limiting number of concurrent processes in a bash script - bash

I am trying to use a bash script to do several jobs in parallel. The jobs are memory intensive so I need to control the number that launch at a time. What I have is below, and it broadly works, but sometimes the delay loop is unaware of a job that has just been started, so several extra jobs get launched, causing the system to run out of memory.
Adding a sleep before the while statement in the delay loop reduces this problem, but does not completely eliminate it. Anyone know of a way to cure this. I'm running on Solaris if that's relevant.
#!/bin/bash
delay(){
while [ 8 -le $(ps -ef |grep myjob |wc -l) ]
do
sleep 1
done
}
./myjob -params1 &
delay
./myjob -params2 &
delay
./myjob -params3 &
delay
./myjob -params4 &
delay
.
.
.

GNU parallel utility http://www.gnu.org/software/parallel/ might be the right tool as it can be said more easy to use than xargs

First, I'll give you a stripped down example of something I do in a couple of my linux scripts. This should work on solaris, but I don't have any systems currently to test on. I modified a couple of things that used /proc, so if anything doesn't work let me know.
#!/bin/bash
# set the max # of threads
max_threads=4
# set the max system load
max_load=4
print_jobs(){
# flush finished jobs messages
jobs > /dev/null
for x in $(jobs -p) ; do
# print all jobs
echo "$x"
done
}
job_count(){
cnt=$(print_jobs $1)
if [ -n "$cnt" ]; then
wc -l <<< "$cnt"
else
echo 0
fi
}
cur_load(){
# get the 1 minute load average integer
uptime |sed 's/.*load average[s]*:[[:space:]]*\([^.]*\)\..*/\1/g'
}
main_function(){
# get current job count and load
jcnow=$(job_count)
loadnow=$(cur_load)
# first, enter a loop waiting for load/threads to be below thresholds
while [ $loadnow -ge $max_load ] || [ $jcnow -ge $max_threads ]; do
if ! [ $firstout ]; then
echo "entering sleep loop. load: $loadnow, threads: $jcnow"
st=$(date +%s)
local firstout=true
else
now=$(date +%s)
# if it's been 5 minutes, echo again:
if [ $(($now - $st)) -ge 300 ]; then
echo "still sleeping. load: $loadnow, threads: $jcnow"
st=$(date +%s)
fi
fi
sleep 5s
# refresh these variables for loop
loadnow=$(cur_load)
jcnow=$(job_count)
unset firstout
done
( ./myjob $# ) &
}
# do some actual work
for jobparams in "params1" "params2" "params3" "params4" "params5" "params6" "params7" ; do
main_function $jobparams
done
wait
A couple of caveats:
you should trap signals so you can kill child processes. I do not know how to do this in solaris, but this works for on linux: trap 'echo "exiting" ; rm -f $lockfile ; kill 0 ; exit' INT TERM EXIT
if load climbs while jobs are already running there's no facility to throttle down
If you're not concerned about load at all, this can be a bit simpler:
#!/bin/bash
# set the max # of threads
max_threads=4
print_jobs(){
# flush finished jobs messages
jobs > /dev/null
for x in $(jobs -p) ; do
# print all jobs
echo "$x"
done
}
job_count(){
cnt=$(print_jobs $1)
if [ -n "$cnt" ]; then
wc -l <<< "$cnt"
else
echo 0
fi
}
main_function(){
# get current job count
jcnow=$(job_count)
# first, enter a loop waiting for threads to be below thresholds
while [ $jcnow -ge $max_threads ]; do
if ! [ $firstout ]; then
echo "entering sleep loop. threads: $jcnow"
st=$(date +%s)
local firstout=true
else
now=$(date +%s)
# if it's been 5 minutes, echo again:
if [ $(($now - $st)) -ge 300 ]; then
echo "still sleeping. threads: $jcnow"
st=$(date +%s)
fi
fi
sleep 5s
# refresh these variables for loop
jcnow=$(job_count)
unset firstout
done
( ./myjob $# ) &
}
# do some actual work
for jobparams in "params1" "params2" "params3" "params4" "params5" "params6" "params7" ; do
main_function $jobparams
done
wait

Use xargs to do this. Pass it -n 1 to indicate one parameter per job and use the --max-jobs parameter to specify the number of concurrent processes.

Formulate your script in terms of a makefile and let make -j N sort it out through Parallel Execution.

Related

Does pushing a block of code to background in Bash result in parallelization? [duplicate]

