Make TCP Keep Alive Timeout configurable

server/kit/config.go

@@ -33,6 +33,10 @@ type Config struct {
 	// The string should be formatted like a time.Duration string.
 	IdleTimeout time.Duration `envconfig:"GIZMO_IDLE_TIMEOUT"`
 
+	// TCPKeepAliveTimeout can be used to override the default tcp timeout of 180s.
+	// The string should be formatted like a time.Duration string.
+	TCPKeepAliveTimeout time.Duration `envconfig:"GIZMO_TCP_KEEP_ALIVE_TIMEOUT"`
+
 	// ShutdownTimeout can be used to override the default http server shutdown timeout
 	// of 5m.
 	ShutdownTimeout time.Duration `envconfig:"GIZMO_SHUTDOWN_TIMEOUT"`
@@ -77,6 +81,9 @@ func loadConfig() Config {
 	if cfg.WriteTimeout.Nanoseconds() == 0 {
 		cfg.WriteTimeout = 10 * time.Second
 	}
+	if cfg.TCPKeepAliveTimeout.Nanoseconds() == 0 {
+		cfg.TCPKeepAliveTimeout = 3 * time.Minute
+	}
 	if cfg.GOMAXPROCS > 0 {
 		runtime.GOMAXPROCS(cfg.GOMAXPROCS)
 	}

server/kit/kitserver.go

@@ -10,6 +10,7 @@ import (
 	"net/http/pprof"
 	"runtime/debug"
 	"strings"
+	"time"
 
 	"cloud.google.com/go/errorreporting"
 	"cloud.google.com/go/profiler"
@@ -307,7 +308,16 @@ func basicDecoder(_ context.Context, r *http.Request) (interface{}, error) {
 
 func (s *Server) start() error {
 	go func() {
-		err := s.svr.ListenAndServe()
+		ln, err := net.Listen("tcp", s.svr.Addr)
+		defer ln.Close()
+		if err != nil {
+			s.logger.Log(
+				"error", err,
+				"message", "HTTP server error - initiating shutting down")
+			s.stop()
+			return
+		}
+		err = s.svr.Serve(tcpKeepAliveListener{ln.(*net.TCPListener), s.cfg.TCPKeepAliveTimeout})
 		if err != nil && err != http.ErrServerClosed {
 			s.logger.Log(
 				"error", err,
@@ -399,3 +409,18 @@ func registerPprof(cfg Config, mx Router) {
 	mx.Handle(http.MethodGet, "/debug/pprof/threadcreate", pprof.Handler("threadcreate"))
 	mx.Handle(http.MethodGet, "/debug/pprof/block", pprof.Handler("block"))
 }
+
+type tcpKeepAliveListener struct {
+	*net.TCPListener
+	Timeout time.Duration
+}
+
+func (ln tcpKeepAliveListener) Accept() (net.Conn, error) {
+	tc, err := ln.AcceptTCP()
+	if err != nil {
+		return nil, err
+	}
+	tc.SetKeepAlive(true)
+	tc.SetKeepAlivePeriod(ln.Timeout)
+	return tc, nil
+}