408 Request Timeout

408 Request Timeout means the server was willing to wait for the client to finish sending a complete request but gave up before that happened, and it may close the connection without waiting for the client's next attempt. It differs from 504 Gateway Timeout, which describes a downstream server taking too long to respond after the request was fully received; 408 is specifically about the request never fully arriving.

Slow or unstable client networks, mobile connections with high latency, and large file uploads sent over a single non-resumable request are the usual causes. It also appears when a client opens a connection early, to reduce perceived latency, but delays sending the body, exceeding the server's read timeout. Load balancers and reverse proxies like nginx and AWS ALB generate their own synthetic 408s independently of the application when an idle connection exceeds their configured timeout.

A subtle variant involves keep-alive connection reuse: if a client reuses a persistent connection to send a second request just as the server's idle-keep-alive timeout is about to close it, the two can race, and the client sees a 408 or a reset connection instead of a normal response, which is why keep-alive timeout values are usually tuned to be longer at the edge, the load balancer, than at the origin server itself.