X-Git-Url: https://git.whamcloud.com/?p=fs%2Flustre-release.git;a=blobdiff_plain;f=lustre%2Ftests%2Fconf-sanity.sh;h=4e60922cb689b1f5a62d7abae09dc2c7427ec9a3;hp=02434ced48e5fa2db0ad3a105fcece0b28754331;hb=7136981bfaad8b95e15e9d2969b94c933491eb98;hpb=26bb579a96efced8825b63b84465334368ecc382 diff --git a/lustre/tests/conf-sanity.sh b/lustre/tests/conf-sanity.sh index 02434ce..4e60922 100644 --- a/lustre/tests/conf-sanity.sh +++ b/lustre/tests/conf-sanity.sh @@ -1,6 +1,4 @@ #!/bin/bash -# -*- mode: Bash; tab-width: 4; indent-tabs-mode: t; -*- -# vim:autoindent:shiftwidth=4:tabstop=4: # FIXME - there is no reason to use all of these different # return codes, espcially when most of them are mapped to something @@ -11,17 +9,39 @@ set -e ONLY=${ONLY:-"$*"} -# bug number for skipped test: -# 15977 -ALWAYS_EXCEPT="$CONF_SANITY_EXCEPT" +# bug number for skipped test: LU-2828 +ALWAYS_EXCEPT="$CONF_SANITY_EXCEPT 59 64" # UPDATE THE COMMENT ABOVE WITH BUG NUMBERS WHEN CHANGING ALWAYS_EXCEPT! +is_sles11() # LU-2181 +{ + if [ -r /etc/SuSE-release ] + then + local vers=`grep VERSION /etc/SuSE-release | awk '{print $3}'` + local patchlev=`grep PATCHLEVEL /etc/SuSE-release \ + | awk '{print $3}'` + if [ $vers -eq 11 ] && [ $patchlev -eq 2 ] + then + return 0 + fi + fi + return 1 +} + +if is_sles11; then # LU-2181 + ALWAYS_EXCEPT="$ALWAYS_EXCEPT 23a 34b" +fi + if [ "$FAILURE_MODE" = "HARD" ]; then CONFIG_EXCEPTIONS="24a " && \ echo "Except the tests: $CONFIG_EXCEPTIONS for FAILURE_MODE=$FAILURE_MODE, bug 23573" && \ ALWAYS_EXCEPT="$ALWAYS_EXCEPT $CONFIG_EXCEPTIONS" fi +# bug number for skipped test: +# a tool to create lustre filesystem images +ALWAYS_EXCEPT="32newtarball $ALWAYS_EXCEPT" + SRCDIR=`dirname $0` PATH=$PWD/$SRCDIR:$SRCDIR:$SRCDIR/../utils:$PATH @@ -29,6 +49,7 @@ PTLDEBUG=${PTLDEBUG:--1} SAVE_PWD=$PWD LUSTRE=${LUSTRE:-`dirname $0`/..} RLUSTRE=${RLUSTRE:-$LUSTRE} +export MULTIOP=${MULTIOP:-multiop} . $LUSTRE/tests/test-framework.sh init_test_env $@ @@ -50,18 +71,13 @@ if [ -n "$MDSSIZE" ]; then fi # pass "-E lazy_itable_init" to mke2fs to speed up the formatting time -for facet in MGS MDS OST; do - opts=${facet}_MKFS_OPTS - if [[ ${!opts} != *lazy_itable_init* ]]; then - eval SAVED_${facet}_MKFS_OPTS=\"${!opts}\" - if [[ ${!opts} != *mkfsoptions* ]]; then - eval ${facet}_MKFS_OPTS=\"${!opts} --mkfsoptions='\\\"-E lazy_itable_init\\\"'\" - else - val=${!opts//--mkfsoptions=\\\"/--mkfsoptions=\\\"-E lazy_itable_init } - eval ${facet}_MKFS_OPTS='${val}' - fi - fi -done +if [[ "$LDISKFS_MKFS_OPTS" != *lazy_itable_init* ]]; then + LDISKFS_MKFS_OPTS=$(csa_add "$LDISKFS_MKFS_OPTS" -E lazy_itable_init) +fi + +[ $(facet_fstype $SINGLEMDS) = "zfs" ] && +# bug number for skipped test: LU-2778 LU-2059 + ALWAYS_EXCEPT="$ALWAYS_EXCEPT 57b 50h" init_logging @@ -74,31 +90,6 @@ require_dsh_ost || exit 0 assert_DIR -reformat() { - formatall -} - -writeconf1() { - local facet=$1 - local dev=$2 - - stop ${facet} -f - rm -f ${facet}active - # who knows if/where $TUNEFS is installed? Better reformat if it fails... - do_facet ${facet} "$TUNEFS --quiet --writeconf $dev" || - { echo "tunefs failed, reformatting instead" && reformat_and_config && return 1; } - return 0 -} - -writeconf() { - # we need ldiskfs - load_modules - # if writeconf fails anywhere, we reformat everything - writeconf1 mds `mdsdevname 1` || return 0 - writeconf1 ost1 `ostdevname 1` || return 0 - writeconf1 ost2 `ostdevname 2` || return 0 -} - gen_config() { # The MGS must be started before the OSTs for a new fs, so start # and stop to generate the startup logs. @@ -117,20 +108,54 @@ reformat_and_config() { gen_config } +writeconf_or_reformat() { + # There are at most 2 OSTs for write_conf test + # who knows if/where $TUNEFS is installed? + # Better reformat if it fails... + writeconf_all $MDSCOUNT 2 || + { echo "tunefs failed, reformatting instead" && + reformat_and_config && return 1; } + return 0 +} + +reformat() { + formatall +} + start_mgs () { echo "start mgs" start mgs $MGSDEV $MGS_MOUNT_OPTS } -start_mds() { - local facet=$SINGLEMDS - # we can not use MDSDEV1 here because SINGLEMDS could be set not to mds1 only - local num=$(echo $facet | tr -d "mds") +start_mdt() { + local num=$1 + local facet=mds$num local dev=$(mdsdevname $num) + shift 1 + echo "start mds service on `facet_active_host $facet`" start $facet ${dev} $MDS_MOUNT_OPTS $@ || return 94 } +stop_mdt() { + local num=$1 + local facet=mds$num + local dev=$(mdsdevname $num) + shift 1 + + echo "stop mds service on `facet_active_host $facet`" + # These tests all use non-failover stop + stop $facet -f || return 97 +} + +start_mds() { + local num + + for num in $(seq $MDSCOUNT); do + start_mdt $num $@ || return 94 + done +} + start_mgsmds() { if ! combined_mgs_mds ; then start_mgs @@ -139,9 +164,10 @@ start_mgsmds() { } stop_mds() { - echo "stop mds service on `facet_active_host $SINGLEMDS`" - # These tests all use non-failover stop - stop $SINGLEMDS -f || return 97 + local num + for num in $(seq $MDSCOUNT); do + stop_mdt $num || return 97 + done } stop_mgs() { @@ -288,9 +314,9 @@ test_1() { run_test 1 "start up ost twice (should return errors)" test_2() { - start_mds + start_mdt 1 echo "start mds second time.." - start_mds && error "2nd MDT start should fail" + start_mdt 1 && error "2nd MDT start should fail" start_ost mount_client $MOUNT check_mount || return 43 @@ -425,6 +451,9 @@ test_5d() { grep " $MOUNT " /etc/mtab && \ error false "unexpected entry in mtab before mount" && return 10 + [ "$(facet_fstype ost1)" = "zfs" ] && + skip "LU-2059: no local config for ZFS OSTs" && return + local rc=0 start_ost start_mds @@ -569,6 +598,11 @@ is_blkdev () { # test_17() { + if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then + skip "Only applicable to ldiskfs-based MDTs" + return + fi + setup check_mount || return 41 cleanup || return $? @@ -591,7 +625,10 @@ test_17() { run_test 17 "Verify failed mds_postsetup won't fail assertion (2936) (should return errs)" test_18() { - [ "$FSTYPE" != "ldiskfs" ] && skip "not needed for FSTYPE=$FSTYPE" && return + if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then + skip "Only applicable to ldiskfs-based MDTs" + return + fi local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/}) @@ -608,8 +645,14 @@ test_18() { log "use STORED_MDSSIZE=$STORED_MDSSIZE" # check if the block device is large enough - [ -z "$OK" ] && $(is_blkdev $SINGLEMDS $MDSDEV $MIN) && OK=1 && - myMDSSIZE=$MIN && log "use device $MDSDEV with MIN=$MIN" + is_blkdev $SINGLEMDS $MDSDEV $MIN + local large_enough=$? + if [ -n "$OK" ]; then + [ $large_enough -ne 0 ] && OK="" + else + [ $large_enough -eq 0 ] && OK=1 && myMDSSIZE=$MIN && + log "use device $MDSDEV with MIN=$MIN" + fi # check if a loopback device has enough space for fs metadata (5%) @@ -625,15 +668,9 @@ test_18() { echo "mount mds with large journal..." - local OLD_MDS_MKFS_OPTS=$MDS_MKFS_OPTS - local opts="--mdt --fsname=$FSNAME --device-size=$myMDSSIZE --param sys.timeout=$TIMEOUT $MDSOPT" - - if combined_mgs_mds ; then - MDS_MKFS_OPTS="--mgs $opts" - else - MDS_MKFS_OPTS="--mgsnode=$MGSNID $opts" - fi + local OLD_MDSSIZE=$MDSSIZE + MDSSIZE=$myMDSSIZE reformat_and_config echo "mount lustre system..." @@ -650,8 +687,8 @@ test_18() { cleanup || return $? - MDS_MKFS_OPTS=$OLD_MDS_MKFS_OPTS - reformat_and_config + MDSSIZE=$OLD_MDSSIZE + reformat_and_config } run_test 18 "check mkfs creates large journals" @@ -662,6 +699,9 @@ test_19a() { run_test 19a "start/stop MDS without OSTs" test_19b() { + [ "$(facet_fstype ost1)" = "zfs" ] && + skip "LU-2059: no local config for ZFS OSTs" && return + start_ost || return 1 stop_ost -f || return 2 } @@ -698,6 +738,9 @@ test_21a() { run_test 21a "start mds before ost, stop ost first" test_21b() { + [ "$(facet_fstype ost1)" = "zfs" ] && + skip "LU-2059: no local config for ZFS OSTs" && return + start_ost start_mds wait_osc_import_state mds ost FULL @@ -715,7 +758,7 @@ test_21c() { stop_ost2 stop_mds #writeconf to remove all ost2 traces for subsequent tests - writeconf + writeconf_or_reformat } run_test 21c "start mds between two osts, stop mds last" @@ -738,18 +781,22 @@ test_21d() { stop_mds stop_mgs #writeconf to remove all ost2 traces for subsequent tests - writeconf + writeconf_or_reformat start_mgs } run_test 21d "start mgs then ost and then mds" test_22() { + local num + start_mds echo Client mount with ost in logs, but none running start_ost # wait until mds connected to ost and open client connection - wait_osc_import_state mds ost FULL + for num in $(seq 1 $MDSCOUNT); do + wait_osc_import_state mds${num} ost FULL + done stop_ost mount_client $MOUNT # check_mount will block trying to contact ost @@ -769,8 +816,10 @@ test_22() { sleep $((TIMEOUT + TIMEOUT + TIMEOUT)) fi mount_client $MOUNT - wait_osc_import_state mds ost FULL - wait_osc_import_state client ost FULL + for num in $(seq 1 $MDSCOUNT); do + wait_osc_import_state mds${num} ost FULL + done + wait_osc_import_state client ost FULL check_mount || return 41 pass @@ -847,6 +896,10 @@ 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_24a() { trap 0 echo "umount $MOUNT2 ..." @@ -867,14 +920,19 @@ test_24a() { [ -n "$ost1_HOST" ] && fs2ost_HOST=$ost1_HOST - local fs2mdsdev=${fs2mds_DEV:-${MDSDEV}_2} - local fs2ostdev=${fs2ost_DEV:-$(ostdevname 1)_2} + local fs2mdsdev=$(mdsdevname 1_2) + local fs2ostdev=$(ostdevname 1_2) + local fs2mdsvdev=$(mdsvdevname 1_2) + local fs2ostvdev=$(ostvdevname 1_2) # test 8-char fsname as well local FSNAME2=test1234 - add fs2mds $MDS_MKFS_OPTS --fsname=${FSNAME2} --nomgs --mgsnode=$MGSNID --reformat $fs2mdsdev || exit 10 - add fs2ost $OST_MKFS_OPTS --fsname=${FSNAME2} --reformat $fs2ostdev || exit 10 + add fs2mds $(mkfs_opts mds1 ${fs2mdsdev} ) --nomgs --mgsnode=$MGSNID \ + --fsname=${FSNAME2} --reformat $fs2mdsdev $fs2mdsvdev || exit 10 + + add fs2ost $(mkfs_opts ost1 ${fs2ostdev}) --fsname=${FSNAME2} \ + --reformat $fs2ostdev $fs2ostvdev || exit 10 setup start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS && trap cleanup_24a EXIT INT @@ -917,9 +975,11 @@ test_24b() { skip_env "mixed loopback and real device not working" && return fi - local fs2mdsdev=${fs2mds_DEV:-${MDSDEV}_2} + local fs2mdsdev=$(mdsdevname 1_2) + local fs2mdsvdev=$(mdsvdevname 1_2) - add fs2mds $MDS_MKFS_OPTS --fsname=${FSNAME}2 --mgs --reformat $fs2mdsdev || exit 10 + add fs2mds $(mkfs_opts mds1 ${fs2mdsdev} ) --mgs --fsname=${FSNAME}2 \ + --reformat $fs2mdsdev $fs2mdsvdev || exit 10 setup start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS && return 2 cleanup || return 6 @@ -945,45 +1005,41 @@ test_26() { lctl get_param -n devices DEVS=$(lctl get_param -n devices | egrep -v MG | wc -l) [ $DEVS -gt 0 ] && return 2 + # start mds to drop writeconf setting + start_mds || return 3 + stop_mds || return 4 unload_modules_conf || return $? } run_test 26 "MDT startup failure cleans LOV (should return errs)" -set_and_check() { - local myfacet=$1 - local TEST=$2 - local PARAM=$3 - local ORIG=$(do_facet $myfacet "$TEST") - if [ $# -gt 3 ]; then - local FINAL=$4 - else - local -i FINAL - FINAL=$(($ORIG + 5)) - fi - echo "Setting $PARAM from $ORIG to $FINAL" - do_facet mgs "$LCTL conf_param $PARAM='$FINAL'" || error conf_param failed - - wait_update $(facet_host $myfacet) "$TEST" "$FINAL" || error check failed! -} - test_27a() { + [ "$(facet_fstype ost1)" = "zfs" ] && + skip "LU-2059: no local config for ZFS OSTs" && return + start_ost || return 1 start_mds || return 2 echo "Requeue thread should have started: " ps -e | grep ll_cfg_requeue - set_and_check ost1 "lctl get_param -n obdfilter.$FSNAME-OST0000.client_cache_seconds" "$FSNAME-OST0000.ost.client_cache_seconds" || return 3 + set_conf_param_and_check ost1 \ + "lctl get_param -n obdfilter.$FSNAME-OST0000.client_cache_seconds" \ + "$FSNAME-OST0000.ost.client_cache_seconds" || return 3 cleanup_nocli } run_test 27a "Reacquire MGS lock if OST started first" test_27b() { # FIXME. ~grev - setup - local device=$(do_facet $SINGLEMDS "lctl get_param -n devices" | awk '($3 ~ "mdt" && $4 ~ "MDT") { print $4 }') + setup + local device=$(do_facet $SINGLEMDS "lctl get_param -n devices" | + awk '($3 ~ "mdt" && $4 ~ "MDT0000") { print $4 }') facet_failover $SINGLEMDS - set_and_check $SINGLEMDS "lctl get_param -n mdt.$device.identity_acquire_expire" "$device.mdt.identity_acquire_expire" || return 3 - set_and_check client "lctl get_param -n mdc.$device-mdc-*.max_rpcs_in_flight" "$device.mdc.max_rpcs_in_flight" || return 4 + set_conf_param_and_check $SINGLEMDS \ + "lctl get_param -n mdt.$device.identity_acquire_expire" \ + "$device.mdt.identity_acquire_expire" || return 3 + set_conf_param_and_check client \ + "lctl get_param -n mdc.$device-mdc-*.max_rpcs_in_flight"\ + "$device.mdc.max_rpcs_in_flight" || return 4 check_mount cleanup } @@ -995,9 +1051,9 @@ test_28() { PARAM="$FSNAME.llite.max_read_ahead_whole_mb" ORIG=$($TEST) FINAL=$(($ORIG + 1)) - set_and_check client "$TEST" "$PARAM" $FINAL || return 3 + set_conf_param_and_check client "$TEST" "$PARAM" $FINAL || return 3 FINAL=$(($FINAL + 1)) - set_and_check client "$TEST" "$PARAM" $FINAL || return 4 + set_conf_param_and_check client "$TEST" "$PARAM" $FINAL || return 4 umount_client $MOUNT || return 200 mount_client $MOUNT RESULT=$($TEST) @@ -1007,7 +1063,7 @@ test_28() { else echo "New config success: got $RESULT" fi - set_and_check client "$TEST" "$PARAM" $ORIG || return 5 + set_conf_param_and_check client "$TEST" "$PARAM" $ORIG || return 5 cleanup } run_test 28 "permanent parameter setting" @@ -1024,7 +1080,8 @@ test_29() { ACTV=$(lctl get_param -n $PROC_ACT) DEAC=$((1 - $ACTV)) - set_and_check client "lctl get_param -n $PROC_ACT" "$PARAM" $DEAC || return 2 + set_conf_param_and_check client \ + "lctl get_param -n $PROC_ACT" "$PARAM" $DEAC || return 2 # also check ost_server_uuid status RESULT=$(lctl get_param -n $PROC_UUID | grep DEACTIV) if [ -z "$RESULT" ]; then @@ -1034,31 +1091,30 @@ test_29() { echo "Live client success: got $RESULT" fi - # check MDT too - local mdtosc=$(get_mdtosc_proc_path $SINGLEMDS $FSNAME-OST0001) - mdtosc=${mdtosc/-MDT*/-MDT\*} - local MPROC="osc.$mdtosc.active" - local MAX=30 - local WAIT=0 - while [ 1 ]; do - sleep 5 - RESULT=`do_facet $SINGLEMDS " lctl get_param -n $MPROC"` - [ ${PIPESTATUS[0]} = 0 ] || error "Can't read $MPROC" - if [ $RESULT -eq $DEAC ]; then - echo "MDT deactivated also after $WAIT sec (got $RESULT)" - break - fi - WAIT=$((WAIT + 5)) - if [ $WAIT -eq $MAX ]; then - echo "MDT not deactivated: wanted $DEAC got $RESULT" - return 4 - fi - echo "Waiting $(($MAX - $WAIT)) secs for MDT deactivated" + # check MDTs too + for num in $(seq $MDSCOUNT); do + local mdtosc=$(get_mdtosc_proc_path mds${num} $FSNAME-OST0001) + local MPROC="osc.$mdtosc.active" + local MAX=30 + local WAIT=0 + while [ 1 ]; do + sleep 5 + RESULT=$(do_facet mds${num} " lctl get_param -n $MPROC") + [ ${PIPESTATUS[0]} = 0 ] || error "Can't read $MPROC" + if [ $RESULT -eq $DEAC ]; then + echo -n "MDT deactivated also after" + echo "$WAIT sec (got $RESULT)" + break + fi + WAIT=$((WAIT + 5)) + if [ $WAIT -eq $MAX ]; then + echo -n "MDT not deactivated: wanted $DEAC" + echo "got $RESULT" + return 4 + fi + echo "Waiting $(($MAX - $WAIT))secs for MDT deactivated" + done done - - # quotacheck should not fail immediately after deactivate - [ -n "$ENABLE_QUOTA" ] && { $LFS quotacheck -ug $MOUNT || error "quotacheck has failed" ; } - # test new client starts deactivated umount_client $MOUNT || return 200 mount_client $MOUNT @@ -1070,17 +1126,15 @@ test_29() { echo "New client success: got $RESULT" fi - # quotacheck should not fail after umount/mount operation - [ -n "$ENABLE_QUOTA" ] && { $LFS quotacheck -ug $MOUNT || error "quotacheck has failed" ; } - # make sure it reactivates - set_and_check client "lctl get_param -n $PROC_ACT" "$PARAM" $ACTV || return 6 + set_conf_param_and_check client \ + "lctl get_param -n $PROC_ACT" "$PARAM" $ACTV || return 6 umount_client $MOUNT stop_ost2 cleanup_nocli #writeconf to remove all ost2 traces for subsequent tests - writeconf + writeconf_or_reformat } run_test 29 "permanently remove an OST" @@ -1092,7 +1146,8 @@ test_30a() { ORIG=$($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 ${LIST[@]}; do - set_and_check client "$TEST" "$FSNAME.llite.max_read_ahead_whole_mb" $i || return 3 + set_conf_param_and_check client "$TEST" \ + "$FSNAME.llite.max_read_ahead_whole_mb" $i || return 3 done # make sure client restart still works umount_client $MOUNT @@ -1126,7 +1181,9 @@ test_30b() { echo "Using fake nid $NEW" TEST="$LCTL get_param -n osc.$FSNAME-OST0000-osc-[^M]*.import | grep failover_nids | sed -n 's/.*\($NEW\).*/\1/p'" - set_and_check client "$TEST" "$FSNAME-OST0000.failover.node" $NEW || error "didn't add failover nid $NEW" + set_conf_param_and_check client "$TEST" \ + "$FSNAME-OST0000.failover.node" $NEW || + error "didn't add failover nid $NEW" NIDS=$($LCTL get_param -n osc.$FSNAME-OST0000-osc-[^M]*.import | grep failover_nids) echo $NIDS NIDCOUNT=$(($(echo "$NIDS" | wc -w) - 1)) @@ -1153,194 +1210,545 @@ test_31() { # bug 10734 } run_test 31 "Connect to non-existent node (shouldn't crash)" -# Use these start32/stop32 fn instead of t-f start/stop fn, -# for local devices, to skip global facet vars init -stop32 () { - local facet=$1 - shift - echo "Stopping local ${MOUNT%/*}/${facet} (opts:$@)" - umount -d $@ ${MOUNT%/*}/${facet} - losetup -a -} - -start32 () { - local facet=$1 - shift - local device=$1 - shift - mkdir -p ${MOUNT%/*}/${facet} - - echo "Starting local ${facet}: $@ $device ${MOUNT%/*}/${facet}" - mount -t lustre $@ ${device} ${MOUNT%/*}/${facet} - local RC=$? - if [ $RC -ne 0 ]; then - echo "mount -t lustre $@ ${device} ${MOUNT%/*}/${facet}" - echo "Start of ${device} of local ${facet} failed ${RC}" - fi - losetup -a - return $RC -} +# +# This is not really a test but a tool to create new disk +# image tarballs for the upgrade tests. +# +# Disk image tarballs should be created on single-node +# clusters by running this test with default configurations +# plus a few mandatory environment settings that are verified +# at the beginning of the test. +# +test_32newtarball() { + local version + local dst=. + local src=/etc/rc.d + local tmp=$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=" + fi -cleanup_nocli32 () { - stop32 mds1 -f - stop32 ost1 -f - wait_exit_ST client -} + mkdir $tmp || { + echo "Found stale $tmp" + return 1 + } -cleanup_32() { - trap 0 - echo "Cleanup test_32 umount $MOUNT ..." - umount -f $MOUNT || true - echo "Cleanup local mds ost1 ..." - cleanup_nocli32 - combined_mgs_mds || start_mgs - unload_modules_conf -} + mkdir $tmp/src + tar cf - -C $src . | tar xf - -C $tmp/src -test_32a() { - client_only && skip "client only testing" && return 0 - [ "$NETTYPE" = "tcp" ] || { skip "NETTYPE != tcp" && return 0; } - [ -z "$TUNEFS" ] && skip_env "No tunefs" && return 0 + formatall + + setupall + tar cf - -C $tmp/src . | tar xf - -C /mnt/$FSNAME + stopall + + mkdir $tmp/img + + setupall + pushd /mnt/$FSNAME + ls -Rni --time-style=+%s >$tmp/img/list + find . ! -name .lustre -type f -exec sha1sum {} \; | + sort -k 2 >$tmp/img/sha1sums + popd + $LCTL get_param -n version | head -n 1 | + sed -e 's/^lustre: *//' >$tmp/img/commit + stopall - local DISK1_8=$LUSTRE/tests/disk1_8.tar.bz2 - [ ! -r $DISK1_8 ] && skip_env "Cannot find $DISK1_8" && return 0 - local tmpdir=$TMP/conf32a - mkdir -p $tmpdir - tar xjvf $DISK1_8 -C $tmpdir || \ - { skip_env "Cannot untar $DISK1_8" && return 0; } + pushd $tmp/src + find -type f -exec sha1sum {} \; | sort -k 2 >$tmp/sha1sums.src + popd - load_modules - $LCTL set_param debug="$PTLDEBUG" + if ! diff -u $tmp/sha1sums.src $tmp/img/sha1sums; then + echo "Data verification failed" + fi - $TUNEFS $tmpdir/mds || error "tunefs failed" + uname -r >$tmp/img/kernel + uname -m >$tmp/img/arch - combined_mgs_mds || stop mgs + mv ${MDSDEV1:-$MDSDEV} $tmp/img + mv $OSTDEV1 $tmp/img - # nids are wrong, so client wont work, but server should start - start32 mds1 $tmpdir/mds "-o loop,exclude=lustre-OST0000" && \ - trap cleanup_32 EXIT INT || return 3 + version=$(sed -e 's/\(^[0-9]\+\.[0-9]\+\)\(.*$\)/\1/' $tmp/img/commit | + sed -e 's/\./_/g') # E.g., "1.8.7" -> "1_8" + dst=$(cd $dst; pwd) + pushd $tmp/img + tar cjvf $dst/disk$version-$(facet_fstype $SINGLEMDS).tar.bz2 -S * + popd - local UUID=$($LCTL get_param -n mdt.lustre-MDT0000.uuid) - echo MDS uuid $UUID - [ "$UUID" == "lustre-MDT0000_UUID" ] || error "UUID is wrong: $UUID" + rm -r $tmp +} +#run_test 32newtarball "Create a new test_32 disk image tarball for this version" - $TUNEFS --mgsnode=$HOSTNAME $tmpdir/ost1 || error "tunefs failed" - start32 ost1 $tmpdir/ost1 "-o loop" || return 5 - UUID=$($LCTL get_param -n obdfilter.lustre-OST0000.uuid) - echo OST uuid $UUID - [ "$UUID" == "lustre-OST0000_UUID" ] || error "UUID is wrong: $UUID" +# +# The list of applicable tarballs is returned via the caller's +# variable "tarballs". +# +t32_check() { + local node=$(facet_active_host $SINGLEMDS) + local r="do_node $node" - local NID=$($LCTL list_nids | head -1) + if [ "$CLIENTONLY" ]; then + skip "Client-only testing" + exit 0 + fi - echo "OSC changes should succeed:" - $LCTL conf_param lustre-OST0000.osc.max_dirty_mb=15 || return 7 - $LCTL conf_param lustre-OST0000.failover.node=$NID || return 8 - echo "ok." + if ! $r which $TUNEFS; then + skip_env "tunefs.lustre required on $node" + exit 0 + fi - echo "MDC changes should succeed:" - $LCTL conf_param lustre-MDT0000.mdc.max_rpcs_in_flight=9 || return 9 - $LCTL conf_param lustre-MDT0000.failover.node=$NID || return 10 - echo "ok." + if [ -n "$($LCTL list_nids | grep -v '\(tcp\|lo\)[[:digit:]]*$')" ]; then + skip "LU-2200: Test cannot run over Infiniband" + exit 0 + fi - echo "LOV changes should succeed:" - $LCTL pool_new lustre.interop || return 11 - $LCTL conf_param lustre-MDT0000.lov.stripesize=4M || return 12 - echo "ok." + local IMGTYPE=$(facet_fstype $SINGLEMDS) - cleanup_32 + tarballs=$($r find $RLUSTRE/tests -maxdepth 1 -name \'disk*-$IMGTYPE.tar.bz2\') - # mount a second time to make sure we didnt leave upgrade flag on - load_modules - $TUNEFS --dryrun $tmpdir/mds || error "tunefs failed" + if [ -z "$tarballs" ]; then + skip "No applicable tarballs found" + exit 0 + fi +} - combined_mgs_mds || stop mgs +t32_test_cleanup() { + local node=$(facet_active_host $SINGLEMDS) + local r="do_node $node" + local tmp=$TMP/t32 + local rc=$? - start32 mds1 $tmpdir/mds "-o loop,exclude=lustre-OST0000" && \ - trap cleanup_32 EXIT INT || return 12 + if $shall_cleanup_lustre; then + umount $tmp/mnt/lustre || rc=$? + fi + if $shall_cleanup_mdt; then + $r umount -d $tmp/mnt/mdt || rc=$? + fi + if $shall_cleanup_mdt1; then + $r umount -d $tmp/mnt/mdt1 || rc=$? + fi + if $shall_cleanup_ost; then + $r umount -d $tmp/mnt/ost || rc=$? + fi + $r rm -rf $tmp || rc=$? + rm -rf $tmp || rc=$? + return $rc +} - cleanup_32 +t32_bits_per_long() { + # + # Yes, this is not meant to be perfect. + # + case $1 in + ppc64|x86_64) + echo -n 64;; + i*86) + echo -n 32;; + esac +} - rm -rf $tmpdir || true # true is only for TMP on NFS +t32_reload_modules() { + local node=$1 + local all_removed=false + local i=0 + + while ((i < 20)); do + echo "Unloading modules on $node: Attempt $i" + do_rpc_nodes $node $LUSTRE_RMMOD $(facet_fstype $SINGLEMDS) && + all_removed=true + do_rpc_nodes $node check_mem_leak || return 1 + if $all_removed; then + load_modules + return 0 + fi + sleep 5 + i=$((i + 1)) + done + echo "Unloading modules on $node: Given up" + return 1 } -run_test 32a "Upgrade from 1.8 (not live)" -test_32b() { - client_only && skip "client only testing" && return 0 - [ "$NETTYPE" = "tcp" ] || { skip "NETTYPE != tcp" && return 0; } - [ -z "$TUNEFS" ] && skip_env "No tunefs" && return +t32_wait_til_devices_gone() { + local node=$1 + local devices + local i=0 + + echo wait for devices to go + while ((i < 20)); do + devices=$(do_rpc_nodes $node $LCTL device_list | wc -l) + echo $device + ((devices == 0)) && return 1 + sleep 5 + i=$((i + 1)) + done + echo "waiting for devices on $node: Given up" + return 1 +} + +t32_test() { + local tarball=$1 + local writeconf=$2 + local dne_upgrade=${dne_upgrade:-"no"} + local shall_cleanup_mdt=false + local shall_cleanup_mdt1=false + local shall_cleanup_ost=false + local shall_cleanup_lustre=false + local node=$(facet_active_host $SINGLEMDS) + local r="do_node $node" + local node2=$(facet_active_host mds2) + local r2="do_node $node2" + local tmp=$TMP/t32 + local img_commit + local img_kernel + local img_arch + local fsname=t32fs + local nid=$($r $LCTL list_nids | head -1) + local mopts + local uuid + local nrpcs_orig + local nrpcs + local list + + trap 'trap - RETURN; t32_test_cleanup' RETURN + + mkdir -p $tmp/mnt/lustre + $r mkdir -p $tmp/mnt/{mdt,ost} + $r tar xjvf $tarball -S -C $tmp || { + error_noexit "Unpacking the disk image tarball" + return 1 + } + img_commit=$($r cat $tmp/commit) + img_kernel=$($r cat $tmp/kernel) + img_arch=$($r cat $tmp/arch) + echo "Upgrading from $(basename $tarball), created with:" + echo " Commit: $img_commit" + echo " Kernel: $img_kernel" + echo " Arch: $img_arch" + + $r $LCTL set_param debug="$PTLDEBUG" + + $r $TUNEFS --dryrun $tmp/mdt || { + error_noexit "tunefs.lustre before mounting the MDT" + return 1 + } + if [ "$writeconf" ]; then + mopts=loop,writeconf + else + mopts=loop,exclude=$fsname-OST0000 + fi - local DISK1_8=$LUSTRE/tests/disk1_8.tar.bz2 - [ ! -r $DISK1_8 ] && skip_env "Cannot find $DISK1_8" && return 0 - local tmpdir=$TMP/conf32b - mkdir -p $tmpdir - tar xjvf $DISK1_8 -C $tmpdir || \ - { skip_env "Cannot untar $DISK1_8" && return ; } + t32_wait_til_devices_gone $node - load_modules - $LCTL set_param debug="config" - local NEWNAME=lustre + $r mount -t lustre -o $mopts $tmp/mdt $tmp/mnt/mdt || { + error_noexit "Mounting the MDT" + return 1 + } + shall_cleanup_mdt=true + + if [ "$dne_upgrade" != "no" ]; then + echo "mkfs new MDT...." + add mds2 $(mkfs_opts mds2 $(mdsdevname 2) $fsname) --reformat \ + $(mdsdevname 2) $(mdsvdevname 2) > /dev/null || { + error_noexit "Mkfs new MDT failed" + return 1 + } + + $r2 $TUNEFS --dryrun $(mdsdevname 2) || { + error_noexit "tunefs.lustre before mounting the MDT" + return 1 + } + + echo "mount new MDT...." + $r2 mkdir -p $tmp/mnt/mdt1 + $r2 mount -t lustre -o $mopts $(mdsdevname 2) $tmp/mnt/mdt1 || { + error_noexit "mount mdt1 failed" + return 1 + } + shall_cleanup_mdt1=true + fi - # writeconf will cause servers to register with their current nids - $TUNEFS --writeconf --fsname=$NEWNAME $tmpdir/mds || error "tunefs failed" - combined_mgs_mds || stop mgs + uuid=$($r $LCTL get_param -n mdt.$fsname-MDT0000.uuid) || { + error_noexit "Getting MDT UUID" + return 1 + } + if [ "$uuid" != $fsname-MDT0000_UUID ]; then + error_noexit "Unexpected MDT UUID: \"$uuid\"" + return 1 + fi - start32 mds1 $tmpdir/mds "-o loop" && \ - trap cleanup_32 EXIT INT || return 3 + $r $TUNEFS --dryrun $tmp/ost || { + error_noexit "tunefs.lustre before mounting the OST" + return 1 + } + if [ "$writeconf" ]; then + mopts=loop,mgsnode=$nid,$writeconf + else + mopts=loop,mgsnode=$nid + fi + $r mount -t lustre -o $mopts $tmp/ost $tmp/mnt/ost || { + error_noexit "Mounting the OST" + return 1 + } + shall_cleanup_ost=true - local UUID=$($LCTL get_param -n mdt.${NEWNAME}-MDT0000.uuid) - echo MDS uuid $UUID - [ "$UUID" == "${NEWNAME}-MDT0000_UUID" ] || error "UUID is wrong: $UUID" + uuid=$($r $LCTL get_param -n obdfilter.