shutdown

shutdown

Package shutdown helps controlling app shutdown and graceful termination of goroutines.

It listens for SIGTERM (e.g. kill command) and SIGINT (e.g. CTRL+C) signals, and also provides a manual way to trigger shutdown.

It publishes a single, shared shutdown channel which is closed when shutdown is about to happen. Modules (goroutines) should monitor this channel using a select statement, and terminate ASAP if it is (gets) closed. Additionally, there is an Initiated() function which tells if a shutdown has been initiated, which basically checks the shared channel in a non-blocking way.

A context.Context is also published which will be cancelled when shutdown is about to happen. Background tasks requiring a context may use this directly or as a parent context.

It also publishes a WaitGroup goroutines may use to "register" themselves should they wish to be patiently waited for and not get terminated abruptly. For this to "work", this shared WaitGroup must be "waited for" in the main() function before returning.

Examples

Simple example

Example #1: If you just want to do something before shutting down:

func main() {
	go func() {
		// This is your app:
		for {
			log.Println("Tick...")
			time.Sleep(time.Second)
		}
	}()

	<-shutdown.C

	log.Println("Doing this before shutting down.")
}

Note that monitoring the shutdown channel must be in the main goroutine and your task in another one (and not the other way), because the app terminates when the main() function returns.

Advanced example

Example #2: A more advanced example where a worker goroutine is to be waited for. This app also self-terminates after 10 seconds:

func main() {
	// Initiate a manual shutdown if we're still running after 10 sec
	time.AfterFunc(10*time.Second, shutdown.InitiateManual)

	// Example generator (job producer)
	jobCh := make(chan int)
	go func() {
		for i := 0; ; i++ {
			jobCh <- i
		}
	}()

	// Example worker goroutine whose completion we will wait for.
	shutdown.Wg.Add(1)
	go func() {
		defer shutdown.Wg.Done()
		for {
			// Receive jobs, listen for shutdown:
			select {
			case jobID := <-jobCh:
				log.Printf("[worker] Doing job #%d...", jobID)
				time.Sleep(time.Second) // Simulate work...
			case <-shutdown.C:
				log.Println("[worker] Aborting. Saving progress...")
				time.Sleep(time.Second) // Simulate work...
				log.Println("[worker] Save complete.")
				return
			}
		}
	}()

	// Wait for a shutdown event (either signal or manual)
	<-shutdown.C

	// Wait for "important" goroutines
	shutdown.Wg.Wait()
}

Note that the above worker goroutine does not guarantee that it won't start execution of a new job after a shutdown has been initiated (because select chooses a "ready" case pseudo-randomly).

Advanced example (variant)

Example #3: If you need guarantee that no new jobs are taken after a shutdown initiation, you may check the shutdown channel first, in a separate select in a non-blocking way, or you may simply add the check as the loop condition like this:

// Example worker goroutine whose completion we will wait for.
shutdown.Wg.Add(1)
go func() {
	defer shutdown.Wg.Done()
	defer func() {
		log.Println("[worker] Aborting. Saving progress...")
		time.Sleep(time.Second) // Simulate work...
		log.Println("[worker] Save complete.")
	}()
	for !shutdown.Initiated() {
		// Receive jobs, listen for shutdown:
		select {
		case jobID := <-jobCh:
			log.Printf("[worker] Doing job #%d...", jobID)
			time.Sleep(time.Second) // Simulate work...
		case <-shutdown.C:
			return
		}
	}
}()

Web server example

Example #4: The following example starts a web server and provides graceful shutdown for it. It also handles abnormal (and silent) termination, in which case it triggers a manual shutdown, making sure the whole app gets terminated (not just its web server):

func main() {
	helloFunc := func(w http.ResponseWriter, r *http.Request) {
		w.Write([]byte("hello"))
	}
	srv := &http.Server{
		Addr:    ":8080",
		Handler: http.HandlerFunc(helloFunc),
	}

	go func() {
		if err := srv.ListenAndServe(); err != nil {
			if err == http.ErrServerClosed {
				log.Println("HTTP Server gracefully shut down.")
				return
			}
			log.Printf("Abnormal HTTP Server shut down with error: %v", err)
		} else {
			log.Println("HTTP Server SILENTLY shut down.")
		}

		// If we got to this point, that's not normal:
		log.Println("Initiating manual system shutdown:")
		shutdown.InitiateManual()
	}()

	// Wait for a shutdown event (either signal or manual)
	<-shutdown.C

	log.Println("Stopping HTTP server (system shutdown)...")

	// Shutdown gracefully, but wait no longer than 20 seconds:
	ctx, cancel := context.WithTimeout(context.Background(), time.Second*20)
	err := srv.Shutdown(ctx)
	cancel() // Call cancel to release resources of the context

	if err != nil {
		log.Printf("Failed to shut down HTTP server gracefully: %v", err)
		// Try forceful shutdown:
		if err := srv.Close(); err != nil {
			log.Printf("HTTP server forceful shutdown error: %v", err)
		}
	}
}

Context example with background worker

Example #5: The following example launches a background worker doing something that uses / requires a context.

func main() {
	// Worker goroutine requiring a context (we'll wait for its completion).
	shutdown.Wg.Add(1)
	go func() {
		defer shutdown.Wg.Done()
		ctx := shutdown.Context
		for !shutdown.Initiated() {
			result, err := dbAdapter.RunQuery(ctx, "some-query")
			if err != nil {
				log.Printf("Query error: %v", err)
			} else {
				log.Printf("Query result: %v", result)
			}
		}
	}()

	// Wait for a shutdown event (either signal or manual)
	<-shutdown.C

	// Wait for the worker to finish
	shutdown.Wg.Wait()
}