[Suggestion] Expose disk footprint of each queue

[the-mikedavis]

RabbitMQ series

4.2.x

Operating system (distribution) used

macOS

How is RabbitMQ deployed?

Other

What would you like to suggest for a future version of RabbitMQ?

Currently it's tough to tell which queue is taking up disk space and that makes it hard to know which queue(s) you should prioritize consuming/purging/deleting if you're in disk alarm. There are metrics for message bytes but message-bytes doesn't necessarily map well to the disk footprint of the queue. For quorum queues for example, you might have a message checked out for a long time which is preventing snapshotting+truncation of later consumed messages. If that queue is published/consumed at fairly high throughput then the disk footprint can grow large despite the message-bytes metric being small.

Ideally there should be a way to determine the disk footprint per queue. I'm not sure if this is necessary or straightforward for all queue types but I think it would be especially useful for quorum queues.

I'm not sure how this should look. A simple way could be a plugin that adds a management API page which lists queues and their disk footprints. That could periodically run du on the queue directory (for QQs and streams). Or we could have queue types track the number of bytes they write to and delete from disk. But that seems like a more invasive change and you'd have to sweep for files on boot-up to get the current sizes.