X-Git-Url: https://git.whamcloud.com/?p=fs%2Flustre-release.git;a=blobdiff_plain;f=lustre%2Ftests%2Fconf-sanity.sh;h=acf020d790c8db0e7aedeb1aa745d3affe51ed2b;hp=3bf85985f57030abcbe08571a22a7cdcf01aa6ea;hb=63e17799a369e2ff0b140fd41dc5d7d8656d2bf0;hpb=e5080203e1358673ad7518c2b86bc9a5fc654b5f diff --git a/lustre/tests/conf-sanity.sh b/lustre/tests/conf-sanity.sh index 3bf8598..acf020d 100644 --- a/lustre/tests/conf-sanity.sh +++ b/lustre/tests/conf-sanity.sh @@ -8,14 +8,14 @@ SRCDIR=$(dirname $0) PTLDEBUG=${PTLDEBUG:--1} LUSTRE=${LUSTRE:-$(dirname $0)/..} . $LUSTRE/tests/test-framework.sh -init_test_env $@ +init_test_env "$@" init_logging # tool to create lustre filesystem images ALWAYS_EXCEPT="$CONF_SANITY_EXCEPT 32newtarball" # bug number for skipped test: LU-11915 -ALWAYS_EXCEPT="$ALWAYS_EXCEPT 110 115" +ALWAYS_EXCEPT="$ALWAYS_EXCEPT 110" # UPDATE THE COMMENT ABOVE WITH BUG NUMBERS WHEN CHANGING ALWAYS_EXCEPT! if $SHARED_KEY; then @@ -98,7 +98,7 @@ reformat() { start_mgs () { echo "start mgs service on $(facet_active_host mgs)" - start mgs $(mgsdevname) $MGS_MOUNT_OPTS $@ + start mgs $(mgsdevname) $MGS_MOUNT_OPTS "$@" } start_mdt() { @@ -108,7 +108,7 @@ start_mdt() { shift 1 echo "start mds service on `facet_active_host $facet`" - start $facet ${dev} $MDS_MOUNT_OPTS $@ || return 94 + start $facet ${dev} $MDS_MOUNT_OPTS "$@" || return 94 } stop_mdt_no_force() { @@ -136,7 +136,7 @@ start_mds() { local num for num in $(seq $MDSCOUNT); do - start_mdt $num $@ || return 94 + start_mdt $num "$@" || return 94 done for num in $(seq $MDSCOUNT); do wait_clients_import_state ${CLIENTS:-$HOSTNAME} mds${num} FULL @@ -147,7 +147,7 @@ start_mgsmds() { if ! combined_mgs_mds ; then start_mgs fi - start_mds $@ + start_mds "$@" } stop_mds() { @@ -165,7 +165,7 @@ stop_mgs() { start_ost() { echo "start ost1 service on `facet_active_host ost1`" - start ost1 $(ostdevname 1) $OST_MOUNT_OPTS $@ || return 95 + start ost1 $(ostdevname 1) $OST_MOUNT_OPTS "$@" || return 95 wait_clients_import_state ${CLIENTS:-$HOSTNAME} ost1 FULL } @@ -177,7 +177,7 @@ stop_ost() { start_ost2() { echo "start ost2 service on `facet_active_host ost2`" - start ost2 $(ostdevname 2) $OST_MOUNT_OPTS $@ || return 92 + start ost2 $(ostdevname 2) $OST_MOUNT_OPTS "$@" || return 92 wait_clients_import_state ${CLIENTS:-$HOSTNAME} ost2 FULL } @@ -199,7 +199,7 @@ umount_client() { local mountpath=$1 shift echo "umount lustre on $mountpath....." - zconf_umount $HOSTNAME $mountpath $@ || return 97 + zconf_umount $HOSTNAME $mountpath "$@" || return 97 } manual_umount_client(){ @@ -332,7 +332,7 @@ test_4() { setup touch $DIR/$tfile || error "touch $DIR/$tfile failed" stop_ost || error "Unable to stop OST1" - umount_client $MOUNT -f || error “unmount $MOUNT failed” + umount_client $MOUNT -f || error "unmount $MOUNT failed" cleanup_nocli eno=$? # ok for ost to fail shutdown @@ -530,6 +530,9 @@ test_5h() { run_test 5h "start mdt failure at mdt_fs_setup()" test_5i() { + (( $MDS1_VERSION >= $(version_code 2.12.54) )) || + skip "Need MDS version at least 2.12.54" + setup stop mds1 @@ -593,6 +596,27 @@ test_9() { } run_test 9 "test ptldebug and subsystem for mkfs" +test_10a() { + setup + + if ! combined_mgs_mds; then + files=$(do_facet mgs "find /{proc/fs,sys/fs,sys/kernel/debug}/lustre -type l -exec test ! -e {} \; -print") + [ -z $files ] || echo "MGS $files is a broken symlink" + fi + + files=$(do_facet mds1 "find /{proc/fs,sys/fs,sys/kernel/debug}/lustre -type l -exec test ! -e {} \; -print") + [ -z $files ] || echo "MDS $files is a broken symlink" + + files=$(do_facet ost1 "find /{proc/fs,sys/fs,sys/kernel/debug}/lustre -type l -exec test ! -e {} \; -print") + [ -z $files ] || echo "OSS $files is a broken symlink" + + files=$(do_facet client "find /{proc/fs,sys/fs,sys/kernel/debug}/lustre -type l -exec test ! -e {} \; -print") + [ -z $files ] || echo "clients $files is a broken symlink" + + cleanup || error "cleanup failed with rc $?" +} +run_test 10a "find lctl param broken symlinks" + # # Test 16 was to "verify that lustre will correct the mode of OBJECTS". # But with new MDS stack we don't care about the mode of local objects @@ -680,13 +704,13 @@ test_18() { setup check_mount || error "check_mount failed" - echo "check journal size..." - local FOUNDSIZE=$(do_facet $SINGLEMDS "$DEBUGFS -c -R 'stat <8>' $MDSDEV" | awk '/Size: / { print $NF; exit;}') - if [ $FOUNDSIZE -gt $((32 * 1024 * 1024)) ]; then - log "Success: mkfs creates large journals. Size: $((FOUNDSIZE >> 20))M" - else - error "expected journal size > 32M, found $((FOUNDSIZE >> 20))M" - fi + echo "check journal size..." + local FOUNDSIZE=$(do_facet $SINGLEMDS "$DEBUGFS -c -R 'stat <8>' $MDSDEV" | awk '/Size: / { print $NF; exit;}') + if [ $FOUNDSIZE -gt $((32 * 1024 * 1024)) ]; then + log "Success: mkfs creates large journals. Size: $((FOUNDSIZE >> 20))M" + else + error "expected journal size > 32M, found $((FOUNDSIZE >> 20))M" + fi cleanup || error "cleanup failed with rc $?" @@ -1229,37 +1253,42 @@ test_30a() { setup_noconfig echo Big config llog - TEST="llite.$FSNAME-*.max_read_ahead_whole_mb" - ORIG=$($LCTL get_param -n $TEST) - LIST=(1 2 3 4 5 4 3 2 1 2 3 4 5 4 3 2 1 2 3 4 5) - for i in ${LIST[@]}; do - set_persistent_param_and_check client "$TEST" \ - "$FSNAME.llite.max_read_ahead_whole_mb" $i + local path="llite.$FSNAME-*" + local cpath="$FSNAME.llite" + local param="max_read_ahead_whole_mb" + + local test="${path}.$param" + local conf="${cpath}.$param" + + local orig=$($LCTL get_param -n $test) + local list=(1 2 3 4 5 4 3 2 1 2 3 4 5 4 3 2 1 2 3 4 5) + for i in ${list[@]}; do + set_persistent_param_and_check client $test $conf $i done # make sure client restart still works umount_client $MOUNT mount_client $MOUNT || error "mount_client $MOUNT failed" - [ "$($LCTL get_param -n $TEST)" -ne "$i" ] && - error "Param didn't stick across restart $($TEST) != $i" + [ "$($LCTL get_param -n $test)" -ne "$i" ] && + error "Param didn't stick across restart $($test) != $i" pass echo Erase parameter setting if [[ $PERM_CMD == *"set_param -P"* ]]; then - do_facet mgs "$PERM_CMD -d $TEST" || - error "Erase param $TEST failed" + do_facet mgs "$PERM_CMD -d $test" || + error "Erase param $test failed" else - do_facet mgs "$PERM_CMD \ - -d $FSNAME.llite.max_read_ahead_whole_mb" || - error "Erase param $FSNAME.llite.max_read_ahead_whole_mb failed" + do_facet mgs "$PERM_CMD -d $conf" || + error "Erase param $conf failed" fi umount_client $MOUNT mount_client $MOUNT || error "mount_client $MOUNT failed" - FINAL=$($LCTL get_param -n $TEST) - echo "deleted (default) value=$FINAL, orig=$ORIG" - ORIG=${ORIG%%.[0-9]*} - FINAL=${FINAL%%.[0-9]*} + local final=$($LCTL get_param -n $test) + echo "deleted (default) value=$final, orig=$orig" + orig=${orig%%.[0-9]*} + final=${final%%.[0-9]*} # assumes this parameter started at the default value - [ "$FINAL" -eq "$ORIG" ] || fail "Deleted value=$FINAL, orig=$ORIG" + [ "$final" -eq "$orig" ] || + error "Deleted value=$final -ne orig=$orig" cleanup || error "cleanup failed with rc $?" } @@ -1338,7 +1367,8 @@ run_test 31 "Connect to non-existent node (shouldn't crash)" T32_QID=60000 T32_BLIMIT=40960 # Kbytes T32_ILIMIT=4 - +T32_PRJID=1000 +T32_PROLIMIT=$((T32_BLIMIT/10)) # # This is not really a test but a tool to create new disk # image tarballs for the upgrade tests. @@ -1356,14 +1386,24 @@ test_32newtarball() { local remote_dir local striped_dir local pushd_dir + local pfl_dir + local pfl_file + local dom_dir + local dom_file + local flr_dir + local flr_file + local pj_quota_dir + local pj_quota_file + local target_dir if [ $FSNAME != t32fs -o \( -z "$MDSDEV" -a -z "$MDSDEV1" \) -o \ - $OSTCOUNT -ne 1 -o -z "$OSTDEV1" ]; then - error "Needs FSNAME=t32fs MDSCOUNT=2 " \ - "MDSDEV1=" \ - "MDSDEV2=" \ - "(or MDSDEV, in the case of b1_8)" \ - "OSTCOUNT=1 OSTDEV1=" + $OSTCOUNT -ne 2 -o -z "$OSTDEV1" ]; then + error "Needs FSNAME=t32fs MDSCOUNT=2 " \ + "MDSDEV1= " \ + "MDSDEV2= " \ + "(or MDSDEV, in the case of b1_8) " \ + "OSTCOUNT=2 OSTDEV1= " \ + "OSTDEV2=" fi mkdir $tmp || { @@ -1373,9 +1413,6 @@ test_32newtarball() { mkdir $tmp/src || return 1 tar cf - -C $src . | tar xf - -C $tmp/src - dd if=/dev/zero of=$tmp/src/t32_qf_old bs=1M \ - count=$(($T32_BLIMIT / 1024 / 4)) - chown $T32_QID.$T32_QID $tmp/src/t32_qf_old # format ost with comma-separated NIDs to verify LU-4460 local failnid="$(h2nettype 1.2.3.4),$(h2nettype 4.3.2.1)" @@ -1388,20 +1425,104 @@ test_32newtarball() { $LFS setquota -u $T32_QID -b 0 -B $T32_BLIMIT -i 0 -I $T32_ILIMIT \ /mnt/$FSNAME - tar cf - -C $tmp/src . | tar xf - -C /mnt/$FSNAME - if [[ $MDSCOUNT -ge 2 ]]; then remote_dir=/mnt/$FSNAME/remote_dir $LFS mkdir -i 1 $remote_dir tar cf - -C $tmp/src . | tar xf - -C $remote_dir - if [[ "$MDS1_VERSION" -ge $(version_code 2.7.0) ]]; then + target_dir=$remote_dir + if [[ $MDS1_VERSION -ge $(version_code 2.7.0) ]]; then striped_dir=/mnt/$FSNAME/striped_dir_old $LFS mkdir -i 1 -c 2 $striped_dir tar cf - -C $tmp/src . | tar xf - -C $striped_dir fi + else + tar cf - -C $tmp/src . | tar xf - -C /mnt/$FSNAME + target_dir=/mnt/$FSNAME fi + # add project quota # + [[ $mds1_FSTYPE == "ldiskfs" && + $MDS1_VERSION -gt $(version_code 2.9.55) ]] || + [[ $mds1_FSTYPE == "zfs" && + $MDS1_VERSION -gt $(version_code 2.10.53) ]] && { + pj_quota_dir=$target_dir/project_quota_dir + pj_quota_file_old=$pj_quota_dir/pj_quota_file_old + + enable_project_quota + set_mdt_qtype ugp || + error "enable mdt quota failed" + + set_ost_qtype ugp || + error "enable ost quota failed" + + mkdir -p $pj_quota_dir + $LFS setquota -p $T32_PRJID -b 0 -B $T32_PROLIMIT -i 0 \ + -I $T32_ILIMIT $pj_quota_dir || + error "setquota -p $T32_PRJID failed" + + $LFS setstripe $pj_quota_file_old -c 1 || + error "setstripe $pj_quota_file_old failed" + + chown $T32_QID:$T32_QID $pj_quota_file_old || + error "chown $pj_quota_file_old failed" + + change_project -p $T32_PRJID $pj_quota_file_old + + mkdir -p $tmp/src/project_quota_dir + cp $pj_quota_file_old $tmp/src/project_quota_dir/ + } + + ##################### + tar cf - -C $tmp/src . | tar xf - -C /mnt/$FSNAME + + #if [[ $MDSCOUNT -ge 2 ]]; then + # remote_dir=/mnt/$FSNAME/remote_dir + # $LFS mkdir -i 1 $remote_dir + # tar cf - -C $tmp/src . | tar xf - -C $remote_dir + + # if [[ "$MDS1_VERSION" -ge $(version_code 2.7.0) ]]; then + # striped_dir=/mnt/$FSNAME/striped_dir_old + # $LFS mkdir -i 1 -c 2 $striped_dir + # tar cf - -C $tmp/src . | tar xf - -C $striped_dir + # fi + #fi + + # PFL file # + if [[ $MDS1_VERSION -ge $(version_code 2.9.51) ]]; then + pfl_dir=$target_dir/pfl_dir + pfl_file=$pfl_dir/pfl_file + mkdir -p $pfl_dir + $LFS setstripe -E 2M -c 1 -o 0 -E -1 -S 2M -c 1 -o 1 \ + $pfl_file || + error "Create PFL file failed" + + dd if=/dev/urandom of=$pfl_file bs=1k count=3k + mkdir -p $tmp/src/pfl_dir + cp $pfl_file $tmp/src/pfl_dir/ + fi + + ############ + # DoM / FLR file # + if [[ $MDS1_VERSION -ge $(version_code 2.10.56) ]]; then + dom_dir=$target_dir/dom_dir + dom_file=$dom_dir/dom_file + flr_dir=$target_dir/flr_dir + flr_file=$flr_dir/flr_file + + mkdir -p $dom_dir + $LFS setstripe -E 1M -L mdt -E -1 -S 4M $dom_file + dd if=/dev/urandom of=$dom_file bs=1k count=2k + mkdir -p $tmp/src/dom_dir + cp $dom_file $tmp/src/dom_dir + # FLR # + mkdir -p $flr_dir + LFS mirror create -N2 $flr_file + dd if=/dev/urandom of=$flr_file bs=1k count=1 + mkdir -p $tmp/src/flr_dir + cp $flr_file $tmp/src/flr_dir + fi + ############ stopall mkdir $tmp/img || return 1 @@ -1450,6 +1571,7 @@ test_32newtarball() { kill -s USR1 $pid wait $pid + ################# stopall pushd $tmp/src @@ -1470,7 +1592,14 @@ test_32newtarball() { devname=$(mdsvdevname $num) mv $devname $tmp/img done - mv $OSTDEV1 $tmp/img + + for num in $(seq $OSTCOUNT); do + local devname=$(ostdevname $num) + local facet=oss$num + [[ $(facet_fstype $facet) != zfs ]] || + devname=$(ostdevname $num) + mv $devname $tmp/img + done version=$(sed -e 's/\(^[0-9]\+\.[0-9]\+\)\(.*$\)/\1/' $tmp/img/commit | sed -e 's/\./_/g') # E.g., "1.8.7" -> "1_8" @@ -1497,8 +1626,9 @@ t32_check() { local IMGTYPE="$mds1_FSTYPE" - tarballs=$($r find $RLUSTRE/tests -maxdepth 1 \ - -name \'disk*-$IMGTYPE.tar.bz2\') + [[ -n "$T32IMAGE" ]] && tarballs=$T32IMAGE || + tarballs=$($r find $RLUSTRE/tests -maxdepth 1 \ + -name \'disk*-$IMGTYPE.tar.bz2\') [ -z "$tarballs" ] && skip "No applicable tarballs found" } @@ -1508,6 +1638,7 @@ t32_test_cleanup() { local facet=$SINGLEMDS local rc=$? + echo "== cleanup with rc=$rc ==" if $shall_cleanup_lustre; then umount $tmp/mnt/lustre || rc=$? fi @@ -1520,6 +1651,9 @@ t32_test_cleanup() { if $shall_cleanup_ost; then $r $UMOUNT $tmp/mnt/ost || rc=$? fi + if $shall_cleanup_ost1; then + $r $UMOUNT $tmp/mnt/ost1 || rc=$? + fi $r rm -rf $tmp rm -rf $tmp @@ -1679,6 +1813,21 @@ t32_verify_quota() { return 0 } +getquota() { + local spec=$4 + local uuid=$3 + local mnt=$5 + + sync_all_data > /dev/null 2>&1 || true + + [ "$uuid" != "global" ] || uuid=$mnt + + $LFS quota -v "$1" "$2" $mnt | + awk 'BEGIN { num='$spec' } { if ($1 == "'$uuid'") \ + { if (NF == 1) { getline } else { num++ } ; print $num;} }' \ + | tr -d "*" +} + t32_test() { local tarball=$1 local writeconf=$2 @@ -1686,11 +1835,17 @@ t32_test() { local dom_upgrade=${dom_upgrade:-"no"} local project_upgrade=${project_upgrade:-"no"} local ff_convert=${ff_convert:-"no"} + local pfl_upgrade=${pfl_upgrade:-"no"} + local project_quota_upgrade=${project_quota_upgrade:-"no"} + local dom_new_upgrade=${dom_new_upgrade:-"no"} + local flr_upgrade=${flr_upgrade:-"no"} local shall_cleanup_mdt=false local shall_cleanup_mdt1=false local shall_cleanup_ost=false + local shall_cleanup_ost1=false local shall_cleanup_lustre=false local mdt2_is_available=false + local ost2_is_available=false local node=$(facet_active_host $SINGLEMDS) local r="do_node $node" local tmp=$TMP/t32 @@ -1711,9 +1866,14 @@ t32_test() { local mdt_dev=$tmp/mdt local mdt2_dev=$tmp/mdt2 local ost_dev=$tmp/ost + local ost2_dev=$tmp/ost2 local stripe_index local stripe_count local dir + local pfl_file=$tmp/mnt/lustre/remote_dir/pfl_dir/pfl_file + local flr_file=$tmp/mnt/lustre/remote_dir/flr_dir/flr_file + local dom_file=$tmp/mnt/lustre/remote_dir/dom_dir/dom_file + local quota_dir=$tmp/mnt/lustre/remote_dir/project_quota_dir combined_mgs_mds || stop_mgs || error "Unable to stop MGS" trap 'trap - RETURN; t32_test_cleanup' RETURN @@ -1722,7 +1882,7 @@ t32_test() { nid=$($r $LCTL list_nids | head -1) mkdir -p $tmp/mnt/lustre || error "mkdir $tmp/mnt/lustre failed" - $r mkdir -p $tmp/mnt/{mdt,mdt1,ost} + $r mkdir -p $tmp/mnt/{mdt,mdt1,ost,ost1} $r tar xjvf $tarball -S -C $tmp || { error_noexit "Unpacking the disk image tarball" return 1 @@ -1755,8 +1915,10 @@ t32_test() { ff_convert="no" ! $r test -f $mdt2_dev || mdt2_is_available=true + ! $r test -f $ost2_dev || ost2_is_available=true if [[ "$mds1_FSTYPE" == zfs ]]; then + echo "== fstype is zfs ==" # import pool first local poolname local poolname_list="t32fs-mdt1 t32fs-ost1" @@ -1776,12 +1938,14 @@ t32_test() { mdt_dev=t32fs-mdt1/mdt1 ost_dev=t32fs-ost1/ost1 ! $mdt2_is_available || mdt2_dev=t32fs-mdt2/mdt2 + ! $ost2_is_available || ost2_dev=t32fs-ost2/ost2 wait_update_facet $SINGLEMDS "$ZPOOL list | awk '/^t32fs-mdt1/ { print \\\$1 }'" "t32fs-mdt1" || { error_noexit "import zfs pool failed" return 1 } elif [ "$project_upgrade" != "no" ]; then + echo "== project upgrade ==" ! $r tune2fs -O project $mdt_dev && error_noexit "enable project on mdt0 failed" && return 1 @@ -1789,7 +1953,10 @@ t32_test() { error_noexit "enable project on mdt failed" && return 1 ! $r tune2fs -O project $ost_dev && - error_noexit "enable project on mdt failed" && + error_noexit "enable project on ost failed" && + return 1 + $ost2_is_available && ! $r tune2fs -O project $ost2_dev && + error_noexit "enable project on ost2 failed" && return 1 fi @@ -1810,6 +1977,7 @@ t32_test() { fi if [ "$writeconf" ]; then + echo "== writeconf mdt tunefs and quota ==" mopts=writeconf if [ "$mds1_FSTYPE" == ldiskfs ]; then mopts="loop,$mopts" @@ -1818,32 +1986,58 @@ t32_test() { error_noexit "Enable mdt quota feature" return 1 } + + if [ "$project_quota_upgrade" != "no" ]; then + echo "== enable mdt2 project quota ==" + $r $TUNE2FS -O project $mdt_dev || { + $r losetup -a + error_noexit "tune2fs $mdt_dev failed" + return 1 + } + fi + if $mdt2_is_available; then $r $TUNEFS --quota $mdt2_dev || { $r losetup -a - error_noexit "Enable mdt quota feature" + error_noexit "Enable mdt2 quota feature" return 1 } + if [ "$project_quota_upgrade" != "no" ]; then + echo "== enable mdt2 project quota ==" + $r $TUNE2FS -O project $mdt2_dev || { + $r losetup -a + error_noexit \ + "tune2fs $mdt2_dev failed" + return 1 + } + fi fi fi else + echo "== replace nids ==" if [ -n "$($LCTL list_nids | grep -v '\(tcp\|lo\)[[:digit:]]*$')" ]; then [[ "$MGS_VERSION" -ge $(version_code 2.3.59) ]] || skip "LU-2200: Cannot run over IB w/o lctl replace_nids " "(Need MGS version at least 2.3.59)" + fi - local osthost=$(facet_active_host ost1) - local ostnid=$(do_node $osthost $LCTL list_nids | head -1) + local osthost=$(facet_active_host ost1) + local ostnid=$(do_node $osthost $LCTL list_nids | head -1) - mopts=nosvc - if [ "$mds1_FSTYPE" == ldiskfs ]; then - mopts="loop,$mopts" - fi - $r $MOUNT_CMD -o $mopts $mdt_dev $tmp/mnt/mdt - $r $LCTL replace_nids $fsname-OST0000 $ostnid - $r $LCTL replace_nids $fsname-MDT0000 $nid - $r $UMOUNT $tmp/mnt/mdt + mopts=nosvc + if [ "$mds1_FSTYPE" == ldiskfs ]; then + mopts="loop,$mopts" + fi + $r $MOUNT_CMD -o $mopts $mdt_dev $tmp/mnt/mdt + $r $LCTL replace_nids $fsname-OST0000 $ostnid + if $ost2_is_available; then + $r $LCTL replace_nids $fsname-OST0001 $ostnid + fi + $r $LCTL replace_nids $fsname-MDT0000 $nid + if $mdt2_is_available; then + $r $LCTL replace_nids $fsname-MDT0001 $nid fi + $r $UMOUNT $tmp/mnt/mdt mopts=exclude=$fsname-OST0000 if [ "$mds1_FSTYPE" == ldiskfs ]; then @@ -1861,29 +2055,33 @@ t32_test() { shall_cleanup_mdt=true if $mdt2_is_available; then + echo "== mdt2 available ==" mopts=mgsnode=$nid,$mopts $r $MOUNT_CMD -o $mopts $mdt2_dev $tmp/mnt/mdt1 || { $r losetup -a error_noexit "Mounting the MDT" return 1 } + shall_cleanup_mdt1=true + echo "== mount new MDT $mdt2_dev ==" - echo "mount new MDT....$mdt2_dev" - $r $LCTL set_param -n mdt.${fsname}*.enable_remote_dir=1 || + $r $LCTL set_param -n mdt.${fsname}*.enable_remote_dir=1 || { error_noexit "enable remote dir create failed" + return 1 + } - shall_cleanup_mdt1=true elif [ "$dne_upgrade" != "no" ]; then local fs2mdsdev=$(mdsdevname 1_2) local fs2mdsvdev=$(mdsvdevname 1_2) - echo "mkfs new MDT on ${fs2mdsdev}...." + echo "== mkfs new MDT on ${fs2mdsdev} ==" if [ "$mds1_FSTYPE" == ldiskfs ]; then mkfsoptions="--mkfsoptions=\\\"-J size=8\\\"" fi - add $SINGLEMDS $(mkfs_opts mds2 $fs2mdsdev $fsname) --reformat \ - $mkfsoptions $fs2mdsdev $fs2mdsvdev > /dev/null || { + add $SINGLEMDS $(mds2failover_HOST="" \ + mkfs_opts mds2 $fs2mdsdev $fsname) --reformat \ + $mkfsoptions $fs2mdsdev $fs2mdsvdev > /dev/null || { error_noexit "Mkfs new MDT failed" return 1 } @@ -1895,7 +2093,7 @@ t32_test() { return 1 } - echo "mount new MDT....$fs2mdsdev" + echo "== mount new MDT....$fs2mdsdev ==" $r $MOUNT_CMD -o $mopts $fs2mdsdev $tmp/mnt/mdt1 || { error_noexit "mount mdt1 failed" return 1 @@ -1920,15 +2118,50 @@ t32_test() { error_noexit "tunefs.lustre before mounting the OST" return 1 } + + if $ost2_is_available; then + $r $TUNEFS --dryrun $ost2_dev || { + error_noexit "tunefs.lustre before mounting the OST" + return 1 + } + fi + if [ "$writeconf" ]; then + echo "== writeconf and ost quota tunefs ==" mopts=mgsnode=$nid,$writeconf - if [ "$mds1_FSTYPE" == ldiskfs ]; then + if [ "$ost1_FSTYPE" == ldiskfs ]; then mopts="loop,$mopts" $r $TUNEFS --quota $ost_dev || { $r losetup -a error_noexit "Enable ost quota feature" return 1 } + + if [ "$project_quota_upgrade" != "no" ]; then + echo "== enable ost project quota ==" + $r $TUNE2FS -O project $ost_dev || { + $r losetup -a + error_noexit "tune2fs $ost_dev failed" + return 1 + } + fi + + if $ost2_is_available; then + $r $TUNEFS --quota $ost2_dev || { + $r losetup -a + error_noexit "Enable ost2 quota feature" + return 1 + } + if [ "$project_quota_upgrade" != "no" ]; then + echo "== enable ost2 project quota ==" + $r $TUNE2FS -O project $ost2_dev || { + $r losetup -a + error_noexit \ + "tune2fs $ost2_dev failed" + return 1 + } + fi + fi fi else mopts=mgsnode=$nid @@ -1941,6 +2174,15 @@ t32_test() { error_noexit "Mounting the OST" return 1 } + + if $ost2_is_available; then + $r $MOUNT_CMD -onomgs -o$mopts $ost2_dev $tmp/mnt/ost1 || { + error_noexit "Mounting the OST2" + return 1 + } + shall_cleanup_ost1=true + fi + shall_cleanup_ost=true uuid=$($r $LCTL get_param -n obdfilter.