Dear All,

I am going to re-install the clustering software in my RH server (CENTOS).

Should there be an issue, we need to roll back.

Is there anyway, i can backup everything on my server, and then revert back to the original system?

Thanks

Just use rsync -a in single user mode to copy everything. Exclude
- /dev*
- /proc/*
- /sys/*
- /tmp/*
- lost+found/
- /media/
and other directories with mountpoints that you don't want to copy.
Create them empty on destination filesystem with the same rights.

Use grub-install after operation, to configure boot loader in destination environment (or restore it after rollback)).

That is all.

You can get a nice progress using rsync -av --progress

RMI

If you want an exact copy of your HD you can use dd.

dd if=/dev/sda (or whatever your device is) of=/dev/file.containing.HD.image

using /dev/sda will do an exact image of that drive wich will contain the partitions on that drive (sda1, sda2, etc...) and all its data.

If this is a true server using hardware this might complicate the usage of something like DD, especially if done without booting to the OS.If this really is critical data I would recommend you do a file level backup to another disk/share, which you can then verify easily from a different server. Then if you can shutdown and boot to a CD or something similar and do a DD to get a block level copy off to a different disk as well that wouldn't hurt.

But what interests me is that if this is a part of a cluster, why even bother? Everything should be replicated across all nodes, otherwise you would have serious problems with failing over or load balancing on the cluster nodes. Realistically all that should be on local dis is the OS and configuration files.

I use dump(8) and restore(8) for exactly this purpose.

1. Backup the filesystems (using snapshots if they're LVM2; otherwise, you may need to boot with a live CD to get good backups) and an easy-to-access copy of /etc/fstab.
2. Upgrade and/or make your changes.
3. If rollback is needed, boot with a live CD, create the filesystems, and restore(8).

Tape device optional. You can do all this over ssh(1).

Quite true, I wasn't thinking of RAID systems, just a simple HD.

Good point, if it were a [Heartbeat|OpenAIS]/Pacemaker cluster all servers should have the same files (clusterwise talking of course). But what about the rest of the files or services (if any) that may not be running through the cluster? maybe that is what he wants to backup too.

Install a need disk and keep the old is another way.