422549 - dup crash reports

Status: RESOLVED WONTFIX

(Socorro :: General, task)

Description

[chris hofmann]

we have seen this with other crash reporting systems, and it seams now to show up in sororro.

here are some what look to be repeated blackbox reports/entries with the identical stack signatures, identical OS and system info, and identical user comments, with back to back submissions at 10:28 and then a couple more submissions 7 minutes later.

http://crash-stats.mozilla.com/report/index/817ff9f0-f016-11dc-a85f-001a4bd43ed6
UUID 817ff9f0-f016-11dc-a85f-001a4bd43ed6
Time	2008-03-12 03:27:15-07:00
2008-03-12 10:35   	Firefox  	3.0b4  	2008030318  	Linux 0.0.0 Linux 2.6.21.7 #1 Sun Sep 9 02:29:54 GMT-8 2007 i686 GNU/Linux  	x86  	SIGSEGV  	0xb75b2096  	Add a comment. Note: Comments are publicly visible

http://crash-stats.mozilla.com/report/index/7f581c17-f016-11dc-a95b-001a4bd43ed6
UUID 7f581c17-f016-11dc-a95b-001a4bd43ed6
Time	2008-03-12 03:27:02-07:00
2008-03-12 10:35 	Firefox 	3.0b4 	2008030318 	Linux 0.0.0 Linux 2.6.21.7 #1 Sun Sep 9 02:29:54 GMT-8 2007 i686 GNU/Linux 	x86 	SIGSEGV 	0xb75ae096 	Add a comment. Note: Comments are publicly visible

http://crash-stats.mozilla.com/report/index/373d7ae8-f016-11dc-917c-001a4bd43ed6
UUID	373d7ae8-f016-11dc-917c-001a4bd43ed6
Time	2008-03-12 03:25:10-07:00
2008-03-12 10:28 	Firefox 	3.0b4 	2008030318 	Linux 0.0.0 Linux 2.6.21.7 #1 Sun Sep 9 02:29:54 GMT-8 2007 i686 GNU/Linux 	x86 	SIGSEGV 	0xb75f9096 	Add a comment. Note: Comments are publicly visible

http://crash-stats.mozilla.com/report/index/34de7019-f016-11dc-8524-001a4bd43ef6
UUID	34de7019-f016-11dc-8524-001a4bd43ef6
Time	2008-03-12 03:24:59-07:00
2008-03-12 10:28 	Firefox 	3.0b4 	2008030318 	Linux 0.0.0 Linux 2.6.21.7 #1 Sun Sep 9 02:29:54 GMT-8 2007 i686 GNU/Linux 	x86 	SIGSEGV 	0xb763e096 	Add a comment. Note: Comments are publicly visible

We try to do some scanning of the db to confirm and detect if this is going on then purge dups if they appear.  if this is going on then the scan/purge should happen before we run nightly/hourly MTBF or other release stat reports.

Comment 1

[Samuel Sidler (old account; do not CC)]

You sure that's a dupe and not just someone crashing on startup? The comment is just the pre-filled text we use on all platforms. I just think the Linux client doesn't delete it onfocus.

Comment 2

[chris hofmann]

not absolutely sure, but the best way to detect and analyze this further is to do some automated scanning of the database to try and detect dups and watch incoming traffic from the same IP addresses.   do we/apache log any part of the transaction as we accept incoming bboxes?   if there is any evidence of retry attempts by the client on some closely connected or regular interval we might find it there too.

Comment 3

[(not currently active) Ted Mielczarek]

The client doesn't have any retry support, fwiw.

Comment 4

[(not currently active) Ted Mielczarek]

Also, this is exactly why we have a unique UserID field in the db. We should be able to take that into account to rule out individual clients that are particularly crashy.

I don't think we should be deleting any crash reports just because they've been submitted multiple times in a row. If someone has submitted a crash report, they may expect it to be there if they go looking for it.

Comment 5

[Michael Morgan [:morgamic]]

There are also unique user stats to help detect loops like this.