Try to increase CallSignalHandlerNumber yet more (10, for example). If you could grab a core file with a stack backtrace from the crash it would help a lot. Regarding PID file - when the application crashes, the application usually does not have an opportunity to delete it. Try to use another monitor application. A very simple way is to add a "respawn" line to /etc/inittab, something like: gk:35:respawn:/usr/sbin/gnugk -c /etc/gnugk.ini -t -o /var/log/gk/gnugk.log ----- Original Message ----- From: <smy@xxxxxxxx> To: <openh323gk-users@xxxxxxxxxxxxxxxxxxxxx>; "Hao Li (V)" <manager@xxxxxxxxxxxxx> Sent: Monday, August 23, 2004 1:49 AM Subject: Re: RE:RE:call limits on 2.0.8 > With Michal's help, I've compiled my RedHat9 system with Pandora release of > openh323 & pwlib, as well as running gnugk 2.0.8 on a Celeron 2.4G with 512MB > RAM. > > My previous situation was that having 3 E1s pointing traffic to my system, and > that the system crashed/stopped running after a few hours. Now that I'm > running the above combination, I've discovered that the system is running > smoother, yet it still crashes/stops running after about 7-8 hours. > > I did specify callsignalhandlernumber=2 as well as using LARGE_FDSET during > compile time, also running in proxy mode. The wired thing is that once gnugk > crashes/stops running, I've checked the pid file in /var/run/, yet the pid > file still exist while the process has been terminated. This problem has > paralyzed my script to monitor Gnugk by pid file. > > Is this a problem of system's performance?? Or something I've missed > compiling/configure the gnugk?? > > PS: Although I have 3 E1's pointing to the server, yet actual concurrent calls > are approximately a T1's worth for now. Actually concurrent calls will > increase as I get more help in stabilizing the system. > > Thanks for any advise/suggestion!! > > Quoting "Hao Li (V)" <manager@xxxxxxxxxxxxx>: > > > Thanks Michal > > > > However, after I changed callsignalhandlernumber=2, the gk crashed after > > reaching 128 concurrect calls, which never happened before. i restarted the > > gk, and it crashed again at that point. Does this mean, i have to recompile > > everything first before i can increase the number to 2? it's not related to > > acct/radius/sql etc, c/z they were not used in this case. > > Would Linux version of 2.0.8 work better with less such call limits > > problem? > > > > Hao > > > > > > From: Zygmuntowicz Michal <m@xxxxx> > > Re: call limits on 2.0.8 > > 2004-08-18 23:39 > > Increase CallSignalHandlerNumber, recompile everything > > (PWLib, OpenH323, GnuGk) with FD_SETSIZE set to a value > > greater than 64. Eventually switch to unix and use LARGE_FDSET > > mode. Check if the bottleneck is not an accounting/authentication > > modules (if you use something sophisticated, like radius/sql). > > > > ----- Original Message ----- > > From: "Hao Li (V)" <manager@xxxxx> > > Sent: Thursday, August 19, 2004 7:07 AM > > > > > > i"ve justed installed 2.0.8 on a win2000 server. seems like we are running > > to call limits > > on it. server can"t handle more than 130 > > (maybe 128) active calls at the same time. does anyone have similar > > experience and have a > > solution? I thought it should be able to > > handle around 1000 lines. > > > > thanks > > > > Hao > > VeniGlobe Inc. ------------------------------------------------------- SF.Net email is sponsored by Shop4tech.com-Lowest price on Blank Media 100pk Sonic DVD-R 4x for only $29 -100pk Sonic DVD+R for only $33 Save 50% off Retail on Ink & Toner - Free Shipping and Free Gift. http://www.shop4tech.com/z/Inkjet_Cartridges/9_108_r285 _______________________________________________________ List: Openh323gk-users@xxxxxxxxxxxxxxxxxxxxx Archive: http://sourceforge.net/mailarchive/forum.php?forum_id=8549 Homepage: http://www.gnugk.org/
