revert: tailscale docker sidecar

2025-11-28 09:31:21 +01:00
parent 99e39ee6e6
commit 19ad30e8c4
2 changed files with 129 additions and 92 deletions
--- a/ai/compose.yaml
+++ b/ai/compose.yaml
@@ -96,7 +96,6 @@ services:
    image: ghcr.io/berriai/litellm:main-latest
    container_name: ${AI_COMPOSE_PROJECT_NAME}_litellm
    restart: unless-stopped
-    network_mode: "service:tailscale"
    environment:
      TZ: ${TIMEZONE:-Europe/Berlin}
      ANTHROPIC_API_KEY: ${ANTHROPIC_API_KEY}
@@ -122,9 +121,21 @@ services:
      ]
    depends_on:
      - ai_postgres
-      - tailscale
    healthcheck:
      disable: true
+    networks:
+      - compose_network
+    labels:
+      - "traefik.enable=true"
+      - "traefik.http.routers.${AI_COMPOSE_PROJECT_NAME}-litellm-web.rule=Host(`${AI_LITELLM_TRAEFIK_HOST}`)"
+      - "traefik.http.routers.${AI_COMPOSE_PROJECT_NAME}-litellm-web.entrypoints=web"
+      - "traefik.http.routers.${AI_COMPOSE_PROJECT_NAME}-litellm-web-secure.rule=Host(`${AI_LITELLM_TRAEFIK_HOST}`)"
+      - "traefik.http.routers.${AI_COMPOSE_PROJECT_NAME}-litellm-web-secure.tls.certresolver=resolver"
+      - "traefik.http.routers.${AI_COMPOSE_PROJECT_NAME}-litellm-web-secure.entrypoints=web-secure"
+      - "traefik.http.routers.${AI_COMPOSE_PROJECT_NAME}-litellm-web-secure.middlewares=security-headers@file"
+      - "traefik.http.services.${AI_COMPOSE_PROJECT_NAME}-litellm.loadbalancer.server.port=4000"
+      - "traefik.docker.network=${NETWORK_NAME}"
+      - "com.centurylinklabs.watchtower.enable=${WATCHTOWER_LABEL_ENABLE}"

  # Facefusion - AI face swapping and enhancement
  facefusion:
@@ -168,130 +179,99 @@ services:
    image: nginx:alpine
    container_name: ${AI_COMPOSE_PROJECT_NAME}_comfyui
    restart: unless-stopped
-    network_mode: "service:tailscale"
    environment:
      TZ: ${TIMEZONE:-Europe/Berlin}
-      GPU_SERVICE_HOST: ${GPU_TAILSCALE_HOST}
+      GPU_SERVICE_HOST: ${GPU_TAILSCALE_HOST:-runpod-ai-orchestrator}
      GPU_SERVICE_PORT: ${COMFYUI_BACKEND_PORT:-8188}
    volumes:
      - ./nginx.conf.template:/etc/nginx/nginx.conf.template:ro
    command: /bin/sh -c "envsubst '$${GPU_SERVICE_HOST},$${GPU_SERVICE_PORT}' < /etc/nginx/nginx.conf.template > /etc/nginx/nginx.conf && exec nginx -g 'daemon off;'"
-    depends_on:
-      - tailscale
-
-  audiocraft:
-    image: nginx:alpine
-    container_name: ${AI_COMPOSE_PROJECT_NAME}_audiocraft
-    restart: unless-stopped
-    network_mode: "service:tailscale"
-    environment:
-      TZ: ${TIMEZONE:-Europe/Berlin}
-      GPU_SERVICE_HOST: ${GPU_TAILSCALE_HOST}
-      GPU_SERVICE_PORT: ${AUDIOCRAFT_BACKEND_PORT:-7860}
-    volumes:
-      - ./nginx.conf.template:/etc/nginx/nginx.conf.template:ro
-    command: /bin/sh -c "envsubst '$${GPU_SERVICE_HOST},$${GPU_SERVICE_PORT}' < /etc/nginx/nginx.conf.template > /etc/nginx/nginx.conf && exec nginx -g 'daemon off;'"
-    depends_on:
-      - tailscale
-
-  upscale:
-    image: nginx:alpine
-    container_name: ${AI_COMPOSE_PROJECT_NAME}_upscale
-    restart: unless-stopped
-    network_mode: "service:tailscale"
-    environment:
-      TZ: ${TIMEZONE:-Europe/Berlin}
-      GPU_SERVICE_HOST: ${GPU_TAILSCALE_HOST}
-      GPU_SERVICE_PORT: ${UPSCALE_BACKEND_PORT:-8080}
-    volumes:
-      - ./nginx.conf.template:/etc/nginx/nginx.conf.template:ro
-    command: /bin/sh -c "envsubst '$${GPU_SERVICE_HOST},$${GPU_SERVICE_PORT}' < /etc/nginx/nginx.conf.template > /etc/nginx/nginx.conf && exec nginx -g 'daemon off;'"
-    depends_on:
-      - tailscale
-
-  # Supervisor UI - Modern web interface for RunPod process management
-  supervisor:
-    image: dev.pivoine.art/valknar/supervisor-ui:latest
-    container_name: ${AI_COMPOSE_PROJECT_NAME}_supervisor_ui
-    restart: unless-stopped
-    network_mode: "service:tailscale"
-    environment:
-      TZ: ${TIMEZONE:-Europe/Berlin}
-      NODE_ENV: production
-      # Connect to RunPod Supervisor via Tailscale
-      SUPERVISOR_HOST: ${GPU_TAILSCALE_HOST}
-      SUPERVISOR_PORT: ${SUPERVISOR_BACKEND_PORT:-9001}
-      # No auth needed - Supervisor has auth disabled (protected by Authelia)
-    depends_on:
-      - tailscale
-
-  # Tailscale VPN - Provides network access to RunPod GPU services
-  tailscale:
-    image: tailscale/tailscale:latest
-    hostname: vps
-    container_name: ${AI_COMPOSE_PROJECT_NAME}_tailscale
-    cap_add:
-      - NET_ADMIN
-      - SYS_MODULE
-    volumes:
-      - tailscale_state:/var/lib/tailscale
-      - /dev/net/tun:/dev/net/tun
-    environment:
-      - TS_AUTHKEY=${TAILSCALE_AUTHKEY}
-      - TS_STATE_DIR=/var/lib/tailscale
-      - TS_USERSPACE=true
-      - TS_ACCEPT_DNS=true
-    restart: unless-stopped
    networks:
      - compose_network
    labels:
      - "traefik.enable=true"
-      - "traefik.docker.network=${NETWORK_NAME}"
-      # === ComfyUI (port 80 - nginx proxy) ===
      - "traefik.http.routers.${AI_COMPOSE_PROJECT_NAME}-comfyui-web.rule=Host(`${AI_COMFYUI_TRAEFIK_HOST:-comfy.ai.pivoine.art}`)"
      - "traefik.http.routers.${AI_COMPOSE_PROJECT_NAME}-comfyui-web.entrypoints=web"
      - "traefik.http.routers.${AI_COMPOSE_PROJECT_NAME}-comfyui-web-secure.rule=Host(`${AI_COMFYUI_TRAEFIK_HOST:-comfy.ai.pivoine.art}`)"
      - "traefik.http.routers.${AI_COMPOSE_PROJECT_NAME}-comfyui-web-secure.tls.certresolver=resolver"
      - "traefik.http.routers.${AI_COMPOSE_PROJECT_NAME}-comfyui-web-secure.entrypoints=web-secure"
      - "traefik.http.routers.${AI_COMPOSE_PROJECT_NAME}-comfyui-web-secure.middlewares=net-authelia,security-headers@file"
-      - "traefik.http.routers.${AI_COMPOSE_PROJECT_NAME}-comfyui-web-secure.service=${AI_COMPOSE_PROJECT_NAME}-comfyui"
      - "traefik.http.services.${AI_COMPOSE_PROJECT_NAME}-comfyui.loadbalancer.server.port=80"
-      # === AudioCraft (port 7860 via nginx on 80) ===
+      - "traefik.docker.network=${NETWORK_NAME}"
+