$fsname-OST0000.uuid) || { + error_noexit "Getting OST UUID" + return 1 + } + if [ "$uuid" != $fsname-OST0000_UUID ]; then + error_noexit "Unexpected OST UUID: \"$uuid\"" + return 1 + fi - $TUNEFS --mgsnode=$HOSTNAME --writeconf --fsname=$NEWNAME $tmpdir/ost1 ||\ - error "tunefs failed" - start32 ost1 $tmpdir/ost1 "-o loop" || return 5 - UUID=$($LCTL get_param -n obdfilter.${NEWNAME}-OST0000.uuid) - echo OST uuid $UUID - [ "$UUID" == "${NEWNAME}-OST0000_UUID" ] || error "UUID is wrong: $UUID" + $r $LCTL conf_param $fsname-OST0000.osc.max_dirty_mb=15 || { + error_noexit "Setting \"max_dirty_mb\"" + return 1 + } + $r $LCTL conf_param $fsname-OST0000.failover.node=$nid || { + error_noexit "Setting OST \"failover.node\"" + return 1 + } + $r $LCTL conf_param $fsname-MDT0000.mdc.max_rpcs_in_flight=9 || { + error_noexit "Setting \"max_rpcs_in_flight\"" + return 1 + } + $r $LCTL conf_param $fsname-MDT0000.failover.node=$nid || { + error_noexit "Setting MDT \"failover.node\"" + return 1 + } + $r $LCTL pool_new $fsname.interop || { + error_noexit "Setting \"interop\"" + return 1 + } + $r $LCTL conf_param $fsname-MDT0000.lov.stripesize=4M || { + error_noexit "Setting \"lov.stripesize\"" + return 1 + } + + if [ "$dne_upgrade" != "no" ]; then + $r2 $LCTL conf_param \ + $fsname-MDT0001.mdc.max_rpcs_in_flight=9 || { + error_noexit "Setting MDT1 \"max_rpcs_in_flight\"" + return 1 + } + $r2 $LCTL conf_param $fsname-MDT0001.failover.node=$nid || { + error_noexit "Setting MDT1 \"failover.node\"" + return 1 + } + $r2 $LCTL conf_param $fsname-MDT0001.lov.stripesize=4M || { + error_noexit "Setting MDT1 \"lov.stripesize\"" + return 1 + } - local NID=$($LCTL list_nids | head -1) + fi - echo "OSC changes should succeed:" - $LCTL conf_param ${NEWNAME}-OST0000.osc.max_dirty_mb=15 || return 7 - $LCTL conf_param ${NEWNAME}-OST0000.failover.node=$NID || return 8 - echo "ok." + if [ "$writeconf" ]; then + mount -t lustre $nid:/$fsname $tmp/mnt/lustre || { + error_noexit "Mounting the client" + return 1 + } + shall_cleanup_lustre=true + $LCTL set_param debug="$PTLDEBUG" + if [ "$dne_upgrade" != "no" ]; then + $LFS mkdir -i 1 $tmp/mnt/lustre/remote_dir || { + error_noexit "set remote dir failed" + return 1 + } + + pushd $tmp/mnt/lustre + tar -cf - . --exclude=./remote_dir | + tar -xvf - -C remote_dir 1>/dev/null || { + error_noexit "cp to remote dir failed" + return 1 + } + popd + fi + + dd if=/dev/zero of=$tmp/mnt/lustre/tmp_file bs=10k count=10 || { + error_noexit "dd failed" + return 1 + } + rm -rf $tmp/mnt/lustre/tmp_file || { + error_noexit "rm failed" + return 1 + } + + if $r test -f $tmp/sha1sums; then + # 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/remote_dir + else + pushd $tmp/mnt/lustre + fi + + find ! -name .lustre -type f -exec sha1sum {} \; | + sort -k 2 >$tmp/sha1sums || { + error_noexit "sha1sum" + return 1 + } + popd + if ! diff -ub $tmp/sha1sums.orig $tmp/sha1sums; then + error_noexit "sha1sum verification failed" + return 1 + fi + else + echo "sha1sum verification skipped" + fi + + if [ "$dne_upgrade" != "no" ]; then + rm -rf $tmp/mnt/lustre/remote_dir || { + error_noexit "remove remote dir failed" + return 1 + } + fi - echo "MDC changes should succeed:" - $LCTL conf_param ${NEWNAME}-MDT0000.mdc.max_rpcs_in_flight=9 || return 9 - $LCTL conf_param ${NEWNAME}-MDT0000.failover.node=$NID || return 10 - echo "ok." + if $r test -f $tmp/list; then + # + # There is not a Test Framework API to copy files to or + # from a remote node. + # + # LU-2393 - do both sorts on same node to ensure locale + # is identical + $r cat $tmp/list | sort -k 6 >$tmp/list.orig + pushd $tmp/mnt/lustre + ls -Rni --time-style=+%s | sort -k 6 >$tmp/list || { + error_noexit "ls" + return 1 + } + popd + # + # 32-bit and 64-bit clients use different algorithms to + # convert FIDs into inode numbers. Hence, remove the inode + # numbers from the lists, if the original list was created + # on an architecture with different number of bits per + # "long". + # + if [ $(t32_bits_per_long $(uname -m)) != \ + $(t32_bits_per_long $img_arch) ]; then + echo "Different number of bits per \"long\" from the disk image" + for list in list.orig list; do + sed -i -e 's/^[0-9]\+[ \t]\+//' $tmp/$list + done + fi + if ! diff -ub $tmp/list.orig $tmp/list; then + error_noexit "list verification failed" + return 1 + fi + else + echo "list verification skipped" + fi - echo "LOV changes should succeed:" - $LCTL pool_new ${NEWNAME}.interop || return 11 - $LCTL conf_param ${NEWNAME}-MDT0000.lov.stripesize=4M || return 12 - echo "ok." + # + # When adding new data verification tests, please check for + # the presence of the required reference files first, like + # the "sha1sums" and "list" tests above, to avoid the need to + # regenerate every image for each test addition. + # + + nrpcs_orig=$($LCTL get_param \ + -n mdc.*MDT0000*.max_rpcs_in_flight) || { + error_noexit "Getting \"max_rpcs_in_flight\"" + return 1 + } + nrpcs=$((nrpcs_orig + 5)) + $r $LCTL conf_param $fsname-MDT0000.mdc.max_rpcs_in_flight=$nrpcs || { + error_noexit "Changing \"max_rpcs_in_flight\"" + return 1 + } + wait_update $HOSTNAME "$LCTL get_param \ + -n mdc.*MDT0000*.max_rpcs_in_flight" $nrpcs || { + error_noexit "Verifying \"max_rpcs_in_flight\"" + return 1 + } + + umount $tmp/mnt/lustre || { + error_noexit "Unmounting the client" + return 1 + } + shall_cleanup_lustre=false + else + $r umount -d $tmp/mnt/mdt || { + error_noexit "Unmounting the MDT" + return 1 + } + shall_cleanup_mdt=false + $r umount -d $tmp/mnt/ost || { + error_noexit "Unmounting the OST" + return 1 + } + shall_cleanup_ost=false + + t32_reload_modules $node || { + error_noexit "Reloading modules" + return 1 + } + + # mount a second time to make sure we didnt leave upgrade flag on + $r $TUNEFS --dryrun $tmp/mdt || { + error_noexit "tunefs.lustre before remounting the MDT" + return 1 + } + $r mount -t lustre -o loop,exclude=$fsname-OST0000 $tmp/mdt \ + $tmp/mnt/mdt || { + error_noexit "Remounting the MDT" + return 1 + } + shall_cleanup_mdt=true + fi +} - # MDT and OST should have registered with new nids, so we should have - # a fully-functioning client - echo "Check client and old fs contents" +test_32a() { + local tarballs + local tarball + local rc=0 - local device=`h2$NETTYPE $HOSTNAME`:/$NEWNAME - echo "Starting local client: $HOSTNAME: $device $MOUNT" - mount -t lustre $device $MOUNT || return 1 + t32_check + for tarball in $tarballs; do + t32_test $tarball || rc=$? + done + return $rc +} +run_test 32a "Upgrade (not live)" - local old=$($LCTL get_param -n mdc.*.max_rpcs_in_flight) - local new=$((old + 5)) - $LCTL conf_param ${NEWNAME}-MDT0000.mdc.max_rpcs_in_flight=$new - wait_update $HOSTNAME "$LCTL get_param -n mdc.*.max_rpcs_in_flight" $new || return 11 +test_32b() { + local tarballs + local tarball + local rc=0 - [ "$(cksum $MOUNT/passwd | cut -d' ' -f 1,2)" == "94306271 1478" ] || return 12 - echo "ok." + t32_check + for tarball in $tarballs; do + t32_test $tarball writeconf || rc=$? + done + return $rc +} +run_test 32b "Upgrade with writeconf" - cleanup_32 +test_32c() { + local tarballs + local tarball + local rc=0 - rm -rf $tmpdir || true # true is only for TMP on NFS + [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return + t32_check + for tarball in $tarballs; do + dne_upgrade=yes t32_test $tarball writeconf || rc=$? + done + return $rc } -run_test 32b "Upgrade from 1.8 with writeconf" +run_test 32c "dne upgrade test" test_33a() { # bug 12333, was test_33 local rc=0 local FSNAME2=test-123 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/}) + local mkfsoptions [ -n "$ost1_HOST" ] && fs2ost_HOST=$ost1_HOST @@ -1351,12 +1759,20 @@ test_33a() { # bug 12333, was test_33 skip_env "mixed loopback and real device not working" && return fi - combined_mgs_mds || mkfs_opts="$mkfs_opts --nomgs" + local fs2mdsdev=$(mdsdevname 1_2) + local fs2ostdev=$(ostdevname 1_2) + local fs2mdsvdev=$(mdsvdevname 1_2) + local fs2ostvdev=$(ostvdevname 1_2) + + if [ $(facet_fstype mds1) == ldiskfs ]; then + mkfsoptions="--mkfsoptions=\\\"-J size=8\\\"" # See bug 17931. + fi - local fs2mdsdev=${fs2mds_DEV:-${MDSDEV}_2} - local fs2ostdev=${fs2ost_DEV:-$(ostdevname 1)_2} - add fs2mds $MDS_MKFS_OPTS --mkfsoptions='\"-J size=8\"' --fsname=${FSNAME2} --reformat $fs2mdsdev || exit 10 - add fs2ost $OST_MKFS_OPTS --fsname=${FSNAME2} --index=8191 --mgsnode=$MGSNID --reformat $fs2ostdev || exit 10 + add fs2mds $(mkfs_opts mds1 ${fs2mdsdev}) --mgs --fsname=${FSNAME2} \ + --reformat $mkfsoptions $fs2mdsdev $fs2mdsvdev || exit 10 + add fs2ost $(mkfs_opts ost1 ${fs2ostdev}) --mgsnode=$MGSNID \ + --fsname=${FSNAME2} --index=8191 --reformat $fs2ostdev \ + $fs2ostvdev || exit 10 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS && trap cleanup_24a EXIT INT start fs2ost $fs2ostdev $OST_MOUNT_OPTS @@ -1371,7 +1787,6 @@ test_33a() { # bug 12333, was test_33 umount -d $MOUNT2 stop fs2ost -f stop fs2mds -f - rm -rf $MOUNT2 $fs2mdsdev $fs2ostdev cleanup_nocli || rc=6 return $rc } @@ -1382,7 +1797,7 @@ test_33b() { # was test_34 do_facet client dd if=/dev/zero of=$MOUNT/24 bs=1024k count=1 # Drop lock cancelation reply during umount - #define OBD_FAIL_LDLM_CANCEL 0x304 + #define OBD_FAIL_LDLM_CANCEL_NET 0x304 do_facet client lctl set_param fail_loc=0x80000304 #lctl