Summary:	ASTERISK-05584: Apparent SIP deadlock with cdr_pgsql || cdr_odbc
Reporter:	inspired (inspired)	Labels:
Date Opened:	2005-11-13 15:50:55.000-0600	Date Closed:	2011-06-07 14:02:38
Priority:	Major	Regression?	No
Status:	Closed/Complete	Components:	Core/General
Versions:		Frequency of Occurrence
Related Issues:
Environment:		Attachments:	( 0) grablock.txt
Description:	At random times, quite often actually, Asterisk just hangs after a call has been completed. First noticed this on Nov 7 after a apt-get dist-upgrade and after I enabled cdr_pgsql.so. I thought that something bad happened since I was using sipfriends and cdr_pgsql.so at the same time. I therefore set cdr_pgsql to noload and started using cdr_odbc.so instead. Now I can see in my logs that the same thing is happening again. The SIP channel hangs for 15-45 minutes until it returns to normal operation. I can also restart asterisk to get it working again, but the same thing will eventually happen again.
Comments:	By: Jason Parker (jparker) 2005-11-13 15:58:33.000-0600 A lot has changed since 1.0.9.  Please try 1.2-rc2, and report your findings. By: Tilghman Lesher (tilghman) 2005-11-13 18:03:16.000-0600 As noted in logger.conf, do NOT report DEBUG messages as problems. By: Mark Spencer (markster) 2005-11-13 19:43:54.000-0600 Also please let us know if this goes away if you turn off database logging entirely. By: inspired (inspired) 2005-11-14 00:35:34.000-0600 Markster: turned off all cdr logging. I will check if it happens again. Corydon76: What clues would I leave to you if I didn't report the DEBUG messages? They are the only messages stating what's happening, and it's clearly that these have something to do with the sip channel crash. By: inspired (inspired) 2005-11-15 03:05:41.000-0600 I can confirm that this hasn't happened again after cdr_odbc or cdr_pgsql are turned off. We do need cdr logging to a db, so I will try cdr_mysql to see if the problem persists. By: Olle Johansson (oej) 2005-11-15 03:10:37.000-0600 A wild shot: Try turning on CDR caching in cdr.conf and see if that helps. It moves the actual writing to the database to a separate thread, so any database issues should not affect the handling of the call tear down. By: inspired (inspired) 2005-11-15 03:40:02.000-0600 oej: I do not have cdr.conf and there is nothing related to cdr caching in any of the other cdr_*.conf files. cdr_mysql.so is turned on now, I'm just waiting to see if anything bad happens. By: inspired (inspired) 2005-11-15 07:46:19.000-0600 Works fine with cdr_mysql. The problem appears to be in cdr_odbc or cdr_pgsql. By: Olle Johansson (oej) 2005-11-15 08:12:28.000-0600 Sorry, that was an option for 1.2, not for 1.0.9. My mistake. By: Russell Bryant (russell) 2005-11-17 15:42:51.000-0600 1.0 will only receive critical security fixes at this point.  I'm glad that you found a reasonable workaround for your problem.  Please consider upgrading to 1.2.