ci: speed up E2E pipeline by running on arm64 runners and skipping redundant cluster rebuild (#278)

Author: drew

.github/workflows/docker-build.yml

@@ -17,6 +17,16 @@ on:
         required: false
         type: boolean
         default: true
+      platform:
+        description: "Target platform(s) for Docker build"
+        required: false
+        type: string
+        default: "linux/amd64,linux/arm64"
+      runner:
+        description: "GitHub Actions runner label"
+        required: false
+        type: string
+        default: "build-amd64"
 
 env:
   MISE_GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
@@ -29,7 +39,7 @@ permissions:
 jobs:
   build:
     name: Build ${{ inputs.component }}
-    runs-on: build-amd64
+    runs-on: ${{ inputs.runner }}
     timeout-minutes: ${{ inputs.timeout-minutes }}
     container:
       image: ghcr.io/nvidia/openshell/ci:latest
@@ -43,7 +53,7 @@ jobs:
       IMAGE_TAG: ${{ github.sha }}
       IMAGE_REGISTRY: ghcr.io/nvidia/openshell
       DOCKER_PUSH: ${{ inputs.push && '1' || '0' }}
-      DOCKER_PLATFORM: linux/amd64,linux/arm64
+      DOCKER_PLATFORM: ${{ inputs.platform }}
     steps:
       - uses: actions/checkout@v4
         with:

.github/workflows/e2e-test.yml

@@ -7,6 +7,11 @@ on:
         description: "Image tag to test (typically the commit SHA)"
         required: true
         type: string
+      runner:
+        description: "GitHub Actions runner label"
+        required: false
+        type: string
+        default: "build-amd64"
 
 permissions:
   contents: read
@@ -15,7 +20,7 @@ permissions:
 jobs:
   e2e:
     name: E2E
-    runs-on: build-amd64
+    runs-on: ${{ inputs.runner }}
     timeout-minutes: 30
     container:
       image: ghcr.io/nvidia/openshell/ci:latest
@@ -50,6 +55,7 @@ jobs:
           GATEWAY_HOST: host.docker.internal
           GATEWAY_PORT: "8080"
           SKIP_IMAGE_PUSH: "1"
+          SKIP_CLUSTER_IMAGE_BUILD: "1"
           OPENSHELL_CLUSTER_IMAGE: ghcr.io/nvidia/openshell/cluster:${{ inputs.image-tag }}
         run: mise run --no-prepare --skip-deps cluster
 

.github/workflows/e2e.yml

@@ -14,15 +14,20 @@ jobs:
     uses: ./.github/workflows/docker-build.yml
     with:
       component: gateway
+      platform: linux/arm64
+      runner: build-arm64
 
   build-cluster:
     if: contains(github.event.pull_request.labels.*.name, 'e2e')
     uses: ./.github/workflows/docker-build.yml
     with:
       component: cluster
+      platform: linux/arm64
+      runner: build-arm64
 
   e2e:
     needs: [build-gateway, build-cluster]
     uses: ./.github/workflows/e2e-test.yml
     with:
       image-tag: ${{ github.sha }}
+      runner: build-arm64