What I do in this article is describe a method of doing backups to inexpensive CDR drives under Linux in a fast and portable way. You can restore a system to a useful state in minimal time, with all your data, on any random computer capable of running Linux with a CDROM drive.

The Problem
If you're somebody who actually uses your computer, as opposed to just playing with it, over time you accumulate information of value. Murphy's Law suggests that this information would be difficult, time consuming, or outright impossible to replace if you lost access to it.

You and your information can be separated in any number of ways. You goof up. Hard disks fail. Houses burn down. Computers get stolen. If you're suspected of a computer related crime, kiss your computer goodbye.

To cover these possibilities you need to keep a backups of your data.

The traditional way to backup for many years has been to backup to tape. Historically tape has been relatively inexpensive per megabyte. Access to individual files is not very fast, but it doesn't need to be.

Backup to tape has disadvantages for the individual user.

Tape drives are expensive for the individual computer user. Currently an entry level tape drive (a IDE interface Seagate TR-5 drive) costs ~$250. Better tape drives (ex. SCSI DAT drives, SLR drives and whatnot) cost $600 and up. Way up.

Tape media is also relatively expensive. A single "20GB" (actually 10GB native, an optimistic 20GB compressed) TR-5 tape will cost you about $40 each. 100GB SLR tapes will cost over $100 each. This is fine for a business, pricey for an individual.

There are too many different types of tape. Capacities range from <100MB to at least 100GB. Compatibility problems are the rule. Any given tape drive will usually read (but not necessarily write to) a few slightly older and lower capacity tape formats. To be able to restore from old tapes, you have to keep have a tape drive of the same type around. Over the years I've used a variety of tape drives for backup -- a 40MB Archive, an Irwin Accutrak, a Connor 250, A Conner QIC-3080, A Tandberg quarter inch 1.2GB, Iomega TR1, Seagate TR3 and others. At this date I would have a hard time reading anything from any of those tapes, because I would have to rustle up a functioning compatible drive to read them with.

If my house burns down, if my computer is stolen or confiscated, if I'm hiding out and on the run, if I for any reason loose access to my computer and my backups were on some obscure tape format I would be paddleless up the proverbial creek.

My Current Solution
IDE CDRW drives are currently very common. Half the new computers out there come with one. If you don't have one they are cheap to buy outright ($60 to $90 or so). Media capacity per disk is 650-700 MB, and price per disk can be as low as $.10/disk.

I do not have a huge stash of pornography, MP3 files or warez to save. 10+ years worth of accumulated email, documents, other files and my personal library of text files and images in my case amounts to about 1.2GB of data to be backed up.

For a while I depended on an inelegant script I wrote that would 1) make a tar archive of half of my stuff, 2)make an iso image of the tar file, 3) burn the iso image to CDR using cdrecord, and 4) do it again with the other half.

In looking for something better I found cdbackup. From the manual page: "cdbackup is a utility to make streaming backups to CD-R(W) disks. It's designed to work with any backup tool which writes the backup to stdout (like tar/cpio/afio)." Cdbackup is not itself a backup program. Cdbackup lets a traditional backup utility like tar write directly to the CDR drive.

An Outline of The Gory Details of Getting This To Work

1. If you haven't already, get your CDR drive to work on your Linux box.

   There's a CDR HOWTO out there (see the Linux Documentation Project page at http://www.tldp.org/HOWTO/CD-Writing-HOWTO.html). Read over the first chapter at least. It's somewhat outdated. You should not, for instance, need to recompile your kernel to get SCSI emulation support -- every modern distribution supports it out of the box. Chances are, all you will need to do is add the line 'append=3D"hdc=3Dide-scsi"' (but see the HOWTO for exactly where and the syntax) to the appropriate section of your /etc/lilo.conf file, run lilo to finalize the change, reboot, and load the the ide-scsi module ("insmod ide-scsi").

2. Install cdbackup.

   Under Debian Linux (my current distribution of choice), while logged in as root (the privileged administrative user) run "apt-get install cdbackup". The apt package manager will install the most recent version it knows about, and any package that it depends on (like cdrecord and mkisofs).

   With a Linux that doesn't use the Debian packaging system, download the cdbackup source tar file from http://www.muempf.de/cdbackup.html, unpack it and follow the instructions for installing it. Also, make sure to install the cdrecord and mkisofs packages.

