HAPRoxy routers experienced out of memory events causing the nodes to not start again

Issue

• We had an outage when our HAProxy received an oom-killer event from the kernel. This occurred during a high memory event on one of the master nodes, where it was unresponsive. The Load balancer for the masters had not taken it out of the pool so that any pods attempting to start would get hung when the node tried to communicate with the master VIP.