Simple skeleton for containerizing your php application using nginx/php-fpm and works great for local development.
For this to work you have to have docker installed with a daemon running. If you can't run docker info on the command line this will never work. Install docker for desktop if youre using mac/windows and docker if you're on linux.
Start by building your base image, the php docker rig compiles everything inline so it's just easier and faster to get that all done once and then copy your php code into the base image. If you need extra packages you can always add them and make a new base image. The following docker command will make the base image we will extend from in the main Dockerfile.
docker build --rm -f ./Dockerfile.php -t base .
The docker-compose.yml file has everything you need to start the sample app and connect php/nginx/mysql via docker networking. The trick is the services are available via dns with their simple names e.g. nginx, php, db so nginx has it's config set to call the php-fpm backend at php:9000 and this is set as an environment variable if you move the container to say kubernetes where you might use different naming conventions.
docker-compose up --build
You will now have http://localhost:8080 pointed to port 80 on the nginx container and a docker ps which looks something like this...
> docker ps
CONTAINER ID IMAGE COMMAND CREATED STATUS PORTS NAMES
31694084f9f4 danday74/nginx-lua "nginx -g 'daemon of…" 30 minutes ago Up 29 minutes 443/tcp, 0.0.0.0:8080->80/tcp php-docker-skeleton_nginx_1
a0b5b24f42b2 php-docker-skeleton_php "docker-php-entrypoi…" 30 minutes ago Up 29 minutes 9000/tcp php-docker-skeleton_php_1
c3b06a217c0f mysql:5.7 "docker-entrypoint.s…" 32 minutes ago Up 29 minutes 0.0.0.0:3306->3306/tcp, 33060/tcp php-docker-skeleton_db_1
There's a sample Github action which you can use to build images on tag and push them into a gitlab repository and auto deploy them to a kubernetes server. Requirements are secrets in your github repository of a gitlab api token and the contents of a kubernetes config yaml file base64 encoded. This means your workflow can be git tag v1.0.0 && git push origin v1.0.0 and the action will make your images and run helm to update the deployments to upgrade their image.
For this to work you will need to build the base image and push it into gitlab to something like registry.gitlab.com/myapp/app/base:latest and update your Dockerfile to extend FROM the full registry.
Included is a sample helm chart for how you could deploy an app and use secrets to store your database credentials. The chart will make a deployment which will contain one pod with the nginx/php-fpm containers running inside. This assumes you are using an external database and you pass the credentials into your app to connect via environment variables much like the docker file.
Since youre now running inside a container you can't run composer from your host and will need to exec into the container to run it. Since the directories are mounted any changes composer makes to the file system will show up on your host and you can use them in your normal workflows like checking them into git. Included are a ps1 and bash script so you can run ./composer like you normally would and it will pass the args and run inside the container. This makes it so you will always run composer with the version of php that is inside the container and ultimately in production.
- For local dev the files are mounted from your host system into the container you can edit files and refresh your browser to see changes.
- nginx is used to serve static files like images/css/js and this is why the Dockerfile.nginx needs the COPY command, you will always build two images with this setup.
- The helm chart is untested, but I do use something similar in production today and copy/pasta'ed it... should be close enough.