Define Automatic scaling policy

[xmunoz]

No description provided.

[cecilia-donnelly]

"What are the conditions that should be met for us to turn this on?"

[xmunoz]

Determining the proper thresholds will require data from monitoring to determine how the machines handle load.

Helpful documentation for how to set up scaling policies: https://docs.aws.amazon.com/autoscaling/ec2/userguide/as-scaling-simple-step.html

[xmunoz]

I think the scaling policy that we need will look at the number of messages in the various sqs queues, and use that to determine how much to scale. Unfortunately, this will require spinning up a new service that just reads and reports that data to Cloudwatch. That is what I will now be working on.

[xmunoz]

Docs: https://docs.aws.amazon.com/autoscaling/ec2/userguide/as-using-sqs-queue.html
Possibly useful repo: https://github.com/MeteoGroup/sqs-count-custom-metric

[xmunoz]

After further discussions, it doesn't look I'll have time to implement this design. It would effectively require spinning up a new service (per-enviroment) that just watches the SQS queues (probably just HIGH_PRIORITY and VIDEO) and triggers a Cloudwatch alarm when the queue lengths get too long. The scaling group will then spin up new instances accordingly.