$fsname-OST0000.uuid) || { @@ -1952,7 +2194,19 @@ t32_test() { return 1 fi - if [[ $PERM_CMD == *"set_param -P"* ]]; then + if $ost2_is_available; then + uuid=$($r $LCTL get_param -n obdfilter.$fsname-OST0001.uuid) ||{ + error_noexit "Getting OST1 UUID" + return 1 + } + if [ "$uuid" != $fsname-OST0001_UUID ]; then + error_noexit "Unexpected OST1 UUID: \"$uuid\"" + return 1 + fi + fi + + if [[ $PERM_CMD =~ "set_param" ]]; then + echo "== perm cmd = '$PERM_CMD' ==" $r $PERM_CMD osc.$fsname-OST0000*.import=connection=$nid || { error_noexit "Setting OST \"failover.node\"" return 1 @@ -1978,6 +2232,7 @@ t32_test() { return 1 } else + echo "== perm cmd = '$PERM_CMD' ==" $r $PERM_CMD $fsname-OST0000.failover.node=$nid || { error_noexit "Setting OST \"failover.node\"" return 1 @@ -2012,6 +2267,7 @@ t32_test() { } if [ "$ff_convert" != "no" -a "$ost1_FSTYPE" == ldiskfs ]; then + echo "== ff convert ($ff_convert) ==" $r $LCTL lfsck_start -M $fsname-OST0000 || { error_noexit "Start OI scrub on OST0" return 1 @@ -2035,8 +2291,9 @@ t32_test() { } fi - if [ "$dne_upgrade" != "no" ]; then - if [[ $PERM_CMD == *"set_param -P"* ]]; then + if [[ "$dne_upgrade" != "no" ]] || $mdt2_is_available; then + echo "== dne upgrate ($dne_upgrade) or mdt2 available ==" + if [[ $PERM_CMD =~ "set_param" ]]; then $r $PERM_CMD mdc.$fsname-MDT0001*.import=connection=$nid || { error_noexit "Setting MDT1 \"failover.node\"" return 1 @@ -2067,6 +2324,7 @@ t32_test() { fi if [ "$writeconf" ]; then + echo "== writeconf and client mount ==" $MOUNT_CMD $nid:/$fsname $tmp/mnt/lustre || { error_noexit "Mounting the client" return 1 @@ -2075,13 +2333,13 @@ t32_test() { shall_cleanup_lustre=true $r $LCTL set_param debug="$PTLDEBUG" - # Leave re-enabling this to a separate patch for LU-11558 - # t32_verify_quota $SINGLEMDS $fsname $tmp/mnt/lustre || { - # error_noexit "verify quota failed" - # return 1 - #} + t32_verify_quota $SINGLEMDS $fsname $tmp/mnt/lustre || { + error_noexit "verify quota failed" + return 1 + } if $r test -f $tmp/list; then + echo "== list verification ==" # # There is not a Test Framework API to copy files to or # from a remote node. @@ -2129,11 +2387,12 @@ t32_test() { return 1 fi else - echo "list verification skipped" + echo "== list verification skipped ==" fi + #non-dom upgrade to dom if [ "$dom_upgrade" != "no" ]; then - echo "Check DoM file can be created" + echo "== dom upgrade ==" $LFS setstripe -E 1M -L mdt -E EOF $tmp/mnt/lustre/dom || { error_noexit "Verify DoM creation" return 1 @@ -2161,12 +2420,41 @@ t32_test() { } fi + #dom upgrade + #$LFS setstripe -E 1M -L mdt -E -1 -S 4M $dom_file + if [ "$dom_new_upgrade" != "no" ]; then + if ! $mdt2_is_available; then + dom_file=$tmp/mnt/lustre/dom_dir/dom_file + fi + echo "== check DoM file can be accessed ==" + [ $($LFS getstripe -I1 -L $dom_file) == "mdt" ] || { + error_noexit "Verify a DoM file" + return 1 + } + [ $(stat -c%s $dom_file) == $((2 * 1024 * 1024)) ] || { + error_noexit "DoM: bad size after write" + return 1 + } + fi + + if [ "$flr_upgrade" != "no" ]; then + if ! $mdt2_is_available; then + flr_file=$tmp/mnt/lustre/flr_dir/flr_file + fi + local mirror_count=$($LFS getstripe -N $flr_file) + echo "== check FLR file ==" + [ $mirror_count == 2 ] || { + error_noexit "FLR mirror count wrong" + return 1 + } + fi + if [ "$dne_upgrade" != "no" ]; then + echo "== dne upgrade striped_dir ==" $LFS mkdir -i 1 -c2 $tmp/mnt/lustre/striped_dir || { error_noexit "set striped dir failed" return 1 } - $LFS setdirstripe -D -c2 $tmp/mnt/lustre/striped_dir pushd $tmp/mnt/lustre @@ -2180,9 +2468,73 @@ t32_test() { popd fi + if [ "$pfl_upgrade" != "no" ]; then + local comp_size + local comp_cnt + local stripe_size + local stripe_cnt + local comp_id + + echo "== check PFL file ==" + if ! $mdt2_is_available; then + pfl_file=$tmp/mnt/lustre/pfl_dir/pfl_file + fi + comp_cnt=$($LFS getstripe --component-count $pfl_file) + [ $comp_cnt == 2 ] || { + error_noexit "wrong comp_cnt $comp_cnt" + return 1 + } + + comp_size=$($LFS getstripe -I1 -E $pfl_file) + [ $comp_size == 2097152 ] || { + error_noexit "wrong component size $comp_size" + return 1 + } + + comp_id=$($LFS getstripe -I1 -i $pfl_file) + [ $comp_id == 0 ] || { + error_noexit "wrong comp id $comp_id" + return 1 + } + + comp_id=$($LFS getstripe -I2 -i $pfl_file) + [ $comp_id -eq 1 ] || { + error_noexit "wrong comp id $comp_id" + return 1 + } + + stripe_size=$($LFS getstripe -I1 -S $pfl_file) + [ $stripe_size -eq 1048576 ] || { + error_noexit "wrong stripe size $stripe_size" + return 1 + } + + stripe_size=$($LFS getstripe -I2 -S $pfl_file) + [ $comp_size -eq 2097152 ] || { + error_noexit "wrong component size $comp_size" + return 1 + } + fi + + if [ "$project_quota_upgrade" != "no" ]; then + if ! $mdt2_is_available; then + quota_dir=$tmp/mnt/lustre/project_quota_dir + fi + local hardlimit + + echo "== check Project Quota ==" + hardlimit=$(getquota -p $T32_PRJID global 3 \ + ${tmp}/mnt/lustre) + [ $hardlimit == $T32_PROLIMIT ] || { + error_noexit "wrong hardlimit $hardlimit" + return 1 + } + fi + # If it is upgrade from DNE (2.5), then rename the remote dir, # which is created in 2.5 to striped dir. if $mdt2_is_available && [[ "$dne_upgrade" != "no" ]]; then + echo "== rename remote_dir ==" stripe_index=$($LFS getdirstripe -i \ $tmp/mnt/lustre/remote_dir) @@ -2201,6 +2553,7 @@ t32_test() { # If it is upgraded from DNE (2.7), then move the striped dir # which was created in 2.7 to the new striped dir. if $mdt2_is_available && [[ "$dne_upgrade" != "no" ]] && + echo "== rename striped dir ==" [[ -d $tmp/mnt/lustre/striped_dir_old ]]; then stripe_count=$($LFS getdirstripe -c \ $tmp/mnt/lustre/striped_dir_old) @@ -2228,24 +2581,40 @@ t32_test() { } if $r test -f $tmp/sha1sums; then + echo "== checking sha1sums ==" # LU-2393 - do both sorts on same node to ensure locale # is identical $r cat $tmp/sha1sums | sort -k 2 >$tmp/sha1sums.orig - if [ "$dne_upgrade" != "no" ]; then - pushd $tmp/mnt/lustre/striped_dir + if [[ "$dne_upgrade" != "no" ]]; then + if [[ -d $tmp/mnt/lustre/striped_dir/remote_dir ]]; then + pushd $tmp/mnt/lustre/striped_dir/remote_dir + else + pushd $tmp/mnt/lustre/striped_dir + fi + elif [[ "$pfl_upgrade" != "no" ]] || + [[ "$flr_upgrade" != "no" ]] || + [[ "$dom_new_upgrade" != "no" ]] || + [[ "$project_quota_upgrade" != "no" ]] || + [[ -d $tmp/mnt/lustre/remote_dir ]]; then + pushd $tmp/mnt/lustre/remote_dir else pushd $tmp/mnt/lustre fi find ! -path "*remote_dir*" ! -path "*striped_dir*" \ - ! -name .lustre -type f -exec sha1sum {} \; | + ! -name .lustre -type f -exec sha1sum {} \; | sort -k 2 >$tmp/sha1sums || { popd error_noexit "sha1sum" return 1 } + local save=$PWD + popd if ! diff -ub $tmp/sha1sums.orig $tmp/sha1sums; then + echo "==** find $save files start **==" + find $save -type f -ls + echo "==** find $save files finish **==" error_noexit "sha1sum verification failed" return 1 fi @@ -2254,6 +2623,7 @@ t32_test() { # if upgrade from DNE(2.7), then check striped directory if $mdt2_is_available && [[ "$dne_upgrade" != "no" ]]; then + echo "== mdt2 upgrade sha1sum check ==" local new_dir="$tmp/mnt/lustre/striped_dir" local striped_dir_old="$new_dir/striped_dir_old" @@ -2262,10 +2632,12 @@ t32_test() { dir_list+=" $striped_dir_old" for dir in $dir_list; do + echo "== dir = $dir ==" pushd $dir - find ! -name .lustre -type f \ + find ! -name .lustre -type f \ -exec sha1sum {} \; | sort -k 2 >$tmp/sha1sums || { + ls -alR popd error_noexit "sha1sum" return 1 @@ -2273,6 +2645,7 @@ t32_test() { popd if ! diff -ub $tmp/sha1sums.orig \ $tmp/sha1sums; then + ls -alR $dir error_noexit "sha1sum $dir" \ "failed" return 1 @@ -2280,7 +2653,20 @@ t32_test() { done fi else - echo "sha1sum verification skipped" + echo "== sha1sum verification skipped ==" + fi + + # PFL write test after sha1sum check + if [ "$pfl_upgrade" != "no" ]; then + local rw_len=$((3 * 1034 * 1024)) + + if ! $mdt2_is_available; then + pfl_file=$tmp/mnt/lustre/pfl_dir/pfl_file + fi + small_write $pfl_file $rw_len || { + error_noexit "PFL RW Failed" + return 1 + } fi if [ "$dne_upgrade" != "no" ]; then @@ -2296,7 +2682,7 @@ t32_test() { $r $LCTL set_param -n \ mdt.