i didn't look at the details, but what's about rdiff-backup? ( http://www.nongnu.org/rdiff-backup/ ) wouldn't it be better to use that instead of rsync and provide a nice interface? especially with huge data directories, rsync could run out of memory, rdiff-backup is better.

What's wrong with the inotify available in all recent Linux kernel versions?

http://en.wikipedia.org/wiki/Inotify

Comments from Robert Love, inotify author:

http://blog.rlove.org/2007/10/leopard-review-notes.html

Isnt soft linking directories a option?

As far as I understand it, soft links will not work because of the ability to remove a previous backup without affecting future backups. If a file is removed, the link is removed and the data stays if at least one hard link exists. Soft links do not have this restriction. If a file is removed the data will (eventually) be removed regardless if no hard links point to it (regardless of soft links).

This is due to the implementation of soft/hard links (one being a file the other being a filesystem entry).

Does it store changes only (like SVN) or whole changed files?

What's different from rsnapshot? rsnapshot worked great for years... BTW I agree that backing up using rdiff-backup would be a huge advance over Time Machine (backing up images, videos, databases and VMware images isn't practical) because rdiff-backup stores only the changed bits of modified files, instead of full copies.

I was thinking about doing a similar thing, but i started from the other end. I have written some python code that talks to inotify, so you don't have to check the whole directory to see if it's changed, if your interested I can give you a copy, or merge it into your system?

Yeah, from the sound of it, you've basically re-invented RSnapshot. I encourage you to read through that to see if there are features that either you've missed, or that rsnapshot has missed.

We've been using rsnapshot for disk-to-disk backups, and I've been quite pleased so far. I chose that instead of rdiff-backup because it is easy to look at the older versions of the files, no special effort needed. This is real handy when you don't know when exactly a file had been removed, for example.

The efficiency hasn't been an issue, with 67GB of data, we've seeing about 1GB of changes per week.

Yes, this or rsnapshot is less space efficient than rdiff-backup, but if you put the backup destination directory on a filesystem that can do data de-duplification at the block level, then you'd have the best of both worlds!!

Someone please tell me there's such a filesystem or block device layer for Linux already.

/djs

So why should I use this instead of TimeVault??

TimeVault? has a daemon backend with a dbus interface, a decent gui, and nautilus integration. It uses inotify, and the author is looking into storing deltas rather than entire files (i.e. the difference between rsnapshot/dirvish and rdiff-backup). It's also written in Python and GPL-licensed.

https://wiki.ubuntu.com/TimeVault

Sounds cool sciurus, im trying it out... never heard of it, then again i switched from RPM-Hell based distros about 4 months ago and am playing catchup

err dude :

http://dirvish.org

I use it at a number of sites... basically You should perhaps consider at this point just writing a gui for it .... would be nicer to see replication of efforts and dirvish has been around for yonks.

ZFS file system... It is in Leopard and allows unlimited constant time snapshots and clones. It is also in FreeBSD ;-)

Another vote for an rdiff-backup GUI, I'll even help. I use rdiff-backup all the time, it's great, very little wasted space, all the meta-data you should need for all "Time Machine"-like functions.

Just another option for the back-end : LBackup http://www.lucidsystems.org/lbackup

Since Time Machine is implemented on an external storage, why not use Ext3Cow? as a backend in linux.

I'm writing a simple python program that permit me to sync via svn (pysvn) the notebook with the server and the server with many other clients. With svn I can have my working copy everywhere I am and a trusted backup on the server.Optionally with some other tool like websvn or svnserve I can get my data from all over the world just with a browser web. Yes, implementing a client/server architecture of this kind is not for all people, but I think it's more flexible than rsync. I can manage more local copy (connected or disconnected) and a centralized backup.

Quick question - how does the automated scheduling get done? I can't see anything in my crontab. Also, is it possible to see the output scheduled runs? Thanks, Alex

OK, found out some more info. the scheduled jobs are written to the root crontab.

It looks like it expects flyback to be installed to yourhome directory, ie the cron line is looking for the file /home/user/flyback.py

I am running v0.3.3 and the restore button does not work, Also in "preferences->backupSchedule" the "delete backups" section is grayed out. Also when I go to "about" screen it says v0.3.1 Is it supposed to be like this or is something not right with my install?

