weird problem with SAN connected device

[Igor Lautar]

Hi,
we have the following configuration:
A device connected to SAN, accessable from windows and hp-ux. When working with device on windows (backup), simple open on the same device from hp-ux causes windows backup to abort (system write error).

HP-UX is 11.00. Did anyone seen such a problem ? What could be wrong ?

thanx,
Igor

[harry d brown jr]

Are you saying that an HPux server is sharing a disk with Windows?

Or are you saying that an HPux server is sharing the same PORT on the Storage array side?

HPux and Windows can not play together on any same ports, mostly because windows does not play well with standards and thus has issues when a new server logs on to the SAN.

live free or die
harry

[Igor Lautar]

This is actually an SCSI tape device, connected to SAN fibre channel SCSI-FC bridge. Windows and HP-UX are both connected to this FC network (SAN environment) and can devices. There are no problems except this one (for now) and this set-up has many many hours of tough backup/restore sessions.

[harry d brown jr]

It makes sense then. TWO computers can NOT use the same tape drive at the same time. When HPux opens the tape drive it is probably issuing a rewind, and that causes the windows box to freak.

If you want to use the same tape drive, then I suggest backing up both systems to DISK, then copying the DISK backup in raw format to tape.

Who is the vendor of your SAN switches and Storage arrays?

live free or die
harry

[Duncan Edmonstone]

You need to install patch PHKL_27303 or a superseding one and set the kernel parameter st_san_safe to 1.

The problem is that the device open'd is a 'rewind-on-close' device, which means when its is closed, it tells the tape drive to rewind - right in the middle of your windows backup... you should also disable the removable storage manager service on the windows hosts that see the tape (they can cause the same problem)

HTH

DUncan

[Igor Lautar]

Hi,
sorry for not mentioning before, but all the latest patches are installed.

And if they waren't, it probably wouldn't make a difference, because we use no-rewind-on-close devices. The problem occurs _BEFORE_ open system call returns, thus no-rewind-on-close isn't matter here.

Very weird...
Tried calling open with various parameters (O_RDONLY, O_RDWR ...) but backup on other host fails before open returns.

We'll try to reproduce problem on 11.23 system...

Regards,
Igor