X-Git-Url: https://git.whamcloud.com/?p=fs%2Flustre-release.git;a=blobdiff_plain;f=lustre%2Ftests%2Ftest-framework.sh;h=332f74065d10773bf3d28eff1cd8196f9325cea1;hp=4771bfbb130868c3b12bb712e299d7b8c5fab0dd;hb=a98d44e0439ad5b20d0d7f273b29e6ea990312f0;hpb=54fe9796ec837698a27420c8a92d9493c733b6a9

diff --git a/lustre/tests/test-framework.sh b/lustre/tests/test-framework.sh
index 4771bfb..332f740 100644
--- a/lustre/tests/test-framework.sh
+++ b/lustre/tests/test-framework.sh
@@ -5,6 +5,7 @@ trap 'print_summary && touch $TF_FAIL && \
 set -e
 #set -x
 
+export LANG=en_US
 export EJOURNAL=${EJOURNAL:-""}
 export REFORMAT=${REFORMAT:-""}
 export WRITECONF=${WRITECONF:-""}
@@ -15,8 +16,8 @@ export GSS_KRB5=false
 export GSS_PIPEFS=false
 export IDENTITY_UPCALL=default
 export QUOTA_AUTO=1
-export JOBSTATS_AUTO=${JOBSTATS_AUTO:-1}
-export JOBID_VAR=${JOBID_VAR:-"procname_uid"}
+# specify environment variable containing batch job name for server statistics
+export JOBID_VAR=${JOBID_VAR:-"procname_uid"}  # or "existing" or "disable"
 
 # LOAD_LLOOP: LU-409: only load llite_lloop module if kernel < 2.6.32 or
 #             LOAD_LLOOP is true. LOAD_LLOOP is false by default.
@@ -118,10 +119,11 @@ print_summary () {
 }
 
 init_test_env() {
-    export LUSTRE=`absolute_path $LUSTRE`
-    export TESTSUITE=`basename $0 .sh`
-    export TEST_FAILED=false
-    export FAIL_ON_SKIP_ENV=${FAIL_ON_SKIP_ENV:-false}
+	export LUSTRE=$(absolute_path $LUSTRE)
+	export TESTSUITE=$(basename $0 .sh)
+	export TEST_FAILED=false
+	export FAIL_ON_SKIP_ENV=${FAIL_ON_SKIP_ENV:-false}
+	export RPC_MODE=${RPC_MODE:-false}
 
     export MKE2FS=$MKE2FS
     if [ -z "$MKE2FS" ]; then
@@ -177,6 +179,15 @@ init_test_env() {
         fi
     fi
 
+	export RESIZE2FS=$RESIZE2FS
+	if [ -z "$RESIZE2FS" ]; then
+		if which resizefs.ldiskfs >/dev/null 2>&1; then
+			export RESIZE2FS=resizefs.ldiskfs
+		else
+			export RESIZE2FS=resize2fs
+		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
@@ -184,6 +195,7 @@ init_test_env() {
 	export ZFS=${ZFS:-zfs}
 	export ZPOOL=${ZPOOL:-zpool}
 	export ZDB=${ZDB:-zdb}
+	export PARTPROBE=${PARTPROBE:-partprobe}
 
     #[ -d /r ] && export ROOT=${ROOT:-/r}
     export TMP=${TMP:-$ROOT/tmp}
@@ -282,8 +294,6 @@ init_test_env() {
             IDENTITY_UPCALL=false
             ;;
     esac
-    USE_OFD=${USE_OFD:-yes}
-    [ "$USE_OFD" = "yes" ] && LOAD_MODULES_REMOTE=true
 
     export LOAD_MODULES_REMOTE=${LOAD_MODULES_REMOTE:-false}
 
@@ -291,13 +301,15 @@ init_test_env() {
     export RLUSTRE=${RLUSTRE:-$LUSTRE}
     export RPWD=${RPWD:-$PWD}
     export I_MOUNTED=${I_MOUNTED:-"no"}
-    if [ ! -f /lib/modules/$(uname -r)/kernel/fs/lustre/mds.ko -a \
-        ! -f /lib/modules/$(uname -r)/updates/kernel/fs/lustre/mds.ko -a \
-        ! -f `dirname $0`/../mds/mds.ko ]; then
-        export CLIENTMODSONLY=yes
-    fi
+	if [ ! -f /lib/modules/$(uname -r)/kernel/fs/lustre/mdt.ko -a \
+	     ! -f /lib/modules/$(uname -r)/updates/kernel/fs/lustre/mdt.ko -a \
+	     ! -f /lib/modules/$(uname -r)/extra/kernel/fs/lustre/mdt.ko -a \
+	     ! -f $LUSTRE/mdt/mdt.ko ]; then
+	    export CLIENTMODSONLY=yes
+	fi
 
-    export SHUTDOWN_ATTEMPTS=${SHUTDOWN_ATTEMPTS:-3}
+	export SHUTDOWN_ATTEMPTS=${SHUTDOWN_ATTEMPTS:-3}
+	export OSD_TRACK_DECLARES_LBUG=${OSD_TRACK_DECLARES_LBUG:-"yes"}
 
     # command line
 
@@ -314,10 +326,12 @@ init_test_env() {
     shift $((OPTIND - 1))
     ONLY=${ONLY:-$*}
 
-    # print the durations of each test if "true"
-    DDETAILS=${DDETAILS:-false}
-    [ "$TESTSUITELOG" ] && rm -f $TESTSUITELOG || true
-    rm -f $TMP/*active
+	# print the durations of each test if "true"
+	DDETAILS=${DDETAILS:-false}
+	[ "$TESTSUITELOG" ] && rm -f $TESTSUITELOG || true
+	if ! $RPC_MODE; then
+		rm -f $TMP/*active
+	fi
 }
 
 check_cpt_number() {
@@ -338,7 +352,7 @@ version_code() {
     # split arguments like "1.8.6-wc3" into "1", "8", "6", "wc3"
     eval set -- $(tr "[:punct:]" " " <<< $*)
 
-    echo -n $((($1 << 16) | ($2 << 8) | $3))
+    echo -n "$((($1 << 16) | ($2 << 8) | $3))"
 }
 
 export LINUX_VERSION=$(uname -r | sed -e "s/[-.]/ /3" -e "s/ .*//")
@@ -373,20 +387,24 @@ load_module() {
         optvar="MODOPTS_$(basename $module | tr a-z A-Z)"
         eval set -- \$$optvar
         if [ $# -eq 0 -a -n "$MODPROBECONF" ]; then
-            # Nothing in $MODOPTS_<MODULE>; try modprobe.conf
-            set -- $(grep -P "^options\\s+${BASE}" $MODPROBECONF)
-            # Get rid of "options $module"
-            (($# > 0)) && shift 2
-
-            # Ensure we have accept=all for lnet
-            if [ $(basename $module) = lnet ]; then
-                # OK, this is a bit wordy...
-                local arg accept_all_present=false
-                for arg in "$@"; do
-                    [ "$arg" = accept=all ] && accept_all_present=true
-                done
-                $accept_all_present || set -- "$@" accept=all
-            fi
+		# Nothing in $MODOPTS_<MODULE>; try modprobe.conf
+		local opt
+		opt=$(awk -v var="^options $BASE" '$0 ~ var \
+			{gsub("'"options $BASE"'",""); print}' $MODPROBECONF)
+		set -- $(echo -n $opt)
+
+		# Ensure we have accept=all for lnet
+		if [ $(basename $module) = lnet ]; then
+			# OK, this is a bit wordy...
+			local arg accept_all_present=false
+
+			for arg in "$@"; do
+				[ "$arg" = accept=all ] && \
+					accept_all_present=true
+			done
+			$accept_all_present || set -- "$@" accept=all
+		fi
+		export $optvar="$*"
         fi
     fi
 
@@ -425,8 +443,6 @@ llite_lloop_enabled() {
 }
 
 load_modules_local() {
-	[ $(facet_fstype ost1) == "zfs" ] && export USE_OFD=yes
-
 	if [ -n "$MODPROBE" ]; then
 		# use modprobe
 		echo "Using modprobe to load modules"
@@ -470,7 +486,6 @@ load_modules_local() {
     load_module ../lnet/lnet/lnet
     LNETLND=${LNETLND:-"socklnd/ksocklnd"}
     load_module ../lnet/klnds/$LNETLND
-    load_module lvfs/lvfs
     load_module obdclass/obdclass
     load_module ptlrpc/ptlrpc
     load_module ptlrpc/gss/ptlrpc_gss
@@ -487,25 +502,21 @@ load_modules_local() {
         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 $SYMLIST || { modprobe jbd2 2>/dev/null || true; }
+		load_module lfsck/lfsck
 		[ "$LQUOTA" != "no" ] && load_module quota/lquota $LQUOTAOPTS
 		if [[ $(node_fstypes $HOSTNAME) == *zfs* ]]; then
 			modprobe zfs
 			load_module osd-zfs/osd_zfs
 		fi
-		load_module mgs/mgs
-		load_module mds/mds
-		load_module mdd/mdd
 		if [[ $(node_fstypes $HOSTNAME) == *ldiskfs* ]]; then
-			#
-			# This block shall be moved up beside osd-zfs as soon
-			# as osd-ldiskfs stops using mdd symbols.
-			#
 			grep -q exportfs_decode_fh $SYMLIST ||
 				{ modprobe exportfs 2> /dev/null || true; }
-			load_module ../ldiskfs/ldiskfs/ldiskfs
-			load_module lvfs/fsfilt_ldiskfs
+			load_module ../ldiskfs/ldiskfs
 			load_module osd-ldiskfs/osd_ldiskfs
 		fi
+		load_module nodemap/nodemap
+		load_module mgs/mgs
+		load_module mdd/mdd
 		load_module mdt/mdt
 		load_module ost/ost
 		load_module lod/lod
@@ -583,6 +594,18 @@ unload_modules() {
     return 0
 }
 
