java memory growth problem with Redhat Enterprise Linux WS release 4
Hi,
Sorry for the cross post. I haven't got any response from JVM forum so I just want to try my luck here.
I have been running one java application on different platforms. The systems are:
A. Dell Inspiron 600M, MS XP Service Pack 2, jre1.5.0_09, 1GB RAM
B. Dell Dimension 9100, Fedora Core 4, Linux version 2.6.17-1.2142_FC4smp (bhcompile@hs20-bc1-4.build.redhat.com) (gcc version 4.0.2 20051125 (Red Hat 4.0.2-8)), jdk1.5.0_06, 2GBRAM,
C. Dell PowerEdge, Linux version 2.6.9-42.0.3.ELsmp (brewbuilder@hs20-bc1-7.build.redhat.com) (gcc version 3.4.6 20060404 (Red Hat 3.4.6-3)) , jdk1.5.0_06, 2GBRAM,
I implemented this application to solve some Integer Programming problems.
I tried to run this java application with exactly same source code and same jvm options as following:
-server -Xms1024m -Xmx1024m -XX:MaxHeapFreeRatio=70 -XX:MinHeapFreeRatio=40.
System C can not handle the task and quits the process with error message:java.lang.OutOfMemoryError: requested 32756 bytes for ChunkPool::allocate. Out of swap space?.
While system A and B can finish the task and give the correct result. Through the windows task manager for system A I found that memory usage level is pretty stable throughout the life span of java application and similar things happened to system B.
For system C, system monitor shows that memory usage is increasing very quickly and it uses up all the available memory before the task is finishes. I don't understand why the behaviors are different for the same java code and similar execution environments. My guess is that system C's OS is redhat linux enterprise version and it needs user to set some special options for jvm before garbage collection can take effects. I actually checked the memory leak of the code on system A using java profiler and couldn't find any obvious problem within the source code.
Does anyone have any suggestion what might go wrong? Sorry I can not post the source code here because it is too big and too complicated. But any suggestion is highly appreciated.