topic Re: SCSI Question in Operating System - HP-UX

SCSI Question

Jason Berendsen — Mon, 17 Feb 2003 16:31:20 GMT

We just connected a second HP 1200EX Optical Jukebox to our N class server and are getting the following in our syslog:

Feb 16 13:05:19 indxprd1 vmunix: scb->cdb: 07 00 00 00 00 00
Feb 16 13:05:20 indxprd1 vmunix: SCSI: Resetting SCSI -- lbolt: 411705, bus: 5
Feb 16 13:05:20 indxprd1 vmunix: SCSI: Reset detected -- lbolt: 411705, bus: 5
Feb 16 13:05:20 indxprd1 syslog: FNSOD:SCSI interface returns SCTL_INCOMPLETE.
Feb 16 13:07:57 indxprd1 ftpd[11867]: exiting on signal 14
Feb 16 13:09:19 indxprd1 vmunix: scb->cdb: 12 00 00 00 80 00
Feb 16 13:09:20 indxprd1 vmunix: SCSI: Resetting SCSI -- lbolt: 435705, bus: 5
Feb 16 13:09:20 indxprd1 vmunix: SCSI: Reset detected -- lbolt: 435705, bus: 5
Feb 16 13:09:20 indxprd1 syslog: FNSOD:SCSI interface returns SCTL_INCOMPLETE.
Feb 16 13:13:25 indxprd1 vmunix: I/O hardware probe timed out
Feb 16 13:13:25 indxprd1 vmunix: Last successful probe at path 0/8/0/0.6.0
Feb 16 13:24:28 indxprd1 vmunix: scb->cdb: 07 00 00 00 00 00
Feb 16 13:24:29 indxprd1 vmunix: SCSI: Resetting SCSI -- lbolt: 526605, bus: 5
Feb 16 13:24:29 indxprd1 vmunix: SCSI: Reset detected -- lbolt: 526605, bus: 5
Feb 16 13:24:29 indxprd1 syslog: FNSOD:SCSI interface returns SCTL_INCOMPLETE.
Feb 16 13:38:28 indxprd1 : su : + ttyp1 bishopcj-fnsw
Feb 16 13:39:43 indxprd1 vmunix: scb->cdb: 16 01 01 00 06 00
Feb 16 13:39:44 indxprd1 vmunix: SCSI: Resetting SCSI -- lbolt: 618105, bus: 5
Feb 16 13:39:44 indxprd1 vmunix: SCSI: Reset detected -- lbolt: 618105, bus: 5
Feb 16 13:39:57 indxprd1 vmunix: scb->cdb: 16 01 03 00 06 00
Feb 16 13:39:58 indxprd1 vmunix: SCSI: Resetting SCSI -- lbolt: 619505, bus: 5
Feb 16 13:39:58 indxprd1 vmunix: SCSI: Reset detected -- lbolt: 619505, bus: 5
Feb 16 13:44:19 indxprd1 : su : + ttyp2 bishopcj-root
Feb 16 13:50:06 indxprd1 vmunix: scb->cdb: a5 00 00 00 01 05 00 10 00 00 00 00
Feb 16 13:50:07 indxprd1 vmunix: SCSI: Resetting SCSI -- lbolt: 680405, bus: 5
Feb 16 13:50:07 indxprd1 vmunix: SCSI: Reset detected -- lbolt: 680405, bus: 5

Can anyone explain scb->cdb and what the associated number is after the :?

Thanks,

Jason

Re: SCSI Question

Michael Duthie — Mon, 17 Feb 2003 17:46:58 GMT

Jason,

you are getting SCSI resets.

1st - it the SCSI bus terminated?

2nd - check for bent pins in the cable & terminator.

Mike

Re: SCSI Question

Tom Jackson — Mon, 17 Feb 2003 18:18:53 GMT

Hi Jason:

If there's only one device plugged into the bus, either the cable, terminator, or device is bad. If you're sure the terminator works, then try a known working cable.

The device may also be self terminating. Check the manual.

Devices can also initiate a bus reset, but typically are not configured to do them. You may want to chack the device manual to make sure device resets are not enabled.

Tom

Re: SCSI Question

T. M. Louah — Mon, 17 Feb 2003 18:41:02 GMT

What do you have attached On Bus 5. Please refer to ITRC doc KBRC00006842, which mention a similar situation about SCSI Driver.

# grep -Ei "dev|Power|fail" syslog.log or the old one for further troubleshooting, for example: dev_T: 0x1F031000 mean that disk c3t1d0 has a power fail msg,
If pvdisplay /dev/dsk/c3t1d0 shows "IO timeout: default" try increasing that by runing:
# pvchage -t 90 /dev/dsk/c3t1d0
this is assuming dd runs fine on disk.
Cheers
T??
PAP! (a.k.a. Pliz Assign Points)

Re: SCSI Question

Steven E. Protter — Mon, 17 Feb 2003 19:28:48 GMT

Possible cause in order of probability, most likely first:

1) bad disk
2) Bad cables
3) Bad or missing termination
4) drive cage issue(model dependent)
5) Bad scsi card

internal disks on D and other classes of boxes have a device called a drive cage. The name is descriptive, it holds the drives and power and scsi bandwidth are routed through it. If it has a problem you should see issues periodically on different drives inside the cage.

lbolt is almost always for us a bad disk. If you have an unused desk on the chain and it fails, no lbolt. But the dead critter can cause lbolt warnings on OTHER disks.

use mstm or in X xstm and exorcize all you disks. Don't do to many at a time or I/O will grind to a halt.

If you have a bad one, besides whats in syslog you will find it.

make sure you've run a make_tape_recovery if you don't have boot disks and such mirrored.

There is ONE way to get an lbolt you don't care about. If you swap out a hot-swapable drive, it will put an lbolt in syslog which will show up on dmesg command.

It will go away next time you boot your machine.

Cheerfully yours.

SEP

Re: SCSI Question

Eugeny Brychkov — Mon, 17 Feb 2003 20:17:43 GMT

Hello Jason,
which card do you use to connect library to inside N-class? Wich interface do you use at library side?
Are you sure that you did NOT connect library's 68-pin HVD (high voltage differential) connector to N-class LVD/SE card using 68-pin LVD (low voltage differential) adapter?
Eugeny