X-Git-Url: https://git.whamcloud.com/?p=fs%2Flustre-release.git;a=blobdiff_plain;f=lustre%2Ftests%2Ftest-framework.sh;h=95ddf56ba0c3517068de209df404c978e7302bd5;hp=1dd2725d364e2f5c36d8752fb30a8ee54013a235;hb=15dd813536ad06a119dfb2358f00281eed22a98b;hpb=7c25eb1ba2b1db3009a0e88b3ecf229134f8ac92

diff --git a/lustre/tests/test-framework.sh b/lustre/tests/test-framework.sh
index 1dd2725..95ddf56 100755
--- a/lustre/tests/test-framework.sh
+++ b/lustre/tests/test-framework.sh
@@ -6,7 +6,6 @@ set -e
 #set -x
 
 export LANG=en_US
-export EJOURNAL=${EJOURNAL:-""}
 export REFORMAT=${REFORMAT:-""}
 export WRITECONF=${WRITECONF:-""}
 export VERBOSE=${VERBOSE:-false}
@@ -24,6 +23,11 @@ export LOAD_LLOOP=${LOAD_LLOOP:-false}
 
 #export PDSH="pdsh -S -Rssh -w"
 export MOUNT_CMD=${MOUNT_CMD:-"mount -t lustre"}
+export UMOUNT=${UMOUNT:-"umount -d"}
+# sles12 umount has a issue with -d option
+[ -e /etc/SuSE-release ] && grep -w VERSION /etc/SuSE-release | grep -wq 12 && {
+	export UMOUNT="umount"
+}
 
 # function used by scripts run on remote nodes
 LUSTRE=${LUSTRE:-$(cd $(dirname $0)/..; echo $PWD)}
@@ -50,16 +54,25 @@ fi
 [ -z "$MODPROBECONF" -a -f /etc/modprobe.conf ] &&
     MODPROBECONF=/etc/modprobe.conf
 
+sanitize_parameters() {
+	for i in DIR DIR1 DIR2 MOUNT MOUNT1 MOUNT2
+	do
+		local path=${!i}
+		if [ -d "$path" ]; then
+			eval export $i=$(echo $path | sed -r 's/\/+$//g')
+		fi
+	done
+}
 assert_DIR () {
-    local failed=""
-    [[ $DIR/ = $MOUNT/* ]] || \
-        { failed=1 && echo "DIR=$DIR not in $MOUNT. Aborting."; }
-    [[ $DIR1/ = $MOUNT1/* ]] || \
-        { failed=1 && echo "DIR1=$DIR1 not in $MOUNT1. Aborting."; }
-    [[ $DIR2/ = $MOUNT2/* ]] || \
-        { failed=1 && echo "DIR2=$DIR2 not in $MOUNT2. Aborting"; }
+	local failed=""
+	[[ $DIR/ = $MOUNT/* ]] ||
+		{ failed=1 && echo "DIR=$DIR not in $MOUNT. Aborting."; }
+	[[ $DIR1/ = $MOUNT1/* ]] ||
+		{ failed=1 && echo "DIR1=$DIR1 not in $MOUNT1. Aborting."; }
+	[[ $DIR2/ = $MOUNT2/* ]] ||
+		{ failed=1 && echo "DIR2=$DIR2 not in $MOUNT2. Aborting"; }
 
-    [ -n "$failed" ] && exit 99 || true
+	[ -n "$failed" ] && exit 99 || true
 }
 
 usage() {
@@ -72,7 +85,6 @@ usage() {
 print_summary () {
     trap 0
 	[ -z "$DEFAULT_SUITES"] && return 0
-    [ "$TESTSUITE" == "lfsck" ] && return 0
     [ -n "$ONLY" ] && echo "WARNING: ONLY is set to $(echo $ONLY)"
     local details
     local form="%-13s %-17s %-9s %s %s\n"
@@ -130,6 +142,7 @@ init_test_env() {
 	export TEST_FAILED=false
 	export FAIL_ON_SKIP_ENV=${FAIL_ON_SKIP_ENV:-false}
 	export RPC_MODE=${RPC_MODE:-false}
+	export DO_CLEANUP=${DO_CLEANUP:-true}
 
     export MKE2FS=$MKE2FS
     if [ -z "$MKE2FS" ]; then
@@ -194,9 +207,8 @@ init_test_env() {
 		fi
 	fi
 
-    export LFSCK_BIN=${LFSCK_BIN:-lfsck}
-    export LFSCK_ALWAYS=${LFSCK_ALWAYS:-"no"} # check fs after each test suite
-    export FSCK_MAX_ERR=4   # File system errors left uncorrected
+	export LFSCK_ALWAYS=${LFSCK_ALWAYS:-"no"} # check fs after test suite
+	export FSCK_MAX_ERR=4   # File system errors left uncorrected
 
 	export ZFS=${ZFS:-zfs}
 	export ZPOOL=${ZPOOL:-zpool}
@@ -230,6 +242,7 @@ init_test_env() {
     # Ubuntu, at least, has a truncate command in /usr/bin
     # so fully path our truncate command.
     export TRUNCATE=${TRUNCATE:-$LUSTRE/tests/truncate}
+	export FSX=${FSX:-$LUSTRE/tests/fsx}
     export MDSRATE=${MDSRATE:-"$LUSTRE/tests/mpi/mdsrate"}
     [ ! -f "$MDSRATE" ] && export MDSRATE=$(which mdsrate 2> /dev/null)
     if ! echo $PATH | grep -q $LUSTRE/tests/racer; then
@@ -281,6 +294,8 @@ init_test_env() {
     export DIR2
     export SAVE_PWD=${SAVE_PWD:-$LUSTRE/tests}
     export AT_MAX_PATH
+    export LDEV=${LDEV:-"$LUSTRE/scripts/ldev"}
+    [ ! -f "$LDEV" ] && export LDEV=$(which ldev 2> /dev/null)
 
     if [ "$ACCEPTOR_PORT" ]; then
         export PORT_OPT="--port $ACCEPTOR_PORT"
@@ -321,20 +336,20 @@ init_test_env() {
 	export SHUTDOWN_ATTEMPTS=${SHUTDOWN_ATTEMPTS:-3}
 	export OSD_TRACK_DECLARES_LBUG=${OSD_TRACK_DECLARES_LBUG:-"yes"}
 
-    # command line
+	# command line
 
-    while getopts "rvwf:" opt $*; do
-        case $opt in
-            f) CONFIG=$OPTARG;;
-            r) REFORMAT=--reformat;;
-            v) VERBOSE=true;;
-            w) WRITECONF=writeconf;;
-            \?) usage;;
-        esac
-    done
+	while getopts "rvwf:" opt $*; do
+		case $opt in
+			f) CONFIG=$OPTARG;;
+			r) REFORMAT=yes;;
+			v) VERBOSE=true;;
+			w) WRITECONF=writeconf;;
+			\?) usage;;
+		esac
+	done
 
-    shift $((OPTIND - 1))
-    ONLY=${ONLY:-$*}
+	shift $((OPTIND - 1))
+	ONLY=${ONLY:-$*}
 
 	# print the durations of each test if "true"
 	DDETAILS=${DDETAILS:-false}
@@ -389,9 +404,7 @@ export LINUX_VERSION_CODE=$(version_code ${LINUX_VERSION//\./ })
 lustre_build_version() {
 	local facet=${1:-client}
 
-	# lustre: 2.6.52
-	# kernel: patchless_client
-	# build: v2_6_92_0-gadb3ee4-2.6.32-431.29.2.el6_lustre.x86_64
+	# lustre: 2.8.52
 	local VER=$(do_facet $facet $LCTL get_param -n version 2> /dev/null |
 		    awk '/lustre: / { print $2 }')
 	# lctl 2.6.50
@@ -557,13 +570,15 @@ load_modules_local() {
     load_module lov/lov
     load_module mgc/mgc
     load_module obdecho/obdecho
-    if ! client_only; then
-        SYMLIST=/proc/kallsyms
-        grep -q crc16 $SYMLIST || { modprobe crc16 2>/dev/null || true; }
-        grep -q -w jbd $SYMLIST || { modprobe jbd 2>/dev/null || true; }
-        grep -q -w jbd2 $SYMLIST || { modprobe jbd2 2>/dev/null || true; }
+	if ! client_only; then
+		SYMLIST=/proc/kallsyms
+		grep -q crc16 $SYMLIST ||
+			{ modprobe crc16 2>/dev/null || true; }
+		grep -q -w jbd2 $SYMLIST ||
+			{ modprobe jbd2 2>/dev/null || true; }
 		load_module lfsck/lfsck
-		[ "$LQUOTA" != "no" ] && load_module quota/lquota $LQUOTAOPTS
+		[ "$LQUOTA" != "no" ] &&
+			load_module quota/lquota $LQUOTAOPTS
 		if [[ $(node_fstypes $HOSTNAME) == *zfs* ]]; then
 			modprobe zfs
 			load_module osd-zfs/osd_zfs
@@ -571,6 +586,8 @@ load_modules_local() {
 		if [[ $(node_fstypes $HOSTNAME) == *ldiskfs* ]]; then
 			grep -q exportfs_decode_fh $SYMLIST ||
 				{ modprobe exportfs 2> /dev/null || true; }
+			grep -q -w mbcache $SYMLIST ||
+				{ modprobe mbcache 2>/dev/null || true; }
 			load_module ../ldiskfs/ldiskfs
 			load_module osd-ldiskfs/osd_ldiskfs
 		fi
@@ -582,7 +599,7 @@ load_modules_local() {
 		load_module osp/osp
 		load_module ofd/ofd
 		load_module osp/osp
-    fi
+	fi
 
 	load_module llite/lustre
 	llite_lloop_enabled && load_module llite/llite_lloop
@@ -594,7 +611,7 @@ load_modules_local() {
 	# 'mount' doesn't look in $PATH, just sbin
 	local mount_lustre=$LUSTRE/utils/mount.lustre
 	if [ -f $mount_lustre ]; then
-		local sbin_mount=/sbin/mount.lustre
+		local sbin_mount=$(readlink -f /sbin)/mount.lustre
 		if grep -qw "$sbin_mount" /proc/mounts; then
 			cmp -s $mount_lustre $sbin_mount || umount $sbin_mount
 		fi
@@ -658,7 +675,7 @@ unload_modules() {
 		fi
 	fi
 
