Connect Nike Model 10 to 2 servers

[Maelbrancke Martial]

Hi,

I'm trying to connect a Nike Model 10 Disk Array to 2 servers, i.e. a K410 and a K420.
The Disk Array is working perfectly if it's only connected to one system, or if both systems are connected to a separate SP.
As soon as we link both SP's the syslog gets filled up with following messages :

Dec 27 15:23:37 zhpux25 vmunix: SCSI: Resetting SCSI -- lbolt: 36609603, bus: 2
Dec 27 15:23:37 zhpux25 vmunix: SCSI: Reset detected -- lbolt: 36609603, bus: 2
Dec 27 15:24:17 zhpux25 vmunix: scb->cdb: 2a 00 00 0f ba 80 00 00 10 00
Dec 27 15:24:17 zhpux25 vmunix: SCSI: Resetting SCSI -- lbolt: 36610842, bus: 2
Dec 27 15:24:17 zhpux25 vmunix: SCSI: Reset detected -- lbolt: 36610842, bus: 2
Dec 27 15:24:17 zhpux25 vmunix: SCSI: Resetting SCSI -- lbolt: 36611785, bus: 2
Dec 27 15:24:17 zhpux25 vmunix: SCSI: Reset detected -- lbolt: 36611785, bus: 2
Dec 27 15:24:17 zhpux25 vmunix: scb->cdb: 2a 00 00 48 5c 50 00 00 10 00
Dec 27 15:24:17 zhpux25 vmunix: SCSI: Resetting SCSI -- lbolt: 36613042, bus: 2
Dec 27 15:24:17 zhpux25 vmunix: SCSI: Reset detected -- lbolt: 36613042, bus: 2


Setup tried :
K420 SCSI-card has address 7
--> Nike SPA (SCSI address 0)
--> Nike SPB (SCSI address 1)
--> K410 SCSI-card has address 6

Both servers are running HP-UX 11.0 with the last patch release installed (sept 2000).

What am I missing ??

Even if you can't help me, I wish you a Merry Christmas and Happy Newyear ;-)

Thierry.

[Thomas J. Harrold · Accepted solution]

Hi.

I have connected model 20 arrays to K420's before. (3 hosts --> 2 model 20's, all in an MC/SG cluster).

I would have suspected that the SCSI ID on your host adapters was set to the same value. Please double-check this! I have seen cases where the dip switches on the HSC cards need to be rocked back and forth, to ensure that they are in the proper position.

I have seen LBOLT errors with a bad host adapter, and bad termination. Are you using the inline termination cables? (usefull if you need to shutdown and remove a host without disrupting I/O to the other host). This involves removing the terminators from the host adapters.

Let me know what you find. I'll keep thinking about other possible problems.

-tjh

[Ovidiu D. Raita]

Model 10 is one of the first Nike models. At that time they used to run under HP-UX 9 and their firmware wasn't supporting multi initiator environments. Please check Nikes firmware revision and Nike disks revisions.

As far as I remember the first disk revision supporting multi initiator env was HP09. Verify then with HP support and ask them for firmware upgrade if needed.

Ovidiu

[Thierry Poels_1]

Thanks for the replies already!

Thomas, the K410 IS set to SCSI-address 6, the K420 is connected to the core I/O which is set to 7.
oooooooopps! Boot disk is at SCSI-address 6!! I think I know where to look ... next year (=Monday, = next working day for me).

Ovidiu, the 2 systems and the Nike were connected together long ago, but the server were running HP-UX 10.20 at that time. I'll definitely check the firmware, I think to have seen HP02!! I can't imagine however that we're running that far behind.

regards and many thanks in advance,
Thierry (now logged in with his own ID :))

[Ovidiu D. Raita]

Looks like Thomas got it right. You just have, then, to solve the problem and assign all the points to him :) :)

Happy New Year,
Ovidiu

[Ajitkumar Rane]

I think you might want to check the cable/ pins properly.Some times a crucial pin might be bent or broken or not making proper contact with the connector and this could send a scsi bus reset.You can do a visual check or try swapping the cables maybe even the scsi card on one of the systems would have problem.
I dont think the system would show such errors for a duplicate scsi id.The system should probably hang during bootup.I think you can halt the boot sequence and try the search utility at the isl prompt.

Hope you problem is solved this year and You all have a Wonderful New Year 2001

[Thierry Poels_1]

Hi,

thanks for the replies. Sorry for the late feedback but I had other problems wich needed my attention.

The cause of problems was indeed a duplicate SCSI address. The RAID is connected on one server to the core I/O, so after changing the SCSI address on the second server to 6 still conflicted with the root disk on the first server :(( The SCSI address on the second server has now been changed to 5, and everything seems to running smoothly.

Glad I didn't make the same stupid mistake this year ;)

regards,
Thierry
(will log in again as Martial and grant some points for the much appreciated help)