It's the testing scenario of QE test case OCP-24405, i.e. after a successful IPI installation, add an additional compute/worker node without infra_id as name prefix. The expectation is, "destroy cluster" could delete the additional compute/worker machine smoothly. But the testing results is, "destroy cluster" seems unaware of the machine. 

    4.18.0-0.nightly-multi-2024-12-17-192034

    Always

1. install an IPI cluster on GCP and make sure it succeeds (see [1])
2. add the additional compute/worker node, and ensure the node's name doesn't have the cluster infra ID (see [2])
3. wait for the node ready and all cluster operators available
4. (optional) scale ingress operator replica into 3 (see [3]), and wait for ingress operator finishing progressing
5. check the new machine on GCP (see [4])
6. "destroy cluster" (see [5])     

    The additional compute/worker node is not deleted, which seems also leading to k8s firewall-rules / forwarding-rule / target-pool / http-health-check not deleted.

    "destroy cluster" should be able to detect the additional compute/worker node by the label "kubernetes-io-cluster-<infra id>: owned" and delete it along with all resources of the cluster.

    Alternatively, we also tested with creating the additional compute/worker machine by a machineset YAML (rather than a machine YAML), and we got the same issue in such case.