Closed Bug 792082 Opened 12 years ago Closed 12 years ago

Ignore browser-hang reports in the collector

Categories

(Socorro :: Backend, task)

x86_64
Linux
task
Not set
normal

Tracking

(Not tracked)

VERIFIED FIXED

People

(Reporter: benjamin, Assigned: lars)

References

Details

Attachments

(1 file)

Browser hang reports are currently worthless and in order to work on the future hang reports we should stop collecting them altogether.
Summary: Ignore browser-hang reports → Ignore browser-hang reports in the collector
https://github.com/mozilla/socorro/pull/837 is the patch. It's untested because I can't get a working vagrant setup, so I'll need somebody (lars?) to test it before pulling.
Assignee: benjamin → nobody
Lars, this needs to be in prod by November 19.  Please review and land.
Assignee: nobody → lars
Target Milestone: --- → 27
bsmedberg:  I'd like to propose an alternative implementation for this.  Rather than hard coding the behavior into the collector, I'd like to have this situation handled by the throttler.  A rule like this would be appropriate:

  ('*', lambda x: 'HangID' in x and x.get('ProcessType', 'browser'), 0)
  
  # [1]: '*' means use free form instead of specific single field
  # [2]: predicate
  # [3]: 0% pass through of crashes matching the predicate

the only difference would be that the collector would return "Discarded=1\n" instead of "Unsupported=1\n"

would that be acceptable?
I don't know, which is why I didn't implement it that way. It's possible that "Discarded" might indicate to the client that it shouldn't submit any more dumps, which is not what we want.

Is it possible to chage the throttling rules slightly so that they can return something other than Discarded?
yes, I can make that happen.  I'll get back to you with a proposal that returns 'Unsupported=\n' using the Throttler...
Commits pushed to master at https://github.com/mozilla/socorro

https://github.com/mozilla/socorro/commit/d4a65ec8a799d4afbb809136787b40ae850c5930
Bug 792082 - Ignore the browser side of plugin-hang reports

https://github.com/mozilla/socorro/commit/8eb46a383c4727c5c851f6d60ea1539b26b87767
Merge pull request #917 from twobraids/hangignore2012

fixes Bug 792082 - Ignore the browser side of plugin-hang reports
Status: NEW → RESOLVED
Closed: 12 years ago
Resolution: --- → FIXED
sample crash that when submitted should be rejected by the collector with a return value of "Unsupported=1\n"
this verified working on dev
staging deployment and testing pending.
sample crash attached to this bug
Depends on: 810241
No longer depends on: 810241
Needs to be in 26
Target Milestone: 27 → 26
(In reply to K Lars Lohn [:lars] [:klohn] from comment #8)
> this verified working on dev
> staging deployment and testing pending.
> sample crash attached to this bug

* works on staging - "Unsupported=1" is returned, if I remove "HangID:" from JSON then crash is accepted (CrashID=bp-39d72aaa-8f8c-4769-a2c9-6710f2121114)
* took known good crash from prod https://crash-stats.mozilla.com/report/index/ba5e807e-1b32-4466-9093-cc17d2121113 and resubmitted to stage (CrashID=bp-ac7eddd2-a0ea-4093-ba6e-3b4fc2121114)
(In reply to Robert Helmer [:rhelmer] from comment #10)
> (In reply to K Lars Lohn [:lars] [:klohn] from comment #8)
> > this verified working on dev
> > staging deployment and testing pending.
> > sample crash attached to this bug
> 
> * works on staging - "Unsupported=1" is returned, if I remove "HangID:" from
> JSON then crash is accepted (CrashID=bp-39d72aaa-8f8c-4769-a2c9-6710f2121114)
> * took known good crash from prod
> https://crash-stats.mozilla.com/report/index/ba5e807e-1b32-4466-9093-
> cc17d2121113 and resubmitted to stage
> (CrashID=bp-ac7eddd2-a0ea-4093-ba6e-3b4fc2121114)

At lars' suggestion, added "ProcessType: plugin" to attachment 678802 [details] and crash is accepted to stage:
CrashID=bp-e8f837c8-0363-4dc3-aacc-f20df2121114
Bumping to QA verified on stage - rhelmer thanks for helping! I couldn't have easily done this without your help.

Both the positive and negative test cases pass -
For the positive "Unsupported=1" is returned
For the negative (ProcessType is missing from the raw crash and one where the ProcessType has a value of something other than "browser"), are accepted.
Status: RESOLVED → VERIFIED
Repeated the tests fro comment 10 and comment 11 on prod, lgtm.
Also, stephend generated a normal crash using crashmenow and looks as expected:
https://crash-stats.mozilla.com/report/index/bp-82d76823-a642-4bc2-bc5c-896f12121115
Blocks: 812511
You need to log in before you can comment on or make changes to this bug.

Attachment

General

Created:
Updated:
Size: