Root Inode Problem

[Jim Lewis_8]

System B-180/Unix HP-UX 10.20

During a backup process, the system could not stat the following files:

/home/hqvmms/MTP_OUT
/home/hqvrmn/.elm

I have attempted to cd and/or remove these directories and I get an I/O error.

I brought the server down into single user mode and did a fsck.

I got the following:

/dev/vg01/rlvhome
Phase 1 - Check Blocks and Sizes
Cannot Read: Blk 32
Continue?

Answer y or yes to the Continue question 2 or 3 times produces the following message:

Failed Read of Block #32, retried 2 times
Unable to Fix Root Inode and you are returned to a UNIX prompt.

The files no longer exist in the "ll" display but show up
as follows:

kamakou $ ll
./.elm not found
total 14
(with a directory listing following.

What can be done to fix the Inode problem. Should I remove the users and add back in?

Thanks.......

[Jeff Schussele · Accepted solution]

Hi Jim,

First off make sure you've got a good tape backup - 2 is better - of this FS

Assuming this is a vxfs FS, run #ncheck -o surface to check the disk surface for bad blocks - here's the link for the ncheck doc:

http://docs.hp.com/cgi-bin/fsearch/framedisplay?top=/hpux/onlinedocs/B2355-90129/B2355-90129_top.html&con=/hpux/onlinedocs/B2355-90129/00/02/208-con.html&toc=/hpux/onlinedocs/B2355-90129/00/02/208-toc.html&searchterms=bad%7cblock%7cinode&queryid=20020314-110106

If it does report a bad block then go to single-user or unmount it if possible & run an fsck -y on it again. (Sometimes multiple fsck s can do the trick)

If it's got a bad block that cannot be deallocated you'll want to replace that disk.

HTH,
Jeff

[Frank Slootweg]

You have got a bad disk. It needs to be checked/fixed/replaced. The clues are the I/O errors that you get and the absence of a "Cannot Seek" error.