RDS Cleanup Idle Clusters

Opportunity Name:

RDS Cleanup Idle Clusters

AWS Resource Type:

RDS

Opportunity Description:

Organizations that use RDS may have idle clusters and/or instances that incur costs unnecessarily. CloudFix can detect and remove these idle resources in cases where metrics collected over an extended period of time indicate little to no activity. A snapshot is taken before deletion, ensuring that the resource can be easily restored if necessary.

Criteria for identifying the opportunity:

• Use CloudWatch to get DatabaseConnections and identify idle clusters/instances
• For cluster instances, check NetworkThroughput  - consider idle when any of the following condition is true:
  • maximum of NetworkThroughput = 0  for 7 days
  • maximum of DatabaseConnections < 1 for 31 days
• For non-cluster instances, check Read/Write IOPS metrics - consider idle when any of the following condition is true:
  • maximum of DatabaseConnections < 1 for 31 days
  • maximum of read/write IOPS < 20.0 for 7 days
• Only delete clusters when all instances are idle
• Ignore idle instances protected from deletion
• Do not delete instances in failed, incompatible-restore, or incompatible-network status

Potential savings (range in % on annual basis):

• For an idle cluster, customers can expect to save 100% of compute costs and almost 80% of storage costs (since storage is billed at $0.1/GB-month, but snapshots are billed at $0.021/GB-month).
• Savings are calculated as [Cost of RDS running hours] + [Cost of storage] - [Cost of snapshot]
• Actual savings will depend on how many idle clusters are in the customer accounts and what size instances they are using.

What happens when the Fixer is executed?

Instance/Cluster is deleted after a snapshot is taken.

Is it possible to rollback once CloudFix implements the fixer?

Rollback is manual because CloudFix cannot know when a resource might be needed in the future. CloudFix makes a snapshot before removing an idle cluster and creates a Rollback Runbook (Cloudfix-Runbook-RDSCleanupIdleClusters-Rollback-prod) to restore the cluster and nodes from the snapshot with the original configuration. 

Runbook parameters:

• ResourceId: DB Cluster/ DB Instance snapshot ID
• ResourceType: 'Cluster' or 'Instance'

The rollback is done by manually executing this Rollback Runbook from the AWS Console.

Can CloudFix implement the fix automatically once I accept the recommendation?

Yes

Does this fix require downtime?

No

Additional Resources:

• CloudFix Knowledge Base
• RestoreClusterFromDBSnapshot
• Create Reader and Writer nodes