1232440 - [prod][stage] /crashes-per-day/ for B2G returns a 500 error

Status: VERIFIED INVALID

(Socorro :: General, task)

Description

[Matt Brandt [:mbrandt]]

Several of the /crashes-per-user/ reports return a 500 error:

https://crash-stats.mozilla.com/crashes-per-user/?p=B2G
https://crash-stats.mozilla.com/crashes-per-user/?p=SeaMonkey

Stacktrace (most recent call last): KeyError
  File "django/core/handlers/base.py", line 111, in get_response
    response = wrapped_callback(request, *callback_args, **callback_kwargs)
  File "newrelic/hooks/framework_django.py", line 499, in wrapper
    return wrapped(*args, **kwargs)
  File "crashstats/crashstats/decorators.py", line 66, in inner
    return view(request, *args, **kwargs)
  File "crashstats/crashstats/views.py", line 1292, in crashes_per_user
    adi_groups = adi_by_version[term]

https://sentry.prod.mozaws.net/operations/socorro-prod/group/157995/#exception

Comment 1

[Matt Brandt [:mbrandt]]

Seamonkey no longer throws an error, however B2G still returns a 500 error.

Comment 2

[Peter Bengtsson [:peterbe]]

The issue is that B2G doesn't have ADI so it doesn't make sense to run that report for B2G. 
People who work on B2G are aware of this. I'm not entirely sure how to deduce which products we have and don't have ADI for. If we knew that, we'd leverage that to disable the Crashes per Day report for those products. 

All in all, it's really a low priority. If it's in your automated QA, I suggest you remove it.

Comment 3

[Matt Brandt [:mbrandt]]

Should we leave this bug open even though this is low priority? A 500 error seems inappropriate.

Comment 4

[Matt Brandt [:mbrandt]]

I should also point out that the B2G report doesn't return a 500 error on stage. 

https://crash-stats.allizom.org/crashes-per-day/?p=B2G

Comment 5

[Robert Kaiser]

In the future, we should end up only reporting pure crash counts (and graphs of those) for products that do not have ADI.

Comment 6

[Peter Bengtsson [:peterbe]]

(In reply to Matt Brandt [:mbrandt] from comment #4)
> I should also point out that the B2G report doesn't return a 500 error on
> stage. 
> 
> https://crash-stats.allizom.org/crashes-per-day/?p=B2G

That's because it has 0 crashes on stage's ES server. Then it doesn't get itself into trouble combining number of crashes with ADI.

Comment 7

[Peter Bengtsson [:peterbe]]

(In reply to Robert Kaiser (:[e-mail]) from comment #5)
> In the future, we should end up only reporting pure crash counts (and graphs
> of those) for products that do not have ADI.

Please take a moment to convince me. 
It just seems out of whack to retrofit B2G into a report that is VERY much about ADI. 

It's like having a sports report called "Tackles per Game" and then include a team that tackles in practice but never plays games.

Comment 8

[Peter Bengtsson [:peterbe]]

How about solving it by doing whatever-it-takes to get ADI numbers for B2G. Even if they have to be guesstimated.

Comment 9

[Robert Kaiser]

(In reply to Peter Bengtsson [:peterbe] from comment #8)
> How about solving it by doing whatever-it-takes to get ADI numbers for B2G.
> Even if they have to be guesstimated.

We will not get ADI for B2G, at least any time in the short- or mid-term future. And non-Mozilla users of Socorro probably would like to see graphs for their products without ADI as well, I guess.

Comment 10

[Peter Bengtsson [:peterbe]]

(In reply to Robert Kaiser (:[e-mail]) from comment #9)
> (In reply to Peter Bengtsson [:peterbe] from comment #8)
> > How about solving it by doing whatever-it-takes to get ADI numbers for B2G.
> > Even if they have to be guesstimated.
> 
> We will not get ADI for B2G, at least any time in the short- or mid-term
> future. And non-Mozilla users of Socorro probably would like to see graphs
> for their products without ADI as well, I guess.

There are, or should be, other ways to see graphs. I don't want to shoehorn a report unnaturally. Especially not to please the few non-Mozilla users we have (the ones we have rarely use the website). 

But if the B2G team really are starving for this report to work, we could just introduce a hack so that the graph isn't ratios between ADI and crashes, but just a plain number per day. And the table the bottom could simply be removed if there is no ADI counts.

Comment 11

[Robert Kaiser]

For one thing there's no "the B2G team" right now, as they are reforming themselves into a collection of small teams. For the other, they are surely not starving for this as they are far away from shipping products.

Comment 12

[Peter Bengtsson [:peterbe]]

Then I'm confident we should see if we can make B2G be removed from the drop-down of products (because it doesn't have ADI). 
And if you get there by changing the `?p=Firefox` to `?p=B2G` we'll simply say "Sorry". 

It's not worth pursuing this any further. Let's work on stuff that matters that people need.

Comment 13

[Matt Brandt [:mbrandt]]

Updating to include stage - stage is hitting the error state that prod is experiencing (see comment #6).

Updating our e2e to reflect this with an xfail for stage.

Comment 14

[Peter Bengtsson [:peterbe]]

B2G not a product any more.

Comment 15

[Matt Brandt [:mbrandt]]

Bumping to verified fixed - both prod and stage no longer have a b2g product