I've got a CI server with a command-line interface that allows me to remotely kick-off a job (jenkins CI server and the jenkins-cli.jar tool).

After I kick the job off I tail -f the log (sorry for the messy command):

ssh -t my-jenkins-host.com "tail -f \"/var/lib/jenkins/jobs/$job_name/builds/\`ls -ltr /var/lib/jenkins/jobs/$job_name/builds/ | grep '^l' | tail -n 1|awk '{print \$9}'\`/log\""

After the job successfully completes, usually after at least 5 minutes, I get the following line on the output:

Finished: SUCCESS

Is there a good way to stop tailing the log at this point? i.e. is there like a tail_until 'some line' my-file.log command?

BONUS: extra credit if you can supply an answer that returns 0 when SUCCESS is matched, 1 when FAILURE is matched, and your solution works on mac! (which i believe is bsd based)


You can pipe the tail -f into sed, telling it to quit when it sees the line you're searching for:

tail -f /path/to/file.log | sed '/^Finished: SUCCESS$/ q'

sed will output each line it processes by default, and exit after it sees that line. The tail process will stop when it tries to write the next line and sees its output pipe is broken

  • booh yea! perfect. ...so by any chance is there a way to exit 0 if i match one thing (say 'SUCCESS') and 1 if i match something else (like maybe 'FAILURE')? – aaronstacy Aug 20 '12 at 21:12
  • 7
    @aaronstacy If you're using GNU grep, the q command takes an optional exit code. So the sed command would be sed '/^Finished: SUCCESS$/ q0; /^Finished: FAILURE$/ q1' – Michael Mrozek Aug 20 '12 at 21:14
  • 5
    This might not work if Finished: SUCCESS is the last line of output – lk- Aug 20 '12 at 21:15
  • @Michael Mrozek aaaand of course i'm not b/c i'm using friggin mac – aaronstacy Aug 20 '12 at 21:16
  • Perfecet answer! – Marslo Nov 15 '16 at 14:49
tail -f my-file.log | grep -qx "Finished: SUCCESS"

-q, meaning quiet, quits as soon as it finds a match

-x makes grep match the whole line

For the second part, try

tail -f my-file.log | grep -m 1 "^Finished: " | grep -q "SUCCESS$"

-m <number> tells grep to stop after number matches

and the grep -q exit status will only be 0 if SUCCESS is found at the end of the line

If you want to see all the output, you can't use grep -q, but you can still do

tail -f my-file.log | grep -m 1 "^Finished: "

which does everything except set the exit status to 1 if FAILURE appears.

  • 5
    I used grep in my answer originally, but if he's using tail -f he probably wants to see the file output; grep isn't going to show all the intermediate lines – Michael Mrozek Aug 20 '12 at 22:36

A variation on @Mikel's answer with @Mrozek's comments (I would have replied on the comment but I think i don't have enough privileges yet)

tail -f my-file.log | tee >( grep -qx "Finished: SUCCESS" )

would allow you to use @Mikel's solution and still see the output on the screen

  • Can we add a timeout intervall to this, like: "if its not read between 60 to 120 seconds, then abort the tail and give an error exit code in the shell" ? – kiltek Mar 15 '18 at 8:50

I didn't like any of the answers here, so decided to roll my own. This bash script meets all the criteria, and includes the BONUS for exit of 1 on failure.

while IFS= read -r LOGLINE || [[ -n "$LOGLINE" ]]; do
    printf '%s\n' "$LOGLINE"
    [[ "${LOGLINE}" == "Finished: SUCCESS" ]] && exit 0
    [[ "${LOGLINE}" == "Finished: FAILURE" ]] && exit 1
done < <(timeout 300 tail -f my-file.log)
exit 3

Also included is a timeout feature, which will result in an exit code of 3. If you don't have the timeout command on your system, grab the timeout.sh script from Anthony Thyssen:


Per the comments below, I updated the log print to stop escape character expansion and included all the features of a standard 'read'. See https://stackoverflow.com/a/10929511 for complete 'read' details. The EOF check isn't required here, but is included for completeness.

  • Very nice. Consider using while IFS= read -r LOGLINE to prevent the shell performing whitespace splitting on the lines from tail. – roaima Sep 21 '18 at 20:47
  • @roaima: read doesn't split when there is only one variable (and it's not an array with -a), but you do need -r if the input data contains backslash. But if the input data contains backslash, then echo "$var" may also screw up depending on your shell and/or system, so better printf '%s\n' "$line" – dave_thompson_085 Sep 22 '18 at 3:55
  • @dave_thompson_085 Understanding "IFS= read -r line" – roaima Sep 22 '18 at 6:57

here's a python script that almost does what i want (see caveats below):

import sys,re

def main():
    re_end = re.compile(sys.argv[1])
    re_fail = re.compile(sys.argv[2]) if len(sys.argv) > 2 else None
    for line in sys.stdin:
        if re_end.match(line):
        elif re_fail and re_fail.match(line):

if __name__ == '__main__': main()


  • the lines aren't printed as they come in... they're printed in groups... seems to be some buffering going on

  • i'd have to install this as a script on my path or something, so it's inconvenient, and i'd prefer a slick one-liner :)

  • update: tail seems to do the same buffering, so i'm guessing it's not something that's worth trying to work around. – aaronstacy Aug 20 '12 at 21:10

I had problems with sed and grep and their options, so I write mine one with bash conditions

tail -f screenlog.* | 
while IFS= read line; 
   echo $line; 
   if [[ $line == *Started\ Application* ]]; 
    then pkill tail; 

You also try

 grep -q 'App Started' <(tail -f /var/log/app/app.log)

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.