A company hosts an application on multiple Amazon EC2 instances. The application processes messages from an Amazon SQS queue, writes for an AmazonRDS table, and...

A company hosts an application on multiple Amazon EC2 instances. The application processes messages from an Amazon SQS queue, writes for an AmazonRDS table, and deletes -the message from the queue. Occasional duplicate records are found in the RDS table. The SQS queue does not contain any duplicate messages.
What should a solutions architect do to ensure messages are being processed once only?