Filter by process name and log CPU usage
LinuxUnixLinux Problem Overview
Is there an option for linux top command where i can filter processes by name and write the CPU usage of that process every 1 second to a log file?
Linux Solutions
Solution 1 - Linux
top
& pgrep
To filter the output of top
by process name, you can use pgrep
to get a list of PIDs by process name then pass them to the -p
option of top
.
For example:
top -p $(pgrep -d',' http)
Note: the -d','
option delimits the PIDs with commas, which is what is expected by the top -p
.
Note 2: top
will return a failure message if there are no running processes that match the name you specify in pgrep
.
To write the results of top
to a file, use the -n 1
option (only one iteration) and redirect the output to your log file.
top -p $(pgrep -d',' http) -n 1 >> your_log_file
To do that every second, perhaps a while
loop with a sleep
would do?
while :; do top -p $(pgrep -d',' http) -n 1 >> your_log_file; sleep 1; done
To timestamp each entry, you can append the output of date
. E.g.
while :; do top -p $(pgrep -d',' http) -n 1 >> log.txt; date >> log.txt; sleep 1; done
Solution 2 - Linux
Another option is:
top -b -d 1 -p $(pgrep -d',' java) -n 120 > log.txt
- The option -d allows to set the frequency used by top to refresh the data.
- The option -b means that the traditional interface of top is not used. Instead, it sends everything to the standard output and then you can use a pipe (|) or a redirection (>).
- The option -n informs about the number of iterations top should execute.
After that you can type:
cat log.txt | grep USER_OF_PROCESS
You will see the execution time of the process and also %CPU, Memory and all that.
Solution 3 - Linux
#You can run following script as ./cpurecorder.sh pid filename
#It will generate output file with memory usage and cpu utilisation.
#You can log other variable by searching man for ps.
`enter code here`filepath=/home/rtcsadm # modify as desired
interval=20 # reports per minute
timelimit=6000 # how long to run, in seconds
mydate=`date "+%H:%M:%S"` # the timestamp
freq=$((60/$interval)) # for sleep function
while [ "$SECONDS" -le "$timelimit" ] ; do
ps -p$1 -opid -opcpu -opmem -ocomm -c | grep $1 | sed "s/^/$mydate /" >> $filepath/$2.txt
sleep 3
mydate=`date "+%H:%M:%S"`
done