Most of my talks contain a demo. A fair share of these demos require multiple "infrastructure" dependencies: a database (or more), Elasticsearch, you name it. To ease my setup and avoid stuffing my machine, I use either Docker Compose or Kubernetes locally on my Mac. Both rely on Docker Desktop.

To expose a cluster Service on my host, I use nodePort. Hence, I set a dedicated node port for each service. I need to remember each of them for each demo. Worse, services might be (are) declared across different manifest files.

For a long time, I wanted to ease my life. I've searched for Kubernetes-based solutions. I found that kube-forward was not stable enough.

My latest attempt was MetalLB. Even though I didn't manage to make it work, it bound port 8080 on my machine: none of my other regular Spring demos could work.

Last week, I decided to take another approach - a regular proxy in front of my local cluster. OSX comes with an existing Apache Web Server installation. You can check it with ls /etc/apache2:



extra                 httpd.conf.pre-update mime.types            other
httpd.conf            magic                 original              users

The following modules are necessary:



#httpd.conf
LoadModule proxy_module libexec/apache2/mod_proxy.so
LoadModule proxy_http_module libexec/apache2/mod_proxy_http.so
LoadModule proxy_balancer_module libexec/apache2/mod_proxy_balancer.so

The requirement is straightforward: proxy calls from http://zerodowntime.hz to http://localhost:$NODE_PORT/. For this, we need to configure a virtual host:



#httpd-vhosts.conf
<VirtualHost *:80>
    ServerName zerodowntime.hz
    ProxyRequests off
    ProxyPass / http://localhost:30002/
    ProxyPassReverse / http://zerodowntime.hz
</VirtualHost>

To make sure everything works fine, we can use apachectl -S:



VirtualHost configuration:
*:80           zerodowntime.hz (/private/etc/apache2/extra/httpd-vhosts.conf:40)

Last but not least, let's configure the host file:



#./etc/hosts
127.0.0.1        zerodowntime.hz

At this point, we can access the application using the zerodowntime.hz URL:

Depending on the deployed application, this step might be the last one. It's unfortunately not my case, as my demo uses a redirect. By default, the redirect location sent to the client is the URL known to the application, http://localhost:30002, defeating the whole purpose. We need to configure the application to use the standard X-Forwarded-* HTTP headers.

I'm using Spring Boot, so that is just a matter of configuration:



#application.yml
server:
  forward-headers-strategy: native

At this point, everything works as expected!

To go further:

Originally published at A Java Geek on November 28^th, 2021

Port management in your local Kubernetes cluster