X-Git-Url: https://git.whamcloud.com/?a=blobdiff_plain;f=lustre%2Ftests%2Fsanity.sh;h=414d95e8238e3efc4d020c88c2cc76ae1498f895;hb=3eff6bada2bc789fe6089e792e4043ef7ec8396f;hp=6f91998c8f0dd7390d9d778657335b49deeaf9d5;hpb=e468c285e0d801e1f2ef6772e43df1465666c649;p=fs%2Flustre-release.git diff --git a/lustre/tests/sanity.sh b/lustre/tests/sanity.sh index 6f91998..414d95e8 100755 --- a/lustre/tests/sanity.sh +++ b/lustre/tests/sanity.sh @@ -1780,17 +1780,24 @@ run_test 27m "create file while OST0 was full" # OSCs keep a NOSPC flag that will be reset after ~5s (qos_maxage) # if the OST isn't full anymore. reset_enospc() { - local OSTIDX=${1:-""} + local ostidx=${1:-""} + local delay + local ready + local get_prealloc local list=$(comma_list $(osts_nodes)) - [ "$OSTIDX" ] && list=$(facet_host ost$((OSTIDX + 1))) + [ "$ostidx" ] && list=$(facet_host ost$((ostidx + 1))) do_nodes $list lctl set_param fail_loc=0 - sync # initiate all OST_DESTROYs from MDS to OST - sleep_maxage + wait_delete_completed # initiate all OST_DESTROYs from MDS to OST + delay=$(do_facet $SINGLEMDS lctl get_param -n lov.*.qos_maxage | + awk '{print $1 * 2;exit;}') + get_prealloc="$LCTL get_param -n osc.*MDT*.prealloc_status | + grep -v \"^0$\"" + wait_update_facet $SINGLEMDS "$get_prealloc" "" $delay } -exhaust_precreations() { +__exhaust_precreations() { local OSTIDX=$1 local FAILLOC=$2 local FAILIDX=${3:-$OSTIDX} @@ -1821,14 +1828,19 @@ exhaust_precreations() { createmany -o $DIR/$tdir/${OST}/f $next_id $((last_id - next_id + 2)) do_facet $mfacet lctl get_param osp.$mdtosc_proc2.prealloc* do_facet $ofacet lctl set_param fail_loc=$FAILLOC +} + +exhaust_precreations() { + __exhaust_precreations $1 $2 $3 sleep_maxage } exhaust_all_precreations() { local i for (( i=0; i < OSTCOUNT; i++ )) ; do - exhaust_precreations $i $1 -1 + __exhaust_precreations $i $1 -1 done + sleep_maxage } test_27n() { @@ -5624,16 +5636,24 @@ test_56c() { local ost_idx=0 local ost_name=$(ostname_from_index $ost_idx) local old_status=$(ost_dev_status $ost_idx) + local p="$TMP/$TESTSUITE-$TESTNAME.parameters" [[ -z "$old_status" ]] || skip_env "OST $ost_name is in $old_status status" do_facet ost1 $LCTL set_param -n obdfilter.$ost_name.degraded=1 - [[ $OST1_VERSION -ge $(version_code 2.12.55) ]] && do_facet ost1 \ + [[ $OST1_VERSION -lt $(version_code 2.12.55) ]] || do_facet ost1 \ $LCTL set_param -n obdfilter.$ost_name.no_precreate=1 + if [[ $OST1_VERSION -ge $(version_code 2.12.57) ]]; then + save_lustre_params ost1 osd-*.$ost_name.nonrotational > $p + do_facet ost1 $LCTL set_param -n osd-*.$ost_name.nonrotational=1 + fi + + [[ $($LFS df -v $MOUNT |& grep -c "inactive device") -eq 0 ]] || + error "$LFS df -v showing inactive devices" sleep_maxage - local new_status=$(ost_dev_status $ost_idx) + local new_status=$(ost_dev_status $ost_idx $MOUNT -v) [[ "$new_status" =~ "D" ]] || error "$ost_name status is '$new_status', missing 'D'" @@ -5641,18 +5661,37 @@ test_56c() { [[ "$new_status" =~ "N" ]] || error "$ost_name status is '$new_status', missing 'N'" fi + if [[ $OST1_VERSION -ge $(version_code 2.12.57) ]]; then + [[ "$new_status" =~ "f" ]] || + error "$ost_name status is '$new_status', missing 'f'" + fi do_facet ost1 $LCTL set_param -n obdfilter.$ost_name.degraded=0 - [[ $OST1_VERSION -ge $(version_code 2.12.55) ]] && do_facet ost1 \ + [[ $OST1_VERSION -lt $(version_code 2.12.55) ]] || do_facet ost1 \ $LCTL set_param -n obdfilter.$ost_name.no_precreate=0 + [[ -z "$p" ]] && restore_lustre_params < $p || true sleep_maxage new_status=$(ost_dev_status $ost_idx) [[ ! "$new_status" =~ "D" && ! "$new_status" =~ "N" ]] || error "$ost_name status is '$new_status', has 'D' and/or 'N'" + # can't check 'f' as devices may actually be on flash } run_test 56c "check 'lfs df' showing device status" +test_56d() { + local mdts=$($LFS df -v $MOUNT | grep -c MDT) + local osts=$($LFS df -v $MOUNT | grep -c OST) + + $LFS df $MOUNT + + (( mdts == MDSCOUNT )) || + error "lfs df -v showed $mdts MDTs, not $MDSCOUNT" + (( osts == OSTCOUNT )) || + error "lfs df -v showed $osts OSTs, not $OSTCOUNT" +} +run_test 56d "'lfs df -v' prints only configured devices" + NUMFILES=3 NUMDIRS=3 setup_56() { @@ -6152,154 +6191,67 @@ test_56r() { } run_test 56r "check lfs find -size works" -test_56ra() { - [[ $MDS1_VERSION -gt $(version_code 2.12.58) ]] || - skip "MDS < 2.12.58 doesn't return LSOM data" - local dir=$DIR/$tdir - - [[ $OSC == "mdc" ]] && skip "DoM files" && return - - setup_56 $dir $NUMFILES $NUMDIRS "-c 1" +test_56ra_sub() { + local expected=$1 + local glimpses=$2 + local cmd="$3" cancel_lru_locks $OSC local rpcs_before=$(calc_stats $OSC.*$OSC*.stats ldlm_glimpse_enqueue) - local expected=12 - local cmd="$LFS find -size 0 -type f -lazy $dir" local nums=$($cmd | wc -l) [ $nums -eq $expected ] || error "'$cmd' wrong: found $nums, expected $expected" local rpcs_after=$(calc_stats $OSC.