Re: Fibre Channel and Debian
On Wed, Jan 15, 2003 at 12:12:04PM -0500, Noah L. Meyerhans wrote:
> On Tue, Jan 14, 2003 at 05:24:26PM -0800, Grant Grundler wrote:
> > I got all the peices in place and ran a storage stress test with
> > mjacob's driver on a 2.4.20 kernel on an HP ZX1 (ia64) box.
> > Immediately found two problems that I didn't see with the qla driver
> > from Qlogic:
> > o NetRAID 2M (LSI/AMI) controller isn't ignored.
> > o data write corruption.
> How were you able to demonstrate this data write corruption?
"Storage stress test" is an in-house (HP) test originally designed
to certify SCSI disks. By default, the test partitions the disk
in two parts and runs several threads of raw IO to one partition
and several threads of FS testing on the other. Data written by each
thread is read back by the same thread to verify data integrity.
A "write" corruption is when two successive reads get the same
(wrong) data back from the mech.
I retried the test to only use raw IO and it also reported write corruption.
When I have a chance, I'll take a look at the driver and see
if it's doing anything obviously non-portable.
(Porting to ZX1 paper at http://gsyprf11.external.hp.com/ols2002/)
> I've not experienced it in production.
I would hope not.
But I'll assume you are using an i386 box and not an ia64 one.