The cluster provisioning log doesn't show output and stays in connecting..

[izaac]

Setup

• Rancher version: v2.15-48f13eb709327344eb541f969422a2b23b3538be-head
• Rancher UI Extensions:
• Browser type & version: Chrome

Describe the bug
The Provisioning log tab of the cluster details page isn't showing any information about the provisioning progress. It's just stuck in connecting.

To Reproduce
Provision an RKE2 Cluster, go to the cluster details and click on the Provisioning log tab.

Result
Stuck with a connecting... message.

Expected Result
Show the logs generated during provisioning.

Screenshots

Additional context

[github-actions]

🎯 Agentic Issue Triage

Summary: The Provisioning Log tab for RKE2 clusters is stuck displaying "Connecting..." and never shows any log output. This appears to be a WebSocket connectivity issue between the dashboard UI and the Rancher backend's /v1/management.cattle.io.clusters/{id}/log endpoint.

📋 Issue Analysis

Type: Bug — regression in provisioning log WebSocket stream
Area: shell/detail/provisioning.cattle.io.cluster.vue — connectLog() method
Milestone: v2.15.0
Severity: High — provisioning logs are critical for diagnosing cluster provisioning failures

Root Cause Candidates:

The connectLog() method in shell/detail/provisioning.cattle.io.cluster.vue opens a WebSocket to the URL returned by this.value.mgmt?.linkFor('log'). The "Connecting..." message is displayed when logOpen === false && logs.length === 0. The logOpen flag is only set to true on EVENT_CONNECTED. Likely failure modes:

1. WebSocket URL is null/undefined: this.value.mgmt could be null or its links.log might not be present, causing url.replace(...) to throw before connection is attempted — though the tab visibility check (showLog) depends on hasLink('log') being truthy, so this is unlikely to be the direct cause.
2. WebSocket connection is refused/failing silently: EVENT_CONNECT_ERROR fires and sets logOpen = false, but the UI shows "Connecting..." rather than an error (the error branch only sets logOpen = false with no user-visible error message beyond the "connecting" text).
3. The backend /log link returns an HTTP 4xx/5xx: The WebSocket upgrade fails, triggering EVENT_CONNECT_ERROR.
4. Regression in URL construction: url.replace(/^http/, 'ws') — if the URL already starts with ws:// or wss:// this is a no-op, but if the URL is an empty string or unexpected format it could break.

🔍 Debugging Steps

1. Open DevTools → Network tab (filter by WS/WebSocket) while the Provisioning Log tab is open. Check:

   • Is a WebSocket connection being attempted?
   • What URL is it connecting to?
   • What is the HTTP status code of the upgrade request (101 = success, 401/403/404/5xx = failure)?

2. Open DevTools → Console — look for Connect Error log message (logged by EVENT_CONNECT_ERROR handler at line 701).

3. Check the raw management cluster object — in the Vue devtools or console, inspect $store.getters['management/byId']('management.cattle.io.cluster', <id>) and verify links.log is set and resolves to a valid URL.

4. Test the log endpoint directly via kubectl proxy or the Rancher API explorer: GET /v1/management.cattle.io.clusters/<id> and check the links.log field.

5. Compare with a working version — check git log around shell/detail/provisioning.cattle.io.cluster.vue for recent changes that may have changed how connectLog() works or how this.value.mgmt is resolved.

🛠️ Potential Fix Areas

• shell/detail/provisioning.cattle.io.cluster.vue — connectLog():

  • Add defensive check: if url is falsy, log an error and return early rather than calling .replace() on undefined.
  • On EVENT_CONNECT_ERROR, surface a more meaningful error to the user (e.g., show an error message instead of perpetually showing "Connecting...").
  • Consider checking if the existing socket is already in a connecting/connected state before creating a new one.

• shell/utils/socket.js: The socket will silently retry on errors (autoReconnect is true). If the backend is consistently rejecting the WS upgrade, this produces an infinite reconnect loop that never surfaces an error to the user — the "Connecting..." message stays forever.

✅ Suggested Sub-tasks

• Verify whether the links.log field is present in the management cluster object for RKE2 clusters on the affected version
• Check if there is a backend regression in the /log WebSocket endpoint for RKE2 clusters
• Add error state UI: when EVENT_CONNECT_ERROR fires after N retries, show an error message instead of "Connecting..."
• Add null guard for url before calling url.replace() in connectLog()
• Check for any recent changes to how mgmt is resolved on the provisioning cluster detail page
• Write/update Cypress E2E test in cypress/e2e/po/lists/cluster-provisioning-log.po.ts to cover the connecting error state

🔗 Relevant Code References

• Main component: shell/detail/provisioning.cattle.io.cluster.vue — connectLog() (~line 673), showLog computed (~line 588)
• Socket utility: shell/utils/socket.js — EVENT_CONNECTED, EVENT_CONNECT_ERROR, EVENT_DISCONNECTED
• Cypress PO: cypress/e2e/po/lists/cluster-provisioning-log.po.ts
• Rancher API: GET /v1/management.cattle.io.clusters/{id} → check links.log

Generated by Agentic Triage for issue #17687 · ● 493.7K · ◷

[github-actions]

🤖 Issue grooming questions

Thanks for the detailed report and screenshots! The problem statement and reproduction steps are clear. A couple of things are needed before this can move to triage:

1. Missing: Acceptance Criteria

Please add a concrete, testable checklist defining what "fixed" looks like. For example:

• The Provisioning log tab displays live log output when provisioning an RKE2 cluster.
• No "connecting..." state persists indefinitely once the cluster provisioning has started.

2. Missing: Technical Details

Could you share any error messages or relevant output from the browser's developer console (Console and Network tabs) while the tab is stuck in "connecting..."? Specifically:

• Are there any WebSocket connection errors or failed API requests visible in the Network tab?
• Are there any JavaScript errors in the Console tab?

This will help pinpoint the component or API endpoint responsible for the hang.

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

Generated by Daily Issue Grooming · ● 450.2K · ◷

[github-actions]

🤖 Issue grooming questions

The issue has been reviewed again following the update. The reproduction steps and problem description remain clear, but the two items previously requested are still outstanding in the issue body:

1. Still missing: Acceptance Criteria

Please add a concrete, testable checklist defining what "fixed" looks like. For example:

• The Provisioning log tab displays live log output when provisioning an RKE2 cluster.
• The tab does not stay in a "connecting..." state indefinitely once provisioning has started.

2. Still missing: Technical Details (browser console/network output)

When the tab is stuck on "connecting...", please share any relevant output from the browser's developer tools:

• Network tab (WS filter): Is a WebSocket connection being attempted? What URL? What is the HTTP status of the upgrade request?
• Console tab: Are there any JavaScript errors or a Connect Error message?

This information will help identify whether this is a UI bug, a backend issue, or a configuration/permissions problem.

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

Generated by Daily Issue Grooming · ● 544.8K · ◷

[github-actions]

🤖 Issue grooming questions

The issue has been reviewed again. The reproduction steps and problem description remain clear. The two items previously requested are still not present in the issue body:

1. Still missing: Acceptance Criteria

Please add a concrete, testable checklist defining what "fixed" looks like. For example:

• The Provisioning log tab displays live log output when provisioning an RKE2 cluster.
• The tab does not stay in a "connecting..." state indefinitely once provisioning has started.

2. Still missing: Technical Details (browser console/network output)

When the tab is stuck on "connecting...", please share any relevant output from the browser's developer tools:

• Network tab (WS filter): Is a WebSocket connection being attempted? What URL? What is the HTTP status of the upgrade request?
• Console tab: Are there any JavaScript errors or a Connect Error message?

This information will help identify whether this is a UI bug, a backend issue, or a configuration/permissions problem.

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

Generated by Daily Issue Grooming · ● 823.9K · ◷