A migration cutover is a bad time to find out that the Availability Zone you planned for has no spare capacity. RDS capacity reservations hold DB instance capacity for a specific engine, instance class and Availability Zone, so the capacity is there when you need it. In this walkthrough, we set one up for an RDS MySQL test database, attached a database to it, and cancelled it after the cutover. Every screenshot comes from our own lab account in us-east-1, using MySQL on db.t3.micro.
Capacity reservations reduce one specific risk: running out of capacity in the zone you planned to use. They do not lower your bill, and they do not guarantee a failover on their own. The limits section below covers what they can and cannot do.
At a glance: capacity reservations vs reserved instances
In the Aurora and RDS console, Capacity reservations has its own page in the left menu, separate from Reserved instances. The two solve different problems.
Before you start
Four settings decide whether a database can use a reservation: the engine, the DB instance class, the license model (for engines that offer more than one), and the Availability Zone. They must match exactly. Plan them before you create the reservation, because the console restricts editing while the reservation is in use. MySQL is one of the engines covered on our MySQL database page, so the steps below use it throughout.
Check your On-Demand DB instance quota as well. Reserved capacity counts once for every Availability Zone you reserve in. One instance reserved across two zones uses two units of quota.
Creating a capacity reservation
Step 1: Open the Capacity reservations page
In the Aurora and RDS console, choose Capacity reservations in the left menu. Our account had none, so the page showed an empty list and a Create capacity reservation button.

Step 2: Choose the engine and instance class
Pick the engine (MySQL in our test) and the DB instance class (db.t3.micro). A warning on the page says that editing is restricted while the reservation is in use, so get these right the first time.
Step 3: Name it, pick the Availability Zone and set the quantity
We named the reservation capacity-reservation-2, chose us-east-1a, and reserved one instance in that zone.
Step 4: Set the start and end times
A reservation can start immediately or at a set time. It can also end at a set time, so billing stops on its own. The console estimated the cost at $0.41 per day for this configuration.

Step 5: Confirm the reservation is active
The reservation became active straight away. It appears in the list with its name and a dbcr- ID. A new database attaches to that ID.
Attaching a database to the reservation
Step 6: Create the database from the reservation
Choose Create database from the reservation. The form is tied to that reservation, and only the matching engine, MySQL, is available to select.
Step 7: Check the reservation in the database settings
The capacity reservation carries into the new database's configuration, alongside the usual settings.

Step 8: Confirm the database landed in the reserved slot
database-1 was created in us-east-1a on db.t3.micro, which matches the reservation exactly.
Step 9: Check the reservation in use
The reservation now shows 0 of 1 capacity available, with database-1 listed under Connected databases. The database runs on the reserved capacity.
Cancelling after the cutover
Step 10: Cancel the reservation
After the cutover, cancel the reservation so you stop paying to hold capacity. Its status changes to Cancelled. Cancelling does not stop a running database. Any DB instance still running in the reservation continues as a normal On-Demand instance.
Step 11: Confirm nothing is left reserved
With cancelled reservations hidden, the list was empty again.
Step 12: Clean up the lab
We deleted the test database. In a real migration, the database would keep running at normal On-Demand rates after you cancel the reservation.
Gotchas that cost time
- A mismatch fails the call. If the engine, class, license model or Availability Zone differs from the reservation, the create or modify request fails with an error saying that zone does not have enough capacity. Check for a mismatch first, then for a real capacity shortage.
- A modify request needs a class change. To move an existing instance into a reservation, the same request must also change the instance class, the Multi-AZ setting or the license model. A request that only names the reservation fails.
- Multi-AZ needs two zones. To place a Multi-AZ instance, reserve in at least two Availability Zones, so that both the primary and the standby have capacity. Our high availability and disaster recovery team can review that layout with you.
- Reserve in every zone you may need. A database in a zone you did not reserve cannot use the reservation, even in the same Region.
Limits to know
Where it fits in a DBA's day
A capacity reservation belongs in the migration runbook. Add it before the cutover window, check its state the day before, and put the cancel step in the post-cutover checklist. A future-dated reservation can hold capacity for a fixed cutover date, but remember the commitment. Watch the cost estimate too: a reservation you forget to cancel keeps billing even when no database uses it. For the wider move, our cloud database migration and database migration services teams can plan the full runbook, and our DBA managed services cover the work after the cutover. For more on AWS work, see our AWS partner page.
Wrapping up
Capacity reservations are simple to set up, and they remove one risk from a cutover: finding no capacity in the zone you planned for. Test the steps in a lab first, and check the limits against your own engine and instance class. If you are planning a migration, talk to us through our contact page, or start with a free database audit to see where your databases stand.
Frequently Asked Questions
1. Am I charged when no database uses a reservation?
Yes. Billing starts when an active reservation is created and continues until you cancel it or it expires. Cancel it right after the cutover, or set an end time when you create it.
2. Can I use a reservation for an engine other than MySQL?
Yes, but each reservation is tied to one engine. Create a separate reservation for each engine and instance class you need.
3. Does a reservation protect a Multi-AZ failover?
It helps, but only if you reserve capacity in both the primary and the standby zones. A Multi-AZ instance needs capacity in both.
4. Can I use a reservation for cross-Region disaster recovery?
No. A reservation applies to one Region and the zones you choose in it. For DR in another Region, create a reservation there.
5. What happens to a running database when I cancel its reservation?
It keeps running as a normal On-Demand instance. You stop paying for the held capacity, not for the database.
6. Does a reservation hold storage?
No. It holds compute capacity for the instance class only, not storage or I/O.
Related reading
- Zero-downtime database migration strategies, challenges and best practices
- Database migration checklist
- Database migration testing
- Designing a reliable MySQL backup and point-in-time recovery strategy with Percona XtraBackup
- MySQL backup and recovery strategies, webinar
Sources: AWS, Amazon RDS User Guide and Aurora User Guide, On-Demand Capacity Reservations (docs.aws.amazon.com/AmazonRDS/latest/UserGuide/capacity-reservations.getting-started.html).


.avif)