[ci] split stage-c-test-4-gpu-b200 to enable a low-disk runner pool (#23417)

Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

.github/workflows/pr-test.yml

@@ -92,6 +92,7 @@ jobs:
       max_parallel_small: ${{ steps.set-parallel.outputs.max_parallel_small }}
       max_parallel_2gpu: ${{ steps.set-parallel.outputs.max_parallel_2gpu }}
       b200_runner: ${{ steps.set-runner.outputs.b200_runner }}
+      b200_low_disk_runner: ${{ steps.set-runner.outputs.b200_low_disk_runner }}
       enable_retry: ${{ steps.set-retry.outputs.enable_retry }}
       continue_on_error: ${{ steps.set-continue-on-error.outputs.continue_on_error }}
     steps:
@@ -273,8 +274,10 @@ jobs:
           target_stage="${{ inputs.target_stage }}"
           if [[ "$sgl_kernel" == "true" && -z "$target_stage" ]]; then
             echo "b200_runner=4-gpu-b200-kernel" >> $GITHUB_OUTPUT
+            echo "b200_low_disk_runner=4-gpu-b200-kernel-low-disk" >> $GITHUB_OUTPUT
           else
             echo "b200_runner=4-gpu-b200" >> $GITHUB_OUTPUT
+            echo "b200_low_disk_runner=4-gpu-b200-low-disk" >> $GITHUB_OUTPUT
           fi
 
       - name: Enable retry for CI
@@ -330,6 +333,7 @@ jobs:
             echo "| detection_method  | ${{ inputs.target_stage && 'GitHub API' || 'dorny/paths-filter' }} |"
             echo "| max_parallel      | ${{ steps.set-parallel.outputs.parallel_level }} (h100=${{ steps.set-parallel.outputs.max_parallel }}, 5090=${{ steps.set-parallel.outputs.max_parallel_small }}, 2gpu=${{ steps.set-parallel.outputs.max_parallel_2gpu }}) |"
             echo "| b200_runner       | ${{ steps.set-runner.outputs.b200_runner }} |"
+            echo "| b200_low_disk_runner | ${{ steps.set-runner.outputs.b200_low_disk_runner }} |"
             echo "| enable_retry      | ${{ steps.set-retry.outputs.enable_retry }} |"
             echo "| continue_on_error | ${{ steps.set-continue-on-error.outputs.continue_on_error }} |"
           } >> $GITHUB_STEP_SUMMARY
@@ -1293,7 +1297,7 @@ jobs:
     strategy:
       fail-fast: false
       matrix:
-        part: [0, 1, 2, 3, 4, 5]
+        part: [0, 1, 2]
 
     steps:
       - name: Checkout code
@@ -1324,7 +1328,69 @@ jobs:
           CONTINUE_ON_ERROR_FLAG: ${{ needs.check-changes.outputs.continue_on_error == 'true' && '--continue-on-error' || '' }}
         run: |
           cd test
-          python3 run_suite.py --hw cuda --suite stage-c-test-4-gpu-b200 --auto-partition-id ${{ matrix.part }} --auto-partition-size 6 --timeout-per-file 1800 $CONTINUE_ON_ERROR_FLAG
+          python3 run_suite.py --hw cuda --suite stage-c-test-4-gpu-b200 --auto-partition-id ${{ matrix.part }} --auto-partition-size 3 --timeout-per-file 1800 $CONTINUE_ON_ERROR_FLAG
+
+      - uses: ./.github/actions/upload-cuda-coredumps
+        if: failure()
+        with:
+          artifact-suffix: ${{ matrix.part }}
+
+      - name: Cleanup venv
+        if: always()
+        run: bash scripts/ci/cuda/ci_cleanup_venv.sh
+
+  stage-c-test-4-gpu-b200-small:
+    needs: [check-changes, call-gate, wait-for-stage-b]
+    if: |
+      always() &&
+      (
+        (inputs.target_stage == 'stage-c-test-4-gpu-b200-small') ||
+        (
+          !inputs.target_stage &&
+          ((github.event_name == 'schedule' || inputs.test_parallel_dispatch == true) || (!failure() && !cancelled())) &&
+          ((needs.check-changes.outputs.main_package == 'true') || (needs.check-changes.outputs.sgl_kernel == 'true'))
+        )
+      )
+    # The `*-low-disk` label (resolved by `set-runner` to `4-gpu-b200-low-disk` or
+    # `4-gpu-b200-kernel-low-disk`) is advertised by both the existing large-disk B200
+    # runners and the new low-disk runner, so this job can land on either pool.
+    runs-on: ${{ needs.check-changes.outputs.b200_low_disk_runner }}
+    timeout-minutes: 240
+    strategy:
+      fail-fast: false
+      matrix:
+        part: [0, 1, 2]
+
+    steps:
+      - name: Checkout code
+        uses: actions/checkout@v4
+        with:
+          ref: ${{ inputs.pr_head_sha || inputs.git_ref || github.sha }}
+
+      - uses: ./.github/actions/check-stage-health
+
+      - uses: ./.github/actions/check-maintenance
+
+      - name: Download artifacts
+        if: needs.check-changes.outputs.sgl_kernel == 'true'
+        uses: actions/download-artifact@v6
+        with:
+          path: sgl-kernel/dist/
+          merge-multiple: true
+          pattern: wheel-python3.10-cuda13.0
+
+      - name: Install dependencies
+        timeout-minutes: 20
+        run: |
+          CUSTOM_BUILD_SGL_KERNEL=${{needs.check-changes.outputs.sgl_kernel}} bash scripts/ci/cuda/ci_install_dependency.sh
+
+      - name: Run test
+        timeout-minutes: 30
+        env:
+          CONTINUE_ON_ERROR_FLAG: ${{ needs.check-changes.outputs.continue_on_error == 'true' && '--continue-on-error' || '' }}
+        run: |
+          cd test
+          python3 run_suite.py --hw cuda --suite stage-c-test-4-gpu-b200-small --auto-partition-id ${{ matrix.part }} --auto-partition-size 3 --timeout-per-file 1800 $CONTINUE_ON_ERROR_FLAG
 
       - uses: ./.github/actions/upload-cuda-coredumps
         if: failure()
@@ -1416,6 +1482,7 @@ jobs:
         stage-c-test-deepep-4-gpu-h100,
         stage-c-test-deepep-8-gpu-h200,
         stage-c-test-4-gpu-b200,
+        stage-c-test-4-gpu-b200-small,
         # stage-c-test-4-gpu-gb200,  # Temporarily disabled — no GB200 runner
       ]
     if: always()