diff --git a/docs/snarfnet_homelab_topology.html b/docs/snarfnet_homelab_topology.html index 497bcc3..896b683 100644 --- a/docs/snarfnet_homelab_topology.html +++ b/docs/snarfnet_homelab_topology.html @@ -3,831 +3,979 @@ - SnarfNet Homelab Topology + SnarfNet Homelab Architecture & Topology - -
+ - -
-
-

Homelab

- snarfnet.net + +
+
+
+ 🪐 +

+ SnarfNet | Infrastructure & Topology +

+ + ALL CLUSTERS ACTIVE +
-
November 2026
+

+ Complete operational schematic mapping dual-WAN routing, VLAN segmentation, 4 compute clusters (Aerodrome, Spaceport, Spacedock, Shipyard), AI inference fabrics, and storage layers. +

- -
- - -
- -
-

- Networks - VLANs & Subnets -

-
-
-
- LAN - Core & Gateway -
- 14.44.4.0/25 -
-
-
- WAN 1 - Fiber ONT -
- 174.106.252.0/24 -
-
-
- WAN 2 - 5G Backup -
- 192.168.225.0/24 -
-
-
- VLAN 20 - Compute Fleet -
- 140.44.4.0/25 -
-
-
- VLAN 3 - Spacedock HCI -
- 140.44.1.0/24 -
-
-
- VLAN 40 - Storage SAN -
- 14.10.10.0/25 -
-
-
- VLAN 50 - Smart Home -
- 14.4.4.0/25 -
-
-
- Tailscale - Mesh Tailnet -
- 100.64.0.0/10 -
-
-
- - -
-

Legend

-
-
- - wired link, colored by network -
-
- - internet uplinks (fiber, 5G) -
-
- - Wi-Fi (not a cable) -
-
- - planned / expansion -
-
- - line width = speed (10G SFP+ vs 1G) -
-
-
app.snarfnet.net served by Ingress
-
:port published service port
-
-
-
+ +
+
+ + ⌘K
- - -
- - -
- -
-
Internet
-
two independent uplinks
-
-
- - - - - - - - - - -
- -
-
- WAN 1 • Fiber ONT -
-
1 Gbps Symmetric
-
10G SFP+ to UDM-SE
-
- - -
-
- WAN 2 • 5G Modem -
-
Cellular Standby
-
GbE RJ45 Failover
-
-
- - - - - - - - -
-
-
-
🛡️
-
-

UDM-SE • gateway & router

-
14.44.4.1 • UniFi OS v4.1
-
-
- 10G SFP+ Backbone -
- - -
-
-
Hardware
-
-
CPU: 4× Cortex-A57 @ 1.7G
-
RAM: 4 GB DDR4
-
Storage: 128 GB SSD
-
Ports: Dual 10G + 8× GbE PoE
-
-
- -
-
Services
-
-
• Suricata IPS/IDS L5
-
• WireGuard Server
-
• Inter-VLAN L3 Routing
-
• Deep Packet Inspection
-
-
- -
-
Uplink Policies
-
-
• WAN 1: Fiber (Primary)
-
• WAN 2: 5G (Hot Standby)
-
• VLAN 20 Compute Priority
-
• Smart Queues FQ-CoDel
-
-
-
-
- - - - - - - - - - - - -
- -
-
- - Wi-Fi clients • wireless -
-
U6-Enterprise & U6-Pro APs
-
Phones, Laptops, Smart TV, IoT
-
- - -
-
- Core Switch • 2.5G/10G -
-
USW-Enterprise-24-PoE
-
10G SFP+ Uplink to Fleet
-
-
- - - - - - - + +
+
Topology Release
+
Phase 8 • Nov 2026
- - -
- -
-

- Ways in from outside - Zero Open Ports -

-
-
-
-
- Cloudflare Tunnel -
- HTTPS -
-
*.snarfnet.net (Zero Inbound Ports)
-
Edge DDoS & WAF Protection
-
- -
-
-
- Tailscale & WireGuard -
- Tailnet -
-
100.64.0.0/10 • Mesh Interconnect
-
Direct Admin Subnet Route
-
-
-
- - -
-

GitOps & CI/CD Pipeline

-
- -
- 1. Author - Developer Commit - git push -
- -
↓
- -
- 2. Repository - Gitea Webhook - git.snarfnet.net -
- -
↓
- -
- 3. Reconcile - Argo CD Sync - selfHeal: true -
- -
↓
- -
- 4. Target - Aerodrome & Spaceport - K3s Deploy -
-
-
- - -
-

Outside services it relies on

-
- Cloudflare - GitHub - Let's Encrypt - Docker Hub - Tailscale DERP -
-
-
-
+
- -
- - -
-
-
⚡
-
-

- snarfnet • core infrastructure fabric + +
+ + +
+ + +
+
+
+

+ + Network VLAN Fabric

-
4 Clusters • 50+ Containers & Pods • Reconciled Declaratively via GitOps
+ 8 Zones
+ +
+
+
+ VLAN 1 LAN / Core +
UDM-SE Gateway, Core Switches
+
+ 14.44.4.0/25 +
+ +
+
+ VLAN 2 Management +
APs, USP RPS, iDRAC, Switches
+
+ 140.44.44.0/24 +
+ +
+
+ VLAN 20 Compute Fleet +
K3s Aerodrome, Spaceport, Swarm
+
+ 140.44.4.0/25 +
+ +
+
+ VLAN 3 Spacedock Lab +
Harvester HCI Virtualization
+
+ 140.44.1.0/24 +
+ +
+
+ VLAN 40 Storage SAN +
SnarfNAS, SN-NAS TrueNAS
+
+ 14.10.10.0/25 +
+ +
+
+ VLAN 50 Smart Home +
Home Assistant, NUT, Tempest
+
+ 14.4.4.0/25 +
+ +
+
+ VLAN 60 HA Interconnect +
Keepalived VIPs & Sync Heartbeats
+
+ 140.4.4.0/25 +
+ +
+
+ VLAN 30 Media Clients +
Smart TVs, Apple TVs, Streaming
+
+ 192.168.1.0/25 +
+
+
+ + +
+

Topology Key

+
+
+ Core Backbone +
+
+ K3s Compute +
+
+ Harvester HCI +
+
+ Swarm Fleet +
+
+ Cloudflare / Tunnel +
+
+ High Availability +
+
+
+
+ + +
+ + +
+
+ +
+
Public Internet
+
Primary WAN: 174.106.252.96
+
+
+ + +
+ +
+
+ WAN 1: Fiber ONT +
+
1 Gbps Symmetric
+
10G SFP+ to UDM-SE
+
+ + +
+
+ WAN 2: 5G Backup +
+
Cellular Standby
+
GbE RJ45 Failover
+
+
+ + +
+ +
+
+ + +
+
+ +
+
+
+ 🛡️ +
+
+

+ UniFi Dream Machine Special Edition (UDM-SE) +

+
Gateway VIP: 14.44.4.1 • UniFi OS v4.1 • WireGuard Server
+
+
+
+ + 10G SFP+ Backbone + +
+
+ + +
+
+
+ ⚡ Routing & QoS +
+
    +
  • • Inter-VLAN Wire-Speed L3
  • +
  • • Smart Queues & FQ-CoDel
  • +
  • • VLAN 20 Compute Priority
  • +
+
+ +
+
+ 🔒 Edge Security +
+
    +
  • • Suricata IPS/IDS Level 5
  • +
  • • Deep Packet Inspection
  • +
  • • Strict LAN/IoT Isolation
  • +
+
+ +
+
+ 🎯 Core Port Forwards +
+
    +
  • • 80/443 → 140.44.4.74 (NPM Host)
  • +
  • • 10051/52 → 140.44.4.51 (Zabbix)
  • +
  • • 8443/27015 → 14.4.4.15 (PC)
  • +
+
+
+
+ + +
+
+
🔀
+
+
USW Enterprise Core Switches
+
2.5G PoE+ to APs, 10G SFP+ to Racks
+
+
+ +
+
📶
+
+
UniFi Wi-Fi 6 Enterprise Fleet
+
SSIDs: SnarfNet-IoT, SnarfNet-5G, Trusted
+
+
+
+ +
+ + +
+ + +
+

+ + Edge Access & Ingress +

+ +
+
+
+ Cloudflare Zero Trust + cloudflared +
+

Outbound-only secure tunnel routes *.snarfnet.net without opening inbound WAN router ports.

+
+ +
+
+ Tailscale Mesh VPN + 100.64.0.0/10 +
+

Subnet routers provide direct wireguard access to all 8 VLANs from mobile devices.

