Weblogic

To trace a java thread from a thread dump to the corresponding thread on Solaris 10, you can find the LWP using pstack by converting the NID (which corresponds to the LWP ID on Solaris) and matching it up in the pstack output, or by finding the TID listed in the pstack output.

For example, I ran a thread dump and I want to trace this thread from the dump:

"ExecuteThread: '2' for queue: 'weblogic.kernel.System'" daemon prio=5 tid=0x00880588 nid=0x35 in Object.wait() [aba7f000..aba7fc30]
	at java.lang.Object.wait(Native Method)
	at java.lang.Object.wait(Object.java:429)

First, I made a few modifications to the nodemanager start/stop scripts:

(on weblogic 8.1)

in startNodeManager.sh, I added a -D tag to be able to easily spot the nodemanager process when doing a 'ps' by adding this line:

----------------------------------------------------------------------------
JAVA_OPTIONS="${JAVA_OPTIONS} -Dnodemanager"
----------------------------------------------------------------------------

and then in the actual java start commands, adding ${JAVA_OPTIONS} to each line that calls nodemanager:

Weblogic Performance Tuning
---------------------------


Generic JVM tuning tips:

        Generally, JVM performance is increased when the initial heap size (-Xms) is set equal to 
        maximum heap size (-Xmx). (Growing or shrinking the heap space is expensive as far as 
        performance goes, so setting them equal to each other prevents growing/shrinking, but also 
        assumes that you know what you're doing when you decided what that one value should be...)  

	The larger the heap, the longer a full garbage collection will take.  Remember, bigger