I want to find out processes running more than 3 hrs, I have written a command for this but it's not returning expected output
ps -u <user> -o pid,stime,pcpu,pmem,etime,cmd --sort=start_time | \
grep <searchString> | grep -v grep| awk '{print $5}' | \
sed 's/:|-/ /g;'| awk '{print $4" "$3" "$2" "$1"}' | \
awk '$1+$2*60+$3*3600+$4*86400 > 10800'
but it's printing the values of etime in output. But expected output is, command should print the values of "pid,stime,pcpu,pmem,etime,cmd"
I am not able to find exact issue with this.
You are executing "awk '{print $5}'" which is taking in the input and printing out only column 5 which in your case is "etime" , everything from this point on is lost.
If your system supports etimes (notice the s on the end), you can easily do this with
ps -eo pid,etimes,etime,comm,user,tty | awk '{if ( $2>10800) print $0}'
on a system not supporting etimes which has a standard output of etime which hh:mm:ss or just mm:ss if no hours have passed
ps -eo pid,etime,comm,user,tty | awk '{seconds_old=10800 ; split($2,a,":",sep) ; if(length(a) < 3) b = (a[1] *60) + (a[2]) ; else b=((a[1]*3600) + (a[2] *60) + (a[3])) ; if(b > seconds_old ) print $0}'
Adjust "seconds_old" to change the age you want to test for:
There are various other methods of doing this using Find for example:
explained here:
https://serverfault.com/questions/181477/how-do-i-kill-processes-older-than-t
However, the solution should match your expected output
Try this:
ps -u <user> -o pid,stime,pcpu,pmem,etime=,cmd --sort=start_time|grep <searchString>|while read z;do tago=$(echo $z|awk '{print $5}'|sed -E 's/(:|-)/ /g'| awk '{print $4+$3*60+$2*3600+$1*86400}');if [ $tago -ge 10800 ];then echo $z;fi;done
It prints only processes >= 10800 secs old.
You can readjust the output further to fit your needs.
Able to find running process for more than 3 hrs with below command.
ps -u <user> -o pid,stime,pcpu,pmem,etime,cmd --sort=start_time |grep -v grep|awk 'substr($0,23,2) > 3'
So I am trying to check the internal CPU and RAM usage from JVM's.
The set up is that we have a single server which hosts 39 JVM's each running its own service and has it's own unique CPU & RAM allocation. So for example it could be set up like the following:
CPU(Cores) RAM(MB)
JVM1 2 300
JVM2 1 50
JVM3 5 1024
These are fictional as I don't have the actual values to hand.
I know the PID's for each of the JVM's but I am wondering how would I see the CPU and RAM usage of each JVM on it's own dissregarding the Host systems usage. Also is it possible to pass multiple PID's into a jstat -gc as I will be looking to script this?.
I know that if I use:
ps -p <PID> -o %CPU %RAM
That will give me the CPU and RAM of that process on the host machine.
Any Help is greatly appreciated.
After playing around for a while I came up with the following script which pulls the memory usage and total memory in use to 2DP:
Script
#!/bin/sh
service=$(mktemp)
jstatop=$(mktemp)
ServiceStatsPre=$(mktemp)
datet=$(date +%Y-%m-%d)
hourt=$(date +%H)
ps -ef | grep java | grep Service | awk '{key=substr($9,31,match($9,"#")-31); serv[key]++}END{for(name in serv){if($9 ~ /Service/){print name}}}'>$service
printf "%-8sService\t\t\t\t%7s PID\t%7s Used\t%7s Total\n" > $ServiceStatsPre
filename=$service
while read -r line; do
if [ ! -z $line ]; then
pid=$(ps -ef | grep java | grep Service |awk -v svc="$line" '{if (match($9,svc)){print $2}}')
rncnt=0
rnag=1
while [ $rnag -eq 1 ]; do
jstat -gc $pid > $jstatop
if [ $? -ne 0 ]; then
sleep 5;
rncnt++;
else
rnag=0
fi
if [ $rncnt -eq 5 ]; then
rnag=0
fi
done
cat $jstatop | awk '{if (NR !=1) print}'|awk -v pid="$pid" -v svc="$line" -v d="$datet" -v h="$hourt" '{printf("%-40s %7d %6.2fMB %6.2fMB %11s %3s\n",svc,pid,($1+$5+$7)/1024,($3+$4+$6+$8)/1024,d,h)}' >> $ServiceStatsPre
fi
done < $filename
#printf "Date,Hour,Service,Used,Total\n" > Service_Stats.csv #Uncomment this line on initial Run to create file with headders
cat $ServiceStatsPre | awk '{if (NR !=1) print}' |
awk '{
printf("%-1s,%1s,%1s,%4.2f,%4.2f\n",$5,$6,$1,$3,$4);
}' >> Service_Stats.csv
rm $service;
rm $jstatop;
rm $ServiceStatsPre;
Breakdown
ps -ef | grep java | grep Service | awk '{key=substr($9,31,match($9,"#")-31); serv[key]++}END{for(name in serv){if($9 ~ /Service/){print name}}}'>$service
This is returning the name of the Service and putting it to a tempary file $service
printf "%-8sService\t\t\t\t%7s PID\t%7s Used\t%7s Total\n" > $ServiceStatsPre
This is just creating Headings in a Temp file $ServiceStatsPre (I know this isn't needed but I was originally using it to debug to a file not a temp file)
Outer While loop
filename=$service
if [ ! -z $line ]; then
while read -r line; do
pid=$(ps -ef | grep java | grep Service |awk -v svc="$line" '{if (match($9,svc)){print $2}}')
rncnt=0
rnag=1
...
cat $jstatop | awk '{if (NR !=1) print}'|awk -v pid="$pid" -v svc="$line" -v d="$datet" -v h="$hourt" '{printf("%-40s %7d %6.2fMB %6.2fMB %11s %3s\n",svc,pid,($1+$5+$7)/1024,($3+$4+$6+$8)/1024,d,h)}' >> $ServiceStatsPre
fi
done < $filename
This is first checking to see if the line that it is passing in is not empty (if [ ! -z $line ]; then) then it is extracting the PID for each of the services in the temp file $service and is setting a check (rnag) + retry counter (rncnt). The final part of this outer while loop is calculating the stats, memory usage in MB (($1+$5+$7)/1024), Total Memory used (($3+$4+$6+$8)/1024) as well as retuning the PID,Service Name,Date and Hour.
Inner While Loop
...
while [ $rnag -eq 1 ]; do
jstat -gc $pid > $jstatop
if [ $? -ne 0 ]; then
sleep 5;
rncnt++;
else
rnag=0
fi
if [ $rncnt -eq 5 ]; then
rnag=0
fi
done
...
This is to basically ensure that the jstat is running correctly and isn't returning an error.
The final part of this script:
#printf "Date,Hour,Service,Used,Total\n" > Service_Stats.csv #Uncomment this line on initial Run to create file with headders
cat $ServiceStatsPre | awk '{if (NR !=1) print}' |
awk '{
printf("%-1s,%1s,%1s,%4.2f,%4.2f\n",$5,$6,$1,$3,$4);
}' >> Service_Stats.csv
is just to resolve the formatting and put it into a format so that it can be used in a CSV.
Output
This script out puts like the following:
Date,Hour,Service,Used,Total
2016-07-03,07,undMFSAdapterServiceVM,331.00,188.87
2016-07-03,07,entSCSServiceVM,332.50,278.19
2016-07-03,07,kServiceVM,457.50,132.91
2016-07-03,07,ServiceTuTrackVm,432.00,282.66
2016-07-03,07,ter1WMSAdapterServiceVm,233.00,77.02
2016-07-03,07,kingMFSAdapterServiceVM,451.50,126.69
2016-07-03,07,erBuilderServiceVM,261.50,211.27
2016-07-03,07,ter3MFSAdapterServiceVM,449.50,210.23
2016-07-03,07,rServiceVM1,1187.00,529.26
2016-07-03,07,rServiceVM2,597.50,398.43
2016-07-03,07,rServiceVM3,2786.00,819.30
2016-07-03,07,rServiceVM4,451.50,163.13
2016-07-03,07,MessagingServiceVm,457.50,357.11
2016-07-03,07,viceVM,444.50,263.59
2016-07-03,07,ServiceVM,1910.50,909.19
2016-07-03,07,undPackingMFSAdapterServiceVM,208.00,113.51
2016-07-03,07,gisticLockServiceVM,245.00,173.05
2016-07-03,07,kingMFSAdapterServiceVM,781.50,327.13
2016-07-03,07,ferWMSAdapterServiceVm,196.00,84.02
2016-07-03,07,geMFSAdapterServiceVM,499.50,256.91
2016-07-03,07,ferMFSAdapterServiceVM,456.50,246.89
2016-07-03,07,kingWMSAdapterServiceVm,195.00,73.70
2016-07-03,07,AdapterServiceVm,149.50,72.62
2016-07-03,07,ter2MFSAdapterServiceVM,455.00,136.02
2016-07-03,07,ionServiceVM,484.00,240.46
2016-07-03,07,ter3WMSAdapterServiceVm,266.00,138.70
2016-07-03,07,iceVm,135.50,106.07
2016-07-03,07,viceVm,3317.00,1882.15
2016-07-03,07,entBCSServiceVM,356.50,143.93
2016-07-03,07,ServiceVm,951.00,227.12
2016-07-03,07,lingServiceVM,145.50,76.61
2016-07-03,07,entDBHServiceVM,182.50,4.63
2016-07-03,07,kingWMSAdapterServiceVm,208.00,103.13
2016-07-03,07,gingServiceVM,1529.50,235.84
2016-07-03,07,ServiceVM,249.50,131.78
2016-07-03,07,ter1MFSAdapterServiceVM,453.00,394.11
2016-07-03,07,AdapterServiceVM,461.00,208.41
2016-07-03,07,ter2WMSAdapterServiceVm,178.50,79.93
2016-07-03,07,AdapterServiceVm,395.00,131.08
2016-07-03,07,ingServiceVM,184.50,126.28
The $3 has been trimmed just to hide service names etc..
