refactor progress

.github/workflows/restart-gateway.yaml

@@ -20,28 +20,22 @@ env:
   IMAGE_NAME: fluxer-gateway
   CONTEXT_DIR: fluxer_gateway
   COMPOSE_STACK: fluxer-gateway
+  RELEASE_CHANNEL: ${{ github.ref_name == 'canary' && 'staging' || 'production' }}
 
 jobs:
   restart:
     name: Restart gateway
-    runs-on: blacksmith-2vcpu-ubuntu-2404
-    timeout-minutes: 10
+    runs-on: blacksmith-8vcpu-ubuntu-2404
+    timeout-minutes: 25
     steps:
       - name: Validate confirmation
         if: ${{ github.event.inputs.confirmation != 'RESTART' }}
-        run: |
-          echo "::error::Confirmation failed. You must type 'RESTART' to proceed with a full restart."
-          echo "::error::For regular updates, use deploy-gateway.yaml instead."
-          exit 1
+        run: python3 scripts/ci/workflows/restart_gateway.py --step validate_confirmation --confirmation "${{ github.event.inputs.confirmation }}"
 
       - uses: actions/checkout@v6
 
       - name: Record deploy commit
-        run: |
-          set -euo pipefail
-          sha=$(git rev-parse HEAD)
-          echo "Deploying commit ${sha}"
-          printf 'DEPLOY_SHA=%s\n' "$sha" >> "$GITHUB_ENV"
+        run: python3 scripts/ci/workflows/restart_gateway.py --step record_deploy_commit
 
       - name: Set up Docker Buildx
         uses: docker/setup-buildx-action@v3
@@ -67,12 +61,7 @@ jobs:
           DOCKER_BUILD_RECORD_UPLOAD: false
 
       - name: Install docker-pussh
-        run: |
-          set -euo pipefail
-          mkdir -p ~/.docker/cli-plugins
-          curl -fsSL https://raw.githubusercontent.com/psviderski/unregistry/v0.3.1/docker-pussh \
-            -o ~/.docker/cli-plugins/docker-pussh
-          chmod +x ~/.docker/cli-plugins/docker-pussh
+        run: python3 scripts/ci/workflows/restart_gateway.py --step install_docker_pussh
 
       - name: Set up SSH agent
         uses: webfactory/ssh-agent@v0.9.1
@@ -80,70 +69,10 @@ jobs:
           ssh-private-key: ${{ secrets.SSH_PRIVATE_KEY_SERVER }}
 
       - name: Add server to known hosts
-        run: |
-          set -euo pipefail
-          mkdir -p ~/.ssh
-          ssh-keyscan -H ${{ secrets.SERVER_IP }} >> ~/.ssh/known_hosts
+        run: python3 scripts/ci/workflows/restart_gateway.py --step add_known_hosts --server-ip ${{ secrets.SERVER_IP }}
 
       - name: Push image and deploy
         env:
           IMAGE_TAG: ${{ env.IMAGE_NAME }}:${{ env.DEPLOY_SHA }}
           SERVER: ${{ secrets.SERVER_USER }}@${{ secrets.SERVER_IP }}
-        run: |
-          set -euo pipefail
-          docker pussh "${IMAGE_TAG}" "${SERVER}"
-
-          ssh "${SERVER}" "IMAGE_TAG=${IMAGE_TAG} SERVICE_NAME=${SERVICE_NAME} COMPOSE_STACK=${COMPOSE_STACK} bash" << 'EOF'
-          set -euo pipefail
-          sudo mkdir -p "/opt/${SERVICE_NAME}"
-          sudo chown -R "${USER}:${USER}" "/opt/${SERVICE_NAME}"
-          cd "/opt/${SERVICE_NAME}"
-
-          cat > compose.yaml << COMPOSEEOF
-          services:
-            app:
-              image: ${IMAGE_TAG}
-              hostname: "{{.Node.Hostname}}-{{.Task.Slot}}"
-              env_file:
-                - /etc/fluxer/fluxer.env
-              environment:
-                - API_HOST=fluxer-api_app:8080
-                - API_CANARY_HOST=fluxer-api-canary_app:8080
-                - RELEASE_NODE=fluxer_gateway@{{.Node.Hostname}}-{{.Task.Slot}}
-                - LOGGER_LEVEL=info
-                - VAPID_PUBLIC_KEY=BEIwQxIwfj6m90tLYAR0AU_GJWU4kw8J_zJcHQG55NCUWSyRy-dzMOgvxk8yEDwdVyJZa6xUL4fmwngijq8T2pY
-                - FLUXER_METRICS_HOST=fluxer-metrics_app:8080
-                - MEDIA_PROXY_ENDPOINT=https://fluxerusercontent.com
-              deploy:
-                replicas: 1
-                endpoint_mode: dnsrr
-                restart_policy:
-                  condition: on-failure
-                  delay: 5s
-                  max_attempts: 3
-                update_config:
-                  parallelism: 1
-                  delay: 10s
-                  order: start-first
-                rollback_config:
-                  parallelism: 1
-                  delay: 10s
-                labels:
-                  - 'caddy_gw=gateway.fluxer.app'
-                  - 'caddy_gw.reverse_proxy={{upstreams 8080}}'
-              networks:
-                - fluxer-shared
-              healthcheck:
-                test: ['CMD', 'curl', '-f', 'http://localhost:8080/_health']
-                interval: 30s
-                timeout: 10s
-                retries: 3
-                start_period: 60s
-
-          networks:
-            fluxer-shared:
-              external: true
-          COMPOSEEOF
-
-          docker stack deploy --with-registry-auth --detach=false --resolve-image never -c compose.yaml "${COMPOSE_STACK}"
-          EOF
+        run: python3 scripts/ci/workflows/restart_gateway.py --step push_and_deploy