Hi Federico,

Refer to your link(https://bugzilla.redhat.com/show_bug.cgi?id=2051738#c16, https://bugzilla.redhat.com/show_bug.cgi?id=2051738#c20), I'm going to ask to collect below data:

1) OVN DBs(nb_db, sb_db)
  sh-4.4# find / -name "ovnnb_db.db"
  /etc/openvswitch/ovnnb_db.db
  sh-4.4# find / -name "ovnsb_db.db"
  /etc/openvswitch/ovnsb_db.db

2) namespace info(where the namespace dmp and cip pod resides)
  oc adm inspect namespace
  ex) oc adm inspect namespace call3

3) When the issue is occurred please collect below info

 3.1) dmp -> cip Request:
      a) What is the srcIP of the client pod?
      b) What is the dstIP of the dstIP of the server pod which is the backend?
      c) <Important> What is the service IP (Cluster IP) of the service accessed by the pod acting as a cip or client?

 3.2) cip -> dmp Reply:
      a) What is the srcIP of the client pod?
      b) What is the dstIP of the dstIP of the server pod which is the backend?
      c) <Important> What is the service IP (Cluster IP) of the service accessed by the pod acting as a cip or client? 


4) For dmp -> cip and cip -> dmp, please capture tcpdump through the tcpdump -neep option (to see IP/MAC, not domain) in each pod

5) a pcap taken on -i any on the two nodes where the pods live taken DURING the client - svc connection (please make sure that there is only one server pod, if there are multiple, take one for each node where the dmp pods are plus the one where the cip0 pod is)

6) sosreports where dmp and cip reside.

7) check if the issue is reproduced on OCP 4.9 cluster.

If you need more data in addition to above, please let me know.

Thank you

Hakyong Do

Technical Account Manager

Looks good, thanks. Another good point for the next steps would be asking if they are able to reproduce with the two pods on the same node.