+fs_log_size() {
+	local facet=${1:-$SINGLEMDS}
+	local fstype=$(facet_fstype $facet)
+	local size=0
+	case $fstype in
+		ldiskfs) size=50;; # largest seen is 44, leave some headroom
+		zfs)     size=400;; # largest seen is 384
+	esac
+
+	echo -n $size
+}
+
 check_gss_daemon_nodes() {
     local list=$1
     dname=$2
@@ -695,10 +718,17 @@ cleanup_gss() {
     fi
 }
 
+facet_svc() {
+	local facet=$1
+	local var=${facet}_svc
+
+	echo -n ${!var}
+}
+
 facet_type() {
 	local facet=$1
 
-	echo -n $facet | sed -e 's/^fs[0-9]\+//' -e 's/[0-9]\+//' |
+	echo -n $facet | sed -e 's/^fs[0-9]\+//' -e 's/[0-9_]\+//' |
 		tr '[:lower:]' '[:upper:]'
 }
 
@@ -733,6 +763,11 @@ facet_fstype() {
 		return
 	fi
 
+	if [[ $facet == mgs ]] && combined_mgs_mds; then
+		facet_fstype mds1
+		return
+	fi
+
 	return 1
 }
 
@@ -789,7 +824,184 @@ ostdevlabel() {
 }
 
 #
-# This and set_obdfilter_param() shall be used to access OSD parameters
+# Get the device of a facet.
+#
+facet_device() {
+	local facet=$1
+	local device
+
+	case $facet in
+		mgs) device=$(mgsdevname) ;;
+		mds*) device=$(mdsdevname $(facet_number $facet)) ;;
+		ost*) device=$(ostdevname $(facet_number $facet)) ;;
+		fs2mds) device=$(mdsdevname 1_2) ;;
+		fs2ost) device=$(ostdevname 1_2) ;;
+		fs3ost) device=$(ostdevname 2_2) ;;
+		*) ;;
+	esac
+
+	echo -n $device
+}
+
+#
+# Get the virtual device of a facet.
+#
+facet_vdevice() {
+	local facet=$1
+	local device
+
+	case $facet in
+		mgs) device=$(mgsvdevname) ;;
+		mds*) device=$(mdsvdevname $(facet_number $facet)) ;;
+		ost*) device=$(ostvdevname $(facet_number $facet)) ;;
+		fs2mds) device=$(mdsvdevname 1_2) ;;
+		fs2ost) device=$(ostvdevname 1_2) ;;
+		fs3ost) device=$(ostvdevname 2_2) ;;
+		*) ;;
+	esac
+
+	echo -n $device
+}
+
+#
+# Re-read the partition table on failover partner host.
+# After a ZFS storage pool is created on a shared device, the partition table
+# on the device may change. However, the operating system on the failover
+# host may not notice the change automatically. Without the up-to-date partition
+# block devices, 'zpool import ..' cannot find the labels, whose positions are
+# relative to partition rather than disk beginnings.
+#
+# This function performs partprobe on the failover host to make it re-read the
+# partition table.
+#
+refresh_partition_table() {
+	local facet=$1
+	local device=$2
+	local host
+
+	host=$(facet_passive_host $facet)
+	if [[ -n "$host" ]]; then
+		do_node $host "$PARTPROBE $device"
+	fi
+}
+
+#
+# Get ZFS storage pool name.
+#
+zpool_name() {
+	local facet=$1
+	local device
+	local poolname
+
+	device=$(facet_device $facet)
+	# poolname is string before "/"
+	poolname="${device%%/*}"
+
+	echo -n $poolname
+}
+
+#
+# Create ZFS storage pool.
+#
+create_zpool() {
+	local facet=$1
+	local poolname=$2
+	local vdev=$3
+	shift 3
+	local opts=${@:-"-o cachefile=none"}
+
+	do_facet $facet "$ZPOOL list -H $poolname >/dev/null 2>&1 ||
+		$ZPOOL create -f $opts $poolname $vdev"
+}
+
+#
+# Create ZFS file system.
+#
+create_zfs() {
+	local facet=$1
+	local dataset=$2
+	shift 2
+	local opts=${@:-"-o mountpoint=legacy"}
+
+	do_facet $facet "$ZFS list -H $dataset >/dev/null 2>&1 ||
+		$ZFS create $opts $dataset"
+}
+
+#
+# Export ZFS storage pool.
+# Before exporting the pool, all datasets within the pool should be unmounted.
+#
+export_zpool() {
+	local facet=$1
+	shift
+	local opts="$@"
+	local poolname
+
+	poolname=$(zpool_name $facet)
+
+	if [[ -n "$poolname" ]]; then
+		do_facet $facet "! $ZPOOL list -H $poolname >/dev/null 2>&1 ||
+			grep -q ^$poolname/ /proc/mounts ||
+			$ZPOOL export $opts $poolname"
+	fi
+}
+
+#
+# Destroy ZFS storage pool.
+# Destroy the given pool and free up any devices for other use. This command
+# tries to unmount any active datasets before destroying the pool.
+# -f    Force any active datasets contained within the pool to be unmounted.
+#
+destroy_zpool() {
+	local facet=$1
+	local poolname=${2:-$(zpool_name $facet)}
+
+	if [[ -n "$poolname" ]]; then
+		do_facet $facet "! $ZPOOL list -H $poolname >/dev/null 2>&1 ||
+			$ZPOOL destroy -f $poolname"
+	fi
+}
+
+#
+# Import ZFS storage pool.
+# Force importing, even if the pool appears to be potentially active.
+#
+import_zpool() {
+	local facet=$1
+	shift
+	local opts=${@:-"-o cachefile=none"}
+	local poolname
+
+	poolname=$(zpool_name $facet)
+
+	if [[ -n "$poolname" ]]; then
+		opts+=" -d $(dirname $(facet_vdevice $facet))"
+		do_facet $facet "$ZPOOL list -H $poolname >/dev/null 2>&1 ||
+			$ZPOOL import -f $opts $poolname"
+	fi
+}
+
+#
+# Set the "cachefile=none" property on ZFS storage pool so that the pool
+# is not automatically imported on system startup.
+#
+# In a failover environment, this will provide resource level fencing which
+# will ensure that the same ZFS storage pool will not be imported concurrently
+# on different nodes.
+#
+disable_zpool_cache() {
+	local facet=$1
+	local poolname
+
+	poolname=$(zpool_name $facet)
+
+	if [[ -n "$poolname" ]]; then
+		do_facet $facet "$ZPOOL set cachefile=none $poolname"
+	fi
+}
+
+#
+# This and set_osd_param() shall be used to access OSD parameters
 # once existed under "obdfilter":
 #
 #   mntdev
@@ -797,7 +1009,7 @@ ostdevlabel() {
 #   read_cache_enable
 #   writethrough_cache_enable
 #
-get_obdfilter_param() {
+get_osd_param() {
 	local nodes=$1
 	local device=${2:-$FSNAME-OST*}
 	local name=$3
@@ -806,7 +1018,7 @@ get_obdfilter_param() {
 		osd-*.$device.$name 2>&1" | grep -v 'Found no match'
 }
 
-set_obdfilter_param() {
+set_osd_param() {
 	local nodes=$1
 	local device=${2:-$FSNAME-OST*}
 	local name=$3
@@ -916,6 +1128,11 @@ mount_facet() {
 		opts=$(csa_add "$opts" -o loop)
 	fi
 
+	if [[ $(facet_fstype $facet) == zfs ]]; then
+		# import ZFS storage pool
+		import_zpool $facet || return ${PIPESTATUS[0]}
+	fi
+
 	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
@@ -962,39 +1179,6 @@ start() {
     return $RC
 }
 
-#
-# When a ZFS OSD is made read-only by replay_barrier(), its pool is "freezed".
-# Because stopping corresponding target may not clear this in-memory state, we
-# need to zap the pool from memory by exporting and reimporting the pool.
-#
-# Although the uberblocks are not updated when a pool is freezed, transactions
-# are still written to the disks.  Modified blocks may be cached in memory when
-# tests try reading them back.  The export-and-reimport process also evicts any
-# cached pool data from memory to provide the correct "data loss" semantics.
-#
-refresh_disk() {
-	local facet=$1
-	local fstype=$(facet_fstype $facet)
-	local _dev
-	local dev
-	local poolname
-
-	if [ "${fstype}" == "zfs" ]; then
-		_dev=$(facet_active $facet)_dev
-		dev=${!_dev} # expand _dev to its value, e.g. ${mds1_dev}
-		poolname="${dev%%/*}" # poolname is string before "/"
-
-		if [ "${poolname}" == "" ]; then
-			echo "invalid dataset name: $dev"
-			return
-		fi
-		do_facet $facet "cp /etc/zfs/zpool.cache /tmp/zpool.cache.back"
-		do_facet $facet "$ZPOOL export ${poolname}"
-		do_facet $facet "$ZPOOL import -f -c /tmp/zpool.cache.back \
-		                 ${poolname}"
-	fi
-}
-
 stop() {
     local running
     local facet=$1
@@ -1009,9 +1193,14 @@ stop() {
         do_facet ${facet} umount -d $@ $mntpt
     fi
 
