Can an Aurora Global Database secondary cluster exist without any DB instances for cost optimization?Hi AWS experts, I have an Amazon Aurora MySQL Global Database configured as follows: Primary Region

Question

Hi AWS experts,
I have an Amazon Aurora MySQL Global Database configured as follows:
Primary Region: Mumbai (ap-south-1)
Secondary (DR) Region: Hyderabad (ap-south-2)
The secondary cluster currently has 3 reader (Aurora Replica) instances. I want to to optimize AWS costs because the DR region is not actively used except during disaster recovery or occasional testing.
I understand that:
Aurora Global Database requires a secondary cluster for cross-region replication.
A DB instance cannot be stopped like a standard RDS instance.
However, I would like clarification on the following points:
Can I delete all DB instances from the secondary cluster while keeping the Global Database configuration intact?
In other words, can the secondary cluster exist with zero DB instances?
If not, what is the minimum requirement?
Is at least one DB instance mandatory in the secondary region for replication?
For cost optimization, what are the AWS-recommended best practices?
Reduce the secondary cluster from 3 instances to 1?
Downsize the remaining instance to a smaller class?
Are there any other recommended approaches?
If the business no longer requires a DR region, is removing the secondary cluster (thereby removing the Global Database) the only way to eliminate the compute cost in the secondary region?
Are there any limitations or risks associated with downsizing the secondary instance, particularly regarding replication lag, failover readiness, or Global Database recovery?
I'm looking for AWS best practices and architectural guidance rather than application-specific recommendations.
Thank you in advance for your help.

re:Post Agent · Answer

Yes, an Aurora Global Database secondary cluster can exist without any DB instances, and this is actually a supported cost optimization strategy called a 'headless' secondary cluster configuration.

With a headless secondary cluster, you maintain cross-Region disaster recovery capability without paying for compute resources in the secondary Region. Since Aurora decouples compute and storage, when you remove all DB instances from the secondary cluster, you're only charged for storage while the storage layer continues replicating data from the primary Region with typically less than 1 second of lag.

To create a headless secondary cluster, you would delete all reader instances from your secondary Aurora DB cluster in Hyderabad. The cluster remains part of the Aurora Global Database and continues receiving replicated data. However, note that you cannot add new DB instances back to a headless cluster using the Amazon RDS console - you must use the AWS CLI or RDS API to provision instances when needed.

For your cost optimization question, AWS-recommended approaches include:

1. Using a headless secondary cluster (zero instances) if your Recovery Time Objective (RTO) allows for the time needed to provision instances during a disaster event
2. Reducing from 3 instances to 1 instance in the secondary Region
3. Downsizing to a smaller instance class based on your actual requirements
4. Considering Reserved Instances or Savings Plans if you maintain provisioned capacity
5. Evaluating Aurora I/O-Optimized configuration for the primary Region if you have high I/O requirements

Regarding your concern about removing the Global Database entirely: if you no longer need DR capability, removing the secondary cluster would eliminate compute costs. However, the headless configuration provides a middle ground - maintaining DR capability while minimizing costs.

One important consideration: when you need to activate your DR environment, you'll need to provision instances in the secondary Region. Be aware that there could be situations of insufficient capacity if the specific DB instance class isn't available in the requested Availability Zone at that time.  
**Sources**  
[Creating a headless Aurora DB cluster in a secondary Region - Amazon Aurora](https://docs.aws.amazon.com/AmazonRDS/latest/AuroraUserGuide/aurora-global-database-attach.console.headless.html)  
[Optimize costs in Amazon Aurora | AWS Database Blog](https://aws.amazon.com/blogs/database/optimize-costs-in-amazon-aurora/)  
[Deleting Aurora DB clusters and DB instances - Amazon Aurora](https://docs.aws.amazon.com/AmazonRDS/latest/AuroraUserGuide/USER_DeleteCluster.html)

Can an Aurora Global Database secondary cluster exist without any DB instances for cost optimization?Hi AWS experts, I have an Amazon Aurora MySQL Global Database configured as follows: Primary Region

Add your answer

Relevant content