X-Git-Url: https://git.whamcloud.com/?a=blobdiff_plain;f=lustre-iokit%2Fobdfilter-survey%2Fobdfilter-survey;h=19cdaf040be209272314d7d1eb64059c22f8d500;hb=8f13bc198125c03dcf79625b310a0f3ac76b00b2;hp=f700a8810f6f39b56270dd1b9be43107bf65350e;hpb=d03c475e63cac8beeb20613d2e42881ae417eeb2;p=fs%2Flustre-release.git diff --git a/lustre-iokit/obdfilter-survey/obdfilter-survey b/lustre-iokit/obdfilter-survey/obdfilter-survey index f700a88..19cdaf0 100755 --- a/lustre-iokit/obdfilter-survey/obdfilter-survey +++ b/lustre-iokit/obdfilter-survey/obdfilter-survey @@ -1,170 +1,146 @@ #!/bin/bash - +set -e ###################################################################### # customize per survey -# specify either the obdecho client names or the obdfilter names -client_names=() -ost_names=(ost{1,2,3,4,5,6,7,8}) - -# result file prefix -rslt=/tmp/obdfilter_survey - -# lustre root (leave blank unless running with own source) -lustre_root= - -# what to do (we always do an initial write) -#tests="rewrite read reread rewrite_again" -tests="rewrite read" - -# total size (MBytes) -# large enough to avoid cache effects -size=8192 - -# record size (KBytes) -rszlo=1024 -rszhi=1024 +# specify obd instances to exercise +# these can be either... +# obdfilter instances (set 'ost_names') +# ...or... +# echo_client instances (set 'client_names') +# ... use 'host:name' for obd instances on other nodes. +# allow these to be passed in via string... +# OR +# one can specify only case=disk or case=network or case=netdisk through +# command line. + +# Perquisite: For "disk" case and "netdisk" case you need to have lustre setup +# with one or more ost's. For "network" case you need to have all +# modules (those llmount.sh loades) loaded in kernel. And the +# 'lctl dl' output must be blank. + +# How to run test: +# case 1 (local disk): +# $ nobjhi=2 thrhi=2 size=1024 case=disk sh obdfilter-survey +# one can also run test with user defined targets as follows, +# $ nobjhi=2 thrhi=2 size=1024 targets="lustre-OST0000 lustre-OST0001 ..." sh obdfilter-survey +# case 2 (network): +# $ nobjhi=2 thrhi=2 size=1024 targets="" case=network sh obdfilter-survey +# where, targets is name or ip address of system, which you want to +# set as server. +# case 3 (network and disk): +# $ nobjhi=2 thrhi=2 size=1024 case=netdisk sh obdfilter-survey +# one can also run test with user defined targets as follows, +# $ nobjhi=2 thrhi=2 size=1024 targets=" ..." sh obdfilter-survey +#[ NOTE: It is advised to have automated login (passwordless entry) between server and +# client systems on which this test runs.] + +# include library +source $(dirname $0)/libecho + +# The following variables can be set in the environment, or on the +# command line +# result file prefix (date/time + hostname makes unique) +# NB ensure path to it exists +rslt_loc=${rslt_loc:-"/tmp"} +rslt=${rslt:-"$rslt_loc/obdfilter_survey_`date +%F@%R`_`uname -n`"} + +# Set this true to check file contents +verify=${verify:-0} + +# test targets +targets=${targets:-""} +# test case +case=${case:-"disk"} + +# total size (MBytes) per obd instance +# large enough to avoid cache effects +# and to make test startup/shutdown