Skip to content

Latest commit

 

History

History
72 lines (57 loc) · 3.26 KB

README.md

File metadata and controls

72 lines (57 loc) · 3.26 KB
Raw

Docker App Proxy

This is a simple Python script we use at the UO Libraries for creating a semi-automated staging server for docker-compose projects.

This project (and this documentation) is mostly for UO internally, so feel free to use this as inspiration, but understand that if your staging server differs significantly enough, you may need to significantly alter the nginx config, the python, or both.

Setup

  • Install docker and docker-compose
  • Install nginx
  • Grab docker-gen and copy the binary to /usr/local/bin or something
  • You may have to make SELinux more permissive if you have containers that expose arbitrary ports:
    • setsebool -P httpd_can_network_connect on
  • Get docker-gen running (this will have to be done as root):
    • /usr/local/bin/docker-gen --watch --notify "nginx -s reload" /usr/local/register/nginx.tmpl /etc/nginx/conf.d/docker.conf
    • Feel free to submit a PR with a systemd setup (we'll probably get to this eventually)
  • Run the "installer": ./deploy.sh staging.mysite.edu. This will make the command register-app available to all users.
  • Instruct users to source setup-dc-staging from their docker project's directory; this allows one to type stage up -d rather than docker-compose -f docker-compose.staging.yml. It also sets up the alias to use docker-compose.override.yml if present, further simplifying usage.

Usage

Automatic

Register an app that uses compose via register-app <path to app>. This will let you specify the app's hostname and which services are exposed by nginx as well as which port in the container needs to be exposed on port 80.

Once you've saved your configuration and exited, you'll have a staging.docker-compose.yml file which you can then use. Assuming the docker-gen watcher is working, simply starting up the stack with the staging file will make everything "just work":

docker-compose -f staging.docker-compose.yml up -d

Part of the process here automatically strips all "ports" definitions from the source compose file when writing out the staging compose file. This ensures that you won't have collisions when running multiple applications / stacks concurrently.

Note that for https to be enabled, you must have certificates in /etc/nginx/certs/<host>.crt and /etc/nginx/certs/<host>.key. If you have /etc/nginx/certs/default.*, that will be used for all hosts if they don't have their own cert.

Manual

You can simply add environment variables to your services if you'd rather not deal with the register-app script. The current list of vars is as follows:

  • STGCONF_X_HOST: This is the prefixed host nginx will use to expose your application. If this is set to "foo", for instance, and the project was deployed with a hostname of "staging.bar.com", the full URL would be "http(s)://foo.staging.bar.com".
  • STGCONF_X_PORT: Specify the service's port which you want to be used when a request is proxied from nginx

NOTE, however, that if you simply adjust an existing compose file, you should be cautious. Many times the compose file will automatically map ports from the services to the host. This can cause conflicts! Consider using the automatic setup to strip "ports" definitions, and then tweaking the output manually.