Migration questions

asked 2019-05-20 11:52:10 -0500

anonymous user

Anonymous

Hi

I have a question of how to properly migrate instances from an existing Openstack to a parallel cloud.

Existing cloud consists of 3 physical compute nodes, and one of them is having a virtualized controller, plus there are 2 GlusterFS servers providing storage. I have a single physical server available at the moment for a parallel cloud installation. Planning to use TripleO deployment method with 1 virtual controller and 1 virtual compute node, plus director - all within one box.

Planned migration:

  1. Deploy new cloud
  2. Migrate instances from old cloud to new one, freeing one compute node at a time and then attaching it to a new cloud.
  3. Once all compute nodes are done, then re-attach GlusterFS servers to a new Openstack farm. New single box has enough storage capacity to accomodate all curently provisioned space in Gluster.

Things that concern me the most:

  1. According to TripleO deployment guide, it says do not use Virtual servers for production. But after some VMs will be migrated from existing cloud to a new one, I can introduce freed up compute node to an Openstack. Estimated planning run time in a virtual environment is about a week or so with only some of the VMs in it. Is this an allowed scenario, or should be considered as a no-go?
  2. During the migration process of a VM, does attached non-root volume also migrates in a snapshot? Where are these volumes will be placed - default location in cinder.conf? How do I move these volumes later on to a gluster volumes later - edit the cinder.conf with new location after all volumes are copied and restart the service?
  3. Networking part: do I need to recreate same VLANs for the same projects, or it might be totally new network scheme? This is probably app dependent, and the way it works within the project. I assume that it'll be safe to keep an existing VLAN.

Thanks in advance!

edit retag flag offensive close merge delete