✨ VirtualMachineReplicaSet CRUD/Scale Up Scale Down support #528

aruneshpa · 2024-05-21T00:30:16Z

What does this PR do, and why is it needed?

This commit adds support for creating, deleting and scaling a VirtualMachineReplicaSet resource. When a VirtualMachineReplicaSet is created, it is reconciled to create the desired number of virtual machine replicas specified in its spec.

Similarly, during scale out/scale in, the controller reconciles the replicas to match the desired state by either creating or deleting VirtualMachine objects. Note that a new replica is only created when the controller detect that it requires a new replica (as opposed to a rollout that is triggered via modification of the template spec in a Deployment).

When the VirtualMachineReplicaSet resource is deleted, all of its owned virtual machine replicas are also cleaned up. The controller relies on the Kubernetes garbage collection mechanism for this.

This change also adds support for deletion policies which can used used to determine how the controller picks a virtual machine to delete from the list of replicas that it controls during a scale in/down. The only supported policy for now is "Random" where all VMs that are not in deleting state get equal weight and the controller can pick any one of them randomly. If a VM is already marked for deletion, that gets the highest deletion priority.

In several places, I have added a TODO to revisit since ReplicaSets offer an ability to signal when an application can be considered "available". That is dependent on a "Ready" condition that is exposed by the replicas. Since we don't have a consistent way to do that, I have removed the support for indicating availability of a VM replica set in the Status. Consequently, the MinReadySeconds field from the spec is also removed since that is used to signal availability. These can be added in a followup.

Testing Done:

Create a VirtualMachineReplicaSet
Get the resource to ensure all replicas ready

k get vmreplicaset -n parunesh-ns test-rs
NAME      REPLICAS   READY   AGE
test-rs   3          3       3m57s

Scale down the replica set

k scale vmreplicaset -n parunesh-ns test-rs --replicas=1
virtualmachinereplicaset.vmoperator.vmware.com/test-rs scaled

Ensure scale down has cleaned up replicas:

k get vmreplicaset -n parunesh-ns test-rs  -o wide
NAME      REPLICAS   READY   AGE
test-rs   1          1       5m15s

k get vm -n parunesh-ns
NAME            POWER-STATE   AGE
test-rs-zmdmm   PoweredOn     3m38s

Scale up the replica set

k scale vmreplicaset -n parunesh-ns test-rs --replicas=2
virtualmachinereplicaset.vmoperator.vmware.com/test-rs scaled

Ensure scaled up has created new replicas

k get vmreplicaset -n parunesh-ns test-rs  -o wide
NAME      REPLICAS   READY   AGE
test-rs   2          2       7m1s

k get vm -n parunesh-ns
NAME            POWER-STATE   AGE
test-rs-r9slq   PoweredOn     44s
test-rs-zmdmm   PoweredOn     4m57s

Delete the replica set and ensure all replicas are deleted.

k delete vmreplicaset -n parunesh-ns test-rs
virtualmachinereplicaset.vmoperator.vmware.com "test-rs" deleted

k get vm -n parunesh-ns
No resources found in parunesh-ns namespace.

Are there any special notes for your reviewer:

None.

Please add a release note if necessary:

Add support for VirtualMachineReplicaSet CRUD and Scale up/down