fix: optimize trigger long running read transactions (#35046)

api/services/credit_pool_service.py

@@ -29,14 +29,15 @@ class CreditPoolService:
     @classmethod
     def get_pool(cls, tenant_id: str, pool_type: str = "trial") -> TenantCreditPool | None:
         """get tenant credit pool"""
-        return db.session.scalar(
-            select(TenantCreditPool)
-            .where(
-                TenantCreditPool.tenant_id == tenant_id,
-                TenantCreditPool.pool_type == pool_type,
+        with sessionmaker(db.engine, expire_on_commit=False).begin() as session:
+            return session.scalar(
+                select(TenantCreditPool)
+                .where(
+                    TenantCreditPool.tenant_id == tenant_id,
+                    TenantCreditPool.pool_type == pool_type,
+                )
+                .limit(1)
             )
-            .limit(1)
-        )
 
     @classmethod
     def check_credits_available(

api/tasks/async_workflow_tasks.py

@@ -162,7 +162,12 @@ def _execute_workflow_common(
                 state_owner_user_id=workflow.created_by,
             )
 
-            # Execute the workflow with the trigger type
+            # NOTE (hj24)
+            # Release the transaction before the blocking generate() call,
+            # otherwise the connection stays "idle in transaction" for hours.
+            session.commit()
+            # NOTE END
+
             generator.generate(
                 app_model=app_model,
                 workflow=workflow,