Lets say I have a loop in Bash:
for foo in `some-command`
do
do-something $foo
done
do-something is cpu bound and I have a nice shiny 4 core processor. I'd like to be able to run up to 4 do-something's at once.
The naive approach seems to be:
for foo in `some-command`
do
do-something $foo &
done
This will run all do-somethings at once, but there are a couple downsides, mainly that do-something may also have some significant I/O which performing all at once might slow down a bit. The other problem is that this code block returns immediately, so no way to do other work when all the do-somethings are finished.
How would you write this loop so there are always X do-somethings running at once?
Depending on what you want to do xargs also can help (here: converting documents with pdf2ps):
cpus=$( ls -d /sys/devices/system/cpu/cpu[[:digit:]]* | wc -w )
find . -name \*.pdf | xargs --max-args=1 --max-procs=$cpus pdf2ps
From the docs:
--max-procs=max-procs
-P max-procs
Run up to max-procs processes at a time; the default is 1.
If max-procs is 0, xargs will run as many processes as possible at a
time. Use the -n option with -P; otherwise chances are that only one
exec will be done.
With GNU Parallel http://www.gnu.org/software/parallel/ you can write:
some-command | parallel do-something
GNU Parallel also supports running jobs on remote computers. This will run one per CPU core on the remote computers - even if they have different number of cores:
some-command | parallel -S server1,server2 do-something
A more advanced example: Here we list of files that we want my_script to run on. Files have extension (maybe .jpeg). We want the output of my_script to be put next to the files in basename.out (e.g. foo.jpeg -> foo.out). We want to run my_script once for each core the computer has and we want to run it on the local computer, too. For the remote computers we want the file to be processed transferred to the given computer. When my_script finishes, we want foo.out transferred back and we then want foo.jpeg and foo.out removed from the remote computer:
cat list_of_files | \
parallel --trc {.}.out -S server1,server2,: \
"my_script {} > {.}.out"
GNU Parallel makes sure the output from each job does not mix, so you can use the output as input for another program:
some-command | parallel do-something | postprocess
See the videos for more examples: https://www.youtube.com/playlist?list=PL284C9FF2488BC6D1
maxjobs=4
parallelize () {
while [ $# -gt 0 ] ; do
jobcnt=(`jobs -p`)
if [ ${#jobcnt[#]} -lt $maxjobs ] ; then
do-something $1 &
shift
else
sleep 1
fi
done
wait
}
parallelize arg1 arg2 "5 args to third job" arg4 ...
Here an alternative solution that can be inserted into .bashrc and used for everyday one liner:
function pwait() {
while [ $(jobs -p | wc -l) -ge $1 ]; do
sleep 1
done
}
To use it, all one has to do is put & after the jobs and a pwait call, the parameter gives the number of parallel processes:
for i in *; do
do_something $i &
pwait 10
done
It would be nicer to use wait instead of busy waiting on the output of jobs -p, but there doesn't seem to be an obvious solution to wait till any of the given jobs is finished instead of a all of them.
Instead of a plain bash, use a Makefile, then specify number of simultaneous jobs with make -jX where X is the number of jobs to run at once.
Or you can use wait ("man wait"): launch several child processes, call wait - it will exit when the child processes finish.
maxjobs = 10
foreach line in `cat file.txt` {
jobsrunning = 0
while jobsrunning < maxjobs {
do job &
jobsrunning += 1
}
wait
}
job ( ){
...
}
If you need to store the job's result, then assign their result to a variable. After wait you just check what the variable contains.
If you're familiar with the make command, most of the time you can express the list of commands you want to run as a a makefile. For example, if you need to run $SOME_COMMAND on files *.input each of which produces *.output, you can use the makefile
INPUT = a.input b.input
OUTPUT = $(INPUT:.input=.output)
%.output : %.input
$(SOME_COMMAND) $< $#
all: $(OUTPUT)
and then just run
make -j<NUMBER>
to run at most NUMBER commands in parallel.
While doing this right in bash is probably impossible, you can do a semi-right fairly easily. bstark gave a fair approximation of right but his has the following flaws:
Word splitting: You can't pass any jobs to it that use any of the following characters in their arguments: spaces, tabs, newlines, stars, question marks. If you do, things will break, possibly unexpectedly.
It relies on the rest of your script to not background anything. If you do, or later you add something to the script that gets sent in the background because you forgot you weren't allowed to use backgrounded jobs because of his snippet, things will break.
Another approximation which doesn't have these flaws is the following:
scheduleAll() {
local job i=0 max=4 pids=()
for job; do
(( ++i % max == 0 )) && {
wait "${pids[#]}"
pids=()
}
bash -c "$job" & pids+=("$!")
done
wait "${pids[#]}"
}
Note that this one is easily adaptable to also check the exit code of each job as it ends so you can warn the user if a job fails or set an exit code for scheduleAll according to the amount of jobs that failed, or something.
The problem with this code is just that:
It schedules four (in this case) jobs at a time and then waits for all four to end. Some might be done sooner than others which will cause the next batch of four jobs to wait until the longest of the previous batch is done.
A solution that takes care of this last issue would have to use kill -0 to poll whether any of the processes have disappeared instead of the wait and schedule the next job. However, that introduces a small new problem: you have a race condition between a job ending, and the kill -0 checking whether it's ended. If the job ended and another process on your system starts up at the same time, taking a random PID which happens to be that of the job that just finished, the kill -0 won't notice your job having finished and things will break again.
A perfect solution isn't possible in bash.
Maybe try a parallelizing utility instead rewriting the loop? I'm a big fan of xjobs. I use xjobs all the time to mass copy files across our network, usually when setting up a new database server.
http://www.maier-komor.de/xjobs.html
function for bash:
parallel ()
{
awk "BEGIN{print \"all: ALL_TARGETS\\n\"}{print \"TARGET_\"NR\":\\n\\t#-\"\$0\"\\n\"}END{printf \"ALL_TARGETS:\";for(i=1;i<=NR;i++){printf \" TARGET_%d\",i};print\"\\n\"}" | make $# -f - all
}
using:
cat my_commands | parallel -j 4
Really late to the party here, but here's another solution.
A lot of solutions don't handle spaces/special characters in the commands, don't keep N jobs running at all times, eat cpu in busy loops, or rely on external dependencies (e.g. GNU parallel).
With inspiration for dead/zombie process handling, here's a pure bash solution:
function run_parallel_jobs {
local concurrent_max=$1
local callback=$2
local cmds=("${#:3}")
local jobs=( )
while [[ "${#cmds[#]}" -gt 0 ]] || [[ "${#jobs[#]}" -gt 0 ]]; do
while [[ "${#jobs[#]}" -lt $concurrent_max ]] && [[ "${#cmds[#]}" -gt 0 ]]; do
local cmd="${cmds[0]}"
cmds=("${cmds[#]:1}")
bash -c "$cmd" &
jobs+=($!)
done
local job="${jobs[0]}"
jobs=("${jobs[#]:1}")
local state="$(ps -p $job -o state= 2>/dev/null)"
if [[ "$state" == "D" ]] || [[ "$state" == "Z" ]]; then
$callback $job
else
wait $job
$callback $job $?
fi
done
}
And sample usage:
function job_done {
if [[ $# -lt 2 ]]; then
echo "PID $1 died unexpectedly"
else
echo "PID $1 exited $2"
fi
}
cmds=( \
"echo 1; sleep 1; exit 1" \
"echo 2; sleep 2; exit 2" \
"echo 3; sleep 3; exit 3" \
"echo 4; sleep 4; exit 4" \
"echo 5; sleep 5; exit 5" \
)
# cpus="$(getconf _NPROCESSORS_ONLN)"
cpus=3
run_parallel_jobs $cpus "job_done" "${cmds[#]}"
The output:
1
2
3
PID 56712 exited 1
4
PID 56713 exited 2
5
PID 56714 exited 3
PID 56720 exited 4
PID 56724 exited 5
For per-process output handling $$ could be used to log to a file, for example:
function job_done {
cat "$1.log"
}
cmds=( \
"echo 1 \$\$ >\$\$.log" \
"echo 2 \$\$ >\$\$.log" \
)
run_parallel_jobs 2 "job_done" "${cmds[#]}"
Output:
1 56871
2 56872
The project I work on uses the wait command to control parallel shell (ksh actually) processes. To address your concerns about IO, on a modern OS, it's possible parallel execution will actually increase efficiency. If all processes are reading the same blocks on disk, only the first process will have to hit the physical hardware. The other processes will often be able to retrieve the block from OS's disk cache in memory. Obviously, reading from memory is several orders of magnitude quicker than reading from disk. Also, the benefit requires no coding changes.
This might be good enough for most purposes, but is not optimal.
#!/bin/bash
n=0
maxjobs=10
for i in *.m4a ; do
# ( DO SOMETHING ) &
# limit jobs
if (( $(($((++n)) % $maxjobs)) == 0 )) ; then
wait # wait until all have finished (not optimal, but most times good enough)
echo $n wait
fi
done
Here is how I managed to solve this issue in a bash script:
#! /bin/bash
MAX_JOBS=32
FILE_LIST=($(cat ${1}))
echo Length ${#FILE_LIST[#]}
for ((INDEX=0; INDEX < ${#FILE_LIST[#]}; INDEX=$((${INDEX}+${MAX_JOBS})) ));
do
JOBS_RUNNING=0
while ((JOBS_RUNNING < MAX_JOBS))
do
I=$((${INDEX}+${JOBS_RUNNING}))
FILE=${FILE_LIST[${I}]}
if [ "$FILE" != "" ];then
echo $JOBS_RUNNING $FILE
./M22Checker ${FILE} &
else
echo $JOBS_RUNNING NULL &
fi
JOBS_RUNNING=$((JOBS_RUNNING+1))
done
wait
done
You can use a simple nested for loop (substitute appropriate integers for N and M below):
for i in {1..N}; do
(for j in {1..M}; do do_something; done & );
done
This will execute do_something N*M times in M rounds, each round executing N jobs in parallel. You can make N equal the number of CPUs you have.
My solution to always keep a given number of processes running, keep tracking of errors and handle ubnterruptible / zombie processes:
function log {
echo "$1"
}
# Take a list of commands to run, runs them sequentially with numberOfProcesses commands simultaneously runs
# Returns the number of non zero exit codes from commands
function ParallelExec {
local numberOfProcesses="${1}" # Number of simultaneous commands to run
local commandsArg="${2}" # Semi-colon separated list of commands
local pid
local runningPids=0
local counter=0
local commandsArray
local pidsArray
local newPidsArray
local retval
local retvalAll=0
local pidState
local commandsArrayPid
IFS=';' read -r -a commandsArray <<< "$commandsArg"
log "Runnning ${#commandsArray[#]} commands in $numberOfProcesses simultaneous processes."
while [ $counter -lt "${#commandsArray[#]}" ] || [ ${#pidsArray[#]} -gt 0 ]; do
while [ $counter -lt "${#commandsArray[#]}" ] && [ ${#pidsArray[#]} -lt $numberOfProcesses ]; do
log "Running command [${commandsArray[$counter]}]."
eval "${commandsArray[$counter]}" &
pid=$!
pidsArray+=($pid)
commandsArrayPid[$pid]="${commandsArray[$counter]}"
counter=$((counter+1))
done
newPidsArray=()
for pid in "${pidsArray[#]}"; do
# Handle uninterruptible sleep state or zombies by ommiting them from running process array (How to kill that is already dead ? :)
if kill -0 $pid > /dev/null 2>&1; then
pidState=$(ps -p$pid -o state= 2 > /dev/null)
if [ "$pidState" != "D" ] && [ "$pidState" != "Z" ]; then
newPidsArray+=($pid)
fi
else
# pid is dead, get it's exit code from wait command
wait $pid
retval=$?
if [ $retval -ne 0 ]; then
log "Command [${commandsArrayPid[$pid]}] failed with exit code [$retval]."
retvalAll=$((retvalAll+1))
fi
fi
done
pidsArray=("${newPidsArray[#]}")
# Add a trivial sleep time so bash won't eat all CPU
sleep .05
done
return $retvalAll
}
Usage:
cmds="du -csh /var;du -csh /tmp;sleep 3;du -csh /root;sleep 10; du -csh /home"
# Execute 2 processes at a time
ParallelExec 2 "$cmds"
# Execute 4 processes at a time
ParallelExec 4 "$cmds"
$DOMAINS = "list of some domain in commands"
for foo in some-command
do
eval `some-command for $DOMAINS` &
job[$i]=$!
i=$(( i + 1))
done
Ndomains=echo $DOMAINS |wc -w
for i in $(seq 1 1 $Ndomains)
do
echo "wait for ${job[$i]}"
wait "${job[$i]}"
done
in this concept will work for the parallelize. important thing is last line of eval is '&'
which will put the commands to backgrounds.