*$OSC*.stats ldlm_glimpse_enqueue) - [ $rpcs_before -eq $rpcs_after ] || - error "'$cmd' should not send glimpse RPCs to OST" - cmd="$LFS find -size 0 -type f $dir" - nums=$($cmd | wc -l) - [ $nums -eq $expected ] || - error "'$cmd' wrong: found $nums, expected $expected" - rpcs_after=$(calc_stats $OSC.*$OSC*.stats ldlm_glimpse_enqueue) - echo "Before: $rpcs_before After: $rpcs_after $NUMFILES" - $LCTL get_param osc.*.stats - [ $rpcs_after -eq $((rpcs_before + 12)) ] || - error "'$cmd' should send 12 glimpse RPCs to OST" - cancel_lru_locks $OSC - rpcs_before=$(calc_stats $OSC.*$OSC*.stats ldlm_glimpse_enqueue) - expected=0 - cmd="$LFS find ! -size 0 -type f -lazy $dir" - nums=$($cmd | wc -l) - [ $nums -eq $expected ] || - error "'$cmd' wrong: found $nums, expected $expected" - rpcs_after=$(calc_stats $OSC.*$OSC*.stats ldlm_glimpse_enqueue) - $LCTL get_param mdc.*.stats - [ $rpcs_before -eq $rpcs_after ] || - error "'$cmd' should not send glimpse RPCs to OST" - cmd="$LFS find ! -size 0 -type f $dir" - nums=$($cmd | wc -l) - [ $nums -eq $expected ] || - error "'$cmd' wrong: found $nums, expected $expected" - rpcs_after=$(calc_stats $OSC.*$OSC*.stats ldlm_glimpse_enqueue) - echo "Before: $rpcs_before After: $rpcs_after $NUMFILES" - [ $rpcs_after -eq $((rpcs_before + 12)) ] || - error "'$cmd' should send 12 glimpse RPCs to OST" + if (( rpcs_before + glimpses != rpcs_after )); then + echo "Before: $rpcs_before After: $rpcs_after $NUMFILES" + $LCTL get_param osc.*.stats | grep ldlm_glimpse_enqueue - echo "test" > $dir/$tfile - echo "test2" > $dir/$tfile.2 && sync - cancel_lru_locks $OSC - rpcs_before=$(calc_stats $OSC.*$OSC*.stats ldlm_glimpse_enqueue) - expected=1 - cmd="$LFS find -size 5 -type f -lazy $dir" - nums=$($cmd | wc -l) - [ $nums -eq $expected ] || - error "'$cmd' wrong: found $nums, expected $expected" - rpcs_after=$(calc_stats $OSC.*$OSC*.stats ldlm_glimpse_enqueue) - [ $rpcs_before -eq $rpcs_after ] || - error "'$cmd' should not send glimpse RPCs to OST" - cmd="$LFS find -size 5 -type f $dir" - nums=$($cmd | wc -l) - [ $nums -eq $expected ] || - error "'$cmd' wrong: found $nums, expected $expected" - rpcs_after=$(calc_stats $OSC.*$OSC*.stats ldlm_glimpse_enqueue) - echo "Before: $rpcs_before After: $rpcs_after $NUMFILES" - [ $rpcs_after -eq $((rpcs_before + 14)) ] || - error "'$cmd' should send 14 glimpse RPCs to OST" + if [[ $glimpses == 0 ]]; then + error "'$cmd' should not send glimpse RPCs to OST" + else + error "'$cmd' should send $glimpses glimpse RPCs to OST" + fi + fi +} - cancel_lru_locks $OSC - rpcs_before=$(calc_stats $OSC.*$OSC*.stats ldlm_glimpse_enqueue) - expected=1 - cmd="$LFS find -size +5 -type f -lazy $dir" - nums=$($cmd | wc -l) - [ $nums -eq $expected ] || - error "'$cmd' wrong: found $nums, expected $expected" - rpcs_after=$(calc_stats $OSC.*$OSC*.stats ldlm_glimpse_enqueue) - [ $rpcs_before -eq $rpcs_after ] || - error "'$cmd' should not send glimpse RPCs to OST" - cmd="$LFS find -size +5 -type f $dir" - nums=$($cmd | wc -l) - [ $nums -eq $expected ] || - error "'$cmd' wrong: found $nums, expected $expected" - rpcs_after=$(calc_stats $OSC.*$OSC*.stats ldlm_glimpse_enqueue) - echo "Before: $rpcs_before After: $rpcs_after $NUMFILES" - [ $rpcs_after -eq $((rpcs_before + 14)) ] || - error "'$cmd' should send 14 glimpse RPCs to OST" +test_56ra() { + [[ $MDS1_VERSION -gt $(version_code 2.12.58) ]] || + skip "MDS < 2.12.58 doesn't return LSOM data" + local dir=$DIR/$tdir - cancel_lru_locks $OSC - rpcs_before=$(calc_stats $OSC.*$OSC*.stats ldlm_glimpse_enqueue) - expected=2 - cmd="$LFS find -size +0 -type f -lazy $dir" - nums=$($cmd | wc -l) - [ $nums -eq $expected ] || - error "'$cmd' wrong: found $nums, expected $expected" - rpcs_after=$(calc_stats $OSC.*$OSC*.stats ldlm_glimpse_enqueue) - [ $rpcs_before -eq $rpcs_after ] || - error "'$cmd' should not send glimpse RPCs to OST" - cmd="$LFS find -size +0 -type f $dir" - nums=$($cmd | wc -l) - [ $nums -eq $expected ] || - error "'$cmd' wrong: found $nums, expected $expected" - rpcs_after=$(calc_stats $OSC.*$OSC*.stats ldlm_glimpse_enqueue) - echo "Before: $rpcs_before After: $rpcs_after $NUMFILES" - [ $rpcs_after -eq $((rpcs_before + 14)) ] || - error "'$cmd' should send 14 glimpse RPCs to OST" + [[ $OSC == "mdc" ]] && skip "DoM files" && return + setup_56 $dir $NUMFILES $NUMDIRS "-c 1" + # open and close all files to ensure LSOM is updated cancel_lru_locks $OSC - rpcs_before=$(calc_stats $OSC.*$OSC*.stats ldlm_glimpse_enqueue) - expected=2 - cmd="$LFS find ! -size -5 -type f -lazy $dir" - nums=$($cmd | wc -l) - [ $nums -eq $expected ] || - error "'$cmd' wrong: found $nums, expected $expected" - rpcs_after=$(calc_stats $OSC.