This project backs up files from the computers in my house. If you have one or more computers that you'd like to keep backed up regularly, flashback might be right for you.
I have a "Pogo Plug" dedicated to doing backups. It's a $12 embedded Linux mini-machine that is about as powerful as a Raspberry Pi. I attached a USB hard disk to it -- that's where the backup files are stored.
Flashback is a Python script that maintains a list of all of the computers and disk volumes that you want to back up. Every ten minutes, it wakes up and goes through the list to see which ones are due for backing up. It tries to ping those machines, and then it tries to do an rsync (an incremental backup with very a efficient copy algorithm). It uses some of rsync's clever "hard link" options so that each daily backup is a complete copy of the backed up disk volume, but files that do not change from day to day are only stored on the disk once.
Although Flashback was designed to be run on a dedicated machine, there is no reason why it can not be run on a normal Linux server. Since it runs OK on the ridiculously underpowered Pogo Plug, it should be fine on a normal Linux desktop or server.
Flashback is suited for making regular backups of many computers in a home or a small office, or even for backing up remote servers.
Flashback can do periodic backups of any size... hourly, daily, weekly, fortnightly, whatever. It can even do more than one period. For example, I back my laptop up daily for 10 days, but I also keep 8 weekly backups. This is done in a super-simple way... I just set up a weekly job that backs up the files from the most recent daily backup.
A word of warning... Flashback is a "no frills" backup package. It takes its orders from a few config files, and it writes its status to two output files. There is no database, no user interface, no web server, nothing. Just those text files.
So note: flashback does not have a fancy user interface for setting up jobs, checking on statuses, or recovering files. You set up jobs by editing a few config files. You check on the status by looking at two status files (via logging in and using 'cat', or by browser if you share those files using a web server). You restore files using rsync.
If you are not comfortable with the shell and command line programs, then flashback is probably not for you.
If you're comfortable with plain text files...
I share those status files via a web server running on the Pogo Plug, so I can check on the status by popping open a browser (even on my phone) and just looking at the current status, in all of its plain ASCII glory.
I also have a script on another machine that reads those status files (using wget), and displays the status on a scrolling LED sign in my office. I have another script that read that status file (again, using wget) and it sends me an email if anything looks out-of-place (this one is included in the git repo).
Flashback has saved my bacon several times. I've had hard disks die suddenly, and have used the very recent backups to recover. I have made mistakes that messed up a server's filesystem, and used the backups to rebuild the server from scratch. I have even used flashback to remotely access recent copies of files from a machine that is currently powered off (wife's laptop is at home, powered off, and I am away from home).
Running on a small (and cheap) (and silent) computer like a Pogo Plug, this little dedicated program just persistently and consistently backs up everything that I care about. When I get home and open up my laptop, flashback notices and starts backing up anything that has changed recently.
Flashback does not do "off-site" backups. That is a much bigger problem than on-site backups, because you have to really think about how much bandwidth you'll be using... especially if you're pushing backups out of your house (where you typically have a very small upload pipe).
That being said, I use flashback to "pull" backups from remote places like web servers in professional data centers, and store them at my house.
With some careful planning, you could also set up flashback to copy to a remote site. But the bottleneck is not going to be flashback itself. It'll be in how much data you're sending upstream.
Flashback started as a python wrapper around 'rsback', which itself is a Perl wrapper around 'rsync'. Rsback does a really nice job of using rsync's hard linking to save multiple backups of a volume without wasting space on files that stay the same day after day. For more info on that feature, see the 'link-dest' option in the man page for rsync.
Rsback is meant to be run as a cron job, but I wanted a daemon that would check every so often (10 minutes) and back up anything that was at least 'so old' (a day). So early versions of flashback simply looped forever, and when it was time to do a backup, it would create a temporary rsback config file and then call rsback. After a while, I decided that it was silly to be setting up rsback for these temporary jobs when I could just call rsync myself. I do owe a lot to the rsback project, since it introduced me to rsync's link-dest capability.
This has been a very fun project, and it is useful as well. I hope you enjoy it. If you have questions or comments, please contact me.
Alan Porter ([email protected])