+
+ +
+
+ Authentik Conditional MFA + SSO & Auth +
+

Client IP evaluation: Internal VLANs bypass MFA for frictionless use; external WAN enforces Passkeys/MFA.

+
+
+
+ + +
+
+

+ 🧠 AI Inference Pipeline +

+ Tesla T4s + Hailo +
+ +
+
+ 1. Frontend + Open WebUI (:8080) +
+
↓
+
+ 2. Reasoning + vLLM DeepSeek-R1 +
+
↓
+
+ 3. Execution + Triton + Qdrant +
+
+
+ +
+ +
+ + +
+
+
+

+ Compute Clusters & Workload Fleet +

+

All compute clusters operate on VLAN 20 (140.44.4.0/25) with dedicated VIPs and Longhorn block fabrics.

- 15/15 Argo Apps Synced - Uptime Kuma: 35 Probes + 29 Physical Nodes + 400+ Containers
- -
+ +
- -
- - -
-
- - Hardware Nodes + +
+
+
+
+
+ +

Cluster A: Aerodrome

+ Rancher Local +
+
+ 10 Nodes • 3 Control Plane + 7 Workers • Debian 12 (arm64) • MetalLB: 140.44.4.28-39 +
+
+
+
28 CPU Cores
+
770 Max Pod Capacity
+
-
-
-
- Aerodrome (k3s Master) - 140.44.4.11 -
-
AMD EPYC • 64 GiB DDR4 • Dual 10GbE SFP+
-
NVIDIA RTX 3060 12GB (Tensor AI Core)
-
-
-
- Spaceport (k3s Worker) - 140.44.4.41 -
-
Intel Core i7 • 32 GiB DDR4 • Dual 2.5GbE
+ +
+
+ ctl01-03: 140.44.4.11-13
- -
-
- Spacedock (Harvester HCI) - 140.44.1.0/24 -
-
Multi-Node Hypervisors • 128 GiB RAM • Ceph
+
+ wrk01-04: 140.44.4.21-24
+
+ wrk05-07: 140.44.4.25-27 +
+
+ Longhorn: Flash NVMe +
+
-
-
- Shipyard (Docker Swarm) - 140.44.4.15 + +
+
Active Workloads & VIPs
+
+ +
+
+
+ 🐄 Rancher Management +
+
aerodrome.snarfnet.net
+
+ :443
-
Container Fleet • 32 GiB RAM • GbE
+ +
+
+
+ 🔑 Authentik SSO Portal +
+
authentik.snarfnet.net
+
+ VIP .29 +
+ +
+
+
+ 📊 Grafana Telemetry +
+
grafana.snarfnet.net
+
+ VIP .32 +
+ +
+
+
+ 📈 Prometheus TSDB +
+
39 Targets (cAdvisor + Nodes)
+
+ :9090 +
+ +
+
+
+ 🟢 Uptime Kuma Probes +
+
uptime.snarfnet.net (35 Mon.)
+
+ VIP .28 +
+ +
+
+
+ 🦅 AWX Ansible Controller +
+
awx.snarfnet.net
+
+ VIP .30 +
+
- - -
-
- - Storage Array Pools -
- -
- -
-
- TrueNAS ZFS Flash Pool - 4.2 / 8.0 TiB (52%) -
-
-
-
-
NVMe Mirror • High-IOPS Databases & Volumes
-
- - -
-
- SnarfNAS Bulk Storage - 28.5 / 64.0 TiB (44%) -
-
-
-
-
HDD RAID-Z2 • Media Library & Long-term Backups
-
- - -
-
- Longhorn Distributed SAN - 850 GiB / 2.0 TiB (42%) -
-
-
-
-
Cross-Cluster Block Replicated Volumes
-
-
-
- - -
-
Runs on Host Fabric
-
-
• K3s v1.28 Engine
-
• Harvester v1.3 HCI
-
• Docker Swarm v24
-
• Keepalived VRRP
-
-
-
- -
- -
- - -
-
Edge & Ingress
- -
+ +
+
+
+
- -
-
NGINX Proxy Mgr
-
Reverse Proxy & SSL
-
+ +

