Webhook Streaming Not Working in Queue Mode (Main + Workers)

Questions
1

Problem Summary:

When running in Regular Mode, both Test URL and Production URL work perfectly with streaming responses. However, after switching to Queue Mode (1 Main + 2 Workers), the Production URL stops working while the Test URL continues to work normally.

Environment:

  • n8n version: 1.123.5

  • Deployment: Kubernetes (self-hosted)

  • Redis: Connected and healthy

    Behavior Comparison:

    Execution Mode Test URL (/webhook-test/ki-cn) Production URL (/webhook/ki-cn)
    Regular Mode :white_check_mark: Works with streaming :white_check_mark: Works with streaming
    Queue Mode :white_check_mark: Works with streaming :cross_mark: Does NOT work

Workflow includes AI Agent nodes with LangChain for streaming chat responses.

Questions:

  1. Is there a way to make Production webhook URLs work in Queue Mode with streaming?

    • Are there specific configurations or environment variables needed?

    • Is there a workaround or recommended architecture?

  2. If webhook streaming in Queue Mode is officially not supported:

    • Can you please confirm this limitation?

    • Are there any plans to support it in future versions?

    • What is the recommended architecture for scalable streaming webhooks?

What I Need:

Either:

  • A solution/configuration to make Production webhooks work in Queue Mode with streaming, OR

  • Official confirmation that this is not supported, so I can plan alternative architectures

Any guidance would be greatly appreciated. This is critical for our AI chatbot deployment where streaming responses and horizontal scalability are both essential requirements.

Thank you!

2

The issue is likely related to how webhook requests are routed in Queue Mode. In Queue Mode, webhook requests need to be processed by the same worker that initiated the workflow, but production webhooks might not be maintaining this session continuity. Try enabling the `EXECUTIONS_MODE=queue` and `WEBHOOK_TUNNEL_URL` environment variables on both main and worker instances, ensuring they point to your main instance’s URL. This helps maintain session consistency across workers.