From c0a395ac899d0fb55b0a0d0338039dcc52472965 Mon Sep 17 00:00:00 2001
From: Ishaan Jaffer <ishaanjaffer0324@gmail.com>
Date: Fri, 20 Feb 2026 17:43:59 -0800
Subject: [PATCH] fix(proxy): add auth reconnect lock timeout budget

---
 litellm/proxy/utils.py | 29 ++++++++++++++++++++++++++++-
 1 file changed, 28 insertions(+), 1 deletion(-)

diff --git a/litellm/proxy/utils.py b/litellm/proxy/utils.py
index 4cbcc7aa713..840d5dfe4a6 100644
--- a/litellm/proxy/utils.py
+++ b/litellm/proxy/utils.py
@@ -2289,6 +2289,10 @@ class PrismaClient:
         self._db_auth_reconnect_timeout_seconds: float = max(
             0.5, float(os.getenv("PRISMA_AUTH_RECONNECT_TIMEOUT_SECONDS", "2.0"))
         )
+        self._db_auth_reconnect_lock_timeout_seconds: float = max(
+            0.0,
+            float(os.getenv("PRISMA_AUTH_RECONNECT_LOCK_TIMEOUT_SECONDS", "0.1")),
+        )
         verbose_proxy_logger.debug("Success - Created Prisma Client")
 
     def get_request_status(
@@ -3587,6 +3591,7 @@ class PrismaClient:
         reason: str,
         force: bool = False,
         timeout_seconds: Optional[float] = None,
+        lock_timeout_seconds: Optional[float] = None,
     ) -> bool:
         """
         Attempt to reconnect the Prisma client in a singleflight manner.
@@ -3606,7 +3611,7 @@ class PrismaClient:
             )
             return False
 
-        async with self._db_reconnect_lock:
+        async def _attempt_reconnect_inside_lock() -> bool:
             now = time.time()
             if (
                 force is False
@@ -3642,6 +3647,28 @@ class PrismaClient:
 
             return reconnect_succeeded
 
+        if lock_timeout_seconds is None:
+            async with self._db_reconnect_lock:
+                return await _attempt_reconnect_inside_lock()
+
+        try:
+            await asyncio.wait_for(
+                self._db_reconnect_lock.acquire(),
+                timeout=lock_timeout_seconds,
+            )
+        except asyncio.TimeoutError:
+            verbose_proxy_logger.debug(
+                "Skipping DB reconnect attempt due to lock acquisition timeout. reason=%s timeout=%ss",
+                reason,
+                lock_timeout_seconds,
+            )
+            return False
+
+        try:
+            return await _attempt_reconnect_inside_lock()
+        finally:
+            self._db_reconnect_lock.release()
+
     async def start_db_health_watchdog_task(self) -> None:
         """
         Start a background task that probes DB health and attempts reconnect on failure.