-	local sbin_mount=/sbin/mount.lustre
+	local sbin_mount=$(readlink -f /sbin)/mount.lustre
 	if grep -qe "$sbin_mount " /proc/mounts; then
 		umount $sbin_mount || true
 		[ -s $sbin_mount ] && ! grep -q "STUB MARK" $sbin_mount ||
@@ -683,6 +700,18 @@ fs_log_size() {
 	echo -n $size
 }
 
+fs_inode_ksize() {
+	local facet=${1:-$SINGLEMDS}
+	local fstype=$(facet_fstype $facet)
+	local size=0
+	case $fstype in
+		ldiskfs) size=4;;  # ~4KB per inode
+		zfs)     size=11;; # 10 to 11KB per inode
+	esac
+
+	echo -n $size
+}
+
 check_gss_daemon_nodes() {
     local list=$1
     dname=$2
@@ -814,7 +843,7 @@ facet_type() {
 facet_number() {
 	local facet=$1
 
-	if [ $facet == mgs ]; then
+	if [ $facet == mgs ] || [ $facet == client ]; then
 		return 1
 	fi
 
@@ -961,6 +990,19 @@ facet_vdevice() {
 	echo -n $device
 }
 
+running_in_vm() {
+	local virt=$(virt-what 2> /dev/null)
+
+	[ $? -eq 0 ] && [ -n "$virt" ] && { echo $virt; return; }
+
+	virt=$(dmidecode -s system-product-name | awk '{print $1}')
+
+	case $virt in
+		VMware|KVM|VirtualBox|Parallels) echo ${virt,,} ;;
+		*) ;;
+	esac
+}
+
 #
 # Re-read the partition table on failover partner host.
 # After a ZFS storage pool is created on a shared device, the partition table
@@ -1113,7 +1155,7 @@ get_osd_param() {
 	local name=$3
 
 	do_nodes $nodes "$LCTL get_param -n obdfilter.$device.$name \
-		osd-*.$device.$name 2>&1" | grep -v 'Found no match'
+		osd-*.$device.$name 2>&1" | grep -v 'error:'
 }
 
 set_osd_param() {
@@ -1123,7 +1165,7 @@ set_osd_param() {
 	local value=$4
 
 	do_nodes $nodes "$LCTL set_param -n obdfilter.$device.$name=$value \
-		osd-*.$device.$name=$value 2>&1" | grep -v 'Found no match'
+		osd-*.$device.$name=$value 2>&1" | grep -v 'error:'
 }
 
 set_debug_size () {
@@ -1132,7 +1174,7 @@ set_debug_size () {
     if [ -f /sys/devices/system/cpu/possible ]; then
         local cpus=$(($(cut -d "-" -f 2 /sys/devices/system/cpu/possible)+1))
     else
-        local cpus=$(getconf _NPROCESSORS_CONF)
+        local cpus=$(getconf _NPROCESSORS_CONF 2>/dev/null)
     fi
 
     # bug 19944, adjust size to be -gt num_possible_cpus()
@@ -1174,6 +1216,15 @@ set_default_debug_facet () {
     set_default_debug_nodes $node
 }
 
+set_hostid () {
+    local hostid=${1:-$(hostid)}
+
+    if [ ! -s /etc/hostid ]; then
+	printf $(echo -n $hostid |
+	    sed 's/\(..\)\(..\)\(..\)\(..\)/\\x\4\\x\3\\x\2\\x\1/') >/etc/hostid
+    fi
+}
+
 # Facet functions
 mount_facets () {
 	local facets=${1:-$(get_facets)}
@@ -1220,6 +1271,10 @@ mount_facet() {
 	local opt=${facet}_opt
 	local mntpt=$(facet_mntpt $facet)
 	local opts="${!opt} $@"
+	local fstype=$(facet_fstype $facet)
+	local devicelabel
+
+	module_loaded lustre || load_modules
 
 	if [ $(facet_fstype $facet) == ldiskfs ] &&
 	   ! do_facet $facet test -b ${!dev}; then
@@ -1231,6 +1286,16 @@ mount_facet() {
 		import_zpool $facet || return ${PIPESTATUS[0]}
 	fi
 
+	case $fstype in
+	ldiskfs)
+		devicelabel=$(do_facet ${facet} "$E2LABEL ${!dev}");;
+	zfs)
+		devicelabel=$(do_facet ${facet} "$ZFS get -H -o value \
+						lustre:svname ${!dev}");;
+	*)
+		error "unknown fstype!";;
+	esac
+
 	echo "Starting ${facet}: $opts ${!dev} $mntpt"
 	# for testing LU-482 error handling in mount_facets() and test_0a()
 	if [ -f $TMP/test-lu482-trigger ]; then
@@ -1240,40 +1305,73 @@ mount_facet() {
 		                   ${!dev} $mntpt"
 		RC=${PIPESTATUS[0]}
 	fi
+
 	if [ $RC -ne 0 ]; then
 		echo "Start of ${!dev} on ${facet} failed ${RC}"
-    else
-        set_default_debug_facet $facet
+		return $RC
+	fi
+
+	health=$(do_facet ${facet} "$LCTL get_param -n health_check")
+	if [[ "$health" != "healthy" ]]; then
+		error "$facet is in a unhealthy state"
+	fi
+
+	set_default_debug_facet $facet
 
 	if [[ $facet == mds* ]]; then
 		do_facet $facet \
-			lctl set_param -n mdt.${FSNAME}*.enable_remote_dir=1 \
-				2>/dev/null
+		lctl set_param -n mdt.${FSNAME}*.enable_remote_dir=1 2>/dev/null
 	fi
 
-		label=$(devicelabel ${facet} ${!dev})
-        [ -z "$label" ] && echo no label for ${!dev} && exit 1
-        eval export ${facet}_svc=${label}
-        echo Started ${label}
-    fi
-    return $RC
+	if [[ $opts =~ .*nosvc.* ]]; then
+		echo "Start ${!dev} without service"
+	else
+
+		case $fstype in
+		ldiskfs)
+			wait_update_facet ${facet} "$E2LABEL ${!dev} \
+				2>/dev/null | grep -E ':[a-zA-Z]{3}[0-9]{4}'" \
+				"" || error "${!dev} failed to initialize!";;
+		zfs)
+			wait_update_facet ${facet} "$ZFS get -H -o value \
+				lustre:svname ${!dev} 2>/dev/null | \
+				grep -E ':[a-zA-Z]{3}[0-9]{4}'" "" ||
+				error "${!dev} failed to initialize!";;
+
+		*)
+			error "unknown fstype!";;
+		esac
+	fi
+
+	# commit the device label change to disk
+	if [[ $devicelabel =~ (:[a-zA-Z]{3}[0-9]{4}) ]]; then
+		do_facet $facet "sync; sync; sync"
+	fi
+
+
+	label=$(devicelabel ${facet} ${!dev})
+	[ -z "$label" ] && echo no label for ${!dev} && exit 1
+	eval export ${facet}_svc=${label}
+	echo Started ${label}
+
+	return $RC
 }
 
 # start facet device options
 start() {
-    local facet=$1
-    shift
-    local device=$1
-    shift
-    eval export ${facet}_dev=${device}
-    eval export ${facet}_opt=\"$@\"
+	local facet=$1
+	shift
+	local device=$1
+	shift
+	eval export ${facet}_dev=${device}
+	eval export ${facet}_opt=\"$@\"
 
-    local varname=${facet}failover_dev
-    if [ -n "${!varname}" ] ; then
-        eval export ${facet}failover_dev=${!varname}
-    else
-        eval export ${facet}failover_dev=$device
-    fi
+	local varname=${facet}failover_dev
+	if [ -n "${!varname}" ] ; then
+		eval export ${facet}failover_dev=${!varname}
+	else
+		eval export ${facet}failover_dev=$device
+	fi
 
 	local mntpt=$(facet_mntpt $facet)
 	do_facet ${facet} mkdir -p $mntpt
@@ -1301,7 +1399,7 @@ stop() {
     running=$(do_facet ${facet} "grep -c $mntpt' ' /proc/mounts") || true
     if [ ${running} -ne 0 ]; then
         echo "Stopping $mntpt (opts:$@) on $HOST"
-        do_facet ${facet} umount -d $@ $mntpt
+	do_facet ${facet} $UMOUNT $@ $mntpt
     fi
 
