Last week I was searching how to manage multiple development environments with the same docker-compose configuration for my Django workflow. I needed to manage a development and a production environment, so this is what I did.
Some descriptions on my data:
- I had around 20 env vars, but some of them where shared among environments.
- I wanted to do it with as little impact as possible.
First, docker-compose help command
The first thing I did was run a simple
docker-compose --help, and it returned this:
Define and run multi-container applications with Docker. Usage: docker-compose [-f <arg>...] [options] [COMMAND] [ARGS...] docker-compose -h|--help Options: -f, --file FILE Specify an alternate compose file (default: docker-compose.yml) # more not necessary stuff --env-file PATH Specify an alternate environment file
I went with the
-f flag, because I also wanted to run some docker images for development. By using the
-f flag I could create a base compose file with the shared env vars (docker-compose.yml) and another one for each of the environments (prod.yml and dev.yml)
So I went to town. I kept the shared variables inside
docker-compose.yml and added the specific variables and configuration to
version: "3" services: app: build: context: . ports: - "8000:8000" volumes: - ./app:/app command: > sh -c "python manage.py migrate && python manage.py runserver 0.0.0.0:8000" environment: - myvar1=myvar1 - myvar2=myvar2 ... - myvarx=myvarx
Since I’m going to connect to a remote RDS and a remote Redis, in my
prod.yml, I don’t need to define a Postgres or Redis image:
version: "3" services: app: environment: # DB connections - DB_HOST=my-host - DB_NAME=db-name - DB_USER=db-user - DB_PASS=mysupersecurepassword ...
dev.yml I added the Postgres database image and Redis image:
version: "3" services: app: depends_on: - db - redis environment: # Basics - DEBUG=True # DB connections - DB_HOST=db - DB_NAME=app - DB_USER=postgres - DB_PASS=supersecretpassword ... db: image: postgres:10-alpine environment: - POSTGRES_DB=app - POSRGRES_USER=postgres - POSTGRES_PASSWORD=supersecretpassword links: - redis:redis redis: image: redis:5.0.7 expose: - "6379"
And to run it between environments, all I have to do is:
# For production docker-compose -f docker-compose.yml -f prod.yml up # For development docker-compose -f docker-compose.yml -f dev.yml up
That’s it! I have multiple
docker-compose files for all my environments, but I could go even further.
Improving the solution
Improving the base docker-compose.yml file
I liked the way it looked, but I knew I could go deeper. A bunch of vars inside the base
docker-compose.yml looked weird, and made the file a little unreadable. So again, I went to the
docker-compose documentation and found what I needed: env files in docker-compose.
So I created a file called
globals.env, and moved all the global env vars to that file:
myvar1=myvar1 mivar2=myvar2 ...
And on the
docker-compose.yml file I called the
app: env_file: globals.env
This is the final result:
version: "3" services: app: build: context: . ports: - "8000:8000" volumes: - ./app:/app command: > sh -c "python manage.py migrate && python manage.py runserver 0.0.0.0:8000" env_file: globals.env
Improving the running command
As I mentioned before, I wanted as little impact as possible, and
docker-compose -f docker-compose.yml -f envfile.yml up was a bit long for me. So I created a couple of bash files to ease the ingestion of
#!/usr/bin/env bash # Run django as production docker-compose -f docker-compose.yml -f prod.yml "[email protected]"
#!/usr/bin/env bash # Run django as development docker-compose -f docker-compose.yml -f dev.yml "[email protected]"
"[email protected]" means “Append all extra arguments here”, so if I ran the
dev command with the
up -d arguments, the full command would be
docker-compose -f docker-compose.yml -f development.yml up -d, so it is exactly what I wanted
A quick permissions management:
chmod +x prod dev
And now I could run my environments as:
./dev up ./prod up
I was satisfied with this solution. I could run both environments wherever I want with only one command instead of moving env vars all over the place. I could go even further by moving the environment variables of each file to its own
.env file, but I don’t think that’s needed for the time being. At least I like to know that I can do that down the road if it is necessary