I wrote a bash script that starts a number of different widgets (various Rails applications) and runs them in the background. I'm now trying to write a complimenting stop script that kills each of the processes started by that start script, but I'm not sure of the best way to approach it.
Following is my start script:
#!/bin/bash
widgets=( widget1 widget2 widget3 ) # Specifies, in order, which widgets to load
port=3000
basePath=$("pwd")
for dir in "${widgets[#]}"
do
cd ${basePath}/widgets/$dir
echo "Starting ${dir} widget."
rails s -p$port &
port=$((port+1))
done
If possible, I was trying to avoid saving the PIDs to a .pid file because they're horribly unreliable. Is there a better way to approach this?
One possibility is to use pkill with the -f switch that is described thus in the man page:
-f The pattern is normally only matched against the process name. When -f is set, the full command line is used.
Hence, if you want to kill rails s -p3002, you can proceed as follows:
pkill -f 'rails s -p3002'
To keep extra dependencies at a minimum and ensure I didn't shut down rails instances that don't belong to me, I ended up going with the following:
Start Script
#!/bin/bash
widgets=( widget1 widget2 widget3 ) # Specifies, in order, which widgets to load
port=3000
basePath=$("pwd")
pidFile="${basePath}/pids.pid"
if [ -f $pidFile ];
then
echo "$pidFile already exists. Stop the process before attempting to start."
else
echo -n "" > $pidFile
for dir in "${widgets[#]}"
do
cd ${basePath}/widgets/$dir
echo "Starting ${dir} widget."
rails s -p$port &
echo -n "$! " >> $pidFile
port=$((port+1))
done
fi
Stop Script
#!/bin/bash
pidFile='pids.pid'
if [ -f $pidFile ];
then
pids=`cat ${pidFile}`
for pid in "${pids[#]}"
do
kill $pid
done
rm $pidFile
else
echo "Process file wasn't found. Aborting..."
fi
Related
So I've looked up other questions and answers for this and as you can imagine, there are lots of ways to find this. However, my situation is kind of different.
I'm able to check whether a bash script is already running or not and I want to kill the script if it's already running.
The problem is that with the below code, -since I'm running this within the same script- the script kills itself too because it sees a script already running.
result=`ps aux | grep -i "myscript.sh" | grep -v "grep" | wc -l`
if [ $result -ge 1 ]
then
echo "script is running"
else
echo "script is not running"
fi
So how can I check if a script is already running besides it's own self and kill itself if there's another instance of the same script is running, else, continue without killing itself.
I thought I could combine the above code with $$ command to find the script's own PID and differentiate them this way but I'm not sure how to do that.
Also a side note, my script can be run multiple times at the same time within the same machine but with different arguments and that's fine. I only need to identify if script is already running with the same arguments.
pid=$(pgrep myscript.sh | grep -x -v $$)
# filter non-existent pids
pid=$(<<<"$pid" xargs -n1 sh -c 'kill -0 "$1" 2>/dev/null && echo "$1"' --)
if [ -n "$pid" ]; then
echo "Other script is running with pid $pid"
echo "Killing him!"
kill $pid
fi
pgrep lists the pids that match the name myscript.sh. From the list we filter current $$ shell with grep -v. It the result is non-empty, then you could kill the other pid.
Without the xargs, it would work, but the pgrep myscript.sh will pick up the temporary pid created for command substitution or the pipe. So the pid will never be empty and the kill will always execute complaining about the non-existent process. To do that, for each pid in pids, I check if the pid exists with kill -0. If it does, then it is outputted, effectively filtering all nonexistent pids.
You could also use a normal for loop to filter the pids:
# filter non-existent pids
pid=$(
for i in $pid; do
if kill -0 "$i" 2>/dev/null; then
echo "$i"
fi
done
)
Alternatively, you could use flock to lock the file and use lsof to list current open files with filtering the current one. As it is now, I think it will kill also editors that are editing the file and such. I believe the lsof output could be better filtered to accommodate this.
if [ "${FLOCKER}" != "$0" ]; then
pids=$(lsof -p "^$$" -- ./myscript.sh | awk 'NR>1{print $2}')
if [ -n "$pids" ]; then
echo "Other processes with $(echo $pids) found. Killing them"
kill $pids
fi
exec env FLOCKER="$0" flock -en "$0" "$0" "$#"
fi
I would go with either of 2 ways to solve this problem.
1st solution: Create a watchdog file lets say a .lck file kind of on a location before starting the script's execution(Make sure we use trap etc commands in case script is aborted so that .lck file should be removed) AND remove it once execution of script is completed successfully.