I take back all I said above. Works good now. V.4 is awesome!!! Keep up the great work.

Never mind the criticism. I have seen some of the proposed alternatives fo Flyback and, while I am sure they are handy for experienced users and sysadmins, Flyback seems like a huge thing to have for the average user (perhaps competing there with TimeVault?, but it is good to have two options to choose between them and, besides, TimeVault? is more Nautilus-oriented).

I like the simplicity that is behind Flyback. Good work!

I have tried TimeVault? and it is good but doesnt work for my senario. TimeVault? seems to use notifications of when files change and then backs them up. This wont work for me because I use laptop with external harddrive for storage and dont alaways have it connected when a file changes so the file wont get backed up. Worse yet is when i do plug my external drive back in it wont get backed up either since it is notification based.

Flyback is schedule based and looks for changes when it runs. This works for me VERY well because if my external drive isn't connected it wont get backed up but as soon as I plug it in it will get backed up the next time it runs.

Plus the interface is much easier to find and restore using flyback ;-)

GREAT WORK!!

and compressing backups, (not just deltas, original files too) would be good. rdiff ( http://linux.die.net/man/1/rdiff ) or librsync may be help

I want flyback to work for me.I did a back up 2 weeks ago of about 25GB and when I run a back up now my 60GB external drive fills up and I have only added emails and some music and work files. It seems that if Flyback was hard linking to unchanged files (incremental backup) I wouldn't have this issue. Maybe I added a couple GBs of data but not more than 25GB! Am I missing something? I'll run it again and see if there is any difference.

This is a great tool! I would like to suggest implementing an individual backup schedule and storage location for each directory backup. For example, I might want to backup /home/user/scripts daily, but /home/user/media only monthly. I might also want to backup one in /home/backup and the other in a remote machine. I believe this would be very useful for many people. Great tool, anyway.

I second the comment of noboruis of 1 Dec, especially the latter point about backing up in different places! Martin

In order to enable inotify support for Python applications, you can use http://pyinotify.sourceforge.net/. It is available for Debian-based distribution via python-pyinotify package.

I just heard of Flyback, and I just wanted to say, that after using it for a week or so, I really like it. I replaced my own rsync script, and it certainly works better than that. Keep up the good work. I will look forward to future updates.

hi

it sounds great but, since its a frontend to rsync, why not make it a frontend to other solutions as well? for example amanda and bacula? Im a bacula user and admit that its UI plain sucks, and they are still working on a real GUI, but they do have linux, windows and mac support for the client, so a nice interface would make for a very powerful solution, just a thought, thanks!

I use dirvish, like someone above also mentioned - but I'll look into flyback as soon as my final thesis is over! It's always good to have a nice simple program with a GUI. For people like my mom :-)

Since flyback uses sqlite, couldn't it use inotify to track files that have changed since the last backup in a sortof pending queue, if you will, and then back them up either at the scheduled time or when a device becomes present. For removable storage locations it could use the UUID of the device to recognize it. If it is not available when a scheduled backup time comes around, mark it as pending and keep queueing files via inotify. When the device is attached and the UUID recognized as the backup media, it can initial the backup and flush the queue. Just some thoughts... :)

...or create the backup locally, and move it to the external storage device when it becomes available.

It needs an OpenGL 3D GUI. You could make the 3D GUI optional for people who don't have good graphics cards but I feel like this needs a unique GUI. It doesn't have to looks like Mac OS X's Time Machine, but it needs to look pretty good. Individual backup schedule is also a must. You all should take down these notes of what we're saying and integrate them into this app! With some time and effort, this thing could kick Time Machine's ass.

An fsevents log equivalent for FreeBSD would rock, as would a harvester plugin for flyback (if you will), to avoid the directory walk. Then one really could emulate Time Machine by taking frequent snapshots.

Maybe ZFS can already do this--I haven't studied it.

Where do the preferences live for flyback? Are there any recommended practices for using it on a server machine that hosts multiple user accounts?

the prefs are in gconf

Thank you for this great tool! I installed it for a friend and it is working great!