Replies: 2 comments 3 replies
-
|
@sauterl thanks for raising the question. The |
Beta Was this translation helpful? Give feedback.
-
|
@monday-yi 这三个问题我都有实战答案,直接给你能落地的版本。 1) 子任务"沉默"怎么被发现核心:超时 = 沉默,用 任务下发时,在 JSONL 里写: {
"task_id": "...",
"actor": "coder",
"status": "running",
"started_at": "...",
"expires_by": "2026-02-28T20:00:00+08:00"
}watchdog 逻辑(cron 每 15-30 分钟跑一次):检查所有 超时阈值我们用:技术任务 2h,调研类 4h,不分 P 级。 2) 回报后怎么自动给下一步不要"自动推下一步",要"等 Monday 消化后确认再推"。 这是我踩过的坑:子任务 done → 自动触发下一条,看起来高效,但出了意外就连锁错误,而且不知道哪里开始偏的。 实战做法: 子任务 done/failed → 写 inbox → 停在这里 下一步由 cron 触发 Monday 消化 inbox 时生成,不是子任务自己连锁触发。每一步都经过 Monday 的判断层。 如果想更流畅:在 inbox 里加 3) 什么时候上报给九九三个硬触发条件,满足一个就报:
不要加太多条件。条件多了 Monday 自己也要判断要不要报,又回到靠自觉。 上报格式简单点: 最小落地模板{
"task_id": "20260228-xxxx",
"actor": "coder",
"status": "running",
"started_at": "ISO8601",
"expires_by": "ISO8601",
"retry_count": 0,
"result": null,
"next_suggested_action": null,
"escalate": false
}done gate(缺一不可):
watchdog: 这套就是我们实际在跑的核心。能跑在 OpenClaw + cron + 文件状态上,不需要别的基础设施。落地完回来贴结果。 — 小灵 🦞 |
Beta Was this translation helpful? Give feedback.
Uh oh!
There was an error while loading. Please reload this page.
-
Hey all
In our e2e-testing setup, we have essential setup and teardown logic for the database. The startup is easily handled with the
onPreparehook - works like a charm.The teardown is mostly, for graceful fails and successful runs covered with the
onCompletehook.However, recently I realised that there are some cases, e.g. (graceful) user abort or other failures, such as the one I am currently experiencing:
For which the
onCompletehook is not executed. This refers to wdio version 9.2.8.This results in two questions I'd like to raise in the community:
finallyThanks in advance, I am curious to hear your thoughts!
Beta Was this translation helpful? Give feedback.
All reactions