Example script for 1st solution: This is just an example a test one. We need to take care of interruptions in the script, lets say script got interrupted by a command or etc then we could use trap in it too, since at that time it would have not been completed but you may need to kick it off again(since last time it was not completed).
cat file.ksh
#!/bin/bash
PWD=`pwd`
watchdog_file="$PWD/script.lck"
if [[ -f "$watchdog_file" ]]
then
echo "Please wait script is still running, exiting from script now.."
exit 1;
else
touch $watchdog_file
fi
while true
do
echo "singh" > test1
done
if [[ -f "$watchdog_file" ]]
then
rm "$watchdog_file"
fi
2nd solution: Take pid of current running shell using $$ save it in a file. Then check if that process is still running come out of script if NOT running then move on to run statements in script.
I have a function in a bash script that runs indefinitely in background and that shall be terminated by running again the same script. It is a sort of switch, when I invoke this script it starts or kills the function if already running. To do this I use a PID file:
#!/bin/bash
background_function() {
...
}
if [[ ! -s myscript.pid ]]
then
background_function &
echo $! > myscript.pid
else
kill $(cat myscript.pid) && rm myscript.pid
fi
Now, I would like to avoid multiple instances running and race conditions. I tried to use flock and I rewrote the above code in this way:
#!/bin/bash
background_function() {
...
}
exec 200>myscript.pid
if flock -n 200
then
background_function &
echo $! > myscript.pid
else
kill $(cat myscript.pid) && rm myscript.pid
fi
In doing so, however, I have a lock on the pid file but every time I launch the script again the pid file is rewritten by exec 200>myscript.pid and therefore I am unable to retrieve the PID of the already running instance and kill it.
What can I do? Should I use two different files, a pid file and a lock file? Or would it be better to implement other lock mechanisms by using mkdir and touch? Thanks.
If an echo $$ is atomic enough for you, you could use:
echo $$ >> lock.pid
lockedby=`head -1 lock.pid`
if [ $$ != $lockedby ] ; then
kill -9 $lockedby
echo $$ > lock.pid
echo "Murdered $lockedby because it had the lock"
fi
# do things in the script
rm lock.pid
So, I want to make a bash script, and I'm going to have it run on boot, but I'd like to update the script if I need to and run it without a reboot, so what I want to do is make the script check if there is any other instances of it running when it is loaded, and terninate any instances of the script other than itself. I want it to check instances of bash and get the path of the scripts that are being ran and kill any instances of scripts that have the same path name as it's own. How can I do this?
Example: If I am in directory /foo/bar and I run the script ../tball/script.sh, it will kill any instances of bash that are running the script /foo/tball/script.sh if they exist.
Here's the basis
kill_others() {
local mypid=$$ # capture this run's pid
declare pids=($(pgrep -f ${0##*/} # get all the pids running this script
for pid in ${pids[#]/$mypid/}; do # cycle through all pids except this one
kill $pid # kill the other pids
sleep 1 # give time to complete
done
}
declare -i count=0
while [[ $(pgrep -f ${0##*/}|wc -l) -gt 1 ]]; do
kill_outhers
((++count))
if [[ $count -gt 10 ]]; then
echo "ERROR: can't kill pids" >&2
exit 1
fi
done
The best approach is a file containing the PID of the process in a volatile filesystem like this:
echo $$ > /run/script.pid
You could refine it further by checking if that PID exists with:
if [ ! -d /proc/$(< /run/script.pid) ] ; then
rm /run/script.pid
fi
In your script you should have something like this, to remove the file on exit or if it receives a signal that kills the process:
trap "rm -f /run/script.pid" EXIT INT QUIT TERM
EDIT: Or you could append the PID to a well known pathname and kill all instances of the script with something like this before saving the PID:
kill $(< /run/script.pid) ; sleep 10 ; kill -9 $(< /run/script.pid)
I am having a hell of a time trying to write a "kill all other daemon processes" function for use within a bash daemon. I do not ever want more than one daemon running at once. Any suggestions? This is what I have:
#!/bin/bash
doService(){
while
do
something
sleep 15
done
}
killOthers(){
otherprocess=`ps ux | awk '/BashScriptName/ && !/awk/ {print $2}'| grep -Ev $$`
WriteLogLine "Checking for running daemons."
if [ "$otherprocess" != "" ]; then
WriteLogLine "There are other daemons running, killing all others."
VAR=`echo "$otherprocess" |grep -Ev $$| sed 's/^/kill /'`
`$VAR`
else
WriteLogLine "There are no daemons running."
fi
}
killOthers
doService
It works some of the time, it doesn't others. There is almost nothing consistent.
