Can someone please help in analysis of the GC logs of WAS 7 (64 Bit) on Solaris (System Configuration: Sun Microsystems sun4v SPARC Enterprise T5120 & Memory size: 65408 Megabytes).
The application seems to be having a unusual large GC.
Can someone provide any hits at where i can start looking at tuning the GC.
The application is a JSF (WAS 7 - JSF 1.2) application. The application was working really well in WAS 6.1 (JSF 1.1)
mx - Min Heap Size 2300
ms - Max Heap Size 2300
MaxNewSize - -XX:MaxNewSize 1000
NewSize --XX:NewSize 1000
We have observed the following in the GC. can this impact the performance ? what are the time specified in like user , sys and real ?
CMS: abort preclean due to time 77664.614: CMS-concurrent-abortable-preclean: 8.588/8.588 secs Times: user=8.65 sys=0.15, real=8.59 secs
Joined: Jul 20, 2010
concurrent-abortable-preclean is a concurrent phase of the garbage collection in CMS, so while it has taken time to run (too much time in fact, since it aborted), it's only impact is on the CPU usage. The stop-the-world phases of CMS are initial-mark and remark, as described in this Oracle documentation.
As for the different durations reported, real is the wall-clock time (indeed, the real time as seen from an observer), user is the processing time across all involved threads in the JVM (so 2 threads running for 3 seconds represent 6 seconds of user time), and sys is the processing time in the OS kernel (underneath malloc & free, for example).