SUMMARY: E450 with PT-PCI450 and DEC StorageWorks problem

From: Sapiro, Vadim (vsapiro@cda.com)
Date: Sat Mar 13 1999 - 16:10:41 CST


Sorry for the belated summary, but we finally put the
problem to rest only a few days ago.

Special thanks to:

   Al Hopper [al@logical-approach.com]
   Steve Vazquez [vazquez@mpi.com]
   Birger Wathne [Birger.Wathne@Ark.no]

for their advice and encouragement :-)

The solution we've settled on was to switch the PTI
card in the E450 to the SUNW emulation mode (leaving
it with SCSI ID of 6) thereby bypassing the driver
supplied by DEC. I am too paranoid to say that the
problem has been solved, so I'll just say that we
haven't been able to replicate the problem so far.

Note: our decision to go with the PTI SCSI card was
driven purely by the fact that SUN has always told
us that you cannot change SCSI ID of the controller
to anything from 7 when using SUN manufactured SCSI
cards. In our quest to solve the original problem,
we've found a handful of documents on sunsolve that
tell how to do just that.

> -----Original Message-----
> From: Sapiro, Vadim [mailto:vsapiro@cda.com]
> Sent: Wednesday, March 03, 1999 2:43 AM
> To: 'sun-managers@sunmanagers.ececs.uc.edu'
> Subject: E450 with PT-PCI450 and DECStorageWorks problem
>
>
> Greetings!
>
> I have an urgent problem with RAID data corruption
> that involves the following hardware: E450, U2, DEC
> StorageWorks, PT-PCI450 UFW SCSI card with DECptisp
> PTIX,isptwo driver.
>
> Both E450 and U2 are running Solaris 2.6, installed
> at the same time (late Jan, 1999), and with the same
> Recommended Patch Cluster. Kernel patch 105181-11.
>
> 1. Background
>
> Here is the setup: E450, U2, and DEC RAID are on the
> same SCSI bus. E450 has Performance Technologies PCI
> UFW SCSI card. With a "Y" cable it connects to U2,
> that has SUN UFW SBUS SCSI card. Another "Y" cable
> connects U2 to DEC where the bus is terminated.
> Termination is disabled on both SUN and PT cards.
> Here is the picture (hope it comes out ok):
>
>
> Terminator
> | |------------|
> | --- -------| DEC St.Wks |
> \ / \ / |------------|
> \/ \/
> ------ ------
> |E450| | U2 |
> |____| |____|
>
> /RAID2 /RAID1
>
> E450 has PTIX,isptwo driver installed (vendor: DEC;
> pkg: DECptisp). SCSI ID of the adapter has been set
> to 6 by the switch on the card.
>
> Raid Array has two RAID 5 sets that are mounted as
> /RAID1 on the U2, and /RAID2 on E450.
>
> 2. Problem
>
> The data on /RAID2 (on E450) appears to be corrupted.
> For example, any tar ball (~1GB, ~2GB, 2Gb+ in size)
> created in that file system cannot be untarred.
> Sometimes ls, file, and other commands fail to read
> or recognize the files.
>
> 3. Other symptoms/facts
>
> - At no time the same raid file system was mounted on
> both systems simultaneously!
> - /RAID1 (on U2) is fine
> - there are no SCSI errors or resets on either system
> - RAID diagnostics report all subsystems (controllers,
> disks, etc.) in "GOOD" state
>
> 4. Testing
>
> - Unmounted /RAID2 from E450, mounted it on U2 and
> did same tests with no problems!
> - Unmounted /RAID2 from U2 and mounted it back on E450
> and could not read the tar balls created on U2.
> - Bottom line is that there are no hardware problems
> with the RAID h/w itself. The problem is elsewhere:
> driver, PT card, conflicting patch, or a combination
> of those.
>
> Note: SUN, DEC, and PT tech supports have not been
> able to figure this out.
>
> Any help/ideas would be greatly appreciated and will
> be summarized to the list.
>
> Thanks!
>



This archive was generated by hypermail 2.1.2 : Fri Sep 28 2001 - 23:13:16 CDT