/var corruption

[Raul Igartua]

I have corruption in the /var fs and I don't have an ignite tape.

I do have a fullbackup and an Installation/update/recovery cd.

Can these two help me recovery a healthy /var fs?

Thanks in advance

[MarkSyder]

It depends on the nature of the corruption. What are you seeing?

Mark Syder (like the drink but spelt different)

[KapilRaj]

single user mod & fsck -y /var . Did you try ?

Regds,

Kaps

[A. Clay Stephenson]

When you don't bother to indicate what kind of backup you have (tar, cpio, fbackup ...), you make it difficult to answer. The good news is that if you used a standard backup tool (e.g. not Data Protector or NetBackup) then you won't even need the recovery media.

All you need to do is bring the machine up in single-user mode. You may simply need to do an fsck on /var or you may run newfs and create a new filesystem, mount /var, and then restore. You then umount /var and reboot and you are fixed. In single-user mode, you need to execute the commands in /sbin because all of these (tar,pax,frecover) are statically linked. You can use /sbin/pax to restore tar, cpio, or pax images although you could mount /usr and then also have cpio itself.

[Raul Igartua]

Sorry, I forgot to say what kind of backup.

I have an fbackup.

The problem started when the EMS reported a a disk error and the server crashed. The disk with problem is in VG00. At the time there was no mirror disk.

I booted the server in single user mode and did the fsck -F vxfs -o full /dev/vg00/lvol8

could not stat.

The server was not booted again until this day. It complained about some files but it booted. I then made a mirror of vg00 and replaced the defective drive and resync'd both drives.

The tested the both drives and they boot ok, but when I tried to run swlist I got the error:

root:/>swlist
# Initializing...
# Contacting target "HPL-2000"...
ERROR: The expected depot or root does not exist at "/".
ERROR: There is currently no installed software on host "HPL-2000" at
location "/".

So, I believe that /var has some corrupted files.

[James R. Ferguson · Accepted solution]

Hi Raul:

You need to restore '/var' from a backup. WIthout the Installed Products Database held in '/var/adm/sw' your server is worthless in terms of any patching. It's not capable anymore.

I would boot up into single user mode; mount '/usr' and '/tmp' and 'newfs' '/var'. Then, use your 'frecover' to recovery the entire contents of the '/var' filesystem. I assume that you have a current backup taken at least after your last patch ('swinstall' and/or 'swremove') session.

Regards!

...JRF...

[Raul Igartua]

Thanks to all. Your resolution of recreating the fs worked fine.

[Raul Igartua]

Everything is working ok