Details
-
Bug
-
Resolution: Test Pending
-
Critical
-
None
-
4.14.0
-
None
-
No
-
5
-
NHE Sprint 239, NHE Sprint 240, NHE Sprint 242, NHE Sprint 243, NHE Sprint 246
-
5
-
False
-
Description
Description of problem:
As of this change to Ovn-Kubernetes (https://github.com/openshift/ovn-kubernetes/commit/63f398bebbed604c74f9a0010c74849d955074c9) node port service traffic fails when run in a DPU mode 2-cluster set up. This is due to the switch from using the l3gatewayAddresses to hostAddresses here: https://github.com/ovn-org/ovn-kubernetes/pull/3557/files#diff-b61468c0028d28eb0e1060d24f53f28b6482acfbfd0cf5c15f9c23c3fd8c9f91R469 In a 2-cluster design, hostAddresses will be the address of the bluefield node running on behalf of the tenant, rather than the tenant node IP which results in this failure.
Version-Release number of selected component (if applicable):
How reproducible:
Everytime
Steps to Reproduce:
1. Deploy a cluster with BF-2 in DPU mode 2. Add a new node port Service to the cluster 3. Try to use the node port service to pass traffic in the tenant cluster (x86)
Actual results:
TCP-05-b-pod2nodePortSvc-podBackend-diffNode.txt 0 bits/sec 0 No Packets Detected On Client Or Server VF Reps Fail 0 bits/sec 0 No Packets Detected On Client Or Server VF Reps Fail
Expected results:
TCP-05-b-pod2nodePortSvc-podBackend-diffNode.txt 3.55Gbits/sec 0 No Packets Detected On Client Or Server VF Reps Pass 9.33Gbits/sec 0 No Packets Detected On Client Or Server VF Reps Pass
Additional info:
