In this tutorial, we'll learn how to use the restart policy with Docker Compose.
First, we'll cover how to restart Docker containers with restart policies. Then we'll cover how Docker Compose defines restart policies in normal mode and swarm mode as a configuration for multi-container Docker applications.
2. Docker Restart Policy
Restart policies are strategies we can use to restart Docker containers automatically and manage their lifecycles.
Given that containers can fail unexpectedly, Docker has safeguards to prevent services from running into a restart loop. In case of a failure, restart policies don't take effect unless the container runs successfully for at least 10 seconds.
We can also assume that manually stopping a container will make Docker restart the service automatically when a restart policy is provided. However, these policies are dismissed in this case to prevent containers from restarting after being stopped arbitrarily.
To use restart policies, Docker provides the following options:
- no: Containers won't restart automatically.
- on-failure[:max-retries]: Restart the container if it exits with a non-zero exit code, and provide a maximum number of attempts for the Docker daemon to restart the container.
- always: Always restart the container if it stops.
- unless-stopped: Always restart the container unless it was stopped arbitrarily, or by the Docker daemon.
Now let's look at an example of how to set a restart policy using the Docker CLI for a single container:
docker run --restart always my-service
From the example above, my-service will always restart if the container stops running. However, if we explicitly stop the container, the restart policy will only take effect when the Docker daemon restarts, or when we use the restart command.
The previous example demonstrates how the restart flag configures the strategy to restart a single container automatically. However, Docker Compose allows us to configure restart policies to manage multiple containers by using the restart or restart_policy properties in normal mode or swarm mode.
Before diving into the restart policies implementation with Docker Compose, let's set up a working environment.
We must have a running Docker container to test the restart policies we'll specify. We'll use a project from a previous article, spring-cloud-docker, which is a dockerized Spring Boot application. This project has two Docker services that we'll use to implement different restart strategies with Docker Compose.
First, we must confirm that we can run both containers by running the following command from the project root:
docker-compose up --detach --build
Now we should be able to see both services running by executing docker-compose ps:
$ docker ps
Name Command State Ports
message-server java -jar /message-server.jar Up 0.0.0.0:18888->8888/tcp
product-server java -jar /product-server.jar Up 0.0.0.0:19999->9999/tcp
Alternatively, we could go to localhost:18888 or localhost:19999 in our browser and verify that we see the messages displayed by the application services.
4. Restart Policy in Docker Compose
Like the restart Docker command, Docker Compose includes the restart property to restart containers automatically.
We can also define restart policies in Docker Compose by providing the restart property to the service in the docker-compose.yml file. Docker Compose uses the same values provided by the Docker CLI restart command to define a restart strategy.
Now let's create a restart policy for our containers. In the spring-cloud-docker project, we must change the docker-compose.yml configuration file by adding the restart policy property:
Notice how we added the restart property to both services. In this case, the message-server will never restart automatically. The product-server will only restart if it exits with a non-zero code as specified by the on-failure value.
Next, let's see how the same policies are declared using Docker Compose in swarm mode.
5. Restart Policy in Docker Compose Swarm Mode
Docker Compose in swarm mode provides a larger set of options when specifying how containers will restart automatically. However, the following implementation only works in Docker Compose v3, which introduces the deploy key-value pair in the configuration.
Below we can find the different properties to further expand the configuration for restart policies in swarm mode:
- condition: none, on-failure, or any (default)
- delay: Duration between restart attempts
- max_attempts: Maximum number of attempts outside of the restart window
- window: Duration for determining if a restart is successful
Now let's define our restart policies. First, we must make sure we're using Docker Compose v3 by changing the version property:
Once we change the version, we can add the restart_policy property to our services. Similar to the previous section, our message-server container will always restart automatically by providing the any value in the condition:
Similarly, we'll add an on-failure restart policy to the product-server:
Notice how the restart_policy property is within the deploy key, which indicates that restart policies are a deployment configuration we provide to manage a cluster of containers in swarm mode.
Also, the restart policies in both services include the additional configuration metadata that makes the policies a more robust restart strategy for the containers.
In this article, we learned how to define restart policies with Docker Compose. After introducing Docker restart policies, we used a previous Baeldung project to demonstrate two ways to configure a restart strategy for containers.
First, we used the Docker Compose restart property configuration, which includes the same options as the native restart command in the Docker CLI. Then we used the restart_policy property, only available in swarm mode and Docker Compose version 3, along with other configuration values that define restart policies.
As always, all the sample code used in this article is available over on GitHub.
res – REST with Spring (eBook) (everywhere)