|
| 1 | +# |
| 2 | +# Copyright (c) 2025 Huawei Technologies Co., Ltd. All Rights Reserved. |
| 3 | +# |
| 4 | +# Licensed under the Apache License, Version 2.0 (the "License"); |
| 5 | +# you may not use this file except in compliance with the License. |
| 6 | +# You may obtain a copy of the License at |
| 7 | +# |
| 8 | +# http://www.apache.org/licenses/LICENSE-2.0 |
| 9 | +# |
| 10 | +# Unless required by applicable law or agreed to in writing, software |
| 11 | +# distributed under the License is distributed on an "AS IS" BASIS, |
| 12 | +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. |
| 13 | +# See the License for the specific language governing permissions and |
| 14 | +# limitations under the License. |
| 15 | +# This file is a part of the vllm-ascend project. |
| 16 | +# |
| 17 | + |
| 18 | +name: Test CPU-32-HK Runner |
| 19 | + |
| 20 | +on: |
| 21 | + pull_request: |
| 22 | + types: |
| 23 | + - opened |
| 24 | + - synchronize |
| 25 | + - reopened |
| 26 | + push: |
| 27 | + branches: |
| 28 | + - 'test-cpu-32-hk*' |
| 29 | + |
| 30 | +defaults: |
| 31 | + run: |
| 32 | + shell: bash -el {0} |
| 33 | + |
| 34 | +jobs: |
| 35 | + test-cpu-32-hk: |
| 36 | + runs-on: linux-amd64-cpu-32-hk |
| 37 | + steps: |
| 38 | + - name: Checkout code |
| 39 | + uses: actions/checkout@v7 |
| 40 | + with: |
| 41 | + fetch-depth: 0 |
| 42 | + |
| 43 | + - name: System Info |
| 44 | + run: | |
| 45 | + echo "==========================================" |
| 46 | + echo "🔍 CPU-32-HK Runner Diagnostic Info" |
| 47 | + echo "==========================================" |
| 48 | + echo "" |
| 49 | + echo "System Information:" |
| 50 | + uname -a |
| 51 | + echo "" |
| 52 | + echo "CPU Cores: $(nproc)" |
| 53 | + echo "Memory: $(free -h | grep Mem | awk '{print $2}')" |
| 54 | + echo "" |
| 55 | + echo "Current User: $(whoami)" |
| 56 | + echo "Current Directory: $(pwd)" |
| 57 | + echo "" |
| 58 | + echo "Environment Variables:" |
| 59 | + env | grep -E "(BUILDKITD|DOCKER|HOME)" || echo "No buildkitd vars found" |
| 60 | + echo "" |
| 61 | +
|
| 62 | + - name: Verify Vault Certificates |
| 63 | + run: | |
| 64 | + echo "==========================================" |
| 65 | + echo "🔐 Checking Vault Certificate Injection" |
| 66 | + echo "==========================================" |
| 67 | + DOCKER_CONFIG="${DOCKER_CONFIG:-.docker}" |
| 68 | + echo "DOCKER_CONFIG path: $DOCKER_CONFIG" |
| 69 | + echo "" |
| 70 | +
|
| 71 | + if [ -f "$HOME/$DOCKER_CONFIG/ca.pem" ]; then |
| 72 | + echo "✅ ca.pem exists: $(ls -lh $HOME/$DOCKER_CONFIG/ca.pem)" |
| 73 | + else |
| 74 | + echo "⚠️ ca.pem NOT found at $HOME/$DOCKER_CONFIG/ca.pem" |
| 75 | + fi |
| 76 | +
|
| 77 | + if [ -f "$HOME/$DOCKER_CONFIG/cert.pem" ]; then |
| 78 | + echo "✅ cert.pem exists: $(ls -lh $HOME/$DOCKER_CONFIG/cert.pem)" |
| 79 | + else |
| 80 | + echo "⚠️ cert.pem NOT found at $HOME/$DOCKER_CONFIG/cert.pem" |
| 81 | + fi |
| 82 | +
|
| 83 | + if [ -f "$HOME/$DOCKER_CONFIG/key.pem" ]; then |
| 84 | + echo "✅ key.pem exists: $(ls -lh $HOME/$DOCKER_CONFIG/key.pem)" |
| 85 | + else |
| 86 | + echo "⚠️ key.pem NOT found at $HOME/$DOCKER_CONFIG/key.pem" |
| 87 | + fi |
| 88 | +
|
| 89 | + if [ -f "$HOME/$DOCKER_CONFIG/config.json" ]; then |
| 90 | + echo "✅ config.json exists: $(ls -lh $HOME/$DOCKER_CONFIG/config.json)" |
| 91 | + else |
| 92 | + echo "⚠️ config.json NOT found at $HOME/$DOCKER_CONFIG/config.json" |
| 93 | + fi |
| 94 | +
|
| 95 | + echo "" |
| 96 | + if [ -d "$HOME/$DOCKER_CONFIG" ]; then |
| 97 | + echo "Docker config directory contents:" |
| 98 | + ls -la "$HOME/$DOCKER_CONFIG/" || echo "Failed to list directory" |
| 99 | + else |
| 100 | + echo "❌ Docker config directory does not exist" |
| 101 | + fi |
| 102 | + echo "" |
| 103 | +
|
| 104 | + - name: Verify buildctl Installation |
| 105 | + run: | |
| 106 | + echo "==========================================" |
| 107 | + echo "🔨 Checking buildctl Installation" |
| 108 | + echo "==========================================" |
| 109 | +
|
| 110 | + if command -v buildctl &> /dev/null; then |
| 111 | + echo "✅ buildctl found in PATH" |
| 112 | + buildctl --version |
| 113 | + echo "" |
| 114 | + echo "buildctl location: $(which buildctl)" |
| 115 | + else |
| 116 | + echo "❌ buildctl NOT found in PATH" |
| 117 | + echo "" |
