Fleet does not deploy to K8S cluster (non-harvester) imported from Virtualization Management page

[slickwarren]

Rancher Server Setup

• Rancher version:v2.6-head (62c101a)
• Installation option (Docker install/Helm Chart): docker
  • If Helm Chart, Kubernetes Cluster and version (RKE1, RKE2, k3s, EKS, etc):
• Proxy/Cert Details:n/a

Information about the Cluster

• Kubernetes version:v1.21.3
• Cluster Type (Local/Downstream):Downstream
  • If downstream, what type of cluster? (Custom/Imported or specify provider for Hosted/Infrastructure Provider):imported

    Fleet does not deploy to K8S cluster (non-harvester) imported from Virtualization Management page. Cluster is not listed in fleet -> git repo -> create -> Deploy To section

    To Reproduce

    * create a stand-alone cluster to import into rancher as a downstream cluster (k8s cluster, not a harvester cluster) * go to `virtualization management` in rancher, get the import command for a downstream cluster by clicking `import existing` and giving the cluster a name * import the cluster into rancher * once the cluster is active, go to `continuous delivery` -> git repo -> create -> `Deploy To` * note that the cluster is not listed here * add a test fleet repo, and select the `all clusters in workspace` option **Result** * Fleet does not deploy to K8S cluster (non-harvester) imported from Virtualization Management page. * Cluster is not listed in fleet -> git repo -> create -> `Deploy To` section **Expected Result** * expected cluster to be listed in fleet -> git repo -> create -> `Deploy To` section **Additional context** * this is use case 3 from this set of tests: https://github.com/rancher/rancher/issues/34880 * related issue https://github.com/rancher/rancher/issues/34954

[StrongMonkey]

This is caused by back-population from clusters.provisioning.cattle.io(v1 cluster) to clusters.management.cattle.io(v3 cluster). Any label that is from v1 cluster will be copied to v3 cluster. The cluster is treated as a harvester cluster because of label override. This is what happened:

1. User creates an imported cluster(v1 cluster) from virtualization management. UI adds provider.cattle.io=harvester label to v1 cluster.
2. For back-compatility issues(ember UI) rancher will create v3 cluster from v1 cluster. Once imported process is finished, it will set status.provider and provider labels(rke/k3s/rke2). The provider detection logic only exists in v3 cluster.
3. Controller resync v1 cluster to v3 cluster, overriding label provider.cattle.io to harvester.

To fix this we need to stop step 3, and figuring out a way to back-populate the correct provider label from v3 cluster to v1 cluster.

After discussion with @yasker we can releate-note this in 2.6.1 and push it back to v2.6.2.

[yasker]

Release note for v2.6.1:

• Harvester cluster is not supported for Continous Delivery page. Importing a cluster in Virtualization Management page will result in the cluster won't show up in the Continous Delivery's Cluster List page, including non-Harvester cluster.

CC @btat

[SheilaghM]

@yasker - Does the Harvester team plan to support Harvester Cluster for the Continuous Delivery page in UI in the near future?

[kkaempf]

Removing team/fleet as this is a provisioning / Harvester issue