sudix / wercker-aws-ecs

Wercker step for AWS ECS

This wercker step allows to deploy Docker containers with AWS ECS service or run a task outside of a service.

Please read the AWS ECS documentation and API before using this step.

The step is written in Python 2.7 and use Pip and Boto3 module.

AWS ECS workflow

To deploy an application with AWS ECS, the Wercker step follow this steps:

There is two different flows depending if the Wercker step is running in "service mode" or in "task only mode". If service-name is provided in the configuration, the service mode is used.

Step Configuring AWS

This initial step consists on configuring AWS.

The following configuration allows to setup this step :

• key (required): AWS Access Key ID
• secret (required): AWS Secret Access Key
• region (optional): Default region name

Step Checking ECS Cluster

This step check a previously created ECS cluster exists.

The following configuration allows to setup this step :

• cluster-name (required): The name of the cluster to deploy the service

Step Checking ECS Service (Skipped in task only mode)

This step check a previously created ECS service exists. The service MUST be created before using this step.

The following configuration allows to setup this step :

• service-name (optional): The name of the service to deploy

Step Create New Task Definition

This step register a new task definition for the service.

The following configuration allows to setup this step :

• task-definition-name (required): The name of the task definition
• task-definition-file (required): The file containing the task definition

In service mode:

Step Downscale ECS Service

This step downscale the service in order to deploy the new revision.

See this thread for explanation :

The service scheduler will always scale up at least 1 new container before shutting down old containers. This means on an update, if your current service uses all resources in your cluster you may actually need to scale your service to N-1 where N is your current running task count for that service. For example, if you have 5 container instances and 5 tasks running bound to host port 80, in order to perform a rolling update you'll need to set the desired count on the service to 4. The scheduler will recognize that it can stop 1 container and still meet your needs. After stopping 1, the free space will be used to start another. You can think of desired count as also the minimum count for now and the scheduler won't remove tasks below that minimum which is likely why you see resources not found.

This step is run only if the number of tasks running is greater than the following configuration :

• minimum-running-tasks (optional default 2): The minimum number of running tasks expected

Step Update ECS Service

This step update the service with the new revision.

Step Upscale ECS Service

This step upscale the service to the initial number of tasks.

In task only mode:

Step Run task

This step run a task on a single node and wait for its completion. The step fail if the task can't be run or if its command exit code is different from 0.

Example

The following example deploy an hello service on ECS :

deploy:
  steps:
    - 1science/aws-ecs:
        key: aws_access_key_id
        secret: aws_access_secret_id
        cluster-name: staging
        task-definition-name: hello-migrate-db
        task-definition-file: /app/hello-migrate-db-task-definition.json
    - 1science/aws-ecs:
        key: aws_access_key_id
        secret: aws_access_secret_id
        cluster-name: staging
        service-name: hello
        task-definition-name: hello
        task-definition-file: /app/hello-task-definition.json