michael_dileo/Keybard-Vagabond-Demo
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity
Files
4351eb6f48873680737d1d051dc6e665cc61f5be
Keybard-Vagabond-Demo/docs/CLOUDFLARE-TUNNEL-NGINX-MIGRATION.md
Michael DiLeo 7327d77dcd redaction (#1) 
Add the redacted source file for demo purposes

Reviewed-on: https://source.michaeldileo.org/michael_dileo/Keybard-Vagabond-Demo/pulls/1
Co-authored-by: Michael DiLeo <michael_dileo@proton.me>
Co-committed-by: Michael DiLeo <michael_dileo@proton.me>
2025-12-24 13:40:47 +00:00

15 KiB
Raw Blame History

Cloudflare Tunnel to Nginx Ingress Migration

Project Overview

Goal: Route Cloudflare Zero Trust tunnel traffic through nginx ingress controller to enable unified request metrics collection for all fediverse applications.

Problem: Currently only Harbor registry shows up in nginx ingress metrics because fediverse apps (PieFed, Mastodon, Pixelfed, BookWyrm) use Cloudflare tunnels that bypass nginx ingress entirely.

Solution: Reconfigure Cloudflare tunnels to route traffic through nginx ingress controller instead of directly to application services.

Current vs Target Architecture

Current Architecture

Internet → Cloudflare Tunnel → Direct to App Services → Fediverse Apps (NO METRICS)
Internet → External IPs → nginx ingress → Harbor (HAS METRICS)

Target Architecture

Internet → Cloudflare Tunnel → nginx ingress → All Applications (UNIFIED METRICS)

Migration Strategy

Approach: Gradual rollout per application to minimize risk and allow monitoring at each stage.

Order: BookWyrm → Pixelfed → PieFed → Mastodon (lowest to highest traffic/criticality)

Application Migration Checklist

Phase 1: BookWyrm (STARTING) ⏳

  • Pre-migration checks
    • Verify BookWyrm ingress configuration
    • Baseline nginx ingress resource usage
    • Test nginx ingress accessibility from within cluster
    • Document current Cloudflare tunnel config for BookWyrm
  • Migration execution
    • Update Cloudflare tunnel: bookwyrm.keyboardvagabond.com → http://ingress-nginx-controller.ingress-nginx.svc.cluster.local:80
    • Test BookWyrm accessibility immediately after change
    • Verify nginx metrics show BookWyrm requests
  • Post-migration monitoring (24-48 hours)
    • Monitor nginx ingress pod CPU/memory usage
    • Check BookWyrm response times and error rates
    • Verify BookWyrm appears in nginx metrics with expected traffic
    • Confirm no nginx ingress errors in logs

Phase 2: Pixelfed (PENDING) 📋

  • Pre-migration checks
    • Verify lessons learned from BookWyrm migration
    • Check nginx resource usage after BookWyrm
    • Baseline Pixelfed performance metrics
  • Migration execution
    • Update Cloudflare tunnel: pixelfed.keyboardvagabond.com → nginx ingress
    • Test and monitor as per BookWyrm process
  • Post-migration monitoring
    • Monitor combined BookWyrm + Pixelfed traffic impact

Phase 3: PieFed (PENDING) 📋

  • Pre-migration checks
    • PieFed has heaviest ActivityPub federation traffic
    • Ensure nginx can handle federation bursts
    • Review PieFed rate limiting configuration
  • Migration execution
    • Update Cloudflare tunnel: piefed.keyboardvagabond.com → nginx ingress
    • Monitor federation traffic patterns closely
  • Post-migration monitoring
    • Watch for ActivityPub federation performance impact
    • Verify rate limiting still works effectively

Phase 4: Mastodon (PENDING) 📋

  • Pre-migration checks
    • Most critical application - proceed with extra caution
    • Verify all previous migrations stable
    • Review Mastodon streaming service impact
  • Migration execution
    • Update Cloudflare tunnel: mastodon.keyboardvagabond.com → nginx ingress
    • Update streaming tunnel: streamingmastodon.keyboardvagabond.com → nginx ingress
  • Post-migration monitoring
    • Monitor Mastodon federation and streaming performance
    • Verify WebSocket connections work correctly

Current Configuration

Nginx Ingress Service

# Main ingress controller service (internal)
kubectl get svc ingress-nginx-controller -n ingress-nginx
# ClusterIP: 10.101.136.40, Port: 80