${fsname}*.enable_remote_dir=1 2>/dev/null - echo "test migration" + echo "== test migration ==" pushd $tmp/mnt/lustre for dir in $(find ! -name .lustre ! -name . -type d); do mdt_index=$($LFS getdirstripe -i $dir) @@ -2395,8 +2781,19 @@ t32_test() { error_noexit "Unmounting the OST" return 1 } - if [[ "$mds1_FSTYPE" == zfs ]]; then - $r "$ZPOOL export t32fs-ost1" + if $ost2_is_available; then + $r $UMOUNT $tmp/mnt/ost1 || { + error_noexit "Unmounting the OST1" + return 1 + } + shall_cleanup_ost1=false + fi + + if [[ $ost1_FSTYPE == zfs ]]; then + $r "$ZPOOL export t32fs-ost1" + if $ost2_is_available; then + $r "$ZPOOL export t32fs-ost2" + fi fi shall_cleanup_ost=false @@ -2468,12 +2865,8 @@ test_32c() { local tarball local rc=0 - [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" t32_check for tarball in $tarballs; do - # Do not support 1_8 and 2_1 direct upgrade to DNE2 anymore */ - [[ "$tarball" =~ "1_8" ]] && echo "skip $tarball" && continue - [[ "$tarball" =~ "2_1" ]] && echo "skip $tarball" && continue banner "testing $tarball upgrade with DNE" load_modules dne_upgrade=yes t32_test $tarball writeconf || rc=$? @@ -2489,7 +2882,7 @@ test_32d() { t32_check for tarball in $tarballs; do - banner "testing $tarball upgrade with ff convert and project upgrade" + banner "testing $tarball upgrade with ff and project upgrade" project_upgrade="no" [[ "$MDS1_VERSION" -ge $(version_code 2.13.54) ]] && [[ "$tarball" =~ "disk2_4-ldiskfs" ]] && @@ -2511,7 +2904,8 @@ test_32e() { t32_check for tarball in $tarballs; do - [[ "$tarball" =~ "2_9" ]] || continue + [[ "$tarball" =~ "2_9" ]] || + { echo "skip $(basename $tarball)"; continue; } #load_modules banner "testing $tarball upgrade with DoM" dom_upgrade=yes t32_test $tarball writeconf || let "rc += $?" @@ -2520,8 +2914,47 @@ test_32e() { } run_test 32e "dom upgrade test" +test_32f() { + [[ $MDS1_VERSION -ge $(version_code 2.10.56) ]] || + skip "Need MDS version at least 2.10.56" + + local tarballs + local tarball + local rc=0 + + t32_check + for tarball in $tarballs; do + [[ "$tarball" =~ "2_10" ]] || + { echo "skip $(basename $tarball)"; continue; } + pfl_upgrade=yes project_quota_upgrade=yes \ + t32_test $tarball writeconf || let "rc += $?" + done + return $rc +} +run_test 32f "pfl upgrade test" + +test_32g() { + [[ $MDS1_VERSION -ge $(version_code 2.10.56) ]] || + skip "Need MDS version at least 2.10.56" + + local tarballs + local tarball + local rc=0 + + t32_check + for tarball in $tarballs; do + [[ $tarball =~ "2_12" ]] || + { echo "skip $(basename $tarball)"; continue; } + flr_upgrade=yes dom_new_upgrade=yes \ + t32_test $tarball writeconf || let "rc += $?" + done + return $rc +} +run_test 32g "flr/dom upgrade test" + test_33a() { # bug 12333, was test_33 - local FSNAME2=test-123 + local FSNAME2=test-$testnum + local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/}) local mkfsoptions @@ -2530,6 +2963,7 @@ test_33a() { # bug 12333, was test_33 if [ -z "$fs2ost_DEV" -o -z "$fs2mds_DEV" ]; then local dev=${SINGLEMDS}_dev local MDSDEV=${!dev} + is_blkdev $SINGLEMDS $MDSDEV && skip_env "mixed loopback and real device not working" fi @@ -2551,7 +2985,7 @@ test_33a() { # bug 12333, was test_33 --reformat $mgs_flag $mkfsoptions $fs2mdsdev $fs2mdsvdev || exit 10 add fs2ost $(mkfs_opts ost1 ${fs2ostdev}) --mgsnode=$MGSNID \ - --fsname=${FSNAME2} --index=8191 --reformat $fs2ostdev \ + --fsname=${FSNAME2} --index=0x1fff --reformat $fs2ostdev \ $fs2ostvdev || exit 10 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS && trap cleanup_fs2 EXIT INT @@ -2573,6 +3007,15 @@ test_33a() { # bug 12333, was test_33 error "$LFS getstripe $MOUNT2/hosts failed" umount $MOUNT2 + + # test lctl del_ost on large index + do_facet mgs "$LCTL del_ost -t ${FSNAME2}-OST1fff" || + error "del_ost failed with $?" + $MOUNT_CMD $MGSNID:/${FSNAME2} $MOUNT2 || error "$MOUNT_CMD failed" + echo "ok." + $LFS df | grep -q OST1fff && error "del_ost did not remove OST1fff!" + umount $MOUNT2 + stop fs2ost -f stop fs2mds -f cleanup_nocli || error "cleanup_nocli failed with $?" @@ -2580,10 +3023,10 @@ test_33a() { # bug 12333, was test_33 run_test 33a "Mount ost with a large index number" test_33b() { # was test_34 - setup + setup - do_facet client dd if=/dev/zero of=$MOUNT/24 bs=1024k count=1 - # Drop lock cancelation reply during umount + do_facet client dd if=/dev/zero of=$MOUNT/24 bs=1024k count=1 + # Drop lock cancelation reply during umount #define OBD_FAIL_LDLM_CANCEL_NET 0x304 do_facet client $LCTL set_param fail_loc=0x80000304 #lctl set_param debug=-1 @@ -2593,7 +3036,7 @@ test_33b() { # was test_34 run_test 33b "Drop cancel during umount" test_34a() { - setup + setup do_facet client "sh runmultiop_bg_pause $DIR/file O_c" manual_umount_client rc=$? @@ -2835,23 +3278,23 @@ test_36() { # 12743 DFAVAIL=$(echo $STRING | cut -d, -f3) DFFREE=$(($DFTOTAL - $DFUSED)) - ALLOWANCE=$((64 * $OSTCOUNT)) - - if [ $DFTOTAL -lt $(($BKTOTAL - $ALLOWANCE)) ] || - [ $DFTOTAL -gt $(($BKTOTAL + $ALLOWANCE)) ] ; then - echo "**** FAIL: df total($DFTOTAL) mismatch OST total($BKTOTAL)" - rc=1 - fi - if [ $DFFREE -lt $(($BKFREE - $ALLOWANCE)) ] || - [ $DFFREE -gt $(($BKFREE + $ALLOWANCE)) ] ; then - echo "**** FAIL: df free($DFFREE) mismatch OST free($BKFREE)" - rc=2 - fi - if [ $DFAVAIL -lt $(($BKAVAIL - $ALLOWANCE)) ] || - [ $DFAVAIL -gt $(($BKAVAIL + $ALLOWANCE)) ] ; then - echo "**** FAIL: df avail($DFAVAIL) mismatch OST avail($BKAVAIL)" - rc=3 - fi + ALLOWANCE=$((64 * $OSTCOUNT)) + + if [ $DFTOTAL -lt $(($BKTOTAL - $ALLOWANCE)) ] || + [ $DFTOTAL -gt $(($BKTOTAL + $ALLOWANCE)) ] ; then + echo "**** FAIL: df total($DFTOTAL) mismatch OST total($BKTOTAL)" + rc=1 + fi + if [ $DFFREE -lt $(($BKFREE - $ALLOWANCE)) ] || + [ $DFFREE -gt $(($BKFREE + $ALLOWANCE)) ] ; then + echo "**** FAIL: df free($DFFREE) mismatch OST free($BKFREE)" + rc=2 + fi + if [ $DFAVAIL -lt $(($BKAVAIL - $ALLOWANCE)) ] || + [ $DFAVAIL -gt $(($BKAVAIL + $ALLOWANCE)) ] ; then + echo "**** FAIL: df avail($DFAVAIL) mismatch OST avail($BKAVAIL)" + rc=3 + fi $UMOUNT $MOUNT2 stop fs3ost -f || error "unable to stop OST3" @@ -2910,6 +3353,7 @@ test_38() { # bug 14222 local COUNT=10 local SRC="/etc /bin" local FILES=$(find $SRC -type f -mtime +1 | head -n $COUNT) + log "copying $(echo $FILES | wc -w) files to $DIR/$tdir" mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed" tar cf - $FILES | tar xf - -C $DIR/$tdir || @@ -3105,6 +3549,7 @@ test_41c() { local rc2=$? wait $pid local rc=$? + do_facet mds1 "$LCTL set_param fail_loc=0x0" if [ $rc -eq 0 ] && [ $rc2 -ne 0 ]; then echo "1st MDT start succeed" @@ -3264,7 +3709,7 @@ test_43a() { echo "222" > $DIR/$tfile-rootfile || error "write 2 failed" chmod go-rw $DIR/$tfile-rootfile || error "chmod 2 faield" - mkdir $DIR/$tdir-rootdir || error "mkdir failed" + mkdir_on_mdt0 $DIR/$tdir-rootdir || error "mkdir failed" chmod go-rwx $DIR/$tdir-rootdir || error "chmod 3 failed" touch $DIR/$tdir-rootdir/tfile-1 || error "touch failed" @@ -3417,12 +3862,12 @@ test_44() { # 16317 check_mount || error "check_mount" UUID=$($LCTL get_param llite.${FSNAME}*.uuid | cut -d= -f2) STATS_FOUND=no - UUIDS=$(do_facet $SINGLEMDS "$LCTL get_param mdt.${FSNAME}*.exports.*.uuid") - for VAL in $UUIDS; do - NID=$(echo $VAL | cut -d= -f1) - CLUUID=$(echo $VAL | cut -d= -f2) - [ "$UUID" = "$CLUUID" ] && STATS_FOUND=yes && break - done + UUIDS=$(do_facet $SINGLEMDS "$LCTL get_param mdt.${FSNAME}*.exports.*.uuid") + for VAL in $UUIDS; do + NID=$(echo $VAL | cut -d= -f1) + CLUUID=$(echo $VAL | cut -d= -f2) + [ "$UUID" = "$CLUUID" ] && STATS_FOUND=yes && break + done [ "$STATS_FOUND" = "no" ] && error "stats not found for client" cleanup || error "cleanup failed with $?" } @@ -3471,7 +3916,7 @@ test_46a() { start_mds || error "unable to start MDS" #first client should see only one ost start_ost || error "Unable to start OST1" - wait_osc_import_state mds ost FULL + wait_osc_import_state mds ost FULL #start_client mount_client $MOUNT || error "mount_client $MOUNT failed" trap "cleanup_46a $OSTCOUNT" EXIT ERR @@ -3517,33 +3962,33 @@ test_47() { #17674 check_mount || error "check_mount failed" $LCTL set_param ldlm.namespaces.$FSNAME-*-*-*.lru_size=100 - local lru_size=[] - local count=0 - for ns in $($LCTL get_param ldlm.namespaces.$FSNAME-*-*-*.lru_size); do - if echo $ns | grep "MDT[[:digit:]]*"; then - continue - fi - lrs=$(echo $ns | sed 's/.*lru_size=//') - lru_size[count]=$lrs - let count=count+1 - done + local lru_size=[] + local count=0 + for ns in $($LCTL get_param ldlm.namespaces.$FSNAME-*-*-*.lru_size); do + if echo $ns | grep "MDT[[:digit:]]*"; then + continue + fi + lrs=$(echo $ns | sed 's/.*lru_size=//') + lru_size[count]=$lrs + let count=count+1 + done facet_failover ost1 facet_failover $SINGLEMDS client_up || error "client_up failed" - count=0 - for ns in $($LCTL get_param ldlm.namespaces.$FSNAME-*-*-*.lru_size); do - if echo $ns | grep "MDT[[:digit:]]*"; then - continue - fi - lrs=$(echo $ns | sed 's/.*lru_size=//') - if ! test "$lrs" -eq "${lru_size[count]}"; then - n=$(echo $ns | sed -e 's/ldlm.namespaces.//' -e 's/.lru_size=.*//') - error "$n has lost lru_size: $lrs vs. ${lru_size[count]}" - fi - let count=count+1 - done + count=0 + for ns in $($LCTL get_param ldlm.namespaces.$FSNAME-*-*-*.lru_size); do + if echo $ns | grep "MDT[[:digit:]]*"; then + continue + fi + lrs=$(echo $ns | sed 's/.*lru_size=//') + if ! test "$lrs" -eq "${lru_size[count]}"; then + n=$(echo $ns | sed -e 's/ldlm.namespaces.//' -e 's/.lru_size=.*//') + error "$n has lost lru_size: $lrs vs. ${lru_size[count]}" + fi + let count=count+1 + done cleanup || error "cleanup failed with $?" } @@ -3594,6 +4039,7 @@ test_48() { # bz-17636 LU-7473 cancel_lru_locks mdc stat $MOUNT/widestripe || error "stat $MOUNT/widestripe failed" local r_count=$(getfacl $MOUNT/widestripe | grep "user:" | wc -l) + count=$((count + 1)) # for the entry "user::rw-" [ $count -eq $r_count ] || @@ -3672,7 +4118,7 @@ run_test 49b "check PARAM_SYS_LDLM_TIMEOUT option of mkfs.lustre" lazystatfs() { # wait long enough to exceed OBD_STATFS_CACHE_SECONDS = 1 sleep 2 - # Test both statfs and lfs df and fail if either one fails + # Test both statfs and lfs df and fail if either one fails multiop_bg_pause $1 f_ RC=$? PID=$! @@ -3733,7 +4179,7 @@ test_50c() { # Wait for client to detect down OST stop_ost || error "Unable to stop OST1" - wait_osc_import_state mds ost DISCONN + wait_osc_import_state mds ost DISCONN lazystatfs $MOUNT || error "lazystatfs failed with one down server" umount_client $MOUNT || error "Unable to unmount client" @@ -3812,10 +4258,10 @@ test_50f() { start_mds || error "Unable to start mds" #first client should see only one ost start_ost || error "Unable to start OST1" - wait_osc_import_state mds ost FULL + wait_osc_import_state mds ost FULL - start_ost2 || error "Unable to start OST2" - wait_osc_import_state mds ost2 FULL + start_ost2 || error "Unable to start OST2" + wait_osc_import_state mds ost2 FULL # Wait for client to detect down OST stop_ost2 || error "Unable to stop OST2" @@ -3932,7 +4378,7 @@ test_50i() { start_ost2 || error "Unable to start OST2" mount_client $MOUNT || error "client start failed" - mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed" + mkdir_on_mdt0 $DIR/$tdir || error "mkdir $DIR/$tdir failed" if [[ $PERM_CMD == *"set_param -P"* ]]; then $PERM_CMD mdc.