Hello, Folks,

I'm in the middle of a migration from NetWare to OES2 Linux. I seem to
be having some problems with a certain model server, and I am hoping the
community might help me out.

The environment:

Hardware- Gateway 9525R Server, 2x 1.8GHz Quad Core Xeon, 8GB Memory,
LSI MegaRAID 8308ELP controller, Drives = 6x Seagate ST3250620NS 250GB
SATA, configured as 5 drive RAID 5 array plus one hotspare. Driver =

OS - SLES 10 SP2 64bit, Acting as XEN Dom0 for two DomU instances of
SLES 10 SP1/OES2 32 bit. One for file serving, one for GroupWise MTA &
PO. Volumes configured: 500MB ext2 for /boot, 100GB reiser for /, 150GB
NSS partition for user data and post office. The volumes for the domU
are LVM volumes on the host which are not mounted there.. but are mapped
as a physical drive in the domU and mounted in the VM.

The problems cropped up when copying large amounts of data to the DomUs
and then when we brought up the migrated GroupWise PO. It seems that
although copying performance is good for a while, it will bog down and
grind to a very slow pace. The NCP server will sometimes disconnect as
if the server had rebooted (but it does not) then come back a short time
later. I thought initially that it was a client setting on my Windows
workstation, but when we brought over the GroupWise PO, similarly, the
processing of files grinds to a halt, worker threads increase, and it
seems stuck forever but actually is extremely slow. If I quit out of the
post office and MTA, the process can take hours to quit while it waits
on threads to complete. The PO also reports many 820E file lock errors
on user databases. While the PO and MTA are ground to a halt, one CPU
maxes out at 100%, and the NCP server drops off the network.

These symptoms appear to happen on a WRITE action. I am now copying for
backup the Domain and PO data reading from this virtual server and it is
going quickly. I have seen a similar behavior on another server, same
model. I did not see it on an HP Proliant 380G5 server, not to this
extent anyway. the HP server has SAS drives instead of SATA. Speed
overall seems slower than it should on writes even on the dom0, which
only writes to the native Linux partitions.. but not as bad as when
copying large files to NSS.

Some thoughts I have on the problem:

1) When installing a Netware VM on the same model server, I needed to
set "XENBLK BARRIERS =0" during installation (not required on the HP
with SAS array drives)... is there an equivalent setting in SLES /OES Linux?

2) Driver problem? I am using the driver that comes with SLES I believe.
I do have LSI's MegaRAID manager on the machine and it sees the RAID
array configuration fine. Has anybody had experiences with the LSI
8308ELP controller? I think LSI has a driver available but being
unfamiliar with linux it seems quite a difficult task to change a driver
for a device. HP has its deployment pack which does this automagically..

3) RAID configuration problem? Something amiss in the configuration of
the RAID array causing a major slowdown in writes?

4) Settings in NCP server or NSS that might improve performance?

Any ideas are very welcome!