Scenario 3: One service at a time with parallel Compute

For most services this scenario is identical to scenario 2, with the exception of the compute service. Rather than upgrading your existing compute nodes as part of this process, you deploy new compute nodes running the Juno compute service. You wait for existing workloads on your Icehouse compute nodes to complete (or migrate them by hand), and when a Icehouse compute node is no longer hosting any instances you upgrade the computer service on that node.

  1. Follow the procedure for scenario 2, but stop after completing the Horizon upgrade (do not upgrade Nova).

    You will want to run the final package upgrade on systems that are not running Nova services.

Deploy the new compute environment

  1. Set up a parallel compute environment: install a new Nova controller and compute nodes using the Juno repositories.

    These systems will use a configuration generally identical to that on your Icehouse Nova nodes. They will be making use of the same Keystone, Glance, Cinder, etc. services.

Move instances to the new environment

  1. Migrate instances to the Juno compute nodes

    The simplest method for "migrating" an instance is to simply stop the instance running in your Icehouse environment and deploy a new one on the Juno infrastructure.

    If re-deployment is not an option, you can move instances from your Icehouse compute nodes to your Juno compute nodes with minimal downtime via the following process:

    1. Snapshot the existing instance.
    2. Delete the existing instance.
    3. Boot a new instance on the Juno compute nodes from the snapshot.

    4. Allocate and assign any necessary floating addresses

      In a Neutron environment, you should be able to re-assign your previously allocated addresses as soon as the instance to which they were assigned is shut down.

      In a Nova networking environment, you will need to create new floating ip pools identical to those in your Havana environment and then explicitly allocate the necessary floating ip addresses before assigning them.

      The nova floating-ip-bulk-create command can be used to allocate explicit addresses.

  2. When you are able to move all the instances from one of your existing Icehouse compute nodes, you can redeploy that node in your Juno compute environment.

  3. When you have moved all your compute nodes into the Juno environment, you can retire any remaining Icehouse Nova services.