X-Git-Url: https://git.whamcloud.com/?a=blobdiff_plain;f=lustre%2Ftests%2Fconf-sanity.sh;h=b2317840d11512fc49fafe6c60014beeaba19cbb;hb=8438f2a7b56fd97aa2207aac375a880df668a61d;hp=51c12d9ee29102ad0274f4bc97a9c79f9c8379d3;hpb=df89c74a320278acac7466a83393af6abd99932b;p=fs%2Flustre-release.git diff --git a/lustre/tests/conf-sanity.sh b/lustre/tests/conf-sanity.sh index 51c12d9..b231784 100644 --- a/lustre/tests/conf-sanity.sh +++ b/lustre/tests/conf-sanity.sh @@ -4,8 +4,8 @@ set -e ONLY=${ONLY:-"$*"} -# bug number for skipped test: LU-2828 -ALWAYS_EXCEPT="$CONF_SANITY_EXCEPT 59 64" +# bug number for skipped test: +ALWAYS_EXCEPT="$CONF_SANITY_EXCEPT" # UPDATE THE COMMENT ABOVE WITH BUG NUMBERS WHEN CHANGING ALWAYS_EXCEPT! is_sles11() # LU-2181 @@ -59,6 +59,14 @@ STORED_OSTSIZE=$OSTSIZE MDSSIZE=200000 OSTSIZE=200000 +fs2mds_HOST=$mds_HOST +fs2ost_HOST=$ost_HOST +fs3ost_HOST=$ost_HOST + +MDSDEV1_2=$fs2mds_DEV +OSTDEV1_2=$fs2ost_DEV +OSTDEV2_2=$fs3ost_DEV + if ! combined_mgs_mds; then # bug number for skipped test: 23954 ALWAYS_EXCEPT="$ALWAYS_EXCEPT 24b" @@ -70,8 +78,8 @@ if [[ "$LDISKFS_MKFS_OPTS" != *lazy_itable_init* ]]; then fi [ $(facet_fstype $SINGLEMDS) = "zfs" ] && -# bug number for skipped test: LU-2778 LU-4444 - ALWAYS_EXCEPT="$ALWAYS_EXCEPT 57b 69" +# bug number for skipped test: LU-4444 + ALWAYS_EXCEPT="$ALWAYS_EXCEPT 69" init_logging @@ -205,16 +213,17 @@ remount_client() { } umount_client() { - local MOUNTPATH=$1 - echo "umount lustre on ${MOUNTPATH}....." - zconf_umount $(hostname) $MOUNTPATH || return 97 + local mountpath=$1 + shift + echo "umount lustre on $mountpath....." + zconf_umount $HOSTNAME $mountpath $@ || return 97 } manual_umount_client(){ local rc local FORCE=$1 echo "manual umount lustre on ${MOUNT}...." - do_facet client "umount -d ${FORCE} $MOUNT" + do_facet client "umount ${FORCE} $MOUNT" rc=$? return $rc } @@ -249,6 +258,16 @@ cleanup() { cleanup_nocli || return $? } +cleanup_fs2() { + trap 0 + echo "umount $MOUNT2 ..." + umount $MOUNT2 || true + echo "stopping fs2mds ..." + stop fs2mds -f || true + echo "stopping fs2ost ..." + stop fs2ost -f || true +} + check_mount() { do_facet client "cp /etc/passwd $DIR/a" || return 71 do_facet client "rm $DIR/a" || return 72 @@ -303,9 +322,9 @@ test_1() { run_test 1 "start up ost twice (should return errors)" test_2() { - start_mdt 1 || error "MDT0 start fail" + start_mds || error "MDT start failed" echo "start mds second time.." - start_mdt 1 && error "2nd MDT start should fail" + start_mds && error "2nd MDT start should fail" start_ost || error "OST start failed" mount_client $MOUNT || error "mount_client failed to start client" check_mount || error "check_mount failed" @@ -326,7 +345,8 @@ test_4() { setup touch $DIR/$tfile || error "touch $DIR/$tfile failed" stop_ost || error "Unable to stop OST1" - cleanup + umount_client $MOUNT -f || error “unmount $MOUNT failed” + cleanup_nocli eno=$? # ok for ost to fail shutdown if [ 202 -ne $eno ] && [ 0 -ne $eno ]; then @@ -345,7 +365,7 @@ test_5a() { # was test_5 # cleanup may return an error from the failed # disconnects; for now I'll consider this successful # if all the modules have unloaded. - umount -d $MOUNT & + $UMOUNT -f $MOUNT & UMOUNT_PID=$! sleep 6 echo "killing umount" @@ -440,7 +460,8 @@ test_5d() { start_mds || error "MDS start failed" stop_ost || error "Unable to stop OST1" mount_client $MOUNT || error "mount_client $MOUNT failed" - cleanup || error "cleanup_nocli failed with $?" + umount_client $MOUNT -f || error "umount_client $MOUNT failed" + cleanup_nocli || error "cleanup_nocli failed with $?" grep " $MOUNT " /etc/mtab && error "$MOUNT entry in mtab after unmount" pass @@ -759,6 +780,58 @@ test_21d() { } run_test 21d "start mgs then ost and then mds" +cleanup_21e() { + MGSNID="$saved_mgsnid" + cleanup_fs2 + echo "stopping fs2mgs ..." + stop $fs2mgs -f || true +} + +test_21e() { # LU-5863 + if [[ -z "$fs3ost_DEV" || -z "$fs2ost_DEV" || -z "$fs2mds_DEV" ]]; then + is_blkdev $SINGLEMDS $(mdsdevname ${SINGLEMDS//mds/}) && + skip_env "mixed loopback and real device not working" && return + fi + + local fs2mdsdev=$(mdsdevname 1_2) + local fs2ostdev=$(ostdevname 1_2) + local fs3ostdev=$(ostdevname 2_2) + + local fs2mdsvdev=$(mdsvdevname 1_2) + local fs2ostvdev=$(ostvdevname 1_2) + local fs3ostvdev=$(ostvdevname 2_2) + + # temporarily use fs3ost as fs2mgs + local fs2mgs=fs3ost + local fs2mgsdev=$fs3ostdev + local fs2mgsvdev=$fs3ostvdev + + local fsname=test1234 + + add $fs2mgs $(mkfs_opts mgs $fs2mgsdev) --fsname=$fsname \ + --reformat $fs2mgsdev $fs2mgsvdev || error "add fs2mgs failed" + start $fs2mgs $fs2mgsdev $MGS_MOUNT_OPTS && trap cleanup_21e EXIT INT || + error "start fs2mgs failed" + + local saved_mgsnid="$MGSNID" + MGSNID=$(do_facet $fs2mgs $LCTL list_nids | xargs | tr ' ' ,) + + add fs2mds $(mkfs_opts mds1 $fs2mdsdev $fsname) \ + --reformat $fs2mdsdev $fs2mdsvdev || error "add fs2mds failed" + add fs2ost $(mkfs_opts ost1 $fs2ostdev $fsname) \ + --reformat $fs2ostdev $fs2ostvdev || error "add fs2ost failed" + + start fs2ost $fs2ostdev $OST_MOUNT_OPTS || error "start fs2ost failed" + start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS || error "start fs2mds failed" + + mkdir -p $MOUNT2 || error "mkdir $MOUNT2 failed" + $MOUNT_CMD $MGSNID:/$fsname $MOUNT2 || error "mount $MOUNT2 failed" + DIR=$MOUNT2 MOUNT=$MOUNT2 check_mount || error "check $MOUNT2 failed" + + cleanup_21e +} +run_test 21e "separate MGS and MDS" + test_22() { start_mds || error "MDS start failed" @@ -771,7 +844,7 @@ test_22() { # check_mount will block trying to contact ost mcreate $DIR/$tfile || error "mcreate $DIR/$tfile failed" rm -f $DIR/$tfile || error "remove $DIR/$tfile failed" - umount_client $MOUNT + umount_client $MOUNT -f pass echo "Client mount with a running ost" @@ -862,23 +935,6 @@ test_23b() { # was test_23 } run_test 23b "Simulate -EINTR during mount" -fs2mds_HOST=$mds_HOST -fs2ost_HOST=$ost_HOST - -MDSDEV1_2=$fs2mds_DEV -OSTDEV1_2=$fs2ost_DEV -OSTDEV2_2=$fs3ost_DEV - -cleanup_fs2() { - trap 0 - echo "umount $MOUNT2 ..." - umount $MOUNT2 || true - echo "stopping fs2mds ..." - stop fs2mds -f || true - echo "stopping fs2ost ..." - stop fs2ost -f || true -} - test_24a() { local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/}) @@ -1099,7 +1155,7 @@ test_28a() { # LU-4221 # Check 3. # prepare a non-symlink parameter in the OSD - name="lma_self_repair" + name="auto_scrub" param="$device.osd.$name" cmd="$LCTL get_param -n osd-*.$device.$name" @@ -1221,30 +1277,35 @@ run_test 30a "Big config llog and conf_param deletion" test_30b() { setup + local orignids=$($LCTL get_param -n \ + osc.$FSNAME-OST0000-osc-[^M]*.import | grep failover_nids) + + local orignidcount=$(echo "$orignids" | wc -w) + # Make a fake nid. Use the OST nid, and add 20 to the least significant # numerical part of it. Hopefully that's not already a failover address # for the server. - OSTNID=$(do_facet ost1 "$LCTL get_param nis" | tail -1 | awk '{print $1}') - ORIGVAL=$(echo $OSTNID | egrep -oi "[0-9]*@") - NEWVAL=$((($(echo $ORIGVAL | egrep -oi "[0-9]*") + 20) % 256)) - NEW=$(echo $OSTNID | sed "s/$ORIGVAL/$NEWVAL@/") + local OSTNID=$(do_facet ost1 "$LCTL get_param nis" | tail -1 | \ + awk '{print $1}') + local ORIGVAL=$(echo $OSTNID | egrep -oi "[0-9]*@") + local NEWVAL=$((($(echo $ORIGVAL | egrep -oi "[0-9]*") + 20) % 256)) + local NEW=$(echo $OSTNID | sed "s/$ORIGVAL/$NEWVAL@/") echo "Using fake nid $NEW" - TEST="$LCTL get_param -n osc.