Get Started, Part 3: Services

Estimated reading time: 9 minutes

Prerequisites

Introduction

In part 3, we scale our application and enable load-balancing. To do this, we must go one level up in the hierarchy of a distributed application: the service.

  • Stack
  • Services (you are here)
  • Container (covered in part 2)

About services

In a distributed application, different pieces of the app are called “services.” For example, if you imagine a video sharing site, it probably includes a service for storing application data in a database, a service for video transcoding in the background after a user uploads something, a service for the front-end, and so on.

Services are really just “containers in production.” A service only runs one image, but it codifies the way that image runs—what ports it should use, how many replicas of the container should run so the service has the capacity it needs, and so on. Scaling a service changes the number of container instances running that piece of software, assigning more computing resources to the service in the process.

Luckily it’s very easy to define, run, and scale services with the Docker platform – just write a docker-compose.yml file.

Your first docker-compose.yml file

A docker-compose.yml file is a YAML file that defines how Docker containers should behave in production.

docker-compose.yml

Save this file as docker-compose.yml wherever you want. Be sure you have pushed the image you created in Part 2 to a registry, and update this .yml by replacing username/repo:tag with your image details.

version: "3"
services:
  web:
    # replace username/repo:tag with your name and image details
    image: username/repo:tag
    deploy:
      replicas: 5
      resources:
        limits:
          cpus: "0.1"
          memory: 50M
      restart_policy:
        condition: on-failure
    ports:
      - "80:80"
    networks:
      - webnet
networks:
  webnet:

This docker-compose.yml file tells Docker to do the following:

  • Pull the image we uploaded in step 2 from the registry.

  • Run 5 instances of that image as a service called web, limiting each one to use, at most, 10% of the CPU (across all cores), and 50MB of RAM.

  • Immediately restart containers if one fails.

  • Map port 80 on the host to web’s port 80.

  • Instruct web’s containers to share port 80 via a load-balanced network called webnet. (Internally, the containers themselves will publish to web’s port 80 at an ephemeral port.)

  • Define the webnet network with the default settings (which is a load-balanced overlay network).

Wondering about Compose file versions, names, and commands?

Notice that we set the Compose file to version: "3". This essentially makes it swarm mode compatible. We can make use of the deploy key (only available on Compose file formats version 3.x and up) and its sub-options to load balance and optimize performance for each service (e.g., web). We can run the file with the docker stack deploy command (also only supported on Compose files version 3.x and up). You could use docker-compose up to run version 3 files with non swarm configurations, but we are focusing on a stack deployment since we are building up to a swarm example.

You can name the Compose file anything you want to make it logically meaningful to you; docker-compose.yml is simply a standard name. We could just as easily have called this file docker-stack.yml or something more specific to our project.

Run your new load-balanced app

Before we can use the docker stack deploy command we’ll first run:

docker swarm init

Note: We’ll get into the meaning of that command in part 4. If you don’t run docker swarm init you’ll get an error that “this node is not a swarm manager.”

Now let’s run it. You have to give your app a name. Here, it is set to getstartedlab:

docker stack deploy -c docker-compose.yml getstartedlab

Our single service stack is running 5 container instances of our deployed image on one host. Let’s investigate.

Get the service ID for the one service in our application:

docker service ls

You’ll see output for the web service, prepended with your app name. If you named it the same as shown in this example, the name will be getstartedlab_web. The service ID is listed as well, along with the number of replicas, image name, and exposed ports.

Docker swarms run tasks that spawn containers. Tasks have state and their own IDs. Let’s list the tasks:

docker service ps <service>

Note: Docker’s support for swarms is built using a project called SwarmKit. SwarmKit tasks do not need to be containers, but Docker swarm tasks are defined to spawn them.

Let’s inspect one of these tasks, and limit the output to container ID:

docker inspect --format='{{.Status.ContainerStatus.ContainerID}}' <task>

Vice versa, you can inspect a container ID, and extract the task ID.

First run docker container ls to get container IDs, then:

docker inspect --format="{{index .Config.Labels \"com.docker.swarm.task.id\"}}" <container>

Now list all 5 containers:

docker container ls -q

You can run curl -4 http://localhost several times in a row, or go to that URL in your browser and hit refresh a few times.

Hello World in browser

Either way, you’ll see the container ID change, demonstrating the load-balancing; with each request, one of the 5 replicas is chosen, in a round-robin fashion, to respond. The container IDs will match your output from the previous command (docker container ls -q).

(Windows 10 PowerShell should already have curl available, but if not you can grab a Linux terminal emulater like Git BASH if you want to try it out. It isn’t critical to the taskflow here.)

Note: At this stage, it may take up to 30 seconds for the containers to respond to HTTP requests. This is not indicative of Docker or swarm performance, but rather an unmet Redis dependency that we will address later in the tutorial. For now, the visitor counter isn’t working for the same reason; we haven’t yet added a service to persist data.

Scale the app

You can scale the app by changing the replicas value in docker-compose.yml, saving the change, and re-running the docker stack deploy command:

docker stack deploy -c docker-compose.yml getstartedlab

Docker will do an in-place update, no need to tear the stack down first or kill any containers.

Now, re-run docker container ls -q to see the deployed instances reconfigured. If you scaled up the replicas, more tasks, and hence, more containers, are started.

Take down the app and the swarm

  • Take the app down with docker stack rm:

    docker stack rm getstartedlab
    

    This removes the app, but our one-node swarm is still up and running (as shown by docker node ls).

  • Take down the swarm.

    docker swarm leave --force
    

    This removes the swarm. The purpose of this is to have a clean slate in preparation for the next steps, where you create a swarm on virtual “Docker machines” starting from the same command shell. This way, there’ll be no confusion as to which Dockerized host you are using or which swarm you are working with.

It’s as easy as that to stand up and scale your app with Docker. You’ve taken a huge step towards learning how to run containers in production. Up next, you will learn how to run this app as a bonafide swarm on a cluster of Docker machines.

Note: Compose files like this are used to define applications with Docker, and can be uploaded to cloud providers using Docker Cloud, or on any hardware or cloud provider you choose with Docker Enterprise Edition.

On to “Part 4” »

Recap and cheat sheet (optional)

Here’s a terminal recording of what was covered on this page:

To recap, while typing docker run is simple enough, the true implementation of a container in production is running it as a service. Services codify a container’s behavior in a Compose file, and this file can be used to scale, limit, and redeploy our app. Changes to the service can be applied in place, as it runs, using the same command that launched the service: docker stack deploy.

Some commands to explore at this stage:

docker stack ls                                            # List stacks or apps
docker stack deploy -c <composefile> <appname>  # Run the specified Compose file
docker service ls                 # List running services associated with an app
docker service ps <service>                  # List tasks associated with an app
docker inspect <task or container>                   # Inspect task or container
docker container ls -q                                      # List container IDs
docker stack rm <appname>                             # Tear down an application
docker swarm leave --force      # Take down a single node swarm from the manager
services, replicas, scale, ports, compose, compose file, stack, networking