Alarms Setting on AWS Cloudwatch

Question

I'm trying to set up alarms on AWS Cloudwatch and would like to someone experienced to help me with "default" settings (eg 80% CPU utilization) so I can set up appropriate alarms for Cloudwatch.

The alerts settings I need are for - EC2 - RDS - Elasticache - ELB - EBS

Would be good if you can also briefly explain impact of crossing threshold for each metric and categorize alerts under Priority 1, 2 and 3?

Thanks for your help.

Answer 1

From personal experience I tend to go with first warnings at 80% and higher warnings at 90% but a lot depends on how quickly things change. I would test and then make alterations as needed.

If you are using Autoscaling on EC2 I'd recommend scaling at more like 70% or 80% as instances take a time to startup and some servers don't like running at higher CPU like 90% and degrade badly.

Answer 2

I'd do the following:

• Priority 1 (>= 90% Alert)
• Priority 2 (>= 85% Alert)
• Priority 3 (>= 75% Alert)

If it is an AutoScale Alert, I'd use the Priority 3 check (>= 75% Alert) to begin scaling out. It gives you time to avoid getting to a Priority 1 or 2.