Comments on: Socorro: Mozilla’s Crash Reporting System http://www.laurathomson.com/2010/05/socorro-mozillas-crash-reporting-system/ Laura Thomson's random thoughts and rants about tech and FOSS Wed, 8 Feb 2012 04:09:48 +0000 http://wordpress.org/?v=2.5 By: Aaron Johnson – Links: 6-1-2010 http://www.laurathomson.com/2010/05/socorro-mozillas-crash-reporting-system/#comment-777 Aaron Johnson – Links: 6-1-2010 Wed, 02 Jun 2010 08:00:34 +0000 http://www.laurathomson.com/?p=124#comment-777 [...] tech ramblings » Blog Archive » Socorro: Mozilla’s Crash Reporting System Mind boggling # of crashes that Mozilla sees on a weekly basis and a cool story about how they’re going to evolve the system to handle more data. (categories: bigdata mozilla hadoop hdfs storage debugging ) [...] […] tech ramblings » Blog Archive » Socorro: Mozilla’s Crash Reporting System Mind boggling # of crashes that Mozilla sees on a weekly basis and a cool story about how they’re going to evolve the system to handle more data. (categories: bigdata mozilla hadoop hdfs storage debugging ) […]

]]> By: admin http://www.laurathomson.com/2010/05/socorro-mozillas-crash-reporting-system/#comment-732 admin Thu, 27 May 2010 17:36:18 +0000 http://www.laurathomson.com/?p=124#comment-732 I almost forgot: 3.6.3 crashes are throttled to 15%, and since 3.6.4 is a beta, we get 100% of crashes. I almost forgot: 3.6.3 crashes are throttled to 15%, and since 3.6.4 is a beta, we get 100% of crashes.

]]> By: admin http://www.laurathomson.com/2010/05/socorro-mozillas-crash-reporting-system/#comment-727 admin Thu, 27 May 2010 15:15:59 +0000 http://www.laurathomson.com/?p=124#comment-727 >* It doesn’t look like it’s possible to compare 3.6.3 with 3.6.4 by showing both graphs at the same time. Any plans for that? >* Why is 3.6.4 listed as a current release when it hasn’t been released yet? Right now, the releases featured and graphed on the homepage vs each other, as well as the current releases, are admin configurable (and generally altered by various members of the Firefox and Platform teams as needed). We haven't had a request for more graph flexibility that I am aware of (either in existing bugs or in the requirements gathering for 2.0). In terms of "current releases" this is really "current releases that are submitting crashes". The ones that usually get the most attention are the current generally available release, plus whatever builds are being worked on at any given time. Right now that's 3.6.4 since it is close to release, and the two alpha releases. It's a little bit different from our end user aimed sites; the main consumers of crash-stats are our own engineers and metrics people. The use cases for it tend to be around finding new crashes, trying to reproduce crashes, and diagnosing crashes. >* If I compare the crashes per ADUs on 3.6.3 and 3.6.4, it looks like 3.6.4 crashes 10x more often! 3.6.3 is about 0.25 and 3.6.4 is around 3.0. I assume this is wrong? The numbers should be right. Note there is a bit of distortion from 3.6.3 -> 3.6.4, so it's not apples to apples. In 3.6.4, we have OOPP support, so crash reports don't necessarily represent a crash to the *end user*. If Flash hangs, a pair of crash reports are generated and submitted, one for the plugin process, and one for the browser process. At present the plugins covered by OOPP represent somewhere in the vicinity of 35-40% of crashes, so there will be that many more reports in the system, and that many less apparent crashes from the user's perspective. The other distortion 3.6.4-wise is that it's not in general release yet, so there are a smaller number of users, and they are likely pushing on it more (testing, QA, and so on). I hope this helps. >* It doesn’t look like it’s possible to compare 3.6.3 with 3.6.4 by showing both graphs at the same time. Any plans for that?
>* Why is 3.6.4 listed as a current release when it hasn’t been released yet?

Right now, the releases featured and graphed on the homepage vs each other, as well as the current releases, are admin configurable (and generally altered by various members of the Firefox and Platform teams as needed). We haven’t had a request for more graph flexibility that I am aware of (either in existing bugs or in the requirements gathering for 2.0).

In terms of “current releases” this is really “current releases that are submitting crashes”. The ones that usually get the most attention are the current generally available release, plus whatever builds are being worked on at any given time. Right now that’s 3.6.4 since it is close to release, and the two alpha releases.

It’s a little bit different from our end user aimed sites; the main consumers of crash-stats are our own engineers and metrics people. The use cases for it tend to be around finding new crashes, trying to reproduce crashes, and diagnosing crashes.

>* If I compare the crashes per ADUs on 3.6.3 and 3.6.4, it looks like 3.6.4 crashes 10x more often! 3.6.3 is about 0.25 and 3.6.4 is around 3.0. I assume this is wrong?

The numbers should be right. Note there is a bit of distortion from 3.6.3 -> 3.6.4, so it’s not apples to apples. In 3.6.4, we have OOPP support, so crash reports don’t necessarily represent a crash to the *end user*. If Flash hangs, a pair of crash reports are generated and submitted, one for the plugin process, and one for the browser process. At present the plugins covered by OOPP represent somewhere in the vicinity of 35-40% of crashes, so there will be that many more reports in the system, and that many less apparent crashes from the user’s perspective. The other distortion 3.6.4-wise is that it’s not in general release yet, so there are a smaller number of users, and they are likely pushing on it more (testing, QA, and so on). I hope this helps.

]]> By: David Tenser http://www.laurathomson.com/2010/05/socorro-mozillas-crash-reporting-system/#comment-714 David Tenser Thu, 27 May 2010 09:47:06 +0000 http://www.laurathomson.com/?p=124#comment-714 Some questions: * It doesn't look like it's possible to compare 3.6.3 with 3.6.4 by showing both graphs at the same time. Any plans for that? * Why is 3.6.4 listed as a current release when it hasn't been released yet? * If I compare the crashes per ADUs on 3.6.3 and 3.6.4, it looks like 3.6.4 crashes 10x more often! 3.6.3 is about 0.25 and 3.6.4 is around 3.0. I assume this is wrong? Some questions:

* It doesn’t look like it’s possible to compare 3.6.3 with 3.6.4 by showing both graphs at the same time. Any plans for that?
* Why is 3.6.4 listed as a current release when it hasn’t been released yet?
* If I compare the crashes per ADUs on 3.6.3 and 3.6.4, it looks like 3.6.4 crashes 10x more often! 3.6.3 is about 0.25 and 3.6.4 is around 3.0. I assume this is wrong?

]]> By: vernon http://www.laurathomson.com/2010/05/socorro-mozillas-crash-reporting-system/#comment-646 vernon Wed, 26 May 2010 18:23:20 +0000 http://www.laurathomson.com/?p=124#comment-646 Yeah, crashes have definitely decreased. 3.6 is a lot stable. Nice work.. Yeah, crashes have definitely decreased. 3.6 is a lot stable. Nice work..

]]>