When exactly is an http.Request's context canceled on the server side, and how do you use it correctly?

For incoming server requests, r.Context() is canceled when any of these happens:

• the client's connection closes;
• the request is canceled (for example an HTTP/2 RST_STREAM);
• the ServeHTTP method returns.

It descends from http.Server.BaseContext (which defaults to Background) and ConnContext, and it carries values such as http.ServerContextKey and http.LocalAddrContextKey.

func report(w http.ResponseWriter, r *http.Request) {
	ctx, cancel := context.WithTimeout(r.Context(), 3*time.Second)
	defer cancel()

	rows, err := db.QueryContext(ctx, `SELECT ... heavy query ...`)
	if err != nil {
		if errors.Is(err, context.Canceled) {
			return // client left; nobody is reading the response
		}
		http.Error(w, "timeout or db error", http.StatusGatewayTimeout)
		return
	}
	defer rows.Close()
	// stream rows...
}

Notes:

• To add values in middleware, use r = r.WithContext(ctx), which makes a shallow copy of the request, and pass the new r down to next.ServeHTTP.
• http.TimeoutHandler and Server.ReadTimeout/WriteTimeout are separate mechanisms. Only TimeoutHandler cancels the context.
• Over HTTP/1.1, the server notices a disconnect only through its background read, so detection can be delayed while the handler is still reading a large request body.