# Public service (external IPs for Harbor)
kubectl get svc ingress-nginx-public -n ingress-nginx  
# LoadBalancer: 10.107.187.45, ExternalIPs: <NODE_1_EXTERNAL_IP>,<NODE_2_EXTERNAL_IP>

Current Cloudflare Tunnel Routes (TO BE CHANGED)

bookwyrm.keyboardvagabond.com → http://bookwyrm-web.bookwyrm-application.svc.cluster.local:80
pixelfed.keyboardvagabond.com → http://pixelfed-web.pixelfed-application.svc.cluster.local:80
piefed.keyboardvagabond.com → http://piefed-web.piefed-application.svc.cluster.local:80
mastodon.keyboardvagabond.com → http://mastodon-web.mastodon-application.svc.cluster.local:3000
streamingmastodon.keyboardvagabond.com → http://mastodon-streaming.mastodon-application.svc.cluster.local:4000

Target Cloudflare Tunnel Routes

bookwyrm.keyboardvagabond.com → http://ingress-nginx-controller.ingress-nginx.svc.cluster.local:80
pixelfed.keyboardvagabond.com → http://ingress-nginx-controller.ingress-nginx.svc.cluster.local:80
piefed.keyboardvagabond.com → http://ingress-nginx-controller.ingress-nginx.svc.cluster.local:80
mastodon.keyboardvagabond.com → http://ingress-nginx-controller.ingress-nginx.svc.cluster.local:80
streamingmastodon.keyboardvagabond.com → http://ingress-nginx-controller.ingress-nginx.svc.cluster.local:80

Monitoring Commands

Pre-Migration Baseline

# Check nginx ingress resource usage
kubectl top pods -n ingress-nginx

# Check current request metrics (should only show Harbor)
# Your existing query:
# (sum(rate(nginx_ingress_controller_requests{status=~"2.."}[5m])) by (host) / sum(rate(nginx_ingress_controller_requests[5m])) by (host)) * 100

# Monitor nginx ingress logs
kubectl logs -n ingress-nginx -l app.kubernetes.io/name=ingress-nginx --tail=50

Post-Migration Verification

# Verify nginx metrics include new application
# Run your metrics query - should now show BookWyrm traffic

# Check nginx ingress is handling traffic
kubectl logs -n ingress-nginx -l app.kubernetes.io/name=ingress-nginx --tail=20 | grep bookwyrm

# Monitor resource impact
kubectl top pods -n ingress-nginx

Rollback Procedures

Quick Rollback (Per Application)

  1. Immediate: Revert Cloudflare tunnel configuration in Zero Trust dashboard
  2. Verify: Test application accessibility
  3. Monitor: Confirm traffic flows correctly

Full Rollback (All Applications)

  1. Revert all Cloudflare tunnel configurations to direct service routing
  2. Verify all applications accessible
  3. Confirm metrics collection returns to Harbor-only state

Risk Mitigation

Resource Monitoring

  • nginx Pod Resources: Watch CPU/memory usage after each migration
  • Response Times: Monitor application response times for degradation
  • Error Rates: Check for increased 5xx errors in nginx logs

Traffic Impact Assessment

  • Federation Traffic: Especially important for PieFed and Mastodon
  • Rate Limiting: Verify existing rate limits still function correctly
  • WebSocket Connections: Critical for Mastodon streaming

Success Criteria

✅ Migration Complete When:

  • All fediverse applications route through nginx ingress
  • Unified metrics show traffic for all applications
  • No performance degradation observed
  • All rate limiting and security policies functional
  • nginx ingress resource usage within acceptable limits

Notes & Lessons Learned

Phase 1 (BookWyrm) - Status: PRE-MIGRATION COMPLETE ✅

Pre-Migration Checks (2025-08-25):

  • ✅ BookWyrm Ingress: Correctly configured with host bookwyrm.keyboardvagabond.com, nginx class, proper CORS settings
  • ✅ BookWyrm Service: bookwyrm-web.bookwyrm-application.svc.cluster.local:80 accessible (ClusterIP: 10.96.26.11)
  • ✅ Nginx Baseline Resources:
    • n1 (625nz): 9m CPU, 174Mi memory
    • n2 (br8rg): 4m CPU, 169Mi memory
    • n3 (rkddn): 14m CPU, 159Mi memory
  • ✅ Nginx Accessibility Test: Successfully accessed BookWyrm through nginx ingress with correct Host header
    • Response: HTTP 200, BookWyrm page served correctly
    • CORS headers applied properly
    • No nginx routing issues

Current Cloudflare Tunnel Config:

bookwyrm.keyboardvagabond.com → http://bookwyrm-web.bookwyrm-application.svc.cluster.local:80

Ready for Migration: All pre-checks passed. Nginx ingress can successfully route BookWyrm traffic.

