Support deployments without duplicating clusters

Nobody has claimed this yet.

The current stable version of the operator creates a new cluster every time a new version of an application is deployed, both clusters run for the same time, and when the new application is running successfully the old cluster is removed. That's a really easy way to support rollbacks and having fast deployments.

For applications that require a lot of resources, that way of deployment is not useful because you need to allocate the double of the resources in order to support deploying new versions, resources that will not be used at all afterwards.

I was wondering if a new deployment could be created to reuse the current cluster, knowing that this will produce slower deployments and possibly without automatic rollbacks.

No files, tests, or entry points are identified. Start by tracing the operator's existing deployment flow and how it creates and removes clusters; define the changes needed to reuse the current cluster, including the expected behavior for slower deployments and rollback support.

Written by the indexing model from the issue text.