226

How can I create a new file and fill it with 1 Gigabyte worth of random data? I need this to test some software.

I would prefer to use /dev/random or /dev/urandom.

Improve this question
0

6 Answers 6

Reset to default
333

On most unices:

head -c 1G </dev/urandom >myfile

If your head doesn't understand the G suffix you can specify the size in bytes:

head -c 1073741824 </dev/urandom >myfile

If your head doesn't understand the -c option (it's common but not POSIX; you probably have OpenBSD):

dd bs=1024 count=1048576 </dev/urandom >myfile

Do not use /dev/random on Linux, use /dev/urandom.

Improve this answer
8
  • 2
    Interesting that head can read /dev/urandom, but tail cannot.
    – Stefan Lasiewski
    Mar 12, 2012 at 23:24
  • 34
    @StefanLasiewski tail first tries to go to the end of the input file, which takes forever (literally).
    – Gilles 'SO- stop being evil'
    Mar 12, 2012 at 23:25
  • 7
    @StefanLasiewski You have /dev/zero as well, if you don't like variety.
    – Gilles 'SO- stop being evil'
    Mar 12, 2012 at 23:55
  • 1
    Correction to my last comment: Arithmetic Expansion is POSIX so it doesn’t have to be Bash.
    – Franklin Yu
    Aug 3, 2020 at 2:29
  • 2
    No need for that first < file redirect, this works: head -c 1G /dev/urandom >myfile
    – agc
    Aug 28, 2020 at 12:41
53

Assuming that pseudo-random data is sufficient, dd if=/dev/urandom of=target-file bs=1M count=1000 will do what you want.

dd(1) will read blocks of data from an input file and write them to an output file. The command line language is a little quirky, but it is one of those really useful tools worth mastering the basics of.

In this case if is input file, of is output file, bs is "block size" - and I used the GNU extension to set the size more conveniently. (You can also use 1048576 if your dd doesn't have GNU extension.) count is the number of blocks to read from if and write to of.

/dev/urandom is a better choice than /dev/random becuase, on Linux, it will fall back to strong pseudo-random data rather than blocking when genuinely random data is exhausted.

You may also want to look at http://www.random.org/ as another path to getting some random data without having to generate it yourself.

Improve this answer
3
  • 4
    A note -- unit specifications like 1M are not available in every standard dd variant. If your version of dd is affected, use bs=1048576.
    – Chris Down
    Mar 7, 2012 at 0:39
  • 4
    The example command will create a file of count 1,000,000 blocks each of size 1 MB. That's about 1 TB (1M x 1MB), not 1 GB (which would be 1K x 1MB). As an aside, up to a point, increasing block size and reducing the block count tends to yield better throughput for a given amount of output data.
    – user
    Mar 7, 2012 at 10:45
  • 1
    GB vs GiB. for dd 1M==1024*1024, use 1MB and count=1000 if you want 1 GB or 1M and count=1024 if you want 1 GiB. Or use 1GB or 1G and count=1.
    – 12431234123412341234123
    Sep 7, 2020 at 16:50
10

An alternative to using /dev/random/ and/or /dev/urandom with an excellent Cryptographically Secure PseudoRandom Number Generator (CSPRNG) is given by openssl. You can even choose the algorithm and encrypting method in such rare cases that that is needed:

size=$( echo 1G | numfmt --from=iec )         # 1 G ==> 1073741824

openssl rand -out myfile "$size"

If you need base64:

openssl rand -base64 -out myfile "$size"

Read man rand for the details.

Improve this answer
6 
while true;do head /dev/urandom | tr -dc A-Za-z0-9;done | head -c 5000K | tee  5000kb

Used this to generate 5MB random character data. If you need different size, change the -c value of head, change the outfile name, execute and wait until the execution completes.

Improve this answer
1
  • 1
    Use base64 to increase recall rate (from 62 / 256 to 62 / 64).
    – iBug
    Mar 12, 2020 at 4:26
2

The UNIX tool jot can do exactly that. No need to pipe data from /dev/urandom into a file with dd.

jot -r -c $FILESIZE > $FILE

The flag -r generates random data based on arc4random(3), -c generates a char based on the ASCII table. Other data can be specified based on printf format types with the -w flag.

By default jot returns to stdout.

Edit:

Another use case of jot could be to generate random passwords:

$ jot -rcs '\'\'' 128 33 126
wFv/Fb7F}iO'|!cGy>?8yyez2$f,vy5vYqa;?s$-wBlVU^QX0|y~M.VeRq7}Jh!l/0'CcveuLTL]_1@P_yV#P)h8VZA\~,s{|69P{YwlL?;U{_yC3$\m*:H2V66RpdJ^
$ jot -rcs '\'\'' 128 33 126
X`&kh3[Z2'ej)w`7z'YXl~VG_MA@^[email protected]/i\u)&g1|Y}%8;r>o>e5$UcyX]Q(F=XH6C@-BOTQ!<}CqI1Ff],7#;kb!GpZ^Ai85NaA,ya|YF+EuH]{FUgC0G7;
Improve this answer
1
  • This one? lib.rs/crates/jot. Since it was not available in my system's software repository (Fedora). So cargo install jot?
    – Pysis
    Jul 17 at 15:13
1

The solution doesn't use random but I use man pages to create sizable data. This does not however guarantee the max size so we might have to loop through a few cats to get the desired size,

man curl > /tmp/man-curl

Or if the size required is larger,

for i in {1..10}; do man curl >> /tmp/man-curl; done
Improve this answer

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .