Controlled removal of hc-demo after creation (TC-HCP-002) and scaling (TC-HCP-003) are complete. Deliberately a separate test case so destroy is not run accidentally too early, and verification of cleanup completeness (namespace gone, VMs gone, pod capacity freed again) is tested as a standalone step — equally important in practice as creation, since leftover remnants directly block pod capacity for the next Hosted Cluster.
Prerequisites checklist
Confirm these before running steps:
1Remove Hosted Cluster
hcp destroy cluster kubevirt --name hc-demo --namespace clustersDestroy often takes 5–15 minutes; occasionally longer if finalisers stick — keep waiting until the HostedCluster object disappears
2Observe until everything is truly gone
watch oc get hostedcluster -n clusters→ object disappears completely from the list, then Ctrl+C
3Verify completeness — worker VMs, namespace
oc get vm -n clusters-hc-demooc get ns clusters-hc-demo→ both 'NotFound' = cleanly removed
4Check MetalLB pool state — pool itself remains, directly reusable for the next Hosted Cluster
oc get ipaddresspool external-pool -n metallb-systemoc get svc -A | grep <external-pool-ip> # or: oc get ipaddresspool external-pool -n metallb-system -o yaml→ no matches = IP free again
5If it hangs (finaliser issue, occurs occasionally when a worker VM sticks during deletion)
oc get hostedcluster hc-demo -n clusters -o jsonpath='{.metadata.finalizers}'oc patch hostedcluster hc-demo -n clusters --type=merge -p '{"metadata":{"finalizers":[]}}'6If multiple Hosted Cluster remnants are suspected in general (e.g. from earlier failed attempts before the SingleReplica fix)
oc get hostedcluster -Aoc get ns | grep clusters-→ remove each no-longer-needed entry likewise with 'hcp destroy cluster kubevirt --name <name> --namespace <namespace>'