overhead insignificant +size=${size:-16384} + +# record size (KBytes) ( 7168 max) +rszlo=${rszlo:-1024} +rszhi=${rszhi:-1024} # number of objects per OST -nobjlo=1 -nobjhi=32 +nobjlo=${nobjlo:-1} +#was nobjhi=${nobjhi:-512} +nobjhi=${nobjhi:-16} # threads per OST (1024 max) -thrlo=1 -thrhi=128 - -# restart from here iff all are defined -restart_rsz= -restart_thr=1 -restart_nobj=1 - -# machine's page size -PAGE_SIZE=64 - -# max buffer_mem (total_threads * buffer size) -# (to avoid lctl ENOMEM problems) -max_buffer_mem=$((256*1024)) - -##################################################################### +thrlo=${thrlo:-1} +thrhi=${thrhi:-16} -snap=1 -verify=1 - -check_obdecho() { - lsmod | grep obdecho > /dev/null 2>&1 -} +export LC_ALL=POSIX -check_obdecho -load_obdecho=$(($? != 0)) - -if [ -z "$lustre_root" ]; then - lctl=lctl - if ((load_obdecho)); then - modprobe obdecho - fi -else - lctl=${lustre_root}/lctl - if ((load_obdecho)); then - if [ -f ${lustre_root}/obdecho/obdecho.ko ]; then - insmod ${lustre_root}/obdecho/obdecho.ko - else - insmod ${lustre_root}/obdecho/obdecho.o - fi - fi -fi - -check_obdecho || (echo "Can't load obdecho"; exit 1) - -get_devno () { - local type=$1 - local name=$2 - $lctl device_list | awk "{if (\$2 == \"UP\" && \$3 == \"$type\" && \$4 == \"$name\") {\ - print \$1; exit}}" -} - -get_ec_devno () { - local idx=$1 - local client_name=${client_names[idx]} - local ost_name=${ost_names[idx]} - if [ -z "$client_name" ]; then - if [ -z "$ost_name" ]; then - echo "client and ost name both null" 1>&2 - return - fi - client_name=${ost_name}_echo_client - fi - ec=`get_devno echo_client $client_name` - if [ -n "$ec" ]; then - echo $ec $client_name - return - fi - if [ -z "$ost_name" ]; then - echo "no echo client and ost_name not set" 1>&2 - return - fi - ost=`get_devno obdfilter $ost_name` - if [ -z "$ost" ]; then - echo "OST $ost_name not setup" 1>&2 - return - fi - $lctl <&2 - return - fi - echo $ec $client_name 1 -} - -teardown_ec_devno () { - local idx=$1 - local client_name=${client_names[$idx]} - if ((do_teardown_ec[$idx])); then - $lctl < $rfile 2>&1 - n=(`awk < $rfile \ - '/is object id/ {obj=strtonum($6);\ - first=!not_first; not_first=1;\ - if (first) first_obj=obj; - else if (obj != prev + 1) exit;\ - prev=obj; n++}\ - END {printf "%d %d\n", first_obj, n}'`) - if ((n[1] != nobj)); then - echo "ERROR" - else - echo ${n[0]} + local host=$1 + local devno=$2 + local nobj=$3 + local rfile=$4 + remote_shell $host $lctl --device $devno create $nobj > $rfile 2>&1 + first=0 + prev=0 + count=0 + error=0 + + # Count number of objects (lines containing " is object id "), and + # ensure that the objects numbers are sequential. + # + exec 3< $rfile + while read -u3 line; do + case "$line" in + ( *' is object id '* ) + set -- $line + if test $(( count += 1 )) -gt 1 ; then + (( $6 != prev + 1 )) && error=1 + else + first=$(( $6 + 0 )) + fi + prev=$6 + ;; + esac + done + exec 3<&- + + if [ $nobj -ne $count ]; then + echo "ERROR: $nobj != $count" >&2 + cat $rfile >&2 + echo "ERROR" + elif [ $error -ne 0 ]; then + echo "ERROR: non contiguous objs found" >&2 + echo ERROR + else + echo $first fi + return $error } +# destroys all objects created in create_objects routine +# parameter: 3. start obj id. destroy_objects () { - local devno=$1 - local obj0=$2 - local nobj=$3 - local rfile=$4 - $lctl --device $devno destroy $obj0 $nobj > $rfile 2>&1 + local host=$1 + local devno=$2 + local obj0=$3 + local nobj=$4 + local rfile=$5 + remote_shell $host $lctl --device $devno destroy $obj0 $nobj > $rfile 2>&1 } get_stats () { local rfile=$1 - awk < $rfile \ + gawk < $rfile \ '/^Selected device [0-9]+$/ {n = 0; next}\ /error/ {n = -1; exit}\ + /^Total/ {next}\ /^[0-9]+\/[0-9]+ Total: [0-9]+\.[0-9]+\/second$/ {n++; v=strtonum($3); \ if (n == 1 || v < min) min = v;\ if (n == 1 || v > max) max = v;\ @@ -184,6 +160,8 @@ get_global_stats () { printf "%d %f %f\n", err, min, max}' } +# enable or disable data check. +# parameter: 1. read/write testname2type () { # 'x' disables data check if ((verify)); then @@ -197,43 +175,244 @@ testname2type () { esac } -start=`date +%F@%R` -rsltf="${rslt}_${start}.summary" -echo -n > $rsltf -workf="${rslt}_${start}.detail" -echo -n > $workf +# for "echo_client + obdfilter" case, "prep + commit" mode should be used +# for "echo_client + osc" case, "BRW" mode should be used +testcase2mode() { + case $case in + disk) echo "p$1";; + *) echo "g";; + esac +} print_summary () { if [ "$1" = "-n" ]; then - minusn=$1; shift + minusn=$1; shift else - minusn="" + minusn="" fi echo $minusn "$*" >> $rsltf echo $minusn "$*" } -ndevs=${#client_names[@]} -if ((ndevs < ${#ost_names[@]} )); then - ndevs=${#ost_names[@]} +# Customisation variables +##################################################################### +# One can change variable values in this section as per requirements + +if [ -n "$targets" ]; then + declare -a ost_names + declare -a client_names + count=0 + for name in $targets; do + if [ $case == "disk" ]; then + ost_names[$count]=$name + else + client_names[$count]=$name + fi + count=$((count+1)) + done +fi + +# what tests to run (first must be write) +tests_str=${tests_str:-""} +if [ -n "$tests_str" ]; then + declare -a tests + count=0 + for name in $tests_str; do + tests[$count]=$name + count=$((count+1)) + done +else + #tests=(write rewrite read reread rewrite_again) + tests=(write rewrite read) fi -for ((idx = 0; idx < ndevs; idx++)); do - devno=(`get_ec_devno $idx`) - if ((${#devno[@]} < 2)); then - exit 1 +# restart from here iff all are defined +restart_rsz= +restart_thr=1 +restart_nobj=1 + +# machine's page size (K) +if [ -z "$PAGE_SIZE" ]; then + if which python >/dev/null; then + PAGE_SIZE=`echo 'import resource; print resource.getpagesize()/1024;' |python` fi - devnos[$idx]=${devno[0]} - client_names[$idx]=${devno[1]} - do_teardown_ec[$idx]=$((${#devno[@]} > 2)) -done +fi +PAGE_SIZE=${PAGE_SIZE:-4} + +# max buffer_mem (total_threads * buffer size) +# (to avoid lctl ENOMEM problems) +max_buffer_mem=$((1024 * 1024)) +snap=1 +clean_srv_OSS=0 +# Customisation variables ends here. +##################################################################### +# leave the rest of this alone unless you know what you're doing... -echo 0 > /proc/sys/portals/debug +# check and insert obdecho module +if ! lsmod | grep obdecho > /dev/null; then + modprobe obdecho +fi +if [ ${#tests[@]} -eq 0 -o "${tests[0]}" != "write" ]; then + echo "tests: ${tests[@]}" + echo "First test must be 'write'" 1>&2 + exit 1 +fi -for ((rsz=$rszlo;rsz<=$rszhi;rsz*=2)); do - for ((nobj=$nobjlo;nobj<=$nobjhi;nobj*=2)); do - for ((thr=$thrlo;thr<=$thrhi;thr*=2)); do - if ((thr < nobj)); then +rsltf="${rslt}.summary" +workf="${rslt}.detail" +cmdsf="${rslt}.script" +vmstatf="${rslt}.vmstat" +echo -n > $rsltf +echo -n > $workf + +# hide a little trick to unset this from the command line +if [ "$lustre_root" == " " ]; then + unset lustre_root +fi + +if [ -z "$lustre_root" ]; then + lctl=lctl +else + lctl=${lustre_root}/utils/lctl +fi + +# split out hostnames from client/ost names +ndevs=0 +for trgt in $targets; do + str=(`split_hostname $trgt`) + host_names[$ndevs]=${str[0]} + client_names[$ndevs]=${str[1]} + ndevs=$((ndevs+1)) +done +if [ $case == "disk" ]; then + if [ $rszhi -gt 1024 ]; then + echo "Test disk case support maximum 1024KB IO data" \ + "(rszhi=$rszhi is too big) please use a smaller value." + exit 1 + fi + for ((i = 0; i < $ndevs; i++)); do + ost_names[$i]=${client_names[$i]} + done +fi +if [ $case == "netdisk" ]; then + if [ "$targets" ]; then + for ((i = 0; i < $ndevs; i++)); do + setup_osc_for_remote_ost ${host_names[$i]} \ + ${client_names[$i]} $i + osc_name=${client_names[$i]}_osc + ec_using_osc $osc_name + cleanup_oscs="$cleanup_oscs $osc_name" + done + else + client_names_str=$($lctl dl | grep -v mdt | \ + awk '{if ($2 == "UP" && $3 == "osc") {print $4} }') + count=0; + for name in $client_names_str; do + client_names[$count]=`echo $name | sed 's/-osc-.*$//'` + count=$((count+1)) + done + + host_names_str=$($lctl dl -t | grep -v mdt | \ + awk '{if ($2 == "UP" && $3 == "osc") {print $7} }') + count=0; + for name in $host_names_str; do + host_names[$count]=`echo $name | sed 's/@.*$//'` + count=$((count+1)) + done + + for (( i = 0; i < $count; i++ )) do + setup_osc_for_remote_ost ${host_names[$i]} \ + ${client_names[$i]} $i + osc_name=${client_names[$i]}_osc + ec_using_osc $osc_name + cleanup_oscs="$cleanup_oscs $osc_name" + done + fi + + echo_clients=$($lctl dl | grep echo_client | awk "{if (\$2 == \"UP\" && \$3 == \"echo_client\") {print \$4} }") + cnt=0; + for name in $echo_clients; do + client_names[$cnt]=$name + host_names[$cnt]=localhost + cnt=$((cnt+1)) + done + ndevs=${#client_names[@]} +fi +if [ $case == "network" ]; then + server_nid=$targets + if [ -z "$server_nid" ]; then + echo "Specify hostname or ip-address of server" + exit 1; + fi + # check for obdecho module on server + if ! dsh $server_nid root "lsmod | grep obdecho > /dev/null"; then + dsh $server_nid root "modprobe obdecho" + fi + # Now do the server setup + setup_srv_obd $server_nid "echo_srv" + oss_on_srv=`dsh $server_nid root "$lctl dl | grep OSS" | awk '{ print $4 }'` + if [ -z $oss_on_srv ]; then + setup_OSS $server_nid + clean_srv_OSS=1 + fi + if ! dsh $server_nid root "$lctl dl | grep obdecho > /dev/null 2>&1"; then + echo "obdecho not setup on server" + exit 1 + fi + if ! dsh $server_nid root "$lctl dl | grep ost > /dev/null 2>&1"; then + echo "ost not setup on