How to control the number of a job executions ended with "&" within a loop [duplicate]

This question already has answers here:
Parallelize Bash script with maximum number of processes
(16 answers)
Closed 1 year ago.
Is there an easy way to limit the number of concurrent jobs in bash? By that I mean making the & block when there are more then n concurrent jobs running in the background.
I know I can implement this with ps | grep -style tricks, but is there an easier way?
If you have GNU Parallel http://www.gnu.org/software/parallel/ installed you can do this:
parallel gzip ::: *.log
which will run one gzip per CPU core until all logfiles are gzipped.
If it is part of a larger loop you can use sem instead:
for i in *.log ; do
echo $i Do more stuff here
sem -j+0 gzip $i ";" echo done
done
sem --wait
It will do the same, but give you a chance to do more stuff for each file.
If GNU Parallel is not packaged for your distribution you can install GNU Parallel simply by:
$ (wget -O - pi.dk/3 || lynx -source pi.dk/3 || curl pi.dk/3/ || \
fetch -o - http://pi.dk/3 ) > install.sh
$ sha1sum install.sh | grep 883c667e01eed62f975ad28b6d50e22a
12345678 883c667e 01eed62f 975ad28b 6d50e22a
$ md5sum install.sh | grep cc21b4c943fd03e93ae1ae49e28573c0
cc21b4c9 43fd03e9 3ae1ae49 e28573c0
$ sha512sum install.sh | grep da012ec113b49a54e705f86d51e784ebced224fdf
79945d9d 250b42a4 2067bb00 99da012e c113b49a 54e705f8 6d51e784 ebced224
fdff3f52 ca588d64 e75f6033 61bd543f d631f592 2f87ceb2 ab034149 6df84a35
$ bash install.sh
It will download, check signature, and do a personal installation if it cannot install globally.
Watch the intro videos for GNU Parallel to learn more:
https://www.youtube.com/playlist?list=PL284C9FF2488BC6D1
A small bash script could help you:
# content of script exec-async.sh
joblist=($(jobs -p))
while (( ${#joblist[*]} >= 3 ))
do
sleep 1
joblist=($(jobs -p))
done
$* &
If you call:
. exec-async.sh sleep 10
...four times, the first three calls will return immediately, the fourth call will block until there are less than three jobs running.
You need to start this script inside the current session by prefixing it with ., because jobs lists only the jobs of the current session.
The sleep inside is ugly, but I didn't find a way to wait for the first job that terminates.
The following script shows a way to do this with functions. You can either put the bgxupdate() and bgxlimit() functions in your script, or have them in a separate file which is sourced from your script with:
. /path/to/bgx.sh
It has the advantage that you can maintain multiple groups of processes independently (you can run, for example, one group with a limit of 10 and another totally separate group with a limit of 3).
It uses the Bash built-in jobs to get a list of sub-processes but maintains them in individual variables. In the loop at the bottom, you can see how to call the bgxlimit() function:
Set up an empty group variable.
Transfer that to bgxgrp.
Call bgxlimit() with the limit and command you want to run.
Transfer the new group back to your group variable.
Of course, if you only have one group, just use bgxgrp variable directly rather than transferring in and out.
#!/bin/bash
# bgxupdate - update active processes in a group.
# Works by transferring each process to new group
# if it is still active.
# in: bgxgrp - current group of processes.
# out: bgxgrp - new group of processes.
# out: bgxcount - number of processes in new group.
bgxupdate() {
bgxoldgrp=${bgxgrp}
bgxgrp=""
((bgxcount = 0))
bgxjobs=" $(jobs -pr | tr '\n' ' ')"
for bgxpid in ${bgxoldgrp} ; do
echo "${bgxjobs}" | grep " ${bgxpid} " >/dev/null 2>&1
if [[ $? -eq 0 ]]; then
bgxgrp="${bgxgrp} ${bgxpid}"
((bgxcount++))
fi
done
}
# bgxlimit - start a sub-process with a limit.
# Loops, calling bgxupdate until there is a free
# slot to run another sub-process. Then runs it
# an updates the process group.
# in: $1 - the limit on processes.
# in: $2+ - the command to run for new process.
# in: bgxgrp - the current group of processes.
# out: bgxgrp - new group of processes
bgxlimit() {
bgxmax=$1; shift
bgxupdate
while [[ ${bgxcount} -ge ${bgxmax} ]]; do
sleep 1
bgxupdate
done
if [[ "$1" != "-" ]]; then
$* &
bgxgrp="${bgxgrp} $!"
fi
}
# Test program, create group and run 6 sleeps with
# limit of 3.
group1=""
echo 0 $(date | awk '{print $4}') '[' ${group1} ']'
echo
for i in 1 2 3 4 5 6; do
bgxgrp=${group1}; bgxlimit 3 sleep ${i}0; group1=${bgxgrp}
echo ${i} $(date | awk '{print $4}') '[' ${group1} ']'
done
# Wait until all others are finished.
echo
bgxgrp=${group1}; bgxupdate; group1=${bgxgrp}
while [[ ${bgxcount} -ne 0 ]]; do
oldcount=${bgxcount}
while [[ ${oldcount} -eq ${bgxcount} ]]; do
sleep 1
bgxgrp=${group1}; bgxupdate; group1=${bgxgrp}
done
echo 9 $(date | awk '{print $4}') '[' ${group1} ']'
done
Here’s a sample run, with blank lines inserted to clearly delineate different time points:
0 12:38:00 [ ]
1 12:38:00 [ 3368 ]
2 12:38:00 [ 3368 5880 ]
3 12:38:00 [ 3368 5880 2524 ]
4 12:38:10 [ 5880 2524 1560 ]
5 12:38:20 [ 2524 1560 5032 ]
6 12:38:30 [ 1560 5032 5212 ]
9 12:38:50 [ 5032 5212 ]
9 12:39:10 [ 5212 ]
9 12:39:30 [ ]
The whole thing starts at 12:38:00 (time t = 0) and, as you can see, the first three processes run immediately.
Each process sleeps for 10n seconds and the fourth process doesn’t start until the first exits (at time t = 10). You can see that process 3368 has disappeared from the list before 1560 is added.
Similarly, the fifth process 5032 starts when 5880 (the second) exits at time t = 20.
And finally, the sixth process 5212 starts when 2524 (the third) exits at time t = 30.
Then the rundown begins, the fourth process exits at time t = 50 (started at 10 with 40 duration).
The fifth exits at time t = 70 (started at 20 with 50 duration).
Finally, the sixth exits at time t = 90 (started at 30 with 60 duration).
Or, if you prefer it in a more graphical time-line form:
Process: 1 2 3 4 5 6
-------- - - - - - -
12:38:00 ^ ^ ^ 1/2/3 start together.
12:38:10 v | | ^ 4 starts when 1 done.
12:38:20 v | | ^ 5 starts when 2 done.
12:38:30 v | | ^ 6 starts when 3 done.
12:38:40 | | |
12:38:50 v | | 4 ends.
12:39:00 | |
12:39:10 v | 5 ends.
12:39:20 |
12:39:30 v 6 ends.
Here's the shortest way:
waitforjobs() {
while test $(jobs -p | wc -w) -ge "$1"; do wait -n; done
}
Call this function before forking off any new job:
waitforjobs 10
run_another_job &
To have as many background jobs as cores on the machine, use $(nproc) instead of a fixed number like 10.
Assuming you'd like to write code like this:
for x in $(seq 1 100); do # 100 things we want to put into the background.
max_bg_procs 5 # Define the limit. See below.
your_intensive_job &
done
Where max_bg_procs should be put in your .bashrc:
function max_bg_procs {
if [[ $# -eq 0 ]] ; then
echo "Usage: max_bg_procs NUM_PROCS. Will wait until the number of background (&)"
echo " bash processes (as determined by 'jobs -pr') falls below NUM_PROCS"
return
fi
local max_number=$((0 + ${1:-0}))
while true; do
local current_number=$(jobs -pr | wc -l)
if [[ $current_number -lt $max_number ]]; then
break
fi
sleep 1
done
}
The following function (developed from tangens answer above, either copy into script or source from file):
job_limit () {
# Test for single positive integer input
if (( $# == 1 )) && [[ $1 =~ ^[1-9][0-9]*$ ]]
then
# Check number of running jobs
joblist=($(jobs -rp))
while (( ${#joblist[*]} >= $1 ))
do
# Wait for any job to finish
command='wait '${joblist[0]}
for job in ${joblist[#]:1}
do
command+=' || wait '$job
done
eval $command
joblist=($(jobs -rp))
done
fi
}
1) Only requires inserting a single line to limit an existing loop
while :
do
task &
job_limit `nproc`
done
2) Waits on completion of existing background tasks rather than polling, increasing efficiency for fast tasks
This might be good enough for most purposes, but is not optimal.
#!/bin/bash
n=0
maxjobs=10
for i in *.m4a ; do
# ( DO SOMETHING ) &
# limit jobs
if (( $(($((++n)) % $maxjobs)) == 0 )) ; then
wait # wait until all have finished (not optimal, but most times good enough)
echo $n wait
fi
done
If you're willing to do this outside of pure bash, you should look into a job queuing system.
For instance, there's GNU queue or PBS. And for PBS, you might want to look into Maui for configuration.
Both systems will require some configuration, but it's entirely possible to allow a specific number of jobs to run at once, only starting newly queued jobs when a running job finishes. Typically, these job queuing systems would be used on supercomputing clusters, where you would want to allocate a specific amount of memory or computing time to any given batch job; however, there's no reason you can't use one of these on a single desktop computer without regard for compute time or memory limits.
It is hard to do without wait -n (for example, shell in busybox does not support it). So here is a workaround, it is not optimal because it calls 'jobs' and 'wc' commands 10x per second. You can reduce the calls to 1x per second for example, if you don't mind waiting a bit longer for each job to complete.
# $1 = maximum concurent jobs
#
limit_jobs()
{
while true; do
if [ "$(jobs -p | wc -l)" -lt "$1" ]; then break; fi
usleep 100000
done
}
# and now start some tasks:
task &
limit_jobs 2
task &
limit_jobs 2
task &
limit_jobs 2
task &
limit_jobs 2
wait
On Linux I use this to limit the bash jobs to the number of available CPUs (possibly overriden by setting the CPU_NUMBER).
[ "$CPU_NUMBER" ] || CPU_NUMBER="`nproc 2>/dev/null || echo 1`"
while [ "$1" ]; do
{
do something
with $1
in parallel
echo "[$# items left] $1 done"
} &
while true; do
# load the PIDs of all child processes to the array
joblist=(`jobs -p`)
if [ ${#joblist[*]} -ge "$CPU_NUMBER" ]; then
# when the job limit is reached, wait for *single* job to finish
wait -n
else
# stop checking when we're below the limit
break
fi
done
# it's great we executed zero external commands to check!
shift
done
# wait for all currently active child processes
wait
Wait command, -n option, waits for the next job to terminate.
maxjobs=10
# wait for the amount of processes less to $maxjobs
jobIds=($(jobs -p))
len=${#jobIds[#]}
while [ $len -ge $maxjobs ]; do
# Wait until one job is finished
wait -n $jobIds
jobIds=($(jobs -p))
len=${#jobIds[#]}
done
Have you considered starting ten long-running listener processes and communicating with them via named pipes?
you can use ulimit -u
see http://ss64.com/bash/ulimit.html
Bash mostly processes files line by line.
So you cap split input file input files by N lines then simple pattern is applicable:
mkdir tmp ; pushd tmp ; split -l 50 ../mainfile.txt
for file in * ; do
while read a b c ; do curl -s http://$a/$b/$c <$file &
done ; wait ; done
popd ; rm -rf tmp;

