######################################################################
# customize per survey
-# the SG device to measure
-dev=/dev/sg6
+# CHOOSE EITHER scsidevs or rawdevs
+# the SCSI devices to measure - WARNING: will be erased.
+# The raw devices to use
+# rawdevs=${rawdevs:-"/dev/raw/raw1"}
+# scsidevs=`ls /dev/sd[a-z] /dev/sd[a-z][a-z]` # all devices, if you use udev
-# result file prefix
+# result file prefix. date/time+hostname makes unique
# NB ensure the path exists if it includes subdirs
-rslt=/tmp/sg_dd_rslt
+rslt_loc=${rslt_loc:-"/tmp"}
+rslt=${rslt:-"$rslt_loc/sgpdd_survey_`date +%F@%R`_`uname -n`"}
# what to do (read or write)
-action=write
+actions=${actions:-"write read"}
-# total size (MBytes)
+# total size per device (MBytes)
# NB bigger than device cache is good
-size=8192
+size=${size:-8192}
# record size (KBytes)
-rszlo=1024
-rszhi=1024
+rszlo=${rszlo:-1024}
+rszhi=${rszhi:-1024}
-# Concurrent regions
-crglo=1
-crghi=4
+# Concurrent regions per device
+crglo=${crglo:-1}
+crghi=${crghi:-256}
-# total numbers of threads to share between concurrent regions
-# NB survey skips over #thr < #regions
-thrlo=1
-thrhi=1024
+# threads to share between concurrent regions per device
+# multiple threads per region simulates a deeper request queue
+# NB survey skips over #thr < #regions and #thr/#regions > SG_MAX_QUEUE
+thrlo=${thrlo:-1}
+thrhi=${thrhi:-4096}
#####################################################################
+# leave the rest of this alone unless you know what you're doing...
-# disk block size (Bytes)
-bs=512
+# and max # threads one instance will spawn
+SG_MAX_QUEUE=16
-if [ $action = read ]; then
- f1="if=$dev"
- f2="of=/dev/null"
- skip=skip
+# map given device names into SG device names
+i=0
+devs=()
+if [ "$scsidevs" ]; then
+ # we will test for a LUN, the test for a partition
+ # if the partition number is > 9 this will fail
+ for d in $scsidevs; do
+ devs[$i]=`sg_map | awk "{if (\\\$2 == \"$d\") print \\\$1}"`
+ if [ -z "${devs[i]}" ]; then
+ echo "Can't find SG device for $d, testing for partition"
+ pt=`echo $d | sed 's/[0-9]$//'`
+ # Try again
+ devs[$i]=`sg_map | awk "{if (\\\$2 == \"$pt\") print \\\$1}"`
+ if [ -z "${devs[i]}" ]; then
+ echo "Can't find SG device $pt"
+ exit 1
+ fi
+ fi
+ i=$((i+1))
+ done
+elif [ "$rawdevs" ]; then
+ for r in $rawdevs; do
+ RES=`raw -q $r`
+ if [ $? -eq 0 ];then
+ devs[$i]=$r
+ i=$((i+1))
+ else
+ echo "Raw device $r not set up"
+ exit 1
+ fi
+ done
else
- f1="if=/dev/zero"
- f2="of=$dev"
- skip=seek
+ echo "Must specify scsidevs or rawdevs"
+ exit 1
fi
-start=`date +%F@%R`
-rsltf=${rslt}_${start}.summary
+ndevs=${#devs[@]}
+
+# determine block size. This should also work for raw devices
+# If it fails, set to 512
+bs=$((`sg_readcap -b ${devs[0]} | awk '{print $2}'`))
+if [ $bs == 0 ];then
+ echo "sg_readcap failed, setting block size to 512"
+ bs=512
+fi
+rsltf=${rslt}.summary
+workf=${rslt}.detail
echo -n > $rsltf
-workf=${rslt}_${start}.detail
echo -n > $workf
print_summary () {
echo $minusn "$*"
}
+print_summary "$(date) sgpdd-survey on $rawdevs$scsidevs from $(hostname)"
+
for ((rsz=$rszlo;rsz<=$rszhi;rsz*=2)); do
for ((crg=$crglo;crg<=$crghi;crg*=2)); do
for ((thr=$thrlo;thr<=$thrhi;thr*=2)); do
- if ((thr < crg)); then
+ if ((thr < crg || thr/crg > SG_MAX_QUEUE)); then
continue
fi
