8

I'm trying to write a simple script for killing a process. I've already read Find and kill a process in one line using bash and regex so please don't redirect me to that.

This is my code:

LINE=$(ps aux | grep '$1')
PROCESS=$LINE | awk '{print $2}'
echo $PROCESS
kill -9 $PROCESS

I want to be able to run something like

sh kill_proc.sh node and have it run

kill -9 node

But instead what I get is

kill_process.sh: line 2: User: command not found

I found out that when I log $PROCESS it is empty. Does anyone know what I'm doing wrong?

Community
  • 1
  • 1
dopatraman
  • 13,416
  • 29
  • 90
  • 154

2 Answers2

4
PROCESS=$(echo "$LINE" | awk '{print $2}')

or

PROCESS=$(ps aux | grep "$1" | awk '{print $2}')

I don't know why you're getting the error you quoted. I can't reproduce it. When you say this:

PROCESS=$LINE | awk '{print $2}'

the shell expands it to something like this:

PROCESS='mayoff  10732 ...' | awk '{print $2}'

(I've shortened the value of $LINE to make the example readable.)

The first subcommand of the pipeline sets variable PROCESS; this variable-setting command has no output so awk reads EOF immediately and prints nothing. And since each subcommand of the pipeline runs in a subshell, the setting of PROCESS takes place only in a subshell, not in the parent shell running the script, so PROCESS is still not set for later commands in your script.

(Note that some versions of bash can run the last subcommand of the pipeline in the current shell instead of in a subshell, but that doesn't affect this example.)

Instead of setting PROCESS in a subshell and feeding nothing to awk on standard input, you want to feed the value of LINE to awk and store the result in PROCESS in the current shell. So you need to run a command that writes the value of LINE to its standard output, and connects that standard output to the standard input of awk. The echo command can do this (or the printf command, as chepner pointed out in his answer).

rob mayoff
  • 375,296
  • 67
  • 796
  • 848
2

You need to use echo (or printf) to actually put the value of $LINE onto the standard input of the awk command.

LINE=$(ps aux | grep "$1")
PROCESS=$(echo "$LINE" | awk '{print $2}')
echo $PROCESS
kill -9 $PROCESS

There's no need use LINE; you can set PROCESS with a single line

PROCESS=$(ps aux | grep "$1" | awk '{print $2}')

or better, skip the grep:

PROCESS=$(ps aux | awk -v pname="$1" '$1 ~ pname {print $2}')

Finally, don't use kill -9; that's a last resort for debugging faulty programs. For any program that you didn't write yourself, kill "$PROCESS" should be sufficient.

chepner
  • 497,756
  • 71
  • 530
  • 681
  • A bit unrelated but... I'm not able to actually kill the process. My script is no longer throwing errors, but the process is not getting killed even when run with `sudo`. thoughts? – dopatraman Aug 27 '14 at 19:53
  • Put `set -x` at the top of the script to get a better idea of what it's doing. This command makes the shell echo each command before running it. – rob mayoff Aug 27 '14 at 19:57
  • @dopatraman show the full process line. Contains an `Z` in the `STAT` field? (Z = Zombie - unkillable). Or maybe, you want kill a process what waits for some kernel I/O. – clt60 Aug 27 '14 at 19:57
  • @chepner the `kill "$PROCESS"` will kill a processes, when the `$PROCESS` will contain multiple `PIDs`? (e.g. the grep founds more processes)... In this case IMHO the quotes are not needed. – clt60 Aug 27 '14 at 20:05
  • Good point. I rarely write code where I would use a single parameter to hold a series of values that I want to split on whitespace. – chepner Aug 27 '14 at 20:14