MVP #1
Description
The simple version of this is to read metrics from https://prometheus.io -compatible agents and send them on to at least one output supported by Telegraf (see https://github.com/influxdata/telegraf/tree/master/plugins/outputs for output plugins). Post MVP, additional output plugins must be supported using environment variables to select the plugin and configure it.
For the MVP, source data will come from ContainerPilot telemetry (described in https://www.joyent.com/blog/containerpilot-telemetry). Later versions will need to support TLS keys as needed to read from Triton Container Monitor (see https://github.com/joyent/rfd/blob/master/rfd/0027/README.md). A running example of applications exposing ContainerPilot telemetry can be found in https://github.com/autopilotpattern/wordpress.
The Telegraf application must be dockerized and use https://www.joyent.com/containerpilot to discover telemetry sources. See an example of how this works in Prometheus itself in https://github.com/autopilotpattern/prometheus.