 	# umount should block, but we should wait for unrelated obd's
@@ -1354,70 +1452,6 @@ quota_type() {
 	return $rc
 }
 
-# XXX This function is kept for interoperability with old server (< 2.3.50),
-#     it should be removed whenever we drop the interoperability for such
-#     server.
-restore_quota_old() {
-	local mntpt=${1:-$MOUNT}
-	local quota_type=$(quota_type $FSNAME | grep MDT | cut -d "=" -f2)
-	if [ ! "$old_QUOTA_TYPE" ] ||
-		[ "$quota_type" = "$old_QUOTA_TYPE" ]; then
-		return
-	fi
-	quota_save_version $old_QUOTA_TYPE
-}
-
-# XXX This function is kept for interoperability with old server (< 2.3.50),
-#     it should be removed whenever we drop the interoperability for such
-#     server.
-setup_quota_old(){
-	local mntpt=$1
-
-	# no quota enforcement for now and accounting works out of the box
-	return
-
-    # We need save the original quota_type params, and restore them after testing
-
-    # Suppose that quota type the same on mds and ost
-    local quota_type=$(quota_type | grep MDT | cut -d "=" -f2)
-    [ ${PIPESTATUS[0]} -eq 0 ] || error "quota_type failed!"
-    echo "[HOST:$HOSTNAME] [old_quota_type:$quota_type] [new_quota_type:$QUOTA_TYPE]"
-    if [ "$quota_type" != "$QUOTA_TYPE" ]; then
-        export old_QUOTA_TYPE=$quota_type
-        quota_save_version $QUOTA_TYPE
-    else
-        qtype=$(tr -c -d "ug" <<< $QUOTA_TYPE)
-        $LFS quotacheck -$qtype $mntpt || error "quotacheck has failed for $type"
-    fi
-
-    local quota_usrs=$QUOTA_USERS
-
-    # get_filesystem_size
-    local disksz=$(lfs_df $mntpt | grep "summary"  | awk '{print $2}')
-    local blk_soft=$((disksz + 1024))
-    local blk_hard=$((blk_soft + blk_soft / 20)) # Go 5% over
-
-    local Inodes=$(lfs_df -i $mntpt | grep "summary"  | awk '{print $2}')
-    local i_soft=$Inodes
-    local i_hard=$((i_soft + i_soft / 20))
-
-    echo "Total disk size: $disksz  block-softlimit: $blk_soft block-hardlimit:
-        $blk_hard inode-softlimit: $i_soft inode-hardlimit: $i_hard"
-
-    local cmd
-    for usr in $quota_usrs; do
-        echo "Setting up quota on $HOSTNAME:$mntpt for $usr..."
-        for type in u g; do
-            cmd="$LFS setquota -$type $usr -b $blk_soft -B $blk_hard -i $i_soft -I $i_hard $mntpt"
-            echo "+ $cmd"
-            eval $cmd || error "$cmd FAILED!"
-        done
-        # display the quota status
-        echo "Quota settings for $usr : "
-        $LFS quota -v -u $usr $mntpt || true
-    done
-}
-
 # get mdt quota type
 mdt_quota_type() {
 	local varsvc=${SINGLEMDS}_svc
@@ -1435,11 +1469,6 @@ ost_quota_type() {
 
 # restore old quota type settings
 restore_quota() {
-	if [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.3.50) ]; then
-		restore_quota_old
-		return
-	fi
-
 	if [ "$old_MDT_QUOTA_TYPE" ]; then
 		do_facet mgs $LCTL conf_param \
 			$FSNAME.quota.mdt=$old_MDT_QUOTA_TYPE
@@ -1475,11 +1504,6 @@ mdt_free_inodes() {
 }
 
 setup_quota(){
-	if [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.3.50) ]; then
-		setup_quota_old $1
-		return
-	fi
-
 	local mntpt=$1
 
 	# save old quota type & set new quota type
@@ -1528,32 +1552,44 @@ setup_quota(){
 }
 
 zconf_mount() {
-    local client=$1
-    local mnt=$2
-    local opts=${3:-$MOUNT_OPTS}
-    opts=${opts:+-o $opts}
-    local flags=${4:-$MOUNT_FLAGS}
-
-    local device=$MGSNID:/$FSNAME
-    if [ -z "$mnt" -o -z "$FSNAME" ]; then
-        echo Bad zconf mount command: opt=$flags $opts dev=$device mnt=$mnt
-        exit 1
-    fi
-
-    echo "Starting client: $client: $flags $opts $device $mnt"
-    do_node $client mkdir -p $mnt
-    do_node $client $MOUNT_CMD $flags $opts $device $mnt || return 1
-
-    set_default_debug_nodes $client
+	local client=$1
+	local mnt=$2
+	local opts=${3:-$MOUNT_OPTS}
+	opts=${opts:+-o $opts}
+	local flags=${4:-$MOUNT_FLAGS}
+
+	local device=$MGSNID:/$FSNAME$FILESET
+	if [ -z "$mnt" -o -z "$FSNAME" ]; then
+		echo "Bad mount command: opt=$flags $opts dev=$device " \
+		     "mnt=$mnt"
+		exit 1
+	fi
+
+	echo "Starting client: $client: $flags $opts $device $mnt"
+	do_node $client mkdir -p $mnt
+	if [ -n "$FILESET" -a -z "$SKIP_FILESET" ];then
+		do_node $client $MOUNT_CMD $flags $opts $MGSNID:/$FSNAME \
+			$mnt || return 1
+		#disable FILESET if not supported
+		do_nodes $client lctl get_param -n \
+			mdc.$FSNAME-MDT0000*.import | grep -q subtree ||
+				device=$MGSNID:/$FSNAME
+		do_node $client mkdir -p $mnt/$FILESET
+		do_node $client "! grep -q $mnt' ' /proc/mounts ||
+			umount $mnt"
+	fi
+	do_node $client $MOUNT_CMD $flags $opts $device $mnt || return 1
+
+	set_default_debug_nodes $client
 
-    return 0
+	return 0
 }
 
 zconf_umount() {
     local client=$1
     local mnt=$2
     local force
-    local busy 
+    local busy
     local need_kill
 
     [ "$3" ] && force=-f
@@ -1594,13 +1630,13 @@ if [ \\\$running -ne \\\$mpts ]; then
     echo \\\$(hostname) env are INSANE!;
     exit 1;
 fi"
-    [ $? -eq 0 ] || rc=1 
+    [ $? -eq 0 ] || rc=1
     done
     return $rc
 }
 
 sanity_mount_check_servers () {
-    [ "$CLIENTONLY" ] && 
+    [ -n "$CLIENTONLY" ] &&
         { echo "CLIENTONLY mode, skip mount_check_servers"; return 0; } || true
     echo Checking servers environments
 
@@ -1636,21 +1672,35 @@ sanity_mount_check () {
 
 # mount clients if not mouted
 zconf_mount_clients() {
-    local clients=$1
-    local mnt=$2
-    local opts=${3:-$MOUNT_OPTS}
-    opts=${opts:+-o $opts}
-    local flags=${4:-$MOUNT_FLAGS}
-
-    local device=$MGSNID:/$FSNAME
-    if [ -z "$mnt" -o -z "$FSNAME" ]; then
-        echo Bad zconf mount command: opt=$flags $opts dev=$device mnt=$mnt
-        exit 1
-    fi
-
-    echo "Starting client $clients: $flags $opts $device $mnt"
-
-    do_nodes $clients "
+	local clients=$1
+	local mnt=$2
+	local opts=${3:-$MOUNT_OPTS}
+	opts=${opts:+-o $opts}
+	local flags=${4:-$MOUNT_FLAGS}
+
+	local device=$MGSNID:/$FSNAME$FILESET
+	if [ -z "$mnt" -o -z "$FSNAME" ]; then
+		echo "Bad conf mount command: opt=$flags $opts dev=$device " \
+		     "mnt=$mnt"
+		exit 1
+	fi
+
+	echo "Starting client $clients: $flags $opts $device $mnt"
+	if [ -n "$FILESET" -a ! -n "$SKIP_FILESET" ]; then
+		do_nodes $clients "! grep -q $mnt' ' /proc/mounts ||
+			umount $mnt"
+		do_nodes $clients $MOUNT_CMD $flags $opts $MGSNID:/$FSNAME \
+			$mnt || return 1
+		#disable FILESET if not supported
+		do_nodes $clients lctl get_param -n \
+			mdc.$FSNAME-MDT0000*.import | grep -q subtree ||
+				device=$MGSNID:/$FSNAME
+		do_nodes $clients mkdir -p $mnt/$FILESET
+		do_nodes $clients "! grep -q $mnt' ' /proc/mounts ||
+			umount $mnt"
+	fi
+
+	do_nodes $clients "
 running=\\\$(mount | grep -c $mnt' ');
 rc=0;
 if [ \\\$running -eq 0 ] ; then
@@ -1660,12 +1710,12 @@ if [ \\\$running -eq 0 ] ; then
 fi;
 exit \\\$rc" || return ${PIPESTATUS[0]}
 
-    echo "Started clients $clients: "
-    do_nodes $clients "mount | grep $mnt' '"
+	echo "Started clients $clients: "
+	do_nodes $clients "mount | grep $mnt' '"
 
-    set_default_debug_nodes $clients
+	set_default_debug_nodes $clients
 
-    return 0
+	return 0
 }
 
 zconf_umount_clients() {
@@ -1708,7 +1758,7 @@ shutdown_node_hard () {
         wait_for_function --quiet "! ping -w 3 -c 1 $host" 5 1 && return 0
         echo "waiting for $host to fail attempts=$attempts"
         [ $i -lt $attempts ] || \
-            { echo "$host still pingable after power down! attempts=$attempts" && return 1; } 
+            { echo "$host still pingable after power down! attempts=$attempts" && return 1; }
     done
 }
 
@@ -2107,16 +2157,19 @@ wait_update () {
 }
 
 wait_update_facet() {
+	local verbose=
+	[ "$1" = "--verbose" ] && verbose="$1" && shift
+
 	local facet=$1
 	shift
-	wait_update $(facet_active_host $facet) "$@"
+	wait_update $verbose $(facet_active_host $facet) "$@"
 }
 
 sync_all_data() {
 	do_nodes $(comma_list $(mdts_nodes)) \
-	    "lctl set_param -n osd*.*MDT*.force_sync 1"
+	    "lctl set_param -n osd*.*MDT*.force_sync=1"
 	do_nodes $(comma_list $(osts_nodes)) \
-	    "lctl set_param -n osd*.*OS*.force_sync 1" 2>&1 |
+	    "lctl set_param -n osd*.*OS*.force_sync=1" 2>&1 |
 		grep -v 'Found no match'
 }
 
@@ -2399,16 +2452,17 @@ client_evicted() {
 }
 
 client_reconnect_try() {
-    uname -n >> $MOUNT/recon
-    if [ -z "$CLIENTS" ]; then
-        df $MOUNT; uname -n >> $MOUNT/recon
-    else
-        do_nodes $CLIENTS "df $MOUNT; uname -n >> $MOUNT/recon" > /dev/null
-    fi
-    echo Connected clients:
-    cat $MOUNT/recon
-    ls -l $MOUNT/recon > /dev/null
-    rm $MOUNT/recon
+	local f=$MOUNT/recon
+
+	uname -n >> $f
+	if [ -z "$CLIENTS" ]; then
+		$LFS df $MOUNT; uname -n >> $f
+	else
+		do_nodes $CLIENTS "$LFS df $MOUNT; uname -n >> $f" > /dev/null
+	fi
+	echo "Connected clients: $(cat $f)"
+	ls -l $f > /dev/null
+	rm $f
 }
 
 client_reconnect() {
@@ -2432,6 +2486,13 @@ affected_facets () {
 }
 
 facet_failover() {
+	local E2FSCK_ON_MDT0=false
+	if [ "$1" == "--fsck" ]; then
+		shift
+		[ $(facet_fstype $SINGLEMDS) == ldiskfs ] &&
+			E2FSCK_ON_MDT0=true
+	fi
+
 	local facets=$1
 	local sleep_time=$2
 	local -a affecteds
@@ -2465,6 +2526,9 @@ facet_failover() {
 		shutdown_facet $facet
 	done
 
