|Summary:||ASTERISK-00886: deadlock undiagnosed|
|Reporter:||Matt Florell (mflorell)||Labels:|
|Date Opened:||2004-01-21 10:02:20.000-0600||Date Closed:||2004-09-25 02:57:08|
|Environment:||Attachments:||( 0) gdb_out_crash_2004-01-22|
( 1) gdb_output_2004-01-21_955am.txt
( 2) messages_2004-01-21_9am
|Description:||I had a deadlock on a fairly isolated system today that had been up for only 19 hours. I cannot figure out what caused it, but I have attached the gdb output and the debug messages file for the hour before I killed the asterisk process.|
****** ADDITIONAL INFORMATION ******
this system has 3 T1s installed on a Digium T400 card with 30 phones configured on 15 sipura SIP adapters, 1 cisco ATA188 and a single Grandstream 102.
|Comments:||By: Matt Florell (mflorell) 2004-01-21 10:19:30.000-0600|
The messages file had to be reduced in size so that this website would handle it. I removed over 4000 lines like this:
Jan 21 09:03:02 DEBUG: DTMF digit: f on Zap/61-1
Jan 21 09:03:02 DEBUG: Fax already handled
One of these lines show up every second of a Zap channel call for some reason. Although it hasn't lead to crashes in the past, I've had this machine running for a week solid at 3000 calls a day last week with no deadlocks.
By: tclark (tclark) 2004-01-21 18:33:20.000-0600
think this is a known issue, with new dsp routines
set in the Makefile to this
#Uncomment this to use the older DSP routines
By: zoa (zoa) 2004-01-21 19:24:59.000-0600
is the deadlock also a known issue ?
Is anyone looking where the new DSP's going wild ?
By: Matt Florell (mflorell) 2004-01-22 10:06:07.000-0600
Had another crash this morning. Don't know if it's related to yesterday or not. see the new GDB_out file
By: zoa (zoa) 2004-01-22 11:59:35.000-0600
coppice will have a look at the new dsp code this week, meanwhile fallback to old dsp routines.
By: Brian West (bkw918) 2004-01-23 14:39:05.000-0600
does the old DSP routines fix this?
By: Matt Florell (mflorell) 2004-01-23 14:47:19.000-0600
nope, when I load the old dsp everything compiles and starts just fine, but the T1's aren't being listened to then, for some reason wink doesn't work when using the old DSPs.
This is true for both of the B8ZS ESF E&M wink T1s and the Robbed-bit SF E&M wink T1 I have on the box
By: Mark Spencer (markster) 2004-01-26 18:35:03.000-0600
DSP routines are fixed. This bug seems to indicate a management program that is going crazy or is becoming unreachable via TCP/IP. Unfortunately it doesn't tell us exactly where in manager it fails. You may have more useful debugging if you do "make clean" and then "make valgrind" and then "make install"
By: jrollyson (jrollyson) 2004-01-27 22:00:36.000-0600
This apparently relates to fax detection - fax detection was fixed in CVS. This should no longer be an issue. Please reopen if you find otherwise.