Connectionpool fix

runpod/serverless/modules/rp_ping.py

@@ -7,6 +7,7 @@
 import threading
 
 import requests
+from urllib3.util.retry import Retry
 
 from runpod.serverless.modules.rp_logger import RunPodLogger
 from .worker_state import Jobs, WORKER_ID
@@ -22,8 +23,27 @@ class Heartbeat:
     PING_URL = PING_URL.replace('$RUNPOD_POD_ID', WORKER_ID)
     PING_INTERVAL = int(os.environ.get('RUNPOD_PING_INTERVAL', 10000))//1000
 
-    _session = requests.Session()
-    _session.headers.update({"Authorization": f"{os.environ.get('RUNPOD_AI_API_KEY')}"})
+    def __init__(self, pool_connections=100, retries=3) -> None:
+        '''
+        Initializes the Heartbeat class.
+        '''
+        self._session = requests.Session()
+        self._session.headers.update({"Authorization": f"{os.environ.get('RUNPOD_AI_API_KEY')}"})
+
+        retry_strategy = Retry(
+            total=retries,
+            status_forcelist=[429, 500, 502, 503, 504],
+            method_whitelist=["GET"],
+            backoff_factor=1
+        )
+
+        adapter = requests.adapters.HTTPAdapter(
+            pool_connections=pool_connections,
+            pool_maxsize=pool_connections,
+            max_retries=retry_strategy
+        )
+        self._session.mount('http://', adapter)
+        self._session.mount('https://', adapter)
 
     def start_ping(self, test=False):
         '''
@@ -40,13 +60,8 @@ def ping_loop(self, test=False):
         Sends heartbeat pings to the Runpod server.
         '''
         while True:
-            try:
-                self._send_ping()
-                time.sleep(self.PING_INTERVAL)
-            except requests.RequestException as err:
-                log.error(f"Ping Error: {err}, attempting to restart ping.")
-                if test:
-                    return
+            self._send_ping()
+            time.sleep(self.PING_INTERVAL)
 
             if test:
                 return
@@ -58,9 +73,13 @@ def _send_ping(self):
         job_ids = jobs.get_job_list()
         ping_params = {'job_id': job_ids} if job_ids is not None else None
 
-        result = self._session.get(
-            self.PING_URL, params=ping_params,
-            timeout=self.PING_INTERVAL
-        )
+        try:
+            result = self._session.get(
+                self.PING_URL, params=ping_params,
+                timeout=self.PING_INTERVAL
+            )
+
+            log.debug(f"Heartbeat Sent | URL: {self.PING_URL} | Status: {result.status_code}")
 
-        log.debug(f"Heartbeat Sent | URL: {self.PING_URL} | Status: {result.status_code}")
+        except requests.RequestException as err:
+            log.error(f"Ping Request Error: {err}, attempting to restart ping.")

runpod/serverless/worker.py

@@ -92,16 +92,18 @@ async def run_worker(config: Dict[str, Any]) -> None:
     Args:
         config (Dict[str, Any]): Configuration parameters for the worker.
     """
-    connector = aiohttp.TCPConnector(limit=None)
-    async with aiohttp.ClientSession(
-            connector=connector, headers=_get_auth_header(), timeout=_TIMEOUT) as session:
+    heartbeat.start_ping()
 
+    client_session = aiohttp.ClientSession(
+        connector=aiohttp.TCPConnector(limit=None),
+        headers=_get_auth_header(), timeout=_TIMEOUT
+    )
+
+    async with client_session as session:
         job_scaler = JobScaler(
             concurrency_controller=config.get('concurrency_controller', None)
         )
 
-        heartbeat.start_ping()
-
         while job_scaler.is_alive():
 
             async for job in job_scaler.get_jobs(session):