| 118 | + echo "Attempting to find buildctl in common locations:" |
| 119 | + find /usr/local/bin -name "buildctl*" 2>/dev/null || echo "Not in /usr/local/bin" |
| 120 | + find /usr/bin -name "buildctl*" 2>/dev/null || echo "Not in /usr/bin" |
| 121 | + find /home -name "buildctl*" 2>/dev/null -o -name "*buildkit*" 2>/dev/null | head -10 || echo "Not found in home" |
| 122 | + fi |
| 123 | + echo "" |
| 124 | +
|
| 125 | + - name: Verify Token Volume Mount |
| 126 | + run: | |
| 127 | + echo "==========================================" |
| 128 | + echo "🔑 Checking Token Volume Mount" |
| 129 | + echo "==========================================" |
| 130 | +
|
| 131 | + TOKEN_PATH="/var/run/secrets/tokens" |
| 132 | +
|
| 133 | + if [ -d "$TOKEN_PATH" ]; then |
| 134 | + echo "✅ Token directory exists: $TOKEN_PATH" |
| 135 | + echo "" |
| 136 | + echo "Directory contents:" |
| 137 | + ls -la "$TOKEN_PATH" || echo "Failed to list directory" |
| 138 | +
|
| 139 | + if [ -f "$TOKEN_PATH/token" ]; then |
| 140 | + echo "" |
| 141 | + echo "✅ Token file exists" |
| 142 | + echo "Token size: $(wc -c < $TOKEN_PATH/token) bytes" |
| 143 | + else |
| 144 | + echo "" |
| 145 | + echo "⚠️ Token file NOT found at $TOKEN_PATH/token" |
| 146 | + fi |
| 147 | + else |
| 148 | + echo "❌ Token directory does not exist: $TOKEN_PATH" |
| 149 | + fi |
| 150 | + echo "" |
| 151 | +
|
| 152 | + - name: Test buildkitd Connectivity |
| 153 | + continue-on-error: true |
| 154 | + run: | |
| 155 | + echo "==========================================" |
| 156 | + echo "🌐 Testing buildkitd Service Connectivity" |
| 157 | + echo "==========================================" |
| 158 | +
|
| 159 | + BUILDKITD_ADDR="${BUILDKITD_ADDR:-tcp://buildkitd-service.buildkitd:1234}" |
| 160 | + echo "BUILDKITD_ADDR: $BUILDKITD_ADDR" |
| 161 | + echo "" |
| 162 | +
|
| 163 | + if command -v buildctl &> /dev/null; then |
| 164 | + echo "Testing buildctl connection..." |
| 165 | + buildctl du || echo "buildctl du failed (expected if service not ready)" |
| 166 | + else |
| 167 | + echo "⚠️ buildctl not available, skipping connectivity test" |
| 168 | + fi |
| 169 | + echo "" |
| 170 | +
|
| 171 | + - name: Verify Cache Volumes |
| 172 | + run: | |
| 173 | + echo "==========================================" |
| 174 | + echo "💾 Checking Cache Volumes" |
| 175 | + echo "==========================================" |
| 176 | +
|
| 177 | + CACHE_DIR="/root/.cache" |
| 178 | +
|
| 179 | + if [ -d "$CACHE_DIR" ]; then |
| 180 | + echo "✅ Cache directory exists: $CACHE_DIR" |
| 181 | + echo "Directory size: $(du -sh $CACHE_DIR 2>/dev/null | cut -f1)" |
| 182 | + else |
| 183 | + echo "❌ Cache directory does not exist: $CACHE_DIR" |
| 184 | + fi |
| 185 | + echo "" |
| 186 | +
|
| 187 | + - name: Keep pod running for manual debugging (15 minutes) |
| 188 | + run: | |
| 189 | + echo "==========================================" |
| 190 | + echo "🛌 Pod Debug Session" |
| 191 | + echo "==========================================" |
| 192 | + echo "" |
| 193 | + echo "Pod is running for manual inspection." |
| 194 | + echo "Debug time limit: 15 minutes (900 seconds)" |
| 195 | + echo "Started: $(date)" |
| 196 | + echo "" |
| 197 | + echo "You can verify setup by running:" |
| 198 | + echo " - buildctl --version" |
| 199 | + echo " - ls ~/.docker/" |
| 200 | + echo " - cat /var/run/secrets/tokens/token" |
| 201 | + echo "" |
| 202 | + sleep 900 |
| 203 | + echo "Debug session ended: $(date)" |
| 204 | + echo "==========================================" |
| 205 | +
|
| 206 | + - name: Final Summary |
| 207 | + if: always() |
| 208 | + run: | |
| 209 | + echo "==========================================" |
| 210 | + echo "✅ CPU-32-HK Runner Test Complete" |
| 211 | + echo "==========================================" |
| 212 | + echo "" |
| 213 | + echo "If this job reached this point, the runner" |
| 214 | + echo "successfully picked up the workflow and" |
| 215 | + echo "created the job pod." |
| 216 | + echo "" |
| 217 | + echo "Check the logs above for:" |
| 218 | + echo " ✓ Vault certificates injected" |
| 219 | + echo " ✓ buildctl installed and available" |
| 220 | + echo " ✓ Token volume mounted" |
| 221 | + echo " ✓ Cache volumes accessible" |
| 222 | + echo "==========================================" |
0 commit comments