Graceful Shutdown

Ply can stop the server without cutting active requests. The simplest option is RunWithGracefulShutdown.

app := ply.New()
app.Get("/", func(c *core.Context) {
	c.Text(200, "ok")
})

if err := app.RunWithGracefulShutdown(":8080"); err != nil {
	log.Fatal(err)
}

This starts the server, waits for SIGINT or SIGTERM, then calls Shutdown with a 10 second timeout. The call returns after shutdown completes or the timeout expires.

Manual shutdown

Use RunServer and Shutdown when you need a different timeout or lifecycle:

app := ply.New()
app.Get("/", func(c *core.Context) {
	c.Text(200, "ok")
})

go func() {
	if err := app.RunServer(":8080"); err != nil {
		log.Printf("server stopped: %v", err)
	}
}()

quit := make(chan os.Signal, 1)
signal.Notify(quit, os.Interrupt, syscall.SIGTERM)
<-quit

ctx, cancel := context.WithTimeout(context.Background(), 30*time.Second)
defer cancel()

if err := app.Shutdown(ctx); err != nil {
	log.Printf("shutdown: %v", err)
}

What happens

• The listener stops accepting new connections.
• Active requests are allowed to finish.
• The call returns when they finish or the context expires.
• Shutdown returns server is not running if the server was never started.

After the context deadline, remaining connections are closed. Choose a timeout that is longer than your slowest expected request but shorter than the orchestrator's termination grace period.

Containers

Kubernetes and most container runtimes send SIGTERM first, then SIGKILL after a grace period. Set the grace period longer than the shutdown timeout. A typical pair is a 30 second shutdown timeout and a 45 second termination grace period.

terminationGracePeriodSeconds: 45

Health checks

Stop reporting ready during shutdown. A readiness endpoint that returns 503 as soon as SIGTERM arrives gives load balancers time to remove the instance before its listener closes.

Ply does not expose readiness hooks. Set an atomic flag in your signal handler and check it in the health route.

Notes

• RunWithGracefulShutdown is not available for TLS. See the TLS page for a manual TLS shutdown example.
• Do not call Shutdown twice. Handle the first error and exit.
• Background goroutines are not tracked. Stop them yourself after the server drains.