Why it matters
Reservations are often created for a launch or a migration and left in place afterwards. The unused part costs the same as a running instance.How nOps finds it
nOps reads the last 30 days of EC2 billing for unused capacity reservation charges (shared and dedicated). A reservation is flagged when those charges were at least $5 over the 30 days, appeared on at least 25 of them, and were still billed in the last 2 days of data. The saving is the unused capacity charge for the 30 days.Before you start
Permission to cancel or modify capacity reservations (ec2:CancelCapacityReservation, ec2:ModifyCapacityReservation).
How to implement
- Check the reservation’s instance type, total and available count, and end date:
aws ec2 describe-capacity-reservations --capacity-reservation-ids ID. - Confirm no upcoming launch, failover plan or event depends on the guaranteed capacity.
- If some of it is used, reduce the count to what is running:
aws ec2 modify-capacity-reservation --capacity-reservation-id ID --instance-count N. - If none of it is needed, cancel it (EC2 console: Capacity Reservations → Actions → Cancel, or
aws ec2 cancel-capacity-reservation --capacity-reservation-id ID). - Remove it from Terraform or CloudFormation if it is defined there.
Risks and considerations
- A cancelled reservation cannot be restored, and the same capacity may not be available if you reserve it again.
- Keep it if a launch, scaling event or disaster-recovery plan needs guaranteed capacity in that Availability Zone.
FAQs
Do Savings Plans or Reserved Instances cover unused capacity?
Do Savings Plans or Reserved Instances cover unused capacity?
They can lower the rate, but the unused capacity is still billed. Cancelling or shrinking the reservation removes the charge.