3. Write a script to automate your backup. Compose it in a text editor and save it to disk.

   The script I currently use consists of a single line: "tar -c -v -f - /root /etc /home/lnf | cdbackup -d /dev/scd0 -r 0,0,0 -s 4 -v -l 640 -m".

   What's happening here?

   "tar -c -v -f - /root /etc /home/lnf" creates an tar archive of the root user's home directory (/root), system configuration directory (/etc) and my personal home directory (/home/lnf), and routes it to the standard output.

   The standard output is piped ("|") to. . . .

   the cdbackup program ("cdbackup") which writes it out to the burner ("-d /dev/scd0 -r 0,0,0"), at 4x speed ("-s 4"), using a media 640MB in size (-l 640). "-m" indicates that the backup can be larger than a single disk. "-v" makes it display enough information for you to follow what's going on.

4. Make the script executable ("chmod 700 backupscript" where "backupscript" is whatever file name you saved your script as).

5. Run it as root ("./backupscript").

The far side -- restoring from backup onto a new system
Assume the worse. You've come home. Someone's been there while you were gone. The place is trashed. And you PC is gone. Buy, beg, borrow or cobble together a replacement.

1. Install Linux. Set up any users that need to be set up.

2. Install cdbackup.

   With Debian or a Debian derived Linux run "apt-get install cdbackup". The cdbackup package includes cdrestore. Cdbackup depends on cdrecord, and the package manager will install it even if you don't have a cd burner on the target system. This is OK.

   If your version of Linux does not use the Debian packaging system (ex. Redhat) you will have to download the source code tar file from http://www.muempf.de/cdbackup.html, unpack it and follow the instructions for installing it.

3. While logged in as root, put in the first backup CD and run cdrestore. Ex. "cdrestore -d /dev/cdrom -l 640 -t 1 -m | tar xvf -"

   What's happening here?

   Cdrestore reads track 1 ("-t 1") of a 640MB disk ("-l 640") in the cdrom drive ("-d /dev/cdrom"). "-m" indicates that the archive can be larger than a single disk. The output is piped ("|") to tar which extracts the archive to the current directory ("tar xvf -"). Restore to a temporary directory, not the original location.

4. Move restored files to their proper places. I find Midnight Commander ("mc") to be the easiest program to use for this. If you've got a new system up and running, you will probably not want to blindly overwrite all the configuration files in your /etc directory. Likewise, be careful overwriting the configuration files in your personal directory.

Caveats and advice:

1. Test your backup. Test your ability to restore. 4X works reliably with most brands of CDR on my cheap old 8X Philips burner in my 500MHz system. You'll need to experiment on your own system to see what is reliable.

2. You should test your backup when you change to a different brand of media. I'm currently working through a stack of Maxell 650MB disks that don't seem to work well on my system at 4X, but work fine at 2X.

3. The backup program does not have to be tar. It can be any program that writes to the standard output. I use tar because I'm familiar with it. Tar is also very portable. Tar does not handle compression very well, so I don't use compression.

4. RTFM, always. Read the cdbackup and cdrestore man pages. The exact options I use may not be what you would use on your system. Read the tar info page ("info tar") -- it's more comprehensive and better organized than the man page.

5. Security: It can be as important to keep your data out of the hands of others as to be able to retrieve it yourself.

   Any random individual trying to read a disk made with cdbackup will see nothing on the disk. This provides some small amount of "security through obscurity".

   The cdbackup man page specifically mentions cpio and afio as alternatives to tar. I've experimented a bit with afio. Afio has some interesting capabilities. Among them is the option to encrypt the backup by running files through pgp or gpg (see the example scripts in /usr/share/doc/afio/examples/script3).

6. Watch sale fliers for computer and office supply stores for deals on media. There will often be rebate deals that bring the price per disk down to $.10 or less. I have yet to pay more than $15 for a 50 pack of CDR disks.

7. What to backup, what not to backup.
   • Back up your personal directory (/home/). Back up the system configuration directory (/etc). If you have anything significant in the root user's directory (/root), back it up. If you need to keep system logs, back up /var/log. If you have mail in the mail spool, back up /var/spool/mail.

   • You should not need to back up /bin, /sbin, /usr/bin, /lib, /boot or any other directory off of / that I haven't mentioned. These will get remade with the new OS install.

   • Do not try to backup /dev and /proc. They're special.

Copyright Anno Domini 2003 by "Lee N. Field" under
the terms of the Gnu Free Documentation License.