how to write a process-pool bash shell

I have more than 10 tasks to execute, and the system restrict that there at most 4 tasks can run at the same time.
My task can be started like:
myprog taskname
How can I write a bash shell script to run these task. The most important thing is that when one task finish, the script can start another immediately, making the running tasks count remain 4 all the time.
Use xargs:
xargs -P <maximum-number-of-process-at-a-time> -n <arguments-per-process> <command>
Details here.
I chanced upon this thread while looking into writing my own process pool and particularly liked Brandon Horsley's solution, though I couldn't get the signals working right, so I took inspiration from Apache and decided to try a pre-fork model with a fifo as my job queue.
The following function is the function that the worker processes run when forked.
# \brief the worker function that is called when we fork off worker processes
# \param[in] id the worker ID
# \param[in] job_queue the fifo to read jobs from
# \param[in] result_log the temporary log file to write exit codes to
function _job_pool_worker()
{
local id=$1
local job_queue=$2
local result_log=$3
local line=
exec 7<> ${job_queue}
while [[ "${line}" != "${job_pool_end_of_jobs}" && -e "${job_queue}" ]]; do
# workers block on the exclusive lock to read the job queue
flock --exclusive 7
read line <${job_queue}
flock --unlock 7
# the worker should exit if it sees the end-of-job marker or run the
# job otherwise and save its exit code to the result log.
if [[ "${line}" == "${job_pool_end_of_jobs}" ]]; then
# write it one more time for the next sibling so that everyone
# will know we are exiting.
echo "${line}" >&7
else
_job_pool_echo "### _job_pool_worker-${id}: ${line}"
# run the job
{ ${line} ; }
# now check the exit code and prepend "ERROR" to the result log entry
# which we will use to count errors and then strip out later.
local result=$?
local status=
if [[ "${result}" != "0" ]]; then
status=ERROR
fi
# now write the error to the log, making sure multiple processes
# don't trample over each other.
exec 8<> ${result_log}
flock --exclusive 8
echo "${status}job_pool: exited ${result}: ${line}" >> ${result_log}
flock --unlock 8
exec 8>&-
_job_pool_echo "### _job_pool_worker-${id}: exited ${result}: ${line}"
fi
done
exec 7>&-
}
You can get a copy of my solution at Github. Here's a sample program using my implementation.
#!/bin/bash
. job_pool.sh
function foobar()
{
# do something
true
}
# initialize the job pool to allow 3 parallel jobs and echo commands
job_pool_init 3 0
# run jobs
job_pool_run sleep 1
job_pool_run sleep 2
job_pool_run sleep 3
job_pool_run foobar
job_pool_run foobar
job_pool_run /bin/false
# wait until all jobs complete before continuing
job_pool_wait
# more jobs
job_pool_run /bin/false
job_pool_run sleep 1
job_pool_run sleep 2
job_pool_run foobar
# don't forget to shut down the job pool
job_pool_shutdown
# check the $job_pool_nerrors for the number of jobs that exited non-zero
echo "job_pool_nerrors: ${job_pool_nerrors}"
Hope this helps!
Using GNU Parallel you can do:
cat tasks | parallel -j4 myprog
If you have 4 cores, you can even just do:
cat tasks | parallel myprog
From http://git.savannah.gnu.org/cgit/parallel.git/tree/README:
Full installation
Full installation of GNU Parallel is as simple as:
./configure && make && make install
Personal installation
If you are not root you can add ~/bin to your path and install in
~/bin and ~/share:
./configure --prefix=$HOME && make && make install
Or if your system lacks 'make' you can simply copy src/parallel
src/sem src/niceload src/sql to a dir in your path.
Minimal installation
If you just need parallel and do not have 'make' installed (maybe the
system is old or Microsoft Windows):
wget http://git.savannah.gnu.org/cgit/parallel.git/plain/src/parallel
chmod 755 parallel
cp parallel sem
mv parallel sem dir-in-your-$PATH/bin/
Test the installation
After this you should be able to do:
parallel -j0 ping -nc 3 ::: foss.org.my gnu.org freenetproject.org
This will send 3 ping packets to 3 different hosts in parallel and print
the output when they complete.
Watch the intro video for a quick introduction:
https://www.youtube.com/playlist?list=PL284C9FF2488BC6D1
I would suggest writing four scripts, each one of which executes a certain number of tasks in series. Then write another script that starts the four scripts in parallel. For instance, if you have scripts, script1.sh, script2.sh, script3.sh, and script4.sh, you could have a script called headscript.sh like so.
#!/bin/sh
./script1.sh &
./script2.sh &
./script3.sh &
./script4.sh &
I found the best solution proposed in A Foo Walks into a Bar... blog using build-in functionality of well know xargs tool
First create a file commands.txt with list of commands you want to execute
myprog taskname1
myprog taskname2
myprog taskname3
myprog taskname4
...
myprog taskname123
and then pipe it to xargs like this to execute in 4 processes pool:
cat commands.txt | xargs -I CMD --max-procs=4 bash -c CMD
you can modify no of process
Following #Parag Sardas' answer and the documentation linked here's a quick script you might want to add on your .bash_aliases.
Relinking the doc link because it's worth a read
#!/bin/bash
# https://stackoverflow.com/a/19618159
# https://stackoverflow.com/a/51861820
#
# Example file contents:
# touch /tmp/a.txt
# touch /tmp/b.txt
if [ "$#" -eq 0 ]; then
echo "$0 <file> [max-procs=0]"
exit 1
fi
FILE=${1}
MAX_PROCS=${2:-0}
cat $FILE | while read line; do printf "%q\n" "$line"; done | xargs --max-procs=$MAX_PROCS -I CMD bash -c CMD
I.e.
./xargs-parallel.sh jobs.txt 4 maximum of 4 processes read from jobs.txt
You could probably do something clever with signals.
Note this is only to illustrate the concept, and thus not thoroughly tested.
#!/usr/local/bin/bash
this_pid="$$"
jobs_running=0
sleep_pid=
# Catch alarm signals to adjust the number of running jobs
trap 'decrement_jobs' SIGALRM
# When a job finishes, decrement the total and kill the sleep process
decrement_jobs()
{
jobs_running=$(($jobs_running - 1))
if [ -n "${sleep_pid}" ]
then
kill -s SIGKILL "${sleep_pid}"
sleep_pid=
fi
}
# Check to see if the max jobs are running, if so sleep until woken
launch_task()
