Viewing and responding to Amazon RDS recommendations - Amazon Relational Database Service

Viewing and responding to Amazon RDS recommendations

Amazon RDS provides automated recommendations for database resources, such as DB instances, read replicas, and DB parameter groups. These recommendations provide best practice guidance by analyzing DB instance configuration, usage, and performance data.

Amazon RDS Performance Insights monitors specific metrics and automatically creates thresholds by analyzing what levels are considered potentially problematic for a specified resource. When new metric values cross a predefined threshold over a given period of time, Performance Insights generates a proactive recommendation. This recommendation helps to prevent future database performance impact. For example, the "Idle In Transaction" recommendation is generated for RDS for PostgreSQL instances when the sessions connected to the database are not performing active work, but can keep database resources blocked. To receive proactive recommendations, you must turn on Performance Insights with a paid tier retention period. For information about turning on Performance Insights, see Turning Performance Insights on and off for Amazon RDS. For information about pricing and data retention for Performance Insights see Pricing and data retention for Performance Insights.

DevOps Guru for RDS monitors certain metrics to detect when the metric's behavior becomes highly unusual or anomalous. These anomalies are reported as reactive insights with recommendations. For example, DevOps Guru for RDS might recommend you to consider increasing CPU capacity or investigate wait events that are contributing to DB load. DevOps Guru for RDS also provides threshold based proactive recommendations. For these recommendations, you must turn on DevOps Guru for RDS. For information about turning on DevOps Guru for RDS, see Turning on DevOps Guru and specifying resource coverage.

Recommendations will be in any of the following status: active, dismissed, pending, or resolved. Resolved recommendations are available for 365 days.

You can view or dismiss the recommendations. You can apply a configuration based active recommendation immediately, schedule it in the next maintenance window, or dismiss it. For threshold based proactive and machine learning based reactive recommendations, you need to review the suggested cause of the issue and then perform the recommended actions to fix the issue.