set_param debug=-1 umount_client $MOUNT @@ -1446,7 +1861,8 @@ test_35a() { # bug 12459 log "Set up a fake failnode for the MDS" FAKENID="127.0.0.2" local device=$(do_facet $SINGLEMDS "lctl get_param -n devices" | awk '($3 ~ "mdt" && $4 ~ "MDT") { print $4 }' | head -1) - do_facet mgs $LCTL conf_param ${device}.failover.node=$FAKENID || return 4 + do_facet mgs "$LCTL conf_param ${device}.failover.node=" \ + "$(h2$NETTYPE $FAKENID)" || return 4 log "Wait for RECONNECT_INTERVAL seconds (10s)" sleep 10 @@ -1454,13 +1870,13 @@ test_35a() { # bug 12459 MSG="conf-sanity.sh test_35a `date +%F%kh%Mm%Ss`" $LCTL clear log "$MSG" - log "Stopping the MDT:" - stop_mds || return 5 + log "Stopping the MDT: $device" + stop_mdt 1 || return 5 df $MOUNT > /dev/null 2>&1 & DFPID=$! - log "Restarting the MDT:" - start_mds || return 6 + log "Restarting the MDT: $device" + start_mdt 1 || return 6 log "Wait for df ($DFPID) ... " wait $DFPID log "done" @@ -1482,7 +1898,7 @@ test_35a() { # bug 12459 [ "$NEXTCONN" != "0" ] && log "The client didn't try to reconnect to the last active server (tried ${NEXTCONN} instead)" && return 7 cleanup # remove nid settings - writeconf + writeconf_or_reformat } run_test 35a "Reconnect to the last active server first" @@ -1500,8 +1916,8 @@ test_35b() { # bug 18674 FAKENID="127.0.0.2" local device=$(do_facet $SINGLEMDS "$LCTL get_param -n devices" | \ awk '($3 ~ "mdt" && $4 ~ "MDT") { print $4 }' | head -1) - do_facet mgs "$LCTL conf_param ${device}.failover.node=$FAKENID" || \ - return 1 + do_facet mgs "$LCTL conf_param ${device}.failover.node=" \ + "$(h2$NETTYPE $FAKENID)" || return 1 local at_max_saved=0 # adaptive timeouts may prevent seeing the issue @@ -1562,7 +1978,7 @@ test_35b() { # bug 18674 cleanup # remove nid settings - writeconf + writeconf_or_reformat } run_test 35b "Continue reconnection retries, if the active server is busy" @@ -1584,14 +2000,21 @@ test_36() { # 12743 skip_env "mixed loopback and real device not working" && return fi - local fs2mdsdev=${fs2mds_DEV:-${MDSDEV}_2} - local fs2ostdev=${fs2ost_DEV:-$(ostdevname 1)_2} - local fs3ostdev=${fs3ost_DEV:-$(ostdevname 2)_2} - add fs2mds $MDS_MKFS_OPTS --fsname=${FSNAME2} --reformat $fs2mdsdev || exit 10 - # XXX after we support non 4K disk blocksize, change following --mkfsoptions with - # other argument - add fs2ost $OST_MKFS_OPTS --mkfsoptions='-b4096' --fsname=${FSNAME2} --mgsnode=$MGSNID --reformat $fs2ostdev || exit 10 - add fs3ost $OST_MKFS_OPTS --mkfsoptions='-b4096' --fsname=${FSNAME2} --mgsnode=$MGSNID --reformat $fs3ostdev || exit 10 + local fs2mdsdev=$(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) + + add fs2mds $(mkfs_opts mds1 ${fs2mdsdev}) --mgs --fsname=${FSNAME2} \ + --reformat $fs2mdsdev $fs2mdsvdev || exit 10 + # XXX after we support non 4K disk blocksize in ldiskfs, specify a + # different one than the default value here. + add fs2ost $(mkfs_opts ost1 ${fs2ostdev}) --mgsnode=$MGSNID \ + --fsname=${FSNAME2} --reformat $fs2ostdev $fs2ostvdev || exit 10 + add fs3ost $(mkfs_opts ost1 ${fs3ostdev}) --mgsnode=$MGSNID \ + --fsname=${FSNAME2} --reformat $fs3ostdev $fs3ostvdev || exit 10 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS start fs2ost $fs2ostdev $OST_MOUNT_OPTS @@ -1634,7 +2057,6 @@ test_36() { # 12743 stop fs3ost -f || return 200 stop fs2ost -f || return 201 stop fs2mds -f || return 202 - rm -rf $MOUNT2 $fs2mdsdev $fs2ostdev $fs3ostdev unload_modules_conf || return 203 return $rc } @@ -1644,6 +2066,13 @@ test_37() { local mntpt=$(facet_mntpt $SINGLEMDS) local mdsdev=$(mdsdevname ${SINGLEMDS//mds/}) local mdsdev_sym="$TMP/sym_mdt.img" + local opts=$MDS_MOUNT_OPTS + local rc=0 + + if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then + skip "Currently only applicable to ldiskfs-based MDTs" + return + fi echo "MDS : $mdsdev" echo "SYMLINK : $mdsdev_sym" @@ -1653,8 +2082,11 @@ test_37() { echo "mount symlink device - $mdsdev_sym" - local rc=0 - mount_op=$(do_facet $SINGLEMDS mount -v -t lustre $MDS_MOUNT_OPTS $mdsdev_sym $mntpt 2>&1 ) + if ! do_facet $SINGLEMDS test -b $mdsdev; then + opts=$(csa_add "$opts" -o loop) + fi + mount_op=$(do_facet $SINGLEMDS mount -v -t lustre $opts \ + $mdsdev_sym $mntpt 2>&1) rc=${PIPESTATUS[0]} echo mount_op=$mount_op @@ -1672,6 +2104,11 @@ test_37() { run_test 37 "verify set tunables works for symlink device" test_38() { # bug 14222 + if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then + skip "Only applicable to ldiskfs-based MDTs" + return + fi + setup # like runtests COUNT=10 @@ -1747,7 +2184,11 @@ test_40() { # bug 15759 run_test 40 "race during service thread startup" test_41a() { #bug 14134 - echo $MDS_MOUNT_OPTS | grep "loop" && skip " loop devices does not work with nosvc option" && return + if [ $(facet_fstype $SINGLEMDS) == ldiskfs ] && + ! do_facet $SINGLEMDS test -b $(mdsdevname 1); then + skip "Loop devices does not work with nosvc option" + return + fi local rc local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/}) @@ -1772,7 +2213,11 @@ test_41a() { #bug 14134 run_test 41a "mount mds with --nosvc and --nomgs" test_41b() { - echo $MDS_MOUNT_OPTS | grep "loop" && skip " loop devices does not work with nosvc option" && return + if [ $(facet_fstype $SINGLEMDS) == ldiskfs ] && + ! do_facet $SINGLEMDS test -b $(mdsdevname 1); then + skip "Loop devices does not work with nosvc option" + return + fi ! combined_mgs_mds && skip "needs combined mgs device" && return 0 @@ -1799,28 +2244,36 @@ test_41b() { run_test 41b "mount mds with --nosvc and --nomgs on first mount" test_42() { #bug 14693 - setup - check_mount || return 2 - do_facet mgs $LCTL conf_param lustre.llite.some_wrong_param=10 - umount_client $MOUNT - mount_client $MOUNT || return 1 - cleanup - return 0 + setup + check_mount || error "client was not mounted" + + do_facet mgs $LCTL conf_param $FSNAME.llite.some_wrong_param=10 + umount_client $MOUNT || + error "unmounting client failed with invalid llite param" + mount_client $MOUNT || + error "mounting client failed with invalid llite param" + + do_facet mgs $LCTL conf_param $FSNAME.sys.some_wrong_param=20 + cleanup || error "stopping $FSNAME failed with invalid sys param" + setup + check_mount || "client was not mounted with invalid sys param" + cleanup || error "stopping $FSNAME failed with invalid sys param" + return 0 } -run_test 42 "invalid config param should not prevent client from mounting" +run_test 42 "allow client/server mount/unmount with invalid config param" test_43() { - [ $UID -ne 0 -o $RUNAS_ID -eq 0 ] && skip_env "run as root" - setup - chmod ugo+x $DIR || error "chmod 0 failed" - set_and_check mds \ - "lctl get_param -n mdt.$FSNAME-MDT0000.root_squash" \ - "$FSNAME.mdt.root_squash" \ - "0:0" - set_and_check mds \ - "lctl get_param -n mdt.$FSNAME-MDT0000.nosquash_nids" \ - "$FSNAME.mdt.nosquash_nids" \ - "NONE" + [ $UID -ne 0 -o $RUNAS_ID -eq 0 ] && skip_env "run as root" + setup + chmod ugo+x $DIR || error "chmod 0 failed" + set_conf_param_and_check mds \ + "lctl get_param -n mdt.$FSNAME-MDT0000.root_squash" \ + "$FSNAME.mdt.root_squash" \ + "0:0" + set_conf_param_and_check mds \ + "lctl get_param -n mdt.$FSNAME-MDT0000.nosquash_nids" \ + "$FSNAME.mdt.nosquash_nids" \ + "NONE" # # create set of test files @@ -1836,15 +2289,15 @@ test_43() { chmod go-rwx $DIR/$tdir-rootdir || error "chmod 3 failed" touch $DIR/$tdir-rootdir/tfile-1 || error "touch failed" - # - # check root_squash: - # set root squash UID:GID to RUNAS_ID - # root should be able to access only files owned by RUNAS_ID - # - set_and_check mds \ - "lctl get_param -n mdt.$FSNAME-MDT0000.root_squash" \ - "$FSNAME.mdt.root_squash" \ - "$RUNAS_ID:$RUNAS_ID" + # + # check root_squash: + # set root squash UID:GID to RUNAS_ID + # root should be able to access only files owned by RUNAS_ID + # + set_conf_param_and_check mds \ + "lctl get_param -n mdt.$FSNAME-MDT0000.root_squash" \ + "$FSNAME.mdt.root_squash" \ + "$RUNAS_ID:$RUNAS_ID" ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-userfile) dd if=$DIR/$tfile-userfile 1>/dev/null 2>/dev/null || \ @@ -1875,18 +2328,18 @@ test_43() { error "$ST: root create permission is granted" echo "$ST: root create permission is denied - ok" - # - # check nosquash_nids: - # put client's NID into nosquash_nids list, - # root should be able to access root file after that - # - local NIDLIST=$(lctl list_nids all | tr '\n' ' ') - NIDLIST="2@elan $NIDLIST 192.168.0.[2,10]@tcp" - NIDLIST=$(echo $NIDLIST | tr -s ' ' ' ') - set_and_check mds \ - "lctl get_param -n mdt.$FSNAME-MDT0000.nosquash_nids" \ - "$FSNAME-MDTall.mdt.nosquash_nids" \ - "$NIDLIST" + # + # check nosquash_nids: + # put client's NID into nosquash_nids list, + # root should be able to access root file after that + # + local NIDLIST=$(lctl list_nids all | tr '\n' ' ') + NIDLIST="2@elan $NIDLIST 192.168.0.[2,10]@tcp" + NIDLIST=$(echo $NIDLIST | tr -s ' ' ' ') + set_conf_param_and_check mds \ + "lctl get_param -n mdt.