Add GA to build A3 Mega VM image

Author: r4victor

.github/workflows/gcp-a3mega-image.yml

@@ -0,0 +1,39 @@
+name: Build GCP A3 Mega VM image
+
+on:
+  - workflow_dispatch
+  - push
+
+env:
+  PACKER_VERSION: "1.9.2"
+  IMAGE_VERSION: ${{ github.run_number }}
+jobs:
+  build-gcp-images:
+    defaults:
+      run:
+        working-directory: scripts/packer
+    runs-on: ubuntu-latest
+    permissions:
+      contents: 'read'
+      id-token: 'write'
+    steps:
+      - uses: actions/checkout@v4
+      - name: Authenticate to Google Cloud
+        uses: google-github-actions/auth@v2
+        with:
+          workload_identity_provider: 'projects/531508670106/locations/global/workloadIdentityPools/github-identity-pool/providers/github-id-provider'
+          service_account: '[email protected]'
+          create_credentials_file: true
+      - name: Set up Cloud SDK
+        uses: google-github-actions/setup-gcloud@v2
+      - name: Download packer
+        run: |
+          wget https://releases.hashicorp.com/packer/${{ env.PACKER_VERSION }}/packer_${{ env.PACKER_VERSION }}_linux_amd64.zip
+          unzip packer_${{ env.PACKER_VERSION }}_linux_amd64.zip
+          chmod +x packer
+      - name: Run packer
+        run: |
+          ./packer build -var image_version=${{ env.IMAGE_VERSION }} -var gcp-a3mega-image.json
+      - name: Publish image
+        run: |
+          gcloud compute images add-iam-policy-binding dstack-a3mega-${{ env.IMAGE_VERSION }} --member='allAuthenticatedUsers' --role='roles/compute.imageUser'

scripts/packer/gcp-a3mega-image.json

@@ -0,0 +1,28 @@
+{
+  "variables": {
+    "image_version": ""
+  },
+  "builders": [
+    {
+      "type": "googlecompute",
+      "project_id": "dstack",
+      "source_image": "dstack-a3mega-20250401t065024z",
+      "image_name": "dstack-a3mega-{{user `image_version`}}",
+      "instance_name": "dstack-a3mega-{{user `image_version`}}",
+      "image_description": "dstack VM image for A3 Mega instances with pre-pulled Docker images. The source image is based on https://cloud.google.com/cluster-toolkit/docs/deploy/deploy-a3-mega-cluster.",
+      "ssh_username": "ubuntu",
+      "zone": "us-central1-a",
+      "disk_size": 100
+    }
+  ],
+  "provisioners": [
+    {
+      "type": "shell",
+      "inline": [
+        "gcloud -q auth configure-docker us-docker.pkg.dev",
+        "docker pull us-docker.pkg.dev/gce-ai-infra/gpudirect-tcpxo/tcpgpudmarxd-dev:v1.0.14",
+        "docker pull us-docker.pkg.dev/gce-ai-infra/gpudirect-tcpxo/nccl-plugin-gpudirecttcpx-dev:v1.0.8-1"
+      ]
+    }
+  ]
+}