And we tried a wide variety of JDK's. This one is 1.4.2_03.
tomcat 4.1.29
RH Linux 7.2
Here's what we did (posted at Sun here [a lot of misery on this thread])
[link|http://forum.java.sun.com/thread.jsp?forum=37&thread=307252&start=30&range=15&hilite=false&q=|http://forum.java.su...5&hilite=false&q=]
>>>>>>>Post
We struggled with the same problem for almost two months. Having tried everything suggested here (and various and sundry other things) with no success, we finally downgraded glibc to version 2.2.4. It works and keeps running for weeks.
We were able to identify that the bug is in the garbage collection of the jvm - at least that's where our jvm was crashing all the time. It looks like libjvm.so is dependent upon a bug in glibc that is not there anymore. With any later version of glibc, we saw crashes in anywhere from minutes to a couple of hours, regardless of which jdk we used. After downgrading glibc, we could use virtually any jdk just fine.
If anyone seeing this problem is running tomcat 4.1.29 on a linux distro - and you can, try moving back your glibc.
HTH.
>>>>>>>>>>>Post
The downgrade to 2.2.4 works on our two dual processor boxen, but we have this test/demo box that is 1 processor and it keeps puking w/Signal 11 errors. I had nothing to do with the development of this app. The consultants my PHB hired to do the app could never get it running for longer than a few minutes on anything other than their little box the developed it on. After countless hours, and with the assistance of a real Linux goob, we got this arrangement to work. We thought we had it nailed, but now the problem is back on our single processor box.
What I know about Java you could put in a thimble and have room for your fist. I know less about tomcat and JSP's. But my experience with this has forever poisoned my view of this platform. I have a notion to just recreate this thing using PHP. It's a simple little app - way less complicated than getting this kludge of stuff running happily together.
I can try to get the hs_err... log file up if that would help. But that thread above at Sun has me thinking that we are all getting virtually the same error log.
Thanks for the help.