-    # umount should block, but we should wait for unrelated obd's
-    # like the MGS or MGC to also stop.
-    wait_exit_ST ${facet}
+	# umount should block, but we should wait for unrelated obd's
+	# like the MGS or MGC to also stop.
+	wait_exit_ST ${facet} || return ${PIPESTATUS[0]}
+
+	if [[ $(facet_fstype $facet) == zfs ]]; then
+		# export ZFS storage pool
+		export_zpool $facet
+	fi
 }
 
 # save quota version (both administrative and operational quotas)
@@ -1044,13 +1233,14 @@ quota_save_version() {
 # 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.
-quota_type () {
-    local fsname=${1:-$FSNAME}
-    local rc=0
-    do_facet mgs lctl get_param mdd.${fsname}-MDT*.quota_type || rc=$?
-    do_nodes $(comma_list $(osts_nodes)) \
-        lctl get_param obdfilter.${fsname}-OST*.quota_type || rc=$?
-    return $rc 
+quota_type() {
+	local fsname=${1:-$FSNAME}
+	local rc=0
+	do_facet $SINGLEMDS lctl get_param mdd.${fsname}-MDT*.quota_type ||
+		rc=$?
+	do_nodes $(comma_list $(osts_nodes)) \
+		lctl get_param obdfilter.${fsname}-OST*.quota_type || rc=$?
+	return $rc 
 }
 
 # XXX This function is kept for interoperability with old server (< 2.3.50),
@@ -1092,11 +1282,11 @@ setup_quota_old(){
     local quota_usrs=$QUOTA_USERS
 
     # get_filesystem_size
-    local disksz=$(lfs df $mntpt | grep "filesystem summary:"  | awk '{print $3}')
+    local disksz=$(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 "filesystem summary:"  | awk '{print $3}')
+    local Inodes=$(lfs_df -i $mntpt | grep "summary"  | awk '{print $2}')
     local i_soft=$Inodes
     local i_hard=$((i_soft + i_soft / 20))
 
@@ -1121,7 +1311,7 @@ setup_quota_old(){
 mdt_quota_type() {
 	local varsvc=${SINGLEMDS}_svc
 	do_facet $SINGLEMDS $LCTL get_param -n \
-		osd-$FSTYPE.${!varsvc}.quota_slave.enabled
+		osd-$(facet_fstype $SINGLEMDS).${!varsvc}.quota_slave.enabled
 }
 
 # get ost quota type
@@ -1129,7 +1319,7 @@ ost_quota_type() {
 	# All OSTs should have same quota type
 	local varsvc=ost1_svc
 	do_facet ost1 $LCTL get_param -n \
-		osd-$FSTYPE.${!varsvc}.quota_slave.enabled
+		osd-$(facet_fstype ost1).${!varsvc}.quota_slave.enabled
 }
 
 # restore old quota type settings
@@ -1149,6 +1339,30 @@ restore_quota() {
 	fi
 }
 
+# Handle the case when there is a space in the lfs df
+# "filesystem summary" line the same as when there is no space.
+# This will allow fixing the "lfs df" summary line in the future.
+lfs_df() {
+	$LFS df $* | sed -e 's/filesystem /filesystem_/'
+}
+
+# Get free inodes on the MDT specified by mdt index, free indoes on
+# the whole filesystem will be returned when index == -1.
+mdt_free_inodes() {
+	local index=$1
+	local free_inodes
+	local mdt_uuid
+
+	if [ $index -eq -1 ]; then
+		mdt_uuid="summary"
+	else
+		mdt_uuid=$(mdtuuid_from_index $index)
+	fi
+
+	free_inodes=$(lfs_df -i $MOUNT | grep $mdt_uuid | awk '{print $4}')
+	echo $free_inodes
+}
+
 setup_quota(){
 	if [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.3.50) ]; then
 		setup_quota_old $1
@@ -1175,13 +1389,11 @@ setup_quota(){
 	local quota_usrs=$QUOTA_USERS
 
 	# get_filesystem_size
-	local disksz=$(lfs df $mntpt | grep "filesystem summary:" |
-		     awk '{print $3}')
+	local disksz=$(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 "filesystem summary:" |
-		     awk '{print $3}')
+	local inodes=$(lfs_df -i $mntpt | grep "summary" | awk '{print $2}')
 	local i_soft=$inodes
 	local i_hard=$((i_soft + i_soft / 20))
 
@@ -1463,7 +1675,6 @@ reboot_facet() {
 	if [ "$FAILURE_MODE" = HARD ]; then
 		reboot_node $(facet_active_host $facet)
 	else
-		refresh_disk ${facet}
 		sleep 10
 	fi
 }
@@ -1526,6 +1737,8 @@ TESTLOG_PREFIX=$TESTLOG_PREFIX \
 TESTNAME=$TESTNAME \
 DBENCH_LIB=$DBENCH_LIB \
 DBENCH_SRC=$DBENCH_SRC \
+CLIENT_COUNT=$((CLIENTCOUNT - 1)) \
+LFS=$LFS \
 run_${load}.sh" &
     local ppid=$!
     log "Started client load: ${load} on $client"
@@ -1734,30 +1947,44 @@ cleanup_check() {
 }
 
 wait_update () {
-    local node=$1
-    local TEST=$2
-    local FINAL=$3
-    local MAX=${4:-90}
-
-        local RESULT
-        local WAIT=0
-        local sleep=1
-        local print=10
-        while [ true ]; do
-            RESULT=$(do_node $node "$TEST")
-            if [ "$RESULT" == "$FINAL" ]; then
-                [ -z "$RESULT" -o $WAIT -le $sleep ] ||
-                    echo "Updated after ${WAIT}s: wanted '$FINAL' got '$RESULT'"
-                return 0
-            fi
-            [ $WAIT -ge $MAX ] && break
-            [ $((WAIT % print)) -eq 0 ] &&
-                echo "Waiting $((MAX - WAIT)) secs for update"
-            WAIT=$((WAIT + sleep))
-            sleep $sleep
-        done
-        echo "Update not seen after ${MAX}s: wanted '$FINAL' got '$RESULT'"
-        return 3
+	local verbose=false
+	if [[ "$1" == "--verbose" ]]; then
+		shift
+		verbose=true
+	fi
+
+	local node=$1
+	local TEST=$2
+	local FINAL=$3
+	local MAX=${4:-90}
+	local RESULT
+	local PREV_RESULT
+	local WAIT=0
+	local sleep=1
+	local print=10
+
+	PREV_RESULT=$(do_node $node "$TEST")
+	while [ true ]; do
+		RESULT=$(do_node $node "$TEST")
+		if [[ "$RESULT" == "$FINAL" ]]; then
+			[[ -z "$RESULT" || $WAIT -le $sleep ]] ||
+				echo "Updated after ${WAIT}s: wanted '$FINAL'"\
+				     "got '$RESULT'"
+			return 0
+		fi
+		if [[ $verbose && "$RESULT" != "$PREV_RESULT" ]]; then
+			echo "Changed after ${WAIT}s: from '$PREV_RESULT'"\
+			     "to '$RESULT'"
+			PREV_RESULT=$RESULT
+		fi
+		[[ $WAIT -ge $MAX ]] && break
+		[[ $((WAIT % print)) -eq 0 ]] &&
+			echo "Waiting $((MAX - WAIT)) secs for update"
+		WAIT=$((WAIT + sleep))
+		sleep $sleep
+	done
+	echo "Update not seen after ${MAX}s: wanted '$FINAL' got '$RESULT'"
+	return 3
 }
 
 wait_update_facet() {
@@ -1767,7 +1994,10 @@ wait_update_facet() {
 }
 
 sync_all_data() {
-	do_node $(osts_nodes) "lctl set_param -n osd*.*OS*.force_sync 1" 2>&1 |
+	do_nodes $(comma_list $(mdts_nodes)) \
+	    "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 |
 		grep -v 'Found no match'
 }
 