$FSNAME-MDT0000.nosquash_nids" \ + "$FSNAME-MDTall.mdt.nosquash_nids" \ + "$NIDLIST" ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-rootfile) dd if=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null || \ @@ -1964,7 +2417,7 @@ cleanup_46a() { stop_mds || rc=$? cleanup_nocli || rc=$? #writeconf to remove all ost2 traces for subsequent tests - writeconf + writeconf_or_reformat return $rc } @@ -1993,7 +2446,7 @@ test_46a() { #second client see all ost's mount_client $MOUNT2 || return 8 - $LFS setstripe $MOUNT2 -c -1 || return 9 + $LFS setstripe -c -1 $MOUNT2 || return 9 $LFS getstripe $MOUNT2 || return 10 echo "ok" > $MOUNT2/widestripe @@ -2061,9 +2514,9 @@ cleanup_48() { test_48() { # bug 17636 reformat setup_noconfig - check_mount || return 2 + check_mount || return 2 - $LFS setstripe $MOUNT -c -1 || return 9 + $LFS setstripe -c -1 $MOUNT || return 9 $LFS getstripe $MOUNT || return 10 echo "ok" > $MOUNT/widestripe @@ -2085,12 +2538,12 @@ run_test 48 "too many acls on file" # check PARAM_SYS_LDLM_TIMEOUT option of MKFS.LUSTRE test_49() { # bug 17710 - local OLD_MDS_MKFS_OPTS=$MDS_MKFS_OPTS - local OLD_OST_MKFS_OPTS=$OST_MKFS_OPTS + local timeout_orig=$TIMEOUT + local ldlm_timeout_orig=$LDLM_TIMEOUT local LOCAL_TIMEOUT=20 - - OST_MKFS_OPTS="--ost --fsname=$FSNAME --device-size=$OSTSIZE --mgsnode=$MGSNID --param sys.timeout=$LOCAL_TIMEOUT --param sys.ldlm_timeout=$LOCAL_TIMEOUT $MKFSOPT $OSTOPT" + LDLM_TIMEOUT=$LOCAL_TIMEOUT + TIMEOUT=$LOCAL_TIMEOUT reformat setup_noconfig @@ -2113,7 +2566,7 @@ test_49() { # bug 17710 stop_ost || return 2 stop_mds || return 3 - OST_MKFS_OPTS="--ost --fsname=$FSNAME --device-size=$OSTSIZE --mgsnode=$MGSNID --param sys.timeout=$LOCAL_TIMEOUT --param sys.ldlm_timeout=$((LOCAL_TIMEOUT - 1)) $MKFSOPT $OSTOPT" + LDLM_TIMEOUT=$((LOCAL_TIMEOUT - 1)) reformat setup_noconfig @@ -2133,8 +2586,8 @@ test_49() { # bug 17710 cleanup || return $? - MDS_MKFS_OPTS=$OLD_MDS_MKFS_OPTS - OST_MKFS_OPTS=$OLD_OST_MKFS_OPTS + LDLM_TIMEOUT=$ldlm_timeout_orig + TIMEOUT=$timeout_orig } run_test 49 "check PARAM_SYS_LDLM_TIMEOUT option of MKFS.LUSTRE" @@ -2206,7 +2659,7 @@ test_50c() { stop_ost2 || error "Unable to stop OST2" stop_mds || error "Unable to stop MDS" #writeconf to remove all ost2 traces for subsequent tests - writeconf + writeconf_or_reformat } run_test 50c "lazystatfs one server down ==========================" @@ -2228,7 +2681,7 @@ test_50d() { stop_ost2 || error "Unable to stop OST2" stop_mds || error "Unable to stop MDS" #writeconf to remove all ost2 traces for subsequent tests - writeconf + writeconf_or_reformat } run_test 50d "lazystatfs client/server conn race ==========================" @@ -2310,7 +2763,7 @@ test_50f() { stop_ost || error "Unable to stop OST1" stop_mds || error "Unable to stop MDS" #writeconf to remove all ost2 traces for subsequent tests - writeconf + writeconf_or_reformat } run_test 50f "normal statfs one server in down ==========================" @@ -2338,10 +2791,46 @@ test_50g() { stop_ost || error "Unable to stop OST1" stop_mds || error "Unable to stop MDS" #writeconf to remove all ost2 traces for subsequent tests - writeconf + writeconf_or_reformat } run_test 50g "deactivated OST should not cause panic=====================" +# LU-642 +test_50h() { + # prepare MDT/OST, make OSC inactive for OST1 + [ "$OSTCOUNT" -lt "2" ] && skip_env "$OSTCOUNT < 2, skipping" && return + do_facet ost1 "$TUNEFS --param osc.active=0 `ostdevname 1`" || + error "tunefs OST1 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 -p $DIR/$tdir + + # activatate OSC for OST1 + local TEST="$LCTL get_param -n osc.${FSNAME}-OST0000-osc-[!M]*.active" + set_conf_param_and_check client \ + "$TEST" "${FSNAME}-OST0000.osc.active" 1 || + error "Unable to activate OST1" + + mkdir -p $DIR/$tdir/2 + $LFS setstripe -c -1 -i 0 $DIR/$tdir/2 + sleep 1 && echo "create a file after OST1 is activated" + # create some file + createmany -o $DIR/$tdir/2/$tfile-%d 1 + + # check OSC import is working + stat $DIR/$tdir/2/* >/dev/null 2>&1 || + error "some OSC imports are still not connected" + + # cleanup + umount_client $MOUNT || error "Unable to umount client" + stop_ost2 || error "Unable to stop OST2" + cleanup_nocli +} +run_test 50h "LU-642: activate deactivated OST ===" + test_51() { local LOCAL_TIMEOUT=20 @@ -2361,7 +2850,7 @@ test_51() { stop_ost2 || return 3 cleanup #writeconf to remove all ost2 traces for subsequent tests - writeconf + writeconf_or_reformat } run_test 51 "Verify that mdt_reint handles RMF_MDT_MD correctly when an OST is added" @@ -2414,6 +2903,11 @@ diff_files_xattrs() } test_52() { + if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then + skip "Only applicable to ldiskfs-based MDTs" + return + fi + start_mds [ $? -eq 0 ] || { error "Unable to start MDS"; return 1; } start_ost @@ -2425,6 +2919,7 @@ test_52() { local ost1mnt=$(facet_mntpt ost1) local ost1node=$(facet_active_host ost1) local ost1tmp=$TMP/conf52 + local loop mkdir -p $DIR/$tdir [ $? -eq 0 ] || { error "Unable to create tdir"; return 4; } @@ -2436,7 +2931,7 @@ test_52() { [ $? -eq 0 ] || { error "Unable to create temporary file"; return 6; } sleep 1 - $LFS setstripe $DIR/$tdir -c -1 -s 1M + $LFS setstripe -c -1 -S 1M $DIR/$tdir [ $? -eq 0 ] || { error "lfs setstripe failed"; return 7; } for (( i=0; i < nrfiles; i++ )); do @@ -2460,14 +2955,19 @@ test_52() { echo mount ost1 as ldiskfs do_node $ost1node mkdir -p $ost1mnt [ $? -eq 0 ] || { error "Unable to create $ost1mnt"; return 23; } - do_node $ost1node mount -t $FSTYPE $ost1_dev $ost1mnt $OST_MOUNT_OPTS + if ! do_node $ost1node test -b $ost1_dev; then + loop="-o loop" + fi + do_node $ost1node mount -t $(facet_fstype ost1) $loop $ost1_dev \ + $ost1mnt [ $? -eq 0 ] || { error "Unable to mount ost1 as ldiskfs"; return 12; } # backup objects echo backup objects to $ost1tmp/objects - local objects=$(do_node $ost1node 'find '$ost1mnt'/O/0 -type f -size +0'\ - '-newer '$ost1tmp'/modified_first -regex ".*\/[0-9]+"') - copy_files_xattrs $ost1node $ost1tmp/objects $ost1tmp/object_xattrs $objects + local objects=$(do_node $ost1node 'find '$ost1mnt'/O/[0-9]* -type f'\ + '-size +0 -newer '$ost1tmp'/modified_first -regex ".*\/[0-9]+"') + copy_files_xattrs $ost1node $ost1tmp/objects $ost1tmp/object_xattrs \ + $objects [ $? -eq 0 ] || { error "Unable to copy objects"; return 13; } # move objects to lost+found @@ -2511,6 +3011,7 @@ thread_sanity() { local facet=$2 local parampat=$3 local opts=$4 + local basethr=$5 local tmin local tmin2 local tmax @@ -2518,9 +3019,10 @@ thread_sanity() { local tstarted local paramp local msg="Insane $modname thread counts" + local ncpts=$(check_cpt_number $facet) + local nthrs shift 4 - setup check_mount || return 41 # We need to expand $parampat, but it may match multiple parameters, so @@ -2539,13 +3041,23 @@ thread_sanity() { tstarted=$(do_facet $facet "lctl get_param -n ${paramp}.threads_started" || echo 0) lassert 23 "$msg (PDSH problems?)" '(($tstarted && $tmin && $tmax))' || return $? lassert 24 "$msg" '(($tstarted >= $tmin && $tstarted <= $tmax ))' || return $? + nthrs=$(expr $tmax - $tmin) + if [ $nthrs -lt $ncpts ]; then + nthrs=0 + else + nthrs=$ncpts + fi + + [ $tmin -eq $tmax -a $tmin -eq $tstarted ] && + skip_env "module parameter forced $facet thread count" && + tmin=3 && tmax=$((3 * tmax)) # Check that we can change min/max - do_facet $facet "lctl set_param ${paramp}.threads_min=$((tmin + 1))" - do_facet $facet "lctl set_param ${paramp}.threads_max=$((tmax - 1))" - tmin2=$(do_facet $facet "lctl get_param -n ${paramp}.threads_min" || echo 0) - tmax2=$(do_facet $facet "lctl get_param -n ${paramp}.threads_max" || echo 0) - lassert 25 "$msg" '(($tmin2 == ($tmin + 1) && $tmax2 == ($tmax -1)))' || return $? + do_facet $facet "lctl set_param ${paramp}.threads_min=$((tmin + nthrs))" + do_facet $facet "lctl set_param ${paramp}.threads_max=$((tmax - nthrs))" + tmin2=$(do_facet $facet "lctl get_param -n ${paramp}.threads_min" || echo 0) + tmax2=$(do_facet $facet "lctl get_param -n ${paramp}.threads_max" || echo 0) + lassert 25 "$msg" '(($tmin2 == ($tmin + $nthrs) && $tmax2 == ($tmax - $nthrs)))' || return $? # Check that we can set min/max to the same value tmin=$(do_facet $facet "lctl get_param -n ${paramp}.threads_min" || echo 0) @@ -2565,7 +3077,8 @@ thread_sanity() { LOAD_MODULES_REMOTE=true cleanup local oldvalue - setmodopts -a $modname "$opts" oldvalue + local newvalue="${opts}=$(expr $basethr \* $ncpts)" + setmodopts -a $modname "$newvalue" oldvalue load_modules setup @@ -2581,25 +3094,37 @@ thread_sanity() { lassert 28 "$msg" '(($tstarted == $tmin && $tstarted == $tmax ))' || return $? cleanup - # Workaround a YALA bug where YALA expects that modules will remain - # loaded on the servers - LOAD_MODULES_REMOTE=false load_modules setup - cleanup } test_53a() { - thread_sanity OST ost1 'ost.*.ost' 'oss_num_threads=64' + setup + thread_sanity OST ost1 'ost.*.ost' 'oss_num_threads' '16' + cleanup } run_test 53a "check OSS thread count params" test_53b() { - thread_sanity MDT $SINGLEMDS 'mdt.*.*.' 'mdt_num_threads=64' + setup + local mds=$(do_facet $SINGLEMDS "lctl get_param -N mds.*.*.threads_max \ + 2>/dev/null") + if [ -z "$mds" ]; then + #running this on an old MDT + thread_sanity MDT $SINGLEMDS 'mdt.*.*.' 'mdt_num_threads' 16 + else + thread_sanity MDT $SINGLEMDS 'mds.*.*.' 