Deleting a pod using sriov vf triggers a node drain

    v4.14.0-202409031910

    The issue was reproduced by the client 3 times

    1. Delete a pod that is using sriov vf 
    2. Watch until node is drained by the operator

    

I can see the port getting released from OVN:
$ oc logs -n openshift-ovn-kubernetes ovnkube-node-6cvqj -c ovn-controller
2024-09-23T10:12:10.790289733Z 2024-09-23T10:12:10.790Z|01622|binding|INFO|Releasing lport duta_lmg-statefulset-set-one-2 from this chassis (sb_readonly=0)
2024-09-23T10:12:10.790289733Z 2024-09-23T10:12:10.790Z|01623|binding|INFO|Setting lport duta_lmg-statefulset-set-one-2 down in Southbound
2024-09-23T10:12:12.555510048Z 2024-09-23T10:12:12.555Z|01624|binding|INFO|Claiming lport duta_lmg-statefulset-set-one-2 for this chassis.
2024-09-23T10:12:12.555510048Z 2024-09-23T10:12:12.555Z|01625|binding|INFO|duta_lmg-statefulset-set-one-2: Claiming 0a:58:0a:81:02:65 10.129.2.101
2024-09-23T10:12:12.567899172Z 2024-09-23T10:12:12.567Z|01626|binding|INFO|Setting lport duta_lmg-statefulset-set-one-2 ovn-installed in OVS
2024-09-23T10:12:12.567899172Z 2024-09-23T10:12:12.567Z|01627|binding|INFO|Setting lport duta_lmg-statefulset-set-one-2 up in Southbound

Then 9 seconds later, we see the node start to Drain per SRIOV side of things with a 'Drain_Required':

2024-09-23T10:12:49.144006714Z 2024-09-23T10:12:49.143927945Z   LEVEL(-2)       controller/controller.go:119    Drain annotations       {"controller": "node", "controllerGroup": "", "controllerKind": "Node", "Node": {"name":"npvmgncs10-allinone-3"}, "namespace": "", "name": "npvmgncs10-allinone-3", "reconcileID": "6580ba2a-6dba-4a62-9128-08009df0d090", "nodeAnnotation": "Drain_Required", "nodeStateAnnotation": "Idle"}

    No node drain when simply deleting a pod