trace is a wrapper for the net/trace package that adds logging and metrics to de-clutter your functions.
trace wraps all of the functionality of net/trace, but also replicates the logs to a structured logger built on Go's standard logging library.
Metrics are exported to Prometheus with trace duration histograms by name/family, trace counts by name/family, and errors by trace name/family.
the net/trace#EventLog is also implemented in the same manner, minus metrics exposition, which doesn't make sense there.
examples/gogrep has an example command-line application that shows usage of the trace functionality to both capture trace information and logs with a single tool.
examples/service has an example web application that shows usage of the trace functionality to both capture trace information and logs with a single tool, combined with the trace.EventLog which serves as a single logging and event source for your application.
2016/10/03 00:27:34 message=found file=../../events.go trace=main
2016/10/03 00:27:34 message=found file=../../examples/gogrep/main.go trace=main
2016/10/03 00:27:34 message=found file=../../log.go trace=main
2016/10/03 00:27:34 message=found file=../../trace.go trace=main
2016/10/03 00:27:34 message=found file=../../metrics.go trace=main
2016/10/03 00:27:34 trace=main : hit count 5
2016/10/03 00:27:34 message=finished hits=5 trace=main
2016/10/03 00:34:05 name=http - Listening on :3000
trace offers two useful and one fun way to expose your metrics.
trace.ServeMetrics() will serve the metrics in Prometheus text format. Use this for long-running apps/services.
trace.PushMetrics() will push the metrics to a Prometheus push server. Use this for command-line utilities.
trace.DumpMetrics() will return a string with the metrics that Prometheus would serve, suitable for inspection, printing, tests.
trace exposes the underlying net/trace /debug/requests and /debug/events endpoints for handy visual representation of the traces, their timing/histograms, and the event log of your application.
