
Region B (Neutron)
- A and B share all other services (Glance, Keystone, Cinder, etc).
- Start new instances in region B
- For instances that need to be migrated, snapshot them into Glance
from region A, and then boot them again in region B.
The downside of course is that you have 2 separate API endpoints for the
compute API for the time of the migration. It's something that we are
at least confident will work.
More information about the OpenStack-operators