fix(redis_cache.py): add 5s default timeout

litellm/_redis.py

@@ -182,9 +182,7 @@ def init_redis_cluster(redis_kwargs) -> redis.RedisCluster:
                 "REDIS_CLUSTER_NODES environment variable is not valid JSON. Please ensure it's properly formatted."
             )
 
-    verbose_logger.debug(
+    verbose_logger.debug("init_redis_cluster: startup nodes are being initialized.")
-        "init_redis_cluster: startup nodes are being initialized."
-    )
     from redis.cluster import ClusterNode
 
     args = _get_redis_cluster_kwargs()
@@ -307,7 +305,6 @@ def get_redis_async_client(
         return _init_async_redis_sentinel(redis_kwargs)
 
     return async_redis.Redis(
-        socket_timeout=5,
         **redis_kwargs,
     )
 

litellm/caching/redis_cache.py

@@ -54,6 +54,7 @@ class RedisCache(BaseCache):
         redis_flush_size: Optional[int] = 100,
         namespace: Optional[str] = None,
         startup_nodes: Optional[List] = None,  # for redis-cluster
+        socket_timeout: Optional[float] = 5.0,  # default 5 second timeout
         **kwargs,
     ):
 
@@ -70,6 +71,9 @@ class RedisCache(BaseCache):
             redis_kwargs["password"] = password
         if startup_nodes is not None:
             redis_kwargs["startup_nodes"] = startup_nodes
+        if socket_timeout is not None:
+            redis_kwargs["socket_timeout"] = socket_timeout
+
         ### HEALTH MONITORING OBJECT ###
         if kwargs.get("service_logger_obj", None) is not None and isinstance(
             kwargs["service_logger_obj"], ServiceLogging
@@ -556,6 +560,7 @@ class RedisCache(BaseCache):
             ## LOGGING ##
             end_time = time.time()
             _duration = end_time - start_time
+
             asyncio.create_task(
                 self.service_logger_obj.async_service_success_hook(
                     service=ServiceTypes.REDIS,

litellm/proxy/_new_secret_config.yaml

@@ -6,4 +6,12 @@ model_list:
     api_base: os.environ/AZURE_API_BASE
     
 litellm_settings:
-  callbacks: ["prometheus"]
+  callbacks: ["prometheus"]
+
+router_settings:
+  routing_strategy: usage-based-routing-v2 # 👈 KEY CHANGE
+  redis_host: os.environ/REDIS_HOST
+  redis_password: os.environ/REDIS_PASSWORD
+  redis_port: os.environ/REDIS_PORT
+
+

tests/litellm/caching/test_redis_cache.py

@@ -1,9 +1,13 @@
+import asyncio
 import json
 import os
 import sys
+import time
 from unittest.mock import MagicMock, patch
 
+import httpx
 import pytest
+import respx
 from fastapi.testclient import TestClient
 
 sys.path.insert(
@@ -39,3 +43,12 @@ async def test_redis_cache_async_increment(namespace):
         mock_redis_instance.incrbyfloat.assert_called_once_with(
             name=expected_key, amount=1
         )
+
+
+@pytest.mark.asyncio
+async def test_redis_client_init_with_socket_timeout():
+    redis_cache = RedisCache(socket_timeout=1.0)
+    assert redis_cache.redis_kwargs["socket_timeout"] == 1.0
+    client = redis_cache.init_async_client()
+    assert client is not None
+    assert client.connection_pool.connection_kwargs["socket_timeout"] == 1.0