Merge pull request #654 from tnfssc/fix/redis-connection-retry

backend/run_agent_background.py

@@ -29,6 +29,10 @@ async def initialize():
     """Initialize the agent API with resources from the main API."""
     global db, instance_id, _initialized
     if _initialized:
+        try: await redis.client.ping()
+        except Exception as e:
+            logger.warning(f"Redis connection failed, re-initializing: {e}")
+            await redis.initialize_async(force=True)
         return
 
     # Use provided instance_id or generate a new one
@@ -55,7 +59,11 @@ async def run_agent_background(
     enable_context_manager: bool
 ):
     """Run the agent in the background using Redis for state."""
+    try:
         await initialize()
+    except Exception as e:
+        logger.critical(f"Failed to initialize Redis connection: {e}")
+        raise e
 
     sentry.sentry.set_tag("thread_id", thread_id)
 

backend/services/redis.py

@@ -6,7 +6,7 @@ from utils.logger import logger
 from typing import List, Any
 
 # Redis client
-client = None
+client: redis.Redis | None = None
 _initialized = False
 _init_lock = asyncio.Lock()
 
@@ -47,11 +47,19 @@ def initialize():
     return client
 
 
-async def initialize_async():
+async def initialize_async(force: bool = False):
     """Initialize Redis connection asynchronously."""
     global client, _initialized
 
     async with _init_lock:
+        if _initialized and force:
+            logger.info("Redis connection already initialized, closing and re-initializing")
+            _initialized = False
+            try:
+                await close()
+            except Exception as e:
+                logger.warning(f"Failed to close Redis connection, proceeding with re-initialization anyway: {e}")
+
         if not _initialized:
             logger.info("Initializing Redis connection")
             initialize()