fix(cd): 序列建構修復 Runner _runner_file_commands 衝突

根因: 並行 Job 的 Set up job 階段會同時寫入 RUNNER_TEMP
解法: build-api needs build-web，確保序列執行
移除: Job-level concurrency groups (不再需要)
更新: ops/runner/README.md v1.0→v2.0

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>

.github/workflows/cd.yaml

@@ -1,11 +1,11 @@
 # =============================================================================
-# AWOOOI CD Pipeline v2.0 (完整沿用 AIOPS 最佳實踐)
+# AWOOOI CD Pipeline v2.1 (序列建構修復 Runner 衝突)
 # =============================================================================
 # 優化項目:
 # 1. Pre-flight Check (10s Fail-Fast)
 # 2. Runner 標籤 [self-hosted, harbor, k8s]
 # 3. dorny/paths-filter 精確路徑偵測
-# 4. API + Web 並行建構
+# 4. Web → API 序列建構 (2026-03-29 修復 _runner_file_commands 衝突)
 # 5. timeout-minutes 防止卡死
 # 6. Telegram + OpenClaw 通知
 # 7. force_deploy 強制重建選項
@@ -57,11 +57,6 @@ jobs:
     name: "Pre-flight Check"
     runs-on: [self-hosted, harbor, k8s]
     timeout-minutes: 1
-    # 2026-03-29: Runner 全局 mutex，確保同一 Runner 不會並行執行任何 CD Job
-    # 使用固定 group 名稱 (非 run_id)，所有 CD Jobs 共用同一把鎖
-    concurrency:
-      group: runner-awoooi-cd-mutex
-      cancel-in-progress: false
     steps:
       # =======================================================================
       # 2026-03-29: Runner _diag/pages 檔案衝突修復 (v3)
@@ -135,9 +130,6 @@ jobs:
     runs-on: [self-hosted, harbor, k8s]
     needs: pre-flight-check
     timeout-minutes: 1
-    concurrency:
-      group: runner-awoooi-cd-mutex
-      cancel-in-progress: false
     outputs:
       api: ${{ inputs.force_deploy == true && 'true' || steps.filter.outputs.api }}
       web: ${{ inputs.force_deploy == true && 'true' || steps.filter.outputs.web }}
@@ -171,15 +163,15 @@ jobs:
             k3s-system:
               - 'k8s/k3s-system/**'
 
-  # ==================== 並行建構 API ====================
+  # ==================== 序列建構 API (必須等 Web 完成) ====================
+  # 2026-03-29 Claude Code: 改為序列執行，修復 Runner _runner_file_commands 衝突
+  # 根因: 並行 Job 的 "Set up job" 階段會互相覆寫 RUNNER_TEMP 檔案
+  # 參考: ops/runner/README.md
   build-api:
     name: "Build API"
     runs-on: [self-hosted, harbor, k8s]
-    needs: detect-changes
+    needs: [detect-changes, build-web]
     timeout-minutes: 20
-    concurrency:
-      group: runner-awoooi-cd-mutex
-      cancel-in-progress: false
     if: |
       !inputs.skip_api && (
         needs.detect-changes.outputs.api == 'true' ||
@@ -214,15 +206,13 @@ jobs:
             --file apps/api/Dockerfile .
           echo "✅ API: ${{ env.REGISTRY }}/${{ env.IMAGE_PREFIX }}-api:${{ steps.tag.outputs.tag }}"
 
-  # ==================== 並行建構 Web ====================
+  # ==================== 建構 Web (先執行) ====================
+  # 2026-03-29 Claude Code: API 依賴 Web 完成，確保序列執行
   build-web:
     name: "Build Web"
     runs-on: [self-hosted, harbor, k8s]
     needs: detect-changes
     timeout-minutes: 20
-    concurrency:
-      group: runner-awoooi-cd-mutex
-      cancel-in-progress: false
     if: |
       !inputs.skip_web && (
         needs.detect-changes.outputs.web == 'true' ||

ops/runner/README.md

@@ -17,19 +17,29 @@ Error: The file '/home/wooo/actions-runner-awoooi/_diag/pages/xxx.log' already e
    - `_work/_temp/_runner_file_commands` 在所有 Jobs 之間共享
    - 清理此目錄會導致 "Missing file at path" 錯誤
 
-### 解決方案 (v3 - 最終版)
+### 解決方案 (v4 - 最終版 2026-03-29)
 
-#### 1. Workflow Concurrency (核心修復)
+#### 1. 序列建構 (核心修復)
+```yaml
+# build-api 必須等 build-web 完成
+build-api:
+  needs: [detect-changes, build-web]  # 關鍵: 依賴 build-web
+```
+
+**根因**: Job 並行時，"Set up job" 階段會同時寫入 `_runner_file_commands`，導致衝突
+**解法**: 改為序列執行，確保同一時間只有一個 Job 在 Runner 上
+
+#### 2. Workflow Concurrency (輔助)
 ```yaml
 concurrency:
   group: cd-${{ github.workflow }}-${{ github.ref }}
-  cancel-in-progress: true  # 必須為 true！
+  cancel-in-progress: true
 ```
 
-**關鍵**: `cancel-in-progress: true` 確保同一時間只有一個 workflow 在執行
+確保同一時間只有一個 workflow 在執行
 
-#### 2. Job 層清理 (輔助)
-每個 Job 開始時只清理 `_diag/pages`，**不碰** `RUNNER_TEMP`：
+#### 3. Job 層清理 (防禦性)
+每個 Job 開始時清理 `_diag/pages`：
 
 ```yaml
 - name: "Clean Runner Diagnostics"
@@ -39,7 +49,7 @@ concurrency:
     mkdir -p "$RUNNER_ROOT/_diag/pages" 2>/dev/null || true
 ```
 
-**警告**: 絕對不要清理 `$RUNNER_TEMP/*`，會破壞其他 Job 的內部通訊
+**警告**: 絕對不要清理 `$RUNNER_TEMP/*`，會破壞 `_runner_file_commands`
 
 #### 2. Systemd Timer (背景清理)
 每 5 分鐘自動清理過期的診斷檔案:
@@ -78,4 +88,5 @@ sudo systemctl start runner-diag-cleanup.service
 - ADR: 待建立 (如果問題持續)
 
 ---
-版本: v1.0 | 建立: 2026-03-29 | 作者: Claude Code
+版本: v2.0 | 更新: 2026-03-29 | 作者: Claude Code
+變更: v1.0→v2.0 序列建構取代 Job Concurrency Groups