How to run the perf profiler in RabbitMQ on K8s

[Rilind Harizaj]

Hello,

I am looking into how I could possibly run the perf profiler tool to analyze the CPU usage of a RabbitMQ node under high loads.

Unfortunately, while I have followed the tutorial here:

https://www.rabbitmq.com/blog/2022/05/31/flame-graphs

I have not been able to successfully launch the perf command with the process ID of RabbitMQ in the containerized environment.

More concretely, running the 'perf record' command produces no output. 'perf stat' however, does.

I am conscious of the fact that this might be a question better directed at the Q&A of the perf tool, I was however wondering if someone had this specific challenge in the context of RabbitMQ.

Is there another official way of doing this that I might be overlooking?

Has someone solved an issue similar to this before?


Thank you for your time.

Best

Rilind

[Michal Kuratczyk]

On Kubernetes, I've been using https://www.parca.dev/, which is a continuous profiler.

On the RabbitMQ side, you just need "+JPperf true" (or "map"): https://github.com/parca-dev/parca-demo/tree/main/erlang

I'm sure it's possible to use perf on a Kubernetes node, but I haven't tried.

--
You received this message because you are subscribed to the Google Groups "rabbitmq-users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to rabbitmq-user...@googlegroups.com.
To view this discussion visit https://groups.google.com/d/msgid/rabbitmq-users/b9e7385d-8191-4e8b-9725-7bbc641ab50en%40googlegroups.com.

--

Michal

RabbitMQ Team

[Rilind Harizaj]

Thank you for your answer. I will look into how I can integrate parca.dev in my current setup.

Best