*$OSC*.stats ldlm_glimpse_enqueue) - [ $rpcs_before -eq $rpcs_after ] || - error "'$cmd' should not send glimpse RPCs to OST" - cmd="$LFS find ! -size -5 -type f $dir" - nums=$($cmd | wc -l) - [ $nums -eq $expected ] || - error "'$cmd' wrong: found $nums, expected $expected" - rpcs_after=$(calc_stats $OSC.*$OSC*.stats ldlm_glimpse_enqueue) - echo "Before: $rpcs_before After: $rpcs_after $NUMFILES" - [ $rpcs_after -eq $((rpcs_before + 14)) ] || - error "'$cmd' should send 14 glimpse RPCs to OST" + find $dir -type f | xargs cat > /dev/null + + # expect_found glimpse_rpcs command_to_run + test_56ra_sub 12 0 "$LFS find -size 0 -type f -lazy $dir" + test_56ra_sub 12 12 "$LFS find -size 0 -type f $dir" + test_56ra_sub 0 0 "$LFS find ! -size 0 -type f -lazy $dir" + test_56ra_sub 0 12 "$LFS find ! -size 0 -type f $dir" + echo "test" > $dir/$tfile + echo "test2" > $dir/$tfile.2 && sync cancel_lru_locks $OSC - rpcs_before=$(calc_stats $OSC.*$OSC*.stats ldlm_glimpse_enqueue) - expected=12 - cmd="$LFS find -size -5 -type f -lazy $dir" - nums=$($cmd | wc -l) - [ $nums -eq $expected ] || - error "'$cmd' wrong: found $nums, expected $expected" - rpcs_after=$(calc_stats $OSC.*$OSC*.stats ldlm_glimpse_enqueue) - [ $rpcs_before -eq $rpcs_after ] || - error "'$cmd' should not send glimpse RPCs to OST" - cmd="$LFS find -size -5 -type f $dir" - nums=$($cmd | wc -l) - [ $nums -eq $expected ] || - error "'$cmd' wrong: found $nums, expected $expected" - rpcs_after=$(calc_stats $OSC.*$OSC*.stats ldlm_glimpse_enqueue) - echo "Before: $rpcs_before After: $rpcs_after $NUMFILES" - [ $rpcs_after -eq $((rpcs_before + 14)) ] || - error "'$cmd' should send 14 glimpse RPCs to OST" + cat $dir/$tfile $dir/$tfile.2 > /dev/null + + test_56ra_sub 1 0 "$LFS find -size 5 -type f -lazy $dir" + test_56ra_sub 1 14 "$LFS find -size 5 -type f $dir" + test_56ra_sub 1 0 "$LFS find -size +5 -type f -lazy $dir" + test_56ra_sub 1 14 "$LFS find -size +5 -type f $dir" + + test_56ra_sub 2 0 "$LFS find -size +0 -type f -lazy $dir" + test_56ra_sub 2 14 "$LFS find -size +0 -type f $dir" + test_56ra_sub 2 0 "$LFS find ! -size -5 -type f -lazy $dir" + test_56ra_sub 2 14 "$LFS find ! -size -5 -type f $dir" + test_56ra_sub 12 0 "$LFS find -size -5 -type f -lazy $dir" + test_56ra_sub 12 14 "$LFS find -size -5 -type f $dir" } run_test 56ra "check lfs find -size -lazy works for data on OSTs" @@ -6844,6 +6796,24 @@ test_56wd() { } run_test 56wd "check lfs_migrate --rsync and --no-rsync work" +test_56we() { + local td=$DIR/$tdir + local tf=$td/$tfile + + test_mkdir $td || error "cannot create $td" + touch $tf || error "cannot touch $tf" + + echo -n "Make sure --non-direct|-D works..." + $LFS_MIGRATE -y --non-direct -v $tf 2>&1 | + grep -q "lfs migrate --non-direct" || + error "--non-direct option cannot work correctly" + $LFS_MIGRATE -y -D -v $tf 2>&1 | + grep -q "lfs migrate -D" || + error "-D option cannot work correctly" + echo "done." +} +run_test 56we "check lfs_migrate --non-direct|-D support" + test_56x() { [[ $OSTCOUNT -lt 2 ]] && skip_env "needs >= 2 OSTs" check_swap_layouts_support @@ -11485,8 +11455,8 @@ test_121() { #bug #10589 } run_test 121 "read cancel race =========" -test_123a() { # was test 123, statahead(bug 11401) - [ $PARALLEL == "yes" ] && skip "skip parallel run" +test_123a_base() { # was test 123, statahead(bug 11401) + local lsx="$1" SLOWOK=0 if ! grep -q "processor.*: 1" /proc/cpuinfo; then @@ -11498,76 +11468,120 @@ test_123a() { # was test 123, statahead(bug 11401) test_mkdir $DIR/$tdir NUMFREE=$(df -i -P $DIR | tail -n 1 | awk '{ print $4 }') [[ $NUMFREE -gt 100000 ]] && NUMFREE=100000 || NUMFREE=$((NUMFREE-1000)) - MULT=10 - for ((i=100, j=0; i<=$NUMFREE; j=$i, i=$((i * MULT)) )); do - createmany -o $DIR/$tdir/$tfile $j $((i - j)) - - max=`lctl get_param -n llite.*.statahead_max | head -n 1` - lctl set_param -n llite.*.statahead_max 0 - lctl get_param llite.*.statahead_max - cancel_lru_locks mdc - cancel_lru_locks osc - stime=`date +%s` - time ls -l $DIR/$tdir | wc -l - etime=`date +%s` - delta=$((etime - stime)) - log "ls $i files without statahead: $delta sec" - lctl set_param llite.*.statahead_max=$max - - swrong=`lctl get_param -n llite.*.statahead_stats | grep "statahead wrong:" | awk '{print $3}'` - lctl get_param -n llite.*.statahead_max | grep '[0-9]' - cancel_lru_locks mdc - cancel_lru_locks osc - stime=`date +%s` - time ls -l $DIR/$tdir | wc -l - etime=`date +%s` - delta_sa=$((etime - stime)) - log "ls $i files with statahead: $delta_sa sec" - lctl get_param -n llite.*.statahead_stats - ewrong=`lctl get_param -n llite.*.statahead_stats | grep "statahead wrong:" | awk '{print $3}'` + MULT=10 + for ((i=100, j=0; i<=$NUMFREE; j=$i, i=$((i * MULT)) )); do + createmany -o $DIR/$tdir/$tfile $j $((i - j)) + + max=$(lctl get_param -n llite.