go-rest-service-example

An example of a cleanly written go service for serving rest requests while implementing basic metrics and handling lifcycle elements gracefully.

This package uses gorilla/mux as the base webserver. In general I prefer mux because it is less clunky than most frameworks in go while providing the flexibility I want for a complex rest server. Also there is no magic. You create a router, register your handlers, and set your middleware and you can test all of that with relative ease.

running with docker-compose

In the docker directory run docker-compose build && docker-compose up. Prometheus will be listening on http://localhost:9090. The service will be receiving rest requests on http://localhost:4445 and it's metrics endpoint will be on http://localhost:4446

package responsibility

• pkg/rest - Contains the lifecycle management for the rest server
• pkg/router - Contains all the routing information. This is where all the middleware and route priority is glued together.
• pkg/middleware - Contains all the middleware. The basics are fingerprinting requests, metrics, and auth. See below for more detail
• pkg/endpoint - Contains the actual end handlers for each route. It also contains the binding from endpoint handlers to authentication method.You simply add a new Endpoint struct and put it in the authed list or public list if you want a new endpoint.
• pkg/config - Contains the configuration logic that could be used to customize the project. I use env variable injection

middleware

• Request pathing is the following:

  fingerprinting -> metrics -> auth -> ... -> handler
  

  Because we want to fingerprint every request but run metrics on as much of the request lifecycle as possible. Right now fingerprinting, metrics, and auth are the only middleware implemented but in a real project it could easily be extended in the middleware package.

metrics

• The current metrics are:
  • http_responses - records the response codes per path
  • http_latency - records the latency of each request
  • http_active_requests - records the number of active requests in memory
• This is really just the beginning of stats and can easily be extended by adding more to the pkg/middleware/metrics.go file.
• There is a nice interplay where the prometheus metrics measure the latency and responses of all the paths, including the metrics endpoint itself. Yo dawg.

Self signed certificates:

• If you want to run this outside a docker container then just run the command

  openssl req -new -newkey rsa:2048 -days 365 -nodes \
      -subj "/O=testcert" \
      -x509 -keyout server.key -out server.crt
  

  to generate self signed certificates to use