@@ -1883,30 +2113,50 @@ wait_recovery_complete () {
 }
 
 wait_mds_ost_sync () {
-    # just because recovery is done doesn't mean we've finished
-    # orphan cleanup. Wait for llogs to get synchronized.
-    echo "Waiting for orphan cleanup..."
-    # MAX value includes time needed for MDS-OST reconnection
-    local MAX=$(( TIMEOUT * 2 ))
-    local WAIT=0
-    while [ $WAIT -lt $MAX ]; do
-        local -a sync=($(do_nodes $(comma_list $(osts_nodes)) \
-            "$LCTL get_param -n obdfilter.*.mds_sync"))
-        local con=1
-        local i
-        for ((i=0; i<${#sync[@]}; i++)); do
-            [ ${sync[$i]} -eq 0 ] && continue
-            # there is a not finished MDS-OST synchronization
-            con=0
-            break;
-        done
-        sleep 2 # increase waiting time and cover statfs cache
-        [ ${con} -eq 1 ] && return 0
-        echo "Waiting $WAIT secs for $facet mds-ost sync done."
-        WAIT=$((WAIT + 2))
-    done
-    echo "$facet recovery not done in $MAX sec. $STATUS"
-    return 1
+	# just because recovery is done doesn't mean we've finished
+	# orphan cleanup. Wait for llogs to get synchronized.
+	echo "Waiting for orphan cleanup..."
+	# MAX value includes time needed for MDS-OST reconnection
+	local MAX=$(( TIMEOUT * 2 ))
+	local WAIT_TIMEOUT=${1:-$MAX}
+	local WAIT=0
+	local new_wait=true
+	local list=$(comma_list $(mdts_nodes))
+	local cmd="$LCTL get_param -n osp.*osc*.old_sync_processed"
+	if ! do_facet $SINGLEMDS \
+		"$LCTL list_param osp.*osc*.old_sync_processed 2> /dev/null"
+	then
+		# old way, use mds_sync
+		new_wait=false
+		list=$(comma_list $(osts_nodes))
+		cmd="$LCTL get_param -n obdfilter.*.mds_sync"
+	fi
+
+	echo "wait $WAIT_TIMEOUT secs maximumly for $list mds-ost sync done."
+	while [ $WAIT -lt $WAIT_TIMEOUT ]; do
+		local -a sync=($(do_nodes $list "$cmd"))
+		local con=1
+		local i
+		for ((i=0; i<${#sync[@]}; i++)); do
+			if $new_wait; then
+				[ ${sync[$i]} -eq 1 ] && continue
+			else
+				[ ${sync[$i]} -eq 0 ] && continue
+			fi
+			# there is a not finished MDS-OST synchronization
+			con=0
+			break;
+		done
+		sleep 2 # increase waiting time and cover statfs cache
+		[ ${con} -eq 1 ] && return 0
+		echo "Waiting $WAIT secs for $list $i mds-ost sync done."
+		WAIT=$((WAIT + 2))
+	done
+
+	# show which nodes are not finished.
+	do_nodes $list "$cmd"
+	echo "$facet recovery node $i not done in $WAIT_TIMEOUT sec. $STATUS"
+	return 1
 }
 
 wait_destroy_complete () {
@@ -2045,32 +2295,58 @@ affected_facets () {
 }
 
 facet_failover() {
-    local facet=$1
-    local sleep_time=$2
-    local host=$(facet_active_host $facet)
-
-    echo "Failing $facet on node $host"
-
-    local affected=$(affected_facets $facet)
+	local facets=$1
+	local sleep_time=$2
+	local -a affecteds
+	local facet
+	local total=0
+	local index=0
+	local skip
+
+	#Because it will only get up facets, we need get affected
+	#facets before shutdown
+	#For HARD Failure mode, it needs make sure facets on the same
+	#HOST will only be shutdown and reboot once
+	for facet in ${facets//,/ }; do
+		local affected_facet
+		skip=0
+		#check whether facet has been included in other affected facets
+		for ((index=0; index<$total; index++)); do
+			[[ *,$facet,* == ,${affecteds[index]}, ]] && skip=1
+		done
 
-    shutdown_facet $facet
+		if [ $skip -eq 0 ]; then
+			affecteds[$total]=$(affected_facets $facet)
+			total=$((total+1))
+		fi
+	done
 
-    echo affected facets: $affected
+	for ((index=0; index<$total; index++)); do
+		facet=$(echo ${affecteds[index]} | tr -s " " | cut -d"," -f 1)
+		local host=$(facet_active_host $facet)
+		echo "Failing ${affecteds[index]} on $host"
+		shutdown_facet $facet
+	done
 
-    [ -n "$sleep_time" ] && sleep $sleep_time
+	for ((index=0; index<$total; index++)); do
+		facet=$(echo ${affecteds[index]} | tr -s " " | cut -d"," -f 1)
+		echo reboot facets: ${affecteds[index]}
 
-    reboot_facet $facet
+		reboot_facet $facet
 
-    change_active $affected
+		change_active ${affecteds[index]}
 
-    wait_for_facet $affected
-    # start mgs first if it is affected
-    if ! combined_mgs_mds && list_member $affected mgs; then
-        mount_facet mgs || error "Restart of mgs failed"
-    fi
-    # FIXME; has to be changed to mount all facets concurrently
-    affected=$(exclude_items_from_list $affected mgs)
-    mount_facets $affected
+		wait_for_facet ${affecteds[index]}
+		# start mgs first if it is affected
+		if ! combined_mgs_mds &&
+			list_member ${affecteds[index]} mgs; then
+			mount_facet mgs || error "Restart of mgs failed"
+		fi
+		# FIXME; has to be changed to mount all facets concurrently
+		affected=$(exclude_items_from_list ${affecteds[index]} mgs)
+		echo mount facets: ${affecteds[index]}
+		mount_facets ${affecteds[index]}
+	done
 }
 
 obd_name() {
@@ -2078,57 +2354,94 @@ obd_name() {
 }
 
 replay_barrier() {
-    local facet=$1
-    do_facet $facet "sync; sync; sync"
-    df $MOUNT
-
-    # make sure there will be no seq change
-    local clients=${CLIENTS:-$HOSTNAME}
-    local f=fsa-\\\$\(hostname\)
-    do_nodes $clients "mcreate $MOUNT/$f; rm $MOUNT/$f"
-    do_nodes $clients "if [ -d $MOUNT2 ]; then mcreate $MOUNT2/$f; rm $MOUNT2/$f; fi"
+	local facet=$1
+	do_facet $facet "sync; sync; sync"
+	df $MOUNT
 
-    local svc=${facet}_svc
-    do_facet $facet $LCTL --device %${!svc} notransno
-    do_facet $facet $LCTL --device %${!svc} readonly
-    do_facet $facet $LCTL mark "$facet REPLAY BARRIER on ${!svc}"
-    $LCTL mark "local REPLAY BARRIER on ${!svc}"
+	# make sure there will be no seq change
+	local clients=${CLIENTS:-$HOSTNAME}
+	local f=fsa-\\\$\(hostname\)
+	do_nodes $clients "mcreate $MOUNT/$f; rm $MOUNT/$f"
+	do_nodes $clients "if [ -d $MOUNT2 ]; then mcreate $MOUNT2/$f; rm $MOUNT2/$f; fi"
+
+	local svc=${facet}_svc
+	do_facet $facet $LCTL --device ${!svc} notransno
+	#
+	# If a ZFS OSD is made read-only here, its pool is "freezed". This
+	# in-memory state has to be cleared by either rebooting the host or
+	# exporting and reimporting the pool.
+	#
+	# Although the uberblocks are not updated when a pool is freezed,
+	# transactions are still written to the disks. Modified blocks may be
+	# cached in memory when tests try reading them back. The
+	# export-and-reimport process also evicts any cached pool data from
+	# memory to provide the correct "data loss" semantics.
+	#
+	# In the test framework, the exporting and importing operations are
+	# handled by stop() and mount_facet() separately, which are used
+	# inside fail() and fail_abort().
+	#
+	do_facet $facet $LCTL --device ${!svc} readonly
+	do_facet $facet $LCTL mark "$facet REPLAY BARRIER on ${!svc}"
+	$LCTL mark "local REPLAY BARRIER on ${!svc}"
 }
 
 replay_barrier_nodf() {
-    local facet=$1    echo running=${running}
-    do_facet $facet "sync; sync; sync"
-    local svc=${facet}_svc
-    echo Replay barrier on ${!svc}
-    do_facet $facet $LCTL --device %${!svc} notransno
-    do_facet $facet $LCTL --device %${!svc} readonly
-    do_facet $facet $LCTL mark "$facet REPLAY BARRIER on ${!svc}"
-    $LCTL mark "local REPLAY BARRIER on ${!svc}"
+	local facet=$1    echo running=${running}
+	do_facet $facet "sync; sync; sync"
+	local svc=${facet}_svc
+	echo Replay barrier on ${!svc}
+	do_facet $facet $LCTL --device ${!svc} notransno
+	do_facet $facet $LCTL --device ${!svc} readonly
+	do_facet $facet $LCTL mark "$facet REPLAY BARRIER on ${!svc}"
+	$LCTL mark "local REPLAY BARRIER on ${!svc}"
 }
 
 replay_barrier_nosync() {
-    local facet=$1    echo running=${running}
-    local svc=${facet}_svc
-    echo Replay barrier on ${!svc}
-    do_facet $facet $LCTL --device %${!svc} notransno
-    do_facet $facet $LCTL --device %${!svc} readonly
-    do_facet $facet $LCTL mark "$facet REPLAY BARRIER on ${!svc}"
-    $LCTL mark "local REPLAY BARRIER on ${!svc}"
+	local facet=$1    echo running=${running}
+	local svc=${facet}_svc
+	echo Replay barrier on ${!svc}
+	do_facet $facet $LCTL --device ${!svc} notransno
+	do_facet $facet $LCTL --device ${!svc} readonly
+	do_facet $facet $LCTL mark "$facet REPLAY BARRIER on ${!svc}"
+	$LCTL mark "local REPLAY BARRIER on ${!svc}"
+}
+
+#
+# Get Lustre client uuid for a given Lustre mount point.
+#
+get_client_uuid() {
+	local mntpnt=${1:-$MOUNT}
+
+	local name=$($LFS getname $mntpnt | cut -d' ' -f1)
+	local uuid=$($LCTL get_param -n llite.$name.uuid)
+
+	echo -n $uuid
 }
 
 mds_evict_client() {
-    UUID=`lctl get_param -n mdc.${mds1_svc}-mdc-*.uuid`
-    do_facet mds1 "lctl set_param -n mdt.${mds1_svc}.evict_client $UUID"
+	local mntpnt=${1:-$MOUNT}
+	local uuid=$(get_client_uuid $mntpnt)
+
+	do_facet $SINGLEMDS \
+		"$LCTL set_param -n mdt.${mds1_svc}.evict_client $uuid"
 }
 
