A legacy app has four problems: web servers hit 90 percent CPU during office hours; the database rejects new connections at its limit; user sessions are held in web server memory, so users are logged out on every deploy; and each release needs two hours of downtime. Which single design fixes all four?
Show answer and explanation
Answer: D
Each problem gets its own fix that works together. Scaling policies add web capacity during peaks. RDS Proxy pools and multiplexes database connections, which relieves the connection ceiling without a bigger instance. Moving sessions to ElastiCache keeps them outside the web servers, so deployments no longer log users out. CodeDeploy blue/green deployments shift traffic to a new fleet and roll back automatically, removing the downtime. The other options leave at least one symptom in place.
Why the other options are wrong
- A. Larger web servers and a higher connection parameter address only two symptoms, since the database still has a hard connection ceiling, and overnight releases do not remove session loss or the deployment downtime.
- B. Read replicas do not pool connections for the writer, sticky sessions still tie users to servers that are replaced during each deployment, and rolling in-place updates keep some downtime and risk.
- C. Containers help packaging, but storing sessions on each task's local disk means they vanish when tasks are replaced, and service discovery does not relieve the database connection limit.
Exam tip. Look for the component that must become stateless: session state belongs in a shared store such as ElastiCache or DynamoDB.