Hi 👋

We've just started exploring Buildkite and I wonder if there's a roadmap available somewhere?
I am wondering if you plan to add more features to the plugin like rollback (on top of what ECS provides when the service rollout fails) and rollout strategies?

ECS introduced multiple target groups for a service and I now have a use case for it but this plugin doesn't currently support an array of target groups.

I've had a look through the code and it looks like I could add this support quite simply with existing functionality but would you accept a PR?

There is an error with this repository's Renovate configuration that needs to be fixed. As a precaution, Renovate will stop PRs until it is resolved.

Error type: Preset name not found within published preset config (monorepo:babel6). Note: this is a nested preset so please contact the preset author if you are unable to fix it yourself.

We have a ECS/fargate service that's managed by terraform. The terraform config includes attaching the service to a target group so it receives traffic from an ALB.

I attempted to use this plugin to deploy a new task definition to the service, and initially I left the target group out of the plugin config as I knew the "create service" code path in the plugin wouldn't be used:

    plugins:
      - ecs-deploy#v1.4.1:
          cluster: "cluster-name"
          service: "my-service"
          task-definition: "deploy/task-container-definitions.json"
          task-family: "thing"
          image: "${ECR_REPO}:${BUILDKITE_BUILD_NUMBER}"
          execution-role: "${ECS_EXECUTION_ROLE_ARN}"

However, deploys failed with this error:

Cannot update a service to add/remove a load balancer. First delete the service
and then run again, or rename the service to force a new one to be created.
Container config differs

I then assumed I might need to include the target group config, even though it wouldn't be used, so I tried this config:

    plugins:
      - ecs-deploy#v1.4.1:
          cluster: "cluster-name"
          service: "my-service"
          task-definition: "deploy/task-container-definitions.json"
          task-family: "thing"
          image: "${ECR_REPO}:${BUILDKITE_BUILD_NUMBER}"
          execution-role: "${ECS_EXECUTION_ROLE_ARN}"
          target-container-name: web
          target-container-port: 3000

However, deploys failed with this error:

Registering new task definition for thing
Registered thing:16
🚨 Error: The command exited with status 1

To confirm the agent running the step has the necessary IAM permissions, I tried a minimal deploy shell script with some sugar around the following commands and it seems to work:

aws ecs register-task-definition --cli-input-json ...
aws ecs update-service ...
aws ecs wait services-stable ...

The test stubs as currently exist are not actually validating the full parameters given, which means that things like expected_container_definition don't really expect anything. If the test behavior is fine as desired, the test file should be cleaned up, otherwise something needs to be corrected with bats-mock. I encountered this issue when updating specs for #62

Example:

Current Contents of "Run a deploy with multiple images" stub:

  stub aws \
    "ecs register-task-definition --family hello-world --container-definitions '\'$expected_container_definition\'' : echo '{\"taskDefinition\":{\"revision\":1}}'" \
    "ecs describe-services --cluster my-cluster --service my-service --query 'services[?status==\`ACTIVE\`].status' --output text : echo '1'" \
    "ecs describe-services --cluster my-cluster --services my-service --query 'services[?status==\`ACTIVE\`]' : echo 'null'" \
    "ecs update-service --cluster my-cluster --service my-service --task-definition hello-world:1 : echo ok" \
    "ecs wait services-stable --cluster my-cluster --services my-service : echo ok" \
    "ecs describe-services --cluster my-cluster --service my-service : echo ok"```

