Whenever one runs ovnkube-trace from an in-cluster pod to a pod in the host network that is in different node, the following spurious error appears despite of the underlying ovn-trace being correct:

ovn-trace indicates failure from ingress-canary-7zhxs to router-default-6758fb465c-s66rv - output to "k8s-worker-0.example.redhat.com" not matched

This is caused because as per[1], if the destination pod is in host network, the outport is expected to be of the form "k8s-${NODE_NAME}", which is true only if either in local gateway or if the source pod is in the same node than the destination pod.

This is already fixed in the master branch[2], but we would need this to be backported to previous releases. 

4.11.4

Always

1. ovnkube-trace from pod in the SDN to pod in host network
2.
3.

Wrong error

No wrong error

References:
[1] - https://github.com/openshift/ovn-kubernetes/blob/release-4.11/go-controller/cmd/ovnkube-trace/ovnkube-trace.go#L771-L777
[2] - https://github.com/openshift/ovn-kubernetes/blob/master/go-controller/cmd/ovnkube-trace/ovnkube-trace.go#L755-L769