fix(install): stop waiting when the gateway service fails

[ericcurtin]

Summary

The installer now stops as soon as the gateway user service fails, instead of waiting 30s.

Related Issue

Closes #4040

Changes

• Check the user service state while waiting for the listener.
• A failed or auto-restarting unit ends the wait with a retry command.
• Starting and running units are ignored. macOS and snap are unchanged.

Testing

• Checks appropriate to the affected code and behavior pass
• Unit tests added/updated (if applicable)
• E2E tests added/updated (if applicable)

mise run test:install-sh passes.

Checklist

• Follows Conventional Commits
• Commits are signed off (DCO)
• Architecture docs updated (if applicable)

[copy-pr-bot]

This pull request requires additional validation before any workflows can run on NVIDIA's runners.

Pull request vetters can view their responsibilities here.

Contributors can view more details about this message here.

[ericcurtin]

If useful, please also try https://github.com/llmmanorg/llmman, which can launch agents in an OpenShell sandbox (--sandbox openshell).

[ericcurtin]

@elezar @drew PTAL when you get a chance. Thank you!

[github-actions]

Label test:e2e applied, but pull-request/4143 does not exist yet. A maintainer needs to comment /ok to test f6ead775ed3520f416e44f8e3b8891b0678e2594 to mirror this PR. Once the mirror exists, re-apply the label or re-run Branch E2E Checks from the Actions tab.

[johntmyers]

/ok to test f6ead77

[johntmyers]

gator-agent

PR Review Status

The independent code review found no blocking findings. This focused installer fix fits validated issue #4040, and Branch Checks, Helm Lint, and Branch E2E Checks are now running or queued for the current head.

Action required: @ericcurtin, add a short note under Linux in docs/about/installation.mdx explaining that the installer exits early when the gateway user service fails, prints diagnostics, and tells users to fix the cause before restarting it. Alternatively, a maintainer can explain here why this behavior change needs no docs update. Gator requires that documentation disposition before pipeline handoff.

Blocking code findings: None.
Carried findings: None.

Gator metadata

• Validation: Concentrated Linux installer bug fix for validated issue bug: install.sh waits 30s when the Linux gateway service fails to start #4040; related PR fix(install): stop waiting on a gateway unit that has failed #4092 is closed without merge.
• Docs: Direct installer behavior change; no docs update or maintainer-authored exemption yet.
• Checks: DCO and vouch pass; no merge conflicts. Current-head Branch Checks and Helm Lint are in progress; Trivy Changes passed.
• E2E: test:e2e applied; /ok to test posted with the freshly fetched full head SHA. The mirror is current and Branch E2E Checks is queued; no rerun is needed.
• Head SHA: f6ead775ed3520f416e44f8e3b8891b0678e2594
• Base SHA: ec49209da25be39840742df29b64ec694d159c2f
• Merge base SHA: ec49209da25be39840742df29b64ec694d159c2f
• Patch ID: d93585d75efc7ef8b4b66b76cea248337d87c628
• Gator payload: 10
• Review mode: initial
• Previous reviewed SHA: none
• Review budget exhausted: no
• Maintainer decision required: no
• Next state: gator:in-review

[johntmyers]

gator-agent

PR Review Status

The independent code review found no blocking findings. This focused installer fix fits validated issue #4040, and Branch Checks, Helm Lint, and Branch E2E Checks are now running or queued for the current head.

Action required: @ericcurtin, add a short note under Linux in docs/about/installation.mdx explaining that the installer exits early when the gateway user service fails, prints diagnostics, and tells users to fix the cause before restarting it. Alternatively, a maintainer can explain here why this behavior change needs no docs update. Gator requires that documentation disposition before pipeline handoff.

Blocking code findings: None. Carried findings: None.

Gator metadata

we can probably skip docs updates 👍

[ethandalebout-beep]

I like that this reports the service failure immediately and prints diagnostics before suggesting a restart. That should make troubleshooting clearer than a generic listener timeout. The test confirming that a running service still gets the full listener wait is a useful safeguard, too.