Measuring performance of application estate is crucial in today's microservice world - it is crucial that dev-ops enabled teams can monitor, react and scale dynamically to changes in the runtime environment. However, because of the plethora of monitoring tools on the market, and because http4k is a toolkit and not a complete "batteries included" framework, it provides a number of integration points to enable monitoring systems to be plugged in as required. Additionally, it is envisaged that users will probably want to provide their own implementations of the http4k
ServerConfig classes (
Undertow etc..) so that tweaking and tuning to their exact requirements is accessible, instead of http4k attempting to provide some generic configuration API to achieve it.
compile group: "org.http4k", name: "http4k-core", version: "3.163.0" compile group: "org.http4k", name: "http4k-metrics-micrometer", version: "3.163.0"
http4k provides module support for monitoring application endpoints using the micrometer metrics abstraction library, which currently enables support for libraries such as Graphite, StatsD, Prometheus and Netflix Atlas. This also provides drop-in classes to record stats such as JVM performance, GC and thread usage.
Alternatively, it's very easy to use a standard
Filter to report on stats:
This is trivial to achieve by using a Filter:
This allows a chain of application calls to be tied together and is generally done through the setting of HTTP headers on each call. http4k supports the OpenZipkin standard for achieving this and provides both Server-side and Client-side
Filters for this purpose. This example shows a chain of two proxies and an endpoint - run it to observe the changes to the tracing headers as the request flows through the system:
Easily wrap an
HttpHandler in a debugging filter to check out what is going on under the covers: