Currently, the only way to extract metrics from the driver is to register for various events (e.g. command events, CMAP events) and calculate statistics in the event handlers.

In addition, there are some statistics exposed by mgo that would not be possible to track even with the current set of events (e.g. total bytes in/out).

This would be very useful for users of the driver for tracking purposes. Finally, if the driver had well designed, consistent statistics in areas, tracking bytes sent and received for purposes of analytics or billing would be vastly easier.

The current requested metrics are:

• total bytes in/out
  • per downstream, per API
• commands (per API, per downstream)
• connections (per downstream)
  • idle
  • total
  • in-use
• latency distribution
  • per API, downstream
  • ideally with percentiles