+	$E2FSCK_ON_MDT0 && (run_e2fsck $(facet_active_host $SINGLEMDS) \
+		$(mdsdevname 1) "-n" || error "Running e2fsck")
+
 	for ((index=0; index<$total; index++)); do
 		facet=$(echo ${affecteds[index]} | tr -s " " | cut -d"," -f 1)
 		echo reboot facets: ${affecteds[index]}
@@ -2493,7 +2557,7 @@ obd_name() {
 replay_barrier() {
 	local facet=$1
 	do_facet $facet "sync; sync; sync"
-	df $MOUNT
+	$LFS df $MOUNT
 
 	# make sure there will be no seq change
 	local clients=${CLIENTS:-$HOSTNAME}
@@ -2578,7 +2642,7 @@ fail() {
 
 	facet_failover $* || error "failover: $?"
 	wait_clients_import_state "$clients" "$facets" FULL
-	clients_up || error "post-failover df: $?"
+	clients_up || error "post-failover stat: $?"
 }
 
 fail_nodf() {
@@ -2592,8 +2656,8 @@ fail_abort() {
 	change_active $facet
 	wait_for_facet $facet
 	mount_facet $facet -o abort_recovery
-	clients_up || echo "first df failed: $?"
-	clients_up || error "post-failover df: $?"
+	clients_up || echo "first stat failed: $?"
+	clients_up || error "post-failover stat: $?"
 }
 
 do_lmc() {
@@ -2934,6 +2998,12 @@ get_env_vars() {
 		fi
 	done
 
+	for var in VERBOSE; do
+		if [ -n "${!var}" ]; then
+			echo -n " $var=${!var}"
+		fi
+	done
+
 	if [ -n "$FSTYPE" ]; then
 		echo -n " FSTYPE=$FSTYPE"
 	fi
@@ -3194,7 +3264,7 @@ facet_mntpt () {
     [[ $facet = mgs ]] && combined_mgs_mds && facet="mds1"
 
     local var=${facet}_MOUNT
-    eval mntpt=${!var:-${MOUNT%/*}/$facet}
+    eval mntpt=${!var:-${MOUNT}-$facet}
 
     echo -n $mntpt
 }
@@ -3202,7 +3272,7 @@ facet_mntpt () {
 mount_ldiskfs() {
 	local facet=$1
 	local dev=$(facet_device $facet)
-	local mnt=$(facet_mntpt $facet)
+	local mnt=${2:-$(facet_mntpt $facet)}
 	local opts
 
 	if ! do_facet $facet test -b $dev; then
@@ -3214,9 +3284,9 @@ mount_ldiskfs() {
 unmount_ldiskfs() {
 	local facet=$1
 	local dev=$(facet_device $facet)
-	local mnt=$(facet_mntpt $facet)
+	local mnt=${2:-$(facet_mntpt $facet)}
 
-	do_facet $facet umount -d $mnt
+	do_facet $facet $UMOUNT $mnt
 }
 
 var_name() {
@@ -3226,7 +3296,7 @@ var_name() {
 mount_zfs() {
 	local facet=$1
 	local ds=$(facet_device $facet)
-	local mnt=$(facet_mntpt $facet)
+	local mnt=${2:-$(facet_mntpt $facet)}
 	local canmnt
 	local mntpt
 
@@ -3249,7 +3319,7 @@ mount_zfs() {
 unmount_zfs() {
 	local facet=$1
 	local ds=$(facet_device $facet)
-	local mnt=$(facet_mntpt $facet)
+	local mnt=${2:-$(facet_mntpt $facet)}
 	local var_mntpt=mz_$(var_name ${facet}_$ds)_mountpoint
 	local var_canmnt=mz_$(var_name ${facet}_$ds)_canmount
 	local mntpt=${!var_mntpt}
@@ -3265,16 +3335,18 @@ unmount_zfs() {
 
 mount_fstype() {
 	local facet=$1
+	local mnt=$2
 	local fstype=$(facet_fstype $facet)
 
-	mount_$fstype $facet
+	mount_$fstype $facet $mnt
 }
 
 unmount_fstype() {
 	local facet=$1
+	local mnt=$2
 	local fstype=$(facet_fstype $facet)
 
-	unmount_$fstype $facet
+	unmount_$fstype $facet $mnt
 }
 
 ########
@@ -3294,7 +3366,8 @@ stopall() {
     zconf_umount_clients $clients $MOUNT "$*" || true
     [ -n "$MOUNT2" ] && zconf_umount_clients $clients $MOUNT2 "$*" || true
 
-    [ "$CLIENTONLY" ] && return
+    [ -n "$CLIENTONLY" ] && return
+
     # The add fn does rm ${facet}active file, this would be enough
     # if we use do_facet <facet> only after the facet added, but
     # currently we use do_facet mds in local.sh
@@ -3317,16 +3390,19 @@ stopall() {
 }
 
 cleanup_echo_devs () {
-    local devs=$($LCTL dl | grep echo | awk '{print $4}')
+	trap 0
+	local dev
+	local devs=$($LCTL dl | grep echo | awk '{print $4}')
 
-    for dev in $devs; do
-        $LCTL --device $dev cleanup
-        $LCTL --device $dev detach
-    done
+	for dev in $devs; do
+		$LCTL --device $dev cleanup
+		$LCTL --device $dev detach
+	done
 }
 
 cleanupall() {
     nfs_client_mode && return
+	cifs_client_mode && return
 
     stopall $*
     cleanup_echo_devs
@@ -3382,7 +3458,6 @@ mkfs_opts() {
 	opts+=${LDLM_TIMEOUT:+" --param=sys.ldlm_timeout=$LDLM_TIMEOUT"}
 
 	if [ $type == MDS ]; then
-		opts+=${SECLEVEL:+" --param=mdt.sec_level"}
 		opts+=${MDSCAPA:+" --param-mdt.capa=$MDSCAPA"}
 		opts+=${STRIPE_BYTES:+" --param=lov.stripesize=$STRIPE_BYTES"}
 		opts+=${STRIPES_PER_OBJ:+" --param=lov.stripecount=$STRIPES_PER_OBJ"}
@@ -3395,20 +3470,26 @@ mkfs_opts() {
 				fs_mkfs_opts+="-O large_xattr"
 			fi
 
-			fs_mkfs_opts+=${MDSJOURNALSIZE:+" -J size=$MDSJOURNALSIZE"}
-			if [ ! -z $EJOURNAL ]; then
-				fs_mkfs_opts+=${MDSJOURNALSIZE:+" device=$EJOURNAL"}
+			var=${facet}_JRN
+			if [ -n "${!var}" ]; then
+				fs_mkfs_opts+=" -J device=${!var}"
+			else
+				fs_mkfs_opts+=${MDSJOURNALSIZE:+" -J size=$MDSJOURNALSIZE"}
 			fi
 			fs_mkfs_opts+=${MDSISIZE:+" -i $MDSISIZE"}
 		fi
 	fi
 
 	if [ $type == OST ]; then
-		opts+=${SECLEVEL:+" --param=ost.sec_level"}
 		opts+=${OSSCAPA:+" --param=ost.capa=$OSSCAPA"}
 
 		if [ $fstype == ldiskfs ]; then
-			fs_mkfs_opts+=${OSTJOURNALSIZE:+" -J size=$OSTJOURNALSIZE"}
+			var=${facet}_JRN
+			if [ -n "${!var}" ]; then
+				fs_mkfs_opts+=" -J device=${!var}"
+			else
+				fs_mkfs_opts+=${OSTJOURNALSIZE:+" -J size=$OSTJOURNALSIZE"}
+			fi
 		fi
 	fi
 
@@ -3453,39 +3534,70 @@ check_ost_indices() {
 	done
 }
 
-formatall() {
+format_mgs() {
+	local quiet
+
+	if ! $VERBOSE; then
+		quiet=yes
+	fi
+	echo "Format mgs: $(mgsdevname)"
+	reformat_external_journal mgs
+	add mgs $(mkfs_opts mgs $(mgsdevname)) --reformat \
+		$(mgsdevname) $(mgsvdevname) ${quiet:+>/dev/null} || exit 10
+}
+
+format_mdt() {
+	local num=$1
 	local quiet
 
 	if ! $VERBOSE; then
 		quiet=yes
 	fi
+	echo "Format mds$num: $(mdsdevname $num)"
+	reformat_external_journal mds$num
+	add mds$num $(mkfs_opts mds$num $(mdsdevname ${num})) \
+		--reformat $(mdsdevname $num) $(mdsvdevname $num) \
+		${quiet:+>/dev/null} || exit 10
+}
 
+format_ost() {
+	local num=$1
+
+	if ! $VERBOSE; then
+		quiet=yes
+	fi
+	echo "Format ost$num: $(ostdevname $num)"
+	reformat_external_journal ost$num
+	add ost$num $(mkfs_opts ost$num $(ostdevname ${num})) \
+		--reformat $(ostdevname $num) $(ostvdevname ${num}) \
+		${quiet:+>/dev/null} || exit 10
+}
+
+formatall() {
 	stopall
+	# Set hostid for ZFS/SPL zpool import protection
+	# (Assumes MDS version is also OSS version)
+	if [ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.8.54) ];
+	then
+	    do_rpc_nodes "$(comma_list $(remote_nodes_list))" set_hostid
+	fi
+
 	# We need ldiskfs here, may as well load them all
 	load_modules
-	[ "$CLIENTONLY" ] && return
+	[ -n "$CLIENTONLY" ] && return
 	echo Formatting mgs, mds, osts
 	if ! combined_mgs_mds ; then
-		echo "Format mgs: $(mgsdevname)"
-		add mgs $(mkfs_opts mgs $(mgsdevname)) --reformat \
-			$(mgsdevname) $(mgsvdevname) ${quiet:+>/dev/null} ||
-			exit 10
+		format_mgs
 	fi
 
 	for num in $(seq $MDSCOUNT); do
-		echo "Format mds$num: $(mdsdevname $num)"
-		add mds$num $(mkfs_opts mds$num $(mdsdevname ${num})) \
-			--reformat $(mdsdevname $num) $(mdsvdevname $num) \
-			${quiet:+>/dev/null} || exit 10
+		format_mdt $num
 	done
 
 	export OST_INDICES=($(hostlist_expand "$OST_INDEX_LIST"))
 	check_ost_indices
 	for num in $(seq $OSTCOUNT); do
-		echo "Format ost$num: $(ostdevname $num)"
-		add ost$num $(mkfs_opts ost$num $(ostdevname ${num})) \
-			--reformat $(ostdevname $num) $(ostvdevname ${num}) \
-			${quiet:+>/dev/null} || exit 10
+		format_ost $num
 	done
 }
 
@@ -3562,20 +3674,19 @@ writeconf_all () {
 }
 
 setupall() {
-    nfs_client_mode && return
+	nfs_client_mode && return
+	cifs_client_mode && return
 
-    sanity_mount_check ||
-        error "environments are insane!"
+	sanity_mount_check || error "environments are insane!"
 
