Skip to content

INC-002 — n8n.atlantisits.co 502 Bad Gateway

Severity: Low (Non-Blocking) Status: Open First Observed: April 21, 2026 (V19 era) Logged: April 29, 2026 Logged By: Ozzy-Chat (Claude — Atlantis AI) Reviewed By: Opus-Ozzy (mental model confirmation)

Summary

The public n8n endpoint at n8n.atlantisits.co returns a 502 Bad Gateway error. Cloudflare Tunnel on Vegeta is unable to reach the n8n instance. This is a non-blocking incident — Nova Phase 2 web search is unaffected because Nova calls Trunks n8n at localhost:5678 directly, not the public endpoint.

Symptoms

Error 1033 — Cloudflare Tunnel error
https://n8n.atlantisits.co

You → Atlanta (Cloudflare) ✅ Working
Cloudflare → n8n.atlantisits.co ❌ Error
  • n8n.atlantisits.co returns 502 Bad Gateway
  • Cloudflare Ray ID visible — tunnel is configured but origin unreachable
  • Error 1033 = tunnel cannot connect to origin server

Suspected Causes (in order of likelihood)

  1. Cloudflare Tunnel down on Vegeta — cloudflared service stopped or token expired
  2. Check: sc query cloudflared (Windows Services on Vegeta)
  3. n8n process crashed on Vegeta — PM2 or Windows service down
  4. Check: pm2 status on Vegeta (if PM2 installed) or Task Manager
  5. Port mismatch in tunnel config — tunnel pointing at wrong port
  6. Should be: localhost:5678
  7. Vegeta n8n was Windows-native install — not WSL — may have different service management

Impact Assessment

Area Impact Notes
Nova Phase 2 web search ✅ Zero Nova calls Trunks localhost:5678 directly
JotForm webhooks ❌ Down Any workflow triggered by JotForm form submissions failing silently
Roofing Lead Engine notifications ❌ Down Discord/Telegram/Google Sheets alerts not firing on new leads
Plumbing/Electrical lead workflows ❌ Down Same as roofing
Internal n8n workflows on Trunks ✅ Unaffected Trunks n8n at localhost:5678 fully operational

Overall: Low immediate operational impact. High revenue impact if Lead Engine demos are live and leads come in without notification.

Resolution Options

Option A — Re-point Cloudflare Tunnel to Trunks (Quick Fix)

Re-configure the existing Cloudflare Tunnel to point at Trunks WSL2 n8n instead of Vegeta.

# On Trunks WSL2 — install cloudflared if not present
curl -L --output cloudflared.deb https://github.com/cloudflare/cloudflared/releases/latest/download/cloudflared-linux-amd64.deb
sudo dpkg -i cloudflared.deb

# Authenticate and create tunnel (or re-use existing token from Vegeta)
cloudflared tunnel login
cloudflared tunnel route dns <tunnel-id> n8n.atlantisits.co

Update tunnel config to point at localhost:5678 on Trunks.

Effort: ~30 minutes Risk: Low Recommended for: Immediate fix before VPS migration

Option B — VPS Migration (Proper Fix)

Deploy n8n on VPS (Hetzner) with a fresh Cloudflare Tunnel.

Effort: Part of Phase 1 VPS deployment (May 2026) Risk: Low — clean slate Recommended for: Permanent solution

Option C — Leave Open (Current Status)

Accept that public endpoint is down. Use localhost:5678 on Trunks for all local workflows. Defer until VPS or Option A when needed.

Effort: Zero Risk: Lead Engine notifications remain silent Recommended for: Until BCs arrive and first demo is scheduled

Resolution Steps (When Ready)

  1. SSH/RDP to Vegeta — check cloudflared service status 
    sc query cloudflared
  2. Check n8n process status on Vegeta
  3. Verify tunnel config points to correct local port (5678)
  4. If Vegeta n8n is dead → choose Option A (re-point to Trunks) or Option B (VPS)
  5. Restart services in order: n8n first, then cloudflared
  6. Retest: curl https://n8n.atlantisits.co/healthz
  7. Verify JotForm webhook fires → Discord/Telegram notification received

Prevention (Post-Resolution)

  • HT-018 (future): Cloudflare Tunnel Health Monitoring — NTFY alert when tunnel drops
  • Move n8n to VPS to eliminate Vegeta-as-public-host fragility
  • PM2 + pm2-logrotate on whichever machine hosts n8n going forward
  • Weekly tunnel health check in Friday Briefing cadence
  • HT-010 — Nova MCP Integration (confirms localhost:5678 as Nova's actual endpoint)
  • HT-014 — Windows Service Management
  • SOP-005 — Forgejo Deployment (pending — VPS context)
  • INC-001 — AAL v1.5 Rollback (resolved — reference for incident format)

Timeline

Date Event
~Apr 21, 2026 First observed — 502 on n8n.atlantisits.co
Apr 21, 2026 Logged in V19/V20 memory file changelog
Apr 29, 2026 Formal INC-002 document created
May 2026 Resolution target — Option A or B

Revision History

Version Date Author Notes
1.0 2026-04-29 Ozzy-Chat Initial formal incident document created

End of INC-002 — n8n.atlantisits.co 502 Bad Gateway