Hi there,

I'm trying to understand the way metrics and monitoring works with Apache
Storm.

My requirements are that I wan't to heartbeat a topology and the storm
daemons to an external processes (for visibility and support purposes) and
gather metrics and collect them in one of the many metrics tool chains that
exist now.

For the heartbeating the storm components I believe I can do this via the
javaagent approach and injecting this via the childopts configuration
(assuming that these arguments are passed before the -jar argument). Is
that correct?

However I'm not sure if I can do a similar approach for topology and their
workers, as I would need to ensure that the heartbeats for topology/workers
would be unique and I'm not sure if that would be possible. Has anyone done
anything similar and can provide guidance on how I could that?

>From a metrics perspective I'm struggling to understand how the v2 metrics,
cluster metrics and metrics consumers and metrics reporters and the
metricstore and the metrics reported in the UI all tie together. Would
anybody be able to provide some guidance as I've read the docs and various
blogs (which all seem to refer to pre 1.2. metrics) and I'm none the wiser

Many Thanks

Edward

Reply via email to