*.statahead_max | head -n 1) + lctl set_param -n llite.*.statahead_max 0 + lctl get_param llite.*.statahead_max + cancel_lru_locks mdc + cancel_lru_locks osc + stime=$(date +%s) + time $lsx $DIR/$tdir | wc -l + etime=$(date +%s) + delta=$((etime - stime)) + log "$lsx $i files without statahead: $delta sec" + lctl set_param llite.*.statahead_max=$max + + swrong=$(lctl get_param -n llite.*.statahead_stats | + grep "statahead wrong:" | awk '{print $3}') + lctl get_param -n llite.*.statahead_max | grep '[0-9]' + cancel_lru_locks mdc + cancel_lru_locks osc + stime=$(date +%s) + time $lsx $DIR/$tdir | wc -l + etime=$(date +%s) + delta_sa=$((etime - stime)) + log "$lsx $i files with statahead: $delta_sa sec" + lctl get_param -n llite.*.statahead_stats + ewrong=$(lctl get_param -n llite.*.statahead_stats | + grep "statahead wrong:" | awk '{print $3}') [[ $swrong -lt $ewrong ]] && log "statahead was stopped, maybe too many locks held!" [[ $delta -eq 0 || $delta_sa -eq 0 ]] && continue - if [ $((delta_sa * 100)) -gt $((delta * 105)) -a $delta_sa -gt $((delta + 2)) ]; then - max=`lctl get_param -n llite.*.statahead_max | head -n 1` - lctl set_param -n llite.*.statahead_max 0 - lctl get_param llite.*.statahead_max - cancel_lru_locks mdc - cancel_lru_locks osc - stime=`date +%s` - time ls -l $DIR/$tdir | wc -l - etime=`date +%s` - delta=$((etime - stime)) - log "ls $i files again without statahead: $delta sec" - lctl set_param llite.*.statahead_max=$max - if [ $((delta_sa * 100)) -gt $((delta * 105)) -a $delta_sa -gt $((delta + 2)) ]; then - if [ $SLOWOK -eq 0 ]; then - error "ls $i files is slower with statahead!" - else - log "ls $i files is slower with statahead!" - fi - break - fi - fi + if [ $((delta_sa * 100)) -gt $((delta * 105)) -a $delta_sa -gt $((delta + 2)) ]; then + max=$(lctl get_param -n llite.*.statahead_max | + head -n 1) + lctl set_param -n llite.*.statahead_max 0 + lctl get_param llite.*.statahead_max + cancel_lru_locks mdc + cancel_lru_locks osc + stime=$(date +%s) + time $lsx $DIR/$tdir | wc -l + etime=$(date +%s) + delta=$((etime - stime)) + log "$lsx $i files again without statahead: $delta sec" + lctl set_param llite.*.statahead_max=$max + if [ $((delta_sa * 100 > delta * 105 && delta_sa > delta + 2)) ]; then + if [ $SLOWOK -eq 0 ]; then + error "$lsx $i files is slower with statahead!" + else + log "$lsx $i files is slower with statahead!" + fi + break + fi + fi - [ $delta -gt 20 ] && break - [ $delta -gt 8 ] && MULT=$((50 / delta)) - [ "$SLOW" = "no" -a $delta -gt 5 ] && break - done - log "ls done" + [ $delta -gt 20 ] && break + [ $delta -gt 8 ] && MULT=$((50 / delta)) + [ "$SLOW" = "no" -a $delta -gt 5 ] && break + done + log "$lsx done" - stime=`date +%s` - rm -r $DIR/$tdir - sync - etime=`date +%s` - delta=$((etime - stime)) - log "rm -r $DIR/$tdir/: $delta seconds" - log "rm done" - lctl get_param -n llite.*.statahead_stats + stime=$(date +%s) + rm -r $DIR/$tdir + sync + etime=$(date +%s) + delta=$((etime - stime)) + log "rm -r $DIR/$tdir/: $delta seconds" + log "rm done" + lctl get_param -n llite.*.statahead_stats +} + +test_123aa() { + [ $PARALLEL == "yes" ] && skip "skip parallel run" + + test_123a_base "ls -l" +} +run_test 123aa "verify statahead work" + +test_123ab() { + [ $PARALLEL == "yes" ] && skip "skip parallel run" + + statx_supported || skip_env "Test must be statx() syscall supported" + + test_123a_base "$STATX -l" +} +run_test 123ab "verify statahead work by using statx" + +test_123ac() { + [ $PARALLEL == "yes" ] && skip "skip parallel run" + + statx_supported || skip_env "Test must be statx() syscall supported" + + local rpcs_before + local rpcs_after + local agl_before + local agl_after + + cancel_lru_locks $OSC + rpcs_before=$(calc_stats $OSC.*$OSC*.stats ldlm_glimpse_enqueue) + agl_before=$($LCTL get_param -n llite.*.statahead_stats | + awk '/agl.total:/ {print $3}') + test_123a_base "$STATX -c \"%n %i %A %h %u %g %W %X %Z\" -D" + test_123a_base "$STATX --cached=always -D" + agl_after=$($LCTL get_param -n llite.*.statahead_stats | + awk '/agl.total:/ {print $3}') + [ $agl_before -eq $agl_after ] || + error "Should not trigger AGL thread - $agl_before:$agl_after" + rpcs_after=$(calc_stats $OSC.*$OSC*.stats ldlm_glimpse_enqueue) + [ $rpcs_after -eq $rpcs_before ] || + error "$STATX should not send glimpse RPCs to $OSC" } -run_test 123a "verify statahead work" +run_test 123ac "verify statahead work by using statx without glimpse RPCs" test_123b () { # statahead(bug 15027) [ $PARALLEL == "yes" ] && skip "skip parallel run" @@ -12092,7 +12106,7 @@ set_dir_limits () { check_mds_dmesg() { local facets=$(get_facets MDS) for facet in ${facets//,/ }; do - do_facet $facet "dmesg | tail -3 | grep -q $1" && return 0 + do_facet $facet "dmesg | tail -3 | grep $1" && return 0 done return 1 } @@ -12107,55 +12121,49 @@ test_129() { remote_mds_nodsh && skip "remote MDS with nodsh" local ENOSPC=28 - local EFBIG=27 local has_warning=false rm -rf $DIR/$tdir mkdir -p $DIR/$tdir # block size of mds1 - local maxsize=$(($($LCTL get_param -n mdc.