Amazon RDS Costs: Stop Windows, Sizing and Reservations
RDS lets you stop a DB instance for up to 7 consecutive days. The rules around that window decide what a schedule can save.
How RDS charges
A provisioned RDS DB instance bills instance hours while it runs, plus provisioned storage (including Provisioned IOPS), backup storage and, for a publicly accessible instance, the public IPv4 address. Stopping removes only the instance-hour line. AWS’s stop guide is explicit that storage and backups keep charging while the database is down.
The rules AWS sets on stopping
- A DB instance can stay stopped for at most 7 consecutive days. After that RDS starts it automatically so it does not miss maintenance.
- An instance that has a read replica, or is one, cannot be stopped.
- RDS for SQL Server in a Multi-AZ deployment cannot be stopped.
- Starting requires instance recovery, which AWS says can take from minutes to hours.
A nightly and weekend schedule fits inside the 7-day window by design, because every weekday start resets it. A database parked for a month does not; that is a snapshot-and-delete decision, not a schedule.
What ZopNight does for RDS
RDS is on ZopNight’s Start/Stop supported list, so a schedule or a manual action can stop and start instances on an account connected with read + write access. Inside a resource group, start runs in tiers (storage, then databases, then compute, then apps) and stop runs in reverse, and you can add a per-resource delay so the database has time to accept connections before the apps start. Remediation never mutates the data inside a customer database; the pause template explicitly excludes RDS and Aurora.
RDS rules that apply to the instance type:
- RDS Instance Right-sizing and RDS Dev/Test Burstable Opportunity for over-provisioned classes.
- Idle RDS Instance, No Connections for databases nobody queries.
- RDS Multi-AZ on Non-Production Instance and RDS PITR Enabled on Non-Production for resilience spend outside production.
- RDS Excessive Backup Retention for long retention windows.
- RDS Reserved Instance Opportunity and RDS RI Coverage Below 50% for commitment gaps.
- RDS Non-Graviton Instance for older processor families.
- RDS Instance Publicly Accessible and RDS Instance Not Using Multi-AZ on the posture side.
Aurora clusters are a separate resource type with their own rules, including Idle Aurora Cluster, No Connections.
See the RDS fact sheet, RDS scheduling and the Aurora cluster schedule notes. For the reservation side, read Reserved Instances.
Questions we get a lot.
Why did my stopped RDS instance start by itself?
AWS starts any DB instance that has been stopped for seven consecutive days. A weekday schedule restarts it well before that; a long parking period needs a different approach.
Does ZopNight change data inside the database?
No. Its remediation pause template excludes RDS and Aurora, and scheduling uses the RDS stop and start operations only.