fix(otel): read the registered v2 logger without importing the proxy (#44485)

* fix(otel): read the registered v2 logger without importing the proxy

phase_span, which the router enters on every deployment pick since #44150, looked up the
proxy's OTel logger by importing litellm.proxy.proxy_server. In an SDK process that import
loads the whole proxy synchronously on the caller's event loop during its first request,
which stalled litellm_router_unit_testing past its 5s wait. Read the module from
sys.modules instead: when the proxy was never imported it has no registered logger.

Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>

* test(otel): pin the no-proxy-import guarantee in-process

Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>

---------

Co-authored-by: mateo <mateo@berri.ai>
Co-authored-by: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>

litellm/integrations/otel/logger.py

@@ -1,5 +1,6 @@
 """``CustomLogger`` adapter on the OpenTelemetry span engine."""
 
+import sys
 from collections import OrderedDict
 from collections.abc import Callable, Iterator, Mapping, Sequence
 from contextlib import contextmanager, nullcontext
@@ -966,10 +967,12 @@ def _v2_configs(in_memory_loggers: Sequence[object], logger: "OpenTelemetryV2")
 
 
 def _registered_v2_logger() -> "OpenTelemetryV2 | None":
-    try:
-        from litellm.proxy import proxy_server
-    except Exception:
-        return None
+    """The proxy's registered V2 logger, read without importing the proxy.
+
+    Request paths call this (the router's ``route`` phase among them), so importing
+    ``proxy_server`` here would load the whole proxy on an SDK caller's event loop.
+    """
+    proxy_server: Final = sys.modules.get("litellm.proxy.proxy_server")
     logger: Final = getattr(proxy_server, "open_telemetry_logger", None)
     return logger if isinstance(logger, OpenTelemetryV2) else None
 

tests/unit/integrations/otel/test_runtime.py

@@ -8,6 +8,8 @@ import lock. These tests pin the import to a single resolution.
 """
 
 import builtins
+import importlib.abc
+import sys
 
 import litellm.integrations.otel.runtime as runtime
 
@@ -69,3 +71,25 @@ def test_phase_event_no_ops_when_runtime_absent(monkeypatch):
 
     assert runtime.phase_event("litellm.request.body_parsed") is None
     assert runtime.phase_event("litellm.request.body_received", {"litellm.request.body_bytes": 3}) is None
+
+
+def test_phase_span_does_not_import_the_proxy_in_an_sdk_process(monkeypatch):
+    import litellm.proxy
+
+    monkeypatch.delitem(sys.modules, "litellm.proxy.proxy_server", raising=False)
+    monkeypatch.delattr(litellm.proxy, "proxy_server", raising=False)
+    proxy_imports: list[str] = []
+
+    class _RefuseProxyImport(importlib.abc.MetaPathFinder):
+        def find_spec(self, fullname, path, target=None):
+            if fullname == "litellm.proxy.proxy_server":
+                proxy_imports.append(fullname)
+                raise ImportError(fullname)
+            return None
+
+    monkeypatch.setattr(sys, "meta_path", [_RefuseProxyImport(), *sys.meta_path])
+
+    with runtime.phase_span("route gpt-5-mini") as span:
+        assert span is None
+
+    assert proxy_imports == []