This issue handles exposing JMX metrics through Jolokia. We may need another issue for strategies on persisting those metrics.

PR: https://github.com/OpenNMS/opennms/pull/2934

Just Some Thoughts,

In regards to OpenShift, pod names are generally dynamic and non-persistent. Configuraton and Image updates are one reason pod names change. Another is by scaling the pod count up and down, possibly with the help of an auto-scaler function inside OpenShift

If the goal is to persist these metrics inside OpenNMS then a bridge/translator would be needed. It is possible to create scripts with a persistent storage device that would cause the individual containers to bind to a "persistent" Sentinel ID. It would not be possible to bind these IDs to the same pod name due to the dynamic nature of the pod names changing over time. OpenNMS would then need to discover all the pods and then walk them to record the metrics. The complication here is defining/discovering a method to use to "proxy" to each individual container.

Secondary to this polling method, another way would be for Sentinel to "push" metrics into OpenNMS via the existing Kafka channel for its own metrics using this persisted Sentinel ID.