What metrics to export?

[anacrolix]

This issue will serve to discuss and collect specific metrics for export, through Prometheus, or Prometheus-compatible tooling (to keep us focused on the kinds of values), such as in #297. It is not a 1:1 list, but rather a list of things we want to and can observe. Units, and types can be figured out in implementation to avoid getting off-topic here.

Here is the list:

• Number of k-buckets
• Number of entries in each k-bucket
• Outbound/inbound message sizes
• Outbound/inbound message counts
• Outbound request latency
• Write latency (time between wanting to send a message, and sending it Add stream pooling #285 and DHT Request pipelining #92)

[anacrolix]

@raulk I'd like some suggestions on aspects of the dial queue to be exported.