$FSNAME-OST0000-osc-[^M]*.import | + local TEST="$LCTL get_param -n osc.$FSNAME-OST0000-osc-[^M]*.import | grep failover_nids | sed -n 's/.*\($NEW\).*/\1/p'" set_conf_param_and_check client "$TEST" \ "$FSNAME-OST0000.failover.node" $NEW || error "didn't add failover nid $NEW" - NIDS=$($LCTL get_param -n osc.$FSNAME-OST0000-osc-[^M]*.import | + local NIDS=$($LCTL get_param -n osc.$FSNAME-OST0000-osc-[^M]*.import | grep failover_nids) echo $NIDS - # The NIDS value is the failover nid strings and "[" and "]". So - # we need to subtract the space taken by the delimiters. This has - # changed from earlier version of Lustre but this test is run only - # locally so this change will not break interop. See LU-3386 - NIDCOUNT=$(($(echo "$NIDS" | wc -w) - 3)) - echo "should have 2 failover nids: $NIDCOUNT" - [ $NIDCOUNT -eq 2 ] || error "Failover nid not added" + local NIDCOUNT=$(echo "$NIDS" | wc -w) + echo "should have $((orignidcount + 1)) entries \ + in failover nids string, have $NIDCOUNT" + [ $NIDCOUNT -eq $((orignidcount + 1)) ] || + error "Failover nid not added" + do_facet mgs "$LCTL conf_param -d $FSNAME-OST0000.failover.node" || error "conf_param delete failed" umount_client $MOUNT @@ -1253,9 +1314,10 @@ test_30b() { NIDS=$($LCTL get_param -n osc.$FSNAME-OST0000-osc-[^M]*.import | grep failover_nids) echo $NIDS - NIDCOUNT=$(($(echo "$NIDS" | wc -w) - 3)) - echo "only 1 final nid should remain: $NIDCOUNT" - [ $NIDCOUNT -eq 1 ] || error "Failover nids not removed" + NIDCOUNT=$(echo "$NIDS" | wc -w) + echo "only $orignidcount final entries should remain \ + in failover nids string, have $NIDCOUNT" + [ $NIDCOUNT -eq $orignidcount ] || error "Failover nids not removed" cleanup || error "cleanup failed with rc $?" } @@ -1288,12 +1350,13 @@ test_32newtarball() { local src=/etc/rc.d local tmp=$TMP/t32_image_create - if [ $FSNAME != t32fs -o $MDSCOUNT -ne 1 -o \ - \( -z "$MDSDEV" -a -z "$MDSDEV1" \) -o $OSTCOUNT -ne 1 -o \ - -z "$OSTDEV1" ]; then - error "Needs FSNAME=t32fs MDSCOUNT=1 MDSDEV1=" \ - "(or MDSDEV, in the case of b1_8) OSTCOUNT=1" \ - "OSTDEV1=" + 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=" fi mkdir $tmp || { @@ -1326,7 +1389,7 @@ test_32newtarball() { setupall pushd /mnt/$FSNAME ls -Rni --time-style=+%s >$tmp/img/list - find . ! -name .lustre -type f -exec sha1sum {} \; | + find ! -name .lustre -type f -exec sha1sum {} \; | sort -k 2 >$tmp/img/sha1sums popd $LCTL get_param -n version | head -n 1 | @@ -1344,6 +1407,11 @@ test_32newtarball() { { if (NF == 1) { getline } else { num++ } ; print $num;} }' \ | tr -d "*" > $tmp/img/ispace + if [ $MDSCOUNT -ge 2 ]; then + $LFS mkdir -i 1 /mnt/$FSNAME/remote_dir + tar cf - -C $tmp/src . | tar xf - -C /mnt/$FSNAME/remote_dir + fi + stopall pushd $tmp/src @@ -1358,6 +1426,10 @@ test_32newtarball() { uname -m >$tmp/img/arch mv ${MDSDEV1:-$MDSDEV} $tmp/img + for num in $(seq 2 $MDSCOUNT); do + local devname=$(mdsdevname $num) + mv $devname $tmp/img + done mv $OSTDEV1 $tmp/img version=$(sed -e 's/\(^[0-9]\+\.[0-9]\+\)\(.*$\)/\1/' $tmp/img/commit | @@ -1408,13 +1480,13 @@ t32_test_cleanup() { umount $tmp/mnt/lustre || rc=$? fi if $shall_cleanup_mdt; then - $r umount -d $tmp/mnt/mdt || rc=$? + $r $UMOUNT $tmp/mnt/mdt || rc=$? fi if $shall_cleanup_mdt1; then - $r umount -d $tmp/mnt/mdt1 || rc=$? + $r $UMOUNT $tmp/mnt/mdt1 || rc=$? fi if $shall_cleanup_ost; then - $r umount -d $tmp/mnt/ost || rc=$? + $r $UMOUNT $tmp/mnt/ost || rc=$? fi $r rm -rf $tmp @@ -1475,7 +1547,7 @@ t32_wait_til_devices_gone() { done echo "waiting for dev on $node: dev $devices loop $loops given up" do_rpc_nodes $node "losetup -a" - do_rpc_nodes $node "$LCTL devices_list" + do_rpc_nodes $node "$LCTL device_list" return 1 } @@ -1586,12 +1658,16 @@ t32_test() { local list local fstype=$(facet_fstype $SINGLEMDS) local mdt_dev=$tmp/mdt + local mdt2_dev=$tmp/mdt2 local ost_dev=$tmp/ost + local stripe_index + local dir trap 'trap - RETURN; t32_test_cleanup' RETURN + load_modules mkdir -p $tmp/mnt/lustre || error "mkdir $tmp/mnt/lustre failed" - $r mkdir -p $tmp/mnt/{mdt,ost} + $r mkdir -p $tmp/mnt/{mdt,mdt1,ost} $r tar xjvf $tarball -S -C $tmp || { error_noexit "Unpacking the disk image tarball" return 1 @@ -1605,7 +1681,7 @@ t32_test() { echo " Commit: $img_commit" echo " Kernel: $img_kernel" echo " Arch: $img_arch" - echo "OST version: $(get_lustre_version ost1)" + echo "OST version: $(lustre_build_version ost1)" # The conversion can be made only when both of the following # conditions are satisfied: @@ -1635,6 +1711,15 @@ t32_test() { error_noexit "tunefs.lustre before mounting the MDT" return 1 } + + if $r test -f $mdt2_dev; then + $r $TUNEFS --dryrun $mdt2_dev || { + $r losetup -a + error_noexit "tunefs.lustre before mounting the MDT" + return 1 + } + fi + if [ "$writeconf" ]; then mopts=writeconf if [ $fstype == "ldiskfs" ]; then @@ -1644,6 +1729,13 @@ t32_test() { error_noexit "Enable mdt quota feature" return 1 } + if $r test -f $mdt2_dev; then + $r $TUNEFS --quota $mdt2_dev || { + $r losetup -a + error_noexit "Enable mdt quota feature" + return 1 + } + fi fi else if [ -n "$($LCTL list_nids | grep -v '\(tcp\|lo\)[[:digit:]]*$')" ]; then @@ -1661,7 +1753,7 @@ t32_test() { $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 -d $tmp/mnt/mdt + $r $UMOUNT $tmp/mnt/mdt fi mopts=exclude=$fsname-OST0000 @@ -1679,7 +1771,20 @@ t32_test() { } shall_cleanup_mdt=true - if [ "$dne_upgrade" != "no" ]; then + if $r test -f $mdt2_dev; then + mopts=mgsnode=$nid,$mopts + $r $MOUNT_CMD -o $mopts $mdt2_dev $tmp/mnt/mdt1 || { + $r losetup -a + error_noexit "Mounting the MDT" + return 1 + } + + echo "mount new MDT....$mdt2_dev" + $r $LCTL set_param -n mdt.${fsname}*.enable_remote_dir=1 || + error_noexit "enable remote dir create failed" + + shall_cleanup_mdt1=true + elif [ "$dne_upgrade" != "no" ]; then local fs2mdsdev=$(mdsdevname 1_2) local fs2mdsvdev=$(mdsvdevname 1_2) @@ -1700,7 +1805,6 @@ t32_test() { } echo "mount new MDT....$fs2mdsdev" - $r mkdir -p $tmp/mnt/mdt1 $r $MOUNT_CMD -o $mopts $fs2mdsdev $tmp/mnt/mdt1 || { error_noexit "mount mdt1 failed" return 1 @@ -1780,6 +1884,10 @@ t32_test() { error_noexit "Setting \"lov.stripesize\"" return 1 } + $r $LCTL conf_param $fsname-MDT0000.mdd.atime_diff=70 || { + error_noexit "Setting \"mdd.atime_diff\"" + return 1 + } if [ "$ff_convert" != "no" -a $(facet_fstype ost1) == "ldiskfs" ]; then $r $LCTL lfsck_start -M $fsname-OST0000 || { @@ -1819,7 +1927,6 @@ t32_test() { error_noexit "Setting MDT1 \"lov.stripesize\"" return 1 } - fi if [ "$writeconf" ]; then @@ -1827,8 +1934,9 @@ t32_test() { error_noexit "Mounting the client" return 1 } + shall_cleanup_lustre=true - $LCTL set_param debug="$PTLDEBUG" + $r $LCTL set_param debug="$PTLDEBUG" t32_verify_quota $node $fsname $tmp/mnt/lustre || { error_noexit "verify quota failed" @@ -1836,20 +1944,39 @@ t32_test() { } if [ "$dne_upgrade" != "no" ]; then - $LFS mkdir -i 1 $tmp/mnt/lustre/remote_dir || { + $LFS mkdir -i 1 -c2 $tmp/mnt/lustre/striped_dir || { error_noexit "set remote dir failed" return 1 } + $LFS setdirstripe -D -c2 $tmp/mnt/lustre/striped_dir + pushd $tmp/mnt/lustre - tar -cf - . --exclude=./remote_dir | - tar -xvf - -C remote_dir 1>/dev/null || { + tar -cf - . --exclude=./striped_dir \ + --exclude=./remote_dir | + tar -xvf - -C striped_dir 1>/dev/null || { error_noexit "cp to remote dir failed" return 1 } popd fi + # If it is upgrade from DNE (2.5), then rename the remote dir, + # which is created in 2.5 to striped dir. + if $r test -f $mdt2_dev; then + stripe_index=$(LFS getdirstripe -i \ + $tmp/mnt/lustre/remote_dir) + [ $stripe_index -eq 1 ] || { + error_noexit "get index $striped_index failed" + return 1 + } + mv $tmp/mnt/lustre/remote_dir \ + $tmp/mnt/lustre/striped_dir/ || { + error_noexit "mv failed" + return 1 + } + fi + dd if=/dev/zero of=$tmp/mnt/lustre/tmp_file bs=10k count=10 || { error_noexit "dd failed" return 1 @@ -1864,13 +1991,15 @@ t32_test() { # is identical $r cat $tmp/sha1sums | sort -k 2 >$tmp/sha1sums.orig if [ "$dne_upgrade" != "no" ]; then - pushd $tmp/mnt/lustre/remote_dir + pushd $tmp/mnt/lustre/striped_dir else pushd $tmp/mnt/lustre fi - find ! -name .lustre -type f -exec sha1sum {} \; | + find ! -path "*remote_dir*" ! -name .lustre -type f \ + -exec sha1sum {} \; | sort -k 2 >$tmp/sha1sums || { + popd error_noexit "sha1sum" return 1 } @@ -1879,12 +2008,30 @@ t32_test() { error_noexit "sha1sum verification failed" return 1 fi + + # if upgrade from DNE (2.5), then check remote directory + if $r test -f $mdt2_dev; then + pushd $tmp/mnt/lustre/striped_dir/remote_dir + find ! -name .lustre -type f \ + -exec sha1sum {} \; | + sort -k 2 >$tmp/sha1sums || { + popd + error_noexit "sha1sum" + return 1 + } + popd + if ! diff -ub $tmp/sha1sums.orig \ + $tmp/sha1sums; then + error_noexit "sha1sum dne failed" + return 1 + fi + fi else echo "sha1sum verification skipped" fi if [ "$dne_upgrade" != "no" ]; then - rm -rf $tmp/mnt/lustre/remote_dir || { + rm -rf $tmp/mnt/lustre/striped_dir || { error_noexit "remove remote dir failed" return 1 } @@ -1926,6 +2073,40 @@ t32_test() { echo "list verification skipped" fi + # migrate files/dirs to remote MDT, then move them back + if [ $(lustre_version_code mds1) -ge $(version_code 2.7.50) -a \ + $dne_upgrade != "no" ]; then + $r $LCTL set_param -n \ + mdt.${fsname}*.enable_remote_dir=1 2>/dev/null + + echo "test migration" + pushd $tmp/mnt/lustre + for dir in $(find ! -name .lustre ! -name . -type d); do + mdt_index=$($LFS getdirstripe -i $dir) + stripe_cnt=$($LFS getdirstripe -c $dir) + if [ $mdt_index = 0 -a $stripe_cnt -le 1 ]; then + $LFS mv -M 1 $dir || { + popd + error_noexit "migrate MDT1 failed" + return 1 + } + fi + done + + for dir in $(find ! -name . ! -name .lustre -type d); do + mdt_index=$($LFS getdirstripe -i $dir) + stripe_cnt=$($LFS getdirstripe -c $dir) + if [ $mdt_index = 1 -a $stripe_cnt -le 1 ]; then + $LFS mv -M 0 $dir || { + popd + error_noexit "migrate MDT0 failed" + return 1 + } + fi + done + popd + fi + # # When adding new data verification tests, please check for # the presence of the required reference files first, like @@ -1956,20 +2137,20 @@ t32_test() { shall_cleanup_lustre=false else if [ "$dne_upgrade" != "no" ]; then - $r umount -d $tmp/mnt/mdt1 || { + $r $UMOUNT $tmp/mnt/mdt1 || { error_noexit "Unmounting the MDT2" return 1 } shall_cleanup_mdt1=false fi - $r umount -d $tmp/mnt/mdt || { + $r $UMOUNT $tmp/mnt/mdt || { error_noexit "Unmounting the MDT" return 1 } shall_cleanup_mdt=false - $r umount -d $tmp/mnt/ost || { + $r $UMOUNT $tmp/mnt/ost || { error_noexit "Unmounting the OST" return 1 } @@ -2033,6 +2214,10 @@ test_32c() { [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return t32_check for tarball in $tarballs; do + # Do not support 1_8 and 2_1 direct upgrade to DNE2 anymore */ + echo $tarball | grep "1_8" && continue + echo $tarball | grep "2_1" && continue + load_modules dne_upgrade=yes t32_test $tarball writeconf || rc=$? done return $rc @@ -2093,7 +2278,7 @@ test_33a() { # bug 12333, was test_33 cp /etc/hosts $MOUNT2/ || error "copy /etc/hosts $MOUNT2/ failed" $GETSTRIPE $MOUNT2/hosts || error "$GETSTRIPE $MOUNT2/hosts failed" - umount -d $MOUNT2 + umount $MOUNT2 stop fs2ost -f stop fs2mds -f cleanup_nocli || error "cleanup_nocli failed with $?" @@ -2293,7 +2478,6 @@ test_36() { # 12743 local rc=0 local FSNAME2=test1234 - local fs3ost_HOST=$ost_HOST local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/}) [ -n "$ost1_HOST" ] && fs2ost_HOST=$ost1_HOST && fs3ost_HOST=$ost1_HOST @@ -2359,7 +2543,7 @@ test_36() { # 12743 rc=3 fi - umount -d $MOUNT2 + $UMOUNT $MOUNT2 stop fs3ost -f || error "unable to stop OST3" stop fs2ost -f || error "unable to stop OST2" stop fs2mds -f || error "unable to stop second MDS" @@ -2397,7 +2581,7 @@ test_37() { echo mount_op=$mount_op - do_facet $SINGLEMDS "umount -d $mntpt && rm -f $mdsdev_sym" + do_facet $SINGLEMDS "$UMOUNT $mntpt && rm -f $mdsdev_sym" if $(echo $mount_op | grep -q "unable to set tunable"); then error "set tunables failed for symlink device" @@ -2511,9 +2695,14 @@ test_41a() { #bug 14134 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/}) - start $SINGLEMDS $MDSDEV $MDS_MOUNT_OPTS -o nosvc -n + start_mdt 1 -o nosvc -n + if [ $MDSCOUNT -ge 2 ]; then + for num in $(seq 2 $MDSCOUNT); do + start_mdt $num || return + done + fi start ost1 $(ostdevname 1) $OST_MOUNT_OPTS - start $SINGLEMDS $MDSDEV $MDS_MOUNT_OPTS -o nomgs,force + start_mdt 1 -o nomgs,force mount_client $MOUNT || error "mount_client $MOUNT failed" sleep 5 @@ -2541,9 +2730,14 @@ test_41b() { reformat local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/}) - start $SINGLEMDS $MDSDEV $MDS_MOUNT_OPTS -o nosvc -n + start_mdt 1 -o nosvc -n + if [ $MDSCOUNT -ge 2 ]; then + for num in $(seq 2 $MDSCOUNT); do + start_mdt $num || return + done + fi start_ost || error "Unable to start OST1" - start $SINGLEMDS $MDSDEV $MDS_MOUNT_OPTS -o nomgs,force + start_mdt 1 -o nomgs,force mount_client $MOUNT || error "mount_client $MOUNT failed" sleep 5 @@ -2568,7 +2762,7 @@ test_41c() { { skip "Need MDS version 2.5.4+ or 2.5.26+ or 2.6.52+"; return; } cleanup - # MDT concurent start + # MDT concurrent start #define OBD_FAIL_TGT_DELAY_CONNECT 0x703 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0x703" start mds1 $(mdsdevname 1) $MDS_MOUNT_OPTS & @@ -2589,10 +2783,17 @@ test_41c() { echo "2nd MDT start succeed" else stop mds1 -f - error "unexpected concurent MDT mounts result, rc=$rc rc2=$rc2" + error "unexpected concurrent MDT mounts result, rc=$rc rc2=$rc2" + fi + + if [ $MDSCOUNT -ge 2 ]; then + for num in $(seq 2 $MDSCOUNT); do + start_mdt $num || return + done fi - # OST concurent start + # OST concurrent start + #define OBD_FAIL_TGT_DELAY_CONNECT 0x703 do_facet ost1 "$LCTL set_param fail_loc=0x703" start ost1 $(ostdevname 1) $OST_MOUNT_OPTS & @@ -2612,26 +2813,26 @@ test_41c() { echo "1st OST start failed with EALREADY" echo "2nd OST start succeed" else - stop mds1 -f + stop_mds -f stop ost1 -f - error "unexpected concurent OST mounts result, rc=$rc rc2=$rc2" + error "unexpected concurrent OST mounts result, rc=$rc rc2=$rc2" fi # cleanup - stop mds1 -f + stop_mds stop ost1 -f # verify everything ok start_mds if [ $? != 0 ] then - stop mds1 -f + stop_mds error "MDT(s) start failed" fi start_ost if [ $? != 0 ] then - stop mds1 -f + stop_mds stop ost1 -f error "OST(s) start failed" fi @@ -2639,20 +2840,20 @@ test_41c() { mount_client $MOUNT if [ $? != 0 ] then - stop mds1 -f + stop_mds stop ost1 -f error "client start failed" fi check_mount if [ $? != 0 ] then - stop mds1 -f + stop_mds stop ost1 -f error "client mount failed" fi cleanup } -run_test 41c "concurent mounts of MDT/OST should all fail but one" +run_test 41c "concurrent mounts of MDT/OST should all fail but one" test_42() { #bug 14693 setup @@ -2793,7 +2994,7 @@ test_43() { # root should be able to access root file after that # local NIDLIST=$($LCTL list_nids all | tr '\n' ' ') - NIDLIST="2@elan $NIDLIST 192.168.0.