A financial services company deploys a RAG application.

Configure Amazon CloudWatch alarms for the OpenSearchDashboardsHealthyNodes metric and the SearchLatency metric. Manually investigate and restart OpenSearch Service instances whenever query response times exceed predefined thresholds based on historical performance patterns.

Use Amazon CloudWatch Container Insights to monitor the OpenSearch Service cluster. Use Amazon EventBridge to invoke AWS Systems Manager Automation to restart the OpenSearch Service instances whenever memory usage exceeds 80%.

Create basic Amazon CloudWatch dashboards to track CPU and memory usage for individual services. Configure alarms based on static thresholds for query latency. Set up email notifications when thresholds are exceeded.

Use Amazon CloudWatch dashboards to monitor end-to-end RAG metrics including vector search latency, foundation model (FM) response times, and document ingestion rates. Set up anomaly detection for the metrics to proactively identify performance degradation before users are affected.