Hi Folks,

Since aquiring OES SP1, every server that I have installed shows
blisteringly pathetic disk I/O, such as:

Linux v20z-7 2.6.5-7.195-smp #1 SMP Wed Jul 20 14:39:18 UTC 2005 i686
athlon i386 GNU/Linux

v20z-7:~ # hdparm -tT /dev/sda

Timing buffer-cache reads: 2892 MB in 2.00 seconds = 1445.50 MB/sec
Timing buffered disk reads: 8 MB in 3.25 seconds = 2.46 MB/sec

An original OES installed server on IDENTICAL hardware does this:

Linux v201-2 2.6.5-7.147-smp #1 SMP Thu Jan 27 09:19:29 UTC 2005 i686
athlon i386 GNU/Linux

v20z-1:~ # hdparm -tT /dev/sda

Timing buffer-cache reads: 2544 MB in 2.00 seconds = 1271.56 MB/sec
Timing buffered disk reads: 198 MB in 3.03 seconds = 65.38 MB/sec

From above, v20z-7 is an OES SP1 install with NO novell components
installed. V20z-1 is an OES install with almost every Novell service

These are Sun V20z dual Operton servers with mirrored 72GB SCSI disks.
Up until SP1 I was reasonably happy with their performance, but now, the
boxes are literally completely useless.

For example, copying a 100MB onto an SP1 box via ssh/ftp/ncp/smb
(anything) results in the entire box essentially locking up for 5-10
minutes with the hard disk lights stuck on. During this time, local
and remote consoles are non-responsive and running services stop
responding. Once it has gotten over this, it returns to normal until
the next time you want to write more than a few megabytes to the hard disk.

*WHEN* I can get access to a responding console, top shows nothing
eating much CPU at all, but the load often hits 20-30.

Anyone else experiencing any similar problems or have any idea where to
start looking?

I'm suspecting given it happens irrespective of Novell services that it
may be a kernel issue. I am about to attempt downgrading the kernel on
the SP1 server and see what happens.