${FSNAME}-MDT0001-mdc-*.active=0 && @@ -4018,11 +4464,11 @@ copy_files_xattrs() do_node $node mkdir -p $dest [ $? -eq 0 ] || { error "Unable to create directory"; return 1; } - do_node $node 'tar cf - '$@' | tar xf - -C '$dest'; + do_node $node 'tar cf - '$*' | tar xf - -C '$dest'; [ \"\${PIPESTATUS[*]}\" = \"0 0\" ] || exit 1' [ $? -eq 0 ] || { error "Unable to tar files"; return 2; } - do_node $node 'getfattr -d -m "[a-z]*\\." '$@' > '$xattrs + do_node $node 'getfattr -d -m "[a-z]*\\." '$*' > '$xattrs [ $? -eq 0 ] || { error "Unable to read xattrs"; return 3; } } @@ -4038,7 +4484,7 @@ diff_files_xattrs() do_node $node mkdir -p $backup2 [ $? -eq 0 ] || { error "Unable to create directory"; return 1; } - do_node $node 'tar cf - '$@' | tar xf - -C '$backup2'; + do_node $node 'tar cf - '$*' | tar xf - -C '$backup2'; [ \"\${PIPESTATUS[*]}\" = \"0 0\" ] || exit 1' [ $? -eq 0 ] || { error "Unable to tar files to diff"; return 2; } @@ -4046,7 +4492,7 @@ diff_files_xattrs() [ $? -eq 0 ] || { error "contents differ"; return 3; } local xattrs2=${TMP}/xattrs2 - do_node $node 'getfattr -d -m "[a-z]*\\." '$@' > '$xattrs2 + do_node $node 'getfattr -d -m "[a-z]*\\." '$*' > '$xattrs2 [ $? -eq 0 ] || { error "Unable to read xattrs to diff"; return 4; } do_node $node "diff $xattrs $xattrs2" @@ -4151,30 +4597,30 @@ run_test 52 "check recovering objects from lost+found" # Arguments: service name (OST or MDT), facet (e.g., ost1, $SINGLEMDS), and a # parameter pattern prefix like 'ost.*.ost'. thread_sanity() { - local modname=$1 - local facet=$2 - local parampat=$3 - local opts=$4 + local modname=$1 + local facet=$2 + local parampat=$3 + local opts=$4 local basethr=$5 - local tmin - local tmin2 - local tmax - local tmax2 - local tstarted - local paramp - local msg="Insane $modname thread counts" + local tmin + local tmin2 + local tmax + local tmax2 + local tstarted + local paramp + local msg="Insane $modname thread counts" local ncpts=$(check_cpt_number $facet) local nthrs - shift 4 + shift 4 - check_mount || return 41 + check_mount || return 41 - # We need to expand $parampat, but it may match multiple parameters, so - # we'll pick the first one - if ! paramp=$(do_facet $facet "lctl get_param -N ${parampat}.threads_min"|head -1); then - error "Couldn't expand ${parampat}.threads_min parameter name" - return 22 - fi + # We need to expand $parampat, but it may match multiple parameters, so + # we'll pick the first one + if ! paramp=$(do_facet $facet "lctl get_param -N ${parampat}.threads_min"|head -1); then + error "Couldn't expand ${parampat}.threads_min parameter name" + return 22 + fi # Remove the .threads_min part paramp=${paramp%.threads_min} @@ -4236,6 +4682,7 @@ thread_sanity() { cleanup local oldvalue local newvalue="${opts}=$(expr $basethr \* $ncpts)" + setmodopts -a $modname "$newvalue" oldvalue setup @@ -4304,6 +4751,7 @@ run_test 54b "test llverfs and partial verify of filesystem" lov_objid_size() { local max_ost_index=$1 + echo -n $(((max_ost_index + 1) * 8)) } @@ -4364,6 +4812,12 @@ test_56a() { echo ok $LFS osts + # test instantiating PFL components with sparse index LU-15513 + mkdir -p $MOUNT/$tdir + $LFS setstripe -E 4M -c 1 -E 1G -c 4 -S4M -E eof -c -1 $MOUNT/$tdir + dd if=/dev/zero of=$MOUNT/$tdir/$tfile bs=4K count=1 seek=10k || + error "dd to second component failed" + if [[ "$MDS1_VERSION" -ge $(version_code 2.6.54) ]] || [[ "$MDS1_VERSION" -ge $(version_code 2.5.4) && "$MDS1_VERSION" -lt $(version_code 2.5.11) ]]; then @@ -4508,7 +4962,7 @@ test_57b() { run_test 57b "initial registration from servicenode should not fail" count_osts() { - do_facet mgs $LCTL get_param mgs.MGS.live.$FSNAME | grep OST | wc -l + do_facet mgs $LCTL get_param mgs.MGS.live.$FSNAME | grep OST | wc -l } test_58() { # bug 22658 @@ -4568,7 +5022,7 @@ test_59() { } run_test 59 "writeconf mount option" -test_60() { # LU-471 +test_60a() { # LU-471 if [ "$mds1_FSTYPE" != ldiskfs ]; then skip "ldiskfs only test" fi @@ -4594,9 +5048,24 @@ test_60() { # LU-471 stop_mds reformat_and_config } -run_test 60 "check mkfs.lustre --mkfsoptions -E -O options setting" +run_test 60a "check mkfs.lustre --mkfsoptions -E -O options setting" + +test_60b() { + [[ "$mds1_FSTYPE" == ldiskfs ]] || skip "ldiskfs only test" + + local features=$(do_facet $SINGLEMDS $DUMPE2FS $(mdsdevname 1) | + grep features) + [ ${PIPESTATUS[0]} -eq 0 ] || error "$DUMPE2FS $(mdsdevname 1) failed" + + echo $features + # ea_inode feature should be enabled by default for MDTs + [[ "$features" =~ "ea_inode" ]] || error "ea_inode is not set" + # large_dir feature should be enabled by default for MDTs + [[ "$features" =~ "large_dir" ]] || error "large_dir is not set" +} +run_test 60b "check mkfs.lustre MDT default features" -test_61() { # LU-80 +test_61a() { # LU-80 local lxattr=$(large_xattr_enabled) [ "$MDS1_VERSION" -ge $(version_code 2.1.53) ] || @@ -4670,7 +5139,79 @@ test_61() { # LU-80 rm -f $file cleanup || error "stopping systems failed" } -run_test 61 "large xattr" +run_test 61a "large xattr" + +test_61b() { # LU-80 + local lxattr=$(large_xattr_enabled) + + [ "$MDS1_VERSION" -ge $(version_code 2.1.53) ] || + skip "Need MDS version at least 2.1.53" + + if [ "$mds1_FSTYPE" != ldiskfs ]; then + skip "ldiskfs specific bug" + fi + + if ! large_xattr_enabled; then + lxattr=true + + for num in $(seq $MDSCOUNT); do + do_facet mds${num} $TUNE2FS -O ea_inode \ + $(mdsdevname $num) || + error "tune2fs on mds $num failed" + done + fi + + setup || error "setting up the filesystem failed" + client_up || error "starting client failed" + + local _file=$MOUNT/panda + local large_value="$(generate_string $(max_xattr_size))" + local name="trusted.big" + + touch ${_file} || error "touch ${_file} failed" + setfattr -n $name -v $large_value ${_file} || + error "saving $name on $file failed" + + MDT_DEV="${FSNAME}-MDT0000" + MDT_DEVNAME=$(mdsdevname ${SINGLEMDS//mds/}) + + stopall || error "stopping for e2fsck run" + + + ino=$(do_facet $SINGLEMDS "$DEBUGFS -R 'stat /ROOT/panda' \ + ${MDT_DEVNAME} | grep trusted.big") + ino=$(echo "${ino}" | awk '{print $2;}') + echo "large ea "${ino} + + do_facet $SINGLEMDS "$DEBUGFS -w -R \\\"ln $ino /lost+found\\\" \ + ${MDT_DEVNAME}" + + setup_noconfig || error "remounting the filesystem failed" + + do_facet $SINGLEMDS $LCTL lfsck_start -M ${MDT_DEV} -t namespace || { + error "can't start lfsck namespace" + } + + sleep 5 + wait_update_facet $SINGLEMDS "$LCTL get_param -n \ + mdd.${MDT_DEV}.lfsck_namespace | + awk '/^status/ { print \\\$2 }'" "completed" 32 || { + error "(2) unexpected status" + } + + stopall || error "stopping for e2fsck run" + for num in $(seq $MDSCOUNT); do + run_e2fsck $(facet_active_host mds$num) \ + $(mdsdevname $num) "-y" || + error "e2fsck MDT$num failed" + done + setup_noconfig || error "remounting the filesystem failed" + + # need to delete this file to avoid problems in other tests + rm -f $file + cleanup || error "stopping systems failed" +} +run_test 61b "large xattr" test_62() { if [ "$mds1_FSTYPE" != ldiskfs ]; then @@ -5011,8 +5552,9 @@ test_69() { if [ $num_create -gt 0 ]; then # Check the number of inodes available on OST0 local files=0 - local ifree=$($LFS df -i $MOUNT | awk '/OST0000/ { print $4 }') - log "On OST0, $ifree inodes available. Want $num_create." + local ifree=$($LFS df -i $MOUNT | + awk '/OST0000/ { print $4 }'; exit ${PIPESTATUS[0]}) + log "On OST0, $ifree inodes available. Want $num_create. rc=$?" $LFS setstripe -i 0 $DIR/$tdir || error "$LFS setstripe -i 0 $DIR/$tdir failed" @@ -5043,14 +5585,16 @@ test_69() { start_ost || error "OST0 restart failure" wait_osc_import_state mds ost FULL + sleep $((TIMEOUT/2)) #object recreation requires some time mount_client $MOUNT || error "mount client failed" touch $DIR/$tdir/$tfile-last || error "create file after reformat" local idx=$($LFS getstripe -i $DIR/$tdir/$tfile-last) [ $idx -ne 0 ] && error "$DIR/$tdir/$tfile-last on $idx not 0" || true - local iused=$($LFS df -i $MOUNT | awk '/OST0000/ { print $3 }') - log "On OST0, $iused used inodes" - [ $iused -ge $((ost_max_pre/2 + 1000)) ] && + local iused=$($LFS df -i $MOUNT | + awk '/OST0000/ { print $3 }'; exit ${PIPESTATUS[0]}) + log "On OST0, $iused used inodes rc=$?" + [ $iused -ge $((ost_max_pre + 1000)) ] && error "OST replacement created too many inodes; $iused" cleanup || error "cleanup failed with $?" } @@ -5996,7 +6540,8 @@ test_82a() { # LU-4665 mount_client $MOUNT || error "mount client $MOUNT failed" wait_osts_up - $LFS df $MOUNT || error "$LFS df $MOUNT failed" + $LFS df $MOUNT + check_lfs_df_ret_val $? || error "$LFS df $MOUNT failed" mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed" stack_trap "do_nodes $(comma_list $(mdts_nodes)) \ @@ -6111,7 +6656,8 @@ test_82b() { # LU-4665 mount_client $MOUNT || error "mount client $MOUNT failed" wait_osts_up - $LFS df $MOUNT || error "$LFS df $MOUNT failed" + $LFS df $MOUNT + check_lfs_df_ret_val $? || error "$LFS df $MOUNT failed" mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed" # Create a new pool and add OSTs into it. @@ -6464,12 +7010,15 @@ test_89() { # LU-7131 stopall - [ "$mds1_FSTYPE" == zfs ] && import_zpool mds1 + if [[ "$mds1_FSTYPE" == zfs ]]; then + import_zpool mds1 || return ${PIPESTATUS[0]} + fi + # Check that parameters are added correctly echo "tunefs --param $key=$val1" - do_facet mds "$TUNEFS --param $key=$val1 $mdsdev >/dev/null" || + do_facet mds1 "$TUNEFS --param $key=$val1 $mdsdev >/dev/null" || error "tunefs --param $key=$val1 failed" - params=$(do_facet mds $TUNEFS --dryrun $mdsdev) || + params=$(do_facet mds1 $TUNEFS --dryrun $mdsdev) || error "tunefs --dryrun failed" params=${params##*Parameters:} params=${params%%exiting*} @@ -6478,9 +7027,9 @@ test_89() { # LU-7131 # Check that parameters replace existing instances when added echo "tunefs --param $key=$val2" - do_facet mds "$TUNEFS --param $key=$val2 $mdsdev >/dev/null" || + do_facet mds1 "$TUNEFS --param $key=$val2 $mdsdev >/dev/null" || error "tunefs --param $key=$val2 failed" - params=$(do_facet mds $TUNEFS --dryrun $mdsdev) || + params=$(do_facet mds1 $TUNEFS --dryrun $mdsdev) || error "tunefs --dryrun failed" params=${params##*Parameters:} params=${params%%exiting*} @@ -6491,9 +7040,9 @@ test_89() { # LU-7131 # Check that a parameter is erased properly echo "tunefs --erase-param $key" - do_facet mds "$TUNEFS --erase-param $key $mdsdev >/dev/null" || + do_facet mds1 "$TUNEFS --erase-param $key $mdsdev >/dev/null" || error "tunefs --erase-param $key failed" - params=$(do_facet mds $TUNEFS --dryrun $mdsdev) || + params=$(do_facet mds1 $TUNEFS --dryrun $mdsdev) || error "tunefs --dryrun failed" params=${params##*Parameters:} params=${params%%exiting*} @@ -6501,22 +7050,25 @@ test_89() { # LU-7131 error "on-disk parameter not erased correctly via tunefs" # Check that all the parameters are erased + do_facet mds1 "$TUNEFS --param $key=$val1 $mdsdev >/dev/null" || + error "tunefs --param $key=$val1 failed" echo "tunefs --erase-params" - do_facet mds "$TUNEFS --erase-params $mdsdev >/dev/null" || + do_facet mds1 "$TUNEFS --erase-params $mdsdev >/dev/null" || error "tunefs --erase-params failed" - params=$(do_facet mds $TUNEFS --dryrun $mdsdev) || + params=$(do_facet mds1 $TUNEFS --dryrun $mdsdev) || error "tunefs --dryrun failed" params=${params##*Parameters:} params=${params%%exiting*} - [ -z $params ] || - error "all on-disk parameters not erased correctly via tunefs" + params=$(echo $params | tr ' ' '\n') + [ -z "$params" ] || + error "all on-disk parameters not erased correctly via tunefs $params" # Check the order of options --erase-params and --param echo "tunefs --param $key=$val1 --erase-params" - do_facet mds \ + do_facet mds1 \ "$TUNEFS --param $key=$val1 --erase-params $mdsdev >/dev/null"|| error "tunefs --param $key=$val1 --erase-params failed" - params=$(do_facet mds $TUNEFS --dryrun $mdsdev) || + params=$(do_facet mds1 $TUNEFS --dryrun $mdsdev) || error "tunefs --dryrun failed" params=${params##*Parameters:} params=${params%%exiting*} @@ -6936,63 +7488,33 @@ generate_ldev_conf() { local ldevconfpath=$1 local fstype= local fsldevformat="" - touch $ldevconfpath + rm -f $ldevconfpath - fstype=$(facet_fstype mgs) - if [ "$fstype" == zfs ]; then - fsldevformat="$fstype:" - else + local facets="mgs,$(get_facets OST),$(get_facets MDS)" + for facet in ${facets//,/ }; do fsldevformat="" - fi - - printf "%s\t-\t%s-MGS0000\t%s%s\n" \ - $mgs_HOST \ - $FSNAME \ - $fsldevformat \ - $(mgsdevname) > $ldevconfpath - - local mdsfo_host=$mdsfailover_HOST; - if [ -z "$mdsfo_host" ]; then - mdsfo_host="-" - fi - - for num in $(seq $MDSCOUNT); do - fstype=$(facet_fstype mds$num) - if [ "$fstype" == zfs ]; then - fsldevformat="$fstype:" - else - fsldevformat="" - fi - - printf "%s\t%s\t%s-MDT%04d\t%s%s\n" \ - $mds_HOST \ - $mdsfo_host \ - $FSNAME \ - $num \ - $fsldevformat \ - $(mdsdevname $num) >> $ldevconfpath - done - - local ostfo_host=$ostfailover_HOST; - if [ -z "$ostfo_host" ]; then - ostfo_host="-" - fi - - for num in $(seq $OSTCOUNT); do - fstype=$(facet_fstype ost$num) - if [ "$fstype" == zfs ]; then - fsldevformat="$fstype:" - else - fsldevformat="" + fstype=$(facet_fstype $facet) + [ "$fstype" = zfs ] && fsldevformat="$fstype:" + local host=$(facet_host $facet) + local fo="-" + local varfo=${facet}failover_HOST + if [ $facet == mgs ] && combined_mgs_mds; then + varfo=mds1failover_HOST fi - - printf "%s\t%s\t%s-OST%04d\t%s%s\n" \ - $ost_HOST \ - $ostfo_host \ + [ -n "${!varfo}" ] && fo=${!varfo} + local type=$(echo $facet | tr -d "[:digit:]" | \ + tr "[:lower:]" "[:upper:]" | sed s/MDS/MDT/ ) + local num=1 + [ ${facet} == mgs ] || + num=$(facet_number $facet) + printf "%s\t%s\t%s-%s%04x\t%s%s\n" \ + ${host} \ + ${fo} \ $FSNAME \ - $num \ + $type \ + $(( num - 1 )) \ $fsldevformat \ - $(ostdevname $num) >> $ldevconfpath + $(facet_device $facet) >> $ldevconfpath done echo "----- $ldevconfpath -----" @@ -7014,8 +7536,8 @@ generate_nids() { } compare_ldev_output() { - ldev_output=$1 - expected_output=$2 + local ldev_output=$1 + local expected_output=$2 sort $expected_output -o $expected_output sort $ldev_output -o $ldev_output @@ -7046,30 +7568,27 @@ test_92() { generate_nids $NIDSPATH # echo the mgs nid and compare it to environment variable MGSNID - # also, ldev.conf and nids is a server side thing, use the OSS - # hostname - local output - output=$($LDEV -c $LDEVCONFPATH -H $ost_HOST -n $NIDSPATH echo %m) - - echo "-- START OF LDEV OUTPUT --" - echo -e "$output" - echo "--- END OF LDEV OUTPUT ---" - - # ldev failed, error - if [ $? -ne 0 ]; then - rm $LDEVCONFPATH $NIDSPATH - error "ldev failed to execute!" - fi - - # need to process multiple lines because of combined MGS and MDS - echo -e $output | awk '{ print $2 }' | while read -r line ; do - if [ "$line" != "$MGSNID" ]; then - rm $LDEVCONFPATH $NIDSPATH - error "ldev failed mgs nid '$line', expected '$MGSNID'" - fi + local facets="$(get_facets OST),$(get_facets MDS),mgs" + for facet in ${facets//,/ }; do + local host=$(facet_host $facet) + local output=$($LDEV -c $LDEVCONFPATH -H $host -n $NIDSPATH echo %m) + + echo "-- START OF LDEV OUTPUT --" + echo -e "$output" + echo "--- END OF LDEV OUTPUT ---" + + [ -z "$output" ] && + error "ldev failed to execute!" + + # need to process multiple lines because of + # several targets on host + echo -e $output | awk '{ print $2 }' | while read -r line ; do + [ "$line" = "$MGSNID" ] || + error "ldev failed mgs nid '$line', \ + expected '$MGSNID'" + done done - - rm $LDEVCONFPATH $NIDSPATH + rm -f $LDEVCONFPATH $NIDSPATH } run_test 92 "ldev returns MGS NID correctly in command substitution" @@ -7128,11 +7647,11 @@ test_94() { printf "%s-MGS0000\n" $FSNAME > $EXPECTED_OUTPUT for num in $(seq $MDSCOUNT); do - printf "%s-MDT%04d\n" $FSNAME $num >> $EXPECTED_OUTPUT + printf "%s-MDT%04x\n" $FSNAME $((num - 1)) >> $EXPECTED_OUTPUT done for num in $(seq $OSTCOUNT); do - printf "%s-OST%04d\n" $FSNAME $num >> $EXPECTED_OUTPUT + printf "%s-OST%04x\n" $FSNAME $((num - 1)) >> $EXPECTED_OUTPUT done compare_ldev_output $LDEV_OUTPUT $EXPECTED_OUTPUT @@ -7260,19 +7779,13 @@ test_96() { echo "$mgs_HOST-$(facet_fstype mgs)" > $EXPECTED_OUTPUT - if [ "$mgs_HOST" == "$mds_HOST" ]; then - for num in $(seq $MDSCOUNT); do - echo "$mds_HOST-$(facet_fstype mds$num)" \ - >> $EXPECTED_OUTPUT - done - fi - - if [ "$mgs_HOST" == "$ost_HOST" ]; then - for num in $(seq $OSTCOUNT); do - echo "$ost_HOST-$(facet_fstype ost$num)" \ + local facets="$(get_facets OST),$(get_facets MDS)" + for facet in ${facets//,/ }; do + local host=$(facet_host $facet) + [ "$mgs_HOST" == "$host" ] && + echo "$host-$(facet_fstype $facet)" \ >> $EXPECTED_OUTPUT - done - fi + done compare_ldev_output $LDEV_OUTPUT $EXPECTED_OUTPUT @@ -7308,7 +7821,7 @@ test_97() { fi for num in $(seq $MDSCOUNT); do - printf "%s-MDT%04d\n" $FSNAME $num >> $EXPECTED_OUTPUT + printf "%s-MDT%04x\n" $FSNAME $((num - 1)) >> $EXPECTED_OUTPUT done compare_ldev_output $LDEV_OUTPUT $EXPECTED_OUTPUT @@ -7328,7 +7841,7 @@ test_97() { rm $EXPECTED_OUTPUT for num in $(seq $OSTCOUNT); do - printf "%s-OST%04d\n" $FSNAME $num >> $EXPECTED_OUTPUT + printf "%s-OST%04x\n" $FSNAME $((num - 1)) >> $EXPECTED_OUTPUT done compare_ldev_output $LDEV_OUTPUT $EXPECTED_OUTPUT @@ -8024,7 +8537,6 @@ test_108b() { } run_test 108b "migrate from ZFS to ldiskfs" - # # set number of permanent parameters # @@ -8473,7 +8985,7 @@ cleanup_115() { trap 0 stopall - rm -f $TMP/$tdir/lustre-mdt + do_facet mds1 rm -f $1 } test_115() { @@ -8482,72 +8994,79 @@ test_115() { fi [ -n "$FILESET" ] && skip "Not functional for FILESET set" - local dbfs_ver=$(do_facet $SINGLEMDS $DEBUGFS -V 2>&1) + local dbfs_ver=$(do_facet mds1 $DEBUGFS -V 2>&1) echo "debugfs version: $dbfs_ver" echo "$dbfs_ver" | egrep -w "1.44.3.wc1|1.44.5.wc1|1.45.2.wc1" && skip_env "This version of debugfs doesn't show inode number" - is_dm_flakey_dev $SINGLEMDS $(mdsdevname 1) && - skip "This test can not be executed on flakey dev" - - IMAGESIZE=$((3072 << 30)) # 3072 GiB + local IMAGESIZE=$((3072 << 30)) # 3072 GiB stopall + FLAKEY=false + echo "client1: " - lctl dl + $LCTL dl mount | grep lustre echo "mds1: " - do_facet mds1 "hostname; ifconfig; lctl dl; mount" + do_facet mds1 "hostname; $LCTL dl; mount" echo "ost1: " - do_facet ost1 "hostname; ifconfig; lctl dl; mount" + do_facet ost1 "hostname; $LCTL dl; mount" # We need MDT size 3072GB, because it is smallest # partition that can store 2B inodes - do_facet $SINGLEMDS "mkdir -p $TMP/$tdir" + do_facet mds1 "mkdir -p $TMP/$tdir" local mdsimgname=$TMP/$tdir/lustre-mdt - do_facet $SINGLEMDS "rm -f $mdsimgname" - do_facet $SINGLEMDS "touch $mdsimgname" - trap cleanup_115 RETURN EXIT - do_facet $SINGLEMDS "$TRUNCATE $mdsimgname $IMAGESIZE" || + + do_facet mds1 "rm -f $mdsimgname" + do_facet mds1 "touch $mdsimgname" + stack_trap "cleanup_115 $mdsimgname" EXIT + do_facet mds1 "$TRUNCATE $mdsimgname $IMAGESIZE" || skip "Backend FS doesn't support sparse files" - local mdsdev=$(do_facet $SINGLEMDS "losetup -f") - do_facet $SINGLEMDS "losetup $mdsdev $mdsimgname" + local mdsdev=$(do_facet mds1 "losetup -f") + + do_facet mds1 "losetup $mdsdev $mdsimgname" local mds_opts="$(mkfs_opts mds1 $(mdsdevname 1)) \ --mkfsoptions='-O ea_inode,^resize_inode,meta_bg \ -N 2247484000 -E lazy_itable_init' --device-size=$IMAGESIZE" add mds1 $mds_opts --mgs --reformat $mdsdev || skip_env "format large MDT failed" - opts="$(mkfs_opts ost1 $(ostdevname 1)) \ - $replace --reformat $(ostdevname 1) $(ostvdevname 1)" + + local ostdev=$(ostdevname 1) + + local opts="$(mkfs_opts ost1 $ostdev) \ + --reformat $ostdev $ostdev" add ost1 $opts || error "add ost1 failed with new params" - start $SINGLEMDS $mdsdev $MDS_MOUNT_OPTS || error "start MDS failed" + start mds1 $mdsdev $MDS_MOUNT_OPTS || error "start MDS failed" start_ost || error "start OSS failed" mount_client $MOUNT || error "mount client failed" mkdir -p $DIR/$tdir || error "mkdir $DIR/$tdir fail" - goal="/sys/fs/ldiskfs/$(basename $mdsdev)/inode_goal" -echo goal: $goal + local goal="/sys/fs/ldiskfs/$(basename $mdsdev)/inode_goal" + echo goal: $goal # 2147483648 is 0x80000000 - do_facet $SINGLEMDS "echo 2147483648 >> $goal; grep . $goal" + do_facet mds1 "echo 2147483648 >> $goal; grep . $goal" touch $DIR/$tdir/$tfile - # attrs from 1 to 15 go to block, 16th - to inode - for i in {1..16}; do + # attrs from 1 to 16 go to block, 17th - to inode + local i + + for i in {1..17}; do local nm="trusted.ea$i" setfattr -n $nm -v $(printf "xattr%0250d" $i) $DIR/$tdir/$tfile done + do_facet mds1 "$DEBUGFS -c -R 'stat ROOT/$tdir/$tfile' $mdsdev" + # inode <2147483649> trusted.ea16 (255) - local inode_num=$(do_facet $SINGLEMDS \ + local inode_num=$(do_facet mds1 \ "$DEBUGFS -c -R 'stat ROOT/$tdir/$tfile' $mdsdev" | - awk '/ea16/ { print $2 }' | + awk '/ea17/ { print $2 }' | sed -e 's/>//' -e 's//dev/null + # mount after writeconf will make "add osp" added to mdt0 config: + # 53 (224)marker 60 (flags=0x01, v2.5.1.0) lustre-MDT0001 'add osp' + # 54 (080)add_uuid nid=... 