creation of new filesystem on a cluster system

[jaminas]

I have finally decided to create a new filesystem called /var (vxfs) on a cluster system (HP-Ux 10.20)

what commands can I use considering that the filesystem was corrupted and need to be restored from backup

Thanks in advance

[Jaime Bolanos Rojas.]

Nasir,

I think we are going to need some more info.

When you said you want to create a new file system is becuase your system is down and you want to start from scratch? or is it that you just want to add a new vg to the system?

If you have a cluster, why don't you create a make_tape_recovery of the one server that is running in the cluster and restore the other server from the tape - just as an option-

Calling a filesystem /var might be a bit confusing, at least for me unless you have a special reason that I did not understand.

Regards,

jaime.

[A. Clay Stephenson]

Because /var is used so much in multi-user mode, you need to move all your existing packages to other nodes (although) I suspect that you has already happened since /var is such a critical filesystem. You then should halt this node, shutdown, and bring this node up in single-user mode. You should then mount /usr and /tmp and run the newfs command to build a new /var filesystem. Mount /var and you should then be able to restore using tar, cpio, frecover, or Data Protector. Once /var is restored, umount all the filesystems except / and reboot normally and allow the node toi join the cluster. You then move packages back as needed.

[jaminas]

I'm begining to have a clue from your suggestions on how to go about it but Look at the following situations, you perhaps might help me to reach the final solution.

1. The system is up and running with the
packages runing on another node.
2. All other filesystems are mounted
e.g /, /stand, /opt, /usr, /opt/sybase,
/home, etc. except /var which is
suspected to be corrupted.
3. The logical volume group for /var
is /dev/vg00/lvol8.

4. I used the following commands to
determine the credentials of the
filesystem:
mkfs -F vxfs -m /dev/vg00/lvol8 and it
returns with the following:
mkfs -F vxfs -o
ninode=unlimited,bsize=1024,
version=3,inosize=256,logsize=1024,
nolargefiles /dev/vg00/lvol8 409600
5. All these were tried in single-user mode


Need to find out the following:
1. What commands can I use to determine the character device that '/var' was created on?
2. I tried the following commands:
newfs -F uxfs /dev/vg00/lvol8 and it reeturns with message that /dev/vg00/lvol8 is not a character device.
3. As I have explained in my previous threads, I need to recreate /var and restore from recent backup.

please assist and very much appreciated for your contributions.

[nanan · Accepted solution]

your command OK
except

newfs -F vxfs /dev/vg00/lvol8

to be /dev/vg00/rlvol8

good luck

[Mohanasundaram_1]

Hi Nasir,

/var may not be a clustered filesystem. The root VG cannot be shared. it has to local to each system. Therefore You suggestion of creating a cluster filesystem at /var confuses me.

If you can provide us more information on what problem you had and how you concluded that /var is corrupted, we may provide a better solution.

The recovery also depends on what type of backup you have. Have you got a fbackup of /var ?

With regards,
Mohan.

[yulianto piyut]

Nasir,

OS partitions (logical volume that use by OS can't be share as cluster. if you want to make new file system, use character device (like /dev/vg00/rlvol8).

if your configurations of clustered server is active-standby, use recovery_tape to fix your /var. make recovery on another node that not corrupted and restore to node that corrupted.


-yut-

[jaminas]

Hi!

I was able to create the corrupted filesystem /var with the following comammands:

# mkfs -F vxfs -o
ninode=unlimited,bsize=1024,
version=3,inosize=256,logsize=1024,
nolargefiles /dev/vg00/lvol8 409600

# mount /var

# frecover -xovf /dev/rmt/6mn -i /var -e / -e
/opt -e /stand -e /tmp -e /opt/sybase -e
/usr -e /home

It restored back from fbackup /var but some directories like /var/sam, /var/mail, /var/news were not recovered.

Can someone tell me why or could it be that the full backup (fbackup -0f...) was done on multi-user mode but the frecover was done on single user-mode? ('bcos I did the frecovery in single-user mode)

As a workaround solution I copied these directories from another node and re-adjusted some files to reflect the current node.

While I await for your response, I sincerely thank all those that made it possible to restore back my /var.

[Jaime Bolanos Rojas.]

Nasir,

The idea behind backups if I remember well, is that usually with fbackup you want to take the full backup in single user mode, becuase if a user or a process is accessing the files or directories that you want to copy a tool like fbackup will skip them, while if you do it in single user mode, you make sure that you are going to take a complete backup of everything.

That is why people create snapshots to do backups... to have an exact copy and still let users access those files, between other reasons.

Regards,

jaime.