Hi everbody,

I hope somebody can help me with my ISCSI-Troubles with the NW6.0

The basic environment was a NW6.0 Cluster with a shared scsi-tower.
Because the scsi-tower had some problems and the performance was poor the
idea was to change to a dedicated iscsi-environment.

We transfered the data from the scsitower to the iscsi-system with
netware mirroring
After transfering, the mirror was broken.

Because of disk-space-need and the basic idea, that Gbit ethernet has a
better performance than a 40Mbit SCSI-bus with old 18GB HDs we were bying
a 2 P4-SATA2 Systems, with 2 250GB in a Linux software raid. The 2
netware cluster machines and the iscsi-machines were connected with 3Com
3200 cards. The target software is SLES 10 with Linux enterprise target.
The first switch was a netgear, the second we tried a DLINK Gbit switch.

The nw6 server are patched with sp5,iscsi sp2,tsaup19,nw6nss5c and

Basicly the system is working, but the performance is poor.
Writing is "acceptabel", but reading is very very slow.

Because of that, we setup jumboframes.

But, after that, when using jumboframes, the iscsi-drives on the nw
servers become deactivated after abour 1 minute and are unaccessable.

Sometimes 1 of 4 volumes were accessible, but maximum for 5 minutes and
very slow.
Switching back to 1514 bytes on the nw server, everything were working

Changing the cards to Intel pro 1000 GT did not change anything.
I tried to change the target to nw6.5 sp5. The difference here were, that
NSSMU on NW6 was unable to load and use the ISCSI drives.

With linux in the background, NSSMU did load, and the iscsi-drives become
deactivated when I tried to access the devices.

The second problem is that the backup is very very slow.
It differs between 50MB and 250 MB per second on the volumes.

Does anybody have an idea what can I do, to get working jumbo frames or
any good idea to get a good performance of our iscsi-environment ?

Thanks for helping,

Regards from Austria,

Reinhold Steininger