New platform branch to support ECS-based docker running on AL2

[equivalent]

Community Note

• Please vote on this issue by adding a 👍 reaction to the original issue to help the community and maintainers prioritize this request.
• Please do not leave "+1" or "me too" comments. They generate extra noise for issue followers and do not help prioritize the request.

Tell us about your request

in Multicontainer Docker platform (Amazon Linux AMI) there was an option to provide essential in Dockerrun.aws.json so that entire EB docker run script re triggers if essential container fails.

Example :

{
  "AWSEBDockerrunVersion": 2,

  "containerDefinitions": [
    {
      "name": "background-worker",
      "image": "my-company/my-bg-worker-app",
      "essential": true
    },
    {
      "name": "health_check_nginx",
      "image": "equivalent/health_check_nginx",
      "portMappings": [
        {
          "hostPort": 80,
          "containerPort": 80
        }
      ]
    }
  ]
}

form health_check_nginx_docker

Now that I've switched to Docker Amazon Linux 2 and docker-compose.yml there is no option for me to specify which containers are essential (docker steps for EB should restart)

Yes docer-compose.yml provides depends_on but this will not prompt EB to restart the cycle to try re-up the containers as it was in Multicontainer Docker Platform with Dockerrun file

---
version: '3'
services:
  healthcheck:
    image: equivalent/health_check_nginx
    ports:
      - '80:80'
    depends_on:
      - worker
  worker:
    image: myprivateapplicationworker
    environment:
      ROLE: 'worker'
    env_file:
      - .env

Nothing will happen if worker container dies

thing to consider

one way to solve this globally for all EB Linux 2 platforms would be to use docker-compose up --abort-on-container-exit (as mentioned in this stack overflow issue ) but that may not be the best for all use cases of all EB applications

[equivalent]

ok here is another scenario that makes EB Docker Amazon Linux 2 inferior for production apps due to missing the essential option

given I have docker-compose.yml file with just one container running
when I kill that container (E.g. memory issue)
then instance will not restart the container

with Multicontainer Docker Platform if I specify "this docker container is essential" if it dies the EB agent would try to restart it which helps a lot if dev-ops person is not available immediately when error happens (many times this feature saved me during a holiday). In my opinion this was the best feature of AWS EB.

[AvihaiSam]

ok here is another scenario that makes EB Docker Amazon Linux 2 inferior for production apps due to missing the essential option

given I have docker-compose.yml file with just one container running
when I kill that container (E.g. memory issue)
then instance will not restart the container

with Multicontainer Docker Platform if I specify "this docker container is essential" if it dies the EB agent would try to restart it which helps a lot if dev-ops person is not available immediately when error happens (many times this feature saved me during a holiday). In my opinion this was the best feature of AWS EB.

I've seen that on AL1 as well...
usually I open a ticket once a month for that...

[1tonyca]

Thanks for bringing this back to our attention.

For customers who prefer to use docker-compose to ensure container running after it was killed, please consider adding restart policy in your docker-compose.yml, more details are here

For customers who prefer essential parameter that is supported by ECS, we are currently working on a new platform branch to support ECS-based docker running on AL2, so it would be simpler port for environments running on Multicontainer docker platform

[1tonyca]

https://docs.aws.amazon.com/elasticbeanstalk/latest/relnotes/release-2022-03-25-ecsal2.html