try multi-pipeline

.azure-pipelines/integration-test.yml

@@ -25,6 +25,13 @@ pr:
       - docs/**
       - '**/*.md'
 
+parameters:
+- name: hostEntries
+  type: string
+  default: |
+    10.0.0.10 mscclit-000000
+    10.0.0.11 mscclit-000001
+
 jobs:
 - job: SGlangTest
   displayName: SGLANG Test
@@ -55,4 +62,11 @@ jobs:
   pool:
     name: mscclpp-it
   container:
-    image: $(containerImage)
+    image: $(containerImage)
+
+steps:
+- template: templates/sglang-test.yml
+  parameters:
+      subscription:  msccl-it
+      vmssName:      mscclit-vmss
+      resourceGroup: msccl-IT

.azure-pipelines/multi-nodes-test.yml

@@ -11,9 +11,19 @@ trigger:
     - docs/**
     - '**/*.md'
 
-# Do not run multi-nodes-test for PR, we can trigger it manually
-pr: none
-
+pr:
+  branches:
+    include:
+    - main
+    - release/*
+  drafts: false
+  paths:
+    exclude:
+      - .devcontainer/**
+      - .github/**
+      - docker/**
+      - docs/**
+      - '**/*.md'
 
 parameters:
 - name: hostEntries
@@ -23,73 +33,40 @@ parameters:
     10.0.0.11 mscclit-000001
 
 jobs:
-- job: MultiNodesTest
-  displayName: Multi nodes test
+- job: SGlangTest
+  displayName: SGLANG Test
   strategy:
     matrix:
-      cuda11:
-        containerImage: ghcr.io/microsoft/mscclpp/mscclpp:base-dev-cuda11.8
-      cuda12:
-        containerImage: ghcr.io/microsoft/mscclpp/mscclpp:base-dev-cuda12.9
+      sglang:
+        containerImage: ghcr.io/microsoft/mscclpp/mscclpp:base-sglang-x86_64
+        
+  pool:
+    name: msccl-ci-h100
+  container:
+    image: $(containerImage)
+
+  steps:
+  - template: templates/sglang-test.yml
+    parameters:
+      subscription:     mscclpp-ci-h100
+      vmssName:         mscclpp-h100-ci
+      gpuArch:          '90'
+
+- job: SGlangMultiNodeTest
+  displayName: SGLANG Multi-Node Test
+  strategy:
+    matrix:
+      sglang:
+        containerImage: ghcr.io/microsoft/mscclpp/mscclpp:base-sglang-x86_64
+
   pool:
     name: mscclpp-it
   container:
-    image: $[ variables['containerImage'] ]
+    image: $(containerImage)
 
-  steps:
-  - task: Bash@3
-    displayName: Add HostEntry
-    inputs:
-      targetType: 'inline'
-      script: |
-        ENTRY="${{ parameters.hostEntries }}"
-        if ! grep -qxF "$ENTRY" /etc/hosts; then
-          echo "Adding to /etc/hosts"
-          echo "$ENTRY" | sudo tee -a /etc/hosts
-        else
-          echo "Entry already exists, nothing to do."
-        fi
-
-  - template: templates/deploy.yml
-    parameters:
+steps:
+- template: templates/sglang-test.yml
+  parameters:
       subscription:  msccl-it
       vmssName:      mscclit-vmss
-      resourceGroup: msccl-IT
-
-  - template: templates/run-remote-task.yml
-    parameters:
-      name: RunMscclppTest
-      displayName: Run multi-nodes mscclpp-test
-      runRemoteArgs: '--hostfile $(System.DefaultWorkingDirectory)/test/deploy/hostfile --host mscclit-000000 --user azureuser'
-      remoteScript: |
-        bash /root/mscclpp/test/deploy/run_tests.sh mscclpp-test
-
-  - template: templates/run-remote-task.yml
-    parameters:
-      name: RunMultiNodeUnitTest
-      displayName: Run multi-nodes unit tests
-      runRemoteArgs: '--hostfile $(System.DefaultWorkingDirectory)/test/deploy/hostfile --host mscclit-000000 --user azureuser'
-      remoteScript: |
-        bash /root/mscclpp/test/deploy/run_tests.sh mp-ut
-
-  - template: templates/run-remote-task.yml
-    parameters:
-      name: RunMultiNodePythonTests
-      displayName: Run multi-nodes python tests
-      runRemoteArgs: '--hostfile $(System.DefaultWorkingDirectory)/test/deploy/hostfile --host mscclit-000000 --user azureuser'
-      remoteScript: |
-        bash /root/mscclpp/test/deploy/run_tests.sh pytests
-
-  - template: templates/run-remote-task.yml
-    parameters:
-      name: RunMultiNodePythonBenchmark
-      displayName: Run multi-nodes python benchmark
-      runRemoteArgs: '--hostfile $(System.DefaultWorkingDirectory)/test/deploy/hostfile --host mscclit-000000 --user azureuser'
-      remoteScript: |
-        bash /root/mscclpp/test/deploy/run_tests.sh py-benchmark
-
-  - template: templates/stop.yml
-    parameters:
-      subscription:  msccl-it
-      vmssName:      mscclit-vmss
-      resourceGroup: msccl-IT
+      resourceGroup: msccl-IT