You might try increasing the size of the jfslogs for a very i/o active
server. I hit what sounds like a similar problem running INN on
a C10 with 2 scsi-2 f/w adapters.
Jim Doherty
-------- Referenced Mail --------------------------------
Date: Tue, 18 Jun 1996 10:30:14 +0200 (METDST)
From: Rainer Toebbicke <[EMAIL PROTECTED]>
To: "Brian W. Spolarich" <[EMAIL PROTECTED]>
cc: [EMAIL PROTECTED]
On Mon, 17 Jun 1996, Brian W. Spolarich wrote:
>
> We're starting to put together the hardware/OS specifications for our
> AFS cell.
>
> The three main contenders are:
>
> Sparc 20/Solaris 2.5
>
> IBM C10/AIX 3.2.5
>
> IBM C20/AIX 4.1.4
>
IBM C10's (perhaps C20s as well) have a serious bug somewhere in the SCSI
handling: independantly of AIX3/AIX4, our C10s crash each with a machine
check roughly every 2-3 months, something that we can provoke with a stress
test within 48 hours. The problem is certainly load dependant, we run these
machines with 18-21 2/4GB disks, some with 2 7137 RAIDs, but other
installations have run the stress test on just 2-4 disks without problems.
The 370s/390s also in use here do not suffer.
The problem has been reproduced in Austin, but they haven't come up with a
fix yet. It hit us (well, our users) so hard that we are currently going
through the trouble of swapping all our C10s against 370s/390s.
=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=
Rainer Toebbicke http://wwwcn1.cern.ch/~rtb -or- [EMAIL PROTECTED] O__
European Laboratory for Particle Physics(CERN) - Geneva, Switzerland > |
Phone: +41 22 767 8985 Fax: +41 22 767 7155 ( )\( )
