Infrastructure/Backup

Goals of the Backup Plan

We have a few servers we need to backup some at Duke University, some at the colo facility in Pheonix. Currently there is 34 GB of data backed up.

• CVS repos, But this will also include what ever VCS that gets implemented in the future.
• Databases there is some mysql and postgres databases that have scripted backups, we just need to pick up the dumps.
• wiki, The fedora wiki needs backing up.
• some other data on various servers.
• need to ensure that linux ACL support is included.
• security in the backup. Preferably ssh tunnel, if not stunnel, but not a vpn setup.
• possibly /home on NetApp but this would need quota limits imposed 5-10MB would be a good starting quota
• Weekly full backups and daily incrementals, keep 4 weeks worth of backups.

Current Options

amanda

Pros -

• ssh support
• Uses standard Unix tools, don't need client to restore can use tar and dump

Cons -

• Requires client on each machine
• complex configuration

bacula

Pros -

• stunnel for security
• Verification of files previously cataloged, permitting a Tripwire like capability
• Saves and restores POSIX ACLs
• Access control lists for Consoles that permit restricting user access to only their data
• built in email reporting.

Cons -

• not currently in Fedora
• requires a daemon to be running on each host (though can be worked around with nfs)
• needs either postgresql, mysql or sqlite for backup database
• differentail and incremental backups are done on time stamps only. files moved in to backup tree with older timestamps will not be backed up

backuppc

Pros -

• Completely web based (for backups and restores)
• Very easy to use
• Can create individual configs on a per host basis or one global config for all hosts
• Stores identical files as hard links on the file system to save space
• Has ACL support on a per host basis (Would integrate well with our current Account system or a future one)
• Supports RSYNC over SSH (we won't have to install client software on the machines to get it to work)
• Perl based

Cons -

• Perl based :-D
• Probably slower than other solutions
• Not currently packaged for Fedora

rdiff-backup

Pros -

• uses ssh
• easily does differential backups
• written in python
• uses rsync
• supports ACL's
• no need for any daemons (
• command line restores

Cons -

• no pretty configuration setup
• command line restores
• rdiff-backup needs to be installed on all machines

rsync and scripts

Pros -

• simple to setup
• currently in place
• can use ssh

Cons -

• harder to scale
• simple to mis-configure and waste disk space