Ability to filter tags out of an attribute?

[osheroff]

Hi,
I'm wiring up jmxfetch to a kafka producer JVM. It includes a "client-id" tag(?) that I'd like to filter out, as historically large number of unique tags have been problematic to our datadog install. Is this possible with jmxfetch?

The log line when it discovers the metric:

2016-06-13 21:55:08,313 | DEBUG| Instance | Getting attributes for bean: kafka.producer:type=producer-node-metrics,client-id=9388b6bb-539b-4bc5-8d60-14be420574d6,node-id=node-8
2016-06-13 21:55:08,319 | DEBUG| Instance | Attribute: kafka.producer:type=producer-node-metrics,client-id=9388b6bb-539b-4bc5-8d60-14be420574d6,node-id=node-8 : javax.management.MBeanAttributeInfo[description=The average number of requests sent per second., name=request-rate, type=double, read-only, descriptor={}] has attributeInfo simple type
2016-06-13 21:55:08,322 | DEBUG| Instance | Attribute: kafka.producer:type=producer-node-metrics,client-id=9388b6bb-539b-4bc5-8d60-14be420574d6,node-id=node-8 : javax.management.MBeanAttributeInfo[description=The average size of all requests in the window.., name=request-size-avg, type=double, read-only, descriptor={}] has attributeInfo simple type
2016-06-13 21:55:08,326 | DEBUG| Instance | Attribute: kafka.producer:type=producer-node-metrics,client-id=9388b6bb-539b-4bc5-8d60-14be420574d6,node-id=node-8 : javax.management.MBeanAttributeInfo[description=, name=incoming-byte-rate, type=double, read-only, descriptor={}] has attributeInfo simple type

[yannmh]

Thanks for reporting the issue and providing a use case @osheroff .

I am afraid this is not something JMXFetch can do right now. I am adding a note to our triage and will follow back to you: I'll believe we should work on such a feature for the next release.

[yannmh]

@osheroff we are going to move forward and add support for tag filtering with Datadog Agent 5.9.

To prevent multiple MBeans from being submitted with the same metric names and tags (resulting from the removal of an attribute's tag) and overriding each other, the corresponding metrics would be submitted with the histogram metric type, i.e.

kafka.producer.$name.avg
kafka.producer.$name.min
kafka.producer.$name.max
kafka.producer.$name.med
kafka.producer.$name.count