Tests also pass with the following (note the munging of the first stub in each):

  stub aws \
    "ecs 1register-task-definition --family hello-world --container-definitions '\'$expected_container_definition\'' : echo '{\"taskDefinition\":{\"revision\":1}}'" \
    "ecs describe-services --cluster my-cluster --service my-service --query 'services[?status==\`ACTIVE\`].status' --output text : echo '1'" \
    "ecs describe-services --cluster my-cluster --services my-service --query 'services[?status==\`ACTIVE\`]' : echo 'null'" \
    "ecs update-service --cluster my-cluster --service my-service --task-definition hello-world:1 : echo ok" \
    "ecs wait services-stable --cluster my-cluster --services my-service : echo ok" \
    "ecs describe-services --cluster my-cluster --service my-service : echo ok"

  stub aws \
    "ecs register-task-definition : echo '{\"taskDefinition\":{\"revision\":1}}'" \
    "ecs describe-services --cluster my-cluster --service my-service --query 'services[?status==\`ACTIVE\`].status' --output text : echo '1'" \
    "ecs describe-services --cluster my-cluster --services my-service --query 'services[?status==\`ACTIVE\`]' : echo 'null'" \
    "ecs update-service --cluster my-cluster --service my-service --task-definition hello-world:1 : echo ok" \
    "ecs wait services-stable --cluster my-cluster --services my-service : echo ok" \
    "ecs describe-services --cluster my-cluster --service my-service : echo ok"

  stub aws \
    "argity blar blar --container-definitions '\'$expected_container_definition\'' : echo '{\"taskDefinition\":{\"revision\":1}}'" \
    "ecs describe-services --cluster my-cluster --service my-service --query 'services[?status==\`ACTIVE\`].status' --output text : echo '1'" \
    "ecs describe-services --cluster my-cluster --services my-service --query 'services[?status==\`ACTIVE\`]' : echo 'null'" \
    "ecs update-service --cluster my-cluster --service my-service --task-definition hello-world:1 : echo ok" \
    "ecs wait services-stable --cluster my-cluster --services my-service : echo ok" \
    "ecs describe-services --cluster my-cluster --service my-service : echo ok"

This one finally fails:

  stub aws \
    "what is this: echo '{\"taskDefinition\":{\"revision\":1}}'" \
    "ecs describe-services --cluster my-cluster --service my-service --query 'services[?status==\`ACTIVE\`].status' --output text : echo '1'" \
    "ecs describe-services --cluster my-cluster --services my-service --query 'services[?status==\`ACTIVE\`]' : echo 'null'" \
    "ecs update-service --cluster my-cluster --service my-service --task-definition hello-world:1 : echo ok" \
    "ecs wait services-stable --cluster my-cluster --services my-service : echo ok" \
    "ecs describe-services --cluster my-cluster --service my-service : echo ok"

Is this plugin support for ecs fargate?

Heya, is it possible to not have to include tasks in your source code and instead fetch them remotely? I have one that's defined in ECS and would like to use it

When using the following step:

- label: Deploy
    concurrency_group: "cicd-eval-deploy"
    concurrency: 1
    plugins:
      - ecs-deploy#v3.0.0:
          cluster: "cicd-eval"
          service: "cicd-eval-buildkite"
          task-family: "cicd-eval-buildkite"
          image: "xxx.dkr.ecr.us-west-2.amazonaws.com/cicd-eval:${BUILDKITE_BUILD_NUMBER}"

deployment stalls:

Updating service for cicd-eval-buildkite | 4m 20s
-- | --
  | {
  | "service": {
  | "serviceArn": "arn:aws:ecs:us-west-2:xxx:service/cicd-eval/cicd-eval-buildkite",
  | "serviceName": "cicd-eval-buildkite",
  | "clusterArn": "arn:aws:ecs:us-west-2:xxx:cluster/cicd-eval",
  | "loadBalancers": [],
  | "serviceRegistries": [],
  | "status": "ACTIVE",
  | "desiredCount": 1,
  | "runningCount": 1,
  | "pendingCount": 0,
  | "capacityProviderStrategy": [
  | {
  | "capacityProvider": "FARGATE",
  | "weight": 1,
  | "base": 0
  | }
  | ],
  | "platformVersion": "LATEST",
  | "platformFamily": "Linux",
  | "taskDefinition": "arn:aws:ecs:us-west-2:xxx:task-definition/cicd-eval-buildkite:8",
  | :

This is caused by paginated output of task definition json, to mitigate I've added:

[default]
cli_pager=

to ~/.aws/config of buildkite-agent user home but this should be addressed by adding --no-paginate to all AWS CLI commands.