-    load_modules
+	load_modules
 
-    if [ -z "$CLIENTONLY" ]; then
-        echo Setup mgs, mdt, osts
-        echo $WRITECONF | grep -q "writeconf" && \
-            writeconf_all
-        if ! combined_mgs_mds ; then
+	if [ -z "$CLIENTONLY" ]; then
+		echo Setup mgs, mdt, osts
+		echo $WRITECONF | grep -q "writeconf" && writeconf_all
+		if ! combined_mgs_mds ; then
 			start mgs $(mgsdevname) $MGS_MOUNT_OPTS
-        fi
+		fi
 
         for num in `seq $MDSCOUNT`; do
             DEVNAME=$(mdsdevname $num)
@@ -3644,7 +3755,7 @@ mounted_lustre_filesystems() {
 }
 
 init_facet_vars () {
-	[ "$CLIENTONLY" ] && return 0
+	[ -n "$CLIENTONLY" ] && return 0
 	local facet=$1
 	shift
 	local device=$1
@@ -3769,14 +3880,17 @@ set_conf_param_and_check() {
 	do_facet mgs "$LCTL conf_param $PARAM='$FINAL'" ||
 		error "conf_param $PARAM failed"
 
-	wait_update $(facet_host $myfacet) "$TEST" "$FINAL" ||
+	wait_update_facet $myfacet "$TEST" "$FINAL" ||
 		error "check $PARAM failed!"
 }
 
 init_param_vars () {
-	remote_mds_nodsh ||
-		TIMEOUT=$(do_facet $SINGLEMDS "lctl get_param -n timeout")
+	TIMEOUT=$(lctl get_param -n timeout)
+	TIMEOUT=${TIMEOUT:-20}
+
+	remote_mds_nodsh && log "Using TIMEOUT=$TIMEOUT" && return 0
 
+	TIMEOUT=$(do_facet $SINGLEMDS "lctl get_param -n timeout")
 	log "Using TIMEOUT=$TIMEOUT"
 
 	osc_ensure_active $SINGLEMDS $TIMEOUT
@@ -3828,18 +3942,23 @@ nfs_client_mode () {
     return 1
 }
 
+cifs_client_mode () {
+	[ x$CIFSCLIENT = xyes ] &&
+		echo "CIFSCLIENT=$CIFSCLIENT mode: setup, cleanup, check config skipped"
+}
+
 check_config_client () {
     local mntpt=$1
 
     local mounted=$(mount | grep " $mntpt ")
-    if [ "$CLIENTONLY" ]; then
+    if [ -n "$CLIENTONLY" ]; then
         # bug 18021
         # CLIENTONLY should not depend on *_HOST settings
         local mgc=$($LCTL device_list | awk '/MGC/ {print $4}')
         # in theory someone could create a new,
         # client-only config file that assumed lustre was already
         # configured and didn't set the MGSNID. If MGSNID is not set,
-        # then we should use the mgs nid currently being used 
+        # then we should use the mgs nid currently being used
         # as the default value. bug 18021
         [[ x$MGSNID = x ]] &&
             MGSNID=${mgc//MGC/}
@@ -3854,9 +3973,9 @@ check_config_client () {
         return 0
     fi
 
-    local myMGS_host=$mgs_HOST   
+    local myMGS_host=$mgs_HOST
     if [ "$NETTYPE" = "ptl" ]; then
-        myMGS_host=$(h2ptl $mgs_HOST | sed -e s/@ptl//) 
+        myMGS_host=$(h2ptl $mgs_HOST | sed -e s/@ptl//)
     fi
 
     echo Checking config lustre mounted on $mntpt
@@ -3875,6 +3994,7 @@ check_config_clients () {
 	local mntpt=$1
 
 	nfs_client_mode && return
+	cifs_client_mode && return
 
 	do_rpc_nodes "$clients" check_config_client $mntpt
 
@@ -3920,20 +4040,22 @@ is_empty_fs() {
 }
 
 check_and_setup_lustre() {
-    nfs_client_mode && return
+	sanitize_parameters
+	nfs_client_mode && return
+	cifs_client_mode && return
+
+	local MOUNTED=$(mounted_lustre_filesystems)
 
-    local MOUNTED=$(mounted_lustre_filesystems)
-
-    local do_check=true
-    # 1.
-    # both MOUNT and MOUNT2 are not mounted
-    if ! is_mounted $MOUNT && ! is_mounted $MOUNT2; then
-        [ "$REFORMAT" ] && formatall
-        # setupall mounts both MOUNT and MOUNT2 (if MOUNT_2 is set)
-        setupall
-        is_mounted $MOUNT || error "NAME=$NAME not mounted"
-        export I_MOUNTED=yes
-        do_check=false
+	local do_check=true
+	# 1.
+	# both MOUNT and MOUNT2 are not mounted
+	if ! is_mounted $MOUNT && ! is_mounted $MOUNT2; then
+		[ "$REFORMAT" = "yes" ] && formatall
+		# setupall mounts both MOUNT and MOUNT2 (if MOUNT_2 is set)
+		setupall
+		is_mounted $MOUNT || error "NAME=$NAME not mounted"
+		export I_MOUNTED=yes
+		do_check=false
     # 2.
     # MOUNT2 is mounted
     elif is_mounted $MOUNT2; then
@@ -3956,7 +4078,7 @@ check_and_setup_lustre() {
                     restore_mount $MOUNT2
                     export I_MOUNTED2=yes
                 fi
-            fi 
+            fi
 
     # 5.
     # MOUNT is mounted MOUNT2 is not mounted
@@ -3977,7 +4099,7 @@ check_and_setup_lustre() {
         set_default_debug_nodes $(comma_list $(nodes_list))
     fi
 
-	if [ $(lower $OSD_TRACK_DECLARES_LBUG) == 'yes' ] ; then
+	if [ -z "$CLIENTONLY" -a $(lower $OSD_TRACK_DECLARES_LBUG) == 'yes' ]; then
 		local facets=""
 		[ "$(facet_fstype ost1)" = "ldiskfs" ] &&
 			facets="$(get_facets OST)"
@@ -3997,12 +4119,14 @@ check_and_setup_lustre() {
 		set_flavor_all $SEC
 	fi
 
-	#Enable remote MDT create for testing
-	for num in $(seq $MDSCOUNT); do
-		do_facet mds$num \
-			lctl set_param -n mdt.${FSNAME}*.enable_remote_dir=1 \
-				2>/dev/null
-	done
+	if [ -z "$CLIENTONLY" ]; then
+		# Enable remote MDT create for testing
+		for num in $(seq $MDSCOUNT); do
+			do_facet mds$num \
+				lctl set_param -n mdt.${FSNAME}*.enable_remote_dir=1 \
+					2>/dev/null
+		done
+	fi
 
 	if [ "$ONLY" == "setup" ]; then
 		exit 0
@@ -4133,88 +4257,31 @@ check_shared_dir() {
 	return 0
 }
 
