909930 - Several products return 500 errors for the /crash_trends/ report

Status: RESOLVED FIXED

(Socorro :: General, task)

Description

[Matt Brandt [:mbrandt]]

https://crash-stats.mozilla.com/crash_trends/products/Fennec
https://errormill.mozilla.org/webtools/socorro-prod/group/95536/
Note, this is not a regression - with bug 907863 landing the visibility of these reports will increase.

https://crash-stats.mozilla.com/crash_trends/products/Thunderbird
https://errormill.mozilla.org/webtools/socorro-prod/group/95536/


https://crash-stats.mozilla.com/crash_trends/products/B2G
https://errormill.mozilla.org/webtools/socorro-prod/group/95536/

https://crash-stats.mozilla.com/crash_trends/products/WebappRuntimeMobile
https://errormill.mozilla.org/webtools/socorro-prod/group/95536/

Comment 1

[Brandon Savage [:brandon]]

Either one of us can look at this; I will be in France this week so I won't be here to look at this, but if you have time espressive feel free. Otherwise I'll fix it when I'm back.

Comment 3

[Matt Brandt [:mbrandt]]

Schalk - interested in taking this boog?

Comment 4

[Schalk Neethling [:espressive]]

(In reply to Matt Brandt [:mbrandt] from comment #3)
> Schalk - interested in taking this boog?

Matt,

So it looks like it has to do with the fact that crash trends only loads featured Nightly releases so, we need to either:

1) no just do featured
2) handle the fact that there are no featured nightly releases for a product better

looping in Laura and KaiRo for some additional feedback.

Comment 5

[Schalk Neethling [:espressive]]

3) do both ;)

So, the only reason I did not suggest this in my previous comment is, if we do 1, the problem will most likely go away but, with that said and thinking about it, it is probably best to be defensive in the code.

So, even if we do not do 1, 2 still applies. Thoughts?

Comment 6

[Robert Kaiser]

I'd prefer:

1) Use featured Nightly.
2) If none available, use highest-versioned active Nightly.
3) If none available, display a reasonable error message (and not just a 500).

Comment 7

[Robert Kaiser]

Oh, and for testing:
Firefox should always have a featured Nightly, so good for testing case 1.
Thunderbird should always have an active Nightly, but it's not featured - good for testing case 2.
Fennec has no active Nightly any more (has inactive ones but we should not use those), good for testing case 3.

Comment 8

[Peter Bengtsson [:peterbe]]

(In reply to Robert Kaiser (:[e-mail]) from comment #6)
> I'd prefer:
> 
> 1) Use featured Nightly.
> 2) If none available, use highest-versioned active Nightly.
> 3) If none available, display a reasonable error message (and not just a
> 500).

Point 2 is hard to parse. :)

Comment 9

[Robert Kaiser]

(In reply to Peter Bengtsson [:peterbe] from comment #8)
> Point 2 is hard to parse. :)

"highest-versioned active Nightly" == filter the active versions for this product for the Nightly channel only, and then take the one highest one in version-number-sort-order (i.e. the one that would come first in the versions dropdown).

Comment 10

[Schalk Neethling [:espressive]]

(In reply to Robert Kaiser (:[e-mail]) from comment #9)
> (In reply to Peter Bengtsson [:peterbe] from comment #8)
> > Point 2 is hard to parse. :)
> 
> "highest-versioned active Nightly" == filter the active versions for this
> product for the Nightly channel only, and then take the one highest one in
> version-number-sort-order (i.e. the one that would come first in the
> versions dropdown).

So, I think I have found an additional bug that we just have not ran into. The call that loads the versions for the selected product (after initial load) loads versions for both Nightly and Aurora versions of the products whilst the main crash_trends function currently, only looks for a featured Nightly.

It this correct or, should it only be loading Nightly versions for the product?

Comment 11

[Schalk Neethling [:espressive]]

(In reply to Robert Kaiser (:[e-mail]) from comment #9)
> (In reply to Peter Bengtsson [:peterbe] from comment #8)
> > Point 2 is hard to parse. :)
> 
> "highest-versioned active Nightly" == filter the active versions for this
> product for the Nightly channel only, and then take the one highest one in
> version-number-sort-order (i.e. the one that would come first in the
> versions dropdown).

To add to what I mentioned in comment 10, it seems that we actually consider Aurora to be part of the Nightlys or has that changed? If not, then this bug might have a different, simpler solution.

Comment 12

[Robert Kaiser]

crash_trends might make sense for Aurora as well, but its primary target is Nightly.

Comment 13

[Schalk Neethling [:espressive]]

https://github.com/mozilla/socorro/pull/1498

Comment 14

[Schalk Neethling [:espressive]]

The best and simplest way to test this, is to use the products from Comment 1 and ensure that these products no longer return a 500 error but now, the page loads successfully and there is a message provided to the user that there was no data found for the provided criteria.

Also, the versions will try to load but, if there are no Nightly versions at all, a message will also be shown above the form stating that no versions was found for the product.

Comment 15

[Robert Kaiser]

(In reply to Schalk Neethling [:espressive] from comment #14)
> The best and simplest way to test this, is to use the products from Comment
> 1 and ensure that these products no longer return a 500 error but now, the
> page loads successfully and there is a message provided to the user that
> there was no data found for the provided criteria.
> 
> Also, the versions will try to load but, if there are no Nightly versions at
> all, a message will also be shown above the form stating that no versions
> was found for the product.

That strategy sounds fine to me.

Comment 16

[Schalk Neethling [:espressive]]

Pull request https://github.com/mozilla/socorro/pull/1498 merged