*MDT0000*.blocksize) * 5)) - set_dir_limits $maxsize $maxsize + local maxsize=$(($($LCTL get_param -n mdc.*MDT0000*.blocksize) * 8)) + set_dir_limits $maxsize $((maxsize * 6 / 8)) + stack_trap "set_dir_limits 0 0" + stack_trap "unlinkmany $DIR/$tdir/file_base_ 2000 || true" local dirsize=$(stat -c%s "$DIR/$tdir") local nfiles=0 - while [[ $dirsize -le $maxsize ]]; do - $MULTIOP $DIR/$tdir/file_base_$nfiles Oc + while (( $dirsize <= $maxsize )); do + $MCREATE $DIR/$tdir/file_base_$nfiles rc=$? - if ! $has_warning; then - check_mds_dmesg '"is approaching"' && has_warning=true - fi # check two errors: - # ENOSPC for new ext4 max_dir_size (kernel commit df981d03ee) - # EFBIG for previous versions included in ldiskfs series - if [ $rc -eq $EFBIG ] || [ $rc -eq $ENOSPC ]; then + # ENOSPC for ext4 max_dir_size, which has been used since + # kernel v3.6-rc1-8-gdf981d03ee, lustre v2_4_50_0-79-gaed82035c0 + if (( rc == ENOSPC )); then set_dir_limits 0 0 - echo "return code $rc received as expected" + echo "rc=$rc returned as expected after $nfiles files" createmany -o $DIR/$tdir/file_extra_$nfiles. 5 || - error_exit "create failed w/o dir size limit" - - check_mds_dmesg '"has reached"' || - error_exit "reached message should be output" + error "create failed w/o dir size limit" - [ $has_warning = "false" ] && - error_exit "warning message should be output" + # messages may be rate limited if test is run repeatedly + check_mds_dmesg '"is approaching max"' || + echo "warning message should be output" + check_mds_dmesg '"has reached max"' || + echo "reached message should be output" dirsize=$(stat -c%s "$DIR/$tdir") [[ $dirsize -ge $maxsize ]] && return 0 - error_exit "current dir size $dirsize, " \ - "previous limit $maxsize" - elif [ $rc -ne 0 ]; then - set_dir_limits 0 0 - error_exit "return $rc received instead of expected " \ - "$EFBIG or $ENOSPC, files in dir $dirsize" + error "dirsize $dirsize < $maxsize after $nfiles files" + elif (( rc != 0 )); then + break fi nfiles=$((nfiles + 1)) dirsize=$(stat -c%s "$DIR/$tdir") done - set_dir_limits 0 0 - error "exceeded dir size limit $maxsize($MDSCOUNT) : $dirsize bytes" + error "rc=$rc, size=$dirsize/$maxsize, mdt=$MDSCOUNT, nfiles=$nfiles" } run_test 129 "test directory size limit ========================" @@ -12882,38 +12890,21 @@ test_133e() { } run_test 133e "Verifying OST {read,write}_bytes nid stats =================" -proc_regexp="/{proc,sys}/{fs,sys,kernel/debug}/{lustre,lnet}/" - -# Some versions of find (4.5.11, 4.5.14) included in CentOS 7.3-7.5 do -# not honor the -ignore_readdir_race option correctly. So we call -# error_ignore() rather than error() in these cases. See LU-11152. -error_133() { - if (find --version; do_facet mds1 find --version) | - grep -q '\b4\.5\.1[1-4]\b'; then - error_ignore LU-11152 "$@" - else - error "$@" - fi -} - test_133f() { - # First without trusting modes. - local proc_dirs=$(eval \ls -d $proc_regexp 2>/dev/null) - echo "proc_dirs='$proc_dirs'" - [ -n "$proc_dirs" ] || error "no proc_dirs on $HOSTNAME" - find $proc_dirs -exec cat '{}' \; &> /dev/null + [[ $(lustre_version_code $facet) -ge $(version_code 2.7.65) ]] || + skip "too old lustre for get_param -R ($facet_ver)" - # Second verifying readability. + # verifying readability. $LCTL get_param -R '*' &> /dev/null # Verifing writability with badarea_io. - find $proc_dirs \ - -ignore_readdir_race \ - -type f \ - -not -name force_lbug \ - -not -name changelog_mask \ - -exec badarea_io '{}' \; || - error_133 "find $proc_dirs failed" + $LCTL list_param -FR '*' | grep '=' | tr -d = | + egrep -v 'force_lbug|changelog_mask' | xargs badarea_io || + error "client badarea_io failed" + + # remount the FS in case writes/reads /proc break the FS + cleanup || error "failed to unmount" + setup || error "failed to setup" } run_test 133f "Check reads/writes of client lustre proc files with bad area io" @@ -12930,10 +12921,10 @@ test_133g() { log "$facet: too old lustre for get_param -R" fi if [ $facet_ver -ge $(version_code 2.5.54) ]; then - do_facet $facet "$LCTL list_param -R '*' | grep '=' | + do_facet $facet "$LCTL list_param -FR '*' | grep '=' | tr -d = | egrep -v 'force_lbug|changelog_mask' | xargs badarea_io" || - error_133 "$facet badarea_io failed" + error "$facet badarea_io failed" else skip_noexit "$facet: too old lustre for get_param -R" fi @@ -12942,7 +12933,6 @@ test_133g() { # remount the FS in case