We need something to figure out where those endpoints are available
from,
gather the metadata about the containers which are exposing those endpoints,
convert the metrics from one type into something Hawkular Metrics can
understand, and then write those metrics (with metadata based tags) into
Hawkular Metrics.
I think this needs to be part of a separate component. There is going to be a
lot of code here dealing specifically with the OpenShift environment.
If we are reading from a jolokia endpoint and we have an agent which knows
how to store jolokia style metrics in Hawkular-Metrics, then we should of
course try and reuse components.
OK, I'll look more into the Prometheus protocol. That's the part I was missing. It
sounds like essentially we are going to be pulling metrics directly from the components
emitting the data (rather than pulling metrics from the Prometheus server itself which is
what I originally thought). Now that I think about it, really this sounds just like
JMX/Jolokia endpoints or WildFly's remote DMR interface). I just have to learn more
about what these Prometheus endpoints look like and what the data looks like over that
protocol.