525785 - Integrate correlation reports into Socorro UI

Status: RESOLVED FIXED

(Socorro :: General, task)

Description

[Michael Morgan [:morgamic]]

David did some fine work here:
http://dbaron.org/mozilla/topcrash-modules

We need to find a way to automate these then integrate them into the existing reporter UI.  This is a meta bug, as many lesser bugs will be spawned to track the effort.

Marking this for 1.3, but dependent bugs will fall into earlier milestones.

Comment 1

[Michael Morgan [:morgamic]]

Austin is going to attempt to integrate the text output with the Socorro UI.

Comment 2

[Austin King [:ozten]]

Bumping to 1.4

Comment 3

[chris hofmann]

some thoughts on integration...

the best place to put all this correlation information is back into the individual reports on some correlation tab.

that way we can look at a crash report and see how it relates to other crashes of the same signature as quickly as possible.  here is an example.

---------------------------------------------------

the crash might have correlation to following areas

  addons in use
   ....
   ....
   ....
   (shows if any addons might be contributing)

 modules (and module versions) in use
   .... 
   ....
   ....
   (shows if any plugings might be contributing)

 os versions in use
   ....
   ....
   ....
   (shows which OSes are in play or not to help reproduce)

 number of cores
   ....
   ....
   ....
   (shows if the problem might be a threading bug)

 other versions of firefox
   ....
   ....
   ....
   ....
   (shows which versions are affected)
    jst's reports could also be integrated into this section)
     http://people.mozilla.com/~jst/new-crashes/Firefox/latest/

 correlation to start up
   x reports inside 3 minutes of start
   y reports inside 5 minutes
   z reports greater than 5 minutes after start up.
   (shows if its user or configuration problem)

 correlatations to domains of sites where the crash is observered
   x reports of this crash happended while the user was on facebook
   y reports happened while user was on gmail.
   z reports happened on random porn site...
    (shows if its a general browsing crash or associated with specific kind of content)

Comment 4

[Austin King [:ozten]]

Attachment: First stab at integrating Correlation reports

You can test this patch at 
http://aking.khan.mozilla.org/reporter/report/index/f9a3aeb0-b0ae-439d-82c5-73b2f2091228#modver

In your own sandbox, match up a current crash on filesystem with the product / version being generated at http://people.mozilla.com/crash_analysis/ for the current day.

Binary files for unit tests can be copied from my sandbox, since they aren't part of the patch.

Comment 5

[chris hofmann]

http://aking.khan.mozilla.org/reporter/report/index/f9a3aeb0-b0ae-439d-82c5-73b2f2091228#modver  looks pretty good.

the modules and addons could have twisties to expand and collapse more details on versions for each item.  

that's kind of the way I use the *-with-versions.txt file as a secondary indicator once an interesting correlation has been drawn to a specific module or addon..

here is a an example of what it might look like

+ 99% (6026/6058) vs.   6% (6245/102490) overlapp32.dll
- 66% (4010/6058) vs.  20% (20236/102490) MSCTFIME.IME
         40% (2423/6058) vs.   0% (110/87631) 5.1.2600.1106
         21% (1272/6058) vs.   7% (6445/87631) 5.1.2600.2180
         19% (1151/6058) vs.   8% (6701/87631) 5.1.2600.5512
          0% (0/6058)    vs.   0% (6/87631) 5.1.2600.2055
          0% (0/6058)    vs.   0% (2/87631) 5.1.2600.2096
          0% (0/6058)    vs.   0% (1/87631) 5.1.2600.2135

+ 79% (4773/6058) vs.  33% (33464/102490) MSCTF.dll
+ 100% (6052/6058) vs.  71% (73170/102490) hnetcfg.dll
+ 100% (6034/6058) vs.  71% (72905/102490) wldap32.dll
+ 100% (6054/6058) vs.  72% (74270/102490) iphlpapi.dll
+ 100% (6053/6058) vs.  73% (74584/102490) wshtcpip.dll
+ 100% (6033/6058) vs.  73% (75003/102490) comres.dll
+ 95% (5733/6058) vs.  68% (70194/102490) xpsp2res.dll
96% (5792/6058) vs.  70% (71583/102490) lz32.dll
100% (6055/6058) vs.  76% (78119/102490) ws2help.dll
100% (6058/6058) vs.  77% (79290/102490) imagehlp.dll
100% (6058/6058) vs.  88% (90352/102490) crypt32.dll
100% (6058/6058) vs.  88% (90363/102490) msasn1.dll
11% (645/6058) vs.   1% (838/102490) correct.dll
100% (6058/6058) vs.  91% (92979/102490) wininet.dll
9% (529/6058) vs.   1% (1249/102490) pshook.dll
7% (428/6058) vs.   1% (560/102490) LHook.dll
99% (6019/6058) vs.  93% (95760/102490) dnsapi.dll
9% (529/6058) vs.   3% (3020/102490) pr_imon.dll
100% (6056/6058) vs.  94% (96561/102490) brwsrcmp.dll
99% (6012/6058) vs.  94% (95878/102490) winrnr.dll
10% (592/6058) vs.   4% (4324/102490) imon.dll
72% (4380/6058) vs.  67% (68972/102490) normaliz.dll

Comment 6

[Austin King [:ozten]]

(In reply to comment #5)
I've broken out the "module: area from the "module with versions" area, because matching them up is non-trivial.

