Other Alarms to Consider
Auto scaling is not a panacea for continually running your application in an optimum state. It is too slow to cope with massive spikes in load and adjust quickly enough and be efficient with resources. It is good at gradually scaling out as load gradually increases and the opposite, scaling in as load gradually decreases.
So you need to be warned when spikes in load are occurring. Some alarms to consider are
- EC2 CPU utilization sustained over 90% for maybe 5 mins and send an email
- If expected to be sustained, manually add desired instances to Web Server ASG.
- RDS CPU utilization sustained over 90% for maybe 30 mins and send an email
- If expected to be sustained, plan a time to scale up the RDS instance.
- Watch Load balancer error rate and web server error rates
- Troubleshoot cause. Likely to be solved by more EC2 instances or scaling up the RDS.