Merge pull request 'fix: saas restart logic' (#13) from furqan into dev

Reviewed-on: https://gitea.maskantech.in/gitea_admin/saas_backend/pulls/13

app/__init__.py

@@ -31,21 +31,41 @@ logging.basicConfig(
 
 logger = logging.getLogger(__name__)
 
+async def wait_for_db(retries: int = 10, base_delay: float = 3.0, max_delay: float = 30.0):
+    """Wait for the database to become available with exponential backoff."""
+    for attempt in range(1, retries + 1):
+        try:
+            with engine.connect() as conn:
+                conn.execute(text("SELECT 1"))
+            logger.info("Database connection successful!")
+            return
+        except Exception as e:
+            if attempt == retries:
+                logger.error(f"Database unreachable after {retries} attempts: {e}")
+                raise
+            delay = min(base_delay * (2 ** (attempt - 1)), max_delay)
+            logger.warning(
+                f"Database not ready (attempt {attempt}/{retries}): {e}. "
+                f"Retrying in {delay:.0f}s..."
+            )
+            await asyncio.sleep(delay)
 
 @asynccontextmanager
 async def lifespan(app: FastAPI):
-    logger.info("Testing database connection...")
-    try:
-        with engine.connect() as conn:
-            conn.execute(text("SELECT 1"))
-            conn.commit()
-        logger.info("Database connection successful!")
-    except Exception as e:
-        logger.error(f"Database connection failed: {e}")
-        raise
+    logger.info("Starting SaaS application...")
 
-    await redis_client.connect()
-    await run_in_threadpool(sync_redis_client.connect)
+    await wait_for_db()
+
+    app.state.redis_available = False
+
+    try:
+        await redis_client.connect()
+        await run_in_threadpool(sync_redis_client.connect)
+        app.state.redis_available = True
+        logger.info("Redis connected successfully.")
+    except Exception as e:
+        logger.error(f"Redis unavailable at startup: {e}")
+        logger.warning("Continuing without Redis — event queue disabled.")
 
     logger.info(
         f"{settings.PROJECT_NAME} v{settings.VERSION} started ({settings.APP_ENV})"
@@ -55,9 +75,9 @@ async def lifespan(app: FastAPI):
         logger.info("Redis Event Consumer STARTED")
         while True:
             try:
-                if not redis_client.client:
-                     await asyncio.sleep(5)
-                     continue
+                if not app.state.redis_available or not redis_client.client:
+                    await asyncio.sleep(5)
+                    continue
 
                 result = await redis_client.client.blpop("saas:events:queue", timeout=5)
 
@@ -69,7 +89,7 @@ async def lifespan(app: FastAPI):
                     except Exception as e:
                         logger.error(f"Error processing event {event_id}: {e}")
 
-            except Exception as e:
+            except Exception:
                 await asyncio.sleep(1)
 
     async def fallback_poller():
@@ -94,9 +114,12 @@ async def lifespan(app: FastAPI):
     for task in background_tasks:
         task.cancel()
     await asyncio.gather(*background_tasks, return_exceptions=True)
-    await redis_client.close()
-    await run_in_threadpool(sync_redis_client.close)
-
+    if app.state.redis_available:
+        try:
+            await redis_client.close()
+            await run_in_threadpool(sync_redis_client.close)
+        except Exception:
+            pass
 
 def create_app() -> FastAPI:
     app = FastAPI(