Description of problem:

The chk_default_ingress script used by keepalived to determine where the ingress VIP should be located is perma-failing at least back to 4.17. In clusters with sharded ingress, this can cause the VIP to be placed on a node hosting a shard and not the default ingress, which breaks access to core ingress services.

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

Initially reported against 4.17.23, but also present on 4.19.

How reproducible:

Always

Steps to Reproduce:

    1. Deploy on-prem IPI cluster
    2. Check keepalived logs on the node holding the ingress VIP
    3.
    

Actual results:

    chk_default_ingress check failing

Expected results:

chk_default_ingress check passing    

Additional info:

    It looks like this is a permission problem. The chroot into /host is failing.