0: 1:... + # 55 (144)attach 0:lustre-MDT0001-osp-MDT0000 1:osp 2:... + # 56 (144)setup 0:lustre-MDT0001-osp-MDT0000 1:... 2:... + # 57 (136)modify_mdc_tgts add 0:lustre-MDT0000-mdtlov 1:... 2:1 3:1 + # duplicate modify_mdc_tgts caused crashes + + debug_size_save + # using larger debug_mb size to avoid lctl dk log truncation + size_mb=$((DEBUG_SIZE_SAVED * 4)) + for i in {1..3}; do + stop_mdt 2 + # though config processing stops after failed attach and setup + # it will proceed after the failed command after each writeconf + # this is the original scenario of the issue + do_facet mds2 "$TUNEFS --writeconf $(mdsdevname 2)" &>/dev/null + do_facet mds1 "$LCTL set_param debug_mb=$size_mb" + start_mdt 2 + + wait_update_facet_cond mds1 \ + "$LCTL dk | grep -c Processed.log.$FSNAME-MDT0000" \ + ">" 1 300 + done + debug_size_restore + + [[ -z "$had_config" ]] && do_facet mds1 lctl set_param debug=-config + + reformat +} +run_test 119 "writeconf on slave mdt shouldn't duplicate mdc/osp and crash" + test_120() { # LU-11130 [ "$MDSCOUNT" -lt 2 ] && skip "mdt count < 2" [ "$mds1_FSTYPE" != ldiskfs ] && @@ -8630,7 +9193,19 @@ test_120() { # LU-11130 } run_test 120 "cross-target rename should not create bad symlinks" -test_122() { +test_121(){ + stopall + start_mgsmds || error "MGS MDS Start failed" + fail mgs + stop_mds || error "Stopping MDSes failed" + #failback + start_mds + fail mgs + stop_mds || error "Stopping MDSes failed" +} +run_test 121 "failover MGS" + +test_122a() { [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" [[ "$OST1_VERSION" -ge $(version_code 2.11.53) ]] || skip "Need OST version at least 2.11.53" @@ -8649,7 +9224,70 @@ test_122() { cleanup } -run_test 122 "Check OST sequence update" +run_test 122a "Check OST sequence update" + +test_122b() { + [[ "$OST1_VERSION" -ge $(version_code 2.11.53) ]] || + skip "Need OST version at least 2.11.53" + local err + + reformat + LOAD_MODULES_REMOTE=true load_modules +#define OBD_FAIL_OFD_SET_OID 0x1e0 + do_facet ost1 $LCTL set_param fail_loc=0x00001e0 + + stack_trap cleanup EXIT + setup_noconfig + do_facet ost1 $LCTL set_param obdfilter.*.precreate_batch=256 + $LFS mkdir -i0 -c1 $DIR/$tdir || error "failed to create directory" + $LFS setstripe -i0 -c1 $DIR/$tdir || error "failed to setstripe" + do_facet ost1 $LCTL set_param fail_loc=0 + # overflow IDIF 32bit and create > OST_MAX_PRECREATE*5 + # so a new wrong sequence would differ from an original with error + #define OST_MAX_PRECREATE 20000 + local ost_max_precreate=20100 + local num_create=$(( ost_max_precreate * 5 )) + + # Check the number of inodes available on OST0 + local files=0 + local ifree=$($LFS df -i $MOUNT | + awk '/OST0000/ { print $4 }'; exit ${PIPESTATUS[0]}) + + log "On OST0, $ifree inodes available. Want $num_create. rc=$?" + + if [ $ifree -lt 10000 ]; then + files=$(( ifree - 50 )) + else + files=10000 + fi + + local j=$((num_create / files + 1)) + + for i in $(seq 1 $j); do + createmany -o $DIR/$tdir/$tfile-$i- $files || + error "createmany fail create $files files: $?" + unlinkmany $DIR/$tdir/$tfile-$i- $files || + error "unlinkmany failed unlink $files files" + done + sync + touch $DIR/$tdir/$tfile + do_facet ost1 sync + #we need a write req during recovery for ofd_seq_load + replay_barrier ost1 + dd if=/dev/urandom of=$DIR/$tdir/$tfile bs=1024k count=1 oflag=sync || + error "failed to write file" + + # OBD_FAIL_OST_CREATE_NET 0x204 + do_facet ost1 $LCTL set_param fail_loc=0x80000204 + fail ost1 + createmany -o $DIR/$tdir/file_ 100 + sync + + err=$(do_facet ost1 dmesg | tac | sed "/Recovery over/,$ d" | + grep "OST replaced or reformatted") + [ -z "$err" ] || error $err +} +run_test 122b "Check OST sequence wouldn't change when IDIF 32bit overflows" test_123aa() { remote_mgs_nodsh && skip "remote MGS with nodsh" @@ -8752,22 +9390,32 @@ test_123ad() { # LU-11566 run_test 123ad "llog_print shows all records" test_123ae() { # LU-11566 + local max + local mgs_arg="" + local log + local id + local orig + local new + local rpcs + remote_mgs_nodsh && skip "remote MGS with nodsh" [ -d $MOUNT/.lustre ] || setupall - local max=$($LCTL get_param -n osc.*-OST0000-*.max_dirty_mb | head -1) - local mgs_arg="" - + max=$($LCTL get_param -n osc.*-OST0000-*.max_dirty_mb | head -1) + pgs=$($LCTL get_param -n osc.*-OST0000-*.max_pages_per_rpc | head -1) [[ $MGS_VERSION -gt $(version_code 2.13.54) ]] || mgs_arg="--device MGS" if do_facet mgs "$LCTL help llog_cancel" 2>&1| grep -q -- --log_id; then # save one set_param -P record in case none exist - do_facet mgs $LCTL set_param -P osc.*.max_dirty_mb=$max - local log=params - local orig=$(do_facet mgs $LCTL $mgs_arg llog_print $log | - tail -1 | awk '{ print $4 }' | tr -d , ) + do_facet mgs $LCTL set_param -P osc.*.max_pages_per_rpc=$pgs + stack_trap "do_facet mgs $LCTL set_param -P -d \ + osc.*.max_pages_per_rpc" + + log=params + orig=$(do_facet mgs $LCTL $mgs_arg llog_print $log | + tail -1 | awk '{ print $4 }' | tr -d , ) do_facet mgs $LCTL set_param -P osc.*.max_dirty_mb=$max do_facet mgs $LCTL $mgs_arg llog_print $log | tail -1 | grep "parameter: osc.*.max_dirty_mb" || @@ -8775,9 +9423,8 @@ test_123ae() { # LU-11566 # - { index: 71, event: set_param, device: general, # param: osc.*.max_dirty_mb, value: 256 } - local id=$(do_facet mgs $LCTL $mgs_arg llog_print $log | - tail -1 | awk '{ print $4 }' | tr -d , ) - + id=$(do_facet mgs $LCTL $mgs_arg llog_print $log | + tail -1 | awk '{ print $4 }' | tr -d , ) do_facet mgs $LCTL $mgs_arg llog_cancel $log --log_idx=$id local new=$(do_facet mgs $LCTL $mgs_arg llog_print $log | tail -1 | awk '{ print $4 }' | tr -d , ) @@ -8788,18 +9435,25 @@ test_123ae() { # LU-11566 # test old positional parameters for a while still if [ "$MGS_VERSION" -le $(version_code 3.1.53) ]; then log=$FSNAME-client + + do_facet mgs $LCTL conf_param \ + $FSNAME-OST0000.osc.max_pages_per_rpc=$pgs + stack_trap "do_facet mgs $LCTL conf_param -d \ + $FSNAME-OST0000.osc.max_pages_per_rpc" + orig=$(do_facet mgs $LCTL --device MGS llog_print $log | tail -1 | awk '{ print $4 }' | tr -d , ) do_facet mgs $LCTL conf_param $FSNAME-OST0000.osc.max_dirty_mb=$max do_facet mgs $LCTL --device MGS llog_print $log | tail -1 | grep "parameter: osc.max_dirty_mb" || error "old conf_param wasn't stored in params log" - + do_facet mgs $LCTL --device MGS llog_print $log # - { index: 71, event: conf_param, device: testfs-OST0000-osc, # param: osc.max_dirty_mb=256 } id=$(do_facet mgs $LCTL --device MGS llog_print $log | tail -1 | awk '{ print $4 }' | tr -d , ) do_facet mgs $LCTL --device MGS llog_cancel $log $id + do_facet mgs $LCTL --device MGS llog_print $log new=$(do_facet mgs $LCTL --device MGS llog_print $log | tail -1 | awk '{ print $4 }' | tr -d , ) (( new == orig )) || @@ -8834,7 +9488,7 @@ test_123af() { #LU-13609 fi orig_clist=($(do_facet $facet $LCTL $cmd | awk '{ print $2 }')) orig_count=${#orig_clist[@]} - echo "orig_clist: ${orig_clist[@]}" + echo "orig_clist: ${orig_clist[*]}" #define OBD_FAIL_CATLIST 0x131b #fetch to llog records from the second one @@ -8842,7 +9496,7 @@ test_123af() { #LU-13609 new_clist=($(do_facet $facet $LCTL $cmd | awk '{ print $2 }')) new_count=${#new_clist[@]} - echo "new_clist: ${new_clist[@]}" + echo "new_clist: ${new_clist[*]}" [ $new_count -eq $((orig_count - 1)) ] || error "$new_count != $orig_count - 1" @@ -8857,17 +9511,104 @@ test_123af() { #LU-13609 } run_test 123af "llog_catlist can show all config files correctly" +test_123ag() { # LU-15142 + local rec + local orig_val + + remote_mgs_nodsh && skip "remote MGS with nodsh" + (( $MGS_VERSION >= $(version_code 2.14.55) )) || + skip "Need server version least 2.14.55" + + [ -d $MOUNT/.lustre ] || setup + + orig_val=$(do_facet mgs $LCTL get_param jobid_name) + stack_trap "do_facet mgs $LCTL set_param -P jobid_name=$orig_val" + + do_facet mgs $LCTL set_param -P jobid_name="TESTNAME1" + do_facet mgs $LCTL set_param -P -d jobid_name + rec=$(do_facet mgs $LCTL --device MGS llog_print params | + grep -c jobid_name) + (( rec == 0 )) || error "parameter was not deleted, check #1" + do_facet mgs $LCTL set_param -P jobid_name="TESTNAME1" + rec=$(do_facet mgs $LCTL --device MGS llog_print params | + grep -c jobid_name) + (( rec == 1)) || error "parameter is not set" + # usage with ordinary set_param format works too + do_facet mgs $LCTL set_param -P -d jobid_name="ANY" + rec=$(do_facet mgs $LCTL --device MGS llog_print params | + grep -c jobid_name) + (( rec == 0 )) || error "parameter was not deleted, check #2" +} +run_test 123ag "llog_print skips values deleted by set_param -P -d" + +test_123ah() { #LU-7668 del_ost + [ "$MGS_VERSION" -ge $(version_code 2.15.50) -a \ + "$MDS1_VERSION" -ge $(version_code 2.15.50) ] || + skip "Need both MGS and MDS version at least 2.15.50" + + [ -d $MOUNT/.lustre ] || setupall + stack_trap "do_facet mds1 $LCTL set_param fail_loc=0" EXIT + + local cmd="--device MGS llog_print" + + cli_llogcnt_orig=$(do_facet mgs $LCTL $cmd $FSNAME-client | + grep -c $FSNAME-OST0000) + mds1_llogcnt_orig=$(do_facet mgs $LCTL $cmd $FSNAME-MDT0000 | + grep -c $FSNAME-OST0000) + + [ $cli_llogcnt_orig -gt 0 ] || + error "$FSNAME-OST0000 not found (client)" + [ $mds1_llogcnt_orig -gt 0 ] || error "$FSNAME-OST0000 not found (MDT)" + + # -n/--dryrun should NOT modify catalog + do_facet mgs "$LCTL del_ost -n -t $FSNAME-OST0000" || + error "del_ost --dryrun failed with $?" + + local cli_llogcnt=$(do_facet mgs $LCTL $cmd $FSNAME-client | + grep -c $FSNAME-OST0000) + local mds1_llogcnt=$(do_facet mgs $LCTL $cmd $FSNAME-MDT0000 | + grep -c $FSNAME-OST0000) + + [ $cli_llogcnt -eq $cli_llogcnt_orig ] || + error "--dryrun error: $cli_llogcnt != $cli_llogcnt_orig" + [ $mds1_llogcnt -eq $mds1_llogcnt_orig ] || + error "--dryrun error: $mds1_llogcnt != $mds1_llogcnt_orig" + + # actual run + do_facet mgs "$LCTL del_ost --target $FSNAME-OST0000" || + error "del_ost failed with $?" + + local cli_llogcnt=$(do_facet mgs $LCTL $cmd $FSNAME-client | + grep -c $FSNAME-OST0000) + local mds1_llogcnt=$(do_facet mgs $LCTL $cmd $FSNAME-MDT0000 | + grep -c $FSNAME-OST0000) + + # every catalog entry for OST0000 should have been cancelled + [ $cli_llogcnt -eq 0 ] || error "$FSNAME-OST0000 not cancelled (cli)" + [ $mds1_llogcnt -eq 0 ] || error "$FSNAME-OST0000 not cancelled (MDT)" + + umount_client $MOUNT + mount_client $MOUNT + + $LFS df | grep -q OST0000 && error "del_ost did not remove OST0000!" + cleanup + reformat_and_config +} +run_test 123ah "del_ost cancels config log entries correctly" + test_123F() { remote_mgs_nodsh && skip "remote MGS with nodsh" [ -d $MOUNT/.lustre ] || setup local yaml_file="$TMP/$tfile.yaml" + do_facet mgs rm "$yaml_file" local cfgfiles=$(do_facet mgs "lctl --device MGS llog_catlist" | sed 's/config_log://') # set jobid_var to a different value for test local orig_val=$(do_facet mgs $LCTL get_param jobid_var) + do_facet mgs $LCTL set_param -P jobid_var="TESTNAME" for i in $cfgfiles params; do @@ -8879,16 +9620,14 @@ test_123F() { echo "Writeconf" writeconf_all echo "Remounting" - mountmgs - mountmds - mountoss - mountcli + setup_noconfig # Reapply the config from before echo "Setting configuration parameters" do_facet mgs "lctl set_param -F $yaml_file" local set_val=$(do_facet mgs $LCTL get_param jobid_var) + do_facet mgs $LCTL set_param -P $orig_val [ $set_val == "jobid_var=TESTNAME" ] || @@ -8905,6 +9644,12 @@ test_124() [ -z $mds2failover_HOST ] && skip "needs MDT failover setup" setup + do_facet mgs $LCTL --device MGS llog_print $FSNAME-client | + grep 1.2.3.4@tcp && error "Should not be fake nid" + do_facet mgs $LCTL conf_param $FSNAME-MDT0001.failover.node=1.2.3.4@tcp\ + || error "Set params error" + do_facet mgs $LCTL --device MGS llog_print $FSNAME-client | + grep 1.2.3.4@tcp || error "Fake nid should be added" cleanup load_modules @@ -8914,6 +9659,7 @@ test_124() fi local nid=$(do_facet mds2 $LCTL list_nids | head -1) local failover_nid=$(do_node $mds2failover_HOST $LCTL list_nids | head -1) + do_facet mgs $LCTL replace_nids $FSNAME-MDT0001 $nid:$failover_nid || error "replace_nids execution error" @@ -8977,6 +9723,7 @@ check_slaves_max_sectors_kb() local slave max_sectors new_max_sectors max_hw_sectors path local rc=0 + for slave in ${slave_devices}; do path="/dev/${slave}" ! is_blkdev ${facet} ${path} && continue @@ -9119,6 +9866,234 @@ test_127() { } run_test 127 "direct io overwrite on full ost" +test_128() +{ + combined_mgs_mds && skip "need separate mgs device" + [ "$ost2_FSTYPE" == zfs ] && import_zpool ost2 + + format_ost 2 + # Try to apply nolocallogs to the virgin OST. Should fail. + do_facet ost2 "$TUNEFS --nolocallogs $(ostdevname 2)" && + error "nolocallogs should not be allowed on the virgin target" + + setupall + stopall + + [ "$ost1_FSTYPE" == zfs ] && import_zpool ost1 + # Start OST without MGS (local configs) + do_facet ost1 "$TUNEFS --dryrun $(ostdevname 1)" + start_ost || error "unable to start OST1" + stop_ost || error "Unable to stop OST1" + + [ "$ost1_FSTYPE" == zfs ] && import_zpool ost1 + # Do not allow reading local configs, should fail + do_facet ost1 "$TUNEFS --nolocallogs $(ostdevname 1)" || + error "Can not set nolocallogs" + start_ost && error "OST1 started, but should fail" + + # Connect to MGS successfully, reset nolocallogs flag + [ "$ost1_FSTYPE" == zfs ] && import_zpool ost1 + start_mgs || error "unable to start MGS" + start_ost || error "unable to start OST1" + + do_facet ost1 "$TUNEFS --dryrun $(ostdevname 1)" | grep "nolocallogs" && + error "nolocallogs expected to be reset" + + stop_ost || error "Unable to stop OST1" +} +run_test 128 "Force using remote logs with --nolocallogs" + +test_129() +{ + stopall + start_mds || error "MDS start failed" + format_ost 1 + start ost1 $(ostdevname 1) $OST_MOUNT_OPTS && + error "start ost1 should fail" || true + start ost1 $(ostdevname 1) $OST_MOUNT_OPTS && + error "second start ost1 should fail" || true + do_facet ost1 "$TUNEFS --writeconf $(ostdevname 1)" + start ost1 $(ostdevname 1) $OST_MOUNT_OPTS || + error "start ost1 failed" + stop ost1 + stop_mds +} +run_test 129 "attempt to connect an OST with the same index should fail" + +test_130() +{ + [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" + setupall + stop_mdt 2 || error "mdt2 stop failed" + do_facet mds2 "$TUNEFS --writeconf $(mdsdevname 2)" + start_mdt 2 || error "mdt2 start failed" + do_facet mds2 "$LCTL dl" | grep MDT0001-osp-MDT0001 && + error "Illegal OSP device created" || true +} +run_test 130 "re-register an MDT after writeconf" + +test_131() { + [ "$mds1_FSTYPE" == "ldiskfs" ] || skip "ldiskfs only test" + do_facet mds1 $DEBUGFS -R features $(mdsdevname 1) | + grep -q project || skip "skip project quota not supported" + + local projid + + setupall + test_mkdir -c $MDSCOUNT -p $DIR/$tdir + $LFS project -p 1000 $DIR/$tdir || error "set dir project id failed" + createmany -o $DIR/$tdir/f 512 + for ((i = 0; i < 512; ++i)); do + $LFS project -p $i $DIR/$tdir/f${i} || + error "set f${i} project id failed" + done + + test_mkdir -c $MDSCOUNT -p $DIR/$tdir.inherit + $LFS project -p 1001 -s $DIR/$tdir.inherit + createmany -o $DIR/$tdir.inherit/f 128 + (( $($LFS project $DIR/$tdir.inherit/f* | + awk '$1 == 1001 { print }' | wc -l) == 128 )) || + error "files did not inherit projid 1001" + + stopall + + for i in $(seq $MDSCOUNT); do + mds_backup_restore mds$i || + error "Backup/restore on mds$i failed" + done + + setupall + + projid=($($LFS project -d $DIR/$tdir)) + [ ${projid[0]} == "1000" ] || + error "projid expected 1000 not ${projid[0]}" + for ((i = 0; i < 512; ++i)); do + projid=($($LFS project $DIR/$tdir/f${i})) + [ ${projid[0]} == "$i" ] || + error "projid expected $i not ${projid[0]}" + done + + (( $($LFS project $DIR/$tdir.inherit/f* | + awk '$1 == 1001 { print }' | wc -l) == 128 )) || + error "restore did not copy projid 1001" +} +run_test 131 "MDT backup restore with project ID" + +test_132() { + local err_cnt + local err_cnt2 + + reformat + combined_mgs_mds || start_mgs || error "unable to start MGS" + start_mdt 1 || error "unable to start mdt1" + + err_cnt=$(do_facet mds1 dmesg | grep -c "cannot take the layout locks") + stop_mdt 1 || error "stop mdt1 failed" + + [ "$mds1_FSTYPE" == zfs ] && import_zpool mds1 + do_facet mds1 $TUNEFS --param mdt.hsm_control=enabled $(mdsdevname 1) || + error "tunefs failed" + start_mdt 1 || error "cannot start mdt1" + + err_cnt2=$(do_facet mds1 dmesg | grep -c "cannot take the layout locks") + [ $err_cnt -eq $err_cnt2 ] || error "Can not take the layout lock" + stop_mdt 1 || error "stop mdt1 failed" +} +run_test 132 "hsm_actions processed after failover" + +# This test verifies we do RR allocation within a pool even if there is a +# significant imbalance vs an OST outside the pool +test_133() { + [[ $OSTCOUNT -lt 4 ]] && skip_env "needs >= 4 OSTs" + # This is the easiest way to ensure OSTs start out balanced + reformat_and_config + setupall + + check_set_fallocate_or_skip + + local testfile=$DIR/$tdir/$tfile + local pool="testpool" + local ostrange=$((OSTCOUNT - 1)) + # Select all but the last OST to add to the pool + local poolostrange=$((OSTCOUNT - 2)) + local filenum=20 + local filecount + local stripecount + + declare -a AVAIL + free_min_max + + [ $MINV -eq 0 ] && error "no free space in OST$MINI" + [ $MAXV -gt $((2 * $MINV)) ] && + error "OSTs badly unbalanced after reformat" + + create_pool $FSNAME.$pool || error "failed to create a pool" + do_facet mgs $LCTL pool_add $FSNAME.$pool OST[0-$poolostrange] || + error "failed to add OST[0-$poolostrange] to the pool" + + test_mkdir -p $DIR/$tdir || error "failed to mkdir $DIR/$tdir" + # Consume space on the OSTs in the pool so they are unbalanced with the + # OST outside of the pool + # fill each OST 90% with fallocate so they are widely + # imbalanced + local size=$(((MINV * 9 / 10) * 1024)) + for ((i = 0; i <= poolostrange; i++)); do + $LFS setstripe -c 1 -i $i $testfile$i || + error "failed to setstripe $testfile$i" + fallocate -l $size $testfile$i || error "fallocate failed" + done + ls -la $DIR/$tdir + sleep_maxage + $LFS df + + # Create files in the pool now that there is an imbalance + filecount=$(((OSTCOUNT - 1) * filenum)) + for ((i = 0; i < filecount; i++)); do + $LFS setstripe -p $pool $testfile-$i || + error "failed to setstripe -p $pool $testfile-$i" + done + $LFS getstripe -i $testfile-* > /tmp/$tfile.log + # Count the number of files with a stripe on each OST to verify the + # pool allocated with round-robin + for ((i = 0; i <= poolostrange; i++)); do + stripecount=$(grep -c $i /tmp/$tfile.log) + # Allow a little leeway + if (( stripecount < filenum - 1 || + stripecount > filenum + 1 )); then + cat /tmp/$tfile.log + error "$stripecount != $filenum files on OST$i" + fi + done + + # Create files across the system now that there is an imbalance + filecount=$((OSTCOUNT * filenum)) + for ((i = 1; i < filecount; i++)); do + $LFS setstripe $testfile-$i.2 || + error "failed to setstripe $testilfe-$i.2" + done + $LFS getstripe -i $testfile-*.2 > /tmp/$tfile.log + local qos_used="" + # Count the number of files with a stripe on each OST to verify the + # files are *NOT* allocated with round-robin + for ((i = 0; i <= ostrange; i++)); do + stripecount=$(grep -c $i /tmp/$tfile.log) + if [[ $stripecount -ne $filenum ]]; then + qos_used="true" + echo "QOS: $stripecount != $filenum files on OST$i" + fi + done + if [ -z "$qos_used" ]; then + error "QOS not used on imbalanced OSTs!" + fi + + rm -rf /tmp/$tfile.log $DIR/$tdir + do_facet mgs $LCTL pool_remove $FSNAME.$pool OST[0-$poolostrange] || + "failed to remove OST[0-$poolostrange] from the pool" + do_facet mgs $LCTL pool_destroy $FSNAME.$pool || + error "failed to destroy pool" +} +run_test 133 "stripe QOS: free space balance in a pool" + if ! combined_mgs_mds ; then stop mgs fi