When managing cloud infrastructure, minimizing costs without compromising data integrity is a continuous priority. Amazon Aurora Serverless v2 offers scalability and high availability, but unlike traditional RDS instances, it introduces nuances in how compute resources are billed. One common question arises: Can an Aurora Serverless v2 database be stopped to save costs?
Understanding Aurora Serverless v2 Billing
Aurora Serverless v2 is designed to scale compute capacity automatically based on application demand. It uses Aurora Capacity Units (ACUs) instead of fixed instance types, billing you per second based on actual usage. However, this also means that ACUs remain allocated even when the database is idle, unless you explicitly stop the cluster.
Stopping the cluster does not mean halting all costs. Instead, it allows you to pause the compute layer while storage and backup charges continue to apply.
How to Stop an Aurora Serverless v2 Cluster Temporarily
As of now, AWS allows you to stop an Aurora cluster for up to 7 days. After that, the cluster is automatically restarted. This feature can be used to suspend development or test environments during off-hours or weekends.
Steps to Stop the Cluster:
- Navigate to the RDS Dashboard in the AWS Management Console.
- Go to Databases and select your Aurora Serverless v2 cluster.
- Click on Actions > Stop temporarily.
- Confirm the acknowledgment: while compute costs are paused, storage and backup charges remain.
- Submit the stop request.
You will be informed of the automatic restart time, exactly 7 days after the stop operation.
What Happens During the Stop?
- Compute billing is paused: No charges for ACUs during the stopped period.
- Storage and backup charges continue: You are still billed for provisioned storage, IOPS (if applicable), automated backups, and manual snapshots.
- Connections are refused: The database is unavailable until it is manually or automatically restarted.
- No scaling events occur: Since compute is inactive, autoscaling is suspended.
Use Cases
- Development environments: Stop the cluster overnight or during weekends.
- Cost-controlled test labs: Pause resources when not under active testing.
- Training or demo environments: Suspend compute until the next session.
Limitations
- You cannot pause for more than 7 consecutive days.
- Aurora Global Databases and clusters with cross-region replication cannot be stopped.
- Aurora Serverless v1 offers automatic pause/resume based on idle time, but it does not support the same PostgreSQL versions as v2.
Final Considerations
Stopping an Aurora Serverless v2 cluster is a practical method to reduce compute-related costs temporarily. However, it does not eliminate all billing, and the benefits are most noticeable in non-production environments. For long-term cost reduction, consider taking a manual snapshot and deleting the cluster if it is not needed for extended periods.