# compute parameters
# show computed parameters
actual_rsz=$((bpt*bs/1024))
actual_size=$((bs*count*crg/1024))
- str=`printf 'total_size %8dK rsz %4d crg %5d thr %3d ' \
- $actual_size $actual_rsz $crg $thr`
+ str=`printf 'total_size %8dK rsz %4d crg %5d thr %5d ' \
+ $((actual_size*ndevs)) $actual_rsz $((crg*ndevs)) $((thr*ndevs))`
echo "==============> $str" >> $workf
print_summary -n "$str"
freemem=`awk < /proc/meminfo '/^MemTotal:/ {printf "%d\n", $2}'`
- if (((actual_rsz * thr /crg + 64) * crg > freemem)); then
+ if (((actual_rsz*thr/crg + 64)*crg*ndevs > freemem)); then
print_summary "ENOMEM"
continue
fi
- # start test
- t0=`date +%s.%N`
- for ((i=0;i<crg;i++)); do
- sgp_dd 2> ${rslt}_tmp${i} \
- $f1 $f2 ${skip}=$((1024+i*blocks)) \
- thr=$((thr/crg)) count=$count bs=$bs bpt=$bpt time=1&
- done
- wait
- t1=`date +%s.%N`
- # collect all results in 1 file
- rfile=${rslt}_thr${thr}_crg${crg}_rsz${rsz}
- echo > $rfile
- ok=0
- for ((i=0;i<crg;i++)); do
- rtmp=${rslt}_tmp${i}
- if grep 'time to transfer data' $rtmp > /dev/null 2>&1; then
- ok=$((ok + 1))
+ # run tests
+ for action in $actions; do
+ print_summary -n "$action "
+ echo "=====> $action" >> $workf
+ tmpf=${workf}_tmp
+ # start test
+ t0=`date +%s.%N`
+ for ((i=0;i<ndevs;i++)); do
+ dev=${devs[i]}
+ devsize=$((bs*`sg_readcap -b ${dev} | awk '{print $1}'`/1024))
+ if [ $devsize -lt $actual_size ]; then
+ echo -e "device $(sg_map | grep $dev | awk '{ print $2; }') not big enough: $devsize < $actual_size.\nConsider reducing \$size"
+ exit 1
+ fi
+ if [ $action = read ]; then
+ inf="if=$dev"
+ outf="of=/dev/null"
+ skip=skip
+ else
+ inf="if=/dev/zero"
+ outf="of=$dev"
+ skip=seek
+ fi
+ for ((j=0;j<crg;j++)); do
+ sgp_dd 2> ${tmpf}_${i}_${j} \
+ $inf $outf ${skip}=$((1024+j*blocks)) \
+ thr=$((thr/crg)) count=$count bs=$bs bpt=$bpt time=1&
+ done
+ done
+ wait
+ t1=`date +%s.%N`
+ # collect/check individual stats
+ echo > $tmpf
+ ok=0
+ for ((i=0;i<ndevs;i++)); do
+ for ((j=0;j<crg;j++)); do
+ rtmp=${tmpf}_${i}_${j}
+ if grep 'error' $rtmp > /dev/null 2>&1; then
+ echo "Error found in $rtmp"
+ elif grep 'time to transfer data' $rtmp > /dev/null 2>&1; then
+ ok=$((ok + 1))
+ fi
+ cat ${rtmp} >> $tmpf
+ cat ${rtmp} >> $workf
+ rm ${rtmp}
+ done
+ done
+ if ((ok != ndevs*crg)); then
+ print_summary -n "$((ndevs*crg - ok)) failed "
+ else
+ # compute MB/sec from elapsed
+ bw=`awk "BEGIN {printf \"%7.2f MB/s\", $actual_size * $ndevs / (( $t1 - $t0 ) * 1024); exit}"`
+ # compute MB/sec from nregions*slowest
+ check=`awk < $tmpf \
+ '/time to transfer data/ {mb=$8/1.048576; if (n == 0 || mb < min) min = mb; n++}\
+ END {printf "%5d x %6.2f = %7.2f MB/s", n, min, min * n}'`
+ print_summary -n "$bw $check "
fi
- cat ${rslt}_tmp${i} >> $rfile
- cat ${rslt}_tmp${i} >> $workf
- rm ${rslt}_tmp${i}
+ rm $tmpf
done
- if [ $ok -ne $crg ]; then
- print_summary `printf "failed %d" $((crg - ok))`
- else
- # compute MB/sec from elapsed
- bw=`awk "BEGIN {printf \"%6.2f MB/s\", $actual_size / (( $t1 - $t0 ) * 1024); exit}"`
- # compute MB/sec from nregions*slowest
- check=`awk < $rfile \
- '/time to transfer data/ {mb=$8/1.048576; if (n == 0 || mb < min) min = mb; n++}\
- END {printf "%3d x %6.2f = %6.2f MB/s", n, min, min * n}'`
- print_summary "$bw $check"
- fi
- rm $rfile
+ print_summary ""
done
done
done
git://git.whamcloud.com / fs / lustre-release.git / blobdiff