each time resin hang, we can find logs like this: [2012-07-31 16:15:51.002] {http--8080-3562$458110589}SessionImpl[abcHefmXCW84PfMKtgCJt,] LRU while in use (use-count=1). Consider increasing session-count. [2012-07-31 16:15:51.007] {http--8080-3135$1712837012}SessionImpl[abcjyOkzo4PRk5OZVtAJt,] LRU while in use (use-count=1). Consider increasing session-count. [2012-07-31 16:15:51.025] {http--8080-3140$1696663856}SessionImpl[abcJXzltaA3iAavORQAJt,] LRU while in use (use-count=1). Consider increasing session-count.
2012/8/3 Chen Bangzhong <bangzh...@gmail.com> > > Attached is the jstack dump. Can you confirm it is the same issue as 3509? > > If it is issue 3509, is there any workaround? Which version should I > upgrade to ? > > thanks > > > > > 2012/8/2 Scott Ferguson <f...@caucho.com> > >> On 08/01/2012 08:19 AM, Chen Bangzhong wrote: >> >> Hi, All >> >> We run resin 3.1.9 open source version in our production environment. The >> visit is about 5M visit/day per resin instance. We ran into the resin hang >> problem several times. We belive it is the same problem as bug 3509 >> >> http://bugs.caucho.com/view.php?id=3509 >> >> All the threads are deadlocked and the number of threads is approaching >> the thread-max setting. >> >> Is there any workaroud to bug 3509 except upgrade to 4.0? Is this bug >> fixed in resin 3.1 ? Can we disable access log to workaroud this problem? >> >> >> Does your thread dump show the same place that it's blocking on as in >> that bug report? i.e. the access log. >> >> >> >> Besides, what's the suggested thread-max for resin ? We set it to 3000, >> is it too big? >> >> >> That's a good value. >> >> >> If we set it to a lower value, what will happen if the concurrent >> requests exceeds the thread-max value? >> >> >> In a normal situation, you shouldn't reach the thread-max ever. (Resin >> 4.0.x enforces that better with a port-thread-max.) >> >> If it hits thread max, Resin will queue new thread requests until one is >> idle. >> >> >> >> Another observation is that when the problem arise, the number of threads >> is approaching the thread-max setting. I want to know under what condition >> will resin start a new thread? >> >> >> There's a thread-idle-min value. When the number of idle threads drops >> below that value, Resin spawns a new one. That generally avoids any actual >> wait for a new thread to be spawned, because they're normally taken from >> the idle pool. >> >> >> >> Our environment: >> >> Jdk 1.6.21 on CentOS 5.3 64bit >> Resin 3.1.9 open source version >> -Xms 4G -Xmx 4G -Xmn 1200m >> >> <jvm-arg>-Xmn1200m</jvm-arg> >> <jvm-arg>-Xms4000m</jvm-arg> >> <jvm-arg>-Xmx4000m</jvm-arg> >> <jvm-arg>-Xss128k</jvm-arg> >> >> <memory-free-min>100M</memory-free-min> >> <thread-max>3000</thread-max> >> <socket-timeout>65s</socket-timeout> >> <keepalive-max>250</keepalive-max> >> <keepalive-timeout>10s</keepalive-timeout> >> >> >> These look fine. >> >> -- Scott >> >> >> thanks. >> >> >> >> >> _______________________________________________ >> resin-interest mailing >> listresin-interest@caucho.comhttp://maillist.caucho.com/mailman/listinfo/resin-interest >> >> >> >> >> _______________________________________________ >> resin-interest mailing list >> resin-interest@caucho.com >> http://maillist.caucho.com/mailman/listinfo/resin-interest >> >> >
_______________________________________________ resin-interest mailing list resin-interest@caucho.com http://maillist.caucho.com/mailman/listinfo/resin-interest