Tape Archiving

… wat…

Sometimes we have to remember that UNIX was designed in the 70s and 80s when hard drive space was either non-existent or at a premium.

So…

What _is_ the deal

Because it’s used everywhere, and often misunderstood

Never remember the flags?

So why tape archive

All tar really does is take many files and make it into one large file. It sets up some metadata, using actual tape markers to delineate when one file done, and the next starts.

Sounds dumb

Yes. But, you’ve probably done this to a file before, or opened a file like this before.

What’s is a zip file? Well, it’s 1 file made out of many… then shrunk via some algorithm.

tar is the first step of that (and they added the second bit of it, too).

`tar.gz`

When we use tar to make 1 file out of many, we usually create a file with the suffix tar. Files that have been gzipped have .gz. Since these can be both, we have both! Or, sometimes file.tgz, which shows that it has been tar‘d and gzipped.

But, we almost always want to shrink the file, too.

tar czvf file_to_write.tar.gz all the files to add to the tar

tar….

c create
z gZip it
v verbose (makes it kind of fun)
f write to a file. This one is special, and has to be the last one because the file name that you’re writing to has to be after it

Example:

$ tar czvf primer.tar.gz cs-tech-primer
cs-tech-primer/
cs-tech-primer/deployment.md
cs-tech-primer/manifest.json
[... snip...]

Then, undo it:

tar xzvf primer.tar.gz

But I was told I could write to tape

Unix… everything is a file…. That actually works for take drives, too. They are “block devices”, which can be roughly translated to “basically a drive”.

tar czvf /dev/tape_device all the files to add to the tape

So, same thing just different!

Backing up a remote

Now, this is amazing…

tar czvf - some_file will take some_file and tar/gz it … and send the file to…. standard out!

Reminder that ssh can be passed an argument to be run, which could… be the tar command.

So… idea: run tar to collect all your files, make them into 1 large file, and then send it to standard out… then locally pipe that to a file with > file redirection.

A backup script that I use with cron on an actual project:

# create a human-readable date
TIME=`date +%Y-%m-%d_%H-%M-%S`
BACKUP_DIR=~/backups

cd $BACKUP_DIR

# - passes the data to stdout, then we can cimply write that to a file
ssh aviary.cs.umanitoba.ca tar -czf - /folder/to/back/up  > backup_$TIME.tgz

This outputs files that have a beautiful date, doesn’t overwrite old backups, and is generally wonderful.