> On Aug. 25, 2015, 6:42 p.m., David McLaughlin wrote:
> > What if they click on an old update and the instance page doesn't reflect 
> > the change made in this instance event? Do we care?
> 
> Joshua Cohen wrote:
>     I don't think we care, there's not much we can do in that case, is there? 
> The only thing I can think of is disabling links for completed updates, but 
> that seems overly broad (some completed updates will link to instances that 
> still exist).
> 
> David McLaughlin wrote:
>     Right. I'm just concerned people click on the link to see what happened 
> on that instance event. I know it's what I would expect, given it's the only 
> link and the first column on the row too. 
>     
>     Basically it comes down - the instance page you added is a 'live' view of 
> that instance. It is potentially misleading to include that link on a table 
> that deals exclusively with historical data.
> 
> Joshua Cohen wrote:
>     Yeah, I understand that, I'm just not sure what the alternative is other 
> than not linking at all (which seems worse to me)? As far as I can see 
> there's no association between an instance in an update and an actual task id 
> (which would let us query to see if the scheduler still has a record of that 
> task existing before displaying the link). That said, even if we *could* 
> conditionally display the link, that might be even more confusing as it would 
> be feasible that only *some* tasks from a historical update have been purged, 
> while others might still remain, leading to a strange inconsistency on the 
> update page where only some instances are links, while others are not.
>     
>     As far as I see it we have three options:
>     
>     1) Always link to the instance page.
>     2) Only link to the instance page for active updates.
>     3) Never link to the instance page.
>     
>     Option 1 seems like the best option in that it provides an easy way to 
> see what happened for an update, and in most likely cases (debugging an 
> active or recently completed update) should provide useful data (though 
> admittedly could prove to be confusing in the cases where the task that was 
> part of an update has already been pruned).
> 
> David McLaughlin wrote:
>     I'm leaning towards (3) but maybe that's because I don't understand where 
> the requirement comes from. The linked ticket suggests the user story is "I 
> see an instance event with a 'bad' event status and I want to click straight 
> through to see why.' I see the main use case for this occuring when you have 
> a bad deploy, but if rollbacks are enabled you'd click the event and get 
> taken to the rolled back instance page. This could be frustrating at best and 
> misleading/confusing at worst.
>     
>     Happy to be overruled by a tie-breaking vote here.

| but if rollbacks are enabled you'd click the event and get taken to the 
rolled back instance page.

The instance page aggregates task history by instance ID, right? This is a 
perfect example when having instance history is actually quite useful when 
debugging a rollback.


- Maxim


-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/37761/#review96394
-----------------------------------------------------------


On Aug. 25, 2015, 3:01 p.m., Joshua Cohen wrote:
> 
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/37761/
> -----------------------------------------------------------
> 
> (Updated Aug. 25, 2015, 3:01 p.m.)
> 
> 
> Review request for Aurora, David McLaughlin and Zameer Manji.
> 
> 
> Bugs: AURORA-1331
>     https://issues.apache.org/jira/browse/AURORA-1331
> 
> 
> Repository: aurora
> 
> 
> Description
> -------
> 
> Add a link to the instance page from instance events on the update page.
> 
> 
> Diffs
> -----
> 
>   src/main/resources/scheduler/assets/update.html 
> 3750aab342e326fc34d254dbfce791da0ca05ec6 
> 
> Diff: https://reviews.apache.org/r/37761/diff/
> 
> 
> Testing
> -------
> 
> See screenshot.
> 
> 
> File Attachments
> ----------------
> 
> Look, the instance numbers are blue, because they're links!
>   
> https://reviews.apache.org/media/uploaded/files/2015/08/25/b0dc6715-be1a-4a81-992f-caf2efd847a6__Screen_Shot_2015-08-25_at_9.59.50_AM.png
> 
> 
> Thanks,
> 
> Joshua Cohen
> 
>

Reply via email to