[2,10]@tcp" + NIDLIST="2@gni $NIDLIST 192.168.0.[2,10]@tcp" NIDLIST=$(echo $NIDLIST | tr -s ' ' ' ') set_conf_param_and_check mds \ "$LCTL get_param -n mdt.$FSNAME-MDT0000.nosquash_nids" \ @@ -2966,7 +3167,7 @@ run_test 47 "server restart does not make client loss lru_resize settings" cleanup_48() { trap 0 - # reformat after this test is needed - if test will failed + # reformat after this test is needed - if the test fails, # we will have unkillable file at FS reformat_and_config } @@ -3230,7 +3431,7 @@ test_50f() { stop_ost2 || error "Unable to stop OST2" fi - umount_client $MOUNT || error "Unable to unmount client" + umount_client $MOUNT -f || error "Unable to unmount client" stop_ost || error "Unable to stop OST1" stop_mds || error "Unable to stop MDS" #writeconf to remove all ost2 traces for subsequent tests @@ -3306,6 +3507,57 @@ test_50h() { } run_test 50h "LU-642: activate deactivated OST" +test_50i() { + # prepare MDT/OST, make OSC inactive for OST1 + [ "$MDSCOUNT" -lt "2" ] && skip_env "$MDSCOUNT < 2, skipping" && return + + [ $(facet_fstype ost1) == zfs ] && import_zpool ost1 + load_modules + do_facet mds2 "$TUNEFS --param mdc.active=0 $(mdsdevname 2)" || + error "tunefs MDT2 failed" + start_mds || error "Unable to start MDT" + start_ost || error "Unable to start OST1" + start_ost2 || error "Unable to start OST2" + mount_client $MOUNT || error "client start failed" + + mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed" + + $LCTL conf_param ${FSNAME}-MDT0000.mdc.active=0 && + error "deactive MDC0 succeeds" + # activate MDC for MDT2 + local TEST="$LCTL get_param -n mdc.${FSNAME}-MDT0001-mdc-[!M]*.active" + set_conf_param_and_check client \ + "$TEST" "${FSNAME}-MDT0001.mdc.active" 1 || + error "Unable to activate MDT2" + + wait_clients_import_state ${CLIENTS:-$HOSTNAME} mds2 FULL + if [ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.7.60) ] + then + wait_dne_interconnect + fi + $LFS mkdir -i1 $DIR/$tdir/2 || error "mkdir $DIR/$tdir/2 failed" + # create some file + createmany -o $DIR/$tdir/2/$tfile-%d 1 || error "create files failed" + + rm -rf $DIR/$tdir/2 || error "unlink dir failed" + + # deactivate MDC for MDT2 + local TEST="$LCTL get_param -n mdc.${FSNAME}-MDT0001-mdc-[!M]*.active" + set_conf_param_and_check client \ + "$TEST" "${FSNAME}-MDT0001.mdc.active" 0 || + error "Unable to deactivate MDT2" + + $LFS mkdir -i1 $DIR/$tdir/2 && + error "mkdir $DIR/$tdir/2 succeeds after deactive MDT" + + # cleanup + umount_client $MOUNT || error "Unable to umount client" + stop_mds + stop_ost + stop_ost 2 +} +run_test 50i "activate deactivated MDT" + test_51() { local LOCAL_TIMEOUT=20 @@ -3324,7 +3576,8 @@ test_51() { start_ost2 || error "Unable to start OST1" wait $pid stop_ost2 || error "Unable to stop OST1" - cleanup || error "cleanup failed with $?" + umount_client $MOUNT -f || error “unmount $MOUNT failed” + cleanup_nocli || error “stop server failed” #writeconf to remove all ost2 traces for subsequent tests writeconf_or_reformat } @@ -3442,9 +3695,12 @@ test_52() { [ $? -eq 0 ] || { error "Unable to move objects"; return 14; } # recover objects dry-run - echo "ll_recover_lost_found_objs dry_run" - do_node $ost1node "ll_recover_lost_found_objs -n -d $ost1mnt/O" || - error "ll_recover_lost_found_objs failed" + if [ $(lustre_version_code ost1) -ge $(version_code 2.5.56) ]; then + echo "ll_recover_lost_found_objs dry_run" + do_node $ost1node \ + "ll_recover_lost_found_objs -n -d $ost1mnt/O" || + error "ll_recover_lost_found_objs failed" + fi # recover objects echo "ll_recover_lost_found_objs fix run" @@ -3672,14 +3928,14 @@ test_55() { run_test 55 "check lov_objid size" test_56() { + local server_version=$(lustre_version_code $SINGLEMDS) local mds_journal_size_orig=$MDSJOURNALSIZE local n MDSJOURNALSIZE=16 for num in $(seq 1 $MDSCOUNT); do - add mds${num} $(mkfs_opts mds${num} $(mdsdevname $num)) \ - --reformat $(mdsdevname $num) $(mdsvdevname $num) + reformat_mdt $num done add ost1 $(mkfs_opts ost1 $(ostdevname 1)) --index=10000 --reformat \ $(ostdevname 1) $(ostvdevname 1) @@ -3692,13 +3948,19 @@ test_56() { mount_client $MOUNT || error "Unable to mount client" echo ok $LFS osts - wait_osc_import_state mds ost1 FULL - wait_osc_import_state mds ost2 FULL - $SETSTRIPE --stripe-count=-1 $DIR/$tfile || - error "Unable to setstripe $DIR/$tfile" - n=$($LFS getstripe --stripe-count $DIR/$tfile) - [ "$n" -eq 2 ] || error "Stripe count not two: $n" - rm $DIR/$tfile + + if [[ $server_version -ge $(version_code 2.6.54) ]] || + [[ $server_version -ge $(version_code 2.5.4) && + $server_version -lt $(version_code 2.5.11) ]]; then + wait_osc_import_state mds ost1 FULL + wait_osc_import_state mds ost2 FULL + $SETSTRIPE --stripe-count=-1 $DIR/$tfile || + error "Unable to setstripe $DIR/$tfile" + n=$($LFS getstripe --stripe-count $DIR/$tfile) + [ "$n" -eq 2 ] || error "Stripe count not two: $n" + rm $DIR/$tfile + fi + stopall MDSJOURNALSIZE=$mds_journal_size_orig reformat @@ -3825,20 +4087,19 @@ test_60() { # LU-471 run_test 60 "check mkfs.lustre --mkfsoptions -E -O options setting" test_61() { # LU-80 - local reformat=false + local lxattr=false [ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.1.53) ] || { skip "Need MDS version at least 2.1.53"; return 0; } if [ $(facet_fstype $SINGLEMDS) == ldiskfs ] && - ! large_xattr_enabled; then - reformat=true - LDISKFS_MKFS_OPTS+=" -O large_xattr" + ! large_xattr_enabled; then + lxattr=true for num in $(seq $MDSCOUNT); do - add mds${num} $(mkfs_opts mds$num $(mdsdevname $num)) \ - --reformat $(mdsdevname $num) $(mdsvdevname $num) || - error "add mds $num failed" + do_facet mds${num} $TUNE2FS -O large_xattr \ + $(mdsdevname $num) || + error "tune2fs on mds $num failed" done fi @@ -3851,45 +4112,59 @@ test_61() { # LU-80 local large_value="$(generate_string $(max_xattr_size))" local small_value="bar" - local name="trusted.big" - log "save large xattr $name on $file" - setfattr -n $name -v $large_value $file || - error "saving $name on $file failed" - - local new_value=$(get_xattr_value $name $file) - [[ "$new_value" != "$large_value" ]] && - error "$name different after saving" - - log "shrink value of $name on $file" - setfattr -n $name -v $small_value $file || - error "shrinking value of $name on $file failed" - - new_value=$(get_xattr_value $name $file) - [[ "$new_value" != "$small_value" ]] && - error "$name different after shrinking" - - log "grow value of $name on $file" - setfattr -n $name -v $large_value $file || - error "growing value of $name on $file failed" - - new_value=$(get_xattr_value $name $file) - [[ "$new_value" != "$large_value" ]] && - error "$name different after growing" - - log "check value of $name on $file after remounting MDS" - fail $SINGLEMDS - new_value=$(get_xattr_value $name $file) - [[ "$new_value" != "$large_value" ]] && - error "$name different after remounting MDS" - - log "remove large xattr $name from $file" - setfattr -x $name $file || error "removing $name from $file failed" - - rm -f $file - stopall - if $reformat; then - LDISKFS_MKFS_OPTS=${LDISKFS_MKFS_OPTS% -O large_xattr} - reformat + local name="trusted.big" + log "save large xattr $name on $file" + setfattr -n $name -v $large_value $file || + error "saving $name on $file failed" + + local new_value=$(get_xattr_value $name $file) + [[ "$new_value" != "$large_value" ]] && + error "$name different after saving" + + log "shrink value of $name on $file" + setfattr -n $name -v $small_value $file || + error "shrinking value of $name on $file failed" + + new_value=$(get_xattr_value $name $file) + [[ "$new_value" != "$small_value" ]] && + error "$name different after shrinking" + + log "grow value of $name on $file" + setfattr -n $name -v $large_value $file || + error "growing value of $name on $file failed" + + new_value=$(get_xattr_value $name $file) + [[ "$new_value" != "$large_value" ]] && + error "$name different after growing" + + log "check value of $name on $file after remounting MDS" + fail $SINGLEMDS + new_value=$(get_xattr_value $name $file) + [[ "$new_value" != "$large_value" ]] && + error "$name different after remounting MDS" + + log "remove large xattr $name from $file" + setfattr -x $name $file || error "removing $name from $file failed" + + if $lxattr; then + 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" + fi + + # need to delete this file to avoid problems in other tests + rm -f $file + stopall || error "stopping systems to turn off large_xattr" + if $lxattr; then + for num in $(seq $MDSCOUNT); do + do_facet mds${num} $TUNE2FS -O ^large_xattr \ + $(mdsdevname $num) || + error "tune2fs on mds $num failed" + done fi } run_test 61 "large xattr" @@ -3908,10 +4183,10 @@ test_62() { { skip "Need MDS version at least 2.2.51"; return 0; } echo "disable journal for mds" - do_facet mds tune2fs -O ^has_journal $mdsdev || error "tune2fs failed" + do_facet mds $TUNE2FS -O ^has_journal $mdsdev || error "tune2fs failed" start_mds && error "MDT start should fail" echo "disable journal for ost" - do_facet ost1 tune2fs -O ^has_journal $ostdev || error "tune2fs failed" + do_facet ost1 $TUNE2FS -O ^has_journal $ostdev || error "tune2fs failed" start_ost && error "OST start should fail" cleanup || error "cleanup failed with rc $?" reformat_and_config @@ -3946,7 +4221,8 @@ test_64() { stop_ost2 || error "Unable to stop second ost" echo "$LFS df" $LFS df --lazy || error "lfs df failed" - cleanup || error "cleanup failed with $?" + umount_client $MOUNT -f || error “unmount $MOUNT failed” + cleanup_nocli || error "cleanup_nocli failed with $?" #writeconf to remove all ost2 traces for subsequent tests writeconf_or_reformat } @@ -3981,7 +4257,7 @@ test_65() { # LU-2237 do_facet $SINGLEMDS \ "mount -t $(facet_fstype $SINGLEMDS) $opts $devname $brpt" do_facet $SINGLEMDS "rm -f ${brpt}/last_rcvd" - do_facet $SINGLEMDS "umount -d $brpt" + do_facet $SINGLEMDS "$UMOUNT $brpt" # restart MDS, the "last_rcvd" file should be recreated. start_mds || error "fail to restart the MDS" @@ -4166,6 +4442,10 @@ test_68() { } run_test 68 "be able to reserve specific sequences in FLDB" +# Test 69: is about the total number of objects ever created on an OST. +# so that when it is reformatted the normal MDS->OST orphan recovery won't +# just "precreate" the missing objects. In the past it might try to recreate +# millions of objects after an OST was reformatted test_69() { local server_version=$(lustre_version_code $SINGLEMDS) @@ -4177,6 +4457,7 @@ test_69() { skip "Need MDS version at least 2.5.0" && return setup + mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed" # use OST0000 since it probably has the most creations local OSTNAME=$(ostname_from_index 0) @@ -4184,21 +4465,44 @@ test_69() { local last_id=$(do_facet mds1 $LCTL get_param -n \ osc.$mdtosc_proc1.prealloc_last_id) - # Want to have OST LAST_ID over 1.5 * OST_MAX_PRECREATE to - # verify that the LAST_ID recovery is working properly. If + # Want to have OST LAST_ID over 5 * OST_MAX_PRECREATE to + # verify that the LAST_ID recovery is working properly. If # not, then the OST will refuse to allow the MDS connect # because the LAST_ID value is too different from the MDS #define OST_MAX_PRECREATE=20000 - local num_create=$((20000 * 5)) + local ost_max_pre=20000 + local num_create=$(( ost_max_pre * 5 + 1 - last_id)) + + # If the LAST_ID is already over 5 * OST_MAX_PRECREATE, we don't + # need to create any files. So, skip this section. + 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." + + $SETSTRIPE -i 0 $DIR/$tdir || + error "$SETSTRIPE -i 0 $DIR/$tdir failed" + 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 + fi - mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed" - $SETSTRIPE -i 0 $DIR/$tdir || error "$SETSTRIPE -i 0 $DIR/$tdir failed" - createmany -o $DIR/$tdir/$tfile- $num_create || - error "createmany: failed to create $num_create files: $?" # delete all of the files with objects on OST0 so the # filesystem is not inconsistent later on - $LFS find $MOUNT --ost 0 | xargs rm + $LFS find $MOUNT --ost 0 -print0 | xargs -0 rm + umount_client $MOUNT || error "umount client failed" stop_ost || error "OST0 stop failure" add ost1 $(mkfs_opts ost1 $(ostdevname 1)) --reformat --replace \ $(ostdevname 1) $(ostvdevname 1) || @@ -4206,10 +4510,15 @@ test_69() { start_ost || error "OST0 restart failure" wait_osc_import_state mds ost FULL + mount_client $MOUNT || error "mount client failed" touch $DIR/$tdir/$tfile-last || error "create file after reformat" local idx=$($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)) ] && + error "OST replacement created too many inodes; $iused" cleanup || error "cleanup failed with $?" } run_test 69 "replace an OST with the same index" @@ -4223,8 +4532,9 @@ test_70a() { start_mdt 1 || error "MDT0 start fail" start_ost || error "OST0 start fail" - - start_mdt 2 || error "MDT1 start fail" + for num in $(seq 2 $MDSCOUNT); do + start_mdt $num || return + done mount_client $MOUNT || error "mount client fails" @@ -4244,8 +4554,7 @@ test_70b() { start_ost || error "OST0 start fail" - start_mdt 1 || error "MDT0 start fail" - start_mdt 2 || error "MDT1 start fail" + start_mds || error "MDS start fail" mount_client $MOUNT || error "mount client fails" @@ -4264,8 +4573,7 @@ test_70c() { [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return local MDTIDX=1 - start_mdt 1 || error "MDT0 start fail" - start_mdt 2 || error "MDT1 start fail" + start_mds || error "MDS start fail" start_ost || error "OST0 start fail" mount_client $MOUNT || error "mount client fails" @@ -4289,8 +4597,7 @@ test_70d() { [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return local MDTIDX=1 - start_mdt 1 || error "MDT0 start fail" - start_mdt 2 || error "MDT1 start fail" + start_mds || error "MDS start fail" start_ost || error "OST0 start fail" mount_client $MOUNT || error "mount client fails" @@ -4322,7 +4629,10 @@ test_71a() { start_mdt 1 || error "MDT0 start fail" start_ost || error "OST0 start fail" - start_mdt 2 || error "MDT1 start fail" + for num in $(seq 2 $MDSCOUNT); do + start_mdt $num || return + done + start_ost2 || error "OST1 start fail" mount_client $MOUNT || error "mount client fails" @@ -4335,8 +4645,7 @@ test_71a() { rm -rf $DIR/$tdir || error "delete dir fail" umount_client $MOUNT || error "umount_client failed" - stop_mdt 1 || error "MDT0 stop fail" - stop_mdt 2 || error "MDT1 stop fail" + stop_mds || error "MDS stop fail" stop_ost || error "OST0 stop fail" stop_ost2 || error "OST1 stop fail" } @@ -4349,7 +4658,9 @@ test_71b() { fi local MDTIDX=1 - start_mdt 2 || error "MDT1 start fail" + for num in $(seq 2 $MDSCOUNT); do + start_mdt $num || return + done start_ost || error "OST0 start fail" start_mdt 1 || error "MDT0 start fail" start_ost2 || error "OST1 start fail" @@ -4364,8 +4675,7 @@ test_71b() { rm -rf $DIR/$tdir || error "delete dir fail" umount_client $MOUNT || error "umount_client failed" - stop_mdt 1 || error "MDT0 stop fail" - stop_mdt 2 || error "MDT1 stop fail" + stop_mds || error "MDT0 stop fail" stop_ost || error "OST0 stop fail" stop_ost2 || error "OST1 stop fail" } @@ -4380,7 +4690,9 @@ test_71c() { start_ost || error "OST0 start fail" start_ost2 || error "OST1 start fail" - start_mdt 2 || error "MDT1 start fail" + for num in $(seq 2 $MDSCOUNT); do + start_mdt $num || return + done start_mdt 1 || error "MDT0 start fail" mount_client $MOUNT || error "mount client fails" @@ -4393,8 +4705,7 @@ test_71c() { rm -rf $DIR/$tdir || error "delete dir fail" umount_client $MOUNT || error "umount_client failed" - stop_mdt 1 || error "MDT0 stop fail" - stop_mdt 2 || error "MDT1 stop fail" + stop_mds || error "MDS stop fail" stop_ost || error "OST0 stop fail" stop_ost2 || error "OST1 stop fail" @@ -4409,7 +4720,9 @@ test_71d() { local MDTIDX=1 start_ost || error "OST0 start fail" - start_mdt 2 || error "MDT0 start fail" + for num in $(seq 2 $MDSCOUNT); do + start_mdt $num || return + done start_mdt 1 || error "MDT0 start fail" start_ost2 || error "OST1 start fail" @@ -4423,8 +4736,7 @@ test_71d() { rm -rf $DIR/$tdir || error "delete dir fail" umount_client $MOUNT || error "umount_client failed" - stop_mdt 1 || error "MDT0 stop fail" - stop_mdt 2 || error "MDT1 stop fail" + stop_mds || error "MDS stop fail" stop_ost || error "OST0 stop fail" stop_ost2 || error "OST1 stop fail" @@ -4439,7 +4751,9 @@ test_71e() { local MDTIDX=1 start_ost || error "OST0 start fail" - start_mdt 2 || error "MDT1 start fail" + for num in $(seq 2 $MDSCOUNT); do + start_mdt $num || return + done start_ost2 || error "OST1 start fail" start_mdt 1 || error "MDT0 start fail" @@ -4453,8 +4767,7 @@ test_71e() { rm -rf $DIR/$tdir || error "delete dir fail" umount_client $MOUNT || error "umount_client failed" - stop_mdt 1 || error "MDT0 stop fail" - stop_mdt 2 || error "MDT1 stop fail" + stop_mds || error "MDS stop fail" stop_ost || error "OST0 stop fail" stop_ost2 || error "OST1 stop fail" @@ -4476,7 +4789,8 @@ test_72() { #LU-2634 add mds${num} $(mkfs_opts mds$num $(mdsdevname $num)) \ --reformat $(mdsdevname $num) $(mdsvdevname $num) || error "add mds $num failed" - $TUNE2FS -O extents $(mdsdevname $num) + do_facet mds${num} "$TUNE2FS -O extents $(mdsdevname $num)" || + error "$TUNE2FS failed on mds${num}" done add ost1 $(mkfs_opts ost1 $ostdev) --reformat $ostdev || @@ -4842,27 +5156,27 @@ test_79() { # LU-4227 add ost1 $opts_ost1 $ostdev1 $ostvdev2 && error "Must specify --mgsnode when formatting an ost" - return 0 + reformat } run_test 79 "format MDT/OST without mgs option (should return errors)" test_80() { - start_mds - start_ost + start_mds || error "Failed to start MDT" + start_ost || error "Failed to start OST1" uuid=$(do_facet ost1 $LCTL get_param -n mgc.*.uuid) #define OBD_FAIL_MGS_PAUSE_TARGET_CON 0x906 do_facet ost1 "$LCTL set_param fail_val=10 fail_loc=0x906" do_facet mgs "$LCTL set_param fail_val=10 fail_loc=0x906" do_facet mgs "$LCTL set_param -n mgs/MGS/evict_client $uuid" sleep 30 - start_ost2 + start_ost2 || error "Failed to start OST2" do_facet ost1 "$LCTL set_param fail_loc=0" stopall } run_test 80 "mgc import reconnect race" -# Save the original values of $OSTCOUNT and $OSTINDEX$i. +#Save the original values of $OSTCOUNT and $OSTINDEX$i. save_ostindex() { local new_ostcount=$1 saved_ostcount=$OSTCOUNT @@ -5166,16 +5480,18 @@ test_82b() { # LU-4665 run_test 82b "specify OSTs for file with --pool and --ost-list options" test_83() { - local dev - local ostmnt - local fstype - local mnt_opts - + [[ $(lustre_version_code ost1) -ge $(version_code 2.6.91) ]] || + { skip "Need OST version at least 2.6.91" && return 0; } if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then skip "Only applicable to ldiskfs-based MDTs" return fi + local dev + local ostmnt + local fstype + local mnt_opts + dev=$(ostdevname 1) ostmnt=$(facet_mntpt ost1) fstype=$(facet_fstype ost1) @@ -5243,24 +5559,26 @@ recovery_time_min() { echo $((2*$RECONNECT_DELAY_MAX)) } -test_83() { +test_84() { local facet=$SINGLEMDS local num=$(echo $facet | tr -d "mds") local dev=$(mdsdevname $num) local time_min=$(recovery_time_min) local recovery_duration local completed_clients + local correct_clients + local wrap_up=5 - echo "start mds service on `facet_active_host $facet`" - start $facet ${dev} $MDS_MOUNT_OPTS \ - "-o recovery_time_hard=$time_min,recovery_time_soft=$time_min"\ - $@ || return 94 + load_modules + echo "start mds service on $(facet_active_host $facet)" + start_mds \ + "-o recovery_time_hard=$time_min,recovery_time_soft=$time_min" $@ || + error "start MDS failed" start_ost start_ost2 - echo "recovery_time_hard $time_min, recovery_time_soft $time_min, \ - timeout $TIMEOUT" + echo "recovery_time=$time_min, timeout=$TIMEOUT, wrap_up=$wrap_up" mount_client $MOUNT1 || error "mount failed" mount_client $MOUNT2 || error "mount failed" @@ -5271,26 +5589,27 @@ test_83() { # We need to catch the end of recovery window to extend it. # Skip 5 requests and add delay to request handling. #define OBD_FAIL_TGT_REPLAY_DELAY 0x709 | FAIL_SKIP - do_facet $SINGLEMDS "lctl set_param fail_loc=0x20000709" - do_facet $SINGLEMDS "lctl set_param fail_val=5" + do_facet $SINGLEMDS "lctl set_param fail_loc=0x20000709 fail_val=5" - facet_failover $SINGLEMDS || error "failover: $?" + facet_failover --fsck $SINGLEMDS || error "failover: $?" client_up echo "recovery status" - do_facet $SINGLEMDS "$LCTL get_param -n \ - mdt.$FSNAME-MDT0000.recovery_status" - - recovery_duration=$(do_facet $SINGLEMDS "$LCTL get_param -n \ - mdt.$FSNAME-MDT0000.recovery_status" | \ - grep recovery_duration |awk '{print $2}') - (($recovery_duration>$time_min)) && \ - error "recovery_duration > recovery_time_hard" - completed_clients=$(do_facet $SINGLEMDS "$LCTL get_param -n \ - mdt.$FSNAME-MDT0000.recovery_status" | \ - grep completed_clients |awk '{print $2}') - [ "$completed_clients" = "1/2" ] || \ - error "completed_clients != 1/2: "$completed_clients + do_facet $SINGLEMDS \ + "$LCTL get_param -n mdt.$FSNAME-MDT0000.recovery_status" + + recovery_duration=$(do_facet $SINGLEMDS \ + "$LCTL get_param -n mdt.$FSNAME-MDT0000.recovery_status" | + awk '/recovery_duration/ { print $2 }') + (( $recovery_duration > $time_min + $wrap_up )) && + error "recovery_duration > recovery_time_hard + wrap up" + completed_clients=$(do_facet $SINGLEMDS \ + "$LCTL get_param -n mdt.