load avg. problems: OWS 3.0.18.0/RDBMS 7.3.2.3/Solaris 2.6

Here's the scenario...

We use OWS 3.0.18.0 which talks to an Oracle database (7.3.2.3). The webservers run (and have run) on Solaris 2.6.

On the Solaris 2.5.1 RDBMS machine, everything is fine. This machine is a UE3000 with 6 338Mhz processors and about 3GB of RAM. Load average is consistently 3-5.

On the Solaris 2.6 test RDBMS machine, we have encountered load avg. problems. This machine is a UE4500 with 10 400Mhz processors and 4GB of RAM. Its load avg. is consistently 4-7 or so.

Question #1: Why is the load higher on the UE4500? If Solaris counts *running* processes (as opposed to "ready to run") as part of the load avg., then I can probably understand it. Unfortunately, nobody at Sun has been able to tell me this for sure.

Problem/Question #2: Whenever we restart a webserver, the load avg. on the RDBMS server shoots up usually to 15 or 20... occasionally as high as 50, 60 or even 100. We have determined that it shoots up whenever the wrb processes are killed on the webserver. Yes, we manually kill them because owsctl stop doesn't work well enough. But we have always killed them this way, even on the Solaris 2.5.1 box that behaves fine. It seems like when the wrb processes go away on the webserver, our Oracle instance sees this and has to clean up the shadow processes. When it does this, several (as many as 35 or more) processes get put into the run queue at once causing the load to shoot up. Also, when logged onto the machine while this happens, I notice that it sort of locks up for about 5 seconds just before the load shoots up.

Any ideas would be greatly appreciated.

Thanks.

Sent via Deja.com http://www.deja.com/
Share what you know. Learn what you don't. Received on Tue Jul 20 1999 - 11:42:06 CDT