Evaluation and effect errors（评估与效果错误）¶
This page describes some common error categories that may be encountered when using Automate.
Evaluation errors¶
An automation may fail to evaluate due to problems with the underlying data. Automations are automatically retried, but some errors may require manual intervention. For example, if the object type being monitored is deleted, automations using that type will fail to evaluate.
Automation out of sync¶
Automations may use a reference to a saved exploration to define the input. This reference is not dynamic, but instead is stored according to the exploration as it exists when the automation is saved. If the exploration changes, the automation will continue to evaluate using the exploration's old state unless the automation is updated. In this case, a warning banner is displayed on the automation:
Permissions¶
Automation evaluation uses the permissions of the owner or recipients. This ensures that condition evaluation and any subsequent effects always reflect data that the user has access to at the time the automation is evaluated. If a user lacks permission to view object types, saved explorations, or the automation, they may see a permission-related error message instead of successful evaluation.
We strongly recommend storing automations and their related resources in shared Projects. Learn more about automation scope and log visibility.
Effect failures and fallback effects¶
To handle effect failures gracefully, you can configure a fallback effect for action and Logic effects. Fallback effects execute when the primary effect fails, allowing you to send notifications, log failures, or trigger alternative workflows.
Action effect errors¶
After a successful condition evaluation, action effects may fail to execute. This failure could happen for a variety of reasons, including the following:
- Changes to the action logic make it incompatible with the saved input configuration on the automation.
- The submission criteria for the action are not met.
- The function backing the action has a runtime execution failure.
If an action effect fails, the history timeline will indicate that one or more actions failed to execute for that event, along with relevant error details.
Note that when per-object execution is enabled, the object identifier surfaced in the error details as the cause of the error represents the object associated with the first request that caused the failure; there may be more hidden failures that are not propagated.
Cycle detection¶
Consider a set of automations defined as follows:
- Automation A: Monitors an object set of all red cars. When a new red car is added to the ontology, an action triggers to change the car color to blue.
- Automation B: Monitors an object set of all blue cars. When a new blue car is added to the ontology, an action triggers to change the car color to red.
Such a sequence of automations would cause an infinite loop, or cycle. A framework has been implemented to automatically detect and disable live automations that cause cycles.
For certain automations, cycle detection may be undesirable. Cycles can be allowed by overriding the configuration in the Settings tab of an automation.
Notification effect errors¶
After a successful condition evaluation, notifications may fail to send. If this occurs, the history event will show a tag indicating that notifications were not sent to subscribers. Additional details will include an error identifier, the error message, and the object or objects that triggered the failure.
中文翻译¶
评估与效果错误¶
本文介绍了使用 Automate 时可能遇到的一些常见错误类别。
评估错误¶
自动化可能因底层数据问题而无法评估。自动化会自动进行重试，但某些错误可能需要手动干预。例如，如果被监控的对象类型被删除，使用该类型的自动化将无法评估。
自动化不同步¶
自动化可能会引用已保存的探索来定义输入。此引用并非动态的，而是根据自动化保存时探索的现有状态进行存储。如果探索发生变化，除非更新自动化，否则自动化将继续使用探索的旧状态进行评估。在这种情况下，自动化上会显示一个警告横幅：
权限¶
自动化评估使用所有者或收件人的权限。这确保了条件评估及后续任何效果始终反映用户在自动化评估时有权访问的数据。如果用户缺乏查看对象类型、已保存探索或自动化的权限，他们可能会看到与权限相关的错误消息，而非成功的评估结果。
我们强烈建议将自动化及其相关资源存储在共享的项目中。了解更多关于自动化范围与日志可见性的信息。
效果失败与回退效果¶
为了优雅地处理效果失败，您可以为操作效果和逻辑效果配置回退效果。当主要效果失败时，回退效果会执行，允许您发送通知、记录失败或触发替代工作流。
操作效果错误¶
在条件评估成功之后，操作效果可能无法执行。这种失败可能由多种原因引起，包括以下情况：
- 操作逻辑的更改使其与自动化上保存的输入配置不兼容。
- 未满足操作的提交标准。
- 支持操作的函数出现运行时执行失败。
如果操作效果失败，历史时间线将指示该事件中有一个或多个操作未能执行，并附带相关的错误详情。
请注意，当启用按对象执行时，错误详情中作为错误原因显示的对象标识符代表与导致失败的第一个请求相关联的对象；可能还存在更多未传播的隐藏失败。
循环检测¶
考虑以下定义的一组自动化：
- 自动化 A： 监控所有红色汽车的对象集。当本体论(Ontology)中添加了一辆新的红色汽车时，触发一个操作将汽车颜色改为蓝色。
- 自动化 B： 监控所有蓝色汽车的对象集。当本体论(Ontology)中添加了一辆新的蓝色汽车时，触发一个操作将汽车颜色改为红色。
这样的一系列自动化会导致无限循环，即循环。已实现一个框架来自动检测并禁用导致循环的实时自动化。
对于某些自动化，循环检测可能是不必要的。可以通过覆盖自动化设置选项卡中的配置来允许循环。
通知效果错误¶
在条件评估成功之后，通知可能无法发送。如果发生这种情况，历史事件将显示一个标签，指示通知未发送给订阅者。附加详情将包括错误标识符、错误消息以及触发失败的一个或多个对象。