'mds_num_threads' 16 + fi + cleanup } -run_test 53b "check MDT thread count params" +run_test 53b "check MDS thread count params" test_54a() { + if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then + skip "Only applicable to ldiskfs-based MDTs" + return + fi + do_rpc_nodes $(facet_host ost1) run_llverdev $(ostdevname 1) -p [ $? -eq 0 ] || error "llverdev failed!" reformat_and_config @@ -2607,6 +3132,11 @@ test_54a() { run_test 54a "test llverdev and partial verify of device" test_54b() { + if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then + skip "Only applicable to ldiskfs-based MDTs" + return + fi + setup run_llverfs $MOUNT -p [ $? -eq 0 ] || error "llverfs failed!" @@ -2621,20 +3151,25 @@ lov_objid_size() } test_55() { + if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then + skip "Only applicable to ldiskfs-based MDTs" + return + fi + local mdsdev=$(mdsdevname 1) - local ostdev=$(ostdevname 1) - local saved_opts=$OST_MKFS_OPTS + local mdsvdev=$(mdsvdevname 1) for i in 1023 2048 do - OST_MKFS_OPTS="$saved_opts --index $i" - reformat - + add mds1 $(mkfs_opts mds1 ${mdsdev}) --reformat $mdsdev \ + $mdsvdev || exit 10 + add ost1 $(mkfs_opts ost1 $(ostdevname 1)) --index=$i \ + --reformat $(ostdevname 1) $(ostvdevname 1) setup_noconfig stopall - setup_noconfig sync + echo checking size of lov_objid for ost index $i LOV_OBJID_SIZE=$(do_facet mds1 "$DEBUGFS -R 'stat lov_objid' $mdsdev 2>/dev/null" | grep ^User | awk '{print $6}') if [ "$LOV_OBJID_SIZE" != $(lov_objid_size $i) ]; then @@ -2645,15 +3180,23 @@ test_55() { stopall done - OST_MKFS_OPTS=$saved_opts reformat } run_test 55 "check lov_objid size" test_56() { - add mds1 $MDS_MKFS_OPTS --mkfsoptions='\"-J size=16\"' --reformat $(mdsdevname 1) - add ost1 $OST_MKFS_OPTS --index=1000 --reformat $(ostdevname 1) - add ost2 $OST_MKFS_OPTS --index=10000 --reformat $(ostdevname 2) + local mds_journal_size_orig=$MDSJOURNALSIZE + + MDSJOURNALSIZE=16 + + for num in $(seq 1 $MDSCOUNT); do + add mds${num} $(mkfs_opts mds${num} $(mdsdevname $num)) \ + --reformat $(mdsdevname $num) $(mdsvdevname $num) + done + add ost1 $(mkfs_opts ost1 $(ostdevname 1)) --index=1000 --reformat \ + $(ostdevname 1) $(ostvdevname 1) + add ost2 $(mkfs_opts ost2 $(ostdevname 2)) --index=10000 --reformat \ + $(ostdevname 2) $(ostvdevname 2) start_mgsmds start_ost @@ -2661,15 +3204,15 @@ test_56() { mount_client $MOUNT || error "Unable to mount client" echo ok $LFS osts - [ -n "$ENABLE_QUOTA" ] && { $LFS quotacheck -ug $MOUNT || error "quotacheck has failed" ; } stopall + MDSJOURNALSIZE=$mds_journal_size_orig reformat } run_test 56 "check big indexes" test_57a() { # bug 22656 local NID=$(do_facet ost1 "$LCTL get_param nis" | tail -1 | awk '{print $1}') - writeconf + writeconf_or_reformat do_facet ost1 "$TUNEFS --failnode=$NID `ostdevname 1`" || error "tunefs failed" start_mgsmds start_ost && error "OST registration from failnode should fail" @@ -2679,7 +3222,7 @@ run_test 57a "initial registration from failnode should fail (should return errs test_57b() { local NID=$(do_facet ost1 "$LCTL get_param nis" | tail -1 | awk '{print $1}') - writeconf + writeconf_or_reformat do_facet ost1 "$TUNEFS --servicenode=$NID `ostdevname 1`" || error "tunefs failed" start_mgsmds start_ost || error "OST registration from servicenode should not fail" @@ -2692,7 +3235,10 @@ count_osts() { } test_58() { # bug 22658 - [ "$FSTYPE" != "ldiskfs" ] && skip "not supported for $FSTYPE" && return + if [ $(facet_fstype mds) != ldiskfs ]; then + skip "Only applicable to ldiskfs-based MDTs" + return + fi setup_noconfig mkdir -p $DIR/$tdir createmany -o $DIR/$tdir/$tfile-%d 100 @@ -2741,12 +3287,24 @@ test_59() { stop_ost2 >> /dev/null cleanup_nocli >> /dev/null #writeconf to remove all ost2 traces for subsequent tests - writeconf + writeconf_or_reformat } run_test 59 "writeconf mount option" test_60() { # LU-471 - add mds1 $MDS_MKFS_OPTS --mkfsoptions='\" -E stride=64 -O ^uninit_bg\"' --reformat $(mdsdevname 1) + local num + + if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then + skip "Only applicable to ldiskfs-based MDTs" + return + fi + + for num in $(seq $MDSCOUNT); do + add mds${num} $(mkfs_opts mds${num} $(mdsdevname $num)) \ + --mkfsoptions='\" -E stride=64 -O ^uninit_bg\"' \ + --reformat $(mdsdevname $num) $(mdsvdevname $num) || + exit 10 + done dump=$(do_facet $SINGLEMDS dumpe2fs $(mdsdevname 1)) rc=${PIPESTATUS[0]} @@ -2763,14 +3321,22 @@ test_60() { # LU-471 run_test 60 "check mkfs.lustre --mkfsoptions -E -O options setting" test_61() { # LU-80 - local reformat=false + local reformat=false + + [ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.1.53) ] || + { skip "Need MDS version at least 2.1.53"; return 0; } - if ! large_xattr_enabled; then - reformat=true - local mds_dev=$(mdsdevname ${SINGLEMDS//mds/}) - add $SINGLEMDS $MDS_MKFS_OPTS --mkfsoptions='\"-O large_xattr\"' \ - --reformat $mds_dev || error "reformatting $mds_dev failed" - fi + if [ $(facet_fstype $SINGLEMDS) == ldiskfs ] && + ! large_xattr_enabled; then + reformat=true + LDISKFS_MKFS_OPTS+=" -O large_xattr" + + 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" + done + fi setup_noconfig || error "setting up the filesystem failed" client_up || error "starting client failed" @@ -2817,23 +3383,466 @@ test_61() { # LU-80 rm -f $file stopall - $reformat && reformat + if $reformat; then + LDISKFS_MKFS_OPTS=${LDISKFS_MKFS_OPTS% -O large_xattr} + reformat + fi } run_test 61 "large xattr" +test_62() { + if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then + skip "Only applicable to ldiskfs-based MDTs" + return + fi + + # MRP-118 + local mdsdev=$(mdsdevname 1) + local ostdev=$(ostdevname 1) + + [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.2.51) ]] || + { 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" + start_mds && error "MDT start should fail" + echo "disable journal for ost" + do_facet ost1 tune2fs -O ^has_journal $ostdev || error "tune2fs failed" + start_ost && error "OST start should fail" + cleanup || return $? + reformat_and_config +} +run_test 62 "start with disabled journal" + +test_63() { + if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then + skip "Only applicable to ldiskfs-based MDTs" + return + fi + + local inode_slab=$(do_facet $SINGLEMDS \ + "awk '/ldiskfs_inode_cache/ { print \\\$5 }' /proc/slabinfo") + if [ -z "$inode_slab" ]; then + skip "ldiskfs module has not been loaded" + return + fi + + echo "$inode_slab ldisk inodes per page" + [ "$inode_slab" -ge "3" ] || + error "ldisk inode size is too big, $inode_slab objs per page" + return +} +run_test 63 "Verify each page can at least hold 3 ldisk inodes" + +test_64() { + start_mds + start_ost + start_ost2 || error "Unable to start second ost" + mount_client $MOUNT || error "Unable to mount client" + stop_ost2 || error "Unable to stop second ost" + echo "$LFS df" + $LFS df --lazy || error "lfs df failed" + cleanup || return $? + #writeconf to remove all ost2 traces for subsequent tests + writeconf_or_reformat +} +run_test 64 "check lfs df --lazy " + +test_65() { # LU-2237 + # Currently, the test is only valid for ldiskfs backend + [ "$(facet_fstype $SINGLEMDS)" != "ldiskfs" ] && + skip "non-ldiskfs backend" && return + + local devname=$(mdsdevname ${SINGLEMDS//mds/}) + local brpt=$(facet_mntpt brpt) + local opts="" + + if ! do_facet $SINGLEMDS "test -b $devname"; then + opts="-o loop" + fi + + stop_mds + local obj=$(do_facet $SINGLEMDS \ + "$DEBUGFS -c -R \\\"stat last_rcvd\\\" $devname" | + grep Inode) + if [ -z "$obj" ]; then + # The MDT may be just re-formatted, mount the MDT for the + # first time to guarantee the "last_rcvd" file is there. + start_mds || error "fail to mount the MDS for the first time" + stop_mds + fi + + # remove the "last_rcvd" file + do_facet $SINGLEMDS "mkdir -p $brpt" + do_facet $SINGLEMDS \ + "mount -t $(facet_fstype $SINGLEMDS) $opts $devname $brpt" + do_facet $SINGLEMDS "rm -f ${brpt}/last_rcvd" + do_facet $SINGLEMDS "umount $brpt" + + # restart MDS, the "last_rcvd" file should be recreated. + start_mds || error "fail to restart the MDS" + stop_mds + obj=$(do_facet $SINGLEMDS \ + "$DEBUGFS -c -R \\\"stat last_rcvd\\\" $devname" | grep Inode) + [ -n "$obj" ] || error "fail to re-create the last_rcvd" +} +run_test 65 "re-create the lost last_rcvd file when server mount" + +test_66() { + [[ $(lustre_version_code mgs) -ge $(version_code 2.3.59) ]] || + { skip "Need MGS version at least 2.3.59"; return 0; } + + setup + local OST1_NID=$(do_facet ost1 $LCTL list_nids | head -1) + local MDS_NID=$(do_facet $SINGLEMDS $LCTL list_nids | head -1) + + echo "replace_nids should fail if MDS, OSTs and clients are UP" + do_facet mgs $LCTL replace_nids $FSNAME-OST0000 $OST1_NID && + error "replace_nids fail" + + umount_client $MOUNT || error "unmounting client failed" + echo "replace_nids should fail if MDS and OSTs are UP" + do_facet mgs $LCTL replace_nids $FSNAME-OST0000 $OST1_NID && + error "replace_nids fail" + + stop_ost + echo "replace_nids should fail if MDS is UP" + do_facet mgs $LCTL replace_nids $FSNAME-OST0000 $OST1_NID && + error "replace_nids fail" + + stop_mds || error "stopping mds failed" + + if combined_mgs_mds; then + start_mds "-o nosvc" || + error "starting mds with nosvc option