How can I repeat a command every interval of time , so that it will allow me to run commands for checking or monitoring directories ?

There is no need for a script, i need just a simple command to be executed in terminal.


8 Answers 8


You can use watch command, watch is used to run any designated command at regular intervals.

Open Terminal and type:

watch -n x <your command>

change x to be the time in seconds you want.

For more help using the watch command and its options, run man watch or visit this Link.

For example : the following will list, every 60s, on the same Terminal, the contents of the Desktop directory so that you can know if any changes took place:

watch -n 60 ls -l ~/Desktop
  • 38
    +1 but be careful when using expansions. For example, try the difference between watch -n 1 'echo $COLUMNS' and watch -n 1 echo $COLUMNS when resizing your terminal - the former is expanded every second, but the latter is expanded only once before watch starts.
    – l0b0
    Mar 6, 2014 at 16:06
  • Problem I have with this solution is that you can't run watch as a process and just leave it running in the background (for example with &disown)
    – Cestarian
    May 13, 2015 at 2:06
  • 2
    Is there any way to use watch with "history enabled" type command? I love using watch, but sometimes I'd prefer to see a log of previous executions as well, instead of just the last one. And yes, I know I can use scripting (while true) to accomplish this, but using the watch utilitiy is so much cleaner!
    – rinogo
    Sep 14, 2015 at 18:44
  • this worked for simpler commands but with pipelined commands chaining this didn't work for me.. following was the command I tried =>cat api.log | grep 'calling' | wc -l Sep 29, 2016 at 7:43
  • 2
    -bash: watch: command not found, macOS 10.14.1 Nov 27, 2018 at 4:22

You can also use this command in terminal, apart from nux's answer:

while true; do <your_command>; sleep <interval_in_seconds>; done


while true; do ls; sleep 2; done

This command will print the output of ls at an interval of 2 sec.

Use Ctrl+C to stop the process.

There are few drawbacks of watch:

  • It cannot use any aliased commands.
  • If the output of any command is quite long, scrolling does not work properly.
  • There is some trouble to set the maximum time interval beyond a certain value.
  • watch will interpret ANSI color sequences passing escape characters using -c or --color option. For example output of pygmentize will work but it will fail for ls --color=auto.

In the above circumstances this may appear as a better option.

  • watch exists for that, this is a bit useless I would say Mar 6, 2014 at 16:36
  • 17
    I am not claiming this answer is to be used at first place. watch is good in most cases. That is why I mentioned "apart from nux's answer" at the beginning. But there are few problems with watch for example One can not use any aliased commands with watch. Take for example ll which is aliased to ls -laF but can not be used with watch. Also in case if the output of any command is quite long you will be in trouble in scrolling using watch. In these few special cases this answer may appear a better option.
    – sourav c.
    Mar 6, 2014 at 17:25
  • 9
    while sleep x is better - it's easier to kill.
    – d33tah
    Mar 8, 2014 at 15:43
  • 7
    This is a nice alternative and, unlike watch, it keeps the command history. Feb 14, 2017 at 18:31
  • 1
    Another weirdness with watch is that if I resize my terminal window, it runs the command again. Sep 25, 2019 at 20:14

Just wanted to pitch in to sourav c.'s and nux's answers:

  1. While watch will work perfectly on Ubuntu, you might want to avoid that if you want your "Unix-fu" to be pure. On FreeBSD for example, watch is a command to "snoop on another tty line".

  2. while true; do command; sleep SECONDS; done also has a caveat: your command might be harder to kill using Ctrl+C. You might prefer:

    while sleep SECONDS; do command; done

    It's not only shorter, but also easier to interrupt. The caveat is that it will first sleep, then run your command, so you'll need to wait some SECONDS before the first occurrence of the command will happen.

  • Why exactly does it matter where you put sleep in the while loop? I couldn't find any difference, Ctrl+C broke the loop instantly no matter what.
    – dessert
    Nov 17, 2017 at 8:03
  • @dessert: depends on what you're trying to break out from I guess. Normally, ctrl+c would just kill your command and sleep and only break if you kill true.
    – d33tah
    Dec 11, 2017 at 17:08
  • This worked, "watch" just sat there. I even made a script to remove the alias'd parts. no dice. The while statement is simple and gives what i need for monitoring when a cron job finishes.
    – dustbuster
    Sep 19 at 14:06

Sounds like the ideal task for the cron daemon which allows for running periodic commands. Run the crontab -e command to start editing your user's cron configuration. Its format is documented in crontab(5). Basically you have five time-related, space-separated fields followed by a command:

The time and date fields are:

       field          allowed values
       -----          --------------
       minute         0-59
       hour           0-23
       day of month   1-31
       month          1-12 (or names, see below)
       day of week    0-7 (0 or 7 is Sunday, or use names)

For example, if you would like to run a Python script on every Tuesday, 11 AM:

0 11 * * 1 python ~/yourscript.py

There are also some special names that replace the time, like @reboot. Very helpful if you need to create a temporary directory. From my crontab (listed with crontab -l):

# Creates a temporary directory for ~/.distcc at boot
@reboot ln -sfn "$(mktemp -d "/tmp/distcc.XXXXXXXX")" "$HOME/.distcc"
  • 5
    The question asks how to run something periodically in the terminal. cron runs behind the scenes rather than in the terminal Dec 15, 2014 at 19:22

you can use crontab. run the command crontab -e and open it with your preferred text editor, then add this line

*/10 * * * *  /path-to-your-command

This will run your command every 10 minutes

* */4 * * *  /path-to-your-command

This will run your command every 4 hours

Another possible solution

$ ..some command...; for i in $(seq X); do $cmd; sleep Y; done

X number of times to repeat.

