[Home]

Summary:ASTERISK-18414: Asterisk DeadLocks After Few Hours of work
Reporter:Sarikov Gabriel (gabriel)Labels:
Date Opened:2011-09-04 09:58:17Date Closed:2011-09-05 07:39:01
Priority:CriticalRegression?No
Status:Closed/CompleteComponents:. I did not set the category correctly.
Versions:1.8.6.0 Frequency of
Occurrence
Constant
Related
Issues:
Environment:Natty 11.04 server with debian packageAttachments:( 0) Asterisk-18414.txt
( 1) Locks.txt
Description:After some time cpu raise to 100%.
sip calls can't enter asterisk.
Asterisk need to be restarted.
we use cpu-limit to limit cpu from reaching 100% or the server not responding well.
we upgraded from 1.6.2 rc2 on 9.04 to 11.04 with 1.8.6-rc3.
the issue is from 1.8.5 (the first version we tried).
iv set up a debug machine to transfer the calls to it but haven't got to use it yet.
we want to probably go back - 1.6.2.0 rc2 ran for more then 1 year without any restarts and issues.
we had one customer that said in a middle of bridged call another channel entered his channel (herd another dial plan).
Comments:By: Gregory Hinton Nietsky (irroot) 2011-09-04 11:34:42.608-0500

You will need to help us debug this further
please see
https://wiki.asterisk.org/wiki/display/AST/Collecting+Debug+Information

Core show locks
Core show threads
and a good backtrace are invaluable


By: Sarikov Gabriel (gabriel) 2011-09-05 02:58:50.197-0500

core show locks

By: Sarikov Gabriel (gabriel) 2011-09-05 03:08:42.418-0500

last 6 minutes before restarted asterisk.

By: Gregory Hinton Nietsky (irroot) 2011-09-05 03:47:17.692-0500

This is a problem with timmer_fd do you have dahdi built and installed ?? if so in modules.conf
do not load res_timing_fd.so

By: Sarikov Gabriel (gabriel) 2011-09-05 03:55:50.005-0500

i do.
we use dahdi dummy
it will work without it?
i already reversed the system to 1.6.2.9 .
is the problems is there too?


By: Sarikov Gabriel (gabriel) 2011-09-05 03:57:03.526-0500

after reversing to 1.6.2.9 the cpu usage dropped about 50% !!
1.8.6 using double the cpu for each sip call.

By: Sarikov Gabriel (gabriel) 2011-09-05 04:03:26.527-0500

this is what i get on the system i tested 1.8.6 rc3 with:
Command 'module unload res_timing_fd.so' failed.
[Sep  5 12:01:58] WARNING[2089]: loader.c:519 ast_unload_resource: Unload failed, 'res_timing_fd.so' could not be found

so i think this module is not loaded the deadlock is not because of this

By: Sarikov Gabriel (gabriel) 2011-09-05 04:04:43.579-0500

but i have this module :

res_timing_timerfd.so

is this the same?

By: Gregory Hinton Nietsky (irroot) 2011-09-05 04:07:25.863-0500

yeah thats the one make sure it does not load its broken at the moment

By: Sarikov Gabriel (gabriel) 2011-09-05 05:47:48.233-0500

Thanks,
iv returned to 1.8.6 rc3
hope it won't crash again.


By: Gregory Hinton Nietsky (irroot) 2011-09-05 06:35:43.704-0500

1.8.6.0 is available there is no change between rc3 and it.

however it will not solve the problem the fix for the timerfd issue is to be included 1.8.7


By: Sarikov Gabriel (gabriel) 2011-09-05 07:39:01.495-0500

add noload = > res_timing_timerfd.so to modules.conf
and use dahdi timing instead(asterisk choose it auto).