Cluster B: Spaceport

+ GitOps Core +
+
+ 10 Nodes • 4 Control Plane + 6 Workers • RPi 5 (arm64) • MetalLB: 140.44.4.57-65
- :81
- -
-
- -
-
Authentik SSO
-
Identity & MFA RBAC
-
-
- :9000 -
- -
-
- -
-
Cloudflare Tunnel
-
Zero-Trust Ingress
-
-
- HTTPS -
- -
-
- -
-
Keepalived VIP
-
HA VRRP Virtual IP
-
-
- .10 +
+
40 CPU Cores
+
52.7 GiB RAM Total
- -
-
GitOps & Automation
- -
-
- -
-
Argo CD
-
Declarative GitOps
-
-
- Synced + +
+
+ ctl01-04: 140.44.4.41-44
- -
-
- -
-
Gitea Repos
-
Git Repositories
-
-
- :3000 +
+ wrk01-03: 140.44.4.51-53
- -
-
- -
-
AWX Automation
-
Ansible Playbooks
-
-
- :8052 +
+ wrk04-06: 140.44.4.54-56
- -
-
- -
-
Semaphore UI
-
Playbook Runner
-
-
- :3001 +
+ Storage: 1.14 TiB Block
- +
-
Observability
- -
-
- +
Active Workloads & VIPs
+
+ +
-
Grafana Analytics
-
Metrics & Dashboards
+
+ 🐙 ArgoCD Reconciler +
+
argo.snarfnet.net (15 Apps)
+ Node .51
- :3000 -
-
-
- +
-
Uptime Kuma
-
Status & Multi-Heartbeat
+
+ 🍵 Gitea Repositories +
+
git.snarfnet.net
+ VIP .62
- 35 Probes -
-
-
- +
-
Zabbix Fleet
-
Hardware & SNMP Alarms
+
+ 🏠 Homepage Dashboard +
+
homepage.snarfnet.net
+ VIP .60
- :8080 -
-
-
- +
-
PeaNUT UPS
-
Battery & Power Telemetry
+
+ 🛡️ Zabbix Core Server +
+
zabbix.snarfnet.net (28 Hosts)
+ VIP .61
- :3493 + +
+
+
+ ⚙️ Semaphore Ansible +
+
semaphore.snarfnet.net
+
+ VIP .58 +
+ +
+
+
+ 👨‍💻 Coder Dev Cloud +
+
coder.snarfnet.net
+
+ VIP .63 +
+
- - -
-
Storage & Platform
- -
-
- -
-
Longhorn Aerodrome
-
Distributed NVMe Flash
-
-
- SAN -
- -
-
- -
-
Longhorn Spaceport
-
Distributed Storage
-
-
- SAN -
- -
-
- -
-
Rancher Manager
-
K8s Multi-Cluster Ops
-
-
- HTTPS -
- -
-
- -
-
Homepage Dashboard
-
Unified Operations Hub
-
-
- :3000 -
-
- - -
-
AI & Inference Fabric
- -
-
- -
-
Ollama Engine
-
Local LLM Acceleration
-
-
- :11434 -
- -
-
- -
-
Open-WebUI
-
Multi-Model AI Chat
-
-
- :8080 -
- -
-
- -
-
NVIDIA DCGM
-
GPU VRAM & Telemetry
-
-
- :9400 -
- -
-
- -
-
Glances Monitor
-
Live Host Metric Stream
-
-
- :61208 -
-
- - -
-
Smart Home & Environment
- -
-
- -
-
Home Assistant
-
Home Automation Core
-
-
- :8123 -
- -
-
- -
-
Tempest Weather
-
Backyard Sensor Array
-
-
- UDP/API -
- -
-
- -
-
UniFi Controller
-
WLAN & Switching SDN
-
-
- :8443 -
- -
-
- -
-
Portainer Swarm
-
Container Management
-
-
- :9000 -
-
-
-
+ +
+
+
+
+
+ +

Cluster C: Spacedock (Harvester HCI)

