fix: harden websocket heartbeat under load

server/hdl_websock.go

@@ -28,14 +28,35 @@ import (
 const (
 	// Time allowed to write a message to the peer.
 	writeWait = 10 * time.Second
-
-	// Time allowed to read the next pong message from the peer.
-	pongWait = idleSessionTimeout
-
-	// Send pings to peer with this period. Must be less than pongWait.
-	pingPeriod = 15 * time.Second
+	// Heartbeat defaults tolerate transient mobile and public-network stalls while
+	// still detecting dead connections in a bounded amount of time.
+	defaultWSPongWait   = 90 * time.Second
+	defaultWSPingPeriod = 15 * time.Second
 )
 
+var (
+	pongWait   = defaultWSPongWait
+	pingPeriod = defaultWSPingPeriod
+)
+
+func configureWSHeartbeat(pingSeconds, pongSeconds int) error {
+	ping := defaultWSPingPeriod
+	pong := defaultWSPongWait
+	if pingSeconds > 0 {
+		ping = time.Duration(pingSeconds) * time.Second
+	}
+	if pongSeconds > 0 {
+		pong = time.Duration(pongSeconds) * time.Second
+	}
+	if ping >= pong {
+		return errors.New("websocket ping period must be shorter than pong wait")
+	}
+	pingPeriod = ping
+	pongWait = pong
+	immetrics.SetWSHeartbeatConfig(ping, pong)
+	return nil
+}
+
 func (sess *Session) closeWS() {
 	if sess.proto == WEBSOCK {
 		sess.ws.Close()
@@ -52,6 +73,7 @@ func (sess *Session) readLoop() {
 	sess.ws.SetReadDeadline(time.Now().Add(pongWait))
 	sess.ws.SetPongHandler(func(string) error {
 		atomic.StoreInt64(&sess.lastAction, types.TimeNow().UnixNano())
+		immetrics.IncWSHeartbeat("pong_received")
 		sess.refreshPresence("ws.pong")
 		sess.ws.SetReadDeadline(time.Now().Add(pongWait))
 		return nil