Hi,
On 2016-03-24 03:06, Nathan Scott wrote:
>
> And, just like pmdajmx.pl would, there will need to be some (config file)
> updates to build in knowledge of semantics/units/text for many of the JVM
> metrics (there's 20-ish sample jvm metrics are there), Wildfly & Cassandra
> metrics, and so on.
I think you've misunderstood one important aspect of pmdajmx. It does
not *require* per-app/per-version config files. It works like some of
the existing PMDAs (e.g., KVM, Samba) in that it initially iterates the
available metrics and then goes on adding them to PMNS. Whatever is
available over JMX, pmdajmx makes it available completely transparently.
It is, however, *possible* to create mappings to provide more precise
metric information than what is used by default if wanted. I think for
something fundamental like the OpenJDK JVM (~300 metrics) that might be
a reasonable goal. But creating such files for the above mentioned
WildFly and Cassandra alone (they already provide over 20k metrics) is
not going to happen anytime soon, if ever. And in some cases (e.g.,
WildFly) it would not even be possible to cover everything as the
configuration changes (and thus available metrics) are reflected over
JMX so what applies to one WildFly instance isn't valid with the next one.
Thanks,
--
Marko Myllynen
|