Move Aurora instances to Graviton
AWS Resource Type:
Amazon Aurora (RDS)
This FF identifies provisioned Aurora instances that can move to Graviton to achieve cost reduction while maintaining the same performance.
Criteria for identifying the opportunity:
This FF finds provisioned Aurora instances (i.e. not Serverless or reserved) that are currently available, running on the Intel platform, and have a database version supporting Graviton:
- Amazon RDS - MySQL: 8.0.17 & higher
- Amazon RDS - PostgreSQL: 13, 12.3 & higher
- Amazon RDS - MariaDB: 10.5, 10.4.13 & higher
- Amazon Aurora - MySQL: 2.09.2 & higher
- Amazon Aurora - PostgreSQL: 12.4, 11.9 & higher
In addition, the target instance type should be compatible with Graviton, respecting the following instance type migration paths:
- db.r5.large & db.r6i.large -> db.r6g.large
- db.r5.xlarge & db.r6i.xlarge -> db.r6g.xlarge
- db.r5.2xlarge & db.r6i.2xlarge -> db.r6g.2xlarge
- db.r5.4xlarge & db.r6i.4xlarge -> db.r6g.4xlarge
- db.r5.8xlarge & db.r6i.8xlarge -> db.r6g.8xlarge
- db.r5.12xlarge & db.r6i.12xlarge -> db.r6g.12xlarge
- db.r5.16xlarge & db.r6i.16xlarge -> db.r6g.16xlarge
- db.t3.medium -> db.t4g.medium
- db.t3.large -> db.t4g.large
Potential savings (range in % on annual basis):
On average, this FF should provide 10% savings based on the current price difference between Graviton and Intel-equivalent instances.
What happens when the Fixer is executed?
The target instance will be immediately re-typed, in place, using the ModifyDBInstance operation.
Is it possible to rollback once CloudFix implements the fixer?
Yes, rollback can be triggered manually by executing the SSM Document "Cloudfix-Runbook-AuroraInstancesToGraviton-Rollback".
- DBClusterIdentifier: The cluster identifier (name) whose instances we are about to rollback
- DBInstanceIdentifier: The instance identifier (name) to rollback
- DBTargetClass: The type to which the instance should be retyped
- IsPrimary: if this is the cluster primary instance or not
- ClusterNumberInstances: Indicates the number of instances in the cluster
Can CloudFix implement the fix automatically once I accept the recommendation?
Yes, approving the FF for the target instance will schedule immediate execution.
Does this fix require downtime?
Yes, even using a fail-over, we expect downtime to be around 30 seconds.