 ost_evict_client() {
-    UUID=`lctl get_param -n devices| grep ${ost1_svc}-osc- | egrep -v 'MDT' | awk '{print $5}'`
-    do_facet ost1 "lctl set_param -n obdfilter.${ost1_svc}.evict_client $UUID"
+	local mntpnt=${1:-$MOUNT}
+	local uuid=$(get_client_uuid $mntpnt)
+
+	do_facet ost1 \
+		"$LCTL set_param -n obdfilter.${ost1_svc}.evict_client $uuid"
 }
 
 fail() {
-    facet_failover $* || error "failover: $?"
-    clients_up || error "post-failover df: $?"
+	local facets=$1
+	local clients=${CLIENTS:-$HOSTNAME}
+
+	facet_failover $* || error "failover: $?"
+	wait_clients_import_state "$clients" "$facets" FULL
+	clients_up || error "post-failover df: $?"
 }
 
 fail_nodf() {
@@ -2139,7 +2452,6 @@ fail_nodf() {
 fail_abort() {
 	local facet=$1
 	stop $facet
-	refresh_disk ${facet}
 	change_active $facet
 	wait_for_facet $facet
 	mount_facet $facet -o abort_recovery
@@ -2165,42 +2477,43 @@ host_nids_address() {
 }
 
 h2name_or_ip() {
-    if [ "$1" = "client" -o "$1" = "'*'" ]; then echo \'*\'; else
-        echo $1"@$2"
-    fi
+	if [ "$1" = "'*'" ]; then echo \'*\'; else
+		echo $1"@$2"
+	fi
 }
 
 h2ptl() {
-   if [ "$1" = "client" -o "$1" = "'*'" ]; then echo \'*\'; else
-       ID=`xtprocadmin -n $1 2>/dev/null | egrep -v 'NID' | awk '{print $1}'`
-       if [ -z "$ID" ]; then
-           echo "Could not get a ptl id for $1..."
-           exit 1
-       fi
-       echo $ID"@ptl"
-   fi
+	if [ "$1" = "'*'" ]; then echo \'*\'; else
+		ID=`xtprocadmin -n $1 2>/dev/null | egrep -v 'NID' | \
+							awk '{print $1}'`
+		if [ -z "$ID" ]; then
+			echo "Could not get a ptl id for $1..."
+			exit 1
+		fi
+		echo $ID"@ptl"
+	fi
 }
 declare -fx h2ptl
 
 h2tcp() {
-    h2name_or_ip "$1" "tcp"
+	h2name_or_ip "$1" "tcp"
 }
 declare -fx h2tcp
 
 h2elan() {
-    if [ "$1" = "client" -o "$1" = "'*'" ]; then echo \'*\'; else
-        if type __h2elan >/dev/null 2>&1; then
-            ID=$(__h2elan $1)
-        else
-            ID=`echo $1 | sed 's/[^0-9]*//g'`
-        fi
-        echo $ID"@elan"
-    fi
+	if [ "$1" = "'*'" ]; then echo \'*\'; else
+		if type __h2elan >/dev/null 2>&1; then
+			ID=$(__h2elan $1)
+		else
+			ID=`echo $1 | sed 's/[^0-9]*//g'`
+		fi
+		echo $ID"@elan"
+	fi
 }
 declare -fx h2elan
 
 h2o2ib() {
-    h2name_or_ip "$1" "o2ib"
+	h2name_or_ip "$1" "o2ib"
 }
 declare -fx h2o2ib
 
@@ -2294,21 +2607,30 @@ hostlist_expand() {
 }
 
 facet_host() {
-    local facet=$1
+	local facet=$1
+	local varname
 
-    [ "$facet" == client ] && echo -n $HOSTNAME && return
-    varname=${facet}_HOST
-    if [ -z "${!varname}" ]; then
-        if [ "${facet:0:3}" == "ost" ]; then
-            eval ${facet}_HOST=${ost_HOST}
-        fi
-    fi
-    echo -n ${!varname}
+	[ "$facet" == client ] && echo -n $HOSTNAME && return
+	varname=${facet}_HOST
+	if [ -z "${!varname}" ]; then
+		if [ "${facet:0:3}" == "ost" ]; then
+			local fh=${facet%failover}_HOST
+			eval export ${facet}_HOST=${!fh}
+			if [ -z "${!varname}" ]; then
+				eval export ${facet}_HOST=${ost_HOST}
+			fi
+		elif [ "${facet:0:3}" == "mdt" -o \
+			"${facet:0:3}" == "mds" -o \
+			"${facet:0:3}" == "mgs" ]; then
+			eval export ${facet}_HOST=${mds_HOST}
+		fi
+	fi
+	echo -n ${!varname}
 }
 
 facet_failover_host() {
 	local facet=$1
-	local var
+	local varname
 
 	var=${facet}failover_HOST
 	if [ -n "${!var}" ]; then
@@ -2316,12 +2638,18 @@ facet_failover_host() {
 		return
 	fi
 
+	if [ "${facet:0:3}" == "mdt" -o "${facet:0:3}" == "mds" -o \
+	     "${facet:0:3}" == "mgs" ]; then
+
+		eval export ${facet}failover_host=${mds_HOST}
+		echo ${mds_HOST}
+		return
+	fi
+
 	if [[ $facet == ost* ]]; then
-		var=ostfailover_HOST
-		if [ -n "${!var}" ]; then
-			echo ${!var}
-			return
-		fi
+		eval export ${facet}failover_host=${ost_HOST}
+		echo ${ost_HOST}
+		return
 	fi
 }
 
@@ -2351,6 +2679,24 @@ facet_active_host() {
     fi
 }
 
+# Get the passive failover partner host of facet.
+facet_passive_host() {
+	local facet=$1
+	[[ $facet = client ]] && return
+
+	local host=${facet}_HOST
+	local failover_host=${facet}failover_HOST
+	local active_host=$(facet_active_host $facet)
+
+	[[ -z ${!failover_host} || ${!failover_host} = ${!host} ]] && return
+
+	if [[ $active_host = ${!host} ]]; then
+		echo -n ${!failover_host}
+	else
+		echo -n ${!host}
+	fi
+}
+
 change_active() {
     local facetlist=$1
     local facet
@@ -2444,8 +2790,6 @@ get_env_vars() {
 		echo -n " ${var}=\"$value\""
 	done
 
-	echo -n " USE_OFD=$USE_OFD"
-
 	for facet in ${facets//,/ }; do
 		var=${facet}_FSTYPE
 		if [ -n "${!var}" ]; then
@@ -2538,18 +2882,32 @@ do_nodesv() {
 }
 
 add() {
-    local facet=$1
-    shift
-    # make sure its not already running
-    stop ${facet} -f
-    rm -f $TMP/${facet}active
-    [[ $facet = mds1 ]] && combined_mgs_mds && rm -f $TMP/mgsactive
-    do_facet ${facet} $MKFS $*
+	local facet=$1
+	shift
+	# make sure its not already running
+	stop ${facet} -f
+	rm -f $TMP/${facet}active
+	[[ $facet = mds1 ]] && combined_mgs_mds && rm -f $TMP/mgsactive
+	do_facet ${facet} $MKFS $* || return ${PIPESTATUS[0]}
+
+	if [[ $(facet_fstype $facet) == zfs ]]; then
+		#
+		# After formatting a ZFS target, "cachefile=none" property will
+		# be set on the ZFS storage pool so that the pool is not
+		# automatically imported on system startup. And then the pool
+		# will be exported so as to leave the importing and exporting
+		# operations handled by mount_facet() and stop() separately.
+		#
+		refresh_partition_table $facet $(facet_vdevice $facet)
+		disable_zpool_cache $facet
+		export_zpool $facet
+	fi
 }
 
+# Device formatted as ost
 ostdevname() {
-    num=$1
-    DEVNAME=OSTDEV$num
+	local num=$1
+	local DEVNAME=OSTDEV$num
 
 	local fstype=$(facet_fstype ost$num)
 
@@ -2558,8 +2916,9 @@ ostdevname() {
 			#if $OSTDEVn isn't defined, default is $OSTDEVBASE + num
 			eval DEVPTR=${!DEVNAME:=${OSTDEVBASE}${num}};;
 		zfs )
-			#dataset name is independent of vdev device names
-			eval DEVPTR=${FSNAME}-ost${num}/ost${num};;
+			#try $OSTZFSDEVn - independent of vdev
+			DEVNAME=OSTZFSDEV$num
+			eval DEVPTR=${!DEVNAME:=${FSNAME}-ost${num}/ost${num}};;
 		* )
 			error "unknown fstype!";;
 	esac
@@ -2567,9 +2926,11 @@ ostdevname() {
     echo -n $DEVPTR
 }
 
+# Physical device location of data
 ostvdevname() {
-	num=$1
-	DEVNAME=OSTDEV$num
+	local num=$1
+	local DEVNAME
+	local VDEVPTR
 
 	local fstype=$(facet_fstype ost$num)
 
@@ -2578,7 +2939,9 @@ ostvdevname() {
 			# vdevs are not supported by ldiskfs
 			eval VDEVPTR="";;
 		zfs )
-			#if $OSTDEVn isn't defined, default is $OSTDEVBASE + num
+			#if $OSTDEVn isn't defined, default is $OSTDEVBASE{n}
+			# Device formated by zfs
+			DEVNAME=OSTDEV$num
 			eval VDEVPTR=${!DEVNAME:=${OSTDEVBASE}${num}};;
 		* )
 			error "unknown fstype!";;
@@ -2587,19 +2950,21 @@ ostvdevname() {
 	echo -n $VDEVPTR
 }
 
