Hi JB,

Thanks for the clarification, it works for me as described now.

However, the problem I wanted to address was that if two different MBeans
would contain the exact same property, I see no way of differentiating
between them. In fact I think only one of the properties will be shown by
the Prometheus servlet.

At the moment that's not a real problem for me, just wanted to let you know
and suggest the possibility of a prefix eg.

java_lang_Memory_HeapMemoryUsage_committed
java_lang_Memory_HeapMemoryUsage_init
java_lang_Memory_HeapMemoryUsage_max
java_lang_Memory_HeapMemoryUsage_used

Best regards,
Steven

On Sun, Sep 19, 2021 at 11:30 AM Jean-Baptiste Onofre <[email protected]>
wrote:

> Hi Steven,
>
> I’m about to write a quick blog post to give some highlights about this
> change.
>
> Basically, I did these two other changes:
>
> 1. For « descendent » properties (the properties inside a Map for
> instance), the property is now named [MAP_PROPERTY]_[ENTRY_PROPERTY]:
>
>
> https://github.com/apache/karaf-decanter/blob/main/appender/prometheus/src/main/java/org/apache/karaf/decanter/appender/prometheus/PrometheusServlet.java#L74
>
> 2. You can specify the property you want to « render » in the Prometheus
> servlet using prometheus.key.foo in
> etc/org.apache.karaf.decanter.appender.prometheus.cfg:
>
>
> https://github.com/apache/karaf-decanter/blob/main/appender/prometheus/src/main/java/org/apache/karaf/decanter/appender/prometheus/PrometheusServlet.java#L69
>
> I hope it helps.
>
> Regards
> JB
>
> > Le 19 sept. 2021 à 10:16, Steven Huypens <[email protected]> a
> écrit :
> >
> > Hi JB,
> >
> > I've tested the new 2.8.0 decanter release and I'm pleased to say your
> fix for https://issues.apache.org/jira/browse/KARAF-7154 works great.
> Thanks for that !
> >
> > However I couldn't figure out from the pull request what changes you
> made related to my second suggestion, to better identify from which MBean
> the properties are (or as you put it 'allow user to define the "exported"
> event properties by configuration'). Can you please explain how I can
> achieve this ?
> >
> > Kind regards,
> > Steven
> >
> > On Wed, May 12, 2021 at 6:57 PM Steven Huypens <[email protected]>
> wrote:
> > Hi Jean-Baptist,
> >
> > 1) You mention the Prometheus appender only exposes the numeric metrics.
> I believe it would be a minor but very useful addition to also expose the
> Objects in a CompositeDataSupport. For example java.lang.memory has a
> HeapMemoryUsage-object which contains 4 values (committed, init, max &
> used) that could easily be exposed as well.
> >
> > 2) I also would like to suggest to prefix the outputted name of a
> property with something that really identifies the MBean, eg. :
> >
> > java_lang_Memory_HeapMemoryUsage_committed
> > java_lang_Memory_HeapMemoryUsage_init
> > java_lang_Memory_HeapMemoryUsage_max
> > java_lang_Memory_HeapMemoryUsage_used
> >
> > Currently MBeans having the same properties will have their values
> overridden in the output.
> >
> > Kind regards,
> > Steven
> >
> > On Mon, May 3, 2021 at 6:14 AM Jean-Baptiste Onofre <[email protected]>
> wrote:
> > Hi Daniel,
> >
> > JMX collector polls all MBeans attributes. However Prometheus appender
> only expose metrics (numeric) on the Prometheus servlet:
> >
> > http://localhost:8181/decanter/prometheus
> >
> > As the generated JMX JSON is "more" than just numeric, it’s possible
> that you don’t have the metrics.
> >
> > You can check the JMX JSON using another kind of appender (like log
> appender or elasticsearch).
> > I can add kind of "json introspection" on the Prometheus appender to
> "force" some JSON fields as metrics (gauge).
> >
> > Regards
> > JB
> >
> > > Le 2 mai 2021 à 22:24, Daniel Las <[email protected]> a écrit :
> > >
> > > Hi,
> > >
> > > I installed Decanter 2.7.0 on Karaf 4.2.11 with JMX collector and
> Prometheus appender features. I uncommented
> "object.name.system=java.lang:*" in
> org.apache.karaf.decanter.collector.jmx-local.cfg.
> > >
> > > Where can I find JVM metrics like current heap memory usage?
> > >
> > > Regards
> > > --
> > > Daniel Łaś
> > >
> >
>
>

Reply via email to