Description of problem:

Deploying BGP configuration. Require External access to CUDN VMs with VRF-Lite configuration - unable to successfully route packets back to the VMs. Traffic is lost in the stack. FRR pods crashloop, unable to map to VRF.

Version-Release number of selected component (if applicable):

OCP 4.20

How reproducible:

Ongoing, single cluster impacted. No replicator internally (hardware limitations).

Steps to Reproduce:

1. See config details in first comment below with specific yaml configurations for replicator/specific configs.

2.

3.

Actual results:

FRR pods crashloop (coredump on zebra process handling - nexthop considered invalid)

BGP advertisments not seen, VMs cannot reach gateway or be routed back.

Expected results:

Platform configurations should allow throughput in this configuration.

Additional info:

• See first comment for additional summary contexts, resources, customer-links and case details for tracking of issue.