Backups

For backups, I now use three methods:

• Dropbox for text and PDF files I want on multiple computers.
• Incremental mirroring (using rsync) from disk to disk on a daily basis. I think of this as a manual replacement for RAID (which maintains the ability to recover from unintended deletions--I have used this feature!).
• Subversion for things I want on multiple computers with version control.

Under Linux, I use fwbackups and am quite happy with it. It just backs up the internal hard drive to an external drive once per day early in the morning.

To arrange this, I merely added the script to /etc/cron.daily/.

I also rsync to a remote server (via DSL in the slow direction) as follows:

  rsync --exclude '.svn' -e /usr/bin/ssh -avvuz --del /home/me/ \
    mirror:~/ | perl \
    -e "while (<>) { print unless (m/uptodate/ or m/newer/ or m/hiding/) }"

where 'mirror' is listed in my /etc/hosts file. (Say 'man rsync' to investigate the options used.) The perl command thins the output so that I see only the files that are being updated--there's probably a better way to do this using rsync options.

I used to burn project-specific CD-ROMs once in a while, particularly when leaving a project for a while. However, now that I have servers in multiple states, I don't seem to get around to making CD-ROMs very often. Other command-line tools that people use for backup on Linux systems include rsnapshot (written in perl, based on rsync itself) and rdiff-backup (written in python). A nice feature of rdiff-backup is that it stores compressed deltas of changed files. However, I am only interested in keeping the latest version of everything, preferring to rename files with a date embedded in the filename if there is any need for that. I hear rsnapshot is a lot like Time Machine on the Mac, but without the GUI. That is, it performs backups at different levels (hourly, daily, weekly, etc.). According to info on the Web,¹²``... rdiff-backup is closer to a version control tool and rsnapshot closer to a traditional backup solution.''

For Windows, the whole virtual machine file is backed up along with my regular Mac OS X backups. Since it is over 100 GB, I ``exclude'' it most of the time, and in the meantime back up my user data files (Eudora, Quicken, TurboTax) into files shared on the Mac and therefore backed up separately by Mac Time-Machine backups.

Before moving Windows into VMware Fusion, I used to run Beyond Compare nightly on Windows to mirror one disk to another. I sometimes also ran rsync under Cygwin, e.g., using the alias

 alias
  mebkp 'rsync -avu /cygdrive/c/me/ /cygdrive/d/me'

(in my  /.tcshrc file). Note how windows disk drives must be addressed under '/cygdrive' (something that took me a while to figure out). Beyond Compare is great when you want to really study the differences before copying across. I also use it over Samba to compare two different directories on my Linux machine.