sometime ago i was wondering how to get rid of some really old bash scripts that we were using to update ecs service. i had an idea to use AWS Lambda along with AWS EventBridge.

basically i wanted to trigger lambda everytime a new docker image was pushed to ECR and that function would update a service with new image. simply put, a new task definition would be created with that new image.

maybe some diagram will help to understand it a little bit better workflow

lambda code

import boto3

client = boto3.client('ecs')

def lambda_handler(event, context):
    repo_name = event.get("detail").get("repository-name")
    format_repo_name = ''.join(repo_name)    
    cluster='name of our cluster',

super simple stuff, we are using boto3 ecs client update

we need to provide the name of our cluster and service. we also force new deployment so we are sure that new task will be created.

i also wanted to make lambda more universal so i am retrieving a repository name from the eventBridge event and i pass it to service parameter. CAUTION: it will only work if your service name is the same as the name of the ECR repository

we also need to create a rule in the EventBridge. we simply need to go to AWS EventBridge console and click create rule. then, there are a few things that are important for us –>

we select an event pattern and then we need to configure our pattern pattern

  • service provider - AWS
  • service name - ECR
  • event type - ecr image action
  • specific action - push, as we want to trigger lambda only on push
  • specific result - success, as we don’t want to trigger lambda if push to ECR fails
  • specific repositories - we select which services EventBridge will observe

this whole configuration can be also done with

  "source": ["aws.ecr"],
  "detail-type": ["ECR Image Action"],
  "detail": {
    "action-type": ["PUSH"],
    "result": ["SUCCESS"],
    "repository-name": ["repository-1", "repository-2"]

then we want to select our lambda as a target for an event lambda target

and that’s basically it. you can test it by pushing new image to ECR.


  • imo much elegant and easier way to manage updates then some bash/python scripts
  • it works if your ECS task is using :latest image which is not generally good idea in production envs
  • devs don’t see update progress in a pipeline - they only see that image was pushed to ECR. they would need to check out AWS console or use AWS CLI. that could be mitigated a little bit by sending notification by lambda to Slack channel for example
  • won’t work if ECS service name is not the same as name of ECR repo