The overnight failsafe
If the appliance stops running, nothing would switch machines off. The failsafe covers that.
- The appliance writes a heartbeat every five minutes.
- Every night (02:00 UTC unless changed in the CloudFormation stack) a small Lambda function checks it. If it's more than 30 minutes old, the function stops every running machine tagged
costcapper:managed=true, and emails the admin address given at install. - Tag a machine
scheduler-failsafe=exemptin AWS to keep the failsafe away from it.
The failsafe can only stop machines; it can't start them, and it never touches the appliance.