{
if [ ${jobs_running} -gt 3 ]
then
(
while true
do
sleep 999
done
) &
sleep_pid=$!
wait ${sleep_pid}
fi
# Launch the requested task, signalling the parent upon completion
(
"$#"
kill -s SIGALRM "${this_pid}"
) &
jobs_running=$((${jobs_running} + 1))
}
# Launch all of the tasks, this can be in a loop, etc.
launch_task task1
launch_task tast2
...
launch_task task99
This tested script runs 5 jobs at a time and will restart a new job as soon as it does (due to the kill of the sleep 10.9 when we get a SIGCHLD. A simpler version of this could use direct polling (change the sleep 10.9 to sleep 1 and get rid of the trap).
#!/usr/bin/bash
set -o monitor
trap "pkill -P $$ -f 'sleep 10\.9' >&/dev/null" SIGCHLD
totaljobs=15
numjobs=5
worktime=10
curjobs=0
declare -A pidlist
dojob()
{
slot=$1
time=$(echo "$RANDOM * 10 / 32768" | bc -l)
echo Starting job $slot with args $time
sleep $time &
pidlist[$slot]=`jobs -p %%`
curjobs=$(($curjobs + 1))
totaljobs=$(($totaljobs - 1))
}
# start
while [ $curjobs -lt $numjobs -a $totaljobs -gt 0 ]
do
dojob $curjobs
done
# Poll for jobs to die, restarting while we have them
while [ $totaljobs -gt 0 ]
do
for ((i=0;$i < $curjobs;i++))
do
if ! kill -0 ${pidlist[$i]} >&/dev/null
then
dojob $i
break
fi
done
sleep 10.9 >&/dev/null
done
wait
Other answer about 4 shell scripts does not fully satisfies me as it assumes that all tasks take approximatelu the same time and because it requires manual set up. But here is how I would improve it.
Main script will create symbolic links to executables following certain namimg convention. For example,
ln -s executable1 ./01-task.01
first prefix is for sorting and suffix identifies batch (01-04).
Now we spawn 4 shell scripts that would take batch number as input and do something like this
for t in $(ls ./*-task.$batch | sort ; do
t
rm t
done
Look at my implementation of job pool in bash: https://github.com/spektom/shell-utils/blob/master/jp.sh
For example, to run at most 3 processes of cURL when downloading from a lot of URLs, you can wrap your cURL commands as follows:
./jp.sh "My Download Pool" 3 curl http://site1/...
./jp.sh "My Download Pool" 3 curl http://site2/...
./jp.sh "My Download Pool" 3 curl http://site3/...
...
Here is my solution. The idea is quite simple. I create a fifo as a semaphore, where each line stands for an available resource. When reading the queue, the main process blocks if there is nothing left. And, we return the resource after the task is done by simply echoing anything to the queue.
function task() {
local task_no="$1"
# doing the actual task...
echo "Executing Task ${task_no}"
# which takes a long time
sleep 1
}
function execute_concurrently() {
local tasks="$1"
local ps_pool_size="$2"
# create an anonymous fifo as a Semaphore
local sema_fifo
sema_fifo="$(mktemp -u)"
mkfifo "${sema_fifo}"
exec 3<>"${sema_fifo}"
rm -f "${sema_fifo}"
# every 'x' stands for an available resource
for i in $(seq 1 "${ps_pool_size}"); do
echo 'x' >&3
done
for task_no in $(seq 1 "${tasks}"); do
read dummy <&3 # blocks util a resource is available
(
trap 'echo x >&3' EXIT # returns the resource on exit
task "${task_no}"
)&
done
wait # wait util all forked tasks have finished
}
execute_concurrently 10 4
The script above will run 10 tasks and 4 each time concurrently. You can change the $(seq 1 "${tasks}") sequence to the actual task queue you want to run.
I made my modifications based on methods introduced in this Writing a process pool in Bash.
#!/bin/bash
#set -e # this doesn't work here for some reason
POOL_SIZE=4 # number of workers running in parallel
#######################################################################
# populate jobs #
#######################################################################
declare -a jobs
for (( i = 1988; i < 2019; i++ )); do
jobs+=($i)
done
echo '################################################'
echo ' Launching jobs'
echo '################################################'
parallel() {
local proc procs jobs cur
jobs=("$#") # input jobs array
declare -a procs=() # processes array
cur=0 # current job idx
morework=true
while $morework; do
# if process array size < pool size, try forking a new proc
if [[ "${#procs[#]}" -lt "$POOL_SIZE" ]]; then
if [[ $cur -lt "${#jobs[#]}" ]]; then
proc=${jobs[$cur]}
echo "JOB ID = $cur; JOB = $proc."
###############
# do job here #
###############
sleep 3 &
# add to current running processes
procs+=("$!")
# move to the next job
((cur++))
else
morework=false
continue
fi
fi
for n in "${!procs[#]}"; do
kill -0 "${procs[n]}" 2>/dev/null && continue
# if process is not running anymore, remove from array
unset procs[n]
done
done
wait
}
parallel "${jobs[#]}"
xargs with -P and -L options does the job.
You can extract the idea from the example below:
#!/usr/bin/env bash
workers_pool_size=10
set -e
function doit {
cmds=""
for e in 4 8 16; do
for m in 1 2 3 4 5 6; do
cmd="python3 ./doit.py --m $m -e $e -m $m"
cmds="$cmd\n$cmds"
done
done
echo -e "All commands:\n$cmds"
echo "Workers pool size = $workers_pool_size"
echo -e "$cmds" | xargs -t -P $workers_pool_size -L 1 time > /dev/null
}
doit
#! /bin/bash
doSomething() {
<...>
}
getCompletedThreads() {
_runningThreads=("$#")
removableThreads=()
for pid in "${_runningThreads[#]}"; do
if ! ps -p $pid > /dev/null; then
removableThreads+=($pid)
fi
done
echo "$removableThreads"
}
releasePool() {
while [[ ${#runningThreads[#]} -eq $MAX_THREAD_NO ]]; do
echo "releasing"
removableThreads=( $(getCompletedThreads "${runningThreads[#]}") )
if [ ${#removableThreads[#]} -eq 0 ]; then
sleep 0.2
else
for removableThread in "${removableThreads[#]}"; do
runningThreads=( ${runningThreads[#]/$removableThread} )
done
echo "released"
fi
done
}
waitAllThreadComplete() {
while [[ ${#runningThreads[#]} -ne 0 ]]; do
removableThreads=( $(getCompletedThreads "${runningThreads[#]}") )
for removableThread in "${removableThreads[#]}"; do
runningThreads=( ${runningThreads[#]/$removableThread} )
done
if [ ${#removableThreads[#]} -eq 0 ]; then
sleep 0.2
fi
done
}
MAX_THREAD_NO=10
runningThreads=()
sequenceNo=0
for i in {1..36}; do
releasePool
((sequenceNo++))
echo "added $sequenceNo"
doSomething &
pid=$!
runningThreads+=($pid)
done
waitAllThreadComplete