Migration Executed (2025-08-25 16:06 UTC): ✅ SUCCESS

  • Cloudflare Tunnel Updated: bookwyrm.keyboardvagabond.com → http://ingress-nginx-controller.ingress-nginx.svc.cluster.local:80
  • Immediate Verification: BookWyrm web UI accessible, no downtime
  • nginx Logs Confirmation: BookWyrm traffic flowing through nginx ingress: 
    136.41.98.74 - "GET / HTTP/1.1" 200 [bookwyrm-application-bookwyrm-web-80]
143.110.147.80 - "POST /inbox HTTP/1.1" 200 [bookwyrm-application-bookwyrm-web-80]
  • Resource Impact: Minimal increase in nginx CPU (9-15m cores), memory stable (~170Mi)
  • Next: Monitor for 24-48 hours, verify metrics collection

METRICS VERIFICATION: ✅ SUCCESS!

  • BookWyrm now appears in nginx metrics query: bookwyrm.keyboardvagabond.com visible alongside <YOUR_REGISTRY_URL>
  • Unified metrics collection achieved: Both Harbor and BookWyrm traffic now measured through nginx ingress
  • Phase 1 COMPLETE: Ready to monitor for stability before Phase 2

Phase 2 (Pixelfed) - Status: PRE-MIGRATION STARTING ⏳

Lessons Learned from BookWyrm:

  • Migration process works flawlessly
  • nginx ingress handles additional load without issues
  • Metrics integration successful
  • Zero downtime achieved

Pre-Migration Checks (2025-08-25): ✅ COMPLETE

  • ✅ Pixelfed Ingress: Correctly configured with host pixelfed.keyboardvagabond.com, nginx class, 20MB upload limit, rate limiting
  • ✅ Pixelfed Service: pixelfed-web.pixelfed-application.svc.cluster.local:80 accessible (ClusterIP: 10.97.130.244)
  • ✅ nginx Post-BookWyrm Resources: Stable performance after BookWyrm migration
    • n1 (625nz): 8m CPU, 173Mi memory
    • n2 (br8rg): 10m CPU, 169Mi memory
    • n3 (rkddn): 11m CPU, 159Mi memory
  • ✅ nginx Accessibility Test: Successfully accessed Pixelfed through nginx ingress with correct Host header
    • Response: HTTP 200, Pixelfed Laravel application served correctly
    • Proper session cookies and security headers
    • No nginx routing issues

Current Cloudflare Tunnel Config:

pixelfed.keyboardvagabond.com → http://pixelfed-web.pixelfed-application.svc.cluster.local:80

Ready for Migration: All pre-checks passed. nginx ingress can successfully route Pixelfed traffic.

Migration Executed (2025-08-25 16:19 UTC): ✅ SUCCESS

  • Cloudflare Tunnel Updated: pixelfed.keyboardvagabond.com → http://ingress-nginx-controller.ingress-nginx.svc.cluster.local:80
  • Immediate Verification: Pixelfed web UI accessible, no downtime
  • nginx Logs Confirmation: Pixelfed traffic flowing through nginx ingress: 
    136.41.98.74 - "HEAD / HTTP/1.1" 200 [pixelfed-application-pixelfed-web-80]
136.41.98.74 - "GET / HTTP/1.1" 302 [pixelfed-application-pixelfed-web-80]
136.41.98.74 - "GET /sw.js HTTP/1.1" 200 [pixelfed-application-pixelfed-web-80]
  • Resource Impact: Stable nginx performance (3-10m CPU cores), memory unchanged
  • Multi-App Success: Both BookWyrm AND Pixelfed now routing through nginx ingress
  • Metrics Fix: Updated query to include 3xx redirects as success (status=~"[23]..")
  • PHASE 2 COMPLETE: Pixelfed metrics now showing correctly in unified dashboard

Phase 3 (PieFed) - Status: PRE-MIGRATION STARTING ⏳

Lessons Learned from BookWyrm + Pixelfed:

