TripleO provides the ability to manage changes over time to a cloud that it has deployed. However, it cannot automatically take over the management of existing OpenStack clouds deployed with another installer. Since there can be no one-size-fits-all procedure for upgrading an existing cloud to use TripleO, it is recommended that a new cloud be deployed with TripleO and any workloads running on an existing cloud be migrated off.
Since the best way of avoiding or handling any downtime associated with moving an application from one cloud to another is application-dependent, it is preferable to have end users migrate their own applications at a time and in the manner of their choosing. This can also help to spread out the network bandwidth requirements, rather than copying a large number of snapshots in bulk.
Ideally applications can be re-created from first principles (an Orchestration tool such as Heat can help make this repeatable) and any data populated after the fact. This allows the new VMs to be backed by a copy-on-write disk image overlaid on the original base image. The alternative is to export and then import snapshots of the VM images. This may require considerably more disk space as each VM’s base image becomes its snapshot, where previously multiple VMs may have shared the same base image.
As workloads are migrated off the previous cloud, compute node hardware can be freed up to reallocate to the new cloud. Since there is likely no guarantee as to the order in which users will migrate, it will be necessary to consolidate the remaining VMs onto a smaller number of machines as utilization drops. This can be done by performing live migration within the old cloud.
Select a compute node to remove from service and follow the procedure for Quiescing a Compute Node. Once this is done, the node can be removed from the old cloud and the hardware reused, possibly by adding it to the new cloud.
As utilization of the new cloud increases and hardware becomes available from the old cloud, additional compute nodes can be added to the new cloud with TripleO.
First, register and introspect the additional hardware with Ironic just as you would have done when initially deploying the cloud with TripleO. Then scale out the ‘Compute’ role in the new overcloud to start making use of the additional capacity.