Mirror Pod Testing

Mirror Pod Testing lets you deploy a temporary copy of a failing pod with the AI-generated fix applied. The mirror runs alongside the original pod under a different name, is excluded from monitoring and security scanning, and auto-deletes after a configurable TTL.

This closes the gap between “the AI says do X” and “I actually merged the fix into git” — you can verify the fix works in your real cluster, with your real config, before touching source control.

How it works

1. Expand a failing pod and click Test Fix
2. The AI generates a fixed YAML manifest based on the failure context (events, logs, current manifest)
3. (Optional) Click Edit Manifest to review and tweak the fix
4. Click Deploy — backend creates a temporary mirror pod with the fix applied
5. Watch the mirror pod status with a live countdown
6. Mirror auto-deletes after the TTL (default: 3 minutes)

Features

• AI-generated manifest fixes based on failure context (events, logs, current manifest)
• Manifest editor to review and modify the fix before deploying
• Cleanup — strips Kubernetes runtime noise (auto-injected tolerations, service-account tokens, scheduler fields) from the fixed manifest
• Configurable TTL (30 seconds to 60 minutes) via the Admin panel
• Live status tracking with a countdown timer
• Auto-excluded from monitoring and security scanning so the mirror doesn’t clutter your dashboard
• Manual delete — clean up early without waiting for the TTL

Configuration

Mirror pod TTL:

1. Admin Panel → Settings
2. Adjust Mirror Pod TTL (default: 180 seconds)
3. Valid range: 30 – 3600 seconds

Mirror pods are deleted as soon as the TTL expires. Cleanup runs as a background task in the backend.

Requirements

The backend ServiceAccount needs pod create/delete permissions. The Helm chart’s RBAC includes these by default:

- apiGroups: [""]
  resources: ["pods"]
  verbs: ["get", "list", "create", "delete"]
- apiGroups: [""]
  resources: ["pods/log"]
  verbs: ["get"]
- apiGroups: [""]
  resources: ["events"]
  verbs: ["list"]

If you installed via raw manifests without the bundled RBAC, mirror pod deploys will fail with 403.

Roles

• read users see the manifest preview but cannot deploy the mirror.
• write users… same as read for mirror pods (admin-only).
• admin users can preview, deploy, and delete mirror pods.

The “Test Fix” button only appears for users with the right role and when the pod failure has an AI solution.

API

All mirror endpoints live under /api/mirror/. See the API Reference for full request and response shapes.

Method	Path	Purpose
POST	/api/mirror/preview/{pod_id}	Generate AI-fixed manifest preview
POST	/api/mirror/deploy/{pod_id}	Deploy mirror pod (admin)
GET	/api/mirror/status/{mirror_id}	Mirror status + countdown
DELETE	/api/mirror/{mirror_id}	Delete early (admin)
GET	/api/mirror/active	List currently active mirrors
GET / PUT	/api/admin/settings/mirror-ttl	Read or change the default TTL

Troubleshooting

Symptom	Cause / fix
”Test Fix” button missing	Logged in as read / write (admin only); or the pod has no AI solution yet
Mirror fails to deploy	Backend RBAC missing pod create/delete; namespace quota exceeded; image pull error (mirror uses the same image as the original)
Stuck in Pending	Node selectors / affinity from the original can’t be satisfied; resources unavailable
Not auto-deleting	Backend pod isn’t running (cleanup is a background task); check backend logs for cleanup errors

See Troubleshooting for full debugging steps.