Y time to wait to repeat.

Example :

$ echo; for i in $(seq 5); do $cmd "This is echo number: $i"; sleep 1;done
  • Why is this an improvement? You just added an extra, needless, step by saving the command as a variable. The only things this does is i) make it longer to type ii) forces you to use only simple commands, no pipes or redirects etc.
    – terdon
    Jul 3, 2014 at 11:52
  • Remove the extra variable
    – Maythux
    May 14, 2015 at 11:24

If you are monitoring the file system, then inotifywait is brilliant and certainly adds less load on your system.

Example :

In 1st terminal type this command :

$ inotifywait .

Then in 2nd terminal, any command that affects the current directory,

$ touch newfile

Then in original terminal inotifywait will wake up and report the event

./ CREATE newfile2

Or in a loop

$ while true ; do inotifywait . ; done
Setting up watches.  
Watches established.
./ OPEN newfile2
Setting up watches.  
Watches established.
./ OPEN newfile2
Setting up watches.  
Watches established.
./ DELETE newfile
Setting up watches.  
Watches established.
./ CREATE,ISDIR newdir
Setting up watches.  
Watches established.
  • the user told you , no script , and maybe he dont want to monitor anything
    – nux
    Mar 6, 2014 at 16:36
  • 3
    I didn't tell him to write a script, I suggested that if they are looping inorder to watch for particular filesystem event, then inotifywait is useful, and uses less resources than repeating a command. I often run several commands on a command line eg grep something InALogFile|less is that a script ?
    – X Tian
    Mar 6, 2014 at 16:43
  • its a good answer , try to edit it to look more simple .
    – nux
    Mar 6, 2014 at 16:45
  • 3
    What could be simpler than . I can't leave out the command.
    – X Tian
    Mar 6, 2014 at 16:48
  • Thanks @XTian, a great command. I also now saw in the man page that you can add -m to continually monitor without a loop.
    – yoniLavi
    Jul 21, 2014 at 16:06

You can create your own repeat command doing the following steps; credits here:

First, open your .bash_aliases file:

$ xdg-open ~/.bash-aliases

Second, paste these lines at the bottom of the file and save:

repeat() {
while [ $(( n -= 1 )) -ge 0 ]

Third, either close and open again your terminal, or type:

$ source ~/.bash_aliases

Et voilà ! You can now use it like this:

$ repeat 5 echo Hello World !!!


$ repeat 5 ./myscript.sh
  • there is a small typo in the line xdg-open ~/.bash-aliases. it should be: xdg-open ~/.bash_aliases (ie: underscore)
    – ssinfod
    Feb 4, 2018 at 1:21

Another concern with the "watch" approach proposed above is that it does display the result only when the process is done. "date;sleep 58;date" will display the 2 dates only after 59 seconds... If you start something running for 4 minutes, that display slowly multiple pages of content, you will not really see it.

On the other hand, the concern with the "while" approach is that it doesn't take the task duration into consideration.

while true; do script_that_take_between_10s_to_50s.sh; sleep 50; done

With this, the script will run sometime every minutes, sometime might take 1m40. So even if a cron will be able to run it every minutes, here, it will not.

So to see the output on the shell as it's generated and wait for the exact request time, you need to look at the time before, and after, and loop with the while.

Something like:

while ( true ); do
  echo Date starting `date`
  before=`date +%s`
  sleep `echo $(( ( RANDOM % 30 )  + 1 ))`
  echo Before waiting `date`
  after=`date +%s`
  DELAY=`echo "60-($after-$before)" | bc`
  sleep $DELAY
  echo Done waiting `date`

This will output this:

As you can see, the command runs every minutes:

Date starting Mon Dec 14 15:49:34 EST 2015
Before waiting Mon Dec 14 15:49:52 EST 2015
Done waiting Mon Dec 14 15:50:34 EST 2015

Date starting Mon Dec 14 15:50:34 EST 2015
Before waiting Mon Dec 14 15:50:39 EST 2015
Done waiting Mon Dec 14 15:51:34 EST 2015

So just replace the "sleep echo $(( ( RANDOM % 30 ) + 1 ))" command with what ever you want and that will be run, on the terminal/shell, exactly every minute. If you want another schedule, just change the "60" seconds with what ever you need.

Shorter version without the debug lines:

while ( true ); do
  before=`date +%s`
  sleep `echo $(( ( RANDOM % 30 )  + 1 ))` # Place you command here
  after=`date +%s`
  DELAY=`echo "60-($after-$before)" | bc`
  sleep $DELAY
  • Replying to myself... If your command takes more than the delay you configure, $DELAY will get a negative value and the sleep command will fail so the script will restart right away. Need to be aware of that.
    – jmspaggi
    Dec 14, 2015 at 21:06

You must log in to answer this question.