norwae / ignifera Edit

Simple akka-http/Prometheus integration

Version Matrix

Ignifera

  1. Prometheus statistics export and collection for akka http routes. The library collects http result codes, timings, and requests in flight. It additionally optionally exposes some basic akka statistics.

  2. Graceful shutdown, health and readiness functions. These routes are provided at a low level to they can be used by both the routing DSL and special case implementations.

  3. Access log collection.

These functions can be freely composed to (e.g.) exclude health check routes from statistics and access log, or include them, depending on the requirements and standards of the user

Features

Ignifera can be inserted into an existing application very simply:

import com.github.norwae.ignifera.{StatsCollector, StandaloneExport}

val myRoutes: Route = ???

Http().bindAndHandle(GracefulShutdownSupport(StatsCollector(myRoutes)), "localhost", 8080) foreach { _ =>
  new StandaloneExport(8081).start()
}

Alternatively, the metrics route can be included in another routing as follows (Be aware that this will slightly modify statics as the metrics requests will be counted in the metrics themselves):

val myRoutes: Route = ???
val fullRoutes = myRoutes ~ get(path("metrics")(IncludedHttpExport.statusRoute))
Http().bindAndHandle(StatsCollector(fullRoutes), "localhost", 8080)

As you can decide to include both metrics, support for the graceful shutdown or either one by itself. If you include the StatsCollector inside the GracefulShutdownSupport, health checks will not be metered. By reversing the order, they will be metered.

If the GracefulShutdownSupport has been included it can be configured to send a final metrics push to a specified endpoint before shutting down. To activate this feature the environment variable PROMETHEUS_PUSH_GATEWAY has to be set to the target which should receive the metrics.

Exported Metrics

http_requests_in_flight

Gauge - Requests currently in flight. Will respond to the number of requests currently en-route between the start and completion of routing. This will not necessarily signal completion of response streaming, only that the response status code and headers have been decided.

http_requests_total

Count - Requests processed by the application. As above, does not require complete transmission of response data (or even request body, if it is ignored).

http_request_duration_microseconds

Summary - Time to response determined. The response time is counted from beginning of routing, and internally has the same resolution as the java method System.nanotime()

Dimensions

  • method - Http method of the requests
  • code - Response status code

http_response_size_bytes

Summary - Response size (estimated). Bytes to be streamed to the client. This includes entity size, and any headers set by the routing layer. Depending on the location the stats collection is included, it may not necessarily include headers added by content negotiation. Detection depends on the presence of a Content-Length header. Responses lacking this header will not be included in the collected statistics.

http_request_size_bytes

Summary - Request size (estimated). Bytes bytes submitted by the client. This includes entity size, and any headers set by the client (including cookies). Detection depends on the presence of a Content-Length header. Responses lacking this header will not be included in the collected statistics.