Harvester ui extension auto install sometimes failed in first run

[a110605]

Setup

• Rancher version: 2.10.0 ~ 2.14.0
• Rancher UI Extensions: harveter ui ext
• Browser type & version: Chrome

Describe the bug
Installing the harvester-ui-extension via the single click button in the Rancher UI always fails on first run for me. Espescilly for fresh installed Rancher.

But it's success for the second or third tries.

To Reproduce

1. Setup a fresh installed Rancher via docker or helm.
2. Go to Virtualization Management page
3. Click install button

Result
Auto install failed.

Expected Result

Should successfully installed if no network or resource race condition.

Screenshots

Additional context

[github-actions]

🎯 Agentic Issue Triage

Summary: The Harvester UI extension one-click install fails intermittently on first attempt (especially on fresh Rancher installations), but succeeds on subsequent retries. The root cause appears to be a race condition between Helm repository creation/sync and the chart install step.

🔍 Root Cause Analysis

The install flow lives in pkg/harvester-manager/list/harvesterhci.io.management.cluster.vue → installHarvesterExtension(), which calls into shell/utils/uiplugins.ts.

Two likely bugs exist in createHelmRepository (shell/utils/uiplugins.ts):

1. Premature loop exit (critical): fetched = true on line 294 sits outside the if (!fetched) block, meaning the polling while-loop always exits after a single iteration regardless of whether the Downloaded condition is True. So the function returns the HelmRepository object before the repo has actually been synced.

// Line 267-295 of shell/utils/uiplugins.ts
while (!fetched) {
  const repo = await store.dispatch(/* ... */);
  tries++;
  const downloaded = repo.status.conditions.find((s: any) => s.type === 'Downloaded');
  if (downloaded) {
    if (downloaded.status === 'True') {
      fetched = true;
    }
  }
  if (!fetched) {
    tries++;           // ← double-increment of `tries`
    if (tries > MAX_RETRIES) throw new Error('...');
    await new Promise((resolve) => setTimeout(resolve, RETRY_WAIT));
  }
  fetched = true;    // ← BUG: always exits after 1st iteration!
}

2. Double tries++: tries is incremented on line 274 and again on line 285 inside if (!fetched), causing the retry counter to advance twice per failed attempt, halving the effective retry window.

3. Insufficient refresh wait: refreshHelmRepository calls waitForState('active', 10000, 1000) + a fixed 2s extra delay. On a freshly bootstrapped Rancher with slow pod startup or network access, the catalog backend may not yet be fully ready, causing getLatestExtensionVersion to return null and abort the install.

🔁 Reproduction Steps

1. Deploy a fresh Rancher (docker or Helm) — this increases the chance of slow initialization.
2. Navigate to Virtualization Management.
3. Click the Install button for the Harvester UI extension.
4. Observe that the install reports failure (screenshot in issue).
5. Click Install again — it likely succeeds on the 2nd or 3rd attempt.

To make the race more deterministic, try on a slow network or with a resource-constrained cluster where the catalog pods take longer to start.

🛠️ Debugging Strategies

• Open the browser DevTools Network tab before clicking Install and capture the API calls. Look for:
  • GET /v1/catalog.cattle.io.clusterrepos/<repo-name> — check how many polls occur and whether Downloaded condition is ever True before proceeding.
  • POST to install the Helm chart — check the request payload and the response for an error body.
• Add console.log / console.error around the catch (error) {} block in installHarvesterExtension to surface the actual error.
• Temporarily increase MAX_RETRIES and RETRY_WAIT in shell/utils/uiplugins.ts and verify that the first-run failure goes away.
• Check Rancher server logs for any Helm operation errors that accompany the UI failure.

✅ Suggested Fix Checklist

• Fix the fetched = true placement in createHelmRepository so it only exits the loop when the repo is actually downloaded:

  // Move `fetched = true` inside the `if (downloaded.status === 'True')` block,
  // and remove the unconditional `fetched = true` at the bottom.

• Fix the double tries++ in createHelmRepository (remove one of the two increments).
• Consider adding a configurable delay or retry after refreshHelmRepository completes before calling getLatestExtensionVersion, to handle slow catalog backend responses on fresh installations.
• Improve error visibility in installHarvesterExtension — the catch (error) {} silently swallows errors; at minimum log them to the console to aid debugging.
• Add/update unit tests in shell/utils/__tests__/uiplugins.test.ts to cover the createHelmRepository polling logic and ensure it does not exit prematurely.
• Verify the fix manually on a fresh Rancher docker install where the bug is most reproducible.

📚 Relevant Code References

File	Relevance
shell/utils/uiplugins.ts	createHelmRepository, refreshHelmRepository, waitForUIExtension, waitForUIPackage
pkg/harvester-manager/list/harvesterhci.io.management.cluster.vue	installHarvesterExtension method — orchestrates the full one-click install flow
shell/dialog/InstallExtensionDialog.vue	Manual install dialog (for comparison with the auto-install path)

Generated by Agentic Triage for issue #17159 · ◷

[github-actions]

🤖 Issue grooming questions

Thanks for the bug report! The problem description and reproduction steps are clear, but a couple of things are missing before this can be fully triaged:

1. Error details: The screenshot shows a generic failure popup. Could you share the full text of the error message and/or any relevant browser console output or network request errors at the time of failure? This is needed to identify the root cause (e.g., a race condition, a missing resource, a timeout).

2. What does "successfully installed" look like? Please clarify the acceptance criteria — for example:

   • Clicking the install button results in the extension being installed on the first attempt, with no error popup
   • No manual retry is required on a fresh Rancher installation

This comment was generated automatically to help gather the information needed to groom this issue.

Generated by Daily Issue Grooming · ◷

[github-actions]

🤖 Issue grooming questions

The reproduction steps and the description of the intermittent first-run failure are clear and appreciated. However, the issue body has not been updated since the previous review, and the two earlier questions remain unanswered:

1. Error details: The screenshot shows a generic failure dialog. Could you share the full error message text and/or any relevant browser console errors or failed network requests visible at the time of the failure? This would help confirm (or rule out) the race condition identified in the triage analysis.

2. Acceptance criteria: What should the fix look like from a user perspective? For example:

   • Clicking the install button on a fresh Rancher installation results in successful extension installation on the first attempt
   • No retry is required under normal network and resource conditions

This comment was generated automatically to help gather the information needed to groom this issue.

Generated by Daily Issue Grooming · ◷