[Home]

Summary:ASTERISK-18211: Asterisk deadlock
Reporter:Morten Larsen (mlarsen)Labels:
Date Opened:2011-07-30 08:31:46Date Closed:2011-08-02 18:46:15
Priority:CriticalRegression?
Status:Closed/CompleteComponents:General
Versions:1.8.5.0 10.0.0-beta1 Frequency of
Occurrence
Occasional
Related
Issues:
Environment:CentOS 5.6 x86_64Attachments:
Description:From time to time our Asterisk servers deadlock. It happens once or twice a week.

I'm not able to reproduce the deadlock, it just happens.

Log file, bt and core show locks can be found at:
http://dev.ipnordic.dk/asteriskbug/locks.dump.txt
http://dev.ipnordic.dk/asteriskbug/thread_bt.txt
http://dev.ipnordic.dk/asteriskbug/thread_info.txt

If you'll need anything from the debug log file then please let me know. The debug log isn't posted due to it's size of about 400 MBytes.

The server is currently running deadlocked, and I wont restart it the next few hours in the hope that someone sees this bug report and needs additional information that can only be obtained from a running deadlocked Asterisk.
Comments:By: Gregory Hinton Nietsky (irroot) 2011-07-30 10:08:57.282-0500

There has been a significant change to the statechange locking in 1.8 after 1.8.5 so any release >= 1.8.6-RC1 or SVN

By: Morten Larsen (mlarsen) 2011-08-01 03:46:43.702-0500

Are those changes in 10.0.0-beta1 as well? (The logs posted here are from a 10.0.0-beta1 installation)

I am running svn versions atm and we'll see if the issue returns.

By: Ernie Dunbar (ernied) 2011-08-02 14:20:29.260-0500

I am also having problems with 1.8.5 with SIP deadlocks, even when res_timing_timerfd.so is unloaded.

Now the official "solution" is "Oh, we've made a bunch of changes that may be somewhat related in the latest Release candidate version and we recommend that you upgrade your **production server** to the latest beta, hope that helps!"?!?!

Sorry if I'm not quite so confident in this resolution, since that has been the response since Asterisk 1.6.2.15, and we are *still* getting SIP deadlocks.

By: Richard Mudgett (rmudgett) 2011-08-02 18:46:15.930-0500

The deadlock captured here has been fixed by the committed change for ASTERISK-17760.