+1 for the feature. Looking at the schema of the log table, I think it's
perfect to store such kind of information.

On Fri, Oct 16, 2020 at 9:55 PM Daniel Imberman <[email protected]>
wrote:

> This could be pretty valuable for future audits. I’d personally rather
> avoid adding fields to the DB in general. Could we store it wherever we
> store normal failure messages?
>
> via Newton Mail
> <https://cloudmagic.com/k/d/mailapp?ct=dx&cv=10.0.51&pv=10.15.6&source=email_footer_2>
>
> On Fri, Oct 16, 2020 at 9:21 AM, Ace Haidrey
> <[email protected]> wrote:
>
> Hi team,
>
> My name is Ace, I’m a data engineer at Pinterest, and I wanted to get some
> feedback from the community on how they’d propose designing a feature that
> has been asked for us multiple times, that we’d like to implement in-house.
>
> The ask is the following:
> - Users want to be able to add an optional reason/description when they
> click on the Mark Success/Failed for both task level in graph views or dag
> level in tree views. This is to clearly state why the actions were taken on
> higher priority flows for when others are stepping in to look in it.
>
> For us the feature makes sense to have, and I was wondering if this is
> something the greater group would like to have upstream as well. And if so
> then we wanted to know what was preferred.
>
> 1. We can add this information in the UI for the confirmation view and an
> optional textbook to add details similar to how users can add conf
> variables when triggering a dag option is selected in the UI.
>
> Then after this, storing this information is the question. Do we want to
> store this information in the action logging extra blob and then add a view
> in the dag view where it has actions just pertaining to this dag and its
> tasks. (We plan to add this personally anyways for ourselves as going to
> the big list of actions for all dags and searching is not convenient
> always).
>
> Or do we want to add a new column in the task Instance model and dag run
> model to store the descriptions and retrieve the information from there.
>
> Tradeoffs here include the addition of a schema change, where the new
> column would generally be sparse, and the data is more resultant from an
> action vs necessarily being stored as part of the task run/dag run.
>
> Please let me know on any feedback, concerns, ideas, and more!
>
> Cheers,
> Ace
>
>

Reply via email to