Google Groups no longer supports new Usenet posts or subscriptions. Historical content remains viewable.
Dismiss

How do you triage website crash bugs?

1 view
Skip to first unread message

Maris Fogels

unread,
May 30, 2014, 12:15:21 PM5/30/14
to mozilla-d...@lists.mozilla.org
Hi everyone,

I'm working to standardize the way we report website crash bugs on MDN,
and I was wondering what other Mozilla web properties do to triage their
crashes (HTTP 5XX errors). If possible, I'd like to reuse practices
other Mozilla website projects for MDN.

Here are some specific questions a bug triager might ask that I'd like
to standardize on our site:

* How do you indicate that a bug involves a server-side error?
* How do you indicate if a website crash bug is reproducible or not?
* How do you indicate if a website crash bug has a stacktrace?
* How do you indicate that a website crash bug is 'complete', ready
for a developer to investigate?

How do you triage website crash bugs like these in your project?


Thanks,

-mars

Wil Clouser

unread,
May 30, 2014, 3:45:11 PM5/30/14
to Maris Fogels, mozilla-d...@lists.mozilla.org
The Marketplace doesn't do any specific indicators on bugs for a server-side
error vs anything else.

When a 5xx error is hit, the details, including the stacktrace are logged to
Sentry. When a bug is filed that says "Oops" or specifies a 500 error someone
(myself, a developer, or QA) will generally log into Sentry and find the
traceback and either fix the bug or paste the traceback into the bug.

We don't run any stats on 5xx errors vs any other kind of error.

Wil

Andy McKay

unread,
May 30, 2014, 4:05:07 PM5/30/14
to Wil Clouser, mozilla-d...@lists.mozilla.org, Maris Fogels

> We don't run any stats on 5xx errors vs any other kind of error.

We do have the graphite stats which allows us to spot if a large number of 5xx (or any other status codes) are occurring.

Andy

Christopher Lonnen

unread,
Jun 7, 2014, 5:19:15 PM6/7/14
to Maris Fogels, mozilla-d...@lists.mozilla.org
On crash-stats and other WebEng websites we normally file the bug with a link to Sentry, which also gives us how often the crash is happening, a whole lot of metadata, and the ability to replay the request generating the crash. NewRelic has similar functionality. The only bugzilla feature we use to provide metadata about the bug is component. This doesn't tell us much, though, since 5xx's tend to all get filed "webapp". Blocking bugs are filed later if the error originates in another component.

----- Original Message -----
> From: "Maris Fogels" <ma...@mozilla.com>
> To: mozilla-d...@lists.mozilla.org
> Sent: Friday, May 30, 2014 9:15:21 AM
> Subject: [webdev] How do you triage website crash bugs?
>
> Hi everyone,
>
> I'm working to standardize the way we report website crash bugs on MDN,
> and I was wondering what other Mozilla web properties do to triage their
> crashes (HTTP 5XX errors). If possible, I'd like to reuse practices
> other Mozilla website projects for MDN.
>
> Here are some specific questions a bug triager might ask that I'd like
> to standardize on our site:
>
> * How do you indicate that a bug involves a server-side error?
> * How do you indicate if a website crash bug is reproducible or not?
> * How do you indicate if a website crash bug has a stacktrace?
> * How do you indicate that a website crash bug is 'complete', ready
> for a developer to investigate?
>
> How do you triage website crash bugs like these in your project?
>
>
> Thanks,
>
> -mars
> _______________________________________________
> dev-webdev mailing list
> dev-w...@lists.mozilla.org
> https://lists.mozilla.org/listinfo/dev-webdev
>
0 new messages