fix(slack_alerting): poll for the router inside the loop instead of a capped pre-wait

A capped pre-wait still burns the first daily pass when the router takes
longer than the cap to appear (a >10 minute boot), and reads the router
in two places. Folding the poll into the loop makes the first alert
unconditional on boot duration and keeps a single read per pass.

litellm/integrations/SlackAlerting/slack_alerting.py

@@ -43,7 +43,6 @@ from litellm.repositories.user_repository import UserRepository
 from litellm.types.integrations.slack_alerting import *
 from litellm.types.proxy.model_deprecation import (
     DEFAULT_DEPRECATION_CHECK_INTERVAL_SECONDS,
-    DEPRECATION_ROUTER_WAIT_ATTEMPTS,
     DEPRECATION_ROUTER_WAIT_SECONDS,
 )
 
@@ -1083,14 +1082,12 @@ Model Info:
         self, get_llm_router: Callable[[], Router | None] = _proxy_llm_router
     ) -> None:
         """Alert once the router is loaded, then daily, re-reading the router and alert types each pass"""
-        for _ in range(DEPRECATION_ROUTER_WAIT_ATTEMPTS):
-            if get_llm_router() is not None:
-                break
-            await asyncio.sleep(DEPRECATION_ROUTER_WAIT_SECONDS)
-
         while True:
+            if (llm_router := get_llm_router()) is None:
+                await asyncio.sleep(DEPRECATION_ROUTER_WAIT_SECONDS)
+                continue
             try:
-                await self.send_model_deprecation_alert(llm_router=get_llm_router())
+                await self.send_model_deprecation_alert(llm_router=llm_router)
             except Exception as e:  # noqa: BLE001  # a failed alert must not kill the daily loop
                 verbose_proxy_logger.exception("Error in model deprecation alert loop: %s", e)
             await asyncio.sleep(DEFAULT_DEPRECATION_CHECK_INTERVAL_SECONDS)

litellm/types/proxy/model_deprecation.py

@@ -11,8 +11,6 @@ DEFAULT_DEPRECATION_CHECK_INTERVAL_SECONDS: Final = 24 * 60 * 60
 
 DEPRECATION_ROUTER_WAIT_SECONDS: Final = 30
 
-DEPRECATION_ROUTER_WAIT_ATTEMPTS: Final = 20
-
 DeprecationStatus = Literal["upcoming", "imminent", "deprecated"]
 
 

tests/test_litellm/integrations/SlackAlerting/test_model_deprecation_alert.py

@@ -13,7 +13,10 @@ sys.path.insert(0, os.path.abspath("../../../.."))
 import litellm
 from litellm.integrations.SlackAlerting.slack_alerting import SlackAlerting
 from litellm.proxy._types import AlertType
-from litellm.types.proxy.model_deprecation import DEPRECATION_ROUTER_WAIT_SECONDS
+from litellm.types.proxy.model_deprecation import (
+    DEFAULT_DEPRECATION_CHECK_INTERVAL_SECONDS,
+    DEPRECATION_ROUTER_WAIT_SECONDS,
+)
 
 
 def _make_router(deployments):
@@ -166,12 +169,13 @@ async def test_should_wait_for_the_router_instead_of_sleeping_a_full_day(monkeyp
             }
         ]
     )
-    routers = chain((None, None), repeat(router))
+    router_absent_passes = 100
+    routers = chain(repeat(None, router_absent_passes), repeat(router))
     slept: list[float] = []
 
     async def record_sleep(seconds):
         slept.append(seconds)
-        if len(slept) > 2:
+        if len(slept) > router_absent_passes:
             raise asyncio.CancelledError
 
     with (
@@ -186,6 +190,8 @@ async def test_should_wait_for_the_router_instead_of_sleeping_a_full_day(monkeyp
             get_llm_router=lambda: next(routers)
         )
 
-    assert slept[:2] == [DEPRECATION_ROUTER_WAIT_SECONDS] * 2
+    assert slept == [DEPRECATION_ROUTER_WAIT_SECONDS] * router_absent_passes + [
+        DEFAULT_DEPRECATION_CHECK_INTERVAL_SECONDS
+    ]
     mock_send_alert.assert_awaited_once()
     assert "dead-alias" in mock_send_alert.await_args.kwargs["message"]