Parametrize modal webrtc worker watchdog timeout

inference/core/env.py

@@ -722,6 +722,8 @@
 # seconds
 WEBRTC_MODAL_RESPONSE_TIMEOUT = int(os.getenv("WEBRTC_MODAL_RESPONSE_TIMEOUT", "60"))
 # seconds
+WEBRTC_MODAL_WATCHDOG_TIMEMOUT = int(os.getenv("WEBRTC_MODAL_WATCHDOG_TIMEMOUT", "60"))
+# seconds
 WEBRTC_MODAL_FUNCTION_TIME_LIMIT = int(
     os.getenv("WEBRTC_MODAL_FUNCTION_TIME_LIMIT", "3600")
 )

inference/core/interfaces/webrtc_worker/modal.py

@@ -43,6 +43,7 @@
     WEBRTC_MODAL_SHUTDOWN_RESERVE,
     WEBRTC_MODAL_TOKEN_ID,
     WEBRTC_MODAL_TOKEN_SECRET,
+    WEBRTC_MODAL_WATCHDOG_TIMEMOUT,
     WORKFLOWS_CUSTOM_PYTHON_EXECUTION_MODE,
 )
 from inference.core.exceptions import RoboflowAPIUnsuccessfulRequestError
@@ -184,6 +185,7 @@ def check_nvidia_smi_gpu() -> str:
             "WEBRTC_MODAL_RTSP_PLACEHOLDER": WEBRTC_MODAL_RTSP_PLACEHOLDER,
             "WEBRTC_MODAL_RTSP_PLACEHOLDER_URL": WEBRTC_MODAL_RTSP_PLACEHOLDER_URL,
             "WEBRTC_MODAL_SHUTDOWN_RESERVE": str(WEBRTC_MODAL_SHUTDOWN_RESERVE),
+            "WEBRTC_MODAL_WATCHDOG_TIMEMOUT": str(WEBRTC_MODAL_WATCHDOG_TIMEMOUT),
         },
         "volumes": {MODEL_CACHE_DIR: rfcache_volume},
     }
@@ -198,7 +200,7 @@ async def run_rtc_peer_connection_with_watchdog(
         )
 
         watchdog = Watchdog(
-            timeout_seconds=30,
+            timeout_seconds=WEBRTC_MODAL_WATCHDOG_TIMEMOUT,
         )
 
         rtc_peer_connection_task = asyncio.create_task(
@@ -211,7 +213,6 @@ async def run_rtc_peer_connection_with_watchdog(
         )
 
         def on_timeout():
-            logger.info("Watchdog timeout reached")
             rtc_peer_connection_task.cancel()
 
         watchdog.on_timeout = on_timeout
@@ -266,6 +267,7 @@ def rtc_peer_connection_modal(
             logger.info("declared_fps: %s", webrtc_request.declared_fps)
             logger.info("rtsp_url: %s", webrtc_request.rtsp_url)
             logger.info("processing_timeout: %s", webrtc_request.processing_timeout)
+            logger.info("watchdog_timeout: %s", WEBRTC_MODAL_WATCHDOG_TIMEMOUT)
             logger.info("requested_plan: %s", webrtc_request.requested_plan)
             logger.info("requested_region: %s", webrtc_request.requested_region)
             logger.info(

inference/core/interfaces/webrtc_worker/watchdog.py

@@ -20,6 +20,7 @@ def __init__(
         self._heartbeats = 0
 
     def start(self):
+        logger.info("Starting watchdog with timeout %s", self.timeout_seconds)
         if not self.on_timeout:
             raise ValueError(
                 "on_timeout callback must be provided before starting the watchdog"
@@ -32,12 +33,13 @@ def stop(self):
             self._thread.join()
 
     def _watchdog_thread(self):
+        logger.info("Watchdog thread started")
         while not self._stopping:
             if not self.is_alive():
                 logger.error("Watchdog timeout reached")
                 self.on_timeout()
                 break
-            time.sleep(0.1)
+            time.sleep(1)
         logger.info("Watchdog stopped")
 
     def heartbeat(self):

inference/core/interfaces/webrtc_worker/webrtc.py

@@ -634,6 +634,7 @@ async def recv(self):
 
 async def _wait_ice_complete(peer_connection: RTCPeerConnectionWithLoop, timeout=2.0):
     if peer_connection.iceGatheringState == "complete":
+        logger.info("ICE gathering state already complete")
         return
     fut = asyncio.get_running_loop().create_future()
 
@@ -662,6 +663,7 @@ async def init_rtc_peer_connection_with_loop(
     shutdown_reserve: int = WEBRTC_MODAL_SHUTDOWN_RESERVE,
     heartbeat_callback: Optional[Callable[[], None]] = None,
 ) -> RTCPeerConnectionWithLoop:
+    logger.info("Initializing RTC peer connection with loop")
     # ice._mdns is instantiated on the module level, it has a lock that is bound to the event loop
     # avoid RuntimeError: asyncio.locks.Lock is bound to a different event loop
     if hasattr(ice, "_mdns"):
@@ -956,10 +958,13 @@ def on_message(message):
     answer = await peer_connection.createAnswer()
     await peer_connection.setLocalDescription(answer)
 
-    logger.debug(f"WebRTC connection status: {peer_connection.connectionState}")
-
     await _wait_ice_complete(peer_connection, timeout=2.0)
 
+    logger.info(
+        "Initialized RTC peer connection with loop (status: %s), sending answer",
+        peer_connection.connectionState,
+    )
+
     send_answer(
         WebRTCWorkerResult(
             answer={
@@ -969,7 +974,9 @@ def on_message(message):
         )
     )
 
+    logger.info("Answer sent, waiting for termination event")
     await terminate_event.wait()
+    logger.info("Termination event received, closing WebRTC connection")
     if player:
         logger.info("Stopping player")
         player.video.stop()