How to setup a continuous deployment from Bitbucket Pipeline to Amazon ECS

For one of my recent projects I had to setup a continuous deployment workflow for a web app, consisting of a dozen microservices. We wanted to use docker to build a image for each service. I already described in my previous article how you can build and push a docker image to Amazon ECR through a Bitbucket Pipeline. You can find this article here:

How to push Docker Images to Amazon ECR with Bitbucket Pipeline

In this article I want to describe the following points:

• How to create a new revision of an task definition in a Bitbucket Pipline
• How to update a service of a ECS cluster to this new revision

My Solution

How does the general workflow looks like?

Workflow of a deployment

1. Commit a new version

I would recommend to use consistent versions for your images, task definitions and services. The best way is to specify a version number in your pipeline or you read it from a file in your git repository. In this example we export an environment variable VERSION_NUMBER and for the target deployment stage the environment variable DEPLOYMENT_ENV. This two combined result in the image name, service name, etc.

In our example we will deploy version 1.0.0 of our microservice, to the “test” stage of the cluster.

# set version and build id
— export VERSION_NUMBER=1.0.0
— export DEPLOYMENT_ENV=test

2. Build & push the Docker image to the Amazon ECR

Please have a look here how to do this:

How to push Docker Images to Amazon ECR with Bitbucket Pipeline

3. Create a new revision of your task definition

First I created a JSON-File containing the container definition of our task definition. This could look like this:

[
  {
    "name": "mytestapp-webgui",
    "image": "xxx.dkr.ecr.eu-central-1.amazonaws.com/mytestapp-webgui:<IMGAGE_VERSION>",
    "cpu": 0,
    "memory": 500,
    "memoryReservation": 75,
    "portMappings": [],
    "essential": true,
    "entryPoint": [],
    "command": [],
    "environment": [
      {
        "name": "ENV_DEBUG",
        "value": "false"
      }
    ],
    "mountPoints": [],
    "volumesFrom": [],
    "workingDirectory": "/usr/share/nginx/html/",
    "disableNetworking": false,
    "privileged": true,
    "readonlyRootFilesystem": false,
    "dnsServers": [],
    "dnsSearchDomains": [],
    "dockerSecurityOptions": [],
    "logConfiguration": {
      "logDriver": "json-file"
    }
  }
]

I saved the file with the name “mytestapp-webgui.container-definition.json”

To create a new revision we will use the “aws ecs register-task-definition” command. We have to specify the task family and pass the content of the JSON file. The command could look like this:

aws ecs register-task-definition --family mytestapp-webgui —- container-definitions <json-file>

This command is included in the “deploy-ecs.sh” script below. I added some clue code in that script.

From the output of the “aws ecs register-task-definition” command we parse the revision number and store it in a variable. If this step was successfully we proceed with the next one.

4. Update the ECS cluster service

If we managed to create a new revision of our task definition, we can update our service. Therefore we use the “aws ecs update-service” command. We have to specify the cluster name, the service and the task definition incl. the revision. The command could look like this:

aws ecs update-service --cluster mytestapp-test-cluster --service mytestapp-test-webgui-service --task-definition mytestapp-webgui:1

Complete solution:

According to the documentation of Pipelines (Link), you can specify in each step one of the following deployment steps:

• test
• staging
• productive

If you don’t want to use this Pipelines feature, you are free to use any other name too.

Like I mentioned in my older post you have to allow all the needed action to your AWS IAM deploy user to be able to deploy your service. A list from my user:

"ecr:GetAuthorizationToken",
"ecr:BatchCheckLayerAvailability",
"ecr:GetDownloadUrlForLayer",
"ecr:GetRepositoryPolicy",
"ecr:DescribeRepositories",
"ecr:ListImages",
"ecr:DescribeImages",
"ecr:BatchGetImage",
"ecr:InitiateLayerUpload",
"ecr:UploadLayerPart",
"ecr:CompleteLayerUpload",
"ecr:PutImage",
"ecs:DeregisterTaskDefinition",
"ecs:RegisterTaskDefinition",
"ecs:UpdateService"

For deployment I wrote the “deploy-ecs.sh” script. The script will create a new revision of our task definition for you and update your service. You only have to specify following parameters:

• the name of your app (“mytestapp” in our case) [-b]
• the service you want to deploy (“webgui” in our case) [-s]
• the deployment environment (“test” in our case) [-e]
• the image name (“test_1.0.0” in our case) [-i]

PS: If you have an idea how to improve this setup, I would be happy If you leave a comment ;)