+ GPU Inference +
+
+ 4 Physical Nodes • AMD64 Server Architecture • Dedicated Tesla T4 AI Compute +
+
+
+
224 CPU Cores
+
4x Tesla T4 GPUs
+
+
+ + +
+
+ Node 01: 140.44.4.81 +
+
+ Node 02: 140.44.4.82 +
+
+ Node 03: 140.44.4.83 +
+
+ Node 04: 140.44.4.84 +
+
+ + +
+
Active Workloads & Inference Services
+
+ +
+
+
+ 🌾 Harvester HCI Console +
+
harvester.spacedock.snarfnet.net
+
+ VLAN 3 +
+ +
+
+
+ 🧠 vLLM DeepSeek-R1-14B +
+
Reasoning Planner (.81)
+
+ :30805 +
+ +
+
+
+ 💎 vLLM Gemma-4-12B +
+
Coding Specialist (.82)
+
+ :30804 +
+ +
+
+
+ 🤖 vLLM Qwen 2.5-7B +
+
Instruction Following (.84)
+
+ :30800 +
+ +
+
+
+ 🔱 Triton Inference Server +
+
Decision Engine (.83)
+
+ :30803 +
+ +
+
+
+ ⚡ Qdrant Vector Engine +
+
Semantic Vector Store (.82)
+
+ :30633 +
+ +
+
+
+
+ + +
+
+
+
+
+ +

Cluster D: Shipyard (Docker Swarm)

+ Swarm VIP: 140.44.4.70 +
+
+ 5 RPi 5 Nodes • Keepalived VRRP VIP • Host-Mode Ingress • Portainer EE 2.45 +
+
+
+
5 Manager Nodes
+
Keepalived Active/Passive
+
+
+ + +
+
+ Swm01: .71 +
+
+ Swm02: .72 +
+
+ Swm03: .73 +
+
+ NPM: .74 +
+
+ Swm05: .75 +
+
+ + +
+
Active Workloads & Stacks
+
+ +
+
+
+ 🚢 Portainer Enterprise +
+
shipyard.snarfnet.net
+
+ :9443 +
+ +
+
+
+ 🌐 NGINX Proxy Manager +
+
Host Mode Ingress (.74)
+
+ :80/:443 +
+ +
+
+
+ 🔔 Gotify Notifications +
+
gotify.snarfnet.net
+
+ :8090 +
+ +
+
+
+ 🧭 Heimdall Landing +
+
portal.snarfnet.net
+
+ :85 +
+ +
+
+
+ 💻 MeshCentral Remote +
+
mesh.snarfnet.net
+
+ :8086 +
+ +
+
+
+ 📦 Homebox Inventory +
+
homebox.snarfnet.net
+
+ :3100 +
+ +
+
+
+
+ +
+
+ + +
+ + +
+
+
+ 💾 +

Storage SAN & Distributed Volumes (VLAN 40)

+
+ 14.10.10.0/25 +
+ +
+
+
+ SnarfNAS (Synology DSM) + 14.10.10.72 +
+

Primary NFS/SMB target, multi-drive RAID, media libraries, Time Machine, off-site replication.

+
+ +
+
+ SN-NAS (TrueNAS SCALE) + 14.10.10.71 +
+

ZFS high-speed pool, iSCSI volumes for hypervisors, automated snapshot mirrors.

+
+ +
+
+ Longhorn Spaceport Storage + 1.14 TiB Total +
+

Replicated block storage across Spaceport workers (262 GiB used, 908 GiB free).

+
+ +
+
+ Longhorn Aerodrome Storage + Flash NVMe +
+

Low-latency distributed storage backing Prometheus, Grafana, and MariaDB databases.

+
+
+
+ + +
+
+
+ 🏡 +

Smart Home & Edge AI Fabric (VLAN 50)

+
+ 14.4.4.0/25 +
+ +
+
+
+ Home Assistant Yellow + 14.4.4.20:8123 +
+

Primary smart home automation hub, Zigbee/Z-Wave networks, Matter, local voice processing.

+
+ +
+
+ Tempest Weather Station + Direct HA Telemetry +
+

Local UDP & cloud backyard weather station (Temp, Feels Like, Humidity, Wind Speed rounded to 0.1°).

+
+ +
+
+ NUT Master Daemon + 14.4.4.44:3493 +
+

Network UPS Tools master daemon monitoring central battery units, broadcast shutdown signals.

+
+ +
+
+ Hailo-8 Edge NPU Node + 140.44.4.7:8000 +
+

Dedicated Raspberry Pi 5 with 26 TOPS Hailo-8 AI acceleration running lightweight local LLMs.

+
+
-
+
+ + + + - \ No newline at end of file + diff --git a/docs/snarfnet_homelab_topology.png b/docs/snarfnet_homelab_topology.png index 171af06..326d937 100644 Binary files a/docs/snarfnet_homelab_topology.png and b/docs/snarfnet_homelab_topology.png differ