性能测试的已知限制

执行持续时间限制

Cloud, serverless robots can run for a maximum of 60 minutes per test. On-premises robots can exceed this limit.

并发限制

• Maximum concurrent jobs: Up to 500 jobs can run concurrently using serverless robots.

• Impact on virtual users (VUs): The number of achievable virtual users depends on multiplexing (multiple VUs per job). For example, with a multiplexing factor of 4, a scenario can scale up to 2,000 concurrent virtual users. The multiplexing capabilities for serverless robots are automatically detected during the dry run, and the resulting factor is reported in the application logs. Unlike on-premises machines, where the multiplexing factor can be manually overridden, this is not supported for serverless robots.

• VPN constraints: When using a VPN, the effective concurrency depends on the VPN or gateway capacity.

  • Default limit: ~250 concurrent jobs
  • The actual limit may vary based on the gateway SKU and infrastructure capacity.
  • If higher concurrency is required, customers should contact UiPath support or their account team to discuss VPN scaling options.

• Execution duration: Serverless robots are limited to 60 minutes per execution

支持的自动化类型

性能测试目前仅限于浏览器、API ( HTTP WebRequest) 和桌面自动化。例如，尚不支持 Integration Service 或编码自动化。

本地部署环境不支持匿名/InPrivate 模式

本地部署计算机上的性能测试不支持 Chrome 隐身模式或 Edge InPrivate 模式下的浏览器自动化。

出现此限制的原因是浏览器安全限制，这些限制要求为隐私浏览手动启用扩展程序。尽管可以安装 UiPath 浏览器扩展程序（例如，通过组策略），但无法在隐身模式下自动启用该扩展程序，这会阻止自动化在测试执行期间与浏览器交互。

Additionally, Performance Testing creates new browser user directories for each execution. Enabling the extension for incognito mode would require manual configuration for each of these directories, making it impractical for automated scenarios. For more information, refer to the Alternative for enabling incognito mode section of the Studio documentation.

此限制仅适用于本地部署环境。

在无服务器环境中，支持匿名模式，因为浏览器配置完全由 UiPath 管理。

Workaround: Run tests in standard browser mode.

性能测试创建自己专用的浏览器配置文件和目录，以确保它不会干扰用户的默认配置文件。此设置允许在沙盒环境中运行测试（类似于隐身模式）。执行后，系统将清理所有为测试创建的浏览器目录，不留下任何残留数据。

其他限制

• Infrastructure gaps: ACR-VM support and elastic provisioning are not yet available, so automatic scaling of resources is currently not possible.
• API access: Direct programmatic access to raw performance test data is not yet supported.
• Historical comparison: Test results from different historical runs cannot yet be compared.
• Autopilot and agentic support: Autopilot and agentic support are not available in the current release.
• Recordings displayed in the detail: Recording can be activated and viewd only in Orchestrator.