server" + exit 1 + fi + # Now start client setup + osc_names_str=$($lctl dl| grep osc | grep -v mdt | grep UP) + if [ -n "$osc_names_str" ]; then + echo "The existing setup must be cleaned"; + exit 0; + fi + ec_using_srv_nid $server_nid "echotmp" "echotmp_UUID" + client_names[0]="echotmp_ecc" +fi +if [ -z "$targets" ]; then + if [ $case == "disk" ]; then + get_targets + ndevs=${#ost_names[@]} + fi +fi +# get vmstat started +# disable portals debug and get obdecho loaded on all relevant hosts +unique_hosts=(`unique ${host_names[@]}`) +load_obdechos +pidcount=0 +for host in ${unique_hosts[@]}; do + host_vmstatf=${vmstatf}_${host} + echo -n > $host_vmstatf + remote_shell $host "vmstat 5 >> $host_vmstatf" &> /dev/null & + pid=$! + vmstatpids[$pidcount]=$pid + pidcount=$((pidcount+1)) +done +# get all the echo_client device numbers and names +for ((i=0; i < $ndevs; i++)); do + host=${host_names[$i]} + devno=(`get_ec_devno $host "${client_names[$i]}" "${ost_names[$i]}"`) + if ((${#devno[@]} != 3)); then + exit 1 + fi + devnos[$i]=${devno[0]} + client_names[$i]=${devno[1]} + do_teardown_ec[$i]=${devno[2]} +done +if (($ndevs <= 0 || ${#host_names[@]} <= 0)); then + echo "no devices or hosts specified" + cleanup 0 $clean_srv_OSS $cleanup_oscs +fi +# Buffers will be spread out among all hosts, so allow for that +max_buffer_mem=$(( ${max_buffer_mem} * ${#unique_hosts[@]} )) +print_summary "$(date) Obdfilter-survey for case=$case from $(hostname)" +for ((rsz = $rszlo; rsz <= $rszhi; rsz*=2)); do + for ((nobj = $nobjlo; nobj <= $nobjhi; nobj*=2)); do + for ((thr = $thrlo; thr <= $thrhi; thr*=2)); do + if ((thr % nobj)); then continue fi # restart? @@ -257,7 +436,7 @@ for ((rsz=$rszlo;rsz<=$rszhi;rsz*=2)); do actual_size=$((actual_rsz*count*thr)) total_size=$((actual_size*ndevs)) # show computed parameters - str=`printf 'ost %2d sz %8dK rsz %4d obj %4d thr %4d ' \ + str=`printf 'ost %2d sz %8dK rsz %4dK obj %4d thr %4d ' \ $ndevs $total_size $actual_rsz $total_nobj $total_thr` echo "=======================> $str" >> $workf print_summary -n "$str" @@ -267,76 +446,117 @@ for ((rsz=$rszlo;rsz<=$rszhi;rsz*=2)); do fi # create the objects tmpf="${workf}_tmp" - for ((idx=0; idx < ndevs; idx++)); do + for ((idx = 0; idx < $ndevs; idx++)); do + host=${host_names[$idx]} devno=${devnos[$idx]} - first_obj=`create_objects $devno $nobj $tmpf` - echo "========> Create [$idx]" >> $workf + client_name="${host}:${client_names[$idx]}" + echo "=============> Create $nobj on $client_name" >> $workf + first_obj=`create_objects $host $devno $nobj $tmpf` cat $tmpf >> $workf rm $tmpf if [ $first_obj = "ERROR" ]; then - print_summary "created object #s [$idx] not contiguous" + print_summary "created object #s on $client_name not contiguous" exit 1 fi first_objs[$idx]=$first_obj done - for test in write $tests; do + # run tests + for test in ${tests[@]}; do + declare -a pidarray + for host in ${unique_hosts[@]}; do + echo "starting run for test: $test rsz: $rsz threads: $thr objects: $nobj" >> ${vmstatf}_${host} + done print_summary -n "$test " - t0=`date +%s.