[Dump-users] problem with restore

[Jean-Yves B. <jy...@er...>]

Hello,

I use dump to backup a large file system to a LTO2 tape changer.

The system is a debian Sarge running a 2.6.8-3-686-smp kernel.
The hardware is a HP LTO-2 tape drive with an Overland LoaderXpress tape 
changer. The system have 2 GB or RAM and a swap of 2 GB too.

Version of dump/restore is the last one (dump 0.4b41 (using libext2fs 
1.37 of 21-Mar-2005)), compiled by my own.

The file system is ext3, its size is 1.3 TB with 300 GB to backup.
Data to backup comes from BackupPC, a simple by efficient backup on disk 
system.

Dump works perfectly. Here is the command I run :

/usr/local/sbin/dump -0 -b 64 -f /dev/st0 -M -F \
     /usr/local/sbin/save_tape_change.sh -I 0 /dev/lvm0/backuppc

The script save_tape_change.sh drives the tape changer.
Data are dumped to tapes in 4h49m.

The problem is when I run restore to check tapes are OK.
Here is the dump command I run :

/usr/local/sbin/restore -r -y -b 64 -f /dev/st0 -M -F  \
     /usr/local/sbin/save_tape_change.sh -N

And restore takes hours, fulling memory and swap.

My question is how could I run restore in the same time dump takes ?
How much memory I need (or how could I compute the memory needed) ?
How could I run dump using less memory ?

Thanks for your answers.