NEW → QUEUED_FOR_RCA → RCA_IN_PROGRESS → RCA_COMPLETED → RCA_GENERATED → Resolved

---

### Insights → Dashboard

**From Dashboard Activity Feed** → Click event → Opens in Insights

**Use case:** - You see "OOMKilled" in Dashboard feed
- Click to see full RCA
- Opens Insights with incident expanded

**Bi-directional navigation** keeps context flowing.

---

### Insights → Memory Engine

**Automatic integration:** - Every resolved incident stored
- Root causes saved to knowledge base
- Resolution patterns learned
- Speeds up future RCA

**You don't see this** - happens in background.

**Benefit**: Each incident makes RubixKube smarter for next time.

---

## Best Practices

<Accordion title="1. Review Insights Daily">
### Morning routine:
1. Open Insights page
2. Check RCA coverage (target: 80%+)
3. Filter by HIGH severity
4. Review new incidents since yesterday
5. Apply recommended actions

**Time required**: 5-10 minutes

**Benefit**: Proactive issue resolution before escalation
</Accordion>

<Accordion title="2. Triage by Severity">
### Priority order:
**critical** (red badge)
- Drop everything, resolve immediately
- System-wide impact
- Data loss risk

**high** (red badge)
- Resolve within hours
- Significant user impact
- Service degradation

**medium** (yellow badge)
- Resolve within 1-2 days
- Moderate impact
- Workarounds exist

**low** (gray badge)
- Resolve next sprint
- Minimal impact
- Informational

**Always check header** "Critical Issues" and "High Priority" counts first.
</Accordion>

<Accordion title="3. Trust High-Confidence RCA">
### When confidence is 70%+:
- Implement recommendations directly
- No need for extensive validation
- RubixKube has solid evidence

### When confidence is below 70%:
- Review Evidence tab carefully
- Cross-check with manual investigation
- Use "Ask AI How" for guidance
- Validate before implementing

**Our example** (40-50% confidence):
- Due to diagnostic tool failures
- Manual verification needed
- Still provides valuable direction
</Accordion>

<Accordion title="4. Use Filters Strategically">
### Common filter combinations:
**Production-only incidents:** - Namespace: production
- Severity: high, critical

**Recent failures:** - Status: Active
- Sort: Newest

**Specific pod issues:** - Search: "pod-name"
- Issue Type: CrashLoop, OOMKilled

**Unresolved RCA:** - Status: Active
- Only show incidents with RCA badge

**Pro tip**: Clear filters between sessions using "Clear filters" button
</Accordion>

<Accordion title="5. Document Resolutions">
### After fixing an incident:
1. Click "Apply" on each implemented recommendation
2. Add notes in "More actions" menu (if available)
3. Take screenshot of RCA for postmortem
4. Share learnings with team

**Why this matters:** - Builds organizational knowledge
- Helps Memory Engine learn faster
- Creates audit trail
- Prevents repeated incidents

**Future enhancement**: RubixKube will auto-detect resolutions
</Accordion>

<Accordion title="6. Leverage Chat Integration">
### Don't investigate alone:
**For every incident:** - Click "Ask AI" button
- Chat has full context already
- Ask for explanations, steps, similar incidents

**Example workflow:** 1. See OOMKilled incident
2. Click "Ask AI"
3. Ask: "Show me memory usage trends"
4. Ask: "What should I set the memory limit to?"
5. Ask: "Has this pod OOMKilled before?"

**Chat makes RCA actionable** with interactive guidance.
</Accordion>

---

## Quick Reference

### Insights Page Elements

| Element | Purpose | Action |
|---------|---------|--------|
| **Health Metrics**  | Overview of incident coverage | Check daily, target 80%+ RCA coverage |
| **Search Bar**  | Find specific incidents | Type pod name, namespace, or error type |
| **Severity Filter**  | Prioritize by impact | Filter for "high" and "critical" first |
| **Incident Cards**  | List all incident groups | Click to expand details |
| **RCA Analysis Tab**  | Root cause findings | Read before taking action |
| **Recommendations**  | Prioritized fixes | Click "Apply" or "Ask AI How" |
| **Timeline Tab**  | Incident progression | Use for debugging RCA issues |
| **Evidence Tab**  | Investigation data | Review for low-confidence RCA |