Helpful Links
A nice explanation of jstat -gc output
I am trying to figure out how to extract the cpu core utilization for a SINGLE process in linux and parse it. I know that I can get the overall core utilization via top and then press "1". I am already able to parse that. However, now I want to do the same thing for a single process. I tried it with ps and calculating the core utilization myself but I am not so sure if my script is accurate enough, something seems off. (Note that this version calculates overall core utilization since it is WIP) I get errors like this after a while in my terminal:
test.sh: line 31: +: syntax error: operand expected (error token is
"+")
I cannot figure out why this error just randomly occurs.
#!/bin/env bash
read -p "Enter PID to observe:" pid
run=false
ps -p $pid -L -o cputime,etime,psr,pcpu
for (( u = 1 ; u <= 100 ; u++ ))
do
lines=$(ps -p $pid -L -o psr,pcpu | awk 'END{print NR}')
for (( i=1; i<=$lines; i++))
do
core=$(ps -p $pid -L -o psr,pcpu | awk 'NR=='$i'{print $1}')
if [ $run == false ]
then cpuTimeS=+$(ps -p $pid -L -o cputime,etime | awk 'NR=='$i'{print $1}' | awk -F : '{ printf("%.2f\n", $1*60+$2*60+$3); }')
elapsedTimeS=+$(ps -p $pid -L -o cputime,etime | awk 'NR=='$i'{print $2}' | awk -F : '{ printf("%.2f\n", $1*60+$2*60+$3); }')
cpuTimeSi=${cpuTimeS%.*}
elapsedTimeSi=${elapsedTimeS%.*}
cpuTimeSiResult=$(( cpuTimeSi + cpuTimeSiResult ))
elapsedTimeSiResult=$(( elapsedTimeSi + elapsedTimeSiResult ))
else
cpuTimeE=+$(ps -p $pid -L -o cputime,etime | awk 'NR=='$i'{print $1}' | awk -F : '{ printf("%.2f\n", $1*60+$2*60+$3); }')
elapsedTimeE=+$(ps -p $pid -L -o cputime,etime | awk 'NR=='$i'{print $2}' | awk -F : '{ printf("%.2f\n", $1*60+$2*60+$3); }')
cpuTimeEi=${cpuTimeE%.*}
elapsedTimeEi=${elapsedTimeE%.*}
cpuTimeEiResult=$(( cpuTimeEi + cpuTimeEiResult ))
elapsedTimeEiResult=$(( elapsedTimeEi + elapsedTimeEiResult ))
fi
done
if [ "$run" = true ]
then result=$( echo "scale=2; ($cpuTimeEiResult - $cpuTimeSiResult) / ($elapsedTimeEiResult - $elapsedTimeSiResult) * 100.0" | bc)
echo "RESULT:" $result
echo "cpuTimeSTART:" $cpuTimeSiResult
echo "elapsedTimeSTART:" $elapsedTimeSiResult
echo "cpuTimeEND:" $cpuTimeEiResult
echo "elapsedTimeEND:" $elapsedTimeEiResult
fi
sleep 1
if [ "$run" = false ]
then run=true
else
run=false
fi
done
Are there any ideas on how I could solve this better?
I am glad for any advice
Looking at your actual request to get the CPU utilization from a particular process; this could be done in less lines; e.g. below is the CPU usage after giving a seq command.
#using the process name
$ ps -C seq -o %cpu
%CPU
10.4
#using the process id
$ ps -p 5710 -o %cpu
%CPU
10.4
I am trying to write a shell script to monitor file system. script logic is,
for each file system from df -H command, read the file system threshold file and get the critical threshold, warning threshold. Based on the condition, it will send notification.
