Archives and compression with tar

Bundle a directory into one file, compress it with gzip, bzip2 or xz, and restore it to the right place.

tar
gzip
gunzip
bzip2
bunzip2
xz
tar -t
tar -x
tar -C

40 min5 exercises

After this lesson you can

Create, list and extract tar archives with and without compression.
Pick gzip, bzip2 or xz and know the trade-off between speed and size.
Extract into a chosen directory, extract one file, and avoid archives that unpack into a mess.

The idea

tar (“tape archive”) bundles many files and their permissions, owners and timestamps into one file. It does not compress by itself; compression is a second step that tar can run for you. The three common compressors trade time for size:

Option	Compressor	Extension	Character
`-z`	gzip	`.tar.gz`, `.tgz`	Fast, good enough, everywhere
`-j`	bzip2	`.tar.bz2`	Slower, smaller
`-J`	xz	`.tar.xz`	Slowest, usually smallest

The three verbs to learn are create, table of contents (list) and extract. Add f and the archive’s file name, and v for a file list as it works. On the exam the usual mistake is the order of arguments after -f: the archive name must come immediately after it.

Try it

mkdir -p ~/lab/data && cd ~/lab
echo one > data/a.txt; echo two > data/b.txt; echo three > data/c.txt
tar -cvf data.tar data
tar -tf data.tar

data/
data/b.txt
data/c.txt
data/a.txt
data/
data/b.txt
data/c.txt
data/a.txt

The first four lines come from -v (verbose, while creating) and the last four from listing. tar stores files in the order the file system returns them, so yours may differ.

tar -czf data.tar.gz data
tar -cJf data.tar.xz data
ls -l data.tar*

-rw-r--r--. 1 alex alex 10240 Nov  3 10:15 data.tar
-rw-r--r--. 1 alex alex   186 Nov  3 10:15 data.tar.gz
-rw-r--r--. 1 alex alex   228 Nov  3 10:15 data.tar.xz

Plain tar pads to blocks of 10 KiB, so even three tiny files take 10240 bytes. For a tiny input xz’s fixed overhead makes it larger than gzip. Try the three compressors on /usr/share/doc in the exercises to see the real difference on bigger data.

mkdir -p ~/lab/restore
tar -xf data.tar -C ~/lab/restore
tar -xf data.tar -C /tmp data/b.txt

Compressing single files

gzip file replaces file with file.gz; gunzip file.gz reverses it. The same pattern applies to bzip2/bunzip2 and xz/unxz. Use -k to keep the original, and zcat, bzcat, xzcat to read a compressed file without unpacking it.

seq 1 100000 > ~/lab/numbers.txt
gzip -k ~/lab/numbers.txt
zcat ~/lab/numbers.txt.gz | head -2
ls -l ~/lab/numbers.txt*

1
2
-rw-r--r--. 1 alex alex 588895 Nov  3 10:20 /home/alex/lab/numbers.txt
-rw-r--r--. 1 alex alex 215169 Nov  3 10:20 /home/alex/lab/numbers.txt.gz

Going further

tar detects the compression when extracting, so tar -xf archive.tar.xz works without -J. --exclude='*.log' skips files, -p preserves permissions when you extract as root, and --selinux --xattrs keep SELinux labels and extended attributes, which you want when you back up system directories. A .zip file needs zip and unzip instead.

Practice

0 of 5 done

Do these on your own system. They get harder. Check each result, then tick it off.

Warm-upDone
Create an archive and read it back
Create ~/lab/data/ with three small text files. Archive the directory to ~/lab/data.tar without compression, then list the archive's contents without extracting it.
Check your result
tar -tf ~/lab/data.tar prints data/ and the three files, with names starting data/.
Hint
-c create, -f file name, -t list.
Solution
```
mkdir -p ~/lab/data && cd ~/lab
echo one > data/a.txt; echo two > data/b.txt; echo three > data/c.txt
tar -cf data.tar data
tar -tf data.tar
```
PracticeDone
Compare gzip, bzip2 and xz
Archive /usr/share/doc three times into ~/lab/doc.tar.gz, ~/lab/doc.tar.bz2 and ~/lab/doc.tar.xz, using the matching tar option each time. Show the three sizes with ls -lh and say which is smallest.
Check your result
ls -lh ~/lab/doc.tar.* shows three files. The .xz file is normally the smallest and the .gz the largest.
Hint
-z gzip, -j bzip2, -J xz. Errors about unreadable files can be ignored: add 2>/dev/null.
Solution
```
cd ~/lab
tar -czf doc.tar.gz  /usr/share/doc 2>/dev/null
tar -cjf doc.tar.bz2 /usr/share/doc 2>/dev/null
tar -cJf doc.tar.xz  /usr/share/doc 2>/dev/null
ls -lh doc.tar.*
```
PracticeDone
Extract into another directory
Extract data.tar into a new directory ~/lab/restore so that a.txt, b.txt and c.txt appear directly under ~/lab/restore/data/. Then extract only data/b.txt into /tmp.
Check your result
ls ~/lab/restore/data lists the three files. cat /tmp/data/b.txt prints two.
Hint
-C dir changes to that directory before extracting. Name the member to extract only it.
Solution
```
mkdir -p ~/lab/restore
tar -xf ~/lab/data.tar -C ~/lab/restore
ls ~/lab/restore/data
tar -xf ~/lab/data.tar -C /tmp data/b.txt
cat /tmp/data/b.txt
```
PracticeDone
Archive without the leading path
Create etc-backup.tar.gz containing /etc/hostname and /etc/hosts so that listing it shows etc/hostname and etc/hosts (relative), not /etc/.... Explain why relative names are safer.
Check your result
tar -tzf etc-backup.tar.gz shows names without a leading slash. tar prints Removing leading '/' from member names when it strips them for you.
Hint
By default GNU tar removes the leading / for you. -C / etc/hostname etc/hosts makes the relative form explicit.
Solution
```
cd ~/lab
tar -czf etc-backup.tar.gz -C / etc/hostname etc/hosts
tar -tzf etc-backup.tar.gz
```
ChallengeDone
Back up, compare, restore
Back up your whole ~/lab directory to /tmp/lab-backup.tar.xz. Verify the archive reads without errors, delete ~/lab/data, then restore only that directory from the archive and prove the files are back.
Check your result
tar -tJf /tmp/lab-backup.tar.xz > /dev/null exits with status 0 (echo $? prints 0). After the restore ls ~/lab/data shows the three files again.
Hint
Archive from your home so names start with lab/. To restore one directory name it exactly as it appears in the listing.
Solution
```
cd ~
tar -cJf /tmp/lab-backup.tar.xz lab
tar -tJf /tmp/lab-backup.tar.xz > /dev/null; echo $?
rm -r ~/lab/data
tar -xJf /tmp/lab-backup.tar.xz -C ~ lab/data
ls ~/lab/data
```

The idea

Try it

Compressing single files

Going further

Create an archive and read it back

Compare gzip, bzip2 and xz

Extract into another directory

Archive without the leading path

Back up, compare, restore