HP-UX 8 Bad Blocks

[Chris Johnson_6]

I have an old Appollo 425e which continues to offer sterling service managing one of our laser cutting machines. Recently, it has complained of an inconsistent file system when booting. I am asked to run fsck manually which I have done. This reports:

/dev/dsk/0s0: CANNOT READ: BLK 259808
CONTINUE?

if I answer yes it repeats this message numerous times then changes to another block & repeats for several other bad blocks. If I stop & restart fsck, it goes throught the same cycle of bad blocks, so I assume they are not being fixed or marked bad. Is there a command to scan & mark bad blocks on the (scsi) disk?

[A. Clay Stephenson]

The mediainit command can be used to spare blocks but it wipes out everything on the disk, so a re-install would be needed. You should really start looking for a replacement disk.

[Peter Godron]

Chris,
fsck cannot fix any bad blocks, it untangles corrupted numbers in the directory and free space table. It can not relocate bad blocks and can not reconsturct missing data due to failed hardware.

You can initially check your disk by dd the disk to /dev/null. That would check reading the disk.

However, it probably best looking at replacement drive.

[Chris Johnson_6]

Thanks for your reply... surprisingly the only media i/o device is a dat drive & we use the cpio command to save all files from the root downwards. If I fit a new disk, I wont have an OS to run the cpio command to reload from the tape!!! Can I recover the os from this tape?

Chris Johnson

[A. Clay Stephenson]

Cpio will not alow you to restore the OS. You would need you original install media.

One technique that might work is to install a second identical disk in this box. You would then use dd to copy from the original to the new. Because I expect dd to have trouble reading these bad blocks, you would have to start dd multiple times specifying oseek= and iseek= to restart the dd's. Finally, remove the original disk and change the SCSI ID of the new disk to match the original and boot. The system would fsck this time as well but should be able to finish.

I honestly don't remember if there is a make_recovery command for HP-UX 8 but you can search for it. You are paying the price for running on unsupported hardware and software.

[A. Clay Stephenson]

After rattling my brain a bit, I seem to remember a "mkrs" command. I know there was one for HP-UX 9 for I just can't remember for 8. Mkrs predated copyutil which was HP-UX 10's tools which in turn was outdated by the Ignite Tools.

[Kent Ostby]

Chris --

Unfortunately the CANNOT READ BLK is the kiss of death and there is not much you can do if you get this other than mediainit and if its the root disk then you would need to install from media.

In the installation, you can select the option to mediainit during the installation, but if you really have a flakey disk then you might end up right back in this situation again.

The problem with the "makers" suggestion above is that the system has to be functioning for that command to run and I'm not even sure if it was there in 8.x.

[Victor BERRIDGE]

Greetings,

If I remember ( I had in older times to fix regularly two series 700 in HPUX 9 and a 822 in 8.. so I can say there is no mkrs for serie 800!...) when you used fsck in single user with the file system unmounted you had to reboot without sync (I forgot the option but remembered if option is omitted you had to start all again not long after...)in order to sort yourself out of trouble...
And Yes ther is a /etc/mkrs utility for creating a recovery system in 8.XX BUT not for serie 800...
http://www4.itrc.hp.com/service/cki/docDisplay.do?docLocale=en_US&docId=200000006442274


All the best
Victor

[A. Clay Stephenson]

But this is an Apollo so it is a 700 series. The news is that this is not an Apollo running Domain_OS (although Domain_OS did file sharing in the '80's as least as well as anyone does today).

[A. Clay Stephenson]

But this is an Apollo so it is a 700 series. The good news is that this is not an Apollo running Domain_OS (although Domain_OS did file sharing in the '80's as least as well as anyone does today).

[OldSchool]

the sysadmin section of the hp-ux porting center lists an app called "BBT" (bootable backup tape). Its version 1.0 from 1996. Says series 800, doesn't list OS and is source code form.

Never tried it, can't vouch for it, but you might give it a try....

[Chris Johnson_6]

Thank you all for your kind help...
I have now found the owners guide
It gives the full model as:

HP Apollo 9000 Model 425e

It also contains a section on making a recovery tape using mkrs!!

I even found a dat tape box labelled "mkrs recovery tape" in my handwriting from a year ago, so I must have made one! (been to bed several times since then).

Alas the box was empty.... the regular user has no idea where it could be!!@##!! :-(

Could I get another mkrs tape from elsewhere? Would it have to be from the same model computer? presumably it needs to be HP-UX version 8 still. If I could find & load such a tape would it just load the os & leave the apps & data alone? If not I guess I could use the cpio backup tape to put all the software (ME10 cad) & data files back.

If I suceeded in all this then I will sort out a replacement disk asap.

Currently the system has deteriorated to the point where it panics & wont boot at all

CHris J.

[A. Clay Stephenson · Accepted solution]

A mkrs tape from another box might work but you would need to find an identical model (or very nearly so) with identical drive(s) (or very nearly so) running HP-UX 8. Most HP-UX software is upwardly compatible within limit so that if you can find something running HP-UX 9.x then your application software will probably work. The problem is that at this point much of your application software and/or data may be trashed; there's just no way to know.

If you can install your disk in another machine as a data disk, you have a far better chance of creating a backup. The idea is that you would create addional mountpoints (e.g. /usrbad) and then use the -f option to force the mount of a dirty hfs filesystem.

[A. Clay Stephenson]

By the way, you are really getting what you deserve. It's one of God's little jokes. Those who have known good backups never need them; those who do not ... .

[Tom Henning]

As I recall from my HP-UX 8.x days, the mkrs tape was just that, a bootable tape with just enough of a system on it to use to restore a known good backup tape. Unlike today's ignite utilities, it did not put a backup of the system onto the same tape. So you will have to find a known good backup tape of the entire system in addition to the recovery tape.

This is one case where the "it's running so don't fool with it" attitude can come back to bite you. With hardware as old as this is it is amazing that the disk is still actually spinning. Disks fail. Bearings fail. Motors fail. Heads crash. With some luck you can possibly find another system that you could attach the failing disk to and get a good backup of what's left.

Good luck!

[Chris Johnson_6]

Latest Update

I left fsck running overnight & have got enough of the system running to get to a logon prompt & keep the laser systems running... Phew!!!

I shall be dumping all available data to tape & preparing to upgrade to alternative hardware/software (not Unix I'm afraid).

A cautionary tale!!
Elsewhere on my site I have 3 severs running business systems which ARE subject to a proper backup regime. I feel rather foolish for overlooking a system lurking on the shop floor.

Its time for a full distaster recovery assessment. I've heard of a BBC micro sitting on a pressbrake machine! What else could I have missed? :-0
How often does anyone else assess risks?

Thanks again for all your help.

Chris J.