ACNP IngressPolicy not working for Flexible IPAM pods

[KMAnju-2021]

Describe the bug

acnp ingress policy not working only for one case in kind cluster:
when src Pod is IPAM-pod without vlan and destination pod is IPAM-pod with vlan (11,12).

To Reproduce

1. create kind cluster
2. apply antrea.yml with enableIPAM feature.
3. apply ingress acnp

apiVersion: crd.antrea.io/v1beta1
kind: ClusterNetworkPolicy
metadata:
name: acnp-with-drop-a
spec:
priority: 1
appliedTo:
- podSelector:
matchLabels:
pod: a
ingress:
- action: Drop
from:
- podSelector: {}
enableLogging: true

4. Run TestAntreaIPAMAntreaPolicy/TestGroupNoK8sNP/Case=ACNPIngressDrop test

Expected

destination pod should drop the packets

Actual behavior

kubectl exec -it antrea-ipam-testa-cf59ccc97-bq5w9 -n antrea-ipam-test -- sh
Defaulted container "c80" out of: c80, c81, c8080, c8081, c8082, c8083, c8084, c8085
/ # curl 192.168.242.100:81
curl: (1) Received HTTP/0.9 when not allowed

/ # /agnhost connect 192.168.242.100:81 --timeout=1s --protocol=tcp && echo "CONNECTED"
CONNECTED
/ # /agnhost connect 192.168.241.100:81 --timeout=1s --protocol=tcp && echo "CONNECTED"
CONNECTED

[antoninbas]

Thanks for reporting this. It's not always the same subtest failing as far as I can tell, but I keep experiencing this issue with the Jenkins job.

[rajnkamr]

It seems to be consistently failing specially for destination pod as vlan IPAM pod with src pod selector configuration.