Example ECS side-car containers.js

{
    // The standard ECS options, place in your own values:
    "cluster": "",
    "serviceName": "my-side-car-container",
    "taskDefinition": "",
    "role": "",

    // desireCount should be greater than the number of instances that are in the ECS cluster.
    "desiredCount": 1000,

    // These two configuration options may be used, but typically you'll deploy these containers once,
    // and then not update them frequently, unless you're updating the entire ECS instance.
    "deploymentConfiguration": {
        "maximumPercent": 0,
        "minimumHealthyPercent": 0
    },
    
    // Ensure that the container instances are spread evenly across multiple availability zones.
    // Not really a concern if you're only running in a single AZ, but you should NOT be doing that.
    "placementStrategy": [{
        "field": "attribute:ecs.availability-zone",
        "type": "spread"
    }],
    
    // Ensure that only one container instance can run on a given ECS instance at any one time,
    // Anymore than one would result in the ECS instance having it's metrics double reported.
    "placementConstraints": [{
        "type": "distinctInstance"
    }]
}

Intro.md

For a former client, I needed to be able to deploy the docker-stats-statsd image to every single instance in my ECS cluster, however, I couldn't have more than one instance of the collector running per ECS instance; Additionally, I needed to be able to scale up and down the number of ECS instances without worrying whether or not they would be monitored. This is roughly what I used, if I recall correctly — there is certainly better ways to do this, and I highly encourage you look at Kubernetes instead of ECS, as it has better support for this sort of monitoring container: https://kubernetes.io/docs/user-guide/monitoring/ & https://kubernetes.io/docs/admin/daemons/