+  audiocraft:
+    image: nginx:alpine
+    container_name: ${AI_COMPOSE_PROJECT_NAME}_audiocraft
+    restart: unless-stopped
+    environment:
+      TZ: ${TIMEZONE:-Europe/Berlin}
+      GPU_SERVICE_HOST: ${GPU_TAILSCALE_HOST:-runpod-ai-orchestrator}
+      GPU_SERVICE_PORT: ${AUDIOCRAFT_BACKEND_PORT:-7860}
+    volumes:
+      - ./nginx.conf.template:/etc/nginx/nginx.conf.template:ro
+    command: /bin/sh -c "envsubst '$${GPU_SERVICE_HOST},$${GPU_SERVICE_PORT}' < /etc/nginx/nginx.conf.template > /etc/nginx/nginx.conf && exec nginx -g 'daemon off;'"
+    networks:
+      - compose_network
+    labels:
+      - "traefik.enable=true"
      - "traefik.http.routers.${AI_COMPOSE_PROJECT_NAME}-audiocraft-web.rule=Host(`${AI_AUDIOCRAFT_TRAEFIK_HOST:-audiocraft.ai.pivoine.art}`)"
      - "traefik.http.routers.${AI_COMPOSE_PROJECT_NAME}-audiocraft-web.entrypoints=web"
      - "traefik.http.routers.${AI_COMPOSE_PROJECT_NAME}-audiocraft-web-secure.rule=Host(`${AI_AUDIOCRAFT_TRAEFIK_HOST:-audiocraft.ai.pivoine.art}`)"
      - "traefik.http.routers.${AI_COMPOSE_PROJECT_NAME}-audiocraft-web-secure.tls.certresolver=resolver"
      - "traefik.http.routers.${AI_COMPOSE_PROJECT_NAME}-audiocraft-web-secure.entrypoints=web-secure"
      - "traefik.http.routers.${AI_COMPOSE_PROJECT_NAME}-audiocraft-web-secure.middlewares=net-authelia,security-headers@file"
-      - "traefik.http.routers.${AI_COMPOSE_PROJECT_NAME}-audiocraft-web-secure.service=${AI_COMPOSE_PROJECT_NAME}-audiocraft"
-      - "traefik.http.services.${AI_COMPOSE_PROJECT_NAME}-audiocraft.loadbalancer.server.port=7860"
-      # === Upscale (port 8080 via nginx on 80) ===
+      - "traefik.http.services.${AI_COMPOSE_PROJECT_NAME}-audiocraft.loadbalancer.server.port=80"
+      - "traefik.docker.network=${NETWORK_NAME}"
+
+  upscale:
+    image: nginx:alpine
+    container_name: ${AI_COMPOSE_PROJECT_NAME}_upscale
+    restart: unless-stopped
+    environment:
+      TZ: ${TIMEZONE:-Europe/Berlin}
+      GPU_SERVICE_HOST: ${GPU_TAILSCALE_HOST:-runpod-ai-orchestrator}
+      GPU_SERVICE_PORT: ${UPSCALE_BACKEND_PORT:-8080}
+    volumes:
+      - ./nginx.conf.template:/etc/nginx/nginx.conf.template:ro
+    command: /bin/sh -c "envsubst '$${GPU_SERVICE_HOST},$${GPU_SERVICE_PORT}' < /etc/nginx/nginx.conf.template > /etc/nginx/nginx.conf && exec nginx -g 'daemon off;'"
+    networks:
+      - compose_network
+    labels:
+      - "traefik.enable=true"
      - "traefik.http.routers.${AI_COMPOSE_PROJECT_NAME}-upscale-web.rule=Host(`${AI_UPSCALE_TRAEFIK_HOST:-upscale.ai.pivoine.art}`)"
      - "traefik.http.routers.${AI_COMPOSE_PROJECT_NAME}-upscale-web.entrypoints=web"
      - "traefik.http.routers.${AI_COMPOSE_PROJECT_NAME}-upscale-web-secure.rule=Host(`${AI_UPSCALE_TRAEFIK_HOST:-upscale.ai.pivoine.art}`)"
      - "traefik.http.routers.${AI_COMPOSE_PROJECT_NAME}-upscale-web-secure.tls.certresolver=resolver"
      - "traefik.http.routers.${AI_COMPOSE_PROJECT_NAME}-upscale-web-secure.entrypoints=web-secure"
