HAProxy is my homelab loadbalancer of choice due to it’s versatility and general ease of configuration. Whether it’s HTTP or just plan TCP traffic I want to land within my lab, a few tweaks to an HAProxy config is all it takes. However, as I deploy more and more random services, which I want available from the internet, having to remote into my various HAProxy ingress servers becomes a pain. Also, since I like to have isolated HAProxy instances depending on what I’m doing, yet again having to remote into boxes to make changes becomes even more tiresome.
After playing with a few possible choices I settled on creating a Ruby on Rails application to manage deploying HAProxy configurations to my various endpoints. This application, which I call HITS (HAProxy Infrastructure Transformation Service), allows for defining rules, frontends, backends, and servers which ultimately go into a building an instance of HAProxy. It also handles issuing and renewing Lets Encrypt SSL certificates, and distributing those certificates to proper endpoints.
Ruby on Rails for the application. Docker for building the HAProxy containers. Modified Docker API gem for doing docker image builds against remote endpoints. And obviously HAProxy.
What does it look like?
Graphical design is a skill that I haven’t been blessed with, so it’s stock Bootstrap 4.
From here you can issue new SSL certificates, delete existing ones, or forcefully renew. This is all enabled thanks to free Lets Encrypt certificates
This is where you define rules (ACLs), such as SSL redirects, path based redirects, etc. These rules help ensure traffic lands where it needs to – such as having multiple websites hosted on one IP, using host headers to decide which backend to land on.
These are backend servers that are available to add to backend pools. Backend servers run the actual applications HAProxy is sending traffic towards. A single backend can have one or more backend servers
Backends are a collection of servers, here you can see your defined backends, and add/remove servers from them. Generally, for a load balancer to actually balance traffic, you want multiple servers in a given backend.
Frontends are ingress points where rules are evaluated and traffic is redirected to backends based on the results on, etc. Frontends define which ports need to take in traffic, whether they are HTTP or regular TCP, and which ACLs apply to each request.
A target host is a server running Docker which will run your desired HAProxy configurations. You attach desired frontends to a host, which generates a new release of your configuration. Each configuration is a stand alone docker image, complete with everything needed to handle the configured traffic workloads. The docker build is automatically performed against the target host.
Finally, you release your generated configuration onto the host, which makes the configuration active. You can easily roll back releases using the release history.
That’s about it. This is a weekend lab project created to make it so I need to SSH into my various load balancers less often. It’s not yet 100% complete (things like Where Used are not coded yet), however it is functional and scratches most of the itches I had around this subject.
Before you say “hey fool, why not just use Traefik or Consul” etc. I’ve used them, they work great, but in my use case I just wanted to quickly be able to toss some endpoints on a HAProxy instance in my lab, or on my dedicated server, and not need to think much about service discovery, having to decom endpoints when I’m done, etc. This is quick and dirty “add endpoints, remove them, press a button, and it’s live” solution.
Thanks for reading!