perf: 遥测上传失败节点

[zmdyy0318]

opus 5.0

Summary by Sourcery

在遥测中将失败的流水线节点作为任务 span 下的子 span 进行附加，以为 Maa 任务提供详细、可追踪的失败链路（包括嵌套流水线和时间信息），同时对每个任务的 span 数量施加限制。

New Features:

• 在任务 span 之上为失败的流水线节点创建子 span，以捕获 Maa 流水线中的细粒度失败点。
• 追踪流水线步骤的开始时间，以记录任务在失败节点上卡住的时长。
• 在遥测中支持将嵌套子流水线的失败归因并标记到其所属任务下。
• 在 span 上暴露 MaaFW 任务 ID 和流水线节点元数据，以帮助与用户日志进行关联。

Enhancements:

• 用结构化的、按节点划分的失败 span 和按任务限制，替换单一失败节点标签，以避免超过 Sentry 的 span 上限。
• 在任务完成时清理与该任务相关的遥测状态，包括步骤计时、失败次数、嵌套流水线条目以及活跃任务跟踪信息。

Original summary in English

Summary by Sourcery

Attach failed pipeline nodes as child spans under task spans in telemetry to provide a detailed, traceable failure chain for Maa tasks, including nested pipelines and timing information, while enforcing per-task span limits.

New Features:

• Create child spans for failed pipeline nodes on top of task spans to capture granular failure points in Maa pipelines.
• Track pipeline step start times to record how long tasks are stuck on failed nodes.
• Support attribution and labeling of failures from nested sub-pipelines under their owning tasks in telemetry.
• Expose MaaFW task IDs and pipeline node metadata on spans to aid correlation with user logs.

Enhancements:

• Replace single failed-node tagging with structured per-node failure spans and per-task limits to avoid exceeding Sentry span caps.
• Clean up per-task telemetry state on task completion, including step timing, failure counts, nested pipeline entries, and active task tracking.

[sourcery-ai]

嗨，我在这里给出了一些总体反馈：

• active_task 和 nested_entries 这部分逻辑假设 Context::run_task 的执行是单线程并且严格串行的；如果存在任务重叠或可重入的可能性，最好在文档中明确说明这一假设，或者增加保护措施，以避免将节点失败错误地归属到错误的父 span。
• 在 record_failed_node 中，failed_nodes 会在检查是否存在 task_span 之前就递增，这意味着有些失败会被计数但不会真正体现为 span；可以考虑在成功获取父 span 之后再进行限流计数，这样该计数器就能反映真正上报的节点 span。

面向 AI Agent 的提示

Please address the comments from this code review:

## Overall Comments
- The `active_task` and `nested_entries` logic assumes a single-threaded, strictly serial `Context::run_task` flow; if there is any chance of overlapping tasks or re-entrancy, it would be safer to document that assumption clearly or add guards to avoid mis-attributing node failures to the wrong parent span.
- In `record_failed_node`, `failed_nodes` is incremented before checking whether a `task_span` exists, which means some failures are counted but never materialize as spans; consider moving the rate-limiting increment after successfully obtaining the parent span so the counter reflects actual reported node spans.

---

Sourcery 对开源项目是免费的——如果你觉得我们的评审有帮助，欢迎分享 ✨

• X
• Mastodon
• LinkedIn
• Facebook

_{帮我变得更有用！请对每条评论点击 👍 或 👎，我会根据你的反馈改进后续的代码评审。}

Original comment in English

Hey - I've left some high level feedback:

• The active_task and nested_entries logic assumes a single-threaded, strictly serial Context::run_task flow; if there is any chance of overlapping tasks or re-entrancy, it would be safer to document that assumption clearly or add guards to avoid mis-attributing node failures to the wrong parent span.
• In record_failed_node, failed_nodes is incremented before checking whether a task_span exists, which means some failures are counted but never materialize as spans; consider moving the rate-limiting increment after successfully obtaining the parent span so the counter reflects actual reported node spans.

Prompt for AI Agents

Please address the comments from this code review:

## Overall Comments
- The `active_task` and `nested_entries` logic assumes a single-threaded, strictly serial `Context::run_task` flow; if there is any chance of overlapping tasks or re-entrancy, it would be safer to document that assumption clearly or add guards to avoid mis-attributing node failures to the wrong parent span.
- In `record_failed_node`, `failed_nodes` is incremented before checking whether a `task_span` exists, which means some failures are counted but never materialize as spans; consider moving the rate-limiting increment after successfully obtaining the parent span so the counter reflects actual reported node spans.

---

Sourcery is free for open source - if you like our reviews please consider sharing them ✨

• X
• Mastodon
• LinkedIn
• Facebook

_{Help me be more useful! Please click 👍 or 👎 on each comment and I'll use the feedback to improve your reviews.}