      - "traefik.http.routers.${AI_COMPOSE_PROJECT_NAME}-upscale-web-secure.middlewares=net-authelia,security-headers@file"
-      - "traefik.http.routers.${AI_COMPOSE_PROJECT_NAME}-upscale-web-secure.service=${AI_COMPOSE_PROJECT_NAME}-upscale"
-      - "traefik.http.services.${AI_COMPOSE_PROJECT_NAME}-upscale.loadbalancer.server.port=8080"
-      # === Supervisor UI (port 3000) ===
+      - "traefik.http.services.${AI_COMPOSE_PROJECT_NAME}-upscale.loadbalancer.server.port=80"
+      - "traefik.docker.network=${NETWORK_NAME}"
+
+  # Supervisor UI - Modern web interface for RunPod process management
+  supervisor:
+    image: dev.pivoine.art/valknar/supervisor-ui:latest
+    container_name: ${AI_COMPOSE_PROJECT_NAME}_supervisor_ui
+    restart: unless-stopped
+    environment:
+      TZ: ${TIMEZONE:-Europe/Berlin}
+      NODE_ENV: production
+      # Connect to RunPod Supervisor via Tailscale (host Tailscale provides DNS)
+      SUPERVISOR_HOST: ${GPU_TAILSCALE_HOST:-runpod-ai-orchestrator}
+      SUPERVISOR_PORT: ${SUPERVISOR_BACKEND_PORT:-9001}
+      # No auth needed - Supervisor has auth disabled (protected by Authelia)
+    networks:
+      - compose_network
+    labels:
+      - "traefik.enable=true"
      - "traefik.http.routers.${AI_COMPOSE_PROJECT_NAME}-supervisor-web.rule=Host(`${AI_SUPERVISOR_TRAEFIK_HOST:-supervisor.ai.pivoine.art}`)"
      - "traefik.http.routers.${AI_COMPOSE_PROJECT_NAME}-supervisor-web.entrypoints=web"
      - "traefik.http.routers.${AI_COMPOSE_PROJECT_NAME}-supervisor-web-secure.rule=Host(`${AI_SUPERVISOR_TRAEFIK_HOST:-supervisor.ai.pivoine.art}`)"
      - "traefik.http.routers.${AI_COMPOSE_PROJECT_NAME}-supervisor-web-secure.tls.certresolver=resolver"
      - "traefik.http.routers.${AI_COMPOSE_PROJECT_NAME}-supervisor-web-secure.entrypoints=web-secure"
      - "traefik.http.routers.${AI_COMPOSE_PROJECT_NAME}-supervisor-web-secure.middlewares=net-authelia,security-headers@file"
-      - "traefik.http.routers.${AI_COMPOSE_PROJECT_NAME}-supervisor-web-secure.service=${AI_COMPOSE_PROJECT_NAME}-supervisor"
      - "traefik.http.services.${AI_COMPOSE_PROJECT_NAME}-supervisor.loadbalancer.server.port=3000"
-      # === LiteLLM (port 4000) ===
-      - "traefik.http.routers.${AI_COMPOSE_PROJECT_NAME}-litellm-web.rule=Host(`${AI_LITELLM_TRAEFIK_HOST}`)"
-      - "traefik.http.routers.${AI_COMPOSE_PROJECT_NAME}-litellm-web.entrypoints=web"
-      - "traefik.http.routers.${AI_COMPOSE_PROJECT_NAME}-litellm-web-secure.rule=Host(`${AI_LITELLM_TRAEFIK_HOST}`)"
-      - "traefik.http.routers.${AI_COMPOSE_PROJECT_NAME}-litellm-web-secure.tls.certresolver=resolver"
-      - "traefik.http.routers.${AI_COMPOSE_PROJECT_NAME}-litellm-web-secure.entrypoints=web-secure"
-      - "traefik.http.routers.${AI_COMPOSE_PROJECT_NAME}-litellm-web-secure.middlewares=security-headers@file"