Bash: limit the number of concurrent jobs? [duplicate]

This question already has answers here:
Parallelize Bash script with maximum number of processes
(16 answers)
Closed 1 year ago.
Is there an easy way to limit the number of concurrent jobs in bash? By that I mean making the & block when there are more then n concurrent jobs running in the background.
I know I can implement this with ps | grep -style tricks, but is there an easier way?
If you have GNU Parallel http://www.gnu.org/software/parallel/ installed you can do this:
parallel gzip ::: *.log
which will run one gzip per CPU core until all logfiles are gzipped.
If it is part of a larger loop you can use sem instead:
for i in *.log ; do
echo $i Do more stuff here
sem -j+0 gzip $i ";" echo done
done
sem --wait
It will do the same, but give you a chance to do more stuff for each file.
If GNU Parallel is not packaged for your distribution you can install GNU Parallel simply by:
$ (wget -O - pi.dk/3 || lynx -source pi.dk/3 || curl pi.dk/3/ || \
fetch -o - http://pi.dk/3 ) > install.sh
$ sha1sum install.sh | grep 883c667e01eed62f975ad28b6d50e22a
12345678 883c667e 01eed62f 975ad28b 6d50e22a
$ md5sum install.sh | grep cc21b4c943fd03e93ae1ae49e28573c0
cc21b4c9 43fd03e9 3ae1ae49 e28573c0
$ sha512sum install.sh | grep da012ec113b49a54e705f86d51e784ebced224fdf
79945d9d 250b42a4 2067bb00 99da012e c113b49a 54e705f8 6d51e784 ebced224
fdff3f52 ca588d64 e75f6033 61bd543f d631f592 2f87ceb2 ab034149 6df84a35
$ bash install.sh
It will download, check signature, and do a personal installation if it cannot install globally.
Watch the intro videos for GNU Parallel to learn more:
https://www.youtube.com/playlist?list=PL284C9FF2488BC6D1
A small bash script could help you:
# content of script exec-async.sh
joblist=($(jobs -p))
while (( ${#joblist[*]} >= 3 ))
do
sleep 1
joblist=($(jobs -p))
done
$* &
If you call:
. exec-async.sh sleep 10
...four times, the first three calls will return immediately, the fourth call will block until there are less than three jobs running.
You need to start this script inside the current session by prefixing it with ., because jobs lists only the jobs of the current session.
The sleep inside is ugly, but I didn't find a way to wait for the first job that terminates.
The following script shows a way to do this with functions. You can either put the bgxupdate() and bgxlimit() functions in your script, or have them in a separate file which is sourced from your script with:
. /path/to/bgx.sh
It has the advantage that you can maintain multiple groups of processes independently (you can run, for example, one group with a limit of 10 and another totally separate group with a limit of 3).
It uses the Bash built-in jobs to get a list of sub-processes but maintains them in individual variables. In the loop at the bottom, you can see how to call the bgxlimit() function:
Set up an empty group variable.
Transfer that to bgxgrp.
Call bgxlimit() with the limit and command you want to run.
Transfer the new group back to your group variable.
Of course, if you only have one group, just use bgxgrp variable directly rather than transferring in and out.
#!/bin/bash
# bgxupdate - update active processes in a group.
# Works by transferring each process to new group
# if it is still active.
# in: bgxgrp - current group of processes.
# out: bgxgrp - new group of processes.
# out: bgxcount - number of processes in new group.
bgxupdate() {
bgxoldgrp=${bgxgrp}
bgxgrp=""
((bgxcount = 0))
bgxjobs=" $(jobs -pr | tr '\n' ' ')"
for bgxpid in ${bgxoldgrp} ; do
echo "${bgxjobs}" | grep " ${bgxpid} " >/dev/null 2>&1
if [[ $? -eq 0 ]]; then
bgxgrp="${bgxgrp} ${bgxpid}"
((bgxcount++))
fi
done
}
# bgxlimit - start a sub-process with a limit.
# Loops, calling bgxupdate until there is a free
# slot to run another sub-process. Then runs it
# an updates the process group.
# in: $1 - the limit on processes.
# in: $2+ - the command to run for new process.
# in: bgxgrp - the current group of processes.
# out: bgxgrp - new group of processes
bgxlimit() {
bgxmax=$1; shift
bgxupdate
while [[ ${bgxcount} -ge ${bgxmax} ]]; do
sleep 1
bgxupdate
done
if [[ "$1" != "-" ]]; then
$* &
bgxgrp="${bgxgrp} $!"
fi
}
# Test program, create group and run 6 sleeps with
# limit of 3.
group1=""
echo 0 $(date | awk '{print $4}') '[' ${group1} ']'
echo
for i in 1 2 3 4 5 6; do
bgxgrp=${group1}; bgxlimit 3 sleep ${i}0; group1=${bgxgrp}
echo ${i} $(date | awk '{print $4}') '[' ${group1} ']'
done
# Wait until all others are finished.
echo
bgxgrp=${group1}; bgxupdate; group1=${bgxgrp}
while [[ ${bgxcount} -ne 0 ]]; do
oldcount=${bgxcount}
while [[ ${oldcount} -eq ${bgxcount} ]]; do
sleep 1
bgxgrp=${group1}; bgxupdate; group1=${bgxgrp}
done
echo 9 $(date | awk '{print $4}') '[' ${group1} ']'
done
Here’s a sample run, with blank lines inserted to clearly delineate different time points:
0 12:38:00 [ ]
1 12:38:00 [ 3368 ]
2 12:38:00 [ 3368 5880 ]
3 12:38:00 [ 3368 5880 2524 ]
4 12:38:10 [ 5880 2524 1560 ]
5 12:38:20 [ 2524 1560 5032 ]
6 12:38:30 [ 1560 5032 5212 ]
9 12:38:50 [ 5032 5212 ]
9 12:39:10 [ 5212 ]
9 12:39:30 [ ]
The whole thing starts at 12:38:00 (time t = 0) and, as you can see, the first three processes run immediately.
Each process sleeps for 10n seconds and the fourth process doesn’t start until the first exits (at time t = 10). You can see that process 3368 has disappeared from the list before 1560 is added.
Similarly, the fifth process 5032 starts when 5880 (the second) exits at time t = 20.
And finally, the sixth process 5212 starts when 2524 (the third) exits at time t = 30.
Then the rundown begins, the fourth process exits at time t = 50 (started at 10 with 40 duration).
The fifth exits at time t = 70 (started at 20 with 50 duration).
Finally, the sixth exits at time t = 90 (started at 30 with 60 duration).
Or, if you prefer it in a more graphical time-line form:
Process: 1 2 3 4 5 6
-------- - - - - - -
12:38:00 ^ ^ ^ 1/2/3 start together.
12:38:10 v | | ^ 4 starts when 1 done.
12:38:20 v | | ^ 5 starts when 2 done.
12:38:30 v | | ^ 6 starts when 3 done.
12:38:40 | | |
12:38:50 v | | 4 ends.
12:39:00 | |
12:39:10 v | 5 ends.
12:39:20 |
12:39:30 v 6 ends.
Here's the shortest way:
waitforjobs() {
while test $(jobs -p | wc -w) -ge "$1"; do wait -n; done
}
Call this function before forking off any new job:
waitforjobs 10
run_another_job &
To have as many background jobs as cores on the machine, use $(nproc) instead of a fixed number like 10.
Assuming you'd like to write code like this:
for x in $(seq 1 100); do # 100 things we want to put into the background.
max_bg_procs 5 # Define the limit. See below.
your_intensive_job &
done
Where max_bg_procs should be put in your .bashrc:
function max_bg_procs {
if [[ $# -eq 0 ]] ; then
echo "Usage: max_bg_procs NUM_PROCS. Will wait until the number of background (&)"
echo " bash processes (as determined by 'jobs -pr') falls below NUM_PROCS"
return
fi
local max_number=$((0 + ${1:-0}))
while true; do
local current_number=$(jobs -pr | wc -l)
if [[ $current_number -lt $max_number ]]; then
break
fi
sleep 1
done
}
The following function (developed from tangens answer above, either copy into script or source from file):
job_limit () {
# Test for single positive integer input
if (( $# == 1 )) && [[ $1 =~ ^[1-9][0-9]*$ ]]
then
# Check number of running jobs
joblist=($(jobs -rp))
while (( ${#joblist[*]} >= $1 ))
do
# Wait for any job to finish
command='wait '${joblist[0]}
for job in ${joblist[#]:1}
do
command+=' || wait '$job
done
eval $command
joblist=($(jobs -rp))
done
fi
}
1) Only requires inserting a single line to limit an existing loop
while :
do
task &
job_limit `nproc`
done
2) Waits on completion of existing background tasks rather than polling, increasing efficiency for fast tasks
This might be good enough for most purposes, but is not optimal.
#!/bin/bash
n=0
maxjobs=10
for i in *.m4a ; do
# ( DO SOMETHING ) &
# limit jobs
if (( $(($((++n)) % $maxjobs)) == 0 )) ; then
wait # wait until all have finished (not optimal, but most times good enough)
echo $n wait
fi
done
If you're willing to do this outside of pure bash, you should look into a job queuing system.
For instance, there's GNU queue or PBS. And for PBS, you might want to look into Maui for configuration.
Both systems will require some configuration, but it's entirely possible to allow a specific number of jobs to run at once, only starting newly queued jobs when a running job finishes. Typically, these job queuing systems would be used on supercomputing clusters, where you would want to allocate a specific amount of memory or computing time to any given batch job; however, there's no reason you can't use one of these on a single desktop computer without regard for compute time or memory limits.
It is hard to do without wait -n (for example, shell in busybox does not support it). So here is a workaround, it is not optimal because it calls 'jobs' and 'wc' commands 10x per second. You can reduce the calls to 1x per second for example, if you don't mind waiting a bit longer for each job to complete.
# $1 = maximum concurent jobs
#
limit_jobs()
{
while true; do
if [ "$(jobs -p | wc -l)" -lt "$1" ]; then break; fi
usleep 100000
done
}
# and now start some tasks:
task &
limit_jobs 2
task &
limit_jobs 2
task &
limit_jobs 2
task &
limit_jobs 2
wait
On Linux I use this to limit the bash jobs to the number of available CPUs (possibly overriden by setting the CPU_NUMBER).
[ "$CPU_NUMBER" ] || CPU_NUMBER="`nproc 2>/dev/null || echo 1`"
while [ "$1" ]; do
{
do something
with $1
in parallel
echo "[$# items left] $1 done"
} &
while true; do
# load the PIDs of all child processes to the array
joblist=(`jobs -p`)
if [ ${#joblist[*]} -ge "$CPU_NUMBER" ]; then
# when the job limit is reached, wait for *single* job to finish
wait -n
else
# stop checking when we're below the limit
break
fi
done
# it's great we executed zero external commands to check!
shift
done
# wait for all currently active child processes
wait
Wait command, -n option, waits for the next job to terminate.
maxjobs=10
# wait for the amount of processes less to $maxjobs
jobIds=($(jobs -p))
len=${#jobIds[#]}
while [ $len -ge $maxjobs ]; do
# Wait until one job is finished
wait -n $jobIds
jobIds=($(jobs -p))
len=${#jobIds[#]}
done
Have you considered starting ten long-running listener processes and communicating with them via named pipes?
you can use ulimit -u
see http://ss64.com/bash/ulimit.html
Bash mostly processes files line by line.
So you cap split input file input files by N lines then simple pattern is applicable:
mkdir tmp ; pushd tmp ; split -l 50 ../mainfile.txt
for file in * ; do
while read a b c ; do curl -s http://$a/$b/$c <$file &
done ; wait ; done
popd ; rm -rf tmp;

