load average and cpu usage too high, why could i do?
 

Hello, I have a host on a client that since two weeks ago the load average has increased too much (it has two cpu's and the load average is 13 or 14..) and the cpu usage is 90-95%, more or less
this is a capture



top - 14:34:56 up 416 days, 21:02, 5 users, load average: 12.98, 11.81, 12.16
Tasks: 158 total, 3 running, 153 sleeping, 0 stopped, 2 zombie
Cpu(s): 97.5% us, 2.1% sy, 0.0% ni, 0.0% id, 0.0% wa, 0.1% hi, 0.3% si


PID USER PR NI VIRT RES SHR S %CPU %MEM TIME+ COMMAND
28112 weblogic 15 0 1363m 1.1g 8656 S 86 4.8 4781:55 java
16446 weblogic 16 0 2248m 2.0g 7904 S 56 8.4 603:58.41 java
etc...

Any suggestion?
Thanks

Looks like PID 28112 is taking 86% of the CPU. It may be stuck; you might kill it if you can restart it later.

How many instances of weblogic are you running at once and do you have a multi-cpu license?

If you are supposed to be running only one instance, then you'll want to kill one.

Is the useage steady across all cpus or does it see-saw?
Which version of the JVM are you using?

Have you tuned the OS or are you using the (windows) defaults?

However, the application server, under high load, is supposed to be running the cpu close to 100%
There are tuning tips:
http://edocs.bea.com/wls/docs81/perform/topten.html

Hello! Thanks for your answers :)
Well, they are running 8 weblogic process, and they said that the cpu load has increased 15-20% since two weeks only...
They have two more hosts with weblogic and the cpu's have a 20% less load... and also they are running 8 weblogic process.
But I realized that the network load on the first host is higher than the others hosts... with the arp command could I see the process that are creating the connections?
Thanks.

... and they all have the same jvm and so on and so forth...With more traffic you will have weblogic doing more work. Realise that david may have a point if one of the weblogic processes is conspicuously higher than the rest all the time.Um...

$ arp -a
? (192.168.1.1) at 00:1C:10:11:1B:19 [ether] on eth0
HBCLUG03.local (192.168.1.108) at 00:0E:35:22:54:FF [ether] on eth0

I'm not sure I understand the question.

Hello! Thanks for your answers :)

With the arp command I mean that If I could see the connections or hosts connected to my host... and I dont know if a high number of connections could increase 20% the load average.
Other interesting data I've found is that in the other two hosts the swap memory is about 300 KB but on the other host is about 2 GB :S
It's normal?? If not, how could I know what process it's taken the swap? and how could I release the swap memory?

Thanks a lot!