kube-apiproxy

If you ever had to set up a Kubernetes cluster you know that it is a fairly simple task, apart from one thing - if API server goes away, you have to bring it back and tell other components where it is. Some people use an ELB with a persistent DNS entry which they update if kubernetes API service is moved somewhere else, others use config management tools to update configuration and restart Kubernetes components, etc.

A lot of people deploy Kubernetes on CoreOS using fleet which is a really great way.

This proxy service talks to fleet daemon to find which node runs Kubernetes API and proxies traffic through to the API server. When that node goes down, fleet will then start the API service on another node which is automatically discovered by kube-apiproxy.

This means that Kubernetes API is always accessible on localhost:8081 by default.

Building

You will need gb tool - http://getgb.io/.

git clone https://github.com/vaijab/kube-apiproxy.git
cd kube-apiproxy
gb build all

Configuration

Configuration is done via command line arguments.

$ kube-apiproxy --help
Usage of bin/kube-apiproxy:
  -api-port="8080": kubernetes api port
  -fleet-endpoint="unix:///run/fleet.sock": fleet endpoint
  -proxy-listen="localhost:8081": proxy listen ip:port
  -unit-name="kube-apiserver.service": fleet unit name for kubernetes api server

Contribution

Please feel free to send PRs and issues my way. I am very new to Go programming, I am sure that the code can be improved a lot by keeping the core principle the same.