Description of problem: Customer is having an outage due to their router pods being unresponsive. Removing health checks allows pods to start up but performing health checks manually (curl http://localhost:1936/healthz) still fail after a long pause with Connection Timeout. Unfortunately, no events or logging output from haproxy. Version-Release number of selected component (if applicable): atomic-openshift-3.3.1.17-1.git.0.b82e86c How reproducible: Happening constantly for customer. Actual results: Unable to utilize pod routes.
I think the iptables errors are a red herring since the router runs with host networking and a service is not being used to access it. What's the cpu utilization of haproxy? Is it under high load?
*** This bug has been marked as a duplicate of bug 1384746 ***