kubernetes-sigs / cluster-api-provider-kubevirt / 30525865769

40%

main: 40%

Build Type

Pull #389

github

Committed by orenc1

Commit Message

OCPBUGS-98218: Fix guest node remaining cordoned after VMI eviction and restart

When CAPK handles a VMI eviction (evictionStrategy=External), it cordons
and drains the guest node before deleting the VMI. After KubeVirt recreates
the VMI on another host, the guest node reconnects but stays cordoned
(spec.unschedulable=true) indefinitely because nothing uncordons it.

Track the cordon with a new annotation (capk.cluster.x-k8s.io/node-cordoned)
on the KubevirtMachine, and uncordon the guest node in DrainNodeIfNeeded
when no eviction is in progress but the annotation is present.

Fixes: OCPBUGS-98218

Signed-off-by: Oren Cohen <ocohen@redhat.com>
Assisted-by: Claude <noreply@anthropic.com>

Pull Request Pull Request #389: OCPBUGS-98218: Fix guest node remaining cordoned after VMI eviction and restart

Coverage Stats

48 of 57 new or added lines in 1 file covered. (84.21%)

70 existing lines in 2 files now uncovered.

1169 of 2947 relevant lines covered (39.67%)

2.24 hits per line