Linux `mpstat` 无法从 crontab 脚本中按预期工作

Question

I have a script running from root crontab:

05 * * * * /bin/bash /root/systat_kpi_tools/systat_kpi.sh

The script has an mpstat output formatted and saved in CSV format in a file:

[[ -e ${BASE_DIR}/cpu_stat_${NOW}.csv ]] || echo "Year,Month,Day,Hr,Min,Sec,%user,%nice,%sys,%iowait,%irq,%soft,%steal,%idle,intr/s" >> ${BASE_DIR}/cpu_stat_${NOW}.csv
mpstat | awk 'NR>3{ print substr($0, index($0,$4)) }' | awk -v dt=$(date +'%Y,%m,%d,%H,%M,%S') -v OFS=, '{$1=dt OFS $1; print}' >> ${BASE_DIR}/cpu_stat_${NOW}.csv

Output of the file:

# cat cpu_stat_160526.csv
Year,Month,Day,Hr,Min,Sec,%user,%nice,%sys,%iowait,%irq,%soft,%steal,%idle,intr/s
2016,05,26,14,05,01,0.00,2.17,0.08,0.01,0.14,0.00,90.29,1093.83
2016,05,26,15,05,02,0.00,2.19,0.08,0.01,0.14,0.00,90.25,1093.82

When I run the command out of the script:

# echo "Year,Month,Day,Hr,Min,Sec,%user,%nice,%sys,%iowait,%irq,%soft,%steal,%idle,intr/s" > tmp
# mpstat | awk 'NR>3{ print substr($0, index($0,$4)) }' | awk -v dt=$(date +'%Y,%m,%d,%H,%M,%S') -v OFS=, '{$1=dt OFS $1; print}' >> tmp
# mpstat | awk 'NR>3{ print substr($0, index($0,$4)) }' | awk -v dt=$(date +'%Y,%m,%d,%H,%M,%S') -v OFS=, '{$1=dt OFS $1; print}' >> tmp
# mpstat | awk 'NR>3{ print substr($0, index($0,$4)) }' | awk -v dt=$(date +'%Y,%m,%d,%H,%M,%S') -v OFS=, '{$1=dt OFS $1; print}' >> tmp

I get this output:

# cat tmp

Year,Month,Day,Hr,Min,Sec,%user,%nice,%sys,%iowait,%irq,%soft,%steal,%idle,intr/s
2016,05,26,15,04,54,7.33,0.00,2.19,0.08,0.01,0.14,0.00,90.25,1093.82
2016,05,26,15,04,57,7.33,0.00,2.19,0.08,0.01,0.14,0.00,90.25,1093.82
2016,05,26,15,04,58,7.33,0.00,2.19,0.08,0.01,0.14,0.00,90.25,1093.82

Watch out for the column %user having values 7.33,7.33,7.33 which is missing out from previous output when the same line executes from script. Why such weird behavior ?

My Linux flavor is RHEL5.10 64 bits : kernel-2.6.18-371.el5

Answer 1

I know this is a bit old, but I just ran into this today. As you noted, sourcing the .bash_profile fixed the problem. In the man page, I found this quote:

ENVIRONMENT The mpstat command takes into account the following environment variable:

 S_TIME_FORMAT If this variable exists and its value is ISO then the current locale will be ignored when printing the date in the report header. The mpstat command will use the ISO 8601 format (YYYY-MM-DD) instead.

For me the major difference was a 12 vs 24 hour clock, hence needing to awk fields $4,$6 vs $3,$5.

Answer 2

You need to use:

mpstat 1 1

Here the explanation why: mpstat with parameters That will give you only 1 line, and the averages at the bottom. Then, you need to parse that one line (line #4). Also, there is no need to parse the date (or time) given by mpstat , you can instead get the current date of the system using $(date +'FORMAT') .

Here the script I made, and it works for me. It creates a file if it doesn't exist, and then it adds the results from mpstat 1 1 . I use \\t as separator; if you like you can use , and change the extension of the file to *.csv . You can then add this script to the crontab to run every x minutes.

#! /bin/bash
DIR=/somepath
FN=$(date +'CPU-%Y-%m.log')
FN=$DIR/$FN
if [ ! -f $FN ]
then
    touch $FN
    chmod 644 $FN
    printf "Server\tDate\t\t\t%%usr\t%%nice\t%%sys\t%%iowait\t%%irq\t%%soft\t%%steal\t%%guest\t%%gnice\t%%idle\n" > $FN
fi
DATE=$(date +'%Y-%m-%d %H:%M:%S')
SERVER="Jupyter"
STAT=$(mpstat 1 1 | awk 'NR==4{printf "%.2f\t%.2f\t%.2f\t%.2f\t%.2f\t%.2f\t%.2f\t%.2f\t%.2f\t%.2f\t", $4, $5, $6, $7, $8, $9, $10, $11, $12, $13}')
echo -e "$SERVER\t$DATE\t$STAT" >> $FN