Question

A company runs a serverless order-processing application where AWS Lambda functions connect directly to an Amazon RDS for PostgreSQL Multi-AZ database. During flash sales, concurrent Lambda executions spike to several thousand, and the application begins throwing 'too many connections' errors and intermittent timeouts. CloudWatch shows the database CPU is moderate but the DatabaseConnections metric is pegged at its maximum. The team wants to improve reliability during traffic bursts with minimal code changes and without over-provisioning the database instance. What should the solutions architect recommend?

Accepted Answer

Deploy Amazon RDS Proxy in front of the database and update the Lambda functions to connect through the proxy endpoint, enabling connection pooling and reuse.

Answer

Increase the max_connections parameter in the RDS parameter group and scale up to a larger instance class to accommodate the connection volume.

Answer

Migrate the database to Amazon DynamoDB to eliminate the relational connection limit entirely.

Answer

Configure reserved concurrency on the Lambda functions to cap the number of concurrent executions below the database connection limit.