-# Run e2fsck on MDT and OST(s) to generate databases used for lfsck.
-generate_db() {
-	local i
-	local ostidx
-	local dev
-	local node
-
-	[[ $(lustre_version_code $SINGLEMDS) -ne $(version_code 2.2.0) ]] ||
-		{ skip "Lustre 2.2.0 lacks the patch for LU-1255"; exit 0; }
-
-	check_shared_dir $SHARED_DIRECTORY ||
-		error "$SHARED_DIRECTORY isn't a shared directory"
-
-	export MDSDB=$SHARED_DIRECTORY/mdsdb
-	export OSTDB=$SHARED_DIRECTORY/ostdb
-
-	# DNE is not supported, so when running e2fsck on a DNE filesystem,
-	# we only pass master MDS parameters.
-	run_e2fsck $MDTNODE $MDTDEV "-n --mdsdb $MDSDB"
-
-    i=0
-    ostidx=0
-    OSTDB_LIST=""
-    for node in $(osts_nodes); do
-        for dev in ${OSTDEVS[i]}; do
-            run_e2fsck $node $dev "-n --mdsdb $MDSDB --ostdb $OSTDB-$ostidx"
-            OSTDB_LIST="$OSTDB_LIST $OSTDB-$ostidx"
-            ostidx=$((ostidx + 1))
-        done
-        i=$((i + 1))
-    done
-}
-
-# Run lfsck on server node if lfsck can't be found on client (LU-2571)
-run_lfsck_remote() {
-	local cmd="$LFSCK_BIN -c -l --mdsdb $MDSDB --ostdb $OSTDB_LIST $MOUNT"
-	local client=$1
-	local mounted=true
-	local rc=0
-
-	#Check if lustre is already mounted
-	do_rpc_nodes $client is_mounted $MOUNT || mounted=false
-	if ! $mounted; then
-		zconf_mount $client $MOUNT ||
-			error "failed to mount Lustre on $client"
-	fi
-	#Run lfsck
-	echo $cmd
-	do_node $client $cmd || rc=$?
-	#Umount if necessary
-	if ! $mounted; then
-		zconf_umount $client $MOUNT ||
-			error "failed to unmount Lustre on $client"
-	fi
-
-	[ $rc -le $FSCK_MAX_ERR ] ||
-		error "$cmd returned $rc, should be <= $FSCK_MAX_ERR"
-	echo "lfsck finished with rc=$rc"
-
-	return $rc
-}
-
 run_lfsck() {
-	local facets="client $SINGLEMDS"
-	local found=false
-	local facet
-	local node
-	local rc=0
-
-	for facet in $facets; do
-		node=$(facet_active_host $facet)
-		if check_progs_installed $node $LFSCK_BIN; then
-			found=true
-			break
-		fi
+	do_nodes $(comma_list $(mdts_nodes) $(osts_nodes)) \
+		$LCTL set_param printk=+lfsck
+	do_facet $SINGLEMDS "$LCTL lfsck_start -M $FSNAME-MDT0000 -r -A -t all"
+
+	for k in $(seq $MDSCOUNT); do
+		# wait up to 10+1 minutes for LFSCK to complete
+		wait_update_facet --verbose mds${k} "$LCTL get_param -n \
+			mdd.$(facet_svc mds${k}).lfsck_layout |
+			awk '/^status/ { print \\\$2 }'" "completed" 600 ||
+			error "MDS${k} layout isn't the expected 'completed'"
+		wait_update_facet --verbose mds${k} "$LCTL get_param -n \
+			mdd.$(facet_svc mds${k}).lfsck_namespace |
+			awk '/^status/ { print \\\$2 }'" "completed" 60 ||
+			error "MDS${k} namespace isn't the expected 'completed'"
 	done
-	! $found && error "None of \"$facets\" supports lfsck"
-
-	run_lfsck_remote $node || rc=$?
-
-	rm -rvf $MDSDB* $OSTDB* || true
-	return $rc
+	local rep_mdt=$(do_nodes $(comma_list $(mdts_nodes)) \
+			$LCTL get_param -n mdd.$FSNAME-*.lfsck_* |
+			awk '/repaired/ { print $2 }' | calc_sum)
+	local rep_ost=$(do_nodes $(comma_list $(osts_nodes)) \
+			$LCTL get_param -n obdfilter.$FSNAME-*.lfsck_* |
+			awk '/repaired/ { print $2 }' | calc_sum)
+	local repaired=$((rep_mdt + rep_ost))
+	[ $repaired -eq 0 ] ||
+		error "lfsck repaired $rep_mdt MDT and $rep_ost OST errors"
 }
 
 dump_file_contents() {
@@ -4272,15 +4339,18 @@ log_zfs_info() {
 }
 
 check_and_cleanup_lustre() {
-    if [ "$LFSCK_ALWAYS" = "yes" -a "$TESTSUITE" != "lfsck" ]; then
-        get_svr_devs
-        generate_db
-        run_lfsck
-    fi
+	if [ "$LFSCK_ALWAYS" = "yes" -a "$TESTSUITE" != "sanity-lfsck" -a \
+	     "$TESTSUITE" != "sanity-scrub" ]; then
+		run_lfsck
+	fi
 
 	if is_mounted $MOUNT; then
-		[ -n "$DIR" ] && rm -rf $DIR/[Rdfs][0-9]* ||
-			error "remove sub-test dirs failed"
+		if $DO_CLEANUP; then
+			[ -n "$DIR" ] && rm -rf $DIR/[Rdfs][0-9]* ||
+				error "remove sub-test dirs failed"
+		else
+			echo "skip cleanup"
+		fi
 		[ "$ENABLE_QUOTA" ] && restore_quota || true
 	fi
 
@@ -4603,37 +4673,15 @@ clear_failloc() {
 }
 
 set_nodes_failloc () {
-	do_nodes $(comma_list $1)  lctl set_param fail_val=0 fail_loc=$2
+	local fv=${3:-0}
+	do_nodes $(comma_list $1)  lctl set_param fail_val=$fv fail_loc=$2
 }
 
 cancel_lru_locks() {
-    $LCTL mark "cancel_lru_locks $1 start"
-
-    if [ $1 != "MGC" ]; then
-	for d in $(lctl get_param -N ldlm.namespaces.*.lru_size |
-		 egrep -i $1); do
-	    $LCTL set_param -n $d=clear
-	done
-	$LCTL get_param ldlm.namespaces.*.lock_unused_count | egrep -i $1 |
-		grep -v '=0'
-    else
-	for d in $(find \
-		/{proc,sys}/fs/lustre/ldlm/namespaces/*$1*/lru_size \
-		2> /dev/null); do
-	    echo "clear" > $d
-	done
-
-	for d in $(find \
-		/{proc,sys}/fs/lustre/ldlm/namespaces/*$1*/lock_unused_count \
-		2> /dev/null); do
-	    if [ $(cat $d) != 0 ]; then
-		echo "ldlm.namespaces.$(echo "$d" |
-			cut -f 7 -d'/').lock_unused_count=$(cat $d)"
-	    fi
-	done
-    fi
-
-    $LCTL mark "cancel_lru_locks $1 stop"
+	#$LCTL mark "cancel_lru_locks $1 start"
+	$LCTL set_param -n ldlm.namespaces.*$1*.lru_size=clear
+	$LCTL get_param ldlm.namespaces.*$1*.lock_unused_count | grep -v '=0'
+	#$LCTL mark "cancel_lru_locks $1 stop"
 }
 
 default_lru_size()
@@ -4673,13 +4721,21 @@ pgcache_empty() {
 }
 
 debugsave() {
-    DEBUGSAVE="$(lctl get_param -n debug)"
+	DEBUGSAVE="$(lctl get_param -n debug)"
+	DEBUGSAVE_SERVER=$(do_facet $SINGLEMDS "$LCTL get_param -n debug")
 }
 
 debugrestore() {
-    [ -n "$DEBUGSAVE" ] && \
-        do_nodes $(comma_list $(nodes_list)) "$LCTL set_param debug=\\\"${DEBUGSAVE}\\\";"
-    DEBUGSAVE=""
+	[ -n "$DEBUGSAVE" ] &&
+		do_nodes $CLIENTS "$LCTL set_param debug=\\\"${DEBUGSAVE}\\\""||
+		true
+	DEBUGSAVE=""
+
+	[ -n "DEBUGSAVE_SERVER" ] &&
+		do_nodes $(comma_list $(all_server_nodes)) \
+			 "$LCTL set_param debug=\\\"${DEBUGSAVE_SERVER}\\\"" ||
+			 true
+	DEBUGSAVE_SERVER=""
 }
 
 debug_size_save() {
@@ -4710,20 +4766,17 @@ stop_full_debug_logging() {
 
 # prints bash call stack
 print_stack_trace() {
+	local skip=${1:-1}
 	echo "  Trace dump:"
-	for (( i=1; i < ${#BASH_LINENO[*]} ; i++ )) ; do
-		local s=${BASH_SOURCE[$i]}
-		local l=${BASH_LINENO[$i-1]}
-		local f=${FUNCNAME[$i]}
-		echo "  = $s:$l:$f()"
+	for (( i=$skip; i < ${#BASH_LINENO[*]} ; i++ )) ; do
+		local src=${BASH_SOURCE[$i]}
+		local lineno=${BASH_LINENO[$i-1]}
+		local funcname=${FUNCNAME[$i]}
+		echo "  = $src:$lineno:$funcname()"
 	done
 }
 
-##################################
-# Test interface
-##################################
-
-error_noexit() {
+report_error() {
 	local TYPE=${TYPE:-"FAIL"}
 
 	local dump=true
@@ -4733,10 +4786,8 @@ error_noexit() {
 		dump=false
 	fi
 
-
 	log " ${TESTSUITE} ${TESTNAME}: @@@@@@ ${TYPE}: $@ "
-	print_stack_trace >&2
-
+	(print_stack_trace 2) >&2
 	mkdir -p $LOGDIR
 	# We need to dump the logs on all nodes
 	if $dump; then
@@ -4755,6 +4806,17 @@ error_noexit() {
 			echo "$@" > $LOGDIR/err
 		fi
 	fi
+
+	# cleanup the env for failed tests
+	reset_fail_loc
+}
+
+##################################
+# Test interface
+##################################
+
+error_noexit() {
+	report_error "$@"
 }
 
 exit_status () {
@@ -4766,12 +4828,13 @@ exit_status () {
 }
 
 error() {
-	error_noexit "$@"
+	report_error "$@"
 	exit 1
 }
 
 error_exit() {
-	error "$@"
+	report_error "$@"
+	exit 1
 }
 
 # use only if we are ignoring failures for this test, bugno required.
@@ -4781,15 +4844,27 @@ error_exit() {
 error_ignore() {
 	local TYPE="IGNORE ($1)"
 	shift
-	error_noexit "$@"
+	report_error "$@"
 }
 
 error_and_remount() {
-	error_noexit "$@"
+	report_error "$@"
 	remount_client $MOUNT
 	exit 1
 }
 
+# Throw an error if it's not running in vm - usually for performance
+# verification
+error_not_in_vm() {
+	local virt=$(running_in_vm)
+	if [[ -n "$virt" ]]; then
+		echo "running in VM '$virt', ignore error"
+		error_ignore env=$virt "$@"
+	else
+		error "$@"
+	fi
+}
+
 skip_env () {
 	$FAIL_ON_SKIP_ENV && error false $@ || skip $@
 }
@@ -4975,17 +5050,17 @@ check_mds() {
 }
 
 reset_fail_loc () {
-    echo -n "Resetting fail_loc on all nodes..."
-    do_nodes $(comma_list $(nodes_list)) "lctl set_param -n fail_loc=0 \
-	    fail_val=0 2>/dev/null || true"
-    echo done.
+	echo -n "Resetting fail_loc on all nodes..."
+	do_nodes $(comma_list $(nodes_list)) "lctl set_param -n fail_loc=0 \
+	    fail_val=0 2>/dev/null" || true
+	echo done.
 }
 
