Export stats to a time-series database

[djs55]

Since we're on the network data path we can measure the traffic as it passes through. We could measure

• packets per second
• bytes per second
• DNS requests latency
• DNS request failure rate
• Number of active UDP NAT rules etc

This data could be sent to an external time-series database such as influxdb and then visualised.

This involves

• writing a client for a time-series database (this may be as easy as connect()ing to it and writing .json messages)
• adding some basic counters to the network stack
• (stretch goal): allow the user to define their own filters via datakit and record these too

[errordeveloper]

From Prometheus it's really quite simple, you'd only need to expose and HTTP /metrics endpoint, see docs about format.

[samoht]

@djs55 have you seen https://github.com/mirage/prometheus ?

[dave-tucker]

Note: It's already being used in DatakitCI courtesy of https://github.com/mirage/prometheus