---

### Keyboard Shortcuts

While on Insights page:
- `R` - Refresh data
- `F` - Focus search bar
- `1-4` - Jump to first 4 incidents
- `Tab` - Navigate between tabs (Overview, RCA, Timeline, Evidence)
- `Esc` - Close expanded incident

*(Note: Keyboard shortcuts may vary by implementation)*

---

## Common Scenarios

### Scenario 1: New OOMKilled Incident

**What you see:** - HIGH severity badge
- "OOMKilled in Pod/memory-hog-demo"
- RCA badge present

### What to do:
1. Click incident to expand
2. Go to RCA Analysis tab
3. Read Root Cause (memory limit too low)
4. Check Recommended Actions
5. Click "Ask AI How" on "Increase memory limit" action
6. Implement suggested limit (e.g., increase from 50Mi to 150Mi)
7. Click "Apply" when done
8. Monitor for resolution

**Expected outcome**: Pod stops crashing, incident auto-resolves.

---

### Scenario 2: Low RCA Confidence

**What you see:** - Incident with 40% confidence
- "Status: RCA_GENERATED" but low certainty

### What to do:
1. Click Evidence tab
2. Check Investigation Completeness (30%)
3. See which tools failed (e.g., get_pod_logs)
4. Perform manual investigation:

5. Click "Ask AI" with manual findings
6. Chat combines RCA + your data for better diagnosis

**Key lesson**: Low confidence doesn't mean wrong, just uncertain.

---

### Scenario 3: Incident Without RCA

**What you see:** - "PodPending in Pod/broken-image-demo"
- No RCA badge
- Only shows Suggestions (not full RCA)

**Why this happens:** - RCA not complete yet (check Timeline)
- RCA failed (check Timeline for errors)
- Incident type doesn't trigger RCA
- RCA Pipeline agent offline

### What to do:
1. Check Timeline tab for RCA status
2. If "QUEUED_FOR_RCA" but no progress:
- RCA Pipeline may be stuck
- Go to Agents page, check RCA Pipeline Agent
3. If no RCA triggered:
- Use Suggestions section for generic fixes
- Click "Ask AI" for Chat investigation
4. Manual investigation:

Look for ImagePullBackOff errors

---

### Scenario 4: Multiple Related Incidents

**What you see:** - "CrashLoop in Pod/memory-hog-demo" (MEDIUM)
- "OOMKilled in Pod/memory-hog-demo" (HIGH)
- Same pod, different incident types

**What this means:** - Pod crashed due to OOM
- OOM is root cause
- CrashLoop is symptom

### What to do:
1. Open HIGH severity incident first (OOMKilled)
2. Read RCA (memory limit too low)
3. Fix memory limit
4. Both incidents should resolve together
5. Mark both as related in notes

**Pro tip**: RubixKube groups related incidents when possible.

---

## Troubleshooting Insights Issues

### Insights page not loading

**Symptoms**: Spinner forever, "Loading insights..." never completes

**Causes:** - Backend API connection issue
- RCA Pipeline not responding
- Database query timeout

### Solutions:
1. Hard refresh: Cmd+Shift+R (Mac) or Ctrl+Shift+R (Windows)
2. Check Dashboard → Agents → Verify RCA Pipeline is active
3. Check browser console for errors
4. Try different browser
5. Contact support if persists

---

### RCA not generating

**Symptoms**: Incidents stuck in "QUEUED_FOR_RCA" for >5 minutes

**Causes:** - RCA Pipeline Agent offline
- Task queue full
- Resource constraints

### Solutions:
1. Go to Agents page
2. Check RCA Pipeline Agent status
3. If degraded, restart:

4. Check Timeline tab for error messages
5. Verify cluster has resources for RCA workload

---

### Low RCA coverage (below 60%)

**Symptoms**: "Health: 55% RCA coverage" in header

**Causes:** - Many incidents without RCA
- RCA failures
- Complex incidents taking longer

### Solutions:
1. Filter by Status: Active, no RCA badge
2. Check those incidents' Timeline tabs for RCA failures
3. Verify diagnostic tools working (get_pod_logs, get_pod_events)
4. Check RCA Pipeline Agent logs:

