A quick-n-dirty memory measurement. #proc-meminfo #memory

README-memory-measurement.md

Take a measurement via /proc/meminfo and pipe it to a parser AWK script to turn it into a data row:

cat /proc/meminfo | parse-meminfo.awk

Maybe append it to a series of measurements (removing the extra header line from each measurement):

cat /proc/meminfo | parse-meminfo.awk | tail -n +2 | tee --append memory-measurements.csv

Plot measurements as time series:

gnuplot -c plot-memory-measurements.gnuplot memory-measurements.csv | display -

parse-meminfo.awk

#!/usr/bin/awk -f

BEGIN {
    FS=":"
    
    total = 0;
    free = 0;
    cached = 0;
    buffers = 0;
    
    timestamp = systime(); # epoch seconds (as in `data +%s`)

    printf ("timestamp;total;free;cached;buffers\n");    
}

/^MemTotal:/ { total = int($2); }

/^MemFree:/ { free = int($2); }

/^Cached:/ { cached = int($2); }

/^Buffers:/ { buffers = int($2); }

END {
    printf("%d;%d;%d;%d;%d\n", timestamp, total, free, cached, buffers);
}

plot-memory-measurements.gnuplot

DATAFILE = ARG1

set grid 
set datafile separator ";"

set terminal png font "sans-serif,10"

set style fill solid 0.33

# the format to read x datapoints from input
set timefmt "%s"

# the format to write datapoints at x axis
set xdata time
#set format x "%b %d\n%H:%M"
set format x "%R"

# place a X tick every 2h
set xtic 7200

TITLE_1 = "Memory Used (%)"

plot DATAFILE using ($1):(100 * ($2 - $3 - $4)/($2)) with lines title TITLE_1