On Wed, October 8, 2003 1at 2:07 am, Volker sent the following > David Rees schrieb: >> On Tue, October 7, 2003 1at 1:59 am, Volker sent the following >>> Following problem can be recognized even if the tomcat process is >>> not in sbwait mode: >>> >>> 1) Sending a -QUIT signal to the tomcat process does not stop the >>> process. >> >> A -QUIT will not shut down Tomcat, but cause the JVM to dump a >> stack trace to stdout of all threads. This will be useful when the >> process hangs up. If the process is not responding to the -QUIT >> command, the JVM has hung up or you have a lot of threads stuck in >> a tight loop! > > I run the jvm in green mode (recommended by John from Volano for > Blackdown Java under FreeBSD!). So the problem should not derive from > too many threads!
You can suffer from scheduler starvation if you get too many threads running in a tight loop! (I know, it's happenned to me with a green JVM!) Symptoms will be 100% CPU, and the JVM will not respond except to the kill signal. > Sending the QUIT-signal does not show any result because stdout is > the terminal and that one does not print out anything when I send a > -QUIT. It is important to emphasize that the process does not show > any reaction related to a -QUIT even when tomcat is running > correctly. My original problem is that tomcat hangs up after a while > (in that case nothing can be done with signals etc.). > > But what I wanted to explain in my last posting: Even when tomcat is > running correctly (handling the servlet requests) you cannot stop it > with shutdown.sh or -QUIT. Shutdown.sh only makes the CPU% increase > over 90% and nothing does work anymore. But the tomcat process still > is in memory. Only a kill -9 helps. It's been a while since I've used Blackdown's JVM, but if we can't get a stack trace from it when Tomcat is hung up (looks like it's easy for you to reproduce by simply trying to shutdown), it will be difficult to figure out what's going on. > Like mentioned above: -QUIT does NOT produce any stack dump on the > stdout! On all Sun JVMs (IBM, too?), sending the JVM the QUIT signal will result in a stack trace like the one described here: http://developer.java.sun.com/developer/technicalArticles/Programming/Stacktrace/ It is the easiest way for us to remote debug an issue like this. BTW, does this hang occur with a stock tomcat with only the default webapps running (IE, no changes to the shipped configuration?) -Dave --------------------------------------------------------------------- To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
