Using rdiff-backup on debian etch

[mister.koz]

Greetings,

I look after 14 linux servers across 3 networks (ipsec joined)

I need to backup all of the data (and hopefully software) on these systems to one of the servers in our main office and then create a replica on one of the other servers.

My problem is that i don't completely understand the metadata of any backup software like this so i don't know if i can simply backup a system to one location and then from that location to the other.

Does rdiff-backup store the metadata in the destination? in this case i would suggest that backup to one server using unison or similar to syncronise the backed up data, thus maintaining the metadata. I haven't tested this yet.

Also, for our scalix server we intend to move the data to an extended partition and do snapshot backups of that drive.... sounds all well and good but i have NO idea how to do this :)

Any suggestions would be greatly appreciated.

ps All of our servers are debian etch or xandros and we run our financial on suse (soon to be etch)

[geek.de.nz]

Hi, I'm not an expert on rdiff-backup but i know how the rdiff algorithm works. In a nutshell it's very efficient and requires very little transfer of data across a link (just hash checksums) and ensures that the files are the same in the destination as in the source.

I always used dirvish for backups with rsync which is very good. You might want to have a look into that as well. I assume rdiff-backup also uses rsync and if it does then you are quite safe in backing up one server to another and then rsynching the result to a third server. The files will be the same, also meta data. rsync works pretty much like the scp or rcp command with a few extras for excluding files and specifying the protocol like over ssh. It only transfers the differences.

Having a quick look at rdiff-backup shows me that it stores the differences in meta data files as you said. Doing another rdiff-backup of that would be overkill as meta data would then be saved redundantly. As I said, you can just mirror the rdiff-backup with rsync and be on the safe side. rdiff-backup is probably slightly more space efficient than dirvish but dirvish is also a bit more transparent because you can just copy the backed up files and space consumption is minimized because only changed files are stored twice which makes the normal cp/scp/rsync commands work. Files that stay the same are stored as hard links which is essentially the same file with two or more links to it.

Hope I could tell you something new. Sorry have been away from Linux servers for a while. Into Software Development now.

[mister.koz]

Awesome thanks mate, yeah backups of backup metadata is not a cool thing :) and overly in-efficient.

I will give this a go with a couple of our smaller servers and report back here.

Thank you.