Prometheus per-object or not per-object

[miksir]

Before all this happened it was standalone rabbitmq and rabbit-exporter.
But cluster coming and we decided to start to use rabbitmq metrics endpoint aka prometheus plugin.
And now I got dilemma. We need metrics per queue. Its really helps to find growing up queue fast, its helps to find queues without consumers. So aggregate mode not for us - there is no queue metrics. Per object mode.... well, lets skip the point of the amount of scraped metrics, it possible to fix with prometheus relabel. But scrape time for us about 25-30 seconds, it's really hard to use especially if we want to scrape each 30 sec.
So, what to do...
Probably switch to aggregate mode back and add default prometheus rabbitmq exporter, but I cant understand how good this exporter for cluster.
Any suggestions?

[daichi703n]

Hi, @miksir
I have the same issue. Did you solved this? Or did you revert to the old exporter?

[daichi703n]

Thank you. I would consider using both as well as you.

[mkuratczyk]

What RabbitMQ version are you using? I believe there are significant improvements made around November last year. Also, since 3.9.8, there is a configurable endpoint which allows to select only some of the per-object metrics: #3520

[daichi703n]

rabbitmq-server-3.9.11-1.el7.noarch.

rabbitmq/cluster-operator#865

• QueueHasNoConsumers
• QueueIsGrowing

These alerts are really what I want to do!

How can I enable detailed metrics? (not yet read docs...)

[mkuratczyk]

With that version, you should just have them (/metrics/detailed endpoint) and you already found the alerts you need that use that endpoint

[daichi703n]

Thanks. I successfully got each detailed metrics with query (/metrics/detailed?family=xxx) described here: https://github.com/rabbitmq/rabbitmq-server/tree/master/deps/rabbitmq_prometheus#selective-querying-of-per-object-metrics .