Help interpret /metrics please!

[petr-larin]

1. Is this in milliseconds?
   postal_message_queue_latency_sum

2. Is it possible to split metrics by server?

[willpower232]

1. It appears to be in seconds according to this comment

postal/app/lib/worker/process.rb

Lines 312 to 313 in d532922

	register_prometheus_histogram :postal_message_queue_latency,
	docstring: "The length of time between a message being queued and being dequeued (in seconds)"

2. It looks like it is powered as a whole rather than by server which makes sense because there is only one queue for your whole Postal instance

postal/app/lib/message_dequeuer/initial_processor.rb

Lines 34 to 39 in d532922

	def increment_dequeue_metric
	time_in_queue = Time.now.to_f - @queued_message.created_at.to_f
	log "queue latency is #{time_in_queue}s"
	observe_prometheus_histogram :postal_message_queue_latency,
	time_in_queue
	end

[petr-larin]

1. Thank you! I recorded these values:
   postal_message_queue_latency_sum 345212.23764801025
   postal_message_queue_latency_count 541.0
   if there are seconds, then I get average queueing time 638s = over 10m, however, all my emails go out very quickly. That made me assume that postal_message_queue_latency_sum is in ms.

2. And my question regarding the split by server was actually about all metrics, not just queues. Eg can I see the overall outgoing volume split by server? So far I haven't seen any such labels coming up in Prometheus.

[willpower232]

1. thats certainly interesting, unless you had a period of time when messages weren't being delivered at all acting as a massive outlier, I'm not sure how to explain that. I don't know much ruby but that certainly appears to be seconds based

2. if its not coming up then I don't think its an option, I haven't seen any customisation in this area so you'd have to do some changes yourself if you were interested in the extra breakdown