$FSNAME-MDT0000.recovery_status" | + awk '/completed_clients/ { print $2 }') + + correct_clients="$MDSCOUNT/$((MDSCOUNT+1))" + [ "$completed_clients" = "${correct_clients}" ] || + error "$completed_clients != $correct_clients" do_facet $SINGLEMDS "lctl set_param fail_loc=0" umount_client $MOUNT1 @@ -5300,7 +5619,495 @@ test_83() { stop_ost2 stop_mds } -run_test 83 "check recovery_hard_time" +run_test 84 "check recovery_hard_time" + +test_85() { + [[ $(lustre_version_code ost1) -ge $(version_code 2.7.55) ]] || + { skip "Need OST version at least 2.7.55" && return 0; } +##define OBD_FAIL_OSD_OST_EA_FID_SET 0x197 + do_facet ost1 "lctl set_param fail_loc=0x197" + start_ost + stop_ost +} +run_test 85 "osd_ost init: fail ea_fid_set" + +test_86() { + [ "$(facet_fstype ost1)" = "zfs" ] && + skip "LU-6442: no such mkfs params for ZFS OSTs" && return + + local OST_OPTS="$(mkfs_opts ost1 $(ostdevname 1)) \ + --reformat $(ostdevname 1) $(ostvdevname 1)" + + local NEWSIZE=1024 + local OLDSIZE=$(do_facet ost1 "$DEBUGFS -c -R stats $(ostdevname 1)" | + awk '/Flex block group size: / { print $NF; exit; }') + + local opts=OST_OPTS + if [[ ${!opts} != *mkfsoptions* ]]; then + eval opts=\"${!opts} \ + --mkfsoptions='\\\"-O flex_bg -G $NEWSIZE\\\"'\" + else + val=${!opts//--mkfsoptions=\\\"/ \ + --mkfsoptions=\\\"-O flex_bg -G $NEWSIZE } + eval opts='${val}' + fi + + echo "params: $opts" + + add ost1 $opts || error "add ost1 failed with new params" + + local FOUNDSIZE=$(do_facet ost1 "$DEBUGFS -c -R stats $(ostdevname 1)" | + awk '/Flex block group size: / { print $NF; exit; }') + + [[ $FOUNDSIZE == $NEWSIZE ]] || + error "Flex block group size: $FOUNDSIZE, expected: $NEWSIZE" + return 0 +} +run_test 86 "Replacing mkfs.lustre -G option" + +test_87() { #LU-6544 + [[ $(lustre_version_code $SINGLEMDS1) -ge $(version_code 2.7.56) ]] || + { skip "Need MDS version at least 2.7.56" && return; } + [[ $(facet_fstype $SINGLEMDS) != ldiskfs ]] && + { skip "Only applicable to ldiskfs-based MDTs" && return; } + [[ $OSTCOUNT -gt 69 ]] && + { skip "Ignore wide striping situation" && return; } + + local mdsdev=$(mdsdevname 1) + local mdsvdev=$(mdsvdevname 1) + local file=$DIR/$tfile + local mntpt=$(facet_mntpt $SINGLEMDS) + local used_xattr_blk=0 + local inode_size=${1:-512} + local left_size=0 + local xtest="trusted.test" + local value + local orig + local i + + #Please see LU-6544 for MDT inode size calculation + if [ $OSTCOUNT -gt 26 ]; then + inode_size=2048 + elif [ $OSTCOUNT -gt 5 ]; then + inode_size=1024 + fi + left_size=$(expr $inode_size - \ + 156 - \ + 32 - \ + 32 - $OSTCOUNT \* 24 - 16 - 3 - \ + 24 - 16 - 3 - \ + 24 - 18 - $(expr length $tfile) - 16 - 4) + if [ $left_size -le 0 ]; then + echo "No space($left_size) is expected in inode." + echo "Try 1-byte xattr instead to verify this." + left_size=1 + else + echo "Estimate: at most $left_size-byte space left in inode." + fi + + unload_modules + reformat + + add mds1 $(mkfs_opts mds1 ${mdsdev}) --stripe-count-hint=$OSTCOUNT \ + --reformat $mdsdev $mdsvdev || error "add mds1 failed" + start_mdt 1 > /dev/null || error "start mdt1 failed" + for i in $(seq $OSTCOUNT); do + start ost$i $(ostdevname $i) $OST_MOUNT_OPTS > /dev/null || + error "start ost$i failed" + done + mount_client $MOUNT > /dev/null || error "mount client $MOUNT failed" + check_mount || error "check client $MOUNT failed" + + #set xattr + $SETSTRIPE -c -1 $file || error "$SETSTRIPE -c -1 $file failed" + $GETSTRIPE $file || error "$GETSTRIPE $file failed" + i=$($GETSTRIPE -c $file) + if [ $i -ne $OSTCOUNT ]; then + left_size=$(expr $left_size + $(expr $OSTCOUNT - $i) \* 24) + echo -n "Since only $i out $OSTCOUNT OSTs are used, " + echo -n "the expected left space is changed to " + echo "$left_size bytes at most." + fi + value=$(generate_string $left_size) + setfattr -n $xtest -v $value $file + orig=$(get_xattr_value $xtest $file) + [[ "$orig" != "$value" ]] && error "$xtest changed" + + #Verify if inode has some expected space left + umount $MOUNT > /dev/null || error "umount $MOUNT failed" + stop_mdt 1 > /dev/null || error "stop mdt1 failed" + mount_ldiskfs $SINGLEMDS || error "mount -t ldiskfs $SINGLEMDS failed" + + do_facet $SINGLEMDS ls -sal $mntpt/ROOT/$tfile + used_xattr_blk=$(do_facet $SINGLEMDS ls -s $mntpt/ROOT/$tfile | + awk '{ print $1 }') + [[ $used_xattr_blk -eq 0 ]] && + error "Please check MDS inode size calculation: \ + more than $left_size-byte space left in inode." + echo "Verified: at most $left_size-byte space left in inode." + + stopall +} +run_test 87 "check if MDT inode can hold EAs with N stripes properly" + +# $1 test directory +# $2 (optional) value of max_mod_rpcs_in_flight to set +check_max_mod_rpcs_in_flight() { + local dir="$1" + local mmr="$2" + local idx + local facet + local tmp + local i + + idx=$(printf "%04x" $($LFS getdirstripe -i $dir)) + facet="mds$((0x$idx + 1))" + + if [ -z "$mmr" ]; then + # get value of max_mod_rcps_in_flight + mmr=$($LCTL get_param -n \ + mdc.$FSNAME-MDT$idx-mdc-*.max_mod_rpcs_in_flight) || + error "Unable to get max_mod_rpcs_in_flight" + echo "max_mod_rcps_in_flight is $mmr" + else + # set value of max_mod_rpcs_in_flight + $LCTL set_param \ + mdc.$FSNAME-MDT$idx-mdc-*.max_mod_rpcs_in_flight=$mmr || + error "Unable to set max_mod_rpcs_in_flight to $mmr" + echo "max_mod_rpcs_in_flight set to $mmr" + fi + + # create mmr+1 files + echo "creating $((mmr + 1)) files ..." + umask 0022 + for i in $(seq $((mmr + 1))); do + touch $dir/file-$i + done + + ### part 1 ### + + # consumes mmr-1 modify RPC slots + #define OBD_FAIL_MDS_REINT_MULTI_NET 0x159 + # drop requests on MDT so that RPC slots are consumed + # during all the request resend interval + do_facet $facet "$LCTL set_param fail_loc=0x159" + echo "launch $((mmr - 1)) chmod in parallel ..." + for i in $(seq $((mmr - 1))); do + chmod 0600 $dir/file-$i & + done + sleep 1 + + # send one additional modify RPC + do_facet $facet "$LCTL set_param fail_loc=0" + echo "launch 1 additional chmod in parallel ..." + chmod 0600 $dir/file-$mmr & + sleep 1 + + # check this additional modify RPC get a modify RPC slot + # and succeed its operation + checkstat -vp 0600 $dir/file-$mmr || + error "Unable to send $mmr modify RPCs in parallel" + wait + + ### part 2 ### + + # consumes mmr modify RPC slots + #define OBD_FAIL_MDS_REINT_MULTI_NET 0x159 + # drop requests on MDT so that RPC slots are consumed + # during all the request resend interval + do_facet $facet "$LCTL set_param fail_loc=0x159" + echo "launch $mmr chmod in parallel ..." + for i in $(seq $mmr); do + chmod 0666 $dir/file-$i & + done + sleep 1 + + # send one additional modify RPC + do_facet $facet "$LCTL set_param fail_loc=0" + echo "launch 1 additional chmod in parallel ..." + chmod 0666 $dir/file-$((mmr + 1)) & + sleep 1 + + # check this additional modify RPC blocked getting a modify RPC slot + checkstat -vp 0644 $dir/file-$((mmr + 1)) || + error "Unexpectedly send $(($mmr + 1)) modify RPCs in parallel" + wait +} + +test_90a() { + reformat + if ! combined_mgs_mds ; then + start_mgs + fi + setup + + [[ $($LCTL get_param mdc.*.import | + grep "connect_flags:.*multi_mod_rpc") ]] || + { skip "Need MDC with 'multi_mod_rpcs' feature"; return 0; } + + # check default value + $LFS mkdir -c1 $DIR/$tdir || error "mkdir $DIR/$tdir failed" + check_max_mod_rpcs_in_flight $DIR/$tdir + + cleanup +} +run_test 90a "check max_mod_rpcs_in_flight is enforced" + +test_90b() { + local idx + local facet + local tmp + local mmrpc + + setup + + [[ $($LCTL get_param mdc.*.import | + grep "connect_flags:.