Rudi Ahlers wrote:
I mainly want to use it as a backup server for hosting servers, so I'll focus on FTP / SSH / SFTP / iSCSI (if possible), and maybe NFS - I don't want SMB (for security reasons). I'll probably also add Webmin to allow users to browse their backups via HTTPS, manage folders, etc.
You might like backuppc (http://backuppc.sourceforge.net/) for a backup system that will let individual machine 'owners' browse/restore their own backups while using compression and linking all duplicate files to use much less disk space than you'd expect. There's some tradeoff in speed compared to straight rsync and it needs more CPU, but the disk savings and ease of use might be worth it.
Yes, Backuppc is one of the programs we'll suggest :)
The pooling won't have the same effect if you run many separate instances sharing the file server. If you run a single instance that backs up many machines, you only actually store one copy of each unique file and all duplicates become hardlinks to that instance whether the duplicates are found across hosts or in different runs of the same host.
If these are real or virtual hosts you can give their owners web access to only their own host's backups. If you have virtual web sites on the same host you have to go through some contortions to split control but it is still possible.
But it will also be used for Linux control panels like cPanel, Plesk, Webmin, etc which use traditional FTP backup (via local LAN only).
And those won't have any pooling.