Delete Idle Aurora Read Replicas

Opportunity Name

RdsDeleteAuroraReadReplicas

AWS Resource Type

Amazon Aurora (RDS Read Replicas)

Opportunity Description

Amazon Aurora clusters often maintain multiple read replicas to support application read scaling and high availability. However, in many cases, these replicas are underutilized, particularly in non-production environments. This CloudFix Finder analyzes historical CPU usage of Aurora read replicas and recommends safe deletion of underutilized nodes — significantly reducing costs while preserving operational reliability.

Criteria for Identifying the Opportunity

This Finder evaluates:

• Clusters with more read replicas than minimally required (1 for production, 0 for non-production).

• Read replicas with low CPU utilization, calculated using a configurable CPU stat (P99, P95, or MAX).

• Predicted CPU load on remaining nodes after replica deletion — ensuring it remains within safe limits (default: 20% buffer).

• Accurate cost data from AWS CUR for each replica.

• Environments classified as production or non-production using integrated tagging or classifier service.

Only clusters where replica removal is safe and cost-effective are recommended.

Potential Savings

Savings are calculated based on the annualized amortized cost of the recommended read replicas for removal. Example:

Cluster Name	Replicas to Remove	Instance Type	Estimated Annual Savings
abcdbnonprod	1	db.r6g.large	$5,246

What Happens When the Fixer is Executed?

This is a Finder-only opportunity. CloudFix does not delete replicas automatically. Instead, it provides:

• A detailed report listing the read replicas recommended for removal.

• The projected impact on CPU load.

• Savings estimate, per-node breakdown, and environmental classification.

These insights help you manually adjust replica counts via the AWS Console or automation tools.

Is It Possible to Roll Back Once CloudFix Implements the Fixer?

Not applicable. CloudFix does not take automated action. Manual deletion must be reversed via instance re-creation or cluster scaling.

Can CloudFix Implement the Fix Automatically Once I Accept the Recommendation?

No. This Finder generates actionable recommendations only. You must manually execute the suggested deletions through your AWS interface or scripts.

Does the Fix Require Downtime?

Deleting a read replica may cause a brief disruption, often under a few minutes.

• Aurora PostgreSQL allows up to 3 minutes for active queries to finish before the replica is shut down.

• Aurora MySQL typically completes failover and rebalancing in under 60–120 seconds.

• Actual impact depends on the engine type, workload, and cluster configuration.

Additional Resources

• Amazon Aurora Replication Documentation