I’m seeking assistance with a persistent hang issue when attempting to run workflows that contain a Code Node using the Manual Execute button.
When the environment variable OFFLOAD_MANUAL_EXECUTIONS_TO_WORKERS is set to true on the Main Node, all manual executions of workflows containing a Code Node (specifically Python) immediately hang in the UI.
Successful Configuration Checks
I’ve confirmed the following critical items are correctly configured and working:
-
Network Connectivity (Task Broker): Workers and Runners can successfully reach the Primary Node’s Task Broker URI (Confirmed via
nc -vz). -
Queue Connectivity: Worker nodes successfully connect to and communicate via the Redis queue
-
Runner Registration: The Runner Pods successfully connect to the Main Node and register their services. The Main Node logs show:
Registered runner "launcher-python" -
Hanging Log State: The Runner logs show the process is stalled waiting for the task offer:
INFO [launcher:py] Waiting for launcher's task offer to be accepted... -
Environment Variables: All runner-related variables are consistent across Main, Workers, and Runners, including:
-
N8N_RUNNERS_MODE: external -
N8N_RUNNERS_AUTH_TOKEN(Confirmed identical via clean secret reference). -
N8N_RUNNERS_TASK_BROKER_URI: http://n8n-primary-service:5679
-
Has anyone encountered this specific failure with OFFLOAD_MANUAL_EXECUTIONS_TO_WORKERS in a K8s queue setup, and is there a known variable that needs to be explicitly set on the Worker or Main node to resolve this possible internal task routing issue?
Information on n8n setup
- n8n version: 1.115.3
- Database: postgres
- n8n setup: Main, Worker, Runner
- Running n8n via: GKE (Google’s Kubernetes)
- Operating system: Standard N8N containers