-      - "traefik.http.routers.${AI_COMPOSE_PROJECT_NAME}-litellm-web-secure.service=${AI_COMPOSE_PROJECT_NAME}-litellm"
-      - "traefik.http.services.${AI_COMPOSE_PROJECT_NAME}-litellm.loadbalancer.server.port=4000"
+      - "traefik.docker.network=${NETWORK_NAME}"
+      - "com.centurylinklabs.watchtower.enable=${WATCHTOWER_LABEL_ENABLE}"

 volumes:
  ai_postgres_data:
@@ -300,5 +280,8 @@ volumes:
    name: ${AI_COMPOSE_PROJECT_NAME}_webui_data
  ai_facefusion_data:
    name: ${AI_COMPOSE_PROJECT_NAME}_facefusion_data
-  tailscale_state:
-    name: ${AI_COMPOSE_PROJECT_NAME}_tailscale_state
+
+networks:
+  compose_network:
+    name: ${NETWORK_NAME}
+    external: true
--- a/arty.yml
+++ b/arty.yml
@@ -262,3 +262,57 @@ scripts:
    docker restart sexy_api &&
    echo "✓ Directus API restarted"
  net/create: docker network create "$NETWORK_NAME"
+  # Setup iptables NAT for Docker containers to reach Tailscale network
+  # Requires Tailscale installed on host: curl -fsSL https://tailscale.com/install.sh | sh
+  tailscale/setup: |
+    echo "Setting up iptables for Docker-to-Tailscale routing..."
+
+    # Enable IP forwarding
+    sudo sysctl -w net.ipv4.ip_forward=1
+    grep -q "net.ipv4.ip_forward=1" /etc/sysctl.conf || echo "net.ipv4.ip_forward=1" | sudo tee -a /etc/sysctl.conf
+
+    # Get Docker network CIDR
+    DOCKER_CIDR=$(docker network inspect ${NETWORK_NAME} --format '{{range .IPAM.Config}}{{.Subnet}}{{end}}' 2>/dev/null || echo "172.18.0.0/16")
+    echo "Docker network CIDR: $DOCKER_CIDR"
+
+    # Add NAT rule (check if already exists)
+    if ! sudo iptables -t nat -C POSTROUTING -s "$DOCKER_CIDR" -o tailscale0 -j MASQUERADE 2>/dev/null; then
+      sudo iptables -t nat -A POSTROUTING -s "$DOCKER_CIDR" -o tailscale0 -j MASQUERADE
+      echo "✓ iptables NAT rule added"
+    else
+      echo "✓ iptables NAT rule already exists"
+    fi
+
+    # Persist rules
+    sudo netfilter-persistent save 2>/dev/null || echo "Install iptables-persistent to persist rules: sudo apt install iptables-persistent"
+
+    echo "✓ Tailscale routing configured"
+
+  # Install and configure Tailscale on host with persistent state
+  tailscale/install: |
+    echo "Installing Tailscale..."
+
+    # Install Tailscale if not present
+    if ! command -v tailscale &> /dev/null; then
+      curl -fsSL https://tailscale.com/install.sh | sh
+    else
+      echo "✓ Tailscale already installed"
+    fi
+
+    # Create state directory for persistence
+    TAILSCALE_STATE="/var/lib/tailscale"
+    sudo mkdir -p "$TAILSCALE_STATE"
+
+    # Start and enable tailscaled service
+    sudo systemctl enable --now tailscaled
+
+    # Connect to Tailscale network
+    echo "Connecting to Tailscale..."
+    sudo tailscale up --authkey="$TAILSCALE_AUTHKEY" --hostname=vps
+
+    # Show status
+    echo ""
+    tailscale status
+    echo ""
+    echo "✓ Tailscale installed and connected"
+    echo "  Run 'arty tailscale/setup' to configure iptables routing for Docker"