Add stats for RabbitMQ's Federation plugin when it is available/in use

[jacquesh]

Feature Request

Would you consider adding reporting for the pre-packaged Federation plugin for RabbitMQ?

Use case:

I have two RabbitMQ clusters with similar setups and queues that exist on both clusters have federation links connecting them. I would like to gain some visibility on how often (and to what extent) these federation links get used. Essentially I'm just looking for a rate of message transfer between the two clusters (per queue & federation upstream).

Current behavior:

No federation-related statsare reported

Desired behavior:

Telegraf reports stats for running federation links.

Admittedly I understand that this might be a bit of slippery slope in terms of data collected from RabbitMQ. For example I can imagine being interested in similar information from the Shovel plugin and I'm sure somebody somewhere has other RabbitMQ plugins that they use that they might want to collect metrics for.

[danielnelson]

I think this would be a nice feature, would be interesting to see an example of the output from GET /api/federation-links.

[jacquesh]

The management API for federation doesn't appear to be particularly well documented (that I can see), but I ran a quick test from my machine to a local server and I got this output

I'm happy to contribute the feature, but the main question is what set of fields should be collected?
Just from looking at that output, as a first approximation I would suggest the following:

Tags

• url
• (local) queue
• upstream queue
• (local) exchange
• upstream exchange
• vhost
• upstream

Fields:

• ACKs uncommitted
• Consumers
• Messages unacknowledged
• Messages uncommitted
• Messages unconfirmed
• Messages confirmed (with rate)
• Messages published (with rate)
• Messages returned (with rate)

I'm not sure that it is useful to have all of those though, thoughts?

[danielnelson]

That looks good, the one change I would make is to skip the rate stats as these can be calculated from the raw counters at query time. Anything we leave out can always be added later, in case we decide we need prefetch, reductions, etc.