writes/reads /proc break the FS cleanup || error "failed to unmount" setup || error "failed to setup" - true } run_test 133g "Check reads/writes of server lustre proc files with bad area io" @@ -12953,24 +12943,21 @@ test_133h() { skip "Need MDS version at least 2.9.54" local facet - for facet in client mds1 ost1; do - local facet_proc_dirs=$(do_facet $facet \ - \\\ls -d $proc_regexp 2> /dev/null) - [ -z "$facet_proc_dirs" ] && error "no proc_dirs on $facet" - echo "${facet}_proc_dirs='$facet_proc_dirs'" # Get the list of files that are missing the terminating newline - local missing=($(do_facet $facet \ - find ${facet_proc_dirs} -type f \| \ - while read F\; do \ - awk -v FS='\v' -v RS='\v\v' \ - "'END { if(NR>0 && \ - \\\$NF !~ /.*\\\n\$/) \ - print FILENAME}'" \ - '\$F'\; \ - done 2>/dev/null)) - [ ${#missing[*]} -eq 0 ] || - error "files do not end with newline: ${missing[*]}" + local plist=$(do_facet $facet + $LCTL list_param -FR '*' | grep '=' | tr -d =) + local ent + for ent in $plist; do + local missing=$(do_facet $facet $LCTL get_param $ent \|\ + awk -v FS='\v' -v RS='\v\v' \ + "'END { if(NR>0 && \\\$NF !~ /.*\\\n\$/) \ + print FILENAME}'" 2>/dev/null) + [ -z $missing ] || { + do_facet $facet $LCTL get_param $ent | od -An -tx1 + error "file does not end with newline: $facet-$ent" + } + done done } run_test 133h "Proc files should end with newlines" @@ -13159,7 +13146,7 @@ test_140() { #bug-17379 } run_test 140 "Check reasonable stack depth (shouldn't LBUG) ====" -test_150() { +test_150a() { [ $PARALLEL == "yes" ] && skip "skip parallel run" local TF="$TMP/$tfile" @@ -13190,7 +13177,60 @@ test_150() { rm -f $TF true } -run_test 150 "truncate/append tests" +run_test 150a "truncate/append tests" + +test_150b() { + [ "$ost1_FSTYPE" != ldiskfs ] && skip "non-ldiskfs backend" + [ $OST1_VERSION -lt $(version_code 2.13.50) ] && + skip "Need OST version at least 2.13.53" + touch $DIR/$tfile + check_fallocate $DIR/$tfile || error "fallocate failed" +} +run_test 150b "Verify fallocate (prealloc) functionality" + +test_150c() { + local bytes + local want + + [ "$ost1_FSTYPE" != ldiskfs ] && skip "non-ldiskfs backend" + [ $OST1_VERSION -lt $(version_code 2.13.50) ] && + skip "Need OST version at least 2.13.53" + + $LFS setstripe -c $OSTCOUNT -S1M $DIR/$tdir || error "setstripe failed" + fallocate -l ${OSTCOUNT}m $DIR/$tdir || error "fallocate failed" + sync; sync_all_data + cancel_lru_locks $OSC + sleep 5 + bytes=$(($(stat -c '%b * %B' $DIR/$tdir))) + want=$((OSTCOUNT * 1048576)) + + # Must allocate all requested space, not more than 5% extra + (( $bytes >= $want && $bytes < $want * 105 / 100 )) || + error "bytes $bytes is not $want" +} +run_test 150c "Verify fallocate Size and Blocks" + +test_150d() { + local bytes + local want + + [ "$ost1_FSTYPE" != ldiskfs ] && skip "non-ldiskfs backend" + [ $OST1_VERSION -lt $(version_code 2.13.50) ] && + skip "Need OST version at least 2.13.53" + + $LFS setstripe -c $OSTCOUNT -S1M $DIR/$tdir || error "setstripe failed" + fallocate -o 1G -l ${OSTCOUNT}m $DIR/$tdir || error "fallocate failed" + sync; sync_all_data + cancel_lru_locks $OSC + sleep 5 + bytes=$(($(stat -c '%b * %B' $DIR/$tdir))) + want=$((OSTCOUNT * 1048576)) + + # Must allocate all requested space, not more than 5% extra + (( $bytes >= $want && $bytes < $want * 105 / 100 )) || + error "bytes $bytes is not $want" +} +run_test 150d "Verify fallocate Size and Blocks - Non zero start" #LU-2902 roc_hit was not able to read all values from lproc function roc_hit_init() { @@ -16313,16 +16353,20 @@ test_205a() { # Job stats } run_test 205a "Verify job stats" -# LU-13117 +# LU-13117, LU-13597 test_205b() { + job_stats="mdt.*.job_stats" + $LCTL set_param $job_stats=clear $LCTL set_param jobid_var=USER jobid_name="%e.%u" env -i USERTESTJOBSTATS=foolish touch $DIR/$tfile.1 - do_facet $SINGLEMDS $LCTL get_param mdt.*.job_stats | - grep job_id: | grep foolish && + do_facet $SINGLEMDS $LCTL get_param $job_stats | + grep "job_id:.*foolish" && error "Unexpected jobid found" - true + do_facet $SINGLEMDS $LCTL get_param $job_stats | + grep "open:.*min.*max.