-
Bug
-
Resolution: Done
-
Critical
-
None
-
4.12
-
None
-
Proposed
-
False
-
Description of problem:
When using cx-5 as default network for pods with ovs hwoffload enabled, pod to pod traffics (same host or different hosts) are blocked.
Version-Release number of selected component (if applicable):
[root@openshift-qe-026 offload]# oc version Client Version: 4.12.0-0.nightly-2022-10-25-210451 Kustomize Version: v4.5.7 Server Version: 4.12.0-0.nightly-2022-10-25-210451 Kubernetes Version: v1.25.2+4bd0702 [root@openshift-qe-026 offload]# [root@openshift-qe-026 offload]# [root@openshift-qe-026 offload]# oc get csv -n openshift-sriov-network-operator NAME DISPLAY VERSION REPLACES PHASE sriov-network-operator.4.12.0-202211021237 SR-IOV Network Operator 4.12.0-202211021237 sriov-network-operator.4.12.0-202210290517 Succeeded [root@openshift-qe-026 offload]#
How reproducible:
1. set up baremetal cluster with cx-5 as default network interface on worker nodes. 2. configure ovs offload following guide https://docs.openshift.com/container-platform/4.10/networking/hardware_networks/configuring-hardware-offloading.html 3. create pods with ovs hwoffloaded VFs as default network interface 4. ping from pod to pod failed. 5. create pods without ovs hwoffload VF. 6. ping from pod to pod succeed.
Steps to Reproduce:
1. 2. 3.
Actual results:
pod to pod traffic is blocked.
Expected results:
pod to pod traffic should pass.
Additional info:
- depends on
-
OCPBUGS-3369 OvS HWOL fails to pass traffic between pods
- Closed