 
 #
-# Log a message (on all nodes) padded with "=" before and after. 
+# Log a message (on all nodes) padded with "=" before and after.
 # Also appends a timestamp and prepends the testsuite name.
-# 
+#
 
 EQUALS="===================================================================================================="
 banner() {
@@ -4997,6 +5072,18 @@ banner() {
     log "$msg== $(date +"%H:%M:%S (%s)")"
 }
 
+check_dmesg_for_errors() {
+	local res
+	local errors="VFS: Busy inodes after unmount of\|\
+ldiskfs_check_descriptors: Checksum for group 0 failed\|\
+group descriptors corrupted"
+
+	res=$(do_nodes $(comma_list $(nodes_list)) "dmesg" | grep "$errors")
+	[ -z "$res" ] && return 0
+	echo "Kernel error detected: $res"
+	return 1
+}
+
 #
 # Run a single test function and cleanup after it.
 #
@@ -5012,12 +5099,17 @@ run_one() {
 	local SAVE_UMASK=`umask`
 	umask 0022
 
+	if ! grep -q $DIR /proc/mounts; then
+		$SETUP
+	fi
+
 	banner "test $testnum: $message"
 	test_${testnum} || error "test_$testnum failed with $?"
 	cd $SAVE_PWD
 	reset_fail_loc
 	check_grant ${testnum} || error "check_grant $testnum failed with $?"
 	check_catastrophe || error "LBUG/LASSERT detected"
+	check_dmesg_for_errors || error "Error in dmesg detected"
 	if [ "$PARALLEL" != "yes" ]; then
 		ps auxww | grep -v grep | grep -q multiop &&
 					error "multiop still running"
@@ -5026,6 +5118,7 @@ run_one() {
 	unset tdir
 	unset tfile
 	umask $SAVE_UMASK
+	$CLEANUP
 	return 0
 }
 
@@ -5218,15 +5311,15 @@ remote_mds ()
 
 remote_mds_nodsh()
 {
-    [ "$CLIENTONLY" ] && return 0 || true
-    remote_mds && [ "$PDSH" = "no_dsh" -o -z "$PDSH" -o -z "$mds_HOST" ]
+	[ -n "$CLIENTONLY" ] && return 0 || true
+	remote_mds && [ "$PDSH" = "no_dsh" -o -z "$PDSH" -o -z "$mds_HOST" ]
 }
 
 require_dsh_mds()
 {
-        remote_mds_nodsh && echo "SKIP: $TESTSUITE: remote MDS with nodsh" && \
-            MSKIPPED=1 && return 1
-        return 0
+	remote_mds_nodsh && echo "SKIP: $TESTSUITE: remote MDS with nodsh" &&
+		MSKIPPED=1 && return 1
+	return 0
 }
 
 remote_ost ()
@@ -5240,8 +5333,8 @@ remote_ost ()
 
 remote_ost_nodsh()
 {
-    [ "$CLIENTONLY" ] && return 0 || true 
-    remote_ost && [ "$PDSH" = "no_dsh" -o -z "$PDSH" -o -z "$ost_HOST" ]
+	[ -n "$CLIENTONLY" ] && return 0 || true
+	remote_ost && [ "$PDSH" = "no_dsh" -o -z "$PDSH" -o -z "$ost_HOST" ]
 }
 
 require_dsh_ost()
@@ -5253,10 +5346,10 @@ require_dsh_ost()
 
 remote_mgs_nodsh()
 {
-	[ "$CLIENTONLY" ] && return 0 || true
-    local MGS 
-    MGS=$(facet_host mgs)
-    remote_node $MGS && [ "$PDSH" = "no_dsh" -o -z "$PDSH" -o -z "$ost_HOST" ]
+	[ -n "$CLIENTONLY" ] && return 0 || true
+	local MGS
+	MGS=$(facet_host mgs)
+	remote_node $MGS && [ "$PDSH" = "no_dsh" -o -z "$PDSH" -o -z "$ost_HOST" ]
 }
 
 local_mode ()
@@ -5417,19 +5510,14 @@ get_random_entry () {
     rnodes=${rnodes//,/ }
 
     local -a nodes=($rnodes)
-    local num=${#nodes[@]} 
+    local num=${#nodes[@]}
     local i=$((RANDOM * num * 2 / 65536))
 
     echo ${nodes[i]}
 }
 
 client_only () {
-    [ "$CLIENTONLY" ] || [ "$CLIENTMODSONLY" = yes ]
-}
-
-is_patchless ()
-{
-    lctl get_param version | grep -q patchless
+	[ -n "$CLIENTONLY" ] || [ "x$CLIENTMODSONLY" = "xyes" ]
 }
 
 check_versions () {
@@ -5778,19 +5866,21 @@ mds_on_old_device() {
 }
 
 get_mdtosc_proc_path() {
-    local mds_facet=$1
-    local ost_label=${2:-"*OST*"}
-
-    [ "$mds_facet" = "mds" ] && mds_facet=$SINGLEMDS
-    local mdt_label=$(convert_facet2label $mds_facet)
-    local mdt_index=$(echo $mdt_label | sed -e 's/^.*-//')
-
-    if [ $(lustre_version_code $mds_facet) -le $(version_code 1.8.0) ] ||
-       mds_on_old_device $mds_facet; then
-        echo "${ost_label}-osc"
-    else
-        echo "${ost_label}-osc-${mdt_index}"
-    fi
+	local mds_facet=$1
+	local ost_label=${2:-"*OST*"}
+
+	[ "$mds_facet" = "mds" ] && mds_facet=$SINGLEMDS
+	local mdt_label=$(convert_facet2label $mds_facet)
+	local mdt_index=$(echo $mdt_label | sed -e 's/^.*-//')
+
+	if [ $(lustre_version_code $mds_facet) -le $(version_code 1.8.0) ] ||
+	   mds_on_old_device $mds_facet; then
+		echo "${ost_label}-osc"
+	elif [[ $ost_label = *OST* ]]; then
+		echo "${ost_label}-osc-${mdt_index}"
+	else
+		echo "${ost_label}-osp-${mdt_index}"
+	fi
 }
 
 get_osc_import_name() {
@@ -5811,6 +5901,7 @@ _wait_import_state () {
     local expected=$1
     local CONN_PROC=$2
     local maxtime=${3:-$(max_recovery_time)}
+    local error_on_failure=${4:-1}
     local CONN_STATE
     local i=0
 
@@ -5823,9 +5914,12 @@ _wait_import_state () {
             # reconnect timeout and test can't see real disconnect
             [ "${CONN_STATE}" == "CONNECTING" ] && return 0
         fi
-        [ $i -ge $maxtime ] && \
-            error "can't put import for $CONN_PROC into ${expected} state after $i sec, have ${CONN_STATE}" && \
+	if [ $i -ge $maxtime ]; then
+	    [ $error_on_failure -ne 0 ] && \
+		error "can't put import for $CONN_PROC into ${expected}" \
+		      "state after $i sec, have ${CONN_STATE}"
             return 1
+	fi
         sleep 1
 	# Add uniq for multi-mount case
 	CONN_STATE=$($LCTL get_param -n $CONN_PROC 2>/dev/null | cut -f2 | uniq)
@@ -5840,10 +5934,11 @@ wait_import_state() {
     local state=$1
     local params=$2
     local maxtime=${3:-$(max_recovery_time)}
+    local error_on_failure=${4:-1}
     local param
 
     for param in ${params//,/ }; do
-        _wait_import_state $state $param $maxtime || return
+	_wait_import_state $state $param $maxtime $error_on_failure || return
     done
 }
 
@@ -5899,8 +5994,8 @@ _wait_osc_import_state() {
 	local facet=$1
 	local ost_facet=$2
 	local expected=$3
-	local ost=$(get_osc_import_name $facet $ost_facet)
-	local param="osc.${ost}.ost_server_uuid"
+	local target=$(get_osc_import_name $facet $ost_facet)
+	local param="osc.${target}.ost_server_uuid"
 	local params=$param
 	local i=0
 
@@ -5924,6 +6019,16 @@ _wait_osc_import_state() {
 			params=$($LCTL list_param $param 2>/dev/null || true)
 		done
 	fi
+
+	if [[ $ost_facet = mds* ]]; then
+		# no OSP connection to itself
+		if [[ $facet = $ost_facet ]]; then
+			return 0
+		fi
+		param="osp.${target}.mdt_server_uuid"
+		params=$param
+	fi
+
 	if ! do_rpc_nodes "$(facet_active_host $facet)" \
 			wait_import_state $expected "$params" $maxtime; then
 		error "import is not in ${expected} state"
@@ -5948,6 +6053,73 @@ wait_osc_import_state() {
 	fi
 }
 
+_wait_mgc_import_state() {
+	local facet=$1
+	local expected=$2
+	local error_on_failure=${3:-1}
+	local param="mgc.*.mgs_server_uuid"
+	local params=$param
+	local i=0
+
+	# 1. wait the deadline of client 1st request (it could be skipped)
+	# 2. wait the deadline of client 2nd request
+	local maxtime=$(( 2 * $(request_timeout $facet)))
+
+	if [[ $facet == client* ]]; then
+		# During setup time, the osc might not be setup, it need wait
+		# until list_param can return valid value. And also if there
+		# are mulitple osc entries we should list all of them before
+		# go to wait.
+		params=$($LCTL list_param $param 2>/dev/null || true)
+		while [ -z "$params" ]; do
+			if [ $i -ge $maxtime ]; then
+				echo "can't get $param in $maxtime secs"
+				return 1
+			fi
+			sleep 1
+			i=$((i + 1))
+			params=$($LCTL list_param $param 2>/dev/null || true)
+		done
+	fi
+	if ! do_rpc_nodes "$(facet_active_host $facet)" \
+			wait_import_state $expected "$params" $maxtime \
+					  $error_on_failure; then
+		if [ $error_on_failure -ne 0 ]; then
+		    error "import is not in ${expected} state"
+		fi
+		return 1
+	fi
+
+	return 0
+}
+
+wait_mgc_import_state() {
+	local facet=$1
+	local expected=$2
+	local error_on_failure=${3:-1}
+	local num
+
+	if [[ $facet = mds ]]; then
+		for num in $(seq $MDSCOUNT); do
+			_wait_mgc_import_state mds$num "$expected" \
+					       $error_on_failure || return
+		done
+	else
+		_wait_mgc_import_state "$facet" "$expected"
+				       $error_on_failure || return
+	fi
+}
+
+wait_dne_interconnect() {
+	local num
+
+	if [ $MDSCOUNT -gt 1 ]; then
+		for num in $(seq $MDSCOUNT); do
+			wait_osc_import_state mds mds$num FULL
+		done
+	fi
+}
+
 get_clientmdc_proc_path() {
     echo "${1}-mdc-*"
 }
@@ -6054,19 +6226,21 @@ create_pool() {
 }
 
 add_pool_to_list () {
-    local fsname=${1%%.*}
-    local poolname=${1##$fsname.}
+	local fsname=${1%%.*}
+	local poolname=${1##$fsname.}
 
