ci: Prod Runners (#5534)

Replacing infra for Github runners

.github/workflows/build-frontend-image.yaml

@@ -58,8 +58,8 @@ jobs:
       fail-fast: false
       matrix:
         platform:
-          - { arch: amd64, runner: gpu-l40-amd64 }
-          - { arch: arm64, runner: cpu-arm-r8g-4xlarge }
+          - { arch: amd64, runner: prod-builder-amd-v1  }
+          - { arch: arm64, runner: prod-builder-arm-v1 }
     runs-on: ${{ matrix.platform.runner }}
     steps:
       - name: Checkout repository

.github/workflows/ci-test-suite.yml

@@ -75,7 +75,7 @@ env:
 jobs:
   build-amd64:
     name: Build ${{ matrix.framework }} (amd64)
-    runs-on: cpu-amd-m5-4xlarge
+    runs-on: prod-builder-amd-v1
     timeout-minutes: 120
     strategy:
       fail-fast: false
@@ -126,7 +126,7 @@ jobs:
 
   build-arm64:
     name: Build ${{ matrix.framework }} (arm64)
-    runs-on: cpu-arm-r8g-4xlarge
+    runs-on: prod-builder-arm-v1
     timeout-minutes: 120
     strategy:
       fail-fast: false
@@ -304,9 +304,9 @@ jobs:
         framework: [vllm, trtllm, sglang]
         arch:
           - arch: amd64
-            runner: gpu-l40-amd64
+            runner: prod-builder-amd-gpu-v1
           - arch: arm64
-            runner: cpu-arm-r8g-4xlarge
+            runner: prod-builder-arm-v1
     steps:
       - uses: actions/checkout@v4
       - name: Check if build succeeded
@@ -437,10 +437,10 @@ jobs:
         framework: [vllm, trtllm, sglang]
         arch:
           - arch: amd64
-            runner: gpu-l40-amd64
+            runner: prod-builder-amd-gpu-v1
             timeout: 90
           - arch: arm64
-            runner: cpu-arm-r8g-4xlarge
+            runner: prod-builder-arm-v1
             timeout: 90
     steps:
       - uses: actions/checkout@v4
@@ -571,10 +571,10 @@ jobs:
         framework: [vllm, trtllm, sglang]
         arch:
           - arch: amd64
-            runner: gpu-l40-amd64
+            runner: prod-builder-amd-gpu-v1
             timeout: 120
           - arch: arm64
-            runner: cpu-arm-r8g-4xlarge
+            runner: prod-builder-arm-v1
             timeout: 120
     steps:
       - uses: actions/checkout@v4
@@ -707,10 +707,10 @@ jobs:
         framework: [vllm, trtllm, sglang]
         arch:
           - arch: amd64
-            runner: gpu-l40-amd64
+            runner: prod-builder-amd-gpu-v1
             timeout: 150
           - arch: arm64
-            runner: cpu-arm-r8g-4xlarge
+            runner: prod-builder-arm-v1
             timeout: 150
     steps:
       - uses: actions/checkout@v4
@@ -836,7 +836,7 @@ jobs:
     name: ${{ matrix.framework.name }}-ft-k8s
     needs: [build-amd64]
     if: always()
-    runs-on: cpu-amd-m5-4xlarge
+    runs-on: prod-builder-amd-v1
     timeout-minutes: 60
     strategy:
       fail-fast: false
@@ -1074,7 +1074,7 @@ jobs:
   ############################## SLACK NOTIFICATION ##############################
   notify-slack:
     name: Notify Slack
-    runs-on: cpu-amd-m5-4xlarge
+    runs-on: prod-builder-amd-v1
     if: always() && inputs.enable_slack_notification && !github.event.repository.fork
     needs: results-summary
     permissions:

.github/workflows/pr.yaml

@@ -63,8 +63,8 @@ jobs:
       fail-fast: false
       matrix:
         platform:
-          - { arch: amd64, runner: cpu-amd-m5-2xlarge }
-          - { arch: arm64, runner: cpu-arm-r8g-4xlarge }
+          - { arch: amd64, runner: prod-builder-amd-v1 }
+          - { arch: arm64, runner: prod-builder-arm-v1 }
     name: operator-build (${{ matrix.platform.arch }})
     runs-on: ${{ matrix.platform.runner }}
     steps:
@@ -81,7 +81,6 @@ jobs:
       - name: Docker Login
         uses: ./.github/actions/docker-login
         with:
-          ngc_ci_access_token: ${{ secrets.NGC_CI_ACCESS_TOKEN }}
           aws_default_region: ${{ secrets.AWS_DEFAULT_REGION }}
           aws_account_id: ${{ secrets.AWS_ACCOUNT_ID }}
           azure_acr_hostname: ${{ secrets.AZURE_ACR_HOSTNAME }}
@@ -141,8 +140,8 @@ jobs:
       fail-fast: false
       matrix:
         platform:
-          - { arch: amd64, runner: gpu-l40-amd64 }
-          - { arch: arm64, runner: cpu-arm-r8g-4xlarge }
+          - { arch: amd64, runner: prod-builder-amd-gpu-v1 }
+          - { arch: arm64, runner: prod-builder-arm-v1 }
         cuda_version:
           - { major_minor: '13.0', major: '13' }
           - { major_minor: '12.9', major: '12' }
@@ -219,8 +218,8 @@ jobs:
       fail-fast: false
       matrix:
         platform:
-          - { arch: amd64, runner: gpu-l40-amd64 }
-          - { arch: arm64, runner: cpu-arm-r8g-4xlarge }
+          - { arch: amd64, runner: prod-builder-amd-gpu-v1 }
+          - { arch: arm64, runner: prod-builder-arm-v1 }
         cuda_version:
           - { major_minor: '13.0', major: '13' }
           - { major_minor: '12.9', major: '12' }
@@ -238,8 +237,8 @@ jobs:
       fail-fast: false
       matrix:
         platform:
-          - { arch: amd64, runner: gpu-l40-amd64 }
-          - { arch: arm64, runner: cpu-arm-r8g-4xlarge }
+          - { arch: amd64, runner: prod-builder-amd-gpu-v1 }
+          - { arch: arm64, runner: prod-builder-arm-v1 }
         cuda_version:
           - { major_minor: '13.0', major: '13' }
     name: trtllm-build-test (cuda${{ matrix.cuda_version.major_minor}}, ${{ matrix.platform.arch }})
@@ -250,7 +249,7 @@ jobs:
     steps: *runtime-container-build-push-test
 
   deploy-operator:
-    runs-on: cpu-amd-m5-2xlarge
+    runs-on: prod-default-v1
     if: needs.changed-files.outputs.core == 'true'
     needs: [changed-files, operator, vllm, sglang, trtllm]
     env:
@@ -325,7 +324,7 @@ jobs:
         timeout 300s kubectl rollout status deployment -n $NAMESPACE --watch
 
   deploy-test-vllm:
-    runs-on: cpu-amd-m5-2xlarge
+    runs-on: prod-default-v1
     # Run if push to main, or manually triggered
     if: ( github.ref_name == 'main' || github.event.inputs.run_deploy_operator )
     needs: [changed-files, deploy-operator, vllm]
@@ -515,7 +514,7 @@ jobs:
   # - Run vllm disagg_router on all commits
   # - Run rest of jobs only on push to main or manual trigger
   deploy-test-vllm-disagg-router:
-    runs-on: cpu-amd-m5-2xlarge
+    runs-on: prod-default-v1
     if: needs.changed-files.outputs.core == 'true' || needs.changed-files.outputs.vllm == 'true' || needs.changed-files.outputs.deploy == 'true'
     needs: [changed-files, deploy-operator, vllm]
     permissions:
@@ -535,7 +534,7 @@ jobs:
     steps: *deploy-test-steps
 
   deploy-test-sglang:
-    runs-on: cpu-amd-m5-2xlarge
+    runs-on: prod-default-v1
     # Run if push to main, or manually triggered
     if: ( github.ref_name == 'main' || github.event.inputs.run_deploy_operator )
     needs: [changed-files, deploy-operator, sglang]
@@ -557,7 +556,7 @@ jobs:
     steps: *deploy-test-steps
 
   deploy-test-trtllm:
-    runs-on: cpu-amd-m5-2xlarge
+    runs-on: prod-default-v1
     # Run if push to main, or manually triggered
     if: ( github.ref_name == 'main' || github.event.inputs.run_deploy_operator )
     needs: [changed-files, deploy-operator, trtllm]
@@ -581,7 +580,7 @@ jobs:
     steps: *deploy-test-steps
 
   cleanup:
-    runs-on: cpu-amd-m5-2xlarge
+    runs-on: prod-default-v1
     if: always()
     needs: [changed-files, deploy-operator, deploy-test-trtllm, deploy-test-sglang, deploy-test-vllm, deploy-test-vllm-disagg-router]
     steps: