Re: [DISCUSS][SUMMER-2020] Modify the search rules of task instances

[JUN GAO <ga...@gmail.com>]

I think it is a good idea

wang <ww...@163.com>于2020年8月16日 周日10:31写道：

> Because of the addition of new features, the task will not have a start
> time before it starts executing. In the original design, the search rule
> for task instances mainly used the start time. For example, the following
> methods in TaskInstanceMapper:
>
>
>
>
> 1. findValidTaskListByProcessId
>
> 2. queryTaskInstanceListPaging
>
> 3. countTaskInstanceStateByUser
>
>
>
>
> For the first two methods, the user experience is not obvious when there
> are fewer task instances. But when there are many instances, the task
> instances without start time are always on the last page.
>
>
>
>
> As for the last method, because it uses whether the start time is a
> standard within a certain period of time, task instances without a start
> time will never be counted. At this time, the user will find that the
> number of tasks in the delayed execution state is always zero.
>
>
>
>
> For the above two reasons, I want to change the above three methods from
> using the start time to using the submit time. My question about this is,
> does such a change have much impact on the system?
>
>
>
>
> ### 修改任务实例的查找规则
>
>
>
> 因为新的特性的加入，导致任务在开始执行前不会拥有开始时间。而在原来的设计中，任务实例的查找规则主要使用了开始时间。例如TaskInstanceMapper中的以下几个方法：
>
>
> 1. findValidTaskListByProcessId
> 2. queryTaskInstanceListPaging
> 3. countTaskInstanceStateByUser
>
>
> 对于前两个方法，在任务实例较少的时候用户感受并不明显。但是实例较多的时候，没有开始时间的任务实例总是在最后一页。
>
>
>
> 而对于最后一个方法，因为是使用开始时间是否在某一个时间段内为标准进行查找的，没有开始时间的任务实例永远不会被统计到。这时用户就会发现处于延迟执行状态的任务数量总是0。
>
>
> 由于以上两个原因，我想把以上三个方法从使用开始时间变为使用提交时间。对此我的疑问是，这样的改动对系统影响大不大？

-- 

DolphinScheduler(Incubator)  PPMC
Jun Gao 高俊
gaojun2048@gmail.com