Skip to content

[WSL2][Onboard] Windows-host Ollama reuse exits during provider configuration #8596

Description

@mercl-lau

Warning

Regression — the equivalent Windows-host Ollama handoff failure was previously fixed in GitHub issue #4208, but is reproducible again in NemoClaw 0.0.102. The exact last-good release was not re-tested.

Description

On WSL2 with Docker Desktop, NemoClaw detects a running Windows-host Ollama service but exits during inference-provider configuration before creating the sandbox. The observed error says that an Ollama systemd loopback override could not be applied, even though Ollama is running on the Windows host rather than as a Linux systemd service inside WSL.

Platform scope: Reproduced on Windows 11 with WSL2 Ubuntu 24.04 and Docker Desktop; other Windows-host Ollama configurations were not tested.

Regression: Yes — the equivalent Windows-host Ollama handoff failure was previously fixed in GitHub issue #4208, but is reproducible again in NemoClaw 0.0.102. The exact last-good release was not re-tested.

Environment

Device:        Windows 11 PC with WSL2
OS:            Ubuntu 24.04 inside WSL2
Architecture:  x86_64
Node.js:       Not captured
npm:           Not captured
Docker:        Docker Desktop (exact version not captured)
OpenShell CLI: openshell 0.0.85
NemoClaw:      nemoclaw v0.0.102
OpenClaw:      N/A (onboard did not complete)

Steps to Reproduce

  1. On Windows 11, enable WSL2 Ubuntu 24.04 and Docker Desktop's WSL2 backend.

  2. Install and start Ollama on the Windows host. Ensure the Ollama executable is not present on the WSL PATH.

  3. From WSL, verify the Windows Ollama API is reachable:

    curl -fsS http://host.docker.internal:11434/api/tags
  4. Run:

    NEMOCLAW_PROVIDER=install-windows-ollama \
    NEMOCLAW_MODEL={existing-model} \
    NEMOCLAW_POLICY_MODE=skip \
    nemoclaw onboard --non-interactive --no-gpu --fresh
  5. Observe inference-provider configuration and check whether onboarding reaches sandbox creation.

Expected Result

NemoClaw reuses the running Windows-host Ollama daemon without attempting to modify a Linux systemd Ollama service. Onboarding exits successfully, creates the sandbox, and reaches Ready.

Actual Result

NemoClaw reaches inference-provider configuration, selects Ollama, attempts a Linux systemd loopback override, and exits with code 1 before creating the sandbox.

Logs

[3/8] Configuring inference provider
[non-interactive] Provider: ollama
Configuring Ollama systemd loopback override...
Failed to apply Ollama systemd loopback override.
Refusing to continue with a potentially non-loopback Ollama bind.

Onboarding did not finish.
Exit code: 1

NVB#6576493

Activity

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

Labels

NV QABugs found by the NVIDIA QA Teamarea: inferenceInference routing, serving, model selection, or outputsarea: onboardingOnboarding FSM, provider setup, sandbox launch, or first-run flowarea: providersInference provider integrations and provider behaviorplatform: windowsAffects native Windows environmentsplatform: wslAffects Windows Subsystem for Linuxprovider: ollamaOllama local model provider behavior

Type

Projects

No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions