ALB Weighted Random Routing and Session Affinity Errors

Why the Answer Is Correct

The correct solution is B, turning on anomaly mitigation for the target group. While the health checks appear green, the 'random application errors' suggest that some instances are failing to handle requests correctly due to statefulness issues or transient glitches that standard HTTP/TCP health checks miss. Anomaly mitigation automatically detects these unhealthy instances based on request failure rates rather than just health check responses, removing them from rotation without manual intervention. This aligns with the community consensus that anomaly detection addresses application failures even when targets are reported healthy.

Why the Other Options Are Wrong

Option A (Least Outstanding Requests) changes the load balancing algorithm but does not inherently fix session affinity issues; if sessions are sticky, the ALB routes to the same instance regardless of the algorithm, so changing to LOR wouldn't resolve errors caused by a specific broken instance handling a sticky session. Option C (Turn off cross-zone LB) restricts traffic flow and might increase latency or imbalance, potentially worsening errors. Option D (Increase deregistration delay) affects how long traffic continues during instance shutdown, which is irrelevant to runtime application errors.

Community Comment Notes

Community users like JSTech and phoenix14 highlight that anomaly detection is designed to address application failures even though underlying registered targets are reported healthy. Some users incorrectly argue for Least Outstanding Requests (A), claiming it makes selection consistent, but this ignores that session affinity locks the user to an instance, making the global algorithm less relevant to the specific error source. The key takeaway is that standard health checks may not catch all application-level faults affecting stateful sessions.