What happens when a back-end server is marked to drain connections?

When a back-end server is marked to drain connections, the primary action taken is to disallow new connections. This process implies that while existing connections can continue to function and finish their ongoing requests, the server will not accept any fresh connection requests. This is a crucial step in maintenance or scaling operations, as it allows for a controlled and graceful withdrawal of the server from active service without abruptly terminating already established sessions.

By disallowing new connections, the server ensures that users with active sessions can complete their tasks without disruption, maintaining service quality for those users currently connected. The server can then be safely updated or removed from the load balancer without impacting ongoing requests.

The other choices do not accurately reflect the behavior of a server in this scenario. For instance, while it may seem that closing existing connections (immediate closure of connections), redirecting connections (alternative routing to other servers), or increasing capacity (adding more connection potential) could be relevant in a broader context of traffic management and scaling, they do not precisely describe the draining process. Draining focuses on preventing any new load while allowing current connections to persist until completion.