From 57629e7383cc042b8371dfc724649f3e14e9b850 Mon Sep 17 00:00:00 2001
From: King Star <mcxin.y@gmail.com>
Date: Wed, 9 Sep 2026 10:00:15 +0800
Subject: [PATCH] test(router): cover failed deployment identity paths

---
 ..._prometheus_requested_model_cardinality.py | 30 +++++++++++++++
 .../test_enforce_model_rate_limits.py         | 37 ++++++++++++++++++-
 2 files changed, 66 insertions(+), 1 deletion(-)

diff --git a/tests/test_litellm/integrations/test_prometheus_requested_model_cardinality.py b/tests/test_litellm/integrations/test_prometheus_requested_model_cardinality.py
index d239764b84f..d84c7e4d705 100644
--- a/tests/test_litellm/integrations/test_prometheus_requested_model_cardinality.py
+++ b/tests/test_litellm/integrations/test_prometheus_requested_model_cardinality.py
@@ -248,6 +248,36 @@ async def test_async_failure_metrics_prefer_stamped_failed_deployment_id():
     assert _model_id_values(logger.litellm_deployment_failure_responses) == {"deployment-a"}
 
 
+def test_deployment_failure_model_id_falls_back_to_nested_metadata():
+    logger = PrometheusLogger()
+
+    model_id = logger._get_deployment_failure_model_id(
+        request_kwargs={
+            "litellm_params": {"litellm_metadata": {"model_info": {"id": "deployment-a"}}},
+        },
+        standard_logging_payload={},
+    )
+
+    assert model_id == "deployment-a"
+
+
+@pytest.mark.parametrize(
+    "request_kwargs",
+    [
+        {},
+        {"litellm_params": {"metadata": {"model_info": {}}}},
+    ],
+)
+def test_deployment_failure_model_id_returns_none_without_a_model_id(request_kwargs):
+    assert (
+        PrometheusLogger._get_deployment_failure_model_id(
+            request_kwargs=request_kwargs,
+            standard_logging_payload={},
+        )
+        is None
+    )
+
+
 @pytest.mark.asyncio
 async def test_sdk_fallback_labels_survive_non_import_errors_from_proxy_module(monkeypatch):
     logger = PrometheusLogger()
diff --git a/tests/test_litellm/test_router/test_enforce_model_rate_limits.py b/tests/test_litellm/test_router/test_enforce_model_rate_limits.py
index 1def253ac93..579fc1fc4dd 100644
--- a/tests/test_litellm/test_router/test_enforce_model_rate_limits.py
+++ b/tests/test_litellm/test_router/test_enforce_model_rate_limits.py
@@ -6,13 +6,14 @@ regardless of the routing strategy being used.
 """
 
 import asyncio
-from unittest.mock import AsyncMock, MagicMock
+from unittest.mock import AsyncMock, MagicMock, patch
 
 import pytest
 
 import litellm
 from litellm import Router
 from litellm.caching.dual_cache import DualCache
+from litellm.integrations.custom_logger import CustomLogger
 from litellm.router_utils.pre_call_checks.model_rate_limit_check import (
     ModelRateLimitingCheck,
 )
@@ -353,3 +354,37 @@ class TestModelRateLimitConcurrency:
 
         assert len(successes) == 2, f"Expected 2 successes, got {len(successes)}"
         assert len(failures) == 2, f"Expected 2 rate limit errors, got {len(failures)}"
+
+
+@pytest.mark.asyncio
+@pytest.mark.parametrize(
+    "hook_error",
+    [
+        litellm.RateLimitError(message="rpm exceeded", llm_provider="openai", model="gpt-5.6"),
+        RuntimeError("pre call check blew up"),
+    ],
+)
+async def test_router_async_pre_call_checks_stamp_the_refusing_deployment(hook_error):
+    class _RaisingPreCallCheck(CustomLogger):
+        async def async_pre_call_check(self, deployment, parent_otel_span):
+            raise hook_error
+
+    router = Router(
+        model_list=[
+            {
+                "model_name": "gpt-5.6",
+                "litellm_params": {"model": "openai/gpt-5.6", "api_key": "sk-fake"},
+                "model_info": {"id": "deployment-a"},
+            }
+        ]
+    )
+    deployment = router.model_list[0]
+
+    with patch.object(litellm, "callbacks", [_RaisingPreCallCheck()]):  # test-quality-ok: router reads this global
+        with pytest.raises(type(hook_error)):
+            await router.async_routing_strategy_pre_call_checks(
+                deployment=deployment,
+                parent_otel_span=None,
+            )
+
+    assert hook_error.failed_deployment_id == "deployment-a"