[CentOS] Backup PC or other solution

[John R Pierce]

On 5/6/2015 1:34 PM, Valeri Galtsev wrote:
> My assistant liked backuppc. It is OK and will do decent job for really
> small number of machines (thinking 3-4 IMHO). I run bacula which has close
> to a hundred of clients; all is stored in files on RAID units, no tapes.
> Once you configure it it is nice. But to make a configuration work for the
> first time is really challenging (says one who still managed to configure
> it

I've been using BackupPC to backup about 25-30 servers and VMs for a 
couple years now.     My backup server has a 20TB raid dedicated to 
BackupPC, using XFS on LVM, on CentOS 6.latest...  That backup raid is 
mirrored to an identical server in a seperate building via drbd for 
disaster recovery.   I keep 12+ months of monthly full backups, and 30+ 
days of daily incrementals.   The deduplicated and compressed backups of 
all this take all of 4800GB, containing 9.1 million files and 4369 
directories.  The full backups WOULD have taken 68TB and the 
incrementals 25TB without dedup.

I'm very happy with it.

its a 'pull' based backup, no agents are required for the clients... it 
can use a variety of methods, I mostly use rsync-over-ssh, all you need 
to configure is a ssh key so the backup server's backuppc user can 
connect to the target via ssh as a user with sufficient privs to backup 
the desired file systems. for my couple windows servers, I install a 
cygwin based rsync.    BackupPC also can use nfs, smb, and tar-over-ssh 
as backup methods.

adding a new host to the backup service takes me about 5 minutes. it 
would probably take even less time if I bothered to document and/or 
automate the process :)

users can be given access to their own backups via the web interface, 
and they can either download single files, a tar or zip of a directory 
tree, or tell the server to push a restore onto the original target.     
you can download or restore ANY version of any file thats in the hive.

the major downside is that ALL the backups have to be stored on one 
monolithic file system, and it uses tons of hard links.  If you use XFS, 
this is not a problem.    maintaining a backup of your backups can be 
done a couple ways, I am using drbd to a mirror server, but there's also 
a provision I haven't explored for generating archives.





-- 
john r pierce, recycling bits in santa cruz