failed" + fi + + echo "command should accept two parameters" + do_facet mgs $LCTL replace_nids $FSNAME-OST0000 && + error "command should accept two params" + + echo "correct device name should be passed" + do_facet mgs $LCTL replace_nids $FSNAME-WRONG0000 $OST1_NID && + error "wrong devname" + + echo "wrong nids list should not destroy the system" + do_facet mgs $LCTL replace_nids $FSNAME-OST0000 "wrong nids list" && + error "wrong parse" + + echo "replace OST nid" + do_facet mgs $LCTL replace_nids $FSNAME-OST0000 $OST1_NID || + error "replace nids failed" + + echo "command should accept two parameters" + do_facet mgs $LCTL replace_nids $FSNAME-MDT0000 && + error "command should accept two params" + + echo "wrong nids list should not destroy the system" + do_facet mgs $LCTL replace_nids $FSNAME-MDT0000 "wrong nids list" && + error "wrong parse" + + echo "replace MDS nid" + do_facet mgs $LCTL replace_nids $FSNAME-MDT0000 $MDS_NID || + error "replace nids failed" + + if ! combined_mgs_mds ; then + stop_mgs + else + stop_mds + fi + + setup_noconfig + check_mount || error "error after nid replace" + cleanup + reformat +} +run_test 66 "replace nids" + +test_70a() { + [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return + local MDTIDX=1 + + start_mdt 1 || error "MDT0 start fail" + + start_ost || error "OST0 start fail" + + start_mdt 2 || error "MDT1 start fail" + + mount_client $MOUNT || error "mount client fails" + + mkdir -p $DIR/$tdir || error "create dir fail" + + $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir || + error "create remote dir fail" + + rm -rf $DIR/$tdir || error "delete dir fail" + cleanup || return $? +} +run_test 70a "start MDT0, then OST, then MDT1" + +test_70b() { + [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return + local MDTIDX=1 + + start_ost || error "OST0 start fail" + + start_mdt 1 || error "MDT0 start fail" + start_mdt 2 || error "MDT1 start fail" + + mount_client $MOUNT || error "mount client fails" + + mkdir -p $DIR/$tdir || error "create dir fail" + + $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir || + error "create remote dir fail" + + rm -rf $DIR/$tdir || error "delete dir fail" + + cleanup || return $? +} +run_test 70b "start OST, MDT1, MDT0" + +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_ost || error "OST0 start fail" + + mount_client $MOUNT || error "mount client fails" + stop_mdt 1 || error "MDT1 start fail" + + local mdc_for_mdt1=$($LCTL dl | grep MDT0000-mdc | awk '{print $4}') + echo "deactivate $mdc_for_mdt1" + $LCTL --device $mdc_for_mdt1 deactivate || return 1 + + mkdir -p $DIR/$tdir && error "mkdir succeed" + + $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir && + error "create remote dir succeed" + + cleanup || return $? +} +run_test 70c "stop MDT0, mkdir fail, create remote dir fail" + +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_ost || error "OST0 start fail" + + mount_client $MOUNT || error "mount client fails" + + stop_mdt 2 || error "MDT1 start fail" + + local mdc_for_mdt2=$($LCTL dl | grep MDT0001-mdc | + awk '{print $4}') + echo "deactivate $mdc_for_mdt2" + $LCTL --device $mdc_for_mdt2 deactivate || + error "set $mdc_for_mdt2 deactivate failed" + + mkdir -p $DIR/$tdir || error "mkdir fail" + $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir && + error "create remote dir succeed" + + rm -rf $DIR/$tdir || error "delete dir fail" + + cleanup || return $? +} +run_test 70d "stop MDT1, mkdir succeed, create remote dir fail" + +test_71a() { + [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return + if combined_mgs_mds; then + skip "needs separate MGS/MDT" && return + fi + local MDTIDX=1 + + start_mdt 1 || error "MDT0 start fail" + start_ost || error "OST0 start fail" + start_mdt 2 || error "MDT1 start fail" + start_ost2 || error "OST1 start fail" + + mount_client $MOUNT || error "mount client fails" + + mkdir -p $DIR/$tdir || error "mkdir fail" + $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir || + error "create remote dir succeed" + + mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed" + rm -rf $DIR/$tdir || error "delete dir fail" + + umount_client $MOUNT + stop_mdt 1 || error "MDT0 stop fail" + stop_mdt 2 || error "MDT1 stop fail" + stop_ost || error "OST0 stop fail" + stop_ost2 || error "OST1 stop fail" +} +run_test 71a "start MDT0 OST0, MDT1, OST1" + +test_71b() { + [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return + if combined_mgs_mds; then + skip "needs separate MGS/MDT" && return + fi + local MDTIDX=1 + + start_mdt 2 || error "MDT1 start fail" + start_ost || error "OST0 start fail" + start_mdt 1 || error "MDT0 start fail" + start_ost2 || error "OST1 start fail" + + mount_client $MOUNT || error "mount client fails" + + mkdir -p $DIR/$tdir || error "mkdir fail" + $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir || + error "create remote dir succeed" + + mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed" + rm -rf $DIR/$tdir || error "delete dir fail" + + umount_client $MOUNT + stop_mdt 1 || error "MDT0 stop fail" + stop_mdt 2 || error "MDT1 stop fail" + stop_ost || error "OST0 stop fail" + stop_ost2 || error "OST1 stop fail" +} +run_test 71b "start MDT1, OST0, MDT0, OST1" + +test_71c() { + [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return + if combined_mgs_mds; then + skip "needs separate MGS/MDT" && return + fi + local MDTIDX=1 + + start_ost || error "OST0 start fail" + start_ost2 || error "OST1 start fail" + start_mdt 2 || error "MDT1 start fail" + start_mdt 1 || error "MDT0 start fail" + + mount_client $MOUNT || error "mount client fails" + + mkdir -p $DIR/$tdir || error "mkdir fail" + $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir || + error "create remote dir succeed" + + mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed" + rm -rf $DIR/$tdir || error "delete dir fail" + + umount_client $MOUNT + stop_mdt 1 || error "MDT0 stop fail" + stop_mdt 2 || error "MDT1 stop fail" + stop_ost || error "OST0 stop fail" + stop_ost2 || error "OST1 stop fail" + +} +run_test 71c "start OST0, OST1, MDT1, MDT0" + +test_71d() { + [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return + if combined_mgs_mds; then + skip "needs separate MGS/MDT" && return + fi + local MDTIDX=1 + + start_ost || error "OST0 start fail" + start_mdt 2 || error "MDT0 start fail" + start_mdt 1 || error "MDT0 start fail" + start_ost2 || error "OST1 start fail" + + mount_client $MOUNT || error "mount client fails" + + mkdir -p $DIR/$tdir || error "mkdir fail" + $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir || + error "create remote dir succeed" + + mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed" + rm -rf $DIR/$tdir || error "delete dir fail" + + umount_client $MOUNT + stop_mdt 1 || error "MDT0 stop fail" + stop_mdt 2 || error "MDT1 stop fail" + stop_ost || error "OST0 stop fail" + stop_ost2 || error "OST1 stop fail" + +} +run_test 71d "start OST0, MDT1, MDT0, OST1" + +test_71e() { + [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return + if combined_mgs_mds; then + skip "needs separate MGS/MDT" && return + fi + local MDTIDX=1 + + start_ost || error "OST0 start fail" + start_mdt 2 || error "MDT1 start fail" + start_ost2 || error "OST1 start fail" + start_mdt 1 || error "MDT0 start fail" + + mount_client $MOUNT || error "mount client fails" + + mkdir -p $DIR/$tdir || error "mkdir fail" + $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir || + error "create remote dir succeed" + + mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed" + rm -rf $DIR/$tdir || error "delete dir fail" + + umount_client $MOUNT + stop_mdt 1 || error "MDT0 stop fail" + stop_mdt 2 || error "MDT1 stop fail" + stop_ost || error "OST0 stop fail" + stop_ost2 || error "OST1 stop fail" + +} +run_test 71e "start OST0, MDT1, OST1, MDT0" + +test_72() { #LU-2634 + local mdsdev=$(mdsdevname 1) + local ostdev=$(ostdevname 1) + local cmd="$E2FSCK -fnvd $mdsdev" + local fn=3 + + #tune MDT with "-O extents" + add $SINGLEMDS \ + $(mkfs_opts $SINGLEMDS ${mdsdev}) --reformat $mdsdev || + error "add $SINGLEMDS failed" + $TUNE2FS -O extents $mdsdev + add ost1 $(mkfs_opts ost1 $ostdev) --reformat $ostdev || + error "add $ostdev failed" + start_mgsmds || error "start mds failed" + start_ost || error "start ost failed" + mount_client $MOUNT || error "mount client failed" + + #create some short symlinks + mkdir -p $DIR/$tdir + createmany -o $DIR/$tdir/$tfile-%d $fn + echo "create $fn short symlinks" + for i in $(seq -w 1 $fn); do + ln -s $DIR/$tdir/$tfile-$i $MOUNT/$tfile-$i + done + ls -al $MOUNT + + #umount + umount_client $MOUNT || error "umount client failed" + stop_mds || error "stop mds failed" + stop_ost || error "stop ost failed" + + #run e2fsck + run_e2fsck $(facet_active_host $SINGLEMDS) $mdsdev "-n" +} +run_test 72 "test fast symlink with extents flag enabled" + if ! combined_mgs_mds ; then stop mgs fi cleanup_gss -# restore the ${facet}_MKFS_OPTS variables -for facet in MGS MDS OST; do - opts=SAVED_${facet}_MKFS_OPTS - if [[ -n ${!opts} ]]; then - eval ${facet}_MKFS_OPTS=\"${!opts}\" - fi -done - -complete $(basename $0) $SECONDS +complete $SECONDS exit_status