Azure DevOps Server 2019并行流水线任务被取消问题求助

我们使用Azure DevOps Server 2019，在同一服务器上创建了代理池并配置了两个代理。当前遇到的问题：

• 单条流水线可正常运行
• 并行运行两条流水线时，其中一条会在数分钟后被取消，仅提示错误The operation was canceled，无其他详细日志
• 怀疑是同一服务器上的两个代理共用MSBuild导致冲突

流水线截图 #

失败流水线的system.debug日志 #

日志1（Build solution步骤失败）

##[debug]Re-evaluate condition on job cancellation for step: 'Build solution **/*.sln'.
##[debug]Leaving Get-SolutionFiles.
##[debug]Leaving C:\agents\w1\_tasks\MSBuild_c6c4c611-aa2e-4a33-b606-5eaba2196824\1.166.2\MSBuild.ps1.
##[error]The operation was canceled.
##[debug]System.OperationCanceledException: The operation was canceled. at System.Threading.CancellationToken.ThrowOperationCanceledException() at Microsoft.VisualStudio.Services.Agent.Util.ProcessInvoker.ExecuteAsync(String workingDirectory, String fileName, String arguments, IDictionary`2 environment, Boolean requireExitCodeZero, Encoding outputEncoding, Boolean killProcessOnCancel, InputQueue`1 redirectStandardIn, Boolean inheritConsoleHandler, Boolean keepStandardInOpen, Boolean highPriorityProcess, CancellationToken cancellationToken) at Microsoft.VisualStudio.Services.Agent.ProcessInvokerWrapper.ExecuteAsync(String workingDirectory, String fileName, String arguments, IDictionary`2 environment, Boolean requireExitCodeZero, Encoding outputEncoding, Boolean killProcessOnCancel, InputQueue`1 redirectStandardIn, Boolean inheritConsoleHandler, Boolean keepStandardInOpen, Boolean highPriorityProcess, CancellationToken cancellationToken) at Microsoft.VisualStudio.Services.Agent.Worker.Handlers.DefaultStepHost.ExecuteAsync(String workingDirectory, String fileName, String arguments, IDictionary`2 environment, Boolean requireExitCodeZero, Encoding outputEncoding, Boolean killProcessOnCancel, Boolean inheritConsoleHandler, CancellationToken cancellationToken) at Microsoft.VisualStudio.Services.Agent.Worker.Handlers.PowerShell3Handler.RunAsync() at Microsoft.VisualStudio.Services.Agent.Worker.TaskRunner.RunAsync() at Microsoft.VisualStudio.Services.Agent.Worker.StepsRunner.RunStepAsync(IStep step, CancellationToken jobCancellationToken)
Finishing: Build solution **/*.sln

日志2（NuGet restore步骤失败）

##[debug]  C:\agents\w1\106\s\ClientApp\node_modules\faker\lib\locales\de_CH\name (directory)
##[error]The operation was canceled.
##[debug]System.OperationCanceledException: The operation was canceled. at System.Threading.CancellationToken.ThrowOperationCanceledException() at Microsoft.VisualStudio.Services.Agent.Util.ProcessInvoker.ExecuteAsync(String workingDirectory, String fileName, String arguments, IDictionary`2 environment, Boolean requireExitCodeZero, Encoding outputEncoding, Boolean killProcessOnCancel, InputQueue`1 redirectStandardIn, Boolean inheritConsoleHandler, Boolean keepStandardInOpen, Boolean highPriorityProcess, CancellationToken cancellationToken) at Microsoft.VisualStudio.Services.Agent.ProcessInvokerWrapper.ExecuteAsync(String workingDirectory, String fileName, String arguments, IDictionary`2 environment, Boolean requireExitCodeZero, Encoding outputEncoding, Boolean killProcessOnCancel, InputQueue`1 redirectStandardIn, Boolean inheritConsoleHandler, Boolean keepStandardInOpen, Boolean highPriorityProcess, CancellationToken cancellationToken) at Microsoft.VisualStudio.Services.Agent.Worker.Handlers.DefaultStepHost.ExecuteAsync(String workingDirectory, String fileName, String arguments, IDictionary`2 environment, Boolean requireExitCodeZero, Encoding outputEncoding, Boolean killProcessOnCancel, Boolean inheritConsoleHandler, CancellationToken cancellationToken) at Microsoft.VisualStudio.Services.Agent.Worker.Handlers.NodeHandler.RunAsync() at Microsoft.VisualStudio.Services.Agent.Worker.TaskRunner.RunAsync() at Microsoft.VisualStudio.Services.Agent.Worker.StepsRunner.RunStepAsync(IStep step, CancellationToken jobCancellationToken)
Finishing: NuGet restore

排查与解决步骤 #

1. 隔离代理运行账户
   确保两个代理使用不同的本地用户账户运行，避免同一账户下的资源锁冲突。卸载现有代理后，用不同账户重新配置两个代理。

2. 验证MSBuild并发兼容性

   • 升级MSBuild到对应VS版本的最新版（如VS2019的MSBuild 16.x），旧版本可能存在单实例限制。
   • 在MSBuild步骤添加参数/m:1禁用并行构建，测试是否仍出现取消问题，排除构建本身的并行冲突。

3. 检查服务器资源瓶颈
   监控并行运行时的CPU、内存、磁盘IO占用，若资源耗尽，Agent会主动终止任务。可限制代理池并行任务数或升级服务器硬件。

4. 确保代理工作目录独立
   配置两个代理时指定不同的工作目录（work文件夹），避免同一目录下的文件锁冲突。

5. 查看详细日志定位原因

   • 服务器端日志路径：C:\Program Files\Azure DevOps Server 2019\Application Tier\Logs，查找任务取消触发源。
   • 代理端日志路径：每个代理的_diag文件夹，检查是否有资源访问拒绝、进程强制终止记录。

6. 禁用代理进程复用
   在代理的config.json中添加"reuseProcess": false，强制每个任务启动新进程，避免进程间资源共享冲突，修改后重启代理服务。

内容的提问来源于stack exchange，提问作者Harish334363