+# Logical device formated for lustre
 mdsdevname() {
-    num=$1
-    DEVNAME=MDSDEV$num
+	local num=$1
+	local DEVNAME=MDSDEV$num
 
 	local fstype=$(facet_fstype mds$num)
 
 	case $fstype in
 		ldiskfs )
-			#if $MDSDEVn isn't defined, default is $MDSDEVBASE + num
+			#if $MDSDEVn isn't defined, default is $MDSDEVBASE{n}
 			eval DEVPTR=${!DEVNAME:=${MDSDEVBASE}${num}};;
 		zfs )
-			#dataset name is independent of vdev device names
-			eval DEVPTR=${FSNAME}-mdt${num}/mdt${num};;
+			# try $MDSZFSDEVn - independent of vdev
+			DEVNAME=MDSZFSDEV$num
+			eval DEVPTR=${!DEVNAME:=${FSNAME}-mdt${num}/mdt${num}};;
 		* )
 			error "unknown fstype!";;
 	esac
@@ -2607,10 +2972,10 @@ mdsdevname() {
 	echo -n $DEVPTR
 }
 
+# Physical location of data
 mdsvdevname() {
-	num=$1
-	DEVNAME=MDSDEV$num
-
+	local VDEVPTR=""
+	local num=$1
 	local fstype=$(facet_fstype mds$num)
 
 	case $fstype in
@@ -2618,7 +2983,9 @@ mdsvdevname() {
 			# vdevs are not supported by ldiskfs
 			eval VDEVPTR="";;
 		zfs )
-			#if $MDSDEVn isn't defined, default is $MDSDEVBASE + num
+			# if $MDSDEVn isn't defined, default is $MDSDEVBASE{n}
+			# Device formated by ZFS
+			local DEVNAME=MDSDEV$num
 			eval VDEVPTR=${!DEVNAME:=${MDSDEVBASE}${num}};;
 		* )
 			error "unknown fstype!";;
@@ -2628,38 +2995,51 @@ mdsvdevname() {
 }
 
 mgsdevname() {
-	DEVNAME=MGSDEV
-
-	local fstype=$(facet_fstype mds$num)
+	local DEVPTR
+	local fstype=$(facet_fstype mgs)
 
 	case $fstype in
-		ldiskfs )
-			#if $MGSDEV isn't defined, default is $MDSDEV1
-			eval DEVPTR=${!DEVNAME:=${MDSDEV1}};;
-		zfs )
-			#dataset name is independent of vdev device names
-			eval DEVPTR=${FSNAME}-mgs/mgs;;
-		* )
-			error "unknown fstype!";;
+	ldiskfs )
+		if [ $(facet_host mgs) = $(facet_host mds1) ] &&
+		   ( [ -z "$MGSDEV" ] || [ $MGSDEV = $(mdsdevname 1) ] ); then
+			DEVPTR=$(mdsdevname 1)
+		else
+			DEVPTR=$MGSDEV
+		fi;;
+	zfs )
+		if [ $(facet_host mgs) = $(facet_host mds1) ] &&
+		    ( [ -z "$MGSZFSDEV" ] &&
+			[ -z "$MGSDEV" -o "$MGSDEV" = $(mdsvdevname 1) ] ); then
+			DEVPTR=$(mdsdevname 1)
+		else
+			DEVPTR=${MGSZFSDEV:-${FSNAME}-mgs/mgs}
+		fi;;
+	* )
+		error "unknown fstype!";;
 	esac
 
-    echo -n $DEVPTR
+	echo -n $DEVPTR
 }
 
 mgsvdevname() {
-	DEVNAME=MGSDEV
+	local VDEVPTR=""
 
-	local fstype=$(facet_fstype mds$num)
+	local fstype=$(facet_fstype mgs)
 
 	case $fstype in
-		ldiskfs )
-			# vdevs are not supported by ldiskfs
-			eval VDEVPTR="";;
-		zfs )
-			#if $MGSDEV isn't defined, default is $MGSDEV1
-			eval VDEVPTR=${!DEVNAME:=${MDSDEV1}};;
-		* )
-			error "unknown fstype!";;
+	ldiskfs )
+		# vdevs are not supported by ldiskfs
+		;;
+	zfs )
+		if [ $(facet_host mgs) = $(facet_host mds1) ] &&
+		   ( [ -z "$MGSDEV" ] &&
+		       [ -z "$MGSZFSDEV" -o "$MGSZFSDEV" = $(mdsdevname 1) ]); then
+			VDEVPTR=$(mdsvdevname 1)
+		elif [ -n "$MGSDEV" ]; then
+			VDEVPTR=$MGSDEV
+		fi;;
+	* )
+		error "unknown fstype!";;
 	esac
 
 	echo -n $VDEVPTR
@@ -2675,6 +3055,84 @@ facet_mntpt () {
     echo -n $mntpt
 }
 
+mount_ldiskfs() {
+	local facet=$1
+	local dev=$(facet_device $facet)
+	local mnt=$(facet_mntpt $facet)
+	local opts
+
+	if ! do_facet $facet test -b $dev; then
+		opts="-o loop"
+	fi
+	do_facet $facet mount -t ldiskfs $opts $dev $mnt
+}
+
+unmount_ldiskfs() {
+	local facet=$1
+	local dev=$(facet_device $facet)
+	local mnt=$(facet_mntpt $facet)
+
+	do_facet $facet umount -d $mnt
+}
+
+var_name() {
+	echo -n "$1" | tr -c '[:alnum:]\n' '_'
+}
+
+mount_zfs() {
+	local facet=$1
+	local ds=$(facet_device $facet)
+	local mnt=$(facet_mntpt $facet)
+	local canmnt
+	local mntpt
+
+	import_zpool $facet
+	canmnt=$(do_facet $facet $ZFS get -H -o value canmount $ds)
+	mntpt=$(do_facet $facet $ZFS get -H -o value mountpoint $ds)
+	do_facet $facet $ZFS set canmount=noauto $ds
+	#
+	# The "legacy" mount method is used here because "zfs unmount $mnt"
+	# calls stat(2) on $mnt/../*, which may include $MOUNT.  If certain
+	# targets are not available at the time, the stat(2) on $MOUNT will
+	# hang.
+	#
+	do_facet $facet $ZFS set mountpoint=legacy $ds
+	do_facet $facet mount -t zfs $ds $mnt
+	eval export mz_$(var_name ${facet}_$ds)_canmount=$canmnt
+	eval export mz_$(var_name ${facet}_$ds)_mountpoint=$mntpt
+}
+
+unmount_zfs() {
+	local facet=$1
+	local ds=$(facet_device $facet)
+	local mnt=$(facet_mntpt $facet)
+	local var_mntpt=mz_$(var_name ${facet}_$ds)_mountpoint
+	local var_canmnt=mz_$(var_name ${facet}_$ds)_canmount
+	local mntpt=${!var_mntpt}
+	local canmnt=${!var_canmnt}
+
+	unset $var_mntpt
+	unset $var_canmnt
+	do_facet $facet umount $mnt
+	do_facet $facet $ZFS set mountpoint=$mntpt $ds
+	do_facet $facet $ZFS set canmount=$canmnt $ds
+	export_zpool $facet
+}
+
+mount_fstype() {
+	local facet=$1
+	local fstype=$(facet_fstype $facet)
+
+	mount_$fstype $facet
+}
+
+unmount_fstype() {
+	local facet=$1
+	local fstype=$(facet_fstype $facet)
+
+	unmount_$fstype $facet
+}
+
 ########
 ## MountConf setup
 
@@ -2734,7 +3192,8 @@ cleanupall() {
 }
 
 combined_mgs_mds () {
-    [[ $MDSDEV1 = $MGSDEV ]] && [[ $mds1_HOST = $mgs_HOST ]]
+	[[ "$(mdsdevname 1)" = "$(mgsdevname)" ]] &&
+		[[ "$(facet_host mds1)" = "$(facet_host mgs)" ]]
 }
 
 lower() {
@@ -2747,9 +3206,12 @@ upper() {
 
 mkfs_opts() {
 	local facet=$1
+	local dev=$2
+	local fsname=${3:-"$FSNAME"}
 	local type=$(facet_type $facet)
 	local index=$(($(facet_number $facet) - 1))
 	local fstype=$(facet_fstype $facet)
+	local host=$(facet_host $facet)
 	local opts
 	local fs_mkfs_opts
 	local var
@@ -2758,14 +3220,17 @@ mkfs_opts() {
 		return 1
 	fi
 