Same with addon and addon with versions.

Once we built this out properly (cloud), twisties would be easy to add.
"Works Sometimes"
For now, adding it would make twisties work sometimes and not others. This is because we load many more signatures from "modules" than we do from "modules with versions".

"Bad Performance"
For a major release, when you click the twisty for the first time, then the backend downloads a 40MB file, pulls it apart and finds the couple lines of text needed to populate the data.

So what will happen is the user will get board waiting for this to load and then want to click another twisty. Unless we build a queue in JavaScript, we'll spawn another request for the same 40MB file...

How important is this. I can think more about a work around if it's high priority. I did want to do this originally, but skipped it as this was supposed to be a quick integration and "not building a Ferrari".

Comment 7

[Ryan Snyder [:ryansnyder] [:rsnyder] [:rysny]]

Comment on attachment 421714 [details] [diff] [review]
First stab at integrating Correlation reports

This is a really solid first effort.  Well done!  Most of the issues I've found are minor and some just plain nitpicky.

==

Correllation_Controller:

Add Mozilla MPL license info to to of file.

$cache = new Cache( Kohana::config('correlation.caching'); -- missing ending parenthesis.

The base URL for "http://people.mozilla.com/crash_analysis/" should be moved into a config file.

Add @return documentation for _getCacheKey() and _getGeneralCacheKey()

In the method _ajax(), you're using var_dump($data)...  Is there a reason for that?  I'm assuming you just forgot to remove it...

Correllation_Library:

Add Mozilla MPL license info to to of file.

handleError() is duplicated in CrashReportDump.php.  Let's only use one method in our code base.  Perhaps we should move it to a library of helper functions or include it in MY_Controller?

Add @return documentation to getTxt()

Comment 8

[Austin King [:ozten]]

Attachment: Seconds stab at integrating Correlation reports

Okay, I've made all of those changes. Nice catches!
I don't think pulling handle error our is that big of a deal, since it's a 2 line function and each class may want to specialize the behavior down the road. That said, I went ahead and made the change, creating an ErrorHandler class for it to live in.
Thanks for your help.

Comment 9

[Ryan Snyder [:ryansnyder] [:rsnyder] [:rysny]]

Comment on attachment 421930 [details] [diff] [review]
Seconds stab at integrating Correlation reports

Looks good!

Comment 10

[Austin King [:ozten]]

(In reply to comment #9)
Thanks!

Adding         webapp-php/application/config/correlation.php
Adding         webapp-php/application/controllers/correlation.php
Sending        webapp-php/application/controllers/report.php
Adding         webapp-php/application/libraries/Correlation.php
Sending        webapp-php/application/libraries/CrashReportDump.php
Adding         webapp-php/application/libraries/ErrorHandler.php
Sending        webapp-php/application/views/report/index.php
Adding         webapp-php/js/socorro/report.js
Adding         webapp-php/tests/CorrelationTest.php
Adding         webapp-php/tests/data
Adding         webapp-php/tests/data/CorrolatedModules
Adding  (bin)  webapp-php/tests/data/CorrolatedModules/20091216_Firefox_3.5.5-core-counts.txt.gz
Adding  (bin)  webapp-php/tests/data/CorrolatedModules/20091216_Firefox_3.5.5-interesting-addons-with-versions.txt.gz
Adding  (bin)  webapp-php/tests/data/CorrolatedModules/20091216_Firefox_3.5.5-interesting-addons.txt.gz
Adding  (bin)  webapp-php/tests/data/CorrolatedModules/20091216_Firefox_3.5.5-interesting-modules-with-versions.txt.gz
Adding  (bin)  webapp-php/tests/data/CorrolatedModules/20091216_Firefox_3.5.5-interesting-modules.txt.gz
Adding         webapp-php/tests/data/CorrolatedModules/20091216_Thunderbird_3.0b4-core-counts.txt
Adding         webapp-php/tests/data/CorrolatedModules/20091216_Thunderbird_3.0b4-interesting-addons-with-versions.txt
Adding         webapp-php/tests/data/CorrolatedModules/20091216_Thunderbird_3.0b4-interesting-addons.txt
Adding         webapp-php/tests/data/CorrolatedModules/20091216_Thunderbird_3.0b4-interesting-modules-with-versions.txt
Adding         webapp-php/tests/data/CorrolatedModules/20091216_Thunderbird_3.0b4-interesting-modules.txt
Transmitting file data ...................
Committed revision 1703.