You've already eliminated the current process ID using grep -v so there's no reason to do it again when you issue the kill. There's also no reason to build the kill in a variable. Just do:
kill $otherprocess
But why not just use:
pkill -v $$ BashScriptName
or
pkill -v $$ $0
without any grep.
Then you can do:
if [[ $? ]]
then
WriteLogLine "Other daemons killed."
else
WriteLogLine "There are no daemons running."
fi
Could you try the old 'lock file' trick here? Test for a file: if it doesn't exists, create it and then startup; otherwise exit.
Like:
#!/bin/bash
LOCKFILE=/TMP/lockfile
if [ -f "$LOCKFILE" ]; then
echo "Lockfile detected, exiting..."
exit 1
fi
touch $LOCKFILE
while :
do
sleep 30
done
rm $LOCKFILE # assuming an exit point here, probably want a 'trap'-based thing here.
The downside is you have to clean-up lock-files from time to time, if an orphan is left behind.
Can you convert this to a 'rc' (or S*/K* script ?) so you can specify 'once' in the inittab (or equivalent method - not sure on MacOS) ?
Like what is described here:
http://aplawrence.com/Unixart/startup.html
EDIT:
Possibly this Apple Doc might help here:
http://developer.apple.com/mac/library/DOCUMENTATION/MacOSX/Conceptual/BPSystemStartup/Articles/StartupItems.html
If you run your service under runit — the service mustn't fork into the background — you'll have a guarantee there is exactly one instance of it running. runit starts the service if it isn't running or if it quit or crashed, stops it if you ask, keeps a pidfile around.
I have a script that runs every 15 minutes but sometimes if the box is busy it hangs and the next process will start before the first one is finished creating a snowball effect. How can I add a couple lines to the bash script to check to see if something is running first before starting?
You can use pidof -x if you know the process name, or kill -0 if you know the PID.
Example:
if pidof -x vim > /dev/null
then
echo "Vim already running"
exit 1
fi
Why don't set a lock file ?
Something like
yourapp.lock
Just remove it when you process is finished, and check for it before to launch it.
It could be done using
if [ -f yourapp.lock ]; then
echo "The process is already launched, please wait..."
fi
In lieu of pidfiles, as long as your script has a uniquely identifiable name you can do something like this:
#!/bin/bash
COMMAND=$0
# exit if I am already running
RUNNING=`ps --no-headers -C${COMMAND} | wc -l`
if [ ${RUNNING} -gt 1 ]; then
echo "Previous ${COMMAND} is still running."
exit 1
fi
... rest of script ...
pgrep -f yourscript >/dev/null && exit
This is how I do it in one of my cron jobs
lockfile=~/myproc.lock
minutes=60
if [ -f "$lockfile" ]
then
filestr=`find $lockfile -mmin +$minutes -print`
if [ "$filestr" = "" ]; then
echo "Lockfile is not older than $minutes minutes! Another $0 running. Exiting ..."
exit 1
else
echo "Lockfile is older than $minutes minutes, ignoring it!"
rm $lockfile
fi
fi
echo "Creating lockfile $lockfile"
touch $lockfile
and delete the lock file at the end of the script
echo "Removing lock $lockfile ..."
rm $lockfile
For a method that does not suffer from parsing bugs and race conditions, check out:
BashFAQ/045 - How can I ensure that only one instance of a script is running at a time (mutual exclusion)?
I had recently the same question and found from above that kill -0 is best for my case:
echo "Starting process..."
run-process > $OUTPUT &
pid=$!
echo "Process started pid=$pid"
while true; do
kill -0 $pid 2> /dev/null || { echo "Process exit detected"; break; }
sleep 1
done
echo "Done."
To expand on what #bgy says, the safe atomic way to create a lock file if it doesn't exist yet, and fail if it doesn't, is to create a temp file, then hard link it to the standard lock file. This protects against another process creating the file in between you testing for it and you creating it.
Here is the lock file code from my hourly backup script:
echo $$ > /tmp/lock.$$
if ! ln /tmp/lock.$$ /tmp/lock ; then
echo "previous backup in process"
rm /tmp/lock.$$
exit
fi
Don't forget to delete both the lock file and the temp file when you're done, even if you exit early through an error.
Use this script:
FILE="/tmp/my_file"
if [ -f "$FILE" ]; then
echo "Still running"
exit
fi
trap EXIT "rm -f $FILE"
touch $FILE
...script here...
This script will create a file and remove it on exit.