*sum" || + error "wrong job_stats format found" } -run_test 205b "Verify job stats jobid parsing" +run_test 205b "Verify job stats jobid and output format" # LU-1480, LU-1773 and LU-1657 test_206() { @@ -16447,6 +16491,27 @@ test_209() { } run_test 209 "read-only open/close requests should be freed promptly" +test_210() { + local pid + + $MULTIOP $DIR/$tfile oO_CREAT:O_RDWR:eW_E+eUc & + pid=$! + sleep 1 + + $LFS getstripe $DIR/$tfile + kill -USR1 $pid + wait $pid || error "multiop failed" + + $MULTIOP $DIR/$tfile oO_RDONLY:eR_E+eUc & + pid=$! + sleep 1 + + $LFS getstripe $DIR/$tfile + kill -USR1 $pid + wait $pid || error "multiop failed" +} +run_test 210 "lfs getstripe does not break leases" + test_212() { size=`date +%s` size=$((size % 8192 + 1)) @@ -17707,7 +17772,7 @@ run_test 230i "lfs migrate -m tolerates trailing slashes" test_230j() { [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" - [ $MDS1_VERSION -lt $(version_code 2.11.52) ] && + [ $MDS1_VERSION -lt $(version_code 2.13.52) ] && skip "Need MDS version at least 2.11.52" $LFS mkdir -m 0 -c 1 $DIR/$tdir || error "mkdir $tdir failed" @@ -17715,7 +17780,7 @@ test_230j() { error "create $tfile failed" cat /etc/passwd > $DIR/$tdir/$tfile - $LFS migrate -m 1 $DIR/$tdir + $LFS migrate -m 1 $DIR/$tdir || error "migrate failed" cmp /etc/passwd $DIR/$tdir/$tfile || error "DoM file mismatch after migration" @@ -17854,6 +17919,156 @@ test_230n() { } run_test 230n "Dir migration with mirrored file" +test_230o() { + [ $MDSCOUNT -ge 2 ] || skip "needs >= 2 MDTs" + [ $MDS1_VERSION -ge $(version_code 2.13.52) ] || + skip "Need MDS version at least 2.13.52" + + local mdts=$(comma_list $(mdts_nodes)) + + local restripe_status + local delta + local i + local j + + # in case "crush" hash type is not set + do_nodes $mdts "$LCTL set_param lod.*.mdt_hash=crush" + + restripe_status=$(do_facet mds1 $LCTL get_param -n \ + mdt.*MDT0000.enable_dir_restripe) + do_nodes $mdts "$LCTL set_param mdt.*.enable_dir_restripe=1" + stack_trap "do_nodes $mdts $LCTL set_param \ + mdt.*.enable_dir_restripe=$restripe_status" + + mkdir $DIR/$tdir + createmany -m $DIR/$tdir/f 100 || + error "create files under remote dir failed $i" + createmany -d $DIR/$tdir/d 100 || + error "create dirs under remote dir failed $i" + + for i in $(seq 2 $MDSCOUNT); do + do_nodes $mdts "$LCTL set_param mdt.*.md_stats=clear > /dev/null" + $LFS setdirstripe -c $i $DIR/$tdir || + error "split -c $i $tdir failed" + wait_update $HOSTNAME \ + "$LFS getdirstripe -H $DIR/$tdir" "crush" 100 || + error "dir split not finished" + delta=$(do_nodes $mdts "lctl get_param -n mdt.*MDT*.md_stats" | + awk '/migrate/ {sum += $2} END { print sum }') + echo "$delta files migrated when dir split from $((i - 1)) to $i stripes" + # delta is around total_files/stripe_count + [ $delta -lt $((200 /(i - 1))) ] || + error "$delta files migrated" + done +} +run_test 230o "dir split" + +test_230p() { + [ $MDSCOUNT -ge 2 ] || skip "needs >= 2 MDTs" + [ $MDS1_VERSION -ge $(version_code 2.13.52) ] || + skip "Need MDS version at least 2.13.52" + + local mdts=$(comma_list $(mdts_nodes)) + + local restripe_status + local delta + local i + local j + + do_nodes $mdts "$LCTL set_param lod.*.mdt_hash=crush" + + restripe_status=$(do_facet mds1 $LCTL get_param -n \ + mdt.*MDT0000.enable_dir_restripe) + do_nodes $mdts "$LCTL set_param mdt.*.enable_dir_restripe=1" + stack_trap "do_nodes $mdts $LCTL set_param \ + mdt.*.enable_dir_restripe=$restripe_status" + + test_mkdir -c $MDSCOUNT -H crush $DIR/$tdir + createmany -m $DIR/$tdir/f 100 || + error "create files under remote dir failed $i" + createmany -d $DIR/$tdir/d 100 || + error "create dirs under remote dir failed $i" + + for i in $(seq $((MDSCOUNT - 1)) -1 1); do + local mdt_hash="crush" + + do_nodes $mdts "$LCTL set_param mdt.*.md_stats=clear > /dev/null" + $LFS setdirstripe -c $i $DIR/$tdir || + error "split -c $i $tdir failed" + [ $i -eq 1 ] && mdt_hash="none" + wait_update $HOSTNAME \ + "$LFS getdirstripe -H $DIR/$tdir" $mdt_hash 100 || + error "dir merge not finished" + delta=$(do_nodes $mdts "lctl get_param -n mdt.*MDT*.md_stats" | + awk '/migrate/ {sum += $2} END { print sum }') + echo "$delta files migrated when dir merge from $((i + 1)) to $i stripes" + # delta is around total_files/stripe_count + [ $delta -lt $((200 / i)) ] || + error "$delta files migrated" + done +} +run_test 230p "dir merge" + +test_230q() { + [ $MDSCOUNT -ge 2 ] || skip "needs >= 2 MDTs" + [ $MDS1_VERSION -ge $(version_code 2.13.52) ] || + skip "Need MDS version at least 2.13.52" + + local mdts=$(comma_list $(mdts_nodes)) + local saved_threshold=$(do_facet mds1 \ + $LCTL get_param -n mdt.*-MDT0000.dir_split_count) + local saved_delta=$(do_facet mds1 \ + $LCTL get_param -n mdt.*-MDT0000.dir_split_delta) + local threshold=100 + local delta=2 + local total=0 + local stripe_count=0 + local stripe_index + local nr_files + + stack_trap "do_nodes $mdts $LCTL set_param \ + mdt.*.dir_split_count=$saved_threshold" + stack_trap "do_nodes $mdts $LCTL set_param \ + mdt.*.dir_split_delta=$saved_delta" + stack_trap "do_nodes $mdts $LCTL set_param mdt.*.dir_restripe_nsonly=1" + do_nodes $mdts "$LCTL set_param mdt.*.enable_dir_auto_split=1" + do_nodes $mdts "$LCTL set_param mdt.*.dir_split_count=$threshold" + do_nodes $mdts "$LCTL set_param mdt.*.dir_split_delta=$delta" + do_nodes $mdts "$LCTL set_param mdt.*.dir_restripe_nsonly=0" + do_nodes $mdts "$LCTL set_param lod.