18

When I run:

watch 'cmd >> output.txt' &

the job gets suspended by the system:

3569 Stopped (tty output)

Is there a workaround?

26

The purpose of watch is to show the results of a command full-screen and update continuously; if you're redirecting the output into a file and backgrounding it there's really no reason to use watch in the first place.

If you want to just run a command over and over again with a delay (watch waits two seconds by default), you can use something like this:

while true; do
    cmd >> output.txt
    sleep 2
done
  • Is there a way to do this in a non-blocking fashion though? For example if I want to put this and other commands in a single bash script, and have it continue to run the code that comes after this (while still doing this in the background)? – Austin Apr 5 at 21:18
13

Here's a way:

watch -n 1 'date' &>/dev/null &

Since you background the process, we can assume you don't need the terminal display and you are fine redirecting to a file. If you do that you will be able to background watch without issue.

sleep, as suggested by Michael Mrozek, will slowly lag which can be undesirable. Aside from a convoluted shell script that monitors system time and executes a command based upon elapsed time, watch -p can be a good option for precise timings.

For precise timings:

watch -n 1 -p 'date' &>/dev/null &
  • sorry, how does this work? why is it possible to just use & two times? – Blauhirn Dec 16 '17 at 18:52
  • Read the pipeline section of man bash to see how the first & works. ;) – Mioriin Jul 21 '18 at 17:25
11

I'm not sure about your motivations, but maybe this would be enough?

while true; do sleep 2; cmd >>output.txt; done &

Otherwise, please explain why you really need watch.

2

tee is your friend. For example, here's how you could show the number of files in the current directory, and append it to count.txt every 2 seconds:

watch 'find . -maxdepth 1 -printf " " | wc -c | tee --append count.txt'
2

Depends on your motivations:

  • You want to run something periodically. Then you could use directly cron. (If a time granularity of 1 minute is enough for your purposes).

    • Note: If the interval is very short or the device is very constrained, and unless you configure syslog to log less messages from crond, it will fill up your syslog pretty quickly.
  • You want to just run something repeatedly:

    while sleep 2 ; do
      cmd
    done >& log.txt
    

    (You can run this from the crontab for example, although using a script is highly recommended).

  • You want to run periodically a command and have access to the output of its last execution.

    • You can write the output to a ramfs (modifying the script before):

      while sleep 2 ; do
        cmd >& last_execution.log
      done 
      

      or even better:

      while sleep 2 ; do
        cmd >& last_execution.log.tmp 
        mv last_execution.log.tmp last_execution.log ## atomically
      done 
      
    • You can combine screen and watch, so that you can always check the latest status, as long as it fits in one screenful:

      screen watch -n 3660 ./make-backup.sh
      
1

You can use screen command.

http://www.cyberciti.biz/tips/linux-screen-command-howto.html

  • with screen you can create a background terminal session, close it and open when you need it. – Roberto Sep 27 '11 at 12:13
0

If you must use watch - Just combine answers from mike dangelo and l0b0

watch -n 1 'cmd | tee -a output.txt' &>/dev/null &

e.g.

watch -n 1 'date | tee -a output.txt' &>/dev/null &

then watch the seconds go by

tail -f output.txt
0

Watch is designed to output to the screen, but it's output can be redirected (both stdout and stderr) then it will run in the background.

watch 'date >>fa' >/dev/null 2>/dev/null &

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.