Bug in listing 7.5 (graceful server shutdown) #11
Description
The sample will only work as intended if the server is shutdown within ~30 seconds after launching, because the context's deadline is set directly at server launch. Running the sample and waiting for > 30 seconds before hitting CTRL+C produces
2022/10/02 14:32:45 I started processing the request
^C2022/10/02 14:32:47 Got signal: interrupt. Server shutting down.
2022/10/02 14:32:47 Waiting for shutdown to complete...
2022/10/02 14:32:47 Error during shutdown: context deadline exceeded
2022/10/02 14:32:47 http: Server closed
The deadline to shutdown the server has been reached, but not because the clean shutdown took too long, but because the deadline was already exceeded when s.Shutdown(ctx) was called.
To fix this, context.WithTimeout() should be called in the shutdown function:
func shutdown(ctx context.Context, s *http.Server, waitForShutdownCompletion chan struct{}) {
sigch := make(chan os.Signal, 1)
signal.Notify(sigch, syscall.SIGINT, syscall.SIGTERM)
sig := <-sigch
log.Printf("Got signal: %v. Server shutting down.", sig)
childCtx, cancel := context.WithTimeout(ctx, 30*time.Second)
defer cancel()
if err := s.Shutdown(childCtx); err != nil {
log.Printf("Error during shutdown: %v", err)
}
waitForShutdownCompletion <- struct{}{}
}
func main() {
listenAddr := os.Getenv("LISTEN_ADDR")
if len(listenAddr) == 0 {
listenAddr = ":8080"
}
waitForShutdownCompletion := make(chan struct{})
mux := http.NewServeMux()
mux.HandleFunc("/api/users/", handleUserAPI)
s := http.Server{
Addr: listenAddr,
Handler: mux,
ReadTimeout: 5 * time.Second,
WriteTimeout: 5 * time.Second,
}
go shutdown(context.Background(), &s, waitForShutdownCompletion)
err := s.ListenAndServe()
log.Println("Waiting for shutdown to complete...")
<-waitForShutdownCompletion
log.Fatal(err)
}