diagnose-staging

Diagnose Staging

Investigate problems in the staging environment using ArgoCD application data, workload logs, and Kubernetes events.

Arguments

• {application} — ArgoCD application name to investigate (required)

Process

Phase 1: Application Overview

Call get_application to capture:

• Sync status and health
• Current and target revision
• Last operation result (success/failure)
• Operation message (if failed)

Phase 2: Resource Health

Call get_application_managed_resources to find unhealthy resources:

• Filter by health.status != Healthy
• Identify failing Deployments, StatefulSets, Jobs
• Note any Missing or Unknown resources

Phase 3: Events

Call get_application_events for application-level events.

For each unhealthy resource found in Phase 2, call get_resource_events to get K8s-level events (ImagePullBackOff, CrashLoopBackOff, FailedScheduling, etc.)

Phase 4: Logs

For unhealthy Deployments/StatefulSets, call get_application_workload_logs:

• Specify the resource reference (kind, name, namespace)
• Look for error patterns: Exception, ERROR, Fatal, Connection refused, OOM

Phase 5: Cross-Reference with Code

# Check recent commits that might have caused the issue
git log --oneline -20

# Check the deployed revision vs HEAD
git log --oneline HEAD...{deployed-revision} 2>/dev/null || true

Phase 6: Root Cause Analysis

Apply 5-Whys:

1. Why is the application unhealthy? → {symptom}
2. Why did {symptom} happen? → {cause}
3. Continue to root cause

Phase 7: Report

## Staging Diagnosis: {application}

**Health:** {status} | **Sync:** {status}
**Revision:** {deployed} (target: {target})

### Root Cause
{One sentence}

### Evidence
| Source | Finding | Relevance |
|--------|---------|-----------|
| App status | {finding} | High |
| Pod events | {finding} | High |
| Workload logs | {finding} | Medium |
| Git history | {finding} | Low |

### Failing Resources
- {kind}/{name}: {status} — {reason}

### Fix
{Specific action: rollback to revision X, fix config Y, increase resource Z}

### Timeline
1. {time} — {what happened}
2. {time} — {what happened}

Related Skills

• /staging-status — Quick health overview
• /drift-check — Check if drift caused the issue
• /compare-environments — Compare with local environment