%N` - for ((idx=0; idx < ndevs; idx++)); do + # create per-host script files + for host in ${unique_hosts[@]}; do + echo -n > ${cmdsf}_${host} + done + for ((idx = 0; idx < $ndevs; idx++)); do + host=${host_names[$idx]} devno=${devnos[$idx]} tmpfi="${tmpf}_$idx" first_obj=${first_objs[$idx]} - $lctl > $tmpfi 2>&1 \ - --threads $thr -$snap $devno \ - test_brw $count `testname2type $test` q $pages ${thr}t${first_obj} & - done - wait - t1=`date +%s.%N` + thr_per_obj=$((${thr}/${nobj})) + echo >> ${cmdsf}_${host} \ + "$lctl > $tmpfi 2>&1 \\ + --threads $thr -$snap $devno \\ + test_brw $count `testname2type $test` q $pages \\ + ${thr_per_obj}t${first_obj} `testcase2mode $pages` &" + done + pidcount=0 + for host in ${unique_hosts[@]}; do + echo "wait" >> ${cmdsf}_${host} + pidarray[$pidcount]=0 + pidcount=$((pidcount+1)) + done + # timed run of all the per-host script files + t0=`date +%s.%N` + pidcount=0 + for host in ${unique_hosts[@]}; do + remote_shell $host bash < ${cmdsf}_${host} & + pidarray[$pidcount]=$! + pidcount=$((pidcount+1)) + done + pidcount=0 + for host in ${unique_hosts[@]}; do + wait ${pidarray[$pidcount]} + pidcount=$((pidcount+1)) + done + #wait + t1=`date +%s.%N` + # clean up per-host script files + for host in ${unique_hosts[@]}; do + rm ${cmdsf}_${host} + done + + # compute bandwidth from total data / elapsed time str=`awk "BEGIN {printf \"%7.2f \",\ $total_size / (( $t1 - $t0 ) * 1024)}"` print_summary -n "$str" + # collect/check individual OST stats echo -n > $tmpf - for ((idx=0; idx < ndevs; idx++)); do + for ((idx = 0; idx < $ndevs; idx++)); do + client_name="${host_names[$idx]}:${client_names[$idx]}" tmpfi="${tmpf}_$idx" - echo "========> $test [$idx]" >> $workf - cat $tmpfi >> $workf - get_stats $tmpfi >> $tmpf - rm $tmpfi + echo "=============> $test $client_name" >> $workf + host="${host_names[$idx]}" + remote_shell $host cat $tmpfi > ${tmpfi}_local + cat ${tmpfi}_local >> $workf + get_stats ${tmpfi}_local >> $tmpf + rm -f $tmpfi ${tmpfi}_local done - echo "========> $test [$idx] global" >> $workf + # compute/display global min/max stats + echo "=============> $test global" >> $workf cat $tmpf >> $workf stats=(`get_global_stats $tmpf`) rm $tmpf if ((stats[0] <= 0)); then if ((stats[0] < 0)); then - str=`printf "%15s " ERROR` + str=`printf "%17s " ERROR` else - str=`printf "%15s " SHORT` + str=`printf "%17s " SHORT` fi else - str=`awk "BEGIN {printf \"[%6.2f,%6.2f] \",\ + str=`awk "BEGIN {printf \"[%7.2f,%7.2f] \",\ (${stats[1]} * $actual_rsz)/1024,\ (${stats[2]} * $actual_rsz)/1024; exit}"` fi print_summary -n "$str" done print_summary "" - for ((idx=0; idx < ndevs; idx++)); do + # destroy objects we created + for ((idx = 0; idx < $ndevs; idx++)); do + host=${host_names[$idx]} devno=${devnos[$idx]} + client_name="${host}:${client_names[$idx]}" first_obj=${first_objs[$idx]} - destroy_objects $devno $first_obj $nobj $tmpf - echo "========> Destroy [$idx]" >> $workf + echo "=============> Destroy $nobj on $client_name" >> $workf + destroy_objects $host $devno $first_obj $nobj $tmpf cat $tmpf >> $workf rm $tmpf done done done done - -for ((idx=0; idx < ndevs; idx++)); do - teardown_ec_devno $idx -done - -if ((load_obdecho)); then - rmmod obdecho -fi +cleanup 0 $clean_srv_OSS $cleanup_oscs +exit 0