Skip to content

[WSL2][Onboard] Ollama installed on Windows host successfully but fails to auto-start — "Timed out waiting for Ollama to start" reproducible on two machines #3262

@zNeill

Description

@zNeill

Description

Description

During nemoclaw onboard on WSL2, selecting "Install Ollama on Windows host (recommended)" downloads and installs Ollama on the Windows side successfully, but the subsequent auto-start fails with "Timed out waiting for Ollama to respond on host.docker.internal". The installer reports "✓ Installed" but Ollama never starts serving. Retrying the same menu option re-downloads and re-installs but still times out.

Reproduced on two independent WSL2 machines (2u1g-x570-1865 and 2u1g-x570-0313), both with Docker Desktop WSL backend. This blocks the entire WSL2 + local Ollama onboard path.
Environment
Machine 1: 2u1g-x570-1865 (10.63.135.21)
Machine 2: 2u1g-x570-0313 (10.176.228.245)
OS:            Windows + WSL2 Ubuntu 24.04.4 LTS
Architecture:  x86_64
Kernel:        6.6.87.2-microsoft-standard-WSL2
Node.js:       v22.22.2
npm:           10.9.7
Docker:        Docker Desktop 29.1.3 / 29.4.1 (WSL backend)
OpenShell CLI: openshell 0.0.36
NemoClaw:      v0.0.37
OpenClaw:      N/A (onboard not completed)
GPU:           None (CPU only)
Steps to Reproduce
1. WSL2 Ubuntu 24.04 with Docker Desktop running, no Ollama on Windows or WSL
2. Run: nemoclaw onboard --fresh --name ollama-wsl-test
3. At step [3/8] inference menu, select "Install Ollama on Windows host (recommended)" (option 7)
4. At security warning prompt (OLLAMA_HOST=0.0.0.0:11434), answer Y
5. Wait for install and auto-start
6. Observe timeout
7. Retry: select option 7 again → same result
Expected Result
Per test case T6002104:
1. Ollama installed on Windows: "✓ Installed: C:\Users\...\Ollama\ollama.exe"
2. Ollama auto-starts with OLLAMA_HOST=0.0.0.0:11434
3. "✓ Ollama responsive on host.docker.internal" printed
4. Onboard continues to step [4/8]
Actual Result
>>> Downloading Ollama for Windows...
######################################## 100.0%
>>> Installing Ollama...
>>> Install complete. Run 'ollama' from the command line.
✓ Installed: C:\Users\local-mercl\AppData\Local\Programs\Ollama\ollama.exe
Waiting for Ollama to respond on host.docker.internal...
Timed out waiting for Ollama to start on the Windows host.

Onboard falls back to the inference menu. Retrying produces the same timeout.
Ollama is installed but never starts serving on the Windows host.
Logs
Reproduced identically on both machines. No additional logs captured —
the timeout message is the only output before falling back to menu.

Bug Details

Field Value
Priority Unprioritized
Action Dev - Open - To fix
Disposition Open issue
Module Machine Learning - NemoClaw
Keyword NemoClaw, NEMOCLAW_GH_SYNC_APPROVAL, NemoClaw_Inference, NemoClaw_Onboard

[NVB#6158545]

Metadata

Metadata

Assignees

Labels

NV QABugs found by the NVIDIA QA Teamarea: local-modelsLocal model providers, downloads, launch, or connectivityarea: providersInference provider integrations and provider behaviorplatform: wslAffects Windows Subsystem for Linux

Type

No fields configured for Bug.

Projects

No projects

Milestone

No milestone

Relationships

None yet

Development

No branches or pull requests

Issue actions