Leo Giordani one of POP's developers has written up a detailed blog on his recent work creating a Cloudwatch metrics dashboard. The dashboard helps visualise our application's use of Celery tasks and to provide alerts where problems in the system occur such as tasks backing up as demand spikes.

Leo's post pretty much steps through the entire process involved including setting up a DynamoDB table in Terraform and creating the Python Lambda to does the work of converting the queries into time-series metrics.

This is a new technique for us and we're looking at how we might use this idea of polling, transforming and writing CloudWatch information more in the future.