fioparser.sh

#!/bin/bash

# This is a wrapper script which will run fioparser 4 times with different IO patterns.
# Courtesy of this article: https://anfcommunity.com/2020/11/20/how-to-nfs-performance-assessment-using-fio-and-fio-parser/

# PREREQUISITES

# 1.    fio (apt install fio || yum install fio)
# 2.    fioparser (git clone https://github.com/jtulak/fio-parser.git)
# 3.    python2.x (apt install python2 || yum install python2 )
# 4.    numpy, hurry (pip2 install numpy && pip2 install hurry)

# OPTIONS:

# --w <working directory>
#       This should point to a directory on whatever volume you're wanting to test. By default it points to /ontap/working,
#       which means it expects you to have a netapp nfs volume mounted to /ontap

# --o <output directory>
#       Where you want the results output to

# EXAMPLE:
#       ./fioparser.sh --w /zfs/working --o /zfs/output

#       In the above example, I am testing an FSX for OpenZFS volume mounted to /zfs

# NOTES:

#   This takes an hour to run. The part that says "for i in 1 2 3 4 ..." those numbers are talking about the IO depth. 
#   You could reduce the runtime of this to 30 min or so by capping it at 10, or 15 min by capping it at 5, but you get
#   less information about how more heavy-hitting workloads will behave

# set defaults here
WORKING="/ontap/working"
OUTPUT="/ontap/output"

w=${w:-$WORKING}
o=${o:-$OUTPUT}

while [ $# -gt 0 ]; do
    if [[ $1 == *"--"* ]]; then
        param="${1/--/}"
        declare $param="$2"
    fi
    shift
done

for i in 1 2 3 4 5 6 7 8 9 10 15 20 25 30 40 50 60 70 80 90 100; do 
    fio --name=fiotest --directory=$WORKING --ioengine=libaio --direct=1 --numjobs=2 --nrfiles=4 --runtime=30 --group_reporting --time_based --stonewall --size=4G --ramp_time=20 --bs=64k --rw=read --iodepth=$i --fallocate=none --output=$OUTPUT/$(uname -n)-seqread-$i; 
done

for i in 1 2 3 4 5 6 7 8 9 10 15 20 25 30 40 50 60 70 80 90 100; do 
    fio --name=fiotest --directory=$WORKING --ioengine=libaio --direct=1 --numjobs=2 --nrfiles=4 --runtime=30 --group_reporting --time_based --stonewall --size=4G --ramp_time=20 --bs=64k --rw=write --iodepth=$i --fallocate=none --output=$OUTPUT/$(uname -n)-seqwrite-$i;
done

for i in 1 2 3 4 5 6 7 8 9 10 15 20 25 30 40 50 60 70 80 90 100; do 
    fio --name=fiotest --directory=$WORKING --ioengine=libaio --direct=1 --numjobs=2 --nrfiles=4 --runtime=30 --group_reporting --time_based --stonewall --size=4G --ramp_time=20 --bs=8k --rw=randread --iodepth=$i --fallocate=none --output=$OUTPUT/$(uname -n)-randread-$i;
done

for i in 1 2 3 4 5 6 7 8 9 10 15 20 25 30 40 50 60 70 80 90 100; do
    fio --name=fiotest --directory=$WORKING --ioengine=libaio --direct=1 --numjobs=2 --nrfiles=4 --runtime=30 --group_reporting --time_based --stonewall --size=4G --ramp_time=20 --bs=8k --rw=randwrite --iodepth=$i --fallocate=none --output=$OUTPUT/$(uname -n)-randwrite-$i;
done

/ontap/fio-parser/fio-parser.py -d $OUTPUT