Description of problem:

The customer has the env of OCP3.11 with 3master + 30infra + 150+ nodes (holding 1900+ pods)
They running router pods with replica=30 across the infra nodes.

This issue is : when they create a route, some of the routers cannot sync it to their haproxy.conf. and after delete the router pod, the recreated one will be normal.

From the issued router pod, the log will keep reporting below errors:
E1228 11:57:38.516286       1 router_controller.go:250] another route has claimed this host


Concern from the customer:

1. They need detailed info about how router sync routes from api and they doubt the RCA is router doesn't treat api as the default input.For example, api "do" update new route to routers but router refuses to update it.

2. Does it make sense to just using replica=30 to run HA routers in the cluster ?


For sosreport, because of the limit of bugzilla , please download it from case 02282929 #5 with password:redhat


Thank you.