Here is my script:
#!/bin/sh
df -H | grep -vE '^Filesystem|none|boot|tmp|tmpfs' | awk '{ print $5 " " $6 }' | while read $output
do
echo $output
fsuse=$(echo $output | awk '{ print $1}' | cut -d'%' -f1 )
fsname=$(echo $output | awk '{ print $2 }' )
server=`cat /workspace/OSE/scripts/fs_alert|grep -w $fsname|awk -F":" '{print $2}'`
fscrit=`cat /workspace/OSE/scripts/fs_alert|grep -w $fsname|awk -F":" '{print $3}'`
fswarn=`cat /workspace/OSE/scripts/fs_alert|grep -w $fsname|awk -F":" '{print $4}'`
serenv=`cat /workspace/OSE/scripts/fs_alert|grep -w $fsname|awk -F":" '{print $5}'`
if [ $fsuse -ge $fscrit ]; then
message="CRITICAL:${server}:${serenv}:$fsname Is $fsuse Filled"
_notify;
elif [ $fsuse -gt $fswarn ] && [ $fsuse -lt $fscrit ]; then
message="WARNING: $fsname is $fsuse Filled"
_notify;
else
echo "File system space looks good"
fi
done
Here is /workspace/OSE/scripts/fs_alert:
/:hlpdbq001:90:80:QA:dba_mail
/dev/shm:hlpdbq001:90:80:QA:dba_mail
/boot:hlpdbq001:90:80:QA:dba_mail
/home:hlpdbq001:90:80:QA:dba_mail
/opt:hlpdbq001:90:80:QA:dba_mail
/opt/security:hlpdbq001:90:80:QA:dba_mail
/tmp:hlpdbq001:90:80:QA:dba_mail
/var:hlpdbq001:90:80:QA:dba_mail
/u01/app:hlpdbq001:90:80:QA:dba_mail
/u01/app/oracle:hlpdbq001:90:80:QA:dba_mail
/oratrace:hlpdbq001:90:80:QA:dba_mail
/u01/app/emagent:hlpdbq001:90:80:QA:dba_mail
/gg:hlpdbq001:90:80:QA:dba_mail
/workspace:hlpdbq001:90:80:QA:dba_mail
/dbaudit:hlpdbq001:90:80:QA:dba_mail
/tools:hlpdbq001:90:80:QA:dba_mail
My problem is when the script is trying to get crit_va, warn_val from the file for /u01 file system, I am getting three results. How do I get/filter one file system at a time?
$ df -H|grep /u01
/dev/mapper/datavg-gridbaselv 53G 12G 39G 24% /u01/app
/dev/mapper/datavg-rdbmsbaselv 53G 9.6G 41G 20% /u01/app/oracle
/dev/mapper/datavg-oemagentlv 22G 980M 20G 5% /u01/app/emagent
what is the best way to handle this issue?
do i need logic based on Filesystem or Mounted on.
Don't reinvent the wheel. There are tools out there that can do this for your. Try monit for example:
http://sysadminman.net/blog/2011/monit-disk-space-monitoring-1716
well, monit is fine ), if you need alternative take a look at df-check - a wrapper for df utility to verify that thresholds are not exceeded , on per partition basis. At least it seems very close to what you started to implement in your bash script, but it's written on perl and has a neat and simple installation layout. ready to use tool.
-- Regards
PS discloser - I am the tool author
I need to calculate CPU usage (user mode, system mode, idle time) of a process in Linux.
I am able to calculate usage in user and system mode using utime and stime values from /proc/PID/stat, but I found nothing which is related to idle time.
I know I can get idle time from /proc/stat but this value is related to machine, not for particular process.
Is it possible to calculate idle time of a process knowing its PID (reading data from /proc directory)?
I don't know much about it but maybe the following works:
1) Get the process start up time. Im sure thats possible
2) Generate time difference (dTime = CurrentTime - TimeProcessStarted)
3) Substract the time the process is running ( dTime - (usageSystemMode + usageUserMode))
Hope this helps! :D
it's too late but, I guessed this command useful:
IFS=$'\n';for i in `ps -eo uname:20,pid,cmd | grep -v "USER\|grep\|root"`; \
do if [ $(id -g `echo $i | cut -d" " -f1`) -gt 1000 ] && \
[ $(echo $((($(date +%s) - $(date -d "$(ll -u \
--time-style=+"%y-%m-%d %H:%M:%S" /proc/$(echo $i | \
awk '{print $2}')/cwd | awk '{print $6" "$7}')" +%s))/3600))) >=1 ]; \
then echo $i; fi; done
to use it in bash file:
#!/bin/bash
IFS=$'\n'
for i in `ps -eo uname:20,pid,cmd | grep -v "USER\|grep\|root"`
do
Name="`echo $i | cut -d' ' -f1`"
Id="$(id -g $Name)"
Pid="`echo $i | awk '{print $2}'`"
Time1=$(date +%s)
Time2=$(date -d "$(/usr/bin/ls -lu --time-style=+"%y-%m-%d %H:%M:%S" \
/proc/$Pid/cwd | awk '{print $6" "$7}')" +%s)/3600
Time=$Time1-$Time2
if [ $Id -gt 1000 ] && [ $Time >=1 ]
then
echo $i
fi
done
you could change grep -v "grep\|root" as you wish.
this one line command list all processes which not root owner or system users.