Is it possible to Ingest druid itself it's own metrics?

Is it possible to Ingest druid itself it’s own metrics?
Druid monitoring is not simple and have many works to do, I think.

So I imagine how about druid can ingest it’s metrics itself by default.

For example elasticsearch have good monitoring system based on elasticsearch itself.

Which architecture can be considered?

I am a beginner of druid, but how about

  • HttpEmitter + Tranquility server

(or even tranquility server can be skipped if emitter can write segments?)

I don’t want to discuss the data format in detail here, but just key-value data format can be an idea

column - value

It is possible (at Imply, we have a metrics analysis product called “Clarity” that uses Druid to ingest Druid metrics).

I would recommend using a separate cluster for the metrics collection vs. the main cluster so that they don’t impact each other performance-wise.

  • HttpEmitter + Tranquility server

This will work.

If you’re concerned about not losing metrics data due to outages/failures, feeding the metrics data into Kafka first and using the Kafka Indexing Service (http://druid.io/docs/latest/development/extensions-core/kafka-ingestion) would be a better approach.

Thanks,

Jon