  • Migration process consistently successful across different app types
  • nginx ingress handles additional load without issues
  • Metrics integration working with proper 2xx+3xx success criteria
  • Zero downtime achieved for both migrations
  • Traffic patterns clearly visible in nginx logs

Pre-Migration Checks (2025-08-25): ✅ COMPLETE

  • ✅ PieFed Ingress: Correctly configured with host piefed.keyboardvagabond.com, nginx class, 20MB upload limit, rate limiting (100/min)
  • ✅ PieFed Service: piefed-web.piefed-application.svc.cluster.local:80 accessible (ClusterIP: 10.104.62.239)
  • ✅ nginx Post-2-Apps Resources: Stable performance after BookWyrm + Pixelfed migrations
    • n1 (625nz): 10m CPU, 173Mi memory
    • n2 (br8rg): 16m CPU, 169Mi memory
    • n3 (rkddn): 3m CPU, 161Mi memory
  • ✅ nginx Accessibility Test: Successfully accessed PieFed through nginx ingress with correct Host header
    • Response: HTTP 200, PieFed application served correctly (343KB response)
    • Proper security headers and CSP policies
    • Flask session handling working correctly
  • ✅ Federation Traffic Assessment: HEAVY ActivityPub load confirmed
    • 58 federation requests in last 30 Cloudflare tunnel logs
    • Constant ActivityPub /inbox POST requests from multiple Lemmy instances
    • Sources: lemmy.dbzer0.com, lemmy.world, and others
    • This will significantly increase nginx ingress load

Current Cloudflare Tunnel Config:

piefed.keyboardvagabond.com → http://piefed-web.piefed-application.svc.cluster.local:80

Ready for Migration: All pre-checks passed. ⚠️ CAUTION: PieFed has the heaviest federation traffic - monitor nginx closely during/after migration.

Migration Executed (2025-08-25 17:26 UTC): ✅ SUCCESS

  • Cloudflare Tunnel Updated: piefed.keyboardvagabond.com → http://ingress-nginx-controller.ingress-nginx.svc.cluster.local:80
  • Immediate Verification: PieFed web UI accessible, no downtime
  • nginx Logs Confirmation: HEAVY federation traffic flowing through nginx ingress: 
    135.181.143.221 - "POST /inbox HTTP/1.1" 200 [piefed-application-piefed-web-80]
135.181.143.221 - "POST /inbox HTTP/1.1" 200 [piefed-application-piefed-web-80]
Multiple ActivityPub federation requests per second from lemmy.world
  • Resource Impact: nginx ingress handling heavy load excellently
    • CPU: 9-17m cores (slight increase, well within limits)
    • Memory: 160-174Mi (stable)
    • Response times: 0.045-0.066s (excellent performance)
  • Load Balancing: Traffic properly distributed across multiple PieFed pods
  • Federation Success: All ActivityPub requests returning HTTP 200
  • PHASE 3 COMPLETE: PieFed successfully migrated with heaviest traffic load

Phase 4 (Mastodon) - Status: COMPLETE ✅

Migration Executed (2025-08-25 17:36 UTC): ✅ SUCCESS

  • Issue Encountered: Complex nginx rate limiting configuration caused host header validation failures
  • Root Cause: server-snippet and configuration-snippet annotations interfered with proper request routing
  • Solution: Simplified ingress configuration by removing complex rate limiting annotations
  • Fix Process:
    1. Suspended Flux applications to prevent config reversion
    2. Deleted and recreated ingress resources to clear nginx cache
    3. Applied clean ingress configuration
  • Cloudflare Tunnel Updated: Both Mastodon routes to nginx ingress:
    • mastodon.keyboardvagabond.com → http://ingress-nginx-controller.ingress-nginx.svc.cluster.local:80
    • streamingmastodon.keyboardvagabond.com → http://ingress-nginx-controller.ingress-nginx.svc.cluster.local:80
  • Immediate Verification: Mastodon web UI accessible, HTTP 200 responses
  • nginx Logs Confirmation: Mastodon traffic flowing through nginx ingress: 
    136.41.98.74 - "HEAD / HTTP/1.1" 200 [mastodon-application-mastodon-web-3000]
  • Performance: Fast response times (0.100s), all security headers working correctly
  • 🎉 MIGRATION COMPLETE: All 4 fediverse applications successfully migrated to unified nginx ingress routing!

Created: 2025-08-25
Last Updated: 2025-08-25
Status: Phase 1 (BookWyrm) Starting

Reference in New Issue View Git Blame Copy Permalink