webserver at 100% cpu

[raviprasad2003]

Hi,
I am running HPUX webserver using java 1.4.1 in hpux 11.11. After a while webserver after ( 4-5 hrs of activity or idle) hangs and the cpu usage is at 100%. Checking with glance there are hight number of calls to sched_yield. I followed the documentation "Performance tuning java other problems". I tried to do kill -s SIGQUIT but it did not produce any log output. when the webserver is running normally Kill -s sigquit will dump information.
I made jvm to dump core file. I used dde and tried to see the stack of the thread.
The following are the stack trace
5. java_java_net_plainsocketimpl_socketaccept
4.jni_callBooleanMethod__FP&JNIEnv....
3.jni_invoke_nonstatic_FP&JNIEnv......
2.os:yield
1.sched_yeid

Also, when the webserver is hung. I cant accept any page from browser. I did a kill -s pid then the cpu usage came down to 0. But then i sent kill -s 3 it went back to 100%, but did not dump any trace.

Can some body point me as to what i should do next. How can i fix this ?

Thanks in advance,

Ravi

[Steven E. Protter]

There may be errors in the web server log.

Its called error_log unless you changed it in httpd.conf. Its located in different places depending on how you got your web server installed(HP, porting center, apache and compile).

Check the access_log file for last access and make sure the hang coincides with the kill.

If java was called by the httpd server it makes sure a kill -9 would nail the web server as the parent.

I know this seems basic, but did you install all of the required OS patches prior to installing java (http://www.hp.com/go.java). Its really important.

There is a recent OS patch on the site and a new release of java is due shortly.

You might want to check that patch and see if it relates to this issue.

Note: itrc forums are going down at 8:00 p.m. Eastern until approximately 8:00 p.m. Sunday night.

SEP