I define my infra in CDK and rely on CodeDeploy to deploy my lambdas with canaries. Each CodeDeploy deployment has a pre-traffic hook and a bunch of alarms linked to it. The CloudWatch alarms are like "lambda is low on memory" or "lambda has high error rate" or "lambda is running out of time" etc... there are quite a few.
I am now looking at a situation where my lambda is already crashing. The alarm is on. I have a new deployment that will fix the lambda. However, I cannot deploy it as CodeDeploy sees that the alarm is already on and rolls back my fix.
Do I need somehow dedicated alarms for the canary deployment? What's the right way to approach it?