Messages delayed count on a cluster rabbitmq setup

[mdmourao]

Describe the bug

Hi! I am unsure if this behavior is an issue or a known limitation.

While using the delayed message exchange, we noticed the 'messages delayed' metric often reports 0 in the Management UI when it should be positive.

I originally commented on an existing issue thinking it was related. However, I performed a test by opening a connection directly to the specific node responsible for those messages. On that specific node, the count was reported correctly, suggesting the issue lies in how the Management UI aggregates or displays this data.

Reproduction steps

1. Setup 3 nodes (we are using RabbitMQ Cluster Kubernetes Operator)
2. Create exchange x-delayed-message
3. Publish N messages with x-delay header
4. Open connection on 2 nodes for example and check the count

Check first node count:

Second node count:

Expected behavior

I would expect to see the value of all messages, not just the messages stored on that node.

Additional context

No response

[mdmourao]

Should be related with: https://github.com/rabbitmq/rabbitmq-server not a exchange problem

It looks like it does not aggregate custom exchange fields from nodes

[michaelklishin]

@mdmourao this README of this plugin does explain that it does not replicate its state.

Now, metrics can be aggregated independently of that but the README also states that Team RabbitMQ no longer maintains this plugin beyond new builds for new RabbitMQ series. If you want metric aggregation, you will have to contribute it.

[mdmourao]

Thanks for the feedback, i did not expect to have the state replicated on all nodes. But i would expect the rabbit mq management UI to aggregate the metrics from all nodes.
I will try to contribute.