From: John M F. <jo...@hj...> - 2005-01-10 15:16:26
|
On Mon, Jan 10, 2005 at 05:33:30PM +1300, Peter wrote: > I have some users running 1.5 Java VMs. They run for a bit. Then=20 > seem=20 > to get 'locked up'. > The process maxes out the UML CPU. They do not respond to a kill=20 > -3=20 > (which normally causes a Java stack trace to be output). The=20 > process=20 > does not output anything when I strace it (after it locks up). > This happens when the 1.5 JDK is used to start Tomcat. And it has=20 > happened running other Java apps as well. I don't see the problem=20 > with=20 > any non-Java apps. The problem also happens with the Update 1=20 > release=20 > of the 1.5 JDK. > I never see the problem with the 'old' 1.4.* releases. I don't see=20 > any=20 > reports like this with JDK 1.5 users on non-UML systems. > So I'm guessing there is a UML/JD1.5 interaction issue here=20 > somewhere. i see the exact same problem with another uml provider (not rimu). they're running 2.4.26-2um. i'm running debian unstable on it. my jboss 4 or 3 instances exhibit this same behavior when i run them=20 in jdk1.5. starting the jvm with -Xint alleviates the problem, but=20 i'm pretty sure it's making the whole app run slower. experimenting locally with the 2.4.26-3um-1 kernel that debian=20 packages, my entire machine locks and i need to kill the whole linux process. i don't know how to fix or debug it, either. if anyone has any ideas, i'd be happy to try them. --=20 John M Flinchbaugh jo...@hj... |