Re: Oracle and Solstice DiskSuite

From: Robin Parker <parkerr_at_logica.com>
Date: 1996/01/17
Message-ID: <parkerr.70.000D1C75_at_logica.com>#1/1


In article <30F6695F.7730_at_mci.com> Chase Hacker <chase.hacker_at_mci.com> writes:

>Hi guys.
 

>We've just made some very interesting findings concerning the use
>of Solstice DiskSuite with oracle. It seems that the added layer
>of devices confuses oracle greatly, and periodically causes the system
>to go into a state where it is basically CPU bound, doing little to
>no I/O work. This in turn causes the load average to spike (like to
>the 20-25 mark) and processing to generally grind down slowly.
 

>This problem has been seemingly baffling us for several weeks. We've
>had Oracle and Sun's best tekkies in here trying to figure it out.
 

>First Oracle says it needs more memory. Then more processors.
>Sun points the finger back and says it's not the problem.
 

>By the way, this is Oracle 7.2.2 running on a SPARCCenter 2000E with
>768Megs of memory, and 6 processors.
 

>The application is an OLTP call-center application, currently running
>a maximum of about 175 users.
 

>Has anyone else seen or heard of such a problem with the combination
>of SDS and Oracle?
 

>By the way, we sorta stumbled upon this when we decided to run the
>stress test on our development server, which is not doing SDS mirroring.
>The production server has 2 18Gbyte Sparc Storage Arrays, each the
>complement mirror of the other.
 

>Any ideas?/Comments?
 

>Thanks!
 

>Please e-mail to chase.hacker_at_mci.com

I had similar problem on a Sparc 1000. There was a memory leak in Sun volume manager. This caused machine to thrash and finally grind to a complete halt (where you couldn't even remotely log on to it!). This was Solaris 2.4 BTW.

Patch for this is 101907-05. I think kernal patch 101945-32 is also for some nasty memory leak.

We have not had this problem lately.

Robin Parker
parkerr_at_logica.com Received on Wed Jan 17 1996 - 00:00:00 CET

Original text of this message