5. May need to tune RCA timeout settings in Settings

---

### Evidence completeness always low

**Symptoms**: Investigation Completeness consistently 30-40%

**Causes:** - Diagnostic tools failing
- Permission issues
- Missing integrations

### Solutions:
1. Check which tools are failing (Evidence tab shows source)
2. Verify RubixKube has proper RBAC permissions:

3. Check integration connections in Settings → Integrations
4. Review RubixKube Observer Agent logs for errors

---

## What You Learned

<CardGroup cols={2}>
<Card title="Health Metrics" icon="gauge-high">
 - 75% RCA coverage
 - 4 total incident groups
 - 0 critical, 1 high priority
 - Daily monitoring targets
</Card>

<Card title="Incident Structure" icon="list">
 - Severity levels (critical, high, medium, low)
 - Incident groups and items
 - Status lifecycle (NEW → RESOLVED)
 - RCA badges and completion
</Card>

<Card title="RCA Analysis" icon="magnifying-glass">
 - Root Cause identification
 - Contributing Factors analysis
 - Impact Assessment scope
 - Confidence levels explained
</Card>

<Card title="Recommendations" icon="list-check">
 - Prioritized actions (HIGH/MEDIUM)
 - Owner assignments
 - Apply/Ask AI How/Dismiss buttons
 - Remediation tracking
</Card>

<Card title="Timeline View" icon="clock">
 - Chronological event progression
 - RCA processing states
 - Investigation duration
 - Actor attribution
</Card>

<Card title="Evidence Collection" icon="folder-open">
 - Investigation completeness percentage
 - Data sources (SearchAgent, logs, events)
 - Diagnostic tool status
 - Copy and share evidence
</Card>

<Card title="Filtering" icon="filter">
 - Search by name/namespace/type
 - Severity, Issue Type, Namespace, Status
 - Multiple selection support
 - Clear filters button
</Card>

<Card title="Workflows" icon="diagram-project">
 - Daily review routine
 - Emergency response steps
 - Troubleshooting approach
 - Integration with Chat
</Card>
</CardGroup>

---

## Next Steps

<CardGroup cols={2}>
<Card
 title="Back to Dashboard"
 icon="gauge"
 href="/using/dashboard"
>
 Monitor overall system health and active incidents
</Card>

<Card
 title="Use Chat for Investigation"
 icon="comments"
 href="/tutorials/chat-troubleshooting"
>
 Ask AI about incidents with full RCA context
</Card>

<Card
 title="View Infrastructure"
 icon="sitemap"
 href="/using/infrastructure"
>
 See affected resources in topology view
</Card>

<Card
 title="Check Agent Status"
 icon="robot"
 href="/using/agents"
>
 Verify RCA Pipeline and Observer agents are healthy
</Card>
</CardGroup>

---

## Related Documentation

<CardGroup cols={2}>
<Card title="Agent Mesh Concepts" icon="network-wired" href="/concepts/agent-mesh">
 Learn how RCA Pipeline generates analysis
</Card>

<Card title="What is SRI?" icon="brain" href="/concepts/what-is-sri">
 Understand Site Reliability Intelligence philosophy
</Card>

<Card title="Memory Engine" icon="database" href="/concepts/memory-engine">
 How RubixKube learns from past incidents
</Card>

<Card title="Guardrails" icon="shield-halved" href="/concepts/guardrails">
 Safety measures and confidence thresholds
</Card>
</CardGroup>

---

## Need Help?

import ContactSupport from '/snippets/contact-support.mdx';

<ContactSupport />

---

## Feedback

Found an issue with this guide or have suggestions?

-**Email** : [[email protected]](mailto:[email protected])
-**Subject** : "Insights Guide Feedback"

---

*Last updated: October 6, 2025*
*Guide version: 2.0*
*Based on RubixKube Console v1.0*

---

## Related Guides

- [Dashboard](/using/dashboard)
- [Infrastructure](/using/infrastructure)
- [Agents](/using/agents)
- [Clusters](/using/clusters)
- [Workspace](/using/workspace)
- [Settings](/using/settings)
- [Integrations](/using/integrations)