Lbolt errors

[TechC]

Hello list.

I have an HP-UX box running 10.20 and I have received scsi Lbolt errors. I know this is a scsi disk error however I think I see 2 disks in the error output below.
I know that the first line caling out LBOLT with
1f004000 is equal to 31 hex and c0t4d0 correct?

But the output also mentions bp->b_dev:1f002000.
Does that mean that c0t2d0 is also having issues? This error actually occured while I was fsck'ing c0t2d0 so I am suspicious.
Here is the output.
10 points to the definitive answer.

Thanks in advance.
----------------------------------------

SCSI: Unhandled interrupt -- lbolt: 1479681, dev: 1f004000
vmunix: lbp->state: 4060
vmunix: lbp->offset: ffffffff
vmunix: lbp->uPhysScript: f2ffe000
vmunix: From most recent interrupt:
vmunix: ISTAT: 2a, SIST0: 18, SIST1: 00, DSTAT: 80, DSPS: 01080000
vmunix: NCR chip register access history (most recent last): 5311984 accesses
vmunix: 1540, ISTAT<-20
vmunix: 30, ISTAT<-20
vmunix: 45, ISTAT<-20
vmunix: 75, ISTAT<-20
vmunix: 63, ISTAT<-20
vmunix: 48, ISTAT<-20
vmunix: 779, ISTAT<-20
vmunix: 1490, ISTAT<-20
vmunix: 1335, ISTAT<-20
vmunix: 1575, ISTAT<-20
vmunix: 1053, ISTAT: 2a
vmunix: 1, SIST0: 18
vmunix: 2, SIST1: 00
vmunix: 2, DSTAT: 80
vmunix: 2, DSPS: 01080000
vmunix: 1, SCRATCHA: ff00086d
vmunix: lsp: 2e31a00
vmunix: bp->b_dev: 1f004000
vmunix: scb->io_id: 390c0d
vmunix: scb->cdb: 28 00 01 29 e8 00 00 00 80 00
vmunix: lbolt_at_timeout: 0, lbolt_at_start: 0
vmunix: lsp->state: 5
vmunix: lbp->owner: 2a20c00

vmunix: bp->b_dev: 1f002000
vmunix: scb->io_id: 390c13
vmunix: scb->cdb: 2a 00 00 43 2c c0 00 00 10 00
vmunix: lbolt_at_timeout: 0, lbolt_at_start: 0
vmunix: lsp->state: 5
vmunix: scratch_lsp: 2e31a00

[VK2COT · Accepted solution]

Hello James,

Sadly, you are right. I wish you were not :)

a) Note that HP-UX 10.20 is really obsolete.
I was involved in recovery of a large-men
clothing company who used HP-UX 10.20 last year.

We even had to look for Y2K patches for it (they did not have proper Ignite backups)!

b) >1f004000 is equal to 31 hex and c0t4d0 correct?

Yes. Here is the mathematics.

1f 00 4 0 00
__ __ _ _ __
| | | | |
major# | target | flags
| |
bus# lun

- major# (1f) is 31 in decimal, and major
number 31 is sdisk.

- bus# (00) is the card instance number to
which the device is attached.

- target# (4) is the device's SCSI id.

- lun# (0) is the device's logical unit
number.

So, this device maps to /dev/dsk/c0t4d0.

c) Similar applies to device 1f002000.
It is /dev/dsk/c0t2d0.

I wish you luck and here is what I propose:

1. Save your data as soon as possible
(Ignite for O/S and whatever else for the applications).

2. Talk to your management and plan the
upgrade immediately.

3. Audit your server to document what runs
and what you will need when you get new hardware and O/S.

Cheers,

VK2COT

[VK2COT]

Ahh,

I dislike that the ITRC screen is not WYSIWYG.
My cut-and-paste for devices got garbled:

1f 00 4 0 00
__ __ _ _ __
| | | | |
major# | target | flags
| |
bus# lun

I now cliked on "Retain format", so I hope it does the right thing.

VK2COT

[TechC]

Thank you for the detailed reply VK2COT.

So indeed the above error is showing that both drives are in jeopardy? And c0t2d0 is also bad, wow that is not good, was hoping it was only c0t4d0. Is it a possibility that this was a bad sector? Or is this definitely a sign of a failed drive?
Just thought I would ask. I know I need to replace this disk anyway but I am trying to gauge how serious or urgent this is.

Unfortunately there is no chance of upgrade as they run legacy software that wont run above 10.20. Also there is no ignite-ux software installed and a previous admin deleted the /var/adm/sw/products folder which I believe leaves unable to install software. Please correct me if I am wrong.

Thanks for the advice,

Please tell me is there any documentation on these errors available?

Thank you again.

James

[Dennis Handly]

>previous admin deleted the /var/adm/sw/products folder which I believe leaves unable to install software.

Probably. This thread mentions about regenerating the INDEX, not sure if that will let you install new software but not patches:
http://forums.itrc.hp.com/service/forums/questionanswer.do?threadId=1301625

[TechC]

Dennis,
Yes unfortunately I had seen the thread you posted and that is where I came to the conclusion that I was in trouble there.
But I am going to use dd to clone the root disk, I have really no choice I think as ignite is no option. Hopefully dd makes it through each block. Off topic a bit here, is there a way to clone a disk without dd maybe with dump command similar to Solaris? Had to ask.

Thanks

[Dennis Handly]

>is there a way to clone a disk without dd

How do you expect to be better than dd(1)? You need to make sure you use a large blocksize:
dd if=/dev/dsk/cxtxdx of=/dev/dsk/cytydy bs=1024k
http://forums.itrc.hp.com/service/forums/questionanswer.do?threadId=774520

[TechC]

Dennis,
Thanks, I did not mean better but different I guess. dd is great if your disk is in perfect shape. But it does not skip over errors if I remember correctly. I have had a disk fail on a dd of a Solaris boot disk due to disk errors. I have been able to use a ufsdump, ufsrestore pipeline to successfiully clone the same Solaris boot disk since it will skip over a certain amount of errors. That is more of what I was asking.

Thanks for the link, good reading.

James

[TechC]

Thanks for all the replies they are appreciated. I will replace these disks as soon as possible.

[Dennis Handly]

>dd is great if your disk is in perfect shape. But it does not skip over errors if I remember correctly.

Unfortunately you are correct.