Difference between links and depends_on in docker_compose.yml

This answer is for docker-compose version 2 and it also works on version 3

You can still access the data when you use depends_on.

If you look at docker docs Docker Compose and Django, you still can access the database like this:

version: '2'
services:
  db:
    image: postgres
  web:
    build: .
    command: python manage.py runserver 0.0.0.0:8000
    volumes:
      - .:/code
    ports:
      - "8000:8000"
    depends_on:
      - db

What is the difference between links and depends_on?

links:

When you create a container for a database, for example:

docker run -d --name=test-mysql --env="MYSQL_ROOT_PASSWORD=mypassword" -P mysql

docker inspect d54cf8a0fb98 |grep HostPort

And you may find

"HostPort": "32777"

This means you can connect the database from your localhost port 32777 (3306 in container) but this port will change every time you restart or remove the container. So you can use links to make sure you will always connect to the database and don't have to know which port it is.

web:
  links:
   - db

depends_on:

I found a nice blog from Giorgio Ferraris Docker-compose.yml: from V1 to V2

When docker-compose executes V2 files, it will automatically build a network between all of the containers defined in the file, and every container will be immediately able to refer to the others just using the names defined in the docker-compose.yml file.

And

So we don’t need links anymore; links were used to start a network communication between our db container and our web-server container, but this is already done by docker-compose

Update

depends_on

Express dependency between services, which has two effects:

docker-compose up will start services in dependency order. In the following example, db and redis will be started before web.

docker-compose up SERVICE will automatically include SERVICE’s dependencies. In the following example, docker-compose up web will also create and start db and redis.

Simple example:

version: '2'
services:
  web:
    build: .
    depends_on:
      - db
      - redis
  redis:
    image: redis
  db:
    image: postgres

Note: depends_on will not wait for db and redis to be “ready” before starting web - only until they have been started. If you need to wait for a service to be ready, see Controlling startup order for more on this problem and strategies for solving it.

---

The post needs an update after the links option is deprecated.

Basically, links is no longer needed because its main purpose, making container reachable by another by adding environment variable, is included implicitly with network. When containers are placed in the same network, they are reachable by each other using their container name and other alias as host.

For docker run, --link is also deprecated and should be replaced by a custom network.

docker network create mynet
docker run -d --net mynet --name container1 my_image
docker run -it --net mynet --name container1 another_image

depends_on expresses start order (and implicitly image pulling order), which was a good side effect of links.

---

[Update Sep 2016]: This answer was intended for docker compose file v1 (as shown by the sample compose file below). For v2, see the other answer by @Windsooon.

[Original answer]:

It is pretty clear in the documentation. depends_on decides the dependency and the order of container creation and links not only does these, but also

Containers for the linked service will be reachable at a hostname identical to the alias, or the service name if no alias was specified.

For example, assuming the following docker-compose.yml file:

web:
  image: example/my_web_app:latest
  links:
    - db
    - cache

db:
  image: postgres:latest

cache:
  image: redis:latest

With links, code inside web will be able to access the database using db:5432, assuming port 5432 is exposed in the db image. If depends_on were used, this wouldn't be possible, but the startup order of the containers would be correct.

---

I think that the answers for this question need updating based on the new Docker compose specification introduced first in v1.27.0, which now allows for a long-form of depends_on:

https://github.com/compose-spec/compose-spec/blob/master/spec.md#long-syntax-1

In this long form, you can specify that you want to wait for a service to be either started, healthy, or completed successfully.

Docker compose knows that a service is healthy if you produce a health_check on that service:

https://github.com/compose-spec/compose-spec/blob/master/spec.md#healthcheck

I'd recommend to read the examples in the documentation for more details, see links above!

For a quick example, this is what I used in a compose file for integration tests:

services:
  cloud-broker:
    image: my.docker.registry/activemq-artemis:latest
    healthcheck:
      test: ["CMD-SHELL", "wget http://localhost:8161/ --delete-after --tries=3 2> /dev/null"]
      interval: 10s
      timeout: 5s
      retries: 5

  postgresql:
    image: postgres
    healthcheck:
      test: ["CMD-SHELL", "pg_isready -U postgres"]
      interval: 10s
      timeout: 5s
      retries: 5
    environment:
      POSTGRES_PASSWORD: "<my-secret>"
      POSTGRES_USER: "postgres"
      POSTGRES_DB: "postgres"
  
  # This service calls a script to create an empty database and the service-user
  postgresql-setup:
    image: postgres
    depends_on:
      postgresql:
        condition: service_healthy
    restart: "no"
    volumes:
      - "./scripts:/scripts"
    environment:
      PGPASSWORD: "<my-secret>"
    entrypoint: "psql -U postgres -d postgres -h postgresql -f /scripts/create-db.sql"

  my-business-service:
    image: my.docker.registry/my-business-service:latest
    depends_on:
      cloud-broker:
        condition: service_healthy
      postgresql-setup:
        condition: service_completed_successfully

---