An example of a service, which "evolves" step by step.
It might be helpful to set up git pre-commit hooks before writing the code. For example, you can use pre-commit by Yelp to manage this kind of hooks.
See .pre-commit-config.yaml
.
The simplest way to make a web-server: register a handler function and listen on some TCP network address:
http.HandleFunc("/", home)
http.ListenAndServe(":8000", nil)
TestHandler
function (see handlers_test.go
) is the simplest way to test our web-server: check if the handler
returns expected result. Use go test -v
command to run test.
We need to be able to process diffrenet HTTP methods and URLs. So, let's use takama/router instead of standard handler registrar.
You can try any other router. For example, one of the most popular solutions is julienschmidt/httprouter.
Don't forget to check if tests are still green (go test -v
) and change them if it is necessary.
Add logger to log requests. Sirupsen/logrus is my favourite!
For this workshop we choose glide.
After installation of glide run glide init
to initialize dependency management and glide install
to install
dependencies. Feel free to use glide up
to check for dependencies updates.
In this example we ignore vendor
directory and glide.lock
file. But in real "production" cases it's
better to commit vendor
or at least glide.lock
to be able to reach dependencies in an emergency situation.
It's very important to store service configuration in the environment. So, ENV variables is, probably, the simplest way to deal with config.
We believe what Dockerfile of any application must be as tiny as it possible.
For simple services like this one, we can use FROM scratch
construction and provide
very tiny Docker image.
Let's compile our service with special tags:
env CGO_ENABLED=0 GOOS=linux go build
and make a container with the binary:
docker build -t step-by-step -f ./Dockerfile .
Run service on Docker:
docker run -p 8000:8000 step-by-step
It might be very helpful to know, what the current version of build is.
Liveness and readiness probes will help us to manage containers.
Liveness probe is more like typical healthcheck. If liveness probe fails, Kubernetes will destroy and deploy again the failed container.
Readiness probe shows that application is ready to receive connections. For example, during the initialization readiness probe may show what the instance of application is not ready for traffic yet, so Kubernetes will not send requests to this instance.