Skip to content

Update an AWS ECS task definition with Docker image and trigger blue/green deployment

Notifications You must be signed in to change notification settings

valiton/ecs-task-deploy

 
 

Repository files navigation

ECS Task Deploy

A script to increment an active task definition on ECS with an updated Docker image, followed by a service update to use it.

Originally forked from mikestead/ecs-task-deploy and extended with the following features:

  • preserves exetutionRoleArns in task definitions(was previously stripped out)
  • allows updating stand-alone task-definitions without service

What does it do?

Sequence of steps performed by the script:

  1. Download the active task definition of an ECS service.
  2. Clone it.
  3. Given the Docker image name provided, find any containers in the task definition with references to it and replace them with the new one. Docker tags are ignored when searching for a match.
  4. Register this new task definition with ECS.
  5. Update the service to use this new task definition, triggering a blue/green deployment.

How to use

npm i @valiton/ecs-task-deploy

Environment Varaiables

For all the container definitions found when looking up your defined image, you have the option to add or update environment variables for these.

This can be done using the --env option.

--env "SERVICE_URL=http://api.somedomain.com"

You can supply as many of these as required.

Spare Capacity

In order to roll a blue/green deployment there must be spare capacity available to spin up a task based on your updated task definition. If there's not capacity to do this your deployment will fail.

This can be done via ECS service minimum healthy percent but as a brute force option this tool supports --kill-task. This will attempt to stop an existing task, making way for the blue/green rollout.

If you're only running a single task you'll experience some down time. Use at your own risk.

Usage

ecs-task-deploy [options]

Options:

-h, --help                output usage information
-V, --version             output the version number
-k, --aws-access-key <k>  aws access key, or via AWS_ACCESS_KEY_ID env variable
-s, --aws-secret-key <k>  aws secret key, or via AWS_SECRET_ACCESS_KEY env variable
-r, --region <r>          aws region, or via AWS_DEFAULT_REGION env variable.
-c, --cluster <c>         ecs cluster, or via AWS_ECS_CLUSTER env variable
-n, --service <n>         ecs service, or via AWS_ECS_SERVICE_NAME env variable
-d, --task-def <d>        task definition name, only needed when service and cluster are not given. can be defined via AWS_ECS_TASK_DEF env variable
-i, --image <i>           docker image for task definition, or via AWS_ECS_TASK_IMAGE env variable
-t, --timeout <t>         max timeout (sec) for ECS service to launch new task, defaults to 90s
-v, --verbose             enable verbose mode
-e, --env <e>             environment variable in "<key>=<value>" format
--kill-task               stop a running task to allow space for a rolling blue/green deployment

Hint: You can pass a service OR a task definition name. If a service is passed that service will be updated, otherwise only the given task definition.

CLI

To run via cli.

npm install -g ecs-task-deploy
ecs-task-deploy \
    -k 'ABCD' \
    -s 'SECRET' \
    -r 'us-west-1' \
    -c 'qa' \
    -n 'website-service' \
    -i '44444444.ddd.ecr.us-east-1.amazonaws.com/website:1.0.2' \
    -e 'SERVICE_URL=http://api.somedomain.com' \
    -e 'API_KEY=clientapikey' \
    -v

To run in code.

npm install @valiton/ecs-task-deploy

Node API

const ecsTaskDeploy = require('@valiton/ecs-task-deploy');

ecsTaskDeploy({
  awsAccessKey: 'ABCD',
  awsSecretKey: 'SECRET',
  region: 'us-east-1',
  cluster: 'cache-cluster',
  service: 'cache-service',
  image: 'redis:2.8',
  env: {
    SERVICE_URL: 'http://api.somedomain.com',
    API_KEY: 'clientapikey'
  }
})
.then(
  newTask => console.info(`Task '${newTask.taskDefinitionArn}' created and deployed`),
  e => console.error(e)
)

About

Update an AWS ECS task definition with Docker image and trigger blue/green deployment

Resources

Stars

Watchers

Forks

Packages

No packages published

Languages

  • JavaScript 97.8%
  • Dockerfile 2.2%