While trying to use this plugin I'm getting an error:

https://buildkite.com/sj26-test/buildkite-ecs-oidc-example/builds/28#01870c6d-495b-4747-9afb-2382a87f3c4a

~/.buildkite-agent/plugins/github-com-buildkite-plugins-ecs-deploy-buildkite-plugin/hooks/command: line 147: aws_default_args[@]: unbound variable
🚨 Error: The command exited with status 1
user command error: The plugin ecs-deploy command hook exited with status 1

Seems to be around here:

https://github.com/buildkite-plugins/ecs-deploy-buildkite-plugin/blob/master/hooks/command#L144-L148

I can see that it's defined earlier in the file, and so should be available:

https://github.com/buildkite-plugins/ecs-deploy-buildkite-plugin/blob/master/hooks/command#L35

My bash-fu is weak, though, so is this something weird to do with arrays? I remember something about bash 5 causing issues?

I'm running an agent on macOS:

bash --version
GNU bash, version 5.2.2(1)-release (aarch64-apple-darwin21.6.0)
Copyright (C) 2022 Free Software Foundation, Inc.
License GPLv3+: GNU GPL version 3 or later <http://gnu.org/licenses/gpl.html>

This is free software; you are free to change and redistribute it.
There is NO WARRANTY, to the extent permitted by law.

When I deploy to fargate instance it throws:

Cannot update a service to add/remove a load balancer. First delete the service and then run again, or rename the service to force a new one to be created. Container config differs | 0s

This is overkill for most folks who just want to deploy new image. Also why is the json require in buildkite different to the one on AWS? I try to copy task definition JSON as "task-definition" and it complained that format is different. Do you expect customer we redefine new definition under your format?

This issue lists Renovate updates and detected dependencies. Read the Dependency Dashboard docs to learn more.

Open

These updates have all been created already. Click a checkbox below to force a retry/rebase of any.

• Update buildkite plugin docker-compose to v4.7.0

Detected dependencies

• Check this box to trigger a request for Renovate to run again on this repository

Full error: /var/lib/buildkite-agent/plugins/github-com-buildkite-plugins-ecs-deploy-buildkite-plugin-v2-0-0/hooks/command: line 136: env_vars[@]: unbound variable

  - label: ":ecs: :rocket:"

    concurrency_group: "my-service-deploy"
    
    concurrency: 1
    
    plugins:
    
      - ecs-deploy#v1.4.1:
      
          cluster: "<existing cluster>"
          
          service: "<non-existent service>"
          
          task-definition: "deploy/containers.json"
          
          task-family: "<task family>"
          
          image: "<arn>/<image>:latest"
          
          target-group: "<existing targetgroup-arn>"
          
          target-container-port: 3000
          

containers.json

[
{ 
"essential": true,
    
    "image": "<taskdef arn>",
    
    "memory": 1024,
    
    "name": "sample",
    
    "portMappings": [
    
      {
        "containerPort": 3000,
        
        "hostPort": 3000
        
      }
      
    ]
  }
]

I have tried pointing to a different service without a target group and ALB in front of it and still get the same sort of issue. I think Im missing some information on the launching of the fargate containers.

Would it be possible to return specific feedback from ECS in the case that the task failed to launch (i.e. could not find image, etc)?

When running a service update that fails, the logs will return Service failed to deploy (image attached).

Having this directly in the Buildkite logs would be very handy as it may negate needing to go into the AWS console & dig around as to why the deployment failed.

In the readme we have The ECS service must have been created before using this plugin. but this doesn't seem to be true 🤔

Hi,
Does this plugin support attaching a service to multiple target groups?

Thanks

Denis