X-Git-Url: https://git.whamcloud.com/?a=blobdiff_plain;f=lustre%2Ftests%2Fha.sh;h=9fc76a47f97dc97b74fff24995794b6cb030d16a;hb=9e8ee5aa0339b06aa53dc00f4911e2019a3ef3a0;hp=0c47ed67e7ef73448aff15a6e35a43ae596e23ce;hpb=1a7c352e97d8b9c41ed8b9c881c813ee87b3789c;p=fs%2Flustre-release.git diff --git a/lustre/tests/ha.sh b/lustre/tests/ha.sh index 0c47ed6..9fc76a4 100755 --- a/lustre/tests/ha.sh +++ b/lustre/tests/ha.sh @@ -82,6 +82,9 @@ # The number of clients run MPI loads is configured by parameter # ha_mpi_instances. Only one client runs MPI workloads by default. # +# MPI workloads can be run from several users. The list of users to use is +# configured by parameter ha_mpi_users, default is "mpiuser". +# # PROCESS STRUCTURE AND IPC # # On the node where this script is run, the processes look like this: @@ -171,6 +174,7 @@ set -eE declare ha_tmp_dir=/tmp/$(basename $0)-$$ declare ha_stop_file=$ha_tmp_dir/stop declare ha_fail_file=$ha_tmp_dir/fail +declare ha_pm_states=$ha_tmp_dir/ha_pm_states declare ha_status_file_prefix=$ha_tmp_dir/status declare -a ha_status_files declare ha_machine_file=$ha_tmp_dir/machine_file @@ -188,22 +192,37 @@ declare ha_lfsck_fail_on_repaired=${LFSCK_FAIL_ON_REPAIRED:-false} declare ha_power_down_cmd=${POWER_DOWN:-"pm -0"} declare ha_power_up_cmd=${POWER_UP:-"pm -1"} declare ha_power_delay=${POWER_DELAY:-60} +declare ha_node_up_delay=${NODE_UP_DELAY:-10} +declare ha_pm_host=${PM_HOST:-$(hostname)} declare ha_failback_delay=${DELAY:-5} declare ha_failback_cmd=${FAILBACK:-""} declare ha_stripe_params=${STRIPEPARAMS:-"-c 0"} +declare ha_test_dir_stripe_count=${TDSTRIPECOUNT:-"1"} +declare ha_test_dir_mdt_index=${TDMDTINDEX:-"0"} +declare ha_test_dir_mdt_index_random=${TDMDTINDEXRAND:-false} declare ha_dir_stripe_count=${DSTRIPECOUNT:-"1"} declare ha_mdt_index=${MDTINDEX:-"0"} declare ha_mdt_index_random=${MDTINDEXRAND:-false} declare -a ha_clients declare -a ha_servers declare -a ha_victims +declare -a ha_victims_pair declare ha_test_dir=/mnt/lustre/$(basename $0)-$$ +declare -a ha_testdirs=(${ha_test_dirs="$ha_test_dir"}) + +for ((i=0; i<${#ha_testdirs[@]}; i++)); do + echo I=$i ${ha_testdirs[i]} + ha_testdirs[i]="${ha_testdirs[i]}/$(basename $0)-$$" + echo i=$i ${ha_testdirs[i]} +done + +declare ha_cleanup=${CLEANUP:-true} declare ha_start_time=$(date +%s) declare ha_expected_duration=$((60 * 60 * 24)) declare ha_max_failover_period=10 declare ha_nr_loops=0 declare ha_stop_signals="SIGINT SIGTERM SIGHUP" -declare ha_load_timeout=$((60 * 10)) +declare ha_load_timeout=${LOAD_TIMEOUT:-$((60 * 10))} declare ha_workloads_only=false declare ha_workloads_dry_run=false declare ha_simultaneous=false @@ -212,19 +231,39 @@ declare ha_mpi_instances=${ha_mpi_instances:-1} declare ha_mpi_loads=${ha_mpi_loads="ior simul mdtest"} declare -a ha_mpi_load_tags=($ha_mpi_loads) +declare -a ha_mpiusers=(${ha_mpi_users="mpiuser"}) +declare -a ha_users +declare -A ha_mpiopts + +for ((i=0; i<${#ha_mpiusers[@]}; i++)); do + u=${ha_mpiusers[i]%%:*} + o="" + # user gets empty option if ha_mpi_users does not specify it explicitly + [[ ${ha_mpiusers[i]} =~ : ]] && o=${ha_mpiusers[i]##*:} + ha_users[i]=$u + ha_mpiopts[$u]+=" $o" +done +ha_users=(${!ha_mpiopts[@]}) declare ha_ior_params=${IORP:-'" -b $ior_blockSize -t 2m -w -W -T 1"'} declare ha_simul_params=${SIMULP:-'" -n 10"'} declare ha_mdtest_params=${MDTESTP:-'" -i 1 -n 1000"'} declare ha_mpirun_options=${MPIRUN_OPTIONS:-""} +declare ha_clients_stripe=${CLIENTSSTRIPE:-'"$STRIPEPARAMS"'} +declare ha_nclientsset=${NCLIENTSSET:-1} +declare ha_ninstmustfail=${NINSTMUSTFAIL:-0} + +declare ha_racer_params=${RACERP:-"MDSCOUNT=1"} eval ha_params_ior=($ha_ior_params) eval ha_params_simul=($ha_simul_params) eval ha_params_mdtest=($ha_mdtest_params) +eval ha_stripe_clients=($ha_clients_stripe) declare ha_nparams_ior=${#ha_params_ior[@]} declare ha_nparams_simul=${#ha_params_simul[@]} declare ha_nparams_mdtest=${#ha_params_mdtest[@]} +declare ha_nstripe_clients=${#ha_stripe_clients[@]} declare -A ha_mpi_load_cmds=( [ior]="$IOR -o {}/f.ior {params}" @@ -232,25 +271,28 @@ declare -A ha_mpi_load_cmds=( [mdtest]="$MDTEST {params} -d {}" ) +declare racer=${RACER:-"$(dirname $0)/racer/racer.sh"} + declare ha_nonmpi_loads=${ha_nonmpi_loads="dd tar iozone"} declare -a ha_nonmpi_load_tags=($ha_nonmpi_loads) -declare -a ha_nonmpi_load_cmds=( - "dd if=/dev/zero of={}/f.dd bs=1M count=256" - "tar cf - /etc | tar xf - -C {}" - "iozone -a -e -+d -s $iozone_SIZE {}/f.iozone" +declare -A ha_nonmpi_load_cmds=( + [dd]="dd if=/dev/zero of={}/f.dd bs=1M count=256" + [tar]="tar cf - /etc | tar xf - -C {}" + [iozone]="iozone -a -e -+d -s $iozone_SIZE {}/f.iozone" + [racer]="$ha_racer_params $racer {}" ) ha_usage() { - ha_info "Usage: $0 -c HOST[,...] -s HOST[,...]" \ - "-v HOST[,...] [-d DIRECTORY] [-u SECONDS]" + ha_info "Usage: $0 -c HOST[,...] -s HOST[,...]" \ + "-v HOST[,...] -f HOST[,...] [-d DIRECTORY] [-u SECONDS]" } ha_process_arguments() { local opt - while getopts hc:s:v:d:p:u:wrm opt; do + while getopts hc:s:v:d:p:u:wrmf: opt; do case $opt in h) ha_usage @@ -283,6 +325,9 @@ ha_process_arguments() m) ha_simultaneous=true ;; + f) + ha_victims_pair=(${OPTARG//,/ }) + ;; \?) ha_usage exit 1 @@ -316,8 +361,8 @@ ha_on() # -S is to be used here to track the # remote command return values # - pdsh -S -w $nodes PATH=/usr/local/sbin:/usr/local/bin:/sbin:\ -/bin:/usr/sbin:/usr/bin "$@" || + pdsh -S -w $nodes "PATH=/usr/local/sbin:/usr/local/bin:/sbin:\ +/bin:/usr/sbin:/usr/bin; $@" || rc=$? return $rc } @@ -327,10 +372,12 @@ ha_trap_exit() touch "$ha_stop_file" trap 0 if [ -e "$ha_fail_file" ]; then - ha_info "Test directory $ha_test_dir not removed" + ha_info "Test directories ${ha_testdirs[@]} not removed" ha_info "Temporary directory $ha_tmp_dir not removed" else - ha_on ${ha_clients[0]} rm -rf "$ha_test_dir" + $ha_cleanup && + ha_on ${ha_clients[0]} rm -rf ${ha_testdirs[@]} || + ha_info "Test directories ${ha_testdirs[@]} not removed" ha_info "Please find the results in the directory $ha_tmp_dir" fi } @@ -404,9 +451,15 @@ ha_repeat_mpi_load() local load=$2 local status=$3 local parameter=$4 + local machines=$5 + local stripeparams=$6 + local mpiuser=$7 + local mustpass=$8 + local mpirunoptions=$9 + local test_dir=${10} local tag=${ha_mpi_load_tags[$load]} local cmd=${ha_mpi_load_cmds[$tag]} - local dir=$ha_test_dir/$client-$tag + local dir=$test_dir/$client-$tag local log=$ha_tmp_dir/$client-$tag local rc=0 local nr_loops=0 @@ -416,10 +469,13 @@ ha_repeat_mpi_load() cmd=${cmd//"{}"/$dir} cmd=${cmd//"{params}"/$parameter} + [[ -n "$ha_postcmd" ]] && ha_postcmd=${ha_postcmd//"{}"/$dir} + [[ -n "$ha_precmd" ]] && ha_precmd=${ha_precmd//"{}"/$dir} ha_info "Starting $tag" - local machines="-machinefile $ha_machine_file" + machines="-machinefile $machines" while [ ! -e "$ha_stop_file" ] && ((rc == 0)); do + ha_info "$client Starts: $mpiuser: $cmd" 2>&1 | tee -a $log { local mdt_index if $ha_mdt_index_random && [ $ha_mdt_index -ne 0 ]; then @@ -427,23 +483,41 @@ ha_repeat_mpi_load() else mdt_index=$ha_mdt_index fi + [[ -n "$ha_precmd" ]] && ha_info "$ha_precmd" && + ha_on $client "$ha_precmd" >>"$log" 2>&1 ha_on $client $LFS mkdir -i$mdt_index -c$ha_dir_stripe_count "$dir" && ha_on $client $LFS getdirstripe "$dir" && + ha_on $client $LFS setstripe $stripeparams $dir && + ha_on $client $LFS getstripe $dir && ha_on $client chmod a+xwr $dir && - ha_on $client "su mpiuser sh -c \" $mpirun $ha_mpirun_options \ - -np $((${#ha_clients[@]} * mpi_threads_per_client )) \ - $machines $cmd \" " && + ha_on $client "su $mpiuser sh -c \" $mpirun $mpirunoptions \ + -np $((${#ha_clients[@]} * mpi_threads_per_client / ha_nclientsset)) \ + $machines $cmd \" " || rc=$? + [[ -n "$ha_postcmd" ]] && ha_info "$ha_postcmd" && + ha_on $client "$ha_postcmd" >>"$log" 2>&1 + (( ((rc == 0)) && (( mustpass != 0 )) )) || + (( ((rc != 0)) && (( mustpass == 0 )) )) && ha_on $client rm -rf "$dir"; } >>"$log" 2>&1 || rc=$? - ha_info rc=$rc - - if ((rc != 0)); then + ha_info $client: rc=$rc mustpass=$mustpass + + # mustpass=0 means that failure is expected + if (( rc !=0 )); then + if (( mustpass != 0 )); then + touch "$ha_fail_file" + touch "$ha_stop_file" + ha_dump_logs "${ha_clients[*]} ${ha_servers[*]}" + else + # Ok to fail + rc=0 + fi + elif (( mustpass == 0 )); then touch "$ha_fail_file" touch "$ha_stop_file" ha_dump_logs "${ha_clients[*]} ${ha_servers[*]}" fi - echo $rc >"$status" + echo rc=$rc mustpass=$mustpass >"$status" nr_loops=$((nr_loops + 1)) done @@ -451,7 +525,8 @@ ha_repeat_mpi_load() [ $nr_loops -ne 0 ] && avg_loop_time=$((($(date +%s) - start_time) / nr_loops)) - ha_info "$tag stopped: rc $rc avg loop time $avg_loop_time" + ha_info "$tag stopped: rc=$rc mustpass=$mustpass \ + avg loop time $avg_loop_time" } ha_start_mpi_loads() @@ -462,24 +537,49 @@ ha_start_mpi_loads() local status local n local nparam + local machines + local m + local -a mach + local mpiuser + local nmpi - for client in ${ha_clients[@]}; do - ha_info ha_machine_file=$ha_machine_file - echo $client >> $ha_machine_file + # ha_mpi_instances defines the number of + # clients start mpi loads; should be <= ${#ha_clients[@]} + # do nothing if + # ha_mpi_instances = 0 + # or + # ${#ha_mpi_load_tags[@]} =0 + local inst=$ha_mpi_instances + (( inst == 0 )) || (( ${#ha_mpi_load_tags[@]} == 0 )) && + ha_info "no mpi load to start" && + return 0 + + (( inst <= ${#ha_clients[@]} )) || inst=${#ha_clients[@]} + + # Define names for machinefiles for each client set + for (( n=0; n < $ha_nclientsset; n++ )); do + mach[$n]=$ha_machine_file$n + done + + for ((n = 0; n < ${#ha_clients[@]}; n++)); do + m=$(( n % ha_nclientsset)) + machines=${mach[m]} + ha_info machine_file=$machines + echo ${ha_clients[n]} >> $machines done local dirname=$(dirname $ha_machine_file) for client in ${ha_clients[@]}; do ha_on $client mkdir -p $dirname - scp $ha_machine_file $client:$ha_machine_file + scp $ha_machine_file* $client:$dirname done - # ha_mpi_instances defines the number of - # clients start mpi loads; should be <= ${#ha_clients[@]} - local inst=$ha_mpi_instances - (( inst <= ${#ha_clients[@]} )) || inst=${#ha_clients[@]} - + local ndir for ((n = 0; n < $inst; n++)); do client=${ha_clients[n]} + nmpi=$((n % ${#ha_users[@]})) + mpiuser=${ha_users[nmpi]} + ndir=$((n % ${#ha_testdirs[@]})) + test_dir=${ha_testdirs[ndir]} for ((load = 0; load < ${#ha_mpi_load_tags[@]}; load++)); do tag=${ha_mpi_load_tags[$load]} status=$ha_status_file_prefix-$tag-$client @@ -489,7 +589,17 @@ ha_start_mpi_loads() nparam=$((n % num)) local aref=ha_params_$tag[nparam] local parameter=${!aref} - ha_repeat_mpi_load $client $load $status "$parameter" & + local nstripe=$((n % ha_nstripe_clients)) + aref=ha_stripe_clients[nstripe] + local stripe=${!aref} + local m=$(( n % ha_nclientsset)) + machines=${mach[m]} + local mustpass=1 + [[ $ha_ninstmustfail == 0 ]] || + mustpass=$(( n % ha_ninstmustfail )) + ha_repeat_mpi_load $client $load $status "$parameter" \ + $machines "$stripe" "$mpiuser" "$mustpass" \ + "${ha_mpiopts[$mpiuser]} $ha_mpirun_options" "$test_dir" & ha_status_files+=("$status") done done @@ -501,19 +611,21 @@ ha_repeat_nonmpi_load() local load=$2 local status=$3 local tag=${ha_nonmpi_load_tags[$load]} - local cmd=${ha_nonmpi_load_cmds[$load]} - local dir=$ha_test_dir/$client-$tag + local cmd=${ha_nonmpi_load_cmds[$tag]} + local test_dir=$4 + local dir=$test_dir/$client-$tag local log=$ha_tmp_dir/$client-$tag local rc=0 local nr_loops=0 local avg_loop_time=0 local start_time=$(date +%s) - cmd=${cmd//"{}"/$dir} + cmd=${cmd//"{}"/$dir} - ha_info "Starting $tag on $client" + ha_info "Starting $tag on $client on $dir" while [ ! -e "$ha_stop_file" ] && ((rc == 0)); do + ha_info "$client Starts: $cmd" 2>&1 | tee -a $log ha_on $client "mkdir -p $dir && \ $cmd && \ rm -rf $dir" >>"$log" 2>&1 || rc=$? @@ -536,19 +648,47 @@ ha_repeat_nonmpi_load() ha_start_nonmpi_loads() { - local client - local load - local tag - local status - - for client in ${ha_clients[@]}; do - for ((load = 0; load < ${#ha_nonmpi_load_tags[@]}; load++)); do - tag=${ha_nonmpi_load_tags[$load]} - status=$ha_status_file_prefix-$tag-$client - ha_repeat_nonmpi_load $client $load $status & - ha_status_files+=("$status") - done - done + local client + local load + local tag + local status + local n + local test_dir + local ndir + + for (( n = 0; n < ${#ha_clients[@]}; n++)); do + client=${ha_clients[n]} + ndir=$((n % ${#ha_testdirs[@]})) + test_dir=${ha_testdirs[ndir]} + for ((load = 0; load < ${#ha_nonmpi_load_tags[@]}; load++)); do + tag=${ha_nonmpi_load_tags[$load]} + status=$ha_status_file_prefix-$tag-$client + ha_repeat_nonmpi_load $client $load $status $test_dir & + ha_status_files+=("$status") + done + done +} + +declare ha_bgcmd=${ha_bgcmd:-""} +declare ha_bgcmd_log=$ha_tmp_dir/bgcmdlog + +ha_cmd_bg () { + [[ -z "$ha_bgcmd" ]] && return 0 + for ((i=0; i<${#ha_testdirs[@]}; i++)); do + ha_bgcmd=${ha_bgcmd//"{}"/${ha_testdirs[i]}} + done + + ha_info "BG cmd: $ha_bgcmd" + while [ true ]; do + [ -f $ha_stop_file ] && + ha_info "$ha_stop_file found! $ha_bgcmd no started" && + break + eval $ha_bgcmd 2>&1 | tee -a $ha_bgcmd_log + sleep 1 + done & + CMD_BG_PID=$! + ha_info CMD BG PID: $CMD_BG_PID + ps aux | grep $CMD_BG_PID } ha_lfsck_bg () { @@ -664,17 +804,22 @@ ha_lfsck_repaired() ha_start_loads() { - trap ha_trap_stop_signals $ha_stop_signals - ha_start_nonmpi_loads - ha_start_mpi_loads + ha_cmd_bg + $ha_lfsck_bg && ha_lfsck_bg + trap ha_trap_stop_signals $ha_stop_signals + ha_start_nonmpi_loads + ha_start_mpi_loads } ha_stop_loads() { - touch $ha_stop_file - trap - $ha_stop_signals - ha_info "Waiting for workloads to stop" - wait + touch $ha_stop_file + [[ -n $CMD_BG_PID ]] && wait $CMD_BG_PID || true + # true because of lfsck_bg could be stopped already + $ha_lfsck_bg && wait $LFSCK_BG_PID || true + trap - $ha_stop_signals + ha_info "Waiting for workloads to stop" + wait } ha_wait_loads() @@ -682,7 +827,7 @@ ha_wait_loads() local file local end=$(($(date +%s) + ha_load_timeout)) - ha_info "Waiting for workload status" + ha_info "Waiting $ha_load_timeout sec for workload status..." rm -f "${ha_status_files[@]}" # @@ -716,24 +861,139 @@ ha_wait_loads() done } +ha_powermanage() +{ + local nodes=$1 + local expected_state=$2 + local state + local -a states + local i + local rc=0 + + # store pm -x -q $nodes results in a file to have + # more information about nodes statuses + ha_on $ha_pm_host pm -x -q $nodes | awk '{print $2 $3}' > $ha_pm_states + rc=${PIPESTATUS[0]} + echo pmrc=$rc + + while IFS=": " read node state; do + [[ "$state" = "$expected_state" ]] && { + nodes=${nodes/$node/} + nodes=${nodes//,,/,} + nodes=${nodes/#,} + nodes=${nodes/%,} + } + done < $ha_pm_states + + if [ -n "$nodes" ]; then + cat $ha_pm_states + return 1 + fi + return 0 +} + +ha_power_down_cmd_fn() +{ + local nodes=$1 + local cmd + + case $ha_power_down_cmd in + # format is: POWER_DOWN=sysrqcrash + sysrqcrash) cmd="pdsh -S -w $nodes 'echo c > /proc/sysrq-trigger' &" ;; + *) cmd="$ha_power_down_cmd $nodes" ;; + esac + + eval $cmd +} + ha_power_down() { local nodes=$1 local rc=1 local i + local state + + case $ha_power_down_cmd in + *pm*) state=off ;; + sysrqcrash) state=off ;; + *) state=on;; + esac if $ha_lfsck_bg && [[ ${nodes//,/ /} =~ $ha_lfsck_node ]]; then ha_info "$ha_lfsck_node down, delay start LFSCK" ha_lock $ha_lfsck_lock fi - ha_info "Powering down $nodes" - for i in $(seq 1 5); do - $ha_power_down_cmd $nodes && rc=0 && break + ha_info "Powering down $nodes : cmd: $ha_power_down_cmd" + for (( i=0; i<10; i++ )) { + ha_info "attempt: $i" + ha_power_down_cmd_fn $nodes && + ha_powermanage $nodes $state && rc=0 && break sleep $ha_power_delay + } + + [ $rc -eq 0 ] || { + ha_info "Failed Powering down in $i attempts:" \ + "$ha_power_down_cmd" + cat $ha_pm_states + exit 1 + } +} + +ha_get_pair() +{ + local node=$1 + local i + + for ((i=0; i<${#ha_victims[@]}; i++)) { + [[ ${ha_victims[i]} == $node ]] && echo ${ha_victims_pair[i]} && + return + } + [[ $i -ne ${#ha_victims[@]} ]] || + ha_error "No pair found!" +} + +ha_power_up_delay() +{ + local nodes=$1 + local end=$(($(date +%s) + ha_node_up_delay)) + local rc + + if [[ ${#ha_victims_pair[@]} -eq 0 ]]; then + ha_sleep $ha_node_up_delay + return 0 + fi + + # Check CRM status on failover pair + while (($(date +%s) <= end)); do + rc=0 + for n in ${nodes//,/ }; do + local pair=$(ha_get_pair $n) + local status=$(ha_on $pair crm_mon -1rQ | \ + grep -w $n | head -1) + + ha_info "$n pair: $pair status: $status" + [[ "$status" == *OFFLINE* ]] || + rc=$((rc + $?)) + ha_info "rc: $rc" + done + + if [[ $rc -eq 0 ]]; then + ha_info "CRM: Got all victims status OFFLINE" + return 0 + fi + sleep 60 done - [ $rc -eq 0 ] || ha_info "Failed Powering down in $i attempts" + ha_info "$nodes CRM status not OFFLINE" + for n in ${nodes//,/ }; do + local pair=$(ha_get_pair $n) + + ha_info "CRM --- $n" + ha_on $pair crm_mon -1rQ + done + ha_error "CRM: some of $nodes are not OFFLINE in $ha_node_up_delay sec" + exit 1 } ha_power_up() @@ -742,13 +1002,20 @@ ha_power_up() local rc=1 local i - ha_info "Powering up $nodes" - for i in $(seq 1 5); do - $ha_power_up_cmd $nodes && rc=0 && break + ha_power_up_delay $nodes + ha_info "Powering up $nodes : cmd: $ha_power_up_cmd" + for (( i=0; i<10; i++ )) { + ha_info "attempt: $i" + $ha_power_up_cmd $nodes && + ha_powermanage $nodes on && rc=0 && break sleep $ha_power_delay - done + } - [ $rc -eq 0 ] || ha_info "Failed Powering up in $i attempts" + [ $rc -eq 0 ] || { + ha_info "Failed Powering up in $i attempts: $ha_power_up_cmd" + cat $ha_pm_states + exit 1 + } } # @@ -874,11 +1141,25 @@ ha_main() "START: $0: $(date +%H:%M:%S' '%s)" trap ha_trap_exit EXIT mkdir "$ha_tmp_dir" - ha_on ${ha_clients[0]} mkdir "$ha_test_dir" - ha_on ${ha_clients[0]} " \ - $LFS setstripe $ha_stripe_params $ha_test_dir" - $ha_lfsck_bg && ha_lfsck_bg + local mdt_index + if $ha_test_dir_mdt_index_random && + [ $ha_test_dir_mdt_index -ne 0 ]; then + mdt_index=$(ha_rand $ha_test_dir_mdt_index) + else + mdt_index=$ha_test_dir_mdt_index + fi + + local dir + test_dir=${ha_testdirs[0]} + ha_on ${ha_clients[0]} "$LFS mkdir -i$mdt_index \ + -c$ha_test_dir_stripe_count $test_dir" + for ((i=0; i<${#ha_testdirs[@]}; i++)); do + test_dir=${ha_testdirs[i]} + ha_on ${ha_clients[0]} $LFS getdirstripe $test_dir + ha_on ${ha_clients[0]} " \ + $LFS setstripe $ha_stripe_params $test_dir" + done ha_start_loads ha_wait_loads @@ -896,9 +1177,6 @@ ha_main() $ha_lfsck_fail_on_repaired && ha_lfsck_repaired - # true because of lfsck_bg could be stopped already - $ha_lfsck_bg && wait $LFSCK_BG_PID || true - if [ -e "$ha_fail_file" ]; then exit 1 else