monitoring

Appkit monitoring

This package provides a basic interface for monitoring request times and other arbitrary data, and recording data into InfluxDB.

Configuration for InfluxDB

Include a field of type monitoring.InfluxMonitorConfig in your app config:

type AppConfig struct {
	// ...

	InfluxDB monitor.InfluxMonitorConfig
}

And configure this field with a value like: https://<username>:<password>@<influxDB host>/<database>.

Create an InfluxDB monitor in main.go:

if monitor, err := monitoring.NewInfluxdbMonitor(appConfig.InfluxDB, l); err != nil {
	// handle the error...
	l.Info().Log(
		"err", err,
		"msg", fmt.Sprintf("error configuring influxdb monitor: %v", err),
	)
}

Middleware

Use included middleware in your stack:

func Routes(monitor monitoring.Monitor, ...) http.Handler {
    // ...

	middleware := server.Compose(
		// ... other middleware
    	monitoring.WithMonitor(monitor),
    	// ... more middleware
    )

    // ...
}

Now request data including request path, method, HTTP response status code, request duration, and request trace ID, will be sent to your InfluxDB instance in the request measurement.

Path scrubbing

The middleware will convert any sequences of 1 or more digits into :id. Eg. GET /api/users/123/comments/456 will be tagged with path of /api/users/:id/comments/:id.

Recording other metrics

To record other metrics, eg counting subscriptions, measuring time of API calls to other services, retrieve the metric from the context with monitoring.ForceContext, and then call methods on the interface:

type Monitor interface {
	InsertRecord(measurement string, value interface{}, tags map[string]string, fields map[string]interface{}, time time.Time)
	Count(measurement string, value float64, tags map[string]string, fields map[string]interface{})
	CountError(measurement string, value float64, err error)
	CountSimple(measurement string, value float64)
}

For example, to count logins:

func Login(w http.ResponseWriter, r *http.Request) {
    monitor := monitoring.ForceContext(r.Context())
    if err := doLogin(); err == nil {
        // ... handle success case
        monitor.CountSimple("login_success", 1)
    } else {
        monitor.CountError("login_error", 1, err)
    }
}

TODO

• Make metrics counted via the context monitor include request tags eg. request ID

• Provide simple wrapper API for context-based metrics:

  func InsertRecord(context.Context, string, interface{}, map[string]string, time.Time)
  func Count(context.Context, measurement string, value float64, tags map[string]string)
  

• Make path-scrubbing more flexible. Eg. scrub out product names: /products/blue-winter-coat -> /products/:product_code.