ELB Connection Draining

Correct Answer: Review CloudWatch metrics and adjust connection draining timeout value.

For a Classic Load Balancer (CLB) with Connection Draining enabled, frequent deregistration of instances (for example, during deployments or scale-in) can leave many connections in a draining state, temporarily reducing available capacity and increasing latency. The right approach is to first confirm the behavior in Amazon CloudWatch by examining ELB metrics such as Latency, SurgeQueueLength, SpilloverCount, HealthyHostCount, and BackendConnectionErrors. If you see rising latency and a growing surge queue or spillovers coinciding with draining events, tune the Connection Draining timeout accordingly: decrease it to free capacity sooner if your application can safely close connections quickly, or increase it if clients need more time to complete in-flight requests to avoid 5xx errors. Disabling connection draining may reduce latency at the expense of abruptly terminating client connections, which can increase errors. Removing instances reduces capacity and typically worsens latency. Adjusting health check intervals does not address latency caused by long draining periods and capacity reduction.

Correct Answer: Enable connection draining on the load balancer.

To enable connection draining, you simply need to enable it on the load balancer. The other options are incorrect because they are either not related to connection draining or involve configurations that are not applicable to connection draining.

Correct Answer: Configure connection draining with an appropriate timeout.

Configuring connection draining with an appropriate timeout allows the ELB to stop sending new requests to the back-end instances that are being taken out of service while they complete their in-flight requests, preventing long termination times. The other options are not related to the issue or do not address the root cause.