app-platform-troubleshooting

App Platform Troubleshooting Skill

Transform debugging from guessing to rapid diagnosis and fix.

Philosophy

Traditional: See error → Guess → Change → Push → Wait 5-7 min → Repeat
With skill:  See error → Diagnose → Fix → Verify → Commit proper fix

Quick Decision

Is the app deployed with running containers?
├── YES → Can we access the shell?
│         ├── YES → LIVE MODE (SDK shell access)
│         └── NO  → LOGS-ONLY MODE (fetch logs)
└── NO (build/deploy failed) → LOGS-ONLY MODE

---

Mode 1: Live Troubleshooting (Quick Start)

from do_app_sandbox import Sandbox

app = Sandbox.get_from_id(app_id="<app-id>", component="web")

# Diagnostics
app.exec("env | grep DATABASE")
app.exec("curl -v localhost:8080/health")
app.exec("ps aux | head -10")

Full guide: See live-troubleshooting.md

---

Mode 2: Logs-Only (Quick Start)

# Runtime logs
doctl apps logs <app_id> <component> --type run

# Build logs
doctl apps logs <app_id> <component> --type build

# Crash logs
doctl apps logs <app_id> --type=run_restarted

Full guide: See logs-analysis.md

---

Debug Container (Infrastructure Issues)

Deploy in ~30-45 seconds to isolate infrastructure from application:

services:
  - name: debug
    image:
      registry_type: GHCR
      registry: ghcr.io
      repository: bikramkgupta/debug-python
      tag: latest
    http_port: 8080
    envs:
      - key: DATABASE_URL
        value: ${db.DATABASE_URL}

# Run validation suite
validate-infra all
validate-infra database
validate-infra kafka

Full guide: See debug-container.md

---

Quick Reference: Exit Codes

Code	Signal	Meaning
0	-	Clean exit (shouldn't exit)
1	-	General error
127	-	Command not found
137	SIGKILL	OOM killed
143	SIGTERM	Graceful shutdown

---

Quick Reference: Common Fixes

Problem	Quick Fix
App exits immediately	Check if listening on $PORT
502 errors	Check health endpoint, verify running
Database connection fails	Use Debug Container, verify trusted sources
Build fails	Check dependencies, review build logs
OOM kills	Upgrade instance size
Health checks fail	Bind to 0.0.0.0, not localhost
Slow startup	Increase initial_delay_seconds

---

Reference Files

• live-troubleshooting.md — SDK shell access, diagnostics, hot fixes
• logs-analysis.md — Log patterns, error codes, health check config
• debug-container.md — Infrastructure validation, validation suite
• networking-issues.md — DNS, CORS, VPC, routing
• app-maintenance.md — Archive/unarchive workflow
• diagnostic-scripts.md — Scripts, testing strategy, performance

---

When to Escalate

Contact DigitalOcean Support when:

• Internal error persists after redeploy
• Resource limit increases needed
• Multiple apps affected (platform issue)
• VPC/networking issues can't be diagnosed

Before escalating, gather:

doctl apps get <app_id> -o json > app_info.json
doctl apps logs <app_id> <component> --type run > runtime.log
doctl apps spec get <app_id> > app_spec.yaml

---

Integration with Other Skills

• → deployment: After fixing, deploy proper changes
• → devcontainers: Reproduce issues locally
• → postgres: Database-specific configuration
• → networking: Comprehensive networking docs