As we continue to develop our newer inhouse software around an AMQP (RabbitMQ) core, we occasionally notice run-away queues (usually a problem on our end) and these are easily spotted in rabbitmqctl and cony output.<br><br>
What we don't seem to be able to determine at the moment, is a breakdown of things like average message size, messages per second or size per second (or whatever period) for specific queues and exchanges.<br><br>The concern we have is that the system might be working reliably, but that we're putting so much data through the system it has an affect on the data quotas of the connections we use as well as potentially slowing down other things going through the system.<br>
<br>In discussing this with a colleague, the only piece of the system that knows everything coming in and going out, would be things like exchanges within a rabbit instance - unless one were to try and implement monitoring on every single endpoint (which would then be another data stream which needs to be handled)...<br>
<br>We looked through the source code of the erlang behind the rabbitmqctl shellscript and couldn't really find any existing obvious ways to extract statistics that would capture details of messages that are being processed.<br>
<br>Google found this:<br><div style="margin-left: 40px;"><a href="https://dev.rabbitmq.com/wiki/ManagementAndMonitoring">https://dev.rabbitmq.com/wiki/ManagementAndMonitoring</a><br></div><br>Tried searching the mailing list on discussion mentioning that page and found a few references but nothing that really talked about the metrics site of what the page mentions. Is there any functionality available at the moment, even if it doesn't have a frontend yet, that would allow us to extract statistics like those mentioned on your dev wiki under requirements?<br>
<br>Anthony<br>