-	if [ $type == MGS ] || ( [ $type == MDS ] && combined_mgs_mds ); then
+	if [ $type == MGS ] || ( [ $type == MDS ] &&
+                                 [ "$dev" == $(mgsdevname) ] &&
+				 [ "$host" == "$(facet_host mgs)" ] ); then
 		opts="--mgs"
 	else
 		opts="--mgsnode=$MGSNID"
 	fi
 
 	if [ $type != MGS ]; then
-		opts+=" --fsname=$FSNAME --$(lower ${type/MDS/MDT}) --index=$index"
+		opts+=" --fsname=$fsname --$(lower ${type/MDS/MDT}) \
+			--index=$index"
 	fi
 
 	var=${facet}failover_HOST
@@ -2785,6 +3250,9 @@ mkfs_opts() {
 
 		if [ $fstype == ldiskfs ]; then
 			fs_mkfs_opts+=${MDSJOURNALSIZE:+" -J size=$MDSJOURNALSIZE"}
+			if [ ! -z $EJOURNAL ]; then
+				fs_mkfs_opts+=${MDSJOURNALSIZE:+" device=$EJOURNAL"}
+			fi
 			fs_mkfs_opts+=${MDSISIZE:+" -i $MDSISIZE"}
 		fi
 	fi
@@ -2835,23 +3303,24 @@ formatall() {
 	echo Formatting mgs, mds, osts
 	if ! combined_mgs_mds ; then
 		echo "Format mgs: $(mgsdevname)"
-		add mgs $(mkfs_opts mgs) --reformat $(mgsdevname) \
-			$(mgsvdevname) ${quiet:+>/dev/null} || exit 10
-		fi
+		add mgs $(mkfs_opts mgs $(mgsdevname)) --reformat \
+			$(mgsdevname) $(mgsvdevname) ${quiet:+>/dev/null} ||
+			exit 10
+	fi
 
-		for num in `seq $MDSCOUNT`; do
-			echo "Format mds$num: $(mdsdevname $num)"
-			add mds$num $(mkfs_opts mds$num) --reformat \
-			$(mdsdevname $num) $(mdsvdevname $num) \
+	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
-		done
+	done
 
-		for num in `seq $OSTCOUNT`; do
-			echo "Format ost$num: $(ostdevname $num)"
-			add ost$num $(mkfs_opts ost$num) --reformat \
-			$(ostdevname $num) $(ostvdevname ${num}) \
+	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
-		done
+	done
 }
 
 mount_client() {
@@ -2899,23 +3368,31 @@ remount_client()
         zconf_mount `hostname` $1 || error "mount failed"
 }
 
-writeconf_facet () {
-    local facet=$1
-    local dev=$2
+writeconf_facet() {
+	local facet=$1
+	local dev=$2
 
-    do_facet $facet "$TUNEFS --writeconf $dev"
+	stop ${facet} -f
+	rm -f $TMP/${facet}active
+	do_facet ${facet} "$TUNEFS --quiet --writeconf $dev" || return 1
+	return 0
 }
 
 writeconf_all () {
-    for num in `seq $MDSCOUNT`; do
-        DEVNAME=$(mdsdevname $num)
-        writeconf_facet mds$num $DEVNAME
-    done
+	local mdt_count=${1:-$MDSCOUNT}
+	local ost_count=${2:-$OSTCOUNT}
+	local rc=0
 
-    for num in `seq $OSTCOUNT`; do
-        DEVNAME=$(ostdevname $num)
-        writeconf_facet ost$num $DEVNAME
-    done
+	for num in $(seq $mdt_count); do
+		DEVNAME=$(mdsdevname $num)
+		writeconf_facet mds$num $DEVNAME || rc=$?
+	done
+
+	for num in $(seq $ost_count); do
+		DEVNAME=$(ostdevname $num)
+		writeconf_facet ost$num $DEVNAME || rc=$?
+	done
+	return $rc
 }
 
 setupall() {
@@ -3044,9 +3521,14 @@ init_facet_vars () {
 
 	local varname=${facet}failover_HOST
 	if [ -z "${!varname}" ]; then
-		eval $varname=$(facet_host $facet)
+		eval export $varname=$(facet_host $facet)
 	fi
 
+	varname=${facet}_HOST
+	if [ -z "${!varname}" ]; then
+		eval export $varname=$(facet_host $facet)
+ 	fi
+
 	# ${facet}failover_dev is set in cfg file
 	varname=${facet}failover_dev
 	if [ -n "${!varname}" ] ; then
@@ -3068,23 +3550,23 @@ init_facet_vars () {
 }
 
 init_facets_vars () {
-    local DEVNAME
+	local DEVNAME
 
-    if ! remote_mds_nodsh; then 
-        for num in `seq $MDSCOUNT`; do
-            DEVNAME=`mdsdevname $num`
-            init_facet_vars mds$num $DEVNAME $MDS_MOUNT_OPTS
-        done
-    fi
+	if ! remote_mds_nodsh; then
+		for num in $(seq $MDSCOUNT); do
+			DEVNAME=`mdsdevname $num`
+			init_facet_vars mds$num $DEVNAME $MDS_MOUNT_OPTS
+		done
+	fi
 
 	combined_mgs_mds || init_facet_vars mgs $(mgsdevname) $MGS_MOUNT_OPTS
 
-    remote_ost_nodsh && return
-
-    for num in `seq $OSTCOUNT`; do
-        DEVNAME=`ostdevname $num`
-        init_facet_vars ost$num $DEVNAME $OST_MOUNT_OPTS
-    done
+	if ! remote_ost_nodsh; then
+		for num in $(seq $OSTCOUNT); do
+			DEVNAME=$(ostdevname $num)
+			init_facet_vars ost$num $DEVNAME $OST_MOUNT_OPTS
+		done
+	fi
 }
 
 osc_ensure_active () {
@@ -3134,26 +3616,20 @@ init_param_vars () {
 	osc_ensure_active $SINGLEMDS $TIMEOUT
 	osc_ensure_active client $TIMEOUT
 
-	local jobid_var
-	if [ -z "$(lctl get_param -n mdc.*.connect_flags | grep jobstats)" ]; then
-		jobid_var="none"
-	elif [ $JOBSTATS_AUTO -ne 0 ]; then
-		echo "enable jobstats, set job scheduler as $JOBID_VAR"
-		jobid_var=$JOBID_VAR
-	else
-		jobid_var=`$LCTL get_param -n jobid_var`
-		if [ $jobid_var != "disable" ]; then
-			echo "disable jobstats as required"
-			jobid_var="disable"
-		else
-			jobid_var="none"
-		fi
-	fi
+	if [ -n "$(lctl get_param -n mdc.*.connect_flags|grep jobstats)" ]; then
+		local current_jobid_var=$($LCTL get_param -n jobid_var)
 
-	if [ $jobid_var == $JOBID_VAR -o $jobid_var == "disable" ]; then
-		do_facet mgs $LCTL conf_param $FSNAME.sys.jobid_var=$jobid_var
-		wait_update $HOSTNAME "$LCTL get_param -n jobid_var" \
-			$jobid_var || return 1
+		if [ $JOBID_VAR = "existing" ]; then
+			echo "keeping jobstats as $current_jobid_var"
+		elif [ $current_jobid_var != $JOBID_VAR ]; then
+			echo "seting jobstats to $JOBID_VAR"
+
+			set_conf_param_and_check client			\
+				"$LCTL get_param -n jobid_var"		\
+				"$FSNAME.sys.jobid_var" $JOBID_VAR
+		fi
+	else
+		echo "jobstats not supported by server"
 	fi
 
 	if [ $QUOTA_AUTO -ne 0 ]; then
@@ -3257,17 +3733,24 @@ is_mounted () {
 }
 
 is_empty_dir() {
-    [ $(find $1 -maxdepth 1 -print | wc -l) = 1 ] && return 0
-    return 1
+	[ $(find $1 -maxdepth 1 -print | wc -l) = 1 ] && return 0
+	return 1
 }
 
 # empty lustre filesystem may have empty directories lost+found and .lustre
 is_empty_fs() {
-    [ $(find $1 -maxdepth 1 -name lost+found -o -name .lustre -prune -o \
-       -print | wc -l) = 1 ] || return 1
-    [ ! -d $1/lost+found ] || is_empty_dir $1/lost+found && return 0
-    [ ! -d $1/.lustre ] || is_empty_dir $1/.lustre && return 0
-    return 1
+	# exclude .lustre & lost+found
+	[ $(find $1 -maxdepth 1 -name lost+found -o -name .lustre -prune -o \
+		-print | wc -l) = 1 ] || return 1
+	[ ! -d $1/lost+found ] || is_empty_dir $1/lost+found || return 1
+	if [ $(lustre_version_code $SINGLEMDS) -gt $(version_code 2.4.0) ]; then
+		# exclude .lustre/fid (LU-2780)
+		[ $(find $1/.lustre -maxdepth 1 -name fid -prune -o \
+			-print | wc -l) = 1 ] || return 1
+	else
+		[ ! -d $1/.lustre ] || is_empty_dir $1/.lustre || return 1
+	fi
+	return 0
 }
 
 check_and_setup_lustre() {
@@ -3328,6 +3811,21 @@ check_and_setup_lustre() {
         set_default_debug_nodes $(comma_list $(nodes_list))
     fi
 
+	if [ $(lower $OSD_TRACK_DECLARES_LBUG) == 'yes' ] ; then
+		local facets=""
+		[ "$(facet_fstype ost1)" = "ldiskfs" ] &&
+			facets="$(get_facets OST)"
+		[ "$(facet_fstype mds1)" = "ldiskfs" ] &&
+			facets="$facets,$(get_facets MDS)"
+		[ "$(facet_fstype mgs)" = "ldiskfs" ] &&
+			facets="$facets,mgs"
+		local nodes="$(facets_hosts ${facets})"
+		if [ -n "$nodes" ] ; then
+			do_nodes $nodes "$LCTL set_param \
+				 osd-ldiskfs.track_declares_assert=1 || true"
+		fi
+	fi
+
 	init_gss
 	if $GSS; then
 		set_flavor_all $SEC
@@ -3371,7 +3869,7 @@ get_mnt_devs() {
 	local dev
 
 	if [ "$type" == ost ]; then
-		devs=$(get_obdfilter_param $node "" mntdev)
+		devs=$(get_osd_param $node "" mntdev)
 	else
 		devs=$(do_node $node \
 		       "lctl get_param -n osd-*.$FSNAME-M*.mntdev")
@@ -3407,7 +3905,7 @@ run_e2fsck() {
     local extra_opts=$3
 
     df > /dev/null      # update statfs data on disk
-    local cmd="$E2FSCK -d -v -t -t -f -n $extra_opts $target_dev"
+    local cmd="$E2FSCK -d -v -t -t -f $extra_opts $target_dev"
     echo $cmd
     local rc=0
     do_node $node $cmd || rc=$?
@@ -3416,13 +3914,27 @@ run_e2fsck() {
     return 0
 }
 
