Scale Fargate service tasks to match CloudWatch metric - amazon-cloudwatch

Question

I'm using CloudWatch Metric Math to calculate the number of workers (tasks) that I want my Fargate service to be scaled to. I planned on creating an alarm in CloudWatch to trigger the scaling action once it rose above or below the target number of 0. However, it doesn't look like there a way I can create an alarm based on CloudWatch Metric Math - or an alarm that does any type of comparison between two numbers (number of tasks needed vs. number of tasks existing).
How can I setup a Fargate scaling policy to scale based on my existing metric of 'Workers Needed'.
Metric Math
Formula: m1-m2-3 == desired scale offset
m1: Active Workers (tasks)
m2: Workers Needed (tasks)
{
"type": "metric",
"x": 0,
"y": 0,
"width": 24,
"height": 6,
"properties": {
"metrics": [
[ { "expression": "m1-m2-3", "label": "Workers/Needed difference", "id": "e1" } ],
[ "AWS/ECS", "MemoryUtilization", "ServiceName", "worker-service", "ClusterName", "my-cluster", { "period": 60, "stat": "SampleCount", "id": "m1", "label": "Active Workers" } ],
[ "LogMetrics", "Workers Needed", { "period": 60, "stat": "Maximum", "id": "m2" } ]
],
"view": "timeSeries",
"stacked": false,
"region": "us-east-1",
"title": "Worker/Lab difference",
"period": 300
}
}

Answer 1

Edit: Alarms based on metric math is now a thing
This doc page about metric math doesn't mention alarms at all, expressions seem to be more about visualizing with a dashboard. I also don't see anything about metric math in the SDK or cli documentation as far as alarms are concerned.
Your next simplest solution is probably paying homage to the great catch-all of all shortcomings of AWS, and write a Lambda that pulls the metrics, does the calculation, then publishes the metric as a custom metric with PutMetric. You can trigger this with a CloudWatch Event if you want a cron-like thing, or many, many other things by integrating it with SNS or just invoking it directly.
It's not the answer you want, but unfortunately I think it's the simplest way to get the functionality you want.