In our non-cloud deployment, we have an F5 load balancer in front of two or more Apache servers running mod_proxy. The single HAProxy implementation for a scaled application represents a single point of failure which we would like to resolve. Does Red Hat have a suggested way of eliminating the single point of failure? Options I've thought of (but not had time to test): a.) allowing multicast broadcast to mod_cluster. b.) tracking the gears by following the HAProxy config file. c.) tracking the gears by subscribing to some event in ActiveMQ. I'm open to discussions are ideas here.
This was addressed in the 2.0 release with External Load Balancing: https://access.redhat.com/site/documentation/en-US/OpenShift_Enterprise/2/html-single/Deployment_Guide/index.html#sect-External_Load_Balancing If you feel there is something else required please let us know.