Need help: fsck read failure on "/usr" logical volume

[Hamid Aboutorabi]

After a power failure the system did not come up.

Used "ISL> hpux -is" and ran fsck on all lvols. All were marked clean except for /dev/vg00/lvol7 which mounts /usr.

The error I get is:
vxfs fsck: fsck read failure bno = 1232, off =0, len = 8192

All help on how to recover is appreciated.

[steven Burgess_2]

Hello

Have you tried

fsck -o full -y /dev/vg00/lvol7

HTH

Steve

[Hamid Aboutorabi]

I used

fsck -o full /dev/vg00/lvol7

It completed pass0 and "read failure" error came up during pass1.

[Deoncia Grayson_1]

Read this thread, looks like the author was experiencing a problem like yours but not completely...

http://forums1.itrc.hp.com/service/forums/questionanswer.do?threadId=767880

[Zeev Schultz]

Could be not so good, phase 1 treats inodes :
pass1 - checking inode sanity and blocks ...

tried to check with dd if=/dev/vg00/rlvol7 of=/dev/null bs=1024k (blocks read test)?

[Stephen Keane]

If you re-run fsck -o full /dev/vg00/lvol7
again, does it still report the same block#, or is it making progress? You have a small chance it will repair itself, otherwise you will probably need to recreate the logical volume/filesystem and restore the files (assuming you don't have a physical media problem).

[Hamid Aboutorabi]

Zeev,

Can't run dd because /usr is not mounted.

Only /sbin commands are available.

[Hamid Aboutorabi]

Stephen,

It is always the same block no.

How do I recraete logical volume/file system, please? I am not a sys admin. Your help is greatly appreciated.

I can mount and access data on all other lvols on the physical drive. This is a C3600 workstation with only one 18G hard drive.

Do I need to replace the drive?

[Zeev Schultz]

Hamid,
Copy dd from another hp-ux (same version) server if you can to /sbin.
The question here is whether the problem with
filesystem or with drive.You can mount /var and
check in /var/adm/syslog/syslog.log (OLDsyslog.log) for error messages on this drive.Many simmilar messages I saw points to disk problem:

http://aa11.cjb.net/hpux_admin/2000/06/0223.html

[Stephen Keane]

Try the following command (as root)

mkfs -m /dev/vg00/lvol7

the -m flag is VERY important. You are not going to actually create a filesystem, you are going to look at the existing filesystem to see how it was created.

You should get back something like this:

mkfs -F vxfs -o ninode=unlimited,bsize=1024,version=4,inosize=256,logsize=1024,nolargefiles /dev/vg00/lvol7 1630208

(all on one line)

IF it gets to the stage where you want to recreate the fileystem, the above reply (your version - not mine!) is the command you should use. Then you need to mount the filesystem as /usr and recover the contents of /usr from your backup media.

But you really need to determine whether the disk itself is faulty first as it may need replacing.

[Bill Hassell]

This cannot be fixed by fsck. The message: "fsck read failure" means that there is a bad spot on the disk. fsck fixes directory structure problems but can never fix a bad sector on the disk. There is nothing you can do except reinstall, possibly after replacing the disk. /usr is absolutely required to run HP-UX so when it goes bad, there is very little you can do. If this disk was mirrored, the problem would not show up as the mirror copy will provide the correct data.

[Hamid Aboutorabi]

Thanks to everyone for their suggestions.

I used /sbin/lvreduce to reduce the size of lvol. This managed to avoid the bad block of the pv. Then I was able to mkfs and load from backup.

Case closed.!!

[Hamid Aboutorabi]

Read my previous post!