Fetch failed on Ollama Node Despite Successful Container-to-Container Network Connection

[upland777]

Hello,

I am running n8n v1.98.2 in the ai-stack Docker environment and have a persistent fetch failed error that I cannot solve.

My Goal: Connect a Basic LLM Chain node to an Ollama Model node to communicate with a local Ollama service running in the same Docker Compose project.

The Error: I consistently get a fetch failed error with an empty error cause.

Here is a complete summary of my troubleshooting:

1. Network Connectivity: The network connection between the containers is working. I have used docker exec to get a shell inside the n8n container and a wget --spider http://ollama:11434 command connects successfully and instantly with a 200 OK response.
2. Timeout Issue: The problem is not a timeout. The fetch failed error occurs with both large models (llama3.2) and tiny models (tinydolphin:1.1b).
3. DNS Issue: I tried adding the NODE_OPTIONS=--dns-result-order=ipv4first environment variable to my docker-compose.yaml for the n8n service and restarted the stack, but the error remains exactly the same.
4. Node Installation: The n8n-nodes-ollama community node is correctly installed via the N8N_EXTENSIONS_INSTALL environment variable and the nodes appear in the n8n UI.

The core question is: Why is the Node.js fetch call inside the n8n node failing when a wget call from the exact same container to the exact same address succeeds?

This seems to be a very low-level issue. Any help or ideas would be greatly appreciated.

[jabbson]

Hey @upland777 hope all is well.

What about native node, the AI Agent, if you connect it to the ollama, does it work?

[upland777]

Thanks for the reply, just figured out problem, the credentials for the Local Ollama Model was set to http://localhost:11434, which did not communicate with the Docker Network. Changed it to http://ollama:11434 and it works fine now, it also works with the dhcp ip of the server.