-    local listvar=${fsname}_CREATED_POOLS
-    eval export ${listvar}=$(expand_list ${!listvar} $poolname)
+	local listvar=${fsname}_CREATED_POOLS
+	local temp=${listvar}=$(expand_list ${!listvar} $poolname)
+	eval export $temp
 }
 
 remove_pool_from_list () {
-    local fsname=${1%%.*}
-    local poolname=${1##$fsname.}
+	local fsname=${1%%.*}
+	local poolname=${1##$fsname.}
 
-    local listvar=${fsname}_CREATED_POOLS
-    eval export ${listvar}=$(exclude_items_from_list ${!listvar} $poolname)
+	local listvar=${fsname}_CREATED_POOLS
+	local temp=${listvar}=$(exclude_items_from_list ${!listvar} $poolname)
+	eval export $temp
 }
 
 destroy_pool_int() {
@@ -6146,7 +6320,7 @@ gather_logs () {
     suffix="$ts.log"
     echo "Dumping lctl log to ${prefix}.*.${suffix}"
 
-    if [ "$CLIENTONLY" -o "$PDSH" == "no_dsh" ]; then
+    if [ -n "$CLIENTONLY" -o "$PDSH" == "no_dsh" ]; then
         echo "Dumping logs only on local client."
         $LCTL dk > ${prefix}.debug_log.$(hostname -s).${suffix}
         dmesg > ${prefix}.dmesg.$(hostname -s).${suffix}
@@ -6203,6 +6377,31 @@ max_recovery_time() {
 	echo -n $service_time
 }
 
+recovery_time_min() {
+	local connection_switch_min=5
+	local connection_switch_inc=5
+	local connection_switch_max
+	local reconnect_delay_max
+	local initial_connect_timeout
+	local max
+	local timout_20
+
+	#connection_switch_max=min(50, max($connection_switch_min,$TIMEOUT)
+	(($connection_switch_min > $TIMEOUT)) &&
+		max=$connection_switch_min || max=$TIMEOUT
+	(($max < 50)) && connection_switch_max=$max || connection_switch_max=50
+
+	#initial_connect_timeout = max(connection_switch_min, obd_timeout/20)
+	timeout_20=$((TIMEOUT/20))
+	(($connection_switch_min > $timeout_20)) &&
+		initial_connect_timeout=$connection_switch_min ||
+		initial_connect_timeout=$timeout_20
+
+	reconnect_delay_max=$((connection_switch_max + connection_switch_inc + \
+			       initial_connect_timeout))
+	echo $((2 * reconnect_delay_max))
+}
+
 get_clients_mount_count () {
     local clients=${CLIENTS:-`hostname`}
 
@@ -6527,29 +6726,36 @@ check_write_access() {
 }
 
 init_logging() {
-    if [[ -n $YAML_LOG ]]; then
-        return
-    fi
-    local SAVE_UMASK=`umask`
-    umask 0000
-
-    export YAML_LOG=${LOGDIR}/results.yml
-    mkdir -p $LOGDIR
-    init_clients_lists
-
-    if [ ! -f $YAML_LOG ]; then       # If the yaml log already exists then we will just append to it
-      if check_shared_dir $LOGDIR; then
-          touch $LOGDIR/shared
-          echo "Logging to shared log directory: $LOGDIR"
-      else
-          echo "Logging to local directory: $LOGDIR"
-      fi
+	[[ -n $YAML_LOG ]] && return
+	local save_umask=$(umask)
+	umask 0000
 
-      yml_nodes_file $LOGDIR >> $YAML_LOG
-      yml_results_file >> $YAML_LOG
-    fi
+	export YAML_LOG=${LOGDIR}/results.yml
+	mkdir -p $LOGDIR
+	init_clients_lists
+
+	# If the yaml log already exists then we will just append to it
+	if [ ! -f $YAML_LOG ]; then
+		if check_shared_dir $LOGDIR; then
+			touch $LOGDIR/shared
+			echo "Logging to shared log directory: $LOGDIR"
+		else
+			echo "Logging to local directory: $LOGDIR"
+		fi
 
-    umask $SAVE_UMASK
+		yml_nodes_file $LOGDIR >> $YAML_LOG
+		yml_results_file >> $YAML_LOG
+	fi
+
+	umask $save_umask
+
+	# If modules are not yet loaded then older "lctl lustre_build_version"
+	# will fail.  Use lctl build version instead.
+	log "Client: $($LCTL lustre_build_version)"
+	log "MDS: $(do_facet $SINGLEMDS $LCTL lustre_build_version 2>/dev/null||
+		    do_facet $SINGLEMDS $LCTL --version)"
+	log "OSS: $(do_facet ost1 $LCTL lustre_build_version 2> /dev/null ||
+		    do_facet ost1 $LCTL --version)"
 }
 
 log_test() {
@@ -6751,11 +6957,10 @@ get_obd_size() {
 #
 get_page_size() {
 	local facet=$1
-	local size
+	local size=$(getconf PAGE_SIZE 2>/dev/null)
 
-	size=$(do_facet $facet getconf PAGE_SIZE)
-	[[ ${PIPESTATUS[0]} = 0 && -n "$size" ]] || size=4096
-	echo -n $size
+	[ -z "$CLIENTONLY" ] && size=$(do_facet $facet getconf PAGE_SIZE)
+	echo -n ${size:-4096}
 }
 
 #
@@ -6766,20 +6971,20 @@ get_block_count() {
 	local device=$2
 	local count
 
-	count=$(do_facet $facet "$DUMPE2FS -h $device 2>&1" |
+	[ -z "$CLIENTONLY" ] && count=$(do_facet $facet "$DUMPE2FS -h $device 2>&1" |
 		awk '/^Block count:/ {print $3}')
-	echo -n $count
+	echo -n ${count:-0}
 }
 
 # Get the block size of the filesystem.
 get_block_size() {
-    local facet=$1
-    local device=$2
-    local size
+	local facet=$1
+	local device=$2
+	local size
 
-    size=$(do_facet $facet "$DUMPE2FS -h $device 2>&1" |
-           awk '/^Block size:/ {print $3}')
-    echo $size
+	[ -z "$CLIENTONLY" ] && size=$(do_facet $facet "$DUMPE2FS -h $device 2>&1" |
+		awk '/^Block size:/ {print $3}')
+	echo -n ${size:-0}
 }
 
 # Check whether the "large_xattr" feature is enabled or not.
@@ -6829,12 +7034,14 @@ generate_string() {
 
 reformat_external_journal() {
 	local facet=$1
+	local var
 
-	if [ ! -z ${EJOURNAL} ]; then
+	var=${facet}_JRN
+	if [ -n "${!var}" ]; then
 		local rcmd="do_facet $facet"
 
-		echo "reformat external journal on $facet:${EJOURNAL}"
-		${rcmd} mke2fs -O journal_dev ${EJOURNAL} || return 1
+		echo "reformat external journal on $facet:${!var}"
+		${rcmd} mke2fs -O journal_dev ${!var} || return 1
 	fi
 }
 
@@ -6874,14 +7081,10 @@ mds_backup_restore() {
 	echo "backup data"
 	${rcmd} tar zcf $metadata -C $mntpt/ . > /dev/null 2>&1 || return 3
 	# step 6: umount
-	${rcmd} umount -d $mntpt || return 4
-	# step 7: reformat external journal if needed
-	reformat_external_journal $facet || return 5
+	${rcmd} $UMOUNT $mntpt || return 4
 	# step 8: reformat dev
 	echo "reformat new device"
-	add $facet $(mkfs_opts $facet ${devname}) --backfstype ldiskfs \
-		--reformat ${devname} $(mdsvdevname $(facet_number $facet)) \
-		> /dev/null || exit 6
+	format_mdt $(facet_number $facet)
 	# step 9: mount dev
 	${rcmd} mount -t ldiskfs $opts $devname $mntpt || return 7
 	# step 10: restore metadata
@@ -6894,7 +7097,7 @@ mds_backup_restore() {
 	echo "remove recovery logs"
 	${rcmd} rm -fv $mntpt/OBJECTS/* $mntpt/CATALOGS
 	# step 13: umount dev
-	${rcmd} umount -d $mntpt || return 10
+	${rcmd} $UMOUNT $mntpt || return 10
 	# step 14: cleanup tmp backup
 	${rcmd} rm -f $metaea $metadata
 	# step 15: reset device label - it's not virgin on
@@ -6934,7 +7137,7 @@ mds_remove_ois() {
 		done
 	fi
 	# step 4: umount
-	${rcmd} umount -d $mntpt || return 2
+	${rcmd} $UMOUNT $mntpt || return 2
 	# OI files will be recreated when mounted as lustre next time.
 }
 
@@ -7330,3 +7533,12 @@ check_start_ost_idx() {
 		error "OST index of the first stripe on $file is" \
 		      "$start_ost_idx, should be $expected"
 }
+
+killall_process () {
+	local clients=${1:-$(hostname)}
+	local name=$2
+	local signal=$3
+	local rc=0
+
+	do_nodes $clients "killall $signal $name"
+}