geekcell / github-action-aws-ecs-run-task Goto Github PK

The current full example task fails as /bin/sh -c only accepts a single argument.

The current example

override-container-command: |
  /bin/sh
  -c
  php artisan migrate --force --ansi && \
  php artisan db:seed --force --ansi && \
  php artisan config:clear --ansi

will produce an ecs run-task equivalent of:

 aws ecs run-task ... \
  --overrides '{ "containerOverrides": [ { "name": "foo", "command": ["/bin/sh", "-c", "php artisan migrate --force --ansi && \\", "php artisan db:seed --force --ansi && \\", "php artisan config:clear --ansi"] } ] }'

which executes the first argument (artisan migrate), but produces line 0 errors on the other commands.

It would be great to support a shell command helper that merges the [multiline] commands into a single argument (when they trail with a \?). For now, I think we must do everything on a single line, e.g.

override-container-command: |
  /bin/sh
  -c
  php artisan migrate --force --ansi && php artisan db:seed --force --ansi && php artisan config:clear --ansi

and should update the example?

I'm trying to use your Github action in a project, as it makes executing ECS tasks very very simple. It works like a charm for short (a couple minutes) ECS tasks, but I'm running into issues when I use it to run ECS task that takes longer to execute (10+ minutes).

ECS task gets started correctly by the Github action, runs till completion (I can see ECS logs output from github action) and then changes its status to Stopped in AWS Console. The Github action execution however just hangs there in running state forever (well, I've waited 20 minutes for it to complete and then cancelled the workflow). For shorter tasks the action execution completes as soon as ECS task goes into Stopped status.

Any advice how to get the action to work with longer ECS tasks?

Task configuration

   - name: Run command
      uses: geekcell/[email protected]
      with:
        cluster: clusterXYZ
        task-definition: taskXYZ
        assign-public-ip: 'DISABLED'
        subnet-ids: subnet-123
        security-group-ids: sg-123
        tail-logs: true
        override-container: Container
        override-container-command: ${{ inputs.command }}
        override-container-environment: |
          NODE_ENV=dev

Cluster/task definitions:

    const image = ecs.EcrImage.fromEcrRepository(repository, 'latest');

    const vpc = ec2.Vpc.fromLookup(this, 'VPC', {
      vpcId: 'vpc-123',
    });

    new ecs.Cluster(this, 'Cluster', {
      clusterName: 'clusterXYZ',
      vpc,
    });

    const taskDefinition = new ecs.FargateTaskDefinition(this, 'TaskDefinition', {
      family: 'taskXYZ',
      memoryLimitMiB: 16384,
      cpu: 4096,
      taskRole: new Role(this, 'TaskRole', {
        ...
      }
   );

    taskDefinition.addContainer('Container', {
      image: image,
      environment: { NODE_ENV: 'dev' },
      logging: new ecs.AwsLogDriver({
        streamPrefix: 'prefixXYZ',
        mode: ecs.AwsLogDriverMode.NON_BLOCKING,
      }),
    });

I have some really long running ones. Yours is the only plugin that works. I turned logs off -- will that just exec the task and go?