🚀 go-pugleaf

Sounds familiar. We do the 'dd' at first to get the drives synced up and
copy over the boot blocks and all the lilo data. Then we unmount the
drive until it's time for another backup. Then why backup everything
again. Just mount the drive and use something like rsync to resyncronize
the drives. Then unmount them and fsck the unmounted drive. All this
runs nicely out of cron and rsync is the key to a much quicker backup.
 The only problem with all this is IF you have a problem.. and you don't
know before the back starts, then you have now duplicated the problem on
the backup drive. But I guess that's what tape backups are for.
 There are also a few hardware solutions with raid 1. But as you said, a
stupid rm and the raid did it's job and ALL copies are gone.

I hope this helps,

Joel

"Olivier M." wrote:
> 
> I have a linux web server with 2 harddisks of the same size, with
> exactely the same partitions. The server should work 24h/day.
> 
> Now I'd like to use these 2 disks to get the most secure system :
> - against harddisk failure : if the first harddisk got a problem,
>   it should be possible to boot (using a lilo bootdisk for example)
>   on the second drive
> - against a "rm -rf /" (from an sysadmin or a hacker). for safety,
>   there won't be an entry in /etc/fstab, and drive will be mounted
>   only when necessary   (-> so RAID is not an option)
> - a complete system backup per night would be enough
> 
> How would you do that ?
> 
> I tried something like that :
> 
> dd IF=/dev/sda1 OF=/dev/sdb1 BS=10000k
> dd IF=/dev/sda2 OF=/dev/sdb2 BS=10000k
> mount /dev/sda1 /mnt
> lilo
> rm -rf /mnt/proc/*
> umount /mnt
> 
> -> it works, but : - it takes about 20 minute (9 Go)
>                    - and the files on the source filesystem changes
>                      during the "dd" (logs, database, etc)
> 
> Any comment or suggestion ? I know a tape drive would be also
> nice, but it's quite expensive, and it's quite difficult to
> boot from a tape... :)
> 
> Regards,
> Olivier