From bc60b49e98f17bbe09a457ba96c6b2d3d0650836 Mon Sep 17 00:00:00 2001 From: yucheng Date: Wed, 16 Sep 2026 00:55:11 +0000 Subject: [PATCH] fix(proxy): key held sign-in attempts on the source while the source is blocked An active source block now takes precedence over a pair block, so every blocked username behind one blocked source shares the source's five held slots instead of getting five each Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com> --- litellm/proxy/auth/login_throttle.py | 4 +- .../proxy/auth/test_login_utils.py | 46 +++++++++++++++++++ 2 files changed, 48 insertions(+), 2 deletions(-) diff --git a/litellm/proxy/auth/login_throttle.py b/litellm/proxy/auth/login_throttle.py index fb708ecf872..393411d0670 100644 --- a/litellm/proxy/auth/login_throttle.py +++ b/litellm/proxy/auth/login_throttle.py @@ -290,10 +290,10 @@ class LoginThrottle: shared: Final = await self._shared_block_ttls(keys) user_ttl: Final = max(local[0], shared[0]) source_ttl: Final = max(local[1], shared[1]) - if user_ttl > 0: - return Block(scope="user", retry_after=user_ttl) if self.source_limit is not None and source_ttl > 0: return Block(scope="source", retry_after=source_ttl) + if user_ttl > 0: + return Block(scope="user", retry_after=user_ttl) return None async def _shared_block_ttls(self, keys: _Keys) -> _BlockTtls: diff --git a/tests/test_litellm/proxy/auth/test_login_utils.py b/tests/test_litellm/proxy/auth/test_login_utils.py index ac40b5364f9..65c388240e5 100644 --- a/tests/test_litellm/proxy/auth/test_login_utils.py +++ b/tests/test_litellm/proxy/auth/test_login_utils.py @@ -1220,6 +1220,52 @@ async def test_held_attempts_from_one_blocked_key_are_capped(monkeypatch): assert lt._HELD_ATTEMPTS.get(slot) is None, "the slots are released once the held attempts answer" +@pytest.mark.asyncio +async def test_a_blocked_source_shares_one_held_slot_pool_across_its_blocked_usernames(monkeypatch): + """Once the source is blocked, a pair block for a username must not hand that username its own five slots.""" + import asyncio + + from litellm.proxy._types import ProxyException + from litellm.proxy.auth import login_throttle as lt + from litellm.proxy.auth.login_throttle import MAX_HELD_ATTEMPTS_PER_KEY + + monkeypatch.setenv("UI_USERNAME", "admin") + monkeypatch.setenv("UI_PASSWORD", "right") + release = asyncio.Event() + + async def _park(_seconds: float) -> None: + await release.wait() + + monkeypatch.setattr(lt, "_sleep", _park) + throttle = _throttle(user_limit=1, source_limit=3, client_ip="203.0.113.45") + assert [await _fail(throttle) for _ in range(2)] == ["401", "401"], "the admin pair is now blocked" + assert [await _fail(throttle, username=f"spray-{i}@corp.com") for i in range(3)] == ["401"] * 3 + source_slot = throttle._keys("admin").source_block + assert throttle._local_block_ttl(source_slot) > 0, "the source is now blocked as well" + + usernames = ["admin", *(f"fresh-{i}@corp.com" for i in range(MAX_HELD_ATTEMPTS_PER_KEY - 1))] + held = [asyncio.create_task(_guess(throttle, username=name)) for name in usernames] + for _ in range(1000): + if lt._HELD_ATTEMPTS.get(source_slot) == MAX_HELD_ATTEMPTS_PER_KEY: + break + await asyncio.sleep(0) + assert lt._HELD_ATTEMPTS == {source_slot: MAX_HELD_ATTEMPTS_PER_KEY} + + try: + for name in ("admin", "fresh-0@corp.com", "never-seen@corp.com"): + with pytest.raises(ProxyException) as over_cap: + await _guess(throttle, username=name) + assert over_cap.value.code == "429" + assert over_cap.value.headers.get("Retry-After") == "30" + finally: + release.set() + for task in held: + with pytest.raises(ProxyException): + await task + + assert lt._HELD_ATTEMPTS == {} + + @pytest.mark.asyncio async def test_disabling_the_control_removes_the_hold_as_well(monkeypatch, login_delays): """The escape hatch has to turn off the whole control, not only the refusal."""