Parallelize Bash script with maximum number of processes

Lets say I have a loop in Bash:
for foo in `some-command`
do
do-something $foo
done
do-something is cpu bound and I have a nice shiny 4 core processor. I'd like to be able to run up to 4 do-something's at once.
The naive approach seems to be:
for foo in `some-command`
do
do-something $foo &
done
This will run all do-somethings at once, but there are a couple downsides, mainly that do-something may also have some significant I/O which performing all at once might slow down a bit. The other problem is that this code block returns immediately, so no way to do other work when all the do-somethings are finished.
How would you write this loop so there are always X do-somethings running at once?
Depending on what you want to do xargs also can help (here: converting documents with pdf2ps):
cpus=$( ls -d /sys/devices/system/cpu/cpu[[:digit:]]* | wc -w )
find . -name \*.pdf | xargs --max-args=1 --max-procs=$cpus pdf2ps
From the docs:
--max-procs=max-procs
-P max-procs
Run up to max-procs processes at a time; the default is 1.
If max-procs is 0, xargs will run as many processes as possible at a
time. Use the -n option with -P; otherwise chances are that only one
exec will be done.
With GNU Parallel http://www.gnu.org/software/parallel/ you can write:
some-command | parallel do-something
GNU Parallel also supports running jobs on remote computers. This will run one per CPU core on the remote computers - even if they have different number of cores:
some-command | parallel -S server1,server2 do-something
A more advanced example: Here we list of files that we want my_script to run on. Files have extension (maybe .jpeg). We want the output of my_script to be put next to the files in basename.out (e.g. foo.jpeg -> foo.out). We want to run my_script once for each core the computer has and we want to run it on the local computer, too. For the remote computers we want the file to be processed transferred to the given computer. When my_script finishes, we want foo.out transferred back and we then want foo.jpeg and foo.out removed from the remote computer:
cat list_of_files | \
parallel --trc {.}.out -S server1,server2,: \
"my_script {} > {.}.out"
GNU Parallel makes sure the output from each job does not mix, so you can use the output as input for another program:
some-command | parallel do-something | postprocess
See the videos for more examples: https://www.youtube.com/playlist?list=PL284C9FF2488BC6D1
maxjobs=4
parallelize () {
while [ $# -gt 0 ] ; do
jobcnt=(`jobs -p`)
if [ ${#jobcnt[#]} -lt $maxjobs ] ; then
do-something $1 &
shift
else
sleep 1
fi
done
wait
}
parallelize arg1 arg2 "5 args to third job" arg4 ...
Here an alternative solution that can be inserted into .bashrc and used for everyday one liner:
function pwait() {
while [ $(jobs -p | wc -l) -ge $1 ]; do
sleep 1
done
}
To use it, all one has to do is put & after the jobs and a pwait call, the parameter gives the number of parallel processes:
for i in *; do
do_something $i &
pwait 10
done
It would be nicer to use wait instead of busy waiting on the output of jobs -p, but there doesn't seem to be an obvious solution to wait till any of the given jobs is finished instead of a all of them.
Instead of a plain bash, use a Makefile, then specify number of simultaneous jobs with make -jX where X is the number of jobs to run at once.
Or you can use wait ("man wait"): launch several child processes, call wait - it will exit when the child processes finish.
maxjobs = 10
foreach line in `cat file.txt` {
jobsrunning = 0
while jobsrunning < maxjobs {
do job &
jobsrunning += 1
}
wait
}
job ( ){
...
}
If you need to store the job's result, then assign their result to a variable. After wait you just check what the variable contains.
If you're familiar with the make command, most of the time you can express the list of commands you want to run as a a makefile. For example, if you need to run $SOME_COMMAND on files *.input each of which produces *.output, you can use the makefile
INPUT = a.input b.input
OUTPUT = $(INPUT:.input=.output)
%.output : %.input
$(SOME_COMMAND) $< $#
all: $(OUTPUT)
and then just run
make -j<NUMBER>
to run at most NUMBER commands in parallel.
While doing this right in bash is probably impossible, you can do a semi-right fairly easily. bstark gave a fair approximation of right but his has the following flaws:
Word splitting: You can't pass any jobs to it that use any of the following characters in their arguments: spaces, tabs, newlines, stars, question marks. If you do, things will break, possibly unexpectedly.
It relies on the rest of your script to not background anything. If you do, or later you add something to the script that gets sent in the background because you forgot you weren't allowed to use backgrounded jobs because of his snippet, things will break.
Another approximation which doesn't have these flaws is the following:
scheduleAll() {
local job i=0 max=4 pids=()
for job; do
(( ++i % max == 0 )) && {
wait "${pids[#]}"
pids=()
}
bash -c "$job" & pids+=("$!")
done
wait "${pids[#]}"
}
Note that this one is easily adaptable to also check the exit code of each job as it ends so you can warn the user if a job fails or set an exit code for scheduleAll according to the amount of jobs that failed, or something.
The problem with this code is just that:
It schedules four (in this case) jobs at a time and then waits for all four to end. Some might be done sooner than others which will cause the next batch of four jobs to wait until the longest of the previous batch is done.
A solution that takes care of this last issue would have to use kill -0 to poll whether any of the processes have disappeared instead of the wait and schedule the next job. However, that introduces a small new problem: you have a race condition between a job ending, and the kill -0 checking whether it's ended. If the job ended and another process on your system starts up at the same time, taking a random PID which happens to be that of the job that just finished, the kill -0 won't notice your job having finished and things will break again.
A perfect solution isn't possible in bash.
