[Feature Request] add Implementation Notes to scs-0214-v2 (as Supplement)

[mbuechse]

Let people know how to implement it... I'm not sure what can reasonably be said, maybe just outline how we do it for KaaS v1 and KaaS v2? Of course, this issue has to wait until we know ourselves.

[tsmado]

The implementation notes should describe the correct solution, how to deploy a cluster compliant with the standard v2 and compliance with the k8s-node-distribution-check.py tests. The inspiration how to achieve the node distribution and mainly the tag for host id on the worker nodes is being tested based on the solution used in cluster-stack repository, where the node labeling is done using ansible ( https://github.com/SovereignCloudStack/cluster-stacks/blob/main/playbooks/tasks/label_nodes.yaml ). I am currently testing it in gx-scs with a cluster spawned by capi/capo, when the cluster will configured for all tests to pass, the steps to achieve such cluster config state will be added to the implementation notes of the standard.

[tsmado]

The solution to have correct implementation guidelines is still to be written, since I faced several issues. I tried 2 different approaches:

• Using the Kubernetes Cluster API Provider OpenStack and the corresponding guide. After some troubleshooting, I was able to spawn a control-plane in gx-scs. The problem faced was, that none of the workers was able to start and after some digging I found the possible issue in the logs of a pod in capi-system:
  “Reconciler error” err=“failed to get client: failed to create cluster accessor: error creating http client and mapper for remote cluster "default/ts-testcluster": error creating client for remote cluster "default/ts-testcluster": cluster is not reachable: client rate limiter Wait returned an error: context deadline exceeded - error from a previous attempt: EOF” controller=“machine” controllerGroup=“cluster.x-k8s.io” controllerKind=“Machine” Machine=“default/ts-testcluster-control-plane-84qxt” namespace=“default” name=“ts-testcluster-control-plane-84qxt”

• Using cluster-stacks and also going by the corresponding guide I managed to get similar results, again being able to spawn the cluster-control-plane but with no success regarding the worker nodes. Also the error logs were of the same kind as above. The control plane instance that was starded reported this issue:
  No valid host was found. There are not enough hosts available.

Other step that we tried was to reduce the used flavor of instances from SCS-2V-4-50 to SCS-1V-4-10 but this didn't bring any improvements on the upper mentioned issue. Next steps will be to try spawning the development cluster under a clear project in the gx-scs, because we suspect there might be some benefit for using a project without any other resources created.