*.mdt_hash=crush" + + $LFS mkdir -i -1 -c 1 $DIR/$tdir || error "mkdir $tdir failed" + stripe_index=$($LFS getdirstripe -i $DIR/$tdir) + + while [ $stripe_count -lt $MDSCOUNT ]; do + createmany -m $DIR/$tdir/f $total $((threshold * 3 / 2)) || + error "create sub files failed" + stat $DIR/$tdir > /dev/null + total=$((total + threshold * 3 / 2)) + stripe_count=$((stripe_count + delta)) + [ $stripe_count -gt $MDSCOUNT ] && stripe_count=$MDSCOUNT + + wait_update $HOSTNAME \ + "$LFS getdirstripe -c $DIR/$tdir" "$stripe_count" 40 || + error "stripe count $($LFS getdirstripe -c $DIR/$tdir) != $stripe_count" + + wait_update $HOSTNAME \ + "$LFS getdirstripe -H $DIR/$tdir" "crush" 200 || + error "stripe hash $($LFS getdirstripe -H $DIR/$tdir) != crush" + + nr_files=$($LFS getstripe -m $DIR/$tdir/* | + grep -w $stripe_index | wc -l) + echo "$nr_files files on MDT$stripe_index after split" + [ $nr_files -lt $((total / (stripe_count - 1))) ] || + error "$nr_files files on MDT$stripe_index after split" + + nr_files=$(ls $DIR/$tdir | wc -w) + [ $nr_files -eq $total ] || + error "total sub files $nr_files != $total" + done +} +run_test 230q "dir auto split" + test_231a() { # For simplicity this test assumes that max_pages_per_rpc @@ -18350,6 +18565,38 @@ test_247e() { } run_test 247e "mount .. as fileset" +test_247f() { + [ $MDSCOUNT -lt 2 ] && skip_env "needs >= 2 MDTs" + [ $MDS1_VERSION -lt $(version_code 2.13.52) ] && + skip "Need at least version 2.13.52" + lctl get_param -n mdc.$FSNAME-MDT0000*.import | + grep -q subtree || + skip "Fileset feature is not supported" + + mkdir $DIR/$tdir || error "mkdir $tdir failed" + $LFS mkdir -i $((MDSCOUNT - 1)) $DIR/$tdir/remote || + error "mkdir remote failed" + mkdir $DIR/$tdir/remote/subdir || error "mkdir remote/subdir failed" + $LFS mkdir -c $MDSCOUNT $DIR/$tdir/striped || + error "mkdir striped failed" + mkdir $DIR/$tdir/striped/subdir || error "mkdir striped/subdir failed" + + local submount=${MOUNT}_$tdir + + mkdir -p $submount || error "mkdir $submount failed" + + local dir + local fileset=$FILESET + + for dir in $tdir/remote $tdir/remote/subdir \ + $tdir/striped $tdir/striped/subdir $tdir/striped/. ; do + FILESET="$fileset/$dir" mount_client $submount || + error "mount $dir failed" + umount_client $submount + done +} +run_test 247f "mount striped or remote directory as fileset" + test_248a() { local fast_read_sav=$($LCTL get_param -n llite.*.fast_read 2>/dev/null) [ -z "$fast_read_sav" ] && skip "no fast read support" @@ -19569,6 +19816,36 @@ test_270g() { } run_test 270g "DoM: default DoM stripe size depends on free space" +test_270h() { + [[ $MDS1_VERSION -ge $(version_code 2.13.53) ]] || + skip "Need MDS version at least 2.13.53" + + local mdtname=${FSNAME}-MDT0000-mdtlov + local dom=$DIR/$tdir/$tfile + local save="$TMP/$TESTSUITE-$TESTNAME.parameters" + + save_lustre_params mds1 "lod.*.dom_stripesize" > $save + stack_trap "restore_lustre_params < $save; rm -f $save" EXIT + + $LFS mkdir -i 0 -c 1 $DIR/$tdir + $LFS setstripe -E 1M -c1 -E -1 -c2 ${dom}_1 || + error "can't create OST file" + # mirrored file with DOM entry in the second mirror + $LFS mirror extend -N -E 1M -L mdt -E eof -c2 ${dom}_1 || + error "can't create mirror with DoM component" + + do_facet mds1 $LCTL set_param -n lod.$mdtname.dom_stripesize=0 + + # DOM component in the middle and has other enries in the same mirror, + # should succeed but lost DoM component + $LFS setstripe --copy=${dom}_1 $dom || + error "Can't create file from OST|DOM mirror layout" + # check new file has no DoM layout after all + [[ $($LFS getstripe -L $dom) != "mdt" ]] || + error "File has DoM component while DoM is disabled" +} +run_test 270h "DoM: DoM stripe removal when disabled on server" + test_271a() { [ $MDS1_VERSION -lt $(version_code 2.10.55) ] && skip "Need MDS version at least 2.10.55" @@ -21573,6 +21850,7 @@ test_401a() { #LU-7437 #count the number of parameters by "list_param -R" local params=$($LCTL list_param -R '*' 2>/dev/null | wc -l) #count the number of parameters by listing proc files + local proc_regexp="/{proc,sys}/{fs,sys,kernel/debug}/{lustre,lnet}/" local proc_dirs=$(eval \ls -d $proc_regexp 2>/dev/null) echo "proc_dirs='$proc_dirs'" [ -n "$proc_dirs" ] || error "no proc_dirs on $HOSTNAME" @@ -22733,6 +23011,14 @@ test_423() { } run_test 423 "statfs should return a right data" +test_424() { +#define OBD_FAIL_PTLRPC_BULK_REPLY_ATTACH 0x522 | OBD_FAIL_ONCE + $LCTL set_param fail_loc=0x80000522 + dd if=/dev/zero of=$DIR/$tfile bs=2M count=1 oflag=sync + rm -f $DIR/$tfile +} +run_test 424 "simulate ENOMEM in ptl_send_rpc bulk reply ME attach" + prep_801() { [[ $MDS1_VERSION -lt $(version_code 2.9.55) ]] || [[ $OST1_VERSION -lt $(version_code 2.9.55) ]] &&