Maybe try a parallelizing utility instead rewriting the loop? I'm a big fan of xjobs. I use xjobs all the time to mass copy files across our network, usually when setting up a new database server.
http://www.maier-komor.de/xjobs.html
function for bash:
parallel ()
{
awk "BEGIN{print \"all: ALL_TARGETS\\n\"}{print \"TARGET_\"NR\":\\n\\t#-\"\$0\"\\n\"}END{printf \"ALL_TARGETS:\";for(i=1;i<=NR;i++){printf \" TARGET_%d\",i};print\"\\n\"}" | make $# -f - all
}
using:
cat my_commands | parallel -j 4
Really late to the party here, but here's another solution.
A lot of solutions don't handle spaces/special characters in the commands, don't keep N jobs running at all times, eat cpu in busy loops, or rely on external dependencies (e.g. GNU parallel).
With inspiration for dead/zombie process handling, here's a pure bash solution:
function run_parallel_jobs {
local concurrent_max=$1
local callback=$2
local cmds=("${#:3}")
local jobs=( )
while [[ "${#cmds[#]}" -gt 0 ]] || [[ "${#jobs[#]}" -gt 0 ]]; do
while [[ "${#jobs[#]}" -lt $concurrent_max ]] && [[ "${#cmds[#]}" -gt 0 ]]; do
local cmd="${cmds[0]}"
cmds=("${cmds[#]:1}")
bash -c "$cmd" &
jobs+=($!)
done
local job="${jobs[0]}"
jobs=("${jobs[#]:1}")
local state="$(ps -p $job -o state= 2>/dev/null)"
if [[ "$state" == "D" ]] || [[ "$state" == "Z" ]]; then
$callback $job
else
wait $job
$callback $job $?
fi
done
}
And sample usage:
function job_done {
if [[ $# -lt 2 ]]; then
echo "PID $1 died unexpectedly"
else
echo "PID $1 exited $2"
fi
}
cmds=( \
"echo 1; sleep 1; exit 1" \
"echo 2; sleep 2; exit 2" \
"echo 3; sleep 3; exit 3" \
"echo 4; sleep 4; exit 4" \
"echo 5; sleep 5; exit 5" \
)
# cpus="$(getconf _NPROCESSORS_ONLN)"
cpus=3
run_parallel_jobs $cpus "job_done" "${cmds[#]}"
The output:
1
2
3
PID 56712 exited 1
4
PID 56713 exited 2
5
PID 56714 exited 3
PID 56720 exited 4
PID 56724 exited 5
For per-process output handling $$ could be used to log to a file, for example:
function job_done {
cat "$1.log"
}
cmds=( \
"echo 1 \$\$ >\$\$.log" \
"echo 2 \$\$ >\$\$.log" \
)
run_parallel_jobs 2 "job_done" "${cmds[#]}"
Output:
1 56871
2 56872
The project I work on uses the wait command to control parallel shell (ksh actually) processes. To address your concerns about IO, on a modern OS, it's possible parallel execution will actually increase efficiency. If all processes are reading the same blocks on disk, only the first process will have to hit the physical hardware. The other processes will often be able to retrieve the block from OS's disk cache in memory. Obviously, reading from memory is several orders of magnitude quicker than reading from disk. Also, the benefit requires no coding changes.
This might be good enough for most purposes, but is not optimal.
#!/bin/bash
n=0
maxjobs=10
for i in *.m4a ; do
# ( DO SOMETHING ) &
# limit jobs
if (( $(($((++n)) % $maxjobs)) == 0 )) ; then
wait # wait until all have finished (not optimal, but most times good enough)
echo $n wait
fi
done
Here is how I managed to solve this issue in a bash script:
#! /bin/bash
MAX_JOBS=32
FILE_LIST=($(cat ${1}))
echo Length ${#FILE_LIST[#]}
for ((INDEX=0; INDEX < ${#FILE_LIST[#]}; INDEX=$((${INDEX}+${MAX_JOBS})) ));
do
JOBS_RUNNING=0
while ((JOBS_RUNNING < MAX_JOBS))
do
I=$((${INDEX}+${JOBS_RUNNING}))
FILE=${FILE_LIST[${I}]}
if [ "$FILE" != "" ];then
echo $JOBS_RUNNING $FILE
./M22Checker ${FILE} &
else
echo $JOBS_RUNNING NULL &
fi
JOBS_RUNNING=$((JOBS_RUNNING+1))
done
wait
done
You can use a simple nested for loop (substitute appropriate integers for N and M below):
for i in {1..N}; do
(for j in {1..M}; do do_something; done & );
done
This will execute do_something N*M times in M rounds, each round executing N jobs in parallel. You can make N equal the number of CPUs you have.
My solution to always keep a given number of processes running, keep tracking of errors and handle ubnterruptible / zombie processes:
function log {
echo "$1"
}
# Take a list of commands to run, runs them sequentially with numberOfProcesses commands simultaneously runs
# Returns the number of non zero exit codes from commands
function ParallelExec {
local numberOfProcesses="${1}" # Number of simultaneous commands to run
local commandsArg="${2}" # Semi-colon separated list of commands
local pid
local runningPids=0
local counter=0
local commandsArray
local pidsArray
local newPidsArray
local retval
local retvalAll=0
local pidState
local commandsArrayPid
IFS=';' read -r -a commandsArray <<< "$commandsArg"
log "Runnning ${#commandsArray[#]} commands in $numberOfProcesses simultaneous processes."
while [ $counter -lt "${#commandsArray[#]}" ] || [ ${#pidsArray[#]} -gt 0 ]; do
while [ $counter -lt "${#commandsArray[#]}" ] && [ ${#pidsArray[#]} -lt $numberOfProcesses ]; do
log "Running command [${commandsArray[$counter]}]."
eval "${commandsArray[$counter]}" &
pid=$!
pidsArray+=($pid)
commandsArrayPid[$pid]="${commandsArray[$counter]}"
counter=$((counter+1))
done
newPidsArray=()
for pid in "${pidsArray[#]}"; do
# Handle uninterruptible sleep state or zombies by ommiting them from running process array (How to kill that is already dead ? :)
if kill -0 $pid > /dev/null 2>&1; then
pidState=$(ps -p$pid -o state= 2 > /dev/null)
if [ "$pidState" != "D" ] && [ "$pidState" != "Z" ]; then
newPidsArray+=($pid)
fi
else
# pid is dead, get it's exit code from wait command
wait $pid
retval=$?
if [ $retval -ne 0 ]; then
log "Command [${commandsArrayPid[$pid]}] failed with exit code [$retval]."
retvalAll=$((retvalAll+1))
fi
fi
done
pidsArray=("${newPidsArray[#]}")
# Add a trivial sleep time so bash won't eat all CPU
sleep .05
done
return $retvalAll
}
Usage:
cmds="du -csh /var;du -csh /tmp;sleep 3;du -csh /root;sleep 10; du -csh /home"
# Execute 2 processes at a time
ParallelExec 2 "$cmds"
# Execute 4 processes at a time
ParallelExec 4 "$cmds"
$DOMAINS = "list of some domain in commands"
for foo in some-command
do
eval `some-command for $DOMAINS` &
job[$i]=$!
i=$(( i + 1))
done
Ndomains=echo $DOMAINS |wc -w
for i in $(seq 1 1 $Ndomains)
do
echo "wait for ${job[$i]}"
wait "${job[$i]}"
done
in this concept will work for the parallelize. important thing is last line of eval is '&'
which will put the commands to backgrounds.

Resources