*multi_mod_rpc") ]] || + { skip "Need MDC with 'multi_mod_rpcs' feature"; return 0; } + + ### test 1. + # update max_mod_rpcs_in_flight + $LFS mkdir -c1 $DIR/${tdir}1 || error "mkdir $DIR/${tdir}1 failed" + check_max_mod_rpcs_in_flight $DIR/${tdir}1 1 + + ### test 2. + # check client is able to send multiple modify RPCs in paralell + tmp=$($LCTL get_param -n mdc.$FSNAME-MDT*-mdc-*.import | + grep -c "multi_mod_rpcs") + if [ "$tmp" -ne $MDSCOUNT ]; then + echo "Client not able to send multiple modify RPCs in parallel" + cleanup + return + fi + + # update max_mod_rpcs_in_flight + $LFS mkdir -c1 $DIR/${tdir}2 || error "mkdir $DIR/${tdir}2 failed" + check_max_mod_rpcs_in_flight $DIR/${tdir}2 5 + + ### test 3. + $LFS mkdir -c1 $DIR/${tdir}3 || error "mkdir $DIR/${tdir}3 failed" + idx=$(printf "%04x" $($LFS getdirstripe -i $DIR/${tdir}3)) + facet="mds$((0x$idx + 1))" + + # save MDT max_mod_rpcs_per_client + mmrpc=$(do_facet $facet \ + cat /sys/module/mdt/parameters/max_mod_rpcs_per_client) + + # update max_mod_rpcs_in_flight + umount_client $MOUNT + do_facet $facet \ + "echo 16 > /sys/module/mdt/parameters/max_mod_rpcs_per_client" + mount_client $MOUNT + $LCTL set_param mdc.$FSNAME-MDT$idx-mdc-*.max_rpcs_in_flight=17 + check_max_mod_rpcs_in_flight $DIR/${tdir}3 16 + + # restore MDT max_mod_rpcs_per_client initial value + do_facet $facet \ + "echo $mmrpc > /sys/module/mdt/parameters/max_mod_rpcs_per_client" + + rm -rf $DIR/${tdir}? + cleanup +} +run_test 90b "check max_mod_rpcs_in_flight is enforced after update" + +test_90c() { + local tmp + local mrif + local mmrpc + + setup + + [[ $($LCTL get_param mdc.*.import | + grep "connect_flags:.*multi_mod_rpc") ]] || + { skip "Need MDC with 'multi_mod_rpcs' feature"; return 0; } + + # check client is able to send multiple modify RPCs in paralell + tmp=$($LCTL get_param -n mdc.$FSNAME-MDT*-mdc-*.import | + grep -c "multi_mod_rpcs") + if [ "$tmp" -ne $MDSCOUNT ]; then + skip "Client not able to send multiple modify RPCs in parallel" + cleanup + return + fi + + # get max_rpcs_in_flight value + mrif=$($LCTL get_param -n mdc.$FSNAME-MDT0000-mdc-*.max_rpcs_in_flight) + echo "max_rpcs_in_flight is $mrif" + + # get MDT max_mod_rpcs_per_client + mmrpc=$(do_facet mds1 \ + cat /sys/module/mdt/parameters/max_mod_rpcs_per_client) + echo "max_mod_rpcs_per_client is $mmrpc" + + # testcase 1 + # attempt to set max_mod_rpcs_in_flight to max_rpcs_in_flight value + # prerequisite: set max_mod_rpcs_per_client to max_rpcs_in_flight value + umount_client $MOUNT + do_facet mds1 \ + "echo $mrif > /sys/module/mdt/parameters/max_mod_rpcs_per_client" + mount_client $MOUNT + + $LCTL set_param \ + mdc.$FSNAME-MDT0000-mdc-*.max_mod_rpcs_in_flight=$mrif && + error "set max_mod_rpcs_in_flight to $mrif should fail" + + umount_client $MOUNT + do_facet mds1 \ + "echo $mmrpc > /sys/module/mdt/parameters/max_mod_rpcs_per_client" + mount_client $MOUNT + + # testcase 2 + # attempt to set max_mod_rpcs_in_flight to max_mod_rpcs_per_client+1 + # prerequisite: set max_rpcs_in_flight to max_mod_rpcs_per_client+2 + $LCTL set_param \ + mdc.$FSNAME-MDT0000-mdc-*.max_rpcs_in_flight=$((mmrpc + 2)) + + $LCTL set_param \ + mdc.$FSNAME-MDT0000-mdc-*.max_mod_rpcs_in_flight=$((mmrpc + 1)) && + error "set max_mod_rpcs_in_flight to $((mmrpc + 1)) should fail" + + cleanup +} +run_test 90c "check max_mod_rpcs_in_flight update limits" + +test_90d() { + local idx + local facet + local mmr + local i + local pid + + setup + + [[ $($LCTL get_param mdc.*.import | + grep "connect_flags:.*multi_mod_rpc") ]] || + { skip "Need MDC with 'multi_mod_rpcs' feature"; return 0; } + + $LFS mkdir -c1 $DIR/$tdir || error "mkdir $DIR/$tdir failed" + idx=$(printf "%04x" $($LFS getdirstripe -i $DIR/$tdir)) + facet="mds$((0x$idx + 1))" + + # check client version supports multislots + tmp=$($LCTL get_param -N \ + mdc.$FSNAME-MDT$idx-mdc-*.max_mod_rpcs_in_flight) + if [ -z "$tmp" ]; then + skip "Client does not support multiple modify RPCs in flight" + cleanup + return + fi + + # get current value of max_mod_rcps_in_flight + mmr=$($LCTL get_param -n \ + mdc.$FSNAME-MDT$idx-mdc-*.max_mod_rpcs_in_flight) + echo "max_mod_rcps_in_flight is $mmr" + + # create mmr files + echo "creating $mmr files ..." + umask 0022 + for i in $(seq $mmr); do + touch $DIR/$tdir/file-$i + done + + # prepare for close RPC + multiop_bg_pause $DIR/$tdir/file-close O_c + pid=$! + + # consumes mmr modify RPC slots + #define OBD_FAIL_MDS_REINT_MULTI_NET 0x159 + # drop requests on MDT so that RPC slots are consumed + # during all the request resend interval + do_facet $facet "$LCTL set_param fail_loc=0x159" + echo "launch $mmr chmod in parallel ..." + for i in $(seq $mmr); do + chmod 0600 $DIR/$tdir/file-$i & + done + + # send one additional close RPC + do_facet $facet "$LCTL set_param fail_loc=0" + echo "launch 1 additional close in parallel ..." + kill -USR1 $pid + cancel_lru_locks mdc + sleep 1 + + # check this additional close RPC get a modify RPC slot + # and multiop process completed + [ -d /proc/$pid ] && + error "Unable to send the additional close RPC in parallel" + wait + rm -rf $DIR/$tdir + cleanup +} +run_test 90d "check one close RPC is allowed above max_mod_rpcs_in_flight" + +check_uuid_on_ost() { + local nid=$1 + do_facet ost1 "$LCTL get_param obdfilter.${FSNAME}*.exports.'$nid'.uuid" +} + +check_uuid_on_mdt() { + local nid=$1 + do_facet $SINGLEMDS "$LCTL get_param mdt.${FSNAME}*.exports.'$nid'.uuid" +} + +test_91() { + local uuid + local nid + local found + + load_modules + + [[ $(lustre_version_code ost1) -ge $(version_code 2.7.63) ]] || + { skip "Need OST version at least 2.7.63" && return 0; } + [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.7.63) ]] || + { skip "Need MDT version at least 2.7.63" && return 0; } + + start_mds || error "MDS start failed" + start_ost || error "unable to start OST" + mount_client $MOUNT || error "client start failed" + check_mount || error "check_mount failed" + + if remote_mds; then + nid=$($LCTL list_nids | head -1 | sed "s/\./\\\./g") + else + nid="0@lo" + fi + uuid=$(get_client_uuid $MOUNT) + + echo "list nids on mdt:" + do_facet $SINGLEMDS "$LCTL list_param mdt.${FSNAME}*.exports.*" + echo "uuid from $nid:" + do_facet $SINGLEMDS "$LCTL get_param mdt.${FSNAME}*.exports.'$nid'.uuid" + + found=$(check_uuid_on_mdt $nid | grep $uuid) + [ -z "$found" ] && error "can't find $uuid $nid on MDT" + found=$(check_uuid_on_ost $nid | grep $uuid) + [ -z "$found" ] && error "can't find $uuid $nid on OST" + + # umount the client so it won't reconnect + manual_umount_client --force || error "failed to umount $?" + # shouldn't disappear on MDS after forced umount + found=$(check_uuid_on_mdt $nid | grep $uuid) + [ -z "$found" ] && error "can't find $uuid $nid" + + echo "evict $nid" + do_facet $SINGLEMDS \ + "$LCTL set_param -n mdt.${mds1_svc}.evict_client nid:$nid" + + found=$(check_uuid_on_mdt $nid | grep $uuid) + [ -n "$found" ] && error "found $uuid $nid on MDT" + found=$(check_uuid_on_ost $nid | grep $uuid) + [ -n "$found" ] && error "found $uuid $nid on OST" + + # check it didn't reconnect (being umounted) + sleep $((TIMEOUT+1)) + found=$(check_uuid_on_mdt $nid | grep $uuid) + [ -n "$found" ] && error "found $uuid $nid on MDT" + found=$(check_uuid_on_ost $nid | grep $uuid) + [ -n "$found" ] && error "found $uuid $nid on OST" + + cleanup +} +run_test 91 "evict-by-nid support" if ! combined_mgs_mds ; then stop mgs