HTTP API: `consumer_utilisation` can be excluded from queue metrics if it hasn't been emitted yet (is undefined)

[chengjoey]

Describe the bug

Our program will request the queue information from the management's /api/queues, but we sometimes find that the consumer_utilisation field is missing from the queue information.

Reproduction steps

It is difficult to reproduce, please ask under what circumstances the consumer_utilisation field may be lost.

Expected behavior

consumer_utilisation could get from /ap/queues

Additional context

No response

[chengjoey]

I tried to directly curl the rabbitmq management's /api/queues interface and found that the consumer_utilisation indeed disappeared in some queue information.

RabbitMQ server version: 3.9.13/3.9.28

Can I add a judgment in rabbit_amqqueue_process.erl, if there is no consumer_utilisation, set it to 0?

[michaelklishin]

@chengjoey thank you for the PR. The change is reasonable but I cannot help but mention that RabbitMQ 3.9 has reached EOL so long ago that it is not even listed on the release series support page.

[michaelklishin]

@chengjoey also, using GET /api/queues without pagination is highly recommended against, it produces 70 or so metrics per queue and once you get into thousands of queues, the JSON responses begin burning a lot of CPU and network bandwidth resources.

Use Prometheus and Grafana instead, at least this way every node reports its own metrics and not 70+ metrics for every queue across the entire cluster.

[chengjoey]

@michaelklishin thanks!

We have a plan to upgrade our rabbitmq server version, currently we are using the classic queue, and we are working on a plan to upgrade to the Quorum queue.

Additionally, I have already passed the cla certification for cluster_operator, may I also sign the cla for rabbitmq-server so that #14552 can be merged into main, at least to keep the fields in the queue info from being missing without modifying our program code.

[chengjoey]

CLA has been signed

rabbitmq/cluster-operator#1954

[michaelklishin]

@chengjoey I conclude that there is nothing to fix in main, v4.1.x. It's a 3.9-specific behavior.

Thank you for taking the time to contribute and pointing out that you have signed the CLA.