What should I do , help me please.

[Fred.Wu]

A hard disk which /usr filesystem located on had bad blocks on it, so I can't login multi-user mode.
Can I get any idea to solve the problem,since there's no backup tape or anything at all,and no mirror disk.
Can I mark the bad block as "fail" and resume?

[monasingh_1]

you can try to boot in single user mode and then try to do fsck on /usr filesystem and see if that recovers the file. one bad block wshould be recoverable.

If this does not work then you can also try to force mount the /usr and then tar the important data to tape and then reload the OS.

Goodluck...

[S.K. Chan]

Before you do anything at this point it may be a good idea to make a full backup of your system just in case everything goes bad. You may not be able to back everything up but at least try to take snapshot of your system configuration so that later on it'll be easier during a system rebuild (if needed). Try running fsck in single user mode (ie fsck on the lvol itself) to see if it can clear it up.
# fsck -y /dev/vg00/rlvolX
If you can't mount /usr after the fsck it has to be severe already and probably will not be able to recover.

[Bill Hassell · Accepted solution]

fsck CANNOT fix bad blocks. It fixes structural errors in directories and inode tables but a hardware error is not fixable through fsck. Indeed, it is not fixable with any software workaround except reformatting or replacing the entire disk. There is no way to resume with a bad block EXCEPT to use mount -f but this risks creating a complete system crash.

You must backup your data as soon as possible using a program that can handle hardware problems (tar, cpio, pax cannot) like fbackup. There are no guarentees that this will succeed if the bad block(s) are in the middle of a directory...the system may crash when trying to access the bad directories or inodes.

[Fred.Wu]

Thanks a lot, fortunately I have a recently ORACLE backup. :)
So,is it the only way out to re-install the system?

[Bill Hassell]

You could spend a few days poking around the /usr directory to see what you can delete to avoid using the bad block but it may render the system unusable.

The problem with a reinstall is that you'll have to configure everything again (mountpoints, networking, security, user logins, etc), install all the patches, then install and configure Oracle and finally restore the Oracle database.

It's too late now but if you had created a recent Ignite/UX vg00 backup, the system would likely be running again by now. And as you now know, there is no substitute for a full backup.

[Steven E. Protter]

If you can get /usr to mount (usable) you can use make_tape_recovery to backup the the system configuration to tape.

Then you boot the box,
intervene at the 10 second prompt

type sea
hit enter

type bo p#

where # is the number displayed of the sequential device(tape)

hit enter

Type N to the question about interacting with the IPL

Right after getting the Ignite system is when you replace the root disk.

This will save you the trouble of reinstalling all your applications.

Take a shot with oracle down at using fbackup to get a current oracle backup.

You must do something about that disk.

I would also suggest purchasing mirror/us and setting your system up to mirror its boot drive and the rest of the os to a second drive to avoid these kind of crisis.

SEP