[Home]

Summary:ASTERISK-14729: Abort/Raise from try_calling/queue_exec
Reporter:Raimund Sacherer (hatrix)Labels:
Date Opened:2009-08-28 06:55:25Date Closed:2011-06-07 14:00:47
Priority:CriticalRegression?No
Status:Closed/CompleteComponents:Applications/app_queue
Versions:Frequency of
Occurrence
Related
Issues:
Environment:Attachments:( 0) 2009-08-28T13:32+0200.txt
Description:Asterisk aborted with core-dump, see attached file.

We use Asterisk Stock debian Version (1.4.21.2) with one applied patch (see attached patch file).

We used to have asterisk 1.4.24, but had lot's of crashes and lockups, then we tried 1.4.26 rc3, rc5 and 1.4.26, we had lot's of crashes in app_queue, as well as different aborts.

Then we downgraded to 1.4.21.2 and we had more than one and a half week a stable system. With yesterday again one memory abort.

The setup is callcentersolution with ~ 60 agents.

Hopefulle I get a patch to apply soon as our client is very disencouraged because of the lot's of hickups we had in the past with asterisk.

Comments:By: David Brillert (aragon) 2009-08-28 10:03:05

I think is yet again another ticket related to 15109
You should try the patch from that ticket to see if it helps.

btw your bt is optimized out leaving not much details for the bug marshalls/developers.

You should choose DONT_OPTIMIZE from make menuselect when you are installing Aasterisk so your bt's are more useful.  There is no noticeable performance hit from the DONT_OPTIMIZE option.

By: David Woolley (davidw) 2009-08-28 11:20:56

I don't think it is ASTERISK-14129.  This one is an assertion failure in pthread_mutex_lock(), not an error in free().  It could be a memory management issue, but only indirectly.

There should have been a message on standard error, although that may not be going anywhere useful.

(I'm not sure the Digium people will be prepared to look at the patch, even though it is only for information.  For official patches, it is better to reference the official patch.  For other ones, it is better to describe what it does.)

By: Raimund Sacherer (hatrix) 2009-08-28 12:17:03

The added patch is only for information, it's one of the patches (i think russell) included in one of the recent crashes regarding app_queue.

It's just there so everybody who is skilled enough (eg, asterisk developer) knows that additional to standard stock debian patches this one is applied as well and I include it as well so a developer could exclude it as a source of the problem.

I have to agree that I do not think it's related to 15109, because the patch russell provided there patches a file for music in mp3 format, which we do not use, nevertheless i will make a noload configuration option for the mp3 stuff so it should be definitly excluded from the list of possible problemsources.

By: Russell Bryant (russell) 2009-08-28 12:49:30

I can not review any submitted code unless it is covered by a contributor license agreement.

By: Raimund Sacherer (hatrix) 2009-08-28 13:16:38

Hi Russell,

again for confirmation, the added patch is just for you to see that it is as well included in my 1.4.21.2 asterisk. It is a pach you (russel) provided for another app_queue crash and has nothing to do with the problem/issue on hand in this bug report.

I added just for the complete picture of my compiled asterisk



By: Leif Madsen (lmadsen) 2009-09-22 08:46:59

Deleted the patch as without the license agreement being signed, it is useless -- and since it isn't the reporters patch, there isn't much point in it being there.

Additionally, what do we need to do here to move this issue forward? Is there more debugging information that is being waited on here?

After reviewing some of the notes, it looks like the reporter has to, at the least:

* provide a new backtrace with DONT_OPTIMIZE enabled
* description of how to reproduce the issue
* some relevant configuration files which may help to reproduce the issue

Thanks!

By: Leif Madsen (lmadsen) 2009-10-26 09:43:31

Closed due to lack of feedback. If the reporter is able to provide the requested information, please reopen this issue. Thanks!