UI incorrectly displays Active state for harvester provisioning cluster

[IsaSih]

Setup

• Rancher version:
  Rancher v2.14.1-alpha2
  Dashboard release-2.14 449e8c2
  Rancher Prime CLI v2.14.0

Describe the bug
UI incorrectly displays Active state while the harvester downstream cluster is still provisioning in the cluster list page and cluster details (Clusters Management)

To Reproduce
1- Provision a harvester downstream cluster
2- Observe the status in the clusters list from the clusters management page
3- Navigate to the clusters details page of the provioned cluster
4- You observe that UI is still displaying Active state while the node is still node ready

Result
UI incorreclty displays Active state while the harvester downstream cluster is still provisioning

Expected Result
Ui must display the Provisioning state while cluster is still provisioning and display Active status only after provisioning is complete.

Screenshots

[github-actions]

🎯 Agentic Issue Triage

Summary: During provisioning of a Harvester downstream cluster, the UI displays Active state instead of Provisioning, giving a misleading impression that the cluster is ready when it is not.

📋 Issue Analysis

Type: Bug — Incorrect state display in the cluster list/details page (Cluster Management)

Versions affected: Rancher v2.14.1-alpha2, Dashboard release-2.14 (449e8c2)

User Impact: Medium — Operators may attempt to use the cluster while it's still provisioning, potentially leading to failed operations or confusion. The cluster state badge is a key signal for readiness.

Likely Root Cause:

The state display for provisioning clusters is controlled by the _stateObj getter in shell/models/provisioning.cattle.io.cluster.js:

get _stateObj() {
  if (!this.isRke2) {
    return this.mgmt?.stateObj || this.metadata?.state;
  }
  return this.metadata?.state;
}

• Harvester downstream clusters are provisioned via RKE2 (spec.rkeConfig is set), so isRke2 returns true.
• For RKE2 clusters, the code returns only this.metadata?.state (the provisioning.cattle.io/Cluster resource), without falling back to this.mgmt?.stateObj (the management.cattle.io/Cluster resource).
• During provisioning, the provisioning.cattle.io/Cluster metadata state may transition to Active before the management cluster has finished and its own state is fully settled — meaning the displayed state goes Active prematurely.
• For non-RKE2 clusters, the code does fall back to this.mgmt?.stateObj, which reflects the actual management-side state more accurately.
• The isHarvester getter (!!this.mgmt?.isHarvester) is used in other parts of the model (e.g., machineProvider) but not in _stateObj, creating a gap for Harvester-hosted clusters.

🔍 Debugging Strategies

1. Confirm which state object is "Active": While reproducing the bug, use browser DevTools → Vue DevTools or the Rancher API Explorer to compare:

   • provisioning.cattle.io/Cluster → .metadata.state
   • management.cattle.io/Cluster → .metadata.state / .status.conditions
     Determine which one prematurely says Active.

2. Add a temporary log in shell/models/provisioning.cattle.io.cluster.js inside _stateObj:

   get _stateObj() {
     if (!this.isRke2) {
       return this.mgmt?.stateObj || this.metadata?.state;
     }
     // Temporary debug
     console.log('provisioning stateObj:', this.metadata?.state, 'mgmt stateObj:', this.mgmt?.stateObj);
     return this.metadata?.state;
   }

   Check if this.mgmt?.stateObj shows Provisioning while this.metadata?.state shows Active.

3. Check the stateParts getter (also in the same file) — it may aggregate sub-states to drive the badge color separately from stateObj.

4. Verify isHarvester returns true for the affected cluster during provisioning — if this.mgmt hasn't been loaded yet, isHarvester would be false.

🛠️ Suggested Fix Approach

The fix likely involves updating _stateObj to also consider the management cluster state for Harvester clusters (similar to non-RKE2 clusters), or applying an additional transitioning/provisioning check:

get _stateObj() {
  if (!this.isRke2) {
    return this.mgmt?.stateObj || this.metadata?.state;
  }

  // Harvester downstream clusters may show premature Active state from the provisioning resource
  if (this.isHarvester && this.mgmt?.stateObj) {
    return this.mgmt.stateObj;
  }

  return this.metadata?.state;
}

⚠️ This is a suggested starting point — care should be taken not to regress state display for non-Harvester RKE2 clusters. The condition should be narrowed based on what is confirmed during debugging.

📁 Relevant Files

• shell/models/provisioning.cattle.io.cluster.js — Primary: _stateObj, stateObj, isHarvester, isRke2 getters
• shell/models/management.cattle.io.cluster.js — Management cluster stateObj/isHarvester

✅ Suggested Sub-tasks

• Reproduce the issue on a Harvester downstream cluster in a dev environment
• Confirm which API resource (provisioning vs management) reports Active prematurely via DevTools / API Explorer
• Inspect the _stateObj getter behavior for a Harvester cluster during provisioning
• Implement a targeted fix in _stateObj that avoids regressing RKE2 non-Harvester clusters
• Add/update unit tests for _stateObj covering the Harvester provisioning scenario
• Verify the fix on both the cluster list and cluster detail pages

🔗 Relevant Resources

• provisioning.cattle.io.cluster.js model — _stateObj, isHarvester, isRke2 getters
• [Rancher Dashboard unit test guide]((extensions.rancher.io/redacted)

Generated by Agentic Triage for issue #17291 · ◷

[github-actions]

🤖 Issue grooming questions

Thanks for the clear reproduction steps and screenshots! The bug is well-described, but the following are needed before this can be triaged:

1. Acceptance criteria: Please add concrete, testable criteria that define what "fixed" looks like. For example:

   • The cluster list displays "Provisioning" state while a Harvester downstream cluster's nodes are not yet Ready
   • The cluster status transitions to "Active" only after provisioning is fully complete (cluster management page and cluster details page)

2. Technical details: Does the backend API already return the correct provisioning state, or is this a UI-only display issue? Any relevant API response fields or specific UI components returning the wrong state would help scope the fix.

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

Generated by Daily Issue Grooming · ◷

[richard-cox]

we might want to pause on this / retest once #16602 is resolved

[github-actions]

🤖 Issue grooming questions

Thanks for the note linking to #16602 — that's useful context. The following items are still needed before this issue can be triaged:

1. Acceptance criteria: Please add concrete, testable criteria that define what "fixed" looks like. For example:

   • The cluster list displays "Provisioning" state while a Harvester downstream cluster's nodes are not yet Ready
   • The cluster status transitions to "Active" only after provisioning is fully complete (both the cluster list and cluster details pages)

2. Technical details: Is this a UI-only display issue, or does the backend API already return the correct provisioning state? Any information about which API resource (provisioning.cattle.io/Cluster vs management.cattle.io/Cluster) reports the incorrect state would help scope the fix.

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

Generated by Daily Issue Grooming · ◷