+#
+# Run resize2fs on MDT or OST device.
+#
+run_resize2fs() {
+	local facet=$1
+	local device=$2
+	local size=$3
+	shift 3
+	local opts="$@"
+
+	do_facet $facet "$RESIZE2FS $opts $device $size"
+}
+
 # verify a directory is shared among nodes.
 check_shared_dir() {
 	local dir=$1
+	local list=${2:-$(comma_list $(nodes_list))}
 
 	[ -z "$dir" ] && return 1
-	do_rpc_nodes "$(comma_list $(nodes_list))" check_logdir $dir
-	check_write_access $dir || return 1
+	do_rpc_nodes "$list" check_logdir $dir
+	check_write_access $dir "$list" || return 1
 	return 0
 }
 
@@ -3443,14 +3955,14 @@ generate_db() {
 
     [ $MDSCOUNT -eq 1 ] || error "CMD is not supported"
 
-    run_e2fsck $(mdts_nodes) $MDTDEV "--mdsdb $MDSDB"
+    run_e2fsck $(mdts_nodes) $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 "--mdsdb $MDSDB --ostdb $OSTDB-$ostidx"
+            run_e2fsck $node $dev "-n --mdsdb $MDSDB --ostdb $OSTDB-$ostidx"
             OSTDB_LIST="$OSTDB_LIST $OSTDB-$ostidx"
             ostidx=$((ostidx + 1))
         done
@@ -3458,17 +3970,55 @@ generate_db() {
     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 $node $cmd || rc=$?
+	#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 $rc
+}
+
 run_lfsck() {
-    local cmd="$LFSCK_BIN -c -l --mdsdb $MDSDB --ostdb $OSTDB_LIST $MOUNT"
-    echo $cmd
-    local rc=0
-    eval $cmd || rc=$?
-    [ $rc -le $FSCK_MAX_ERR ] || \
-        error "$cmd returned $rc, should be <= $FSCK_MAX_ERR"
-    echo "lfsck finished with rc=$rc"
+	local facets="client $SINGLEMDS"
+	local found=false
+	local facet
+	local node
+	local rc=0
 
-    rm -rvf $MDSDB* $OSTDB* || true
-    return 0
+	for facet in $facets; do
+		node=$(facet_active_host $facet)
+		if check_progs_installed $node $LFSCK_BIN; then
+			found=true
+			break
+		fi
+	done
+	! $found && error "None of \"$facets\" supports lfsck"
+
+	run_lfsck_remote $node || rc=$?
+
+	rm -rvf $MDSDB* $OSTDB* || true
+	return $rc
 }
 
 check_and_cleanup_lustre() {
@@ -3619,7 +4169,7 @@ get_facets () {
 
         case $type in
                 MGS ) list="$list $name";;
-            MDS|OST ) local count=${type}COUNT
+            MDS|OST|AGT ) local count=${type}COUNT
                        for ((i=1; i<=${!count}; i++)) do
                           list="$list ${name}$i"
                       done;;
@@ -3709,14 +4259,29 @@ drop_reint_reply() {
     return $RC
 }
 
+drop_update_reply() {
+# OBD_FAIL_UPDATE_OBJ_NET_REP
+	local index=$1
+	shift 1
+	RC=0
+	do_facet mds${index} lctl set_param fail_loc=0x1701
+	do_facet client "$@" || RC=$?
+	do_facet mds${index} lctl set_param fail_loc=0
+	return $RC
+}
+
 pause_bulk() {
 #define OBD_FAIL_OST_BRW_PAUSE_BULK      0x214
-    RC=0
-    do_facet ost1 lctl set_param fail_loc=0x214
-    do_facet client "$1" || RC=$?
-    do_facet client "sync"
-    do_facet ost1 lctl set_param fail_loc=0
-    return $RC
+	RC=0
+
+	local timeout=${2:-0}
+	# default is (obd_timeout / 4) if unspecified
+	echo "timeout is $timeout/$2"
+	do_facet ost1 lctl set_param fail_val=$timeout fail_loc=0x80000214
+	do_facet client "$1" || RC=$?
+	do_facet client "sync"
+	do_facet ost1 lctl set_param fail_loc=0
+	return $RC
 }
 
 drop_ldlm_cancel() {
@@ -3733,11 +4298,11 @@ drop_ldlm_cancel() {
 
 drop_bl_callback() {
 #define OBD_FAIL_LDLM_BL_CALLBACK_NET			0x305
-    RC=0
-    do_facet client lctl set_param fail_loc=0x305
-    do_facet client "$@" || RC=$?
-    do_facet client lctl set_param fail_loc=0
-    return $RC
+	RC=0
+	do_facet client lctl set_param fail_loc=0x80000305
+	do_facet client "$@" || RC=$?
+	do_facet client lctl set_param fail_loc=0
+	return $RC
 }
 
 drop_ldlm_reply() {
@@ -3758,7 +4323,7 @@ clear_failloc() {
 }
 
 set_nodes_failloc () {
-    do_nodes $(comma_list $1)  lctl set_param fail_loc=$2
+	do_nodes $(comma_list $1)  lctl set_param fail_val=0 fail_loc=$2
 }
 
 cancel_lru_locks() {
@@ -3851,28 +4416,37 @@ log_trace_dump() {
 ##################################
 
 error_noexit() {
-    local TYPE=${TYPE:-"FAIL"}
+	local TYPE=${TYPE:-"FAIL"}
 
-    local dump=true
-    # do not dump logs if $1=false
-    if [ "x$1" = "xfalse" ]; then
-        shift
-        dump=false
-    fi
+	local dump=true
+	# do not dump logs if $1=false
+	if [ "x$1" = "xfalse" ]; then
+		shift
+		dump=false
+	fi
 
-    log " ${TESTSUITE} ${TESTNAME}: @@@@@@ ${TYPE}: $@ "
-    log_trace_dump
 
-    mkdir -p $LOGDIR
-    # We need to dump the logs on all nodes
-    if $dump; then
-        gather_logs $(comma_list $(nodes_list))
-    fi
+	log " ${TESTSUITE} ${TESTNAME}: @@@@@@ ${TYPE}: $@ "
+	log_trace_dump
+
+	mkdir -p $LOGDIR
+	# We need to dump the logs on all nodes
+	if $dump; then
+		gather_logs $(comma_list $(nodes_list))
+	fi
 
 	debugrestore
 	[ "$TESTSUITELOG" ] &&
 		echo "$TESTSUITE: $TYPE: $TESTNAME $@" >> $TESTSUITELOG
-	echo "$@" > $LOGDIR/err
+	if [ -z "$*" ]; then
+		echo "error() without useful message, please fix" > $LOGDIR/err
+	else
+		if [[ `echo $TYPE | grep ^IGNORE` ]]; then
+			echo "$@" > $LOGDIR/ignore
+		else
+			echo "$@" > $LOGDIR/err
+		fi
+	fi
 }
 
 exit_status () {
@@ -3884,21 +4458,22 @@ exit_status () {
 }
 
 error() {
-    error_noexit "$@"
-    exit 1
+	error_noexit "$@"
+	exit 1
 }
 
 error_exit() {
-    error "$@"
+	error "$@"
 }
 
 # use only if we are ignoring failures for this test, bugno required.
 # (like ALWAYS_EXCEPT, but run the test and ignore the results.)
-# e.g. error_ignore 5494 "your message"
+# e.g. error_ignore bz5494 "your message" or
+# error_ignore LU-5494 "your message"
 error_ignore() {
-    local TYPE="IGNORE (bz$1)"
-    shift
-    error_noexit "$@"
+	local TYPE="IGNORE ($1)"
+	shift
+	error_noexit "$@"
 }
 
 error_and_remount() {
@@ -3908,7 +4483,7 @@ error_and_remount() {
 }
 
 skip_env () {
-    $FAIL_ON_SKIP_ENV && error false $@ || skip $@
+	$FAIL_ON_SKIP_ENV && error false $@ || skip $@
 }
 
 skip() {
@@ -3988,7 +4563,6 @@ run_test() {
             return $?
         fi
         LAST_SKIPPED="y"
-        echo -n "."
         return 0
     fi
 
@@ -4109,31 +4683,35 @@ banner() {
 }
 
 #
-# Run a single test function and cleanup after it.  
+# Run a single test function and cleanup after it.
 #
 # This function should be run in a subshell so the test func can
 # exit() without stopping the whole script.
 #
 run_one() {
-    local testnum=$1
-    local message=$2
-    tfile=f.${TESTSUITE}.${testnum}
-    export tdir=d0.${TESTSUITE}/d${base}
-    export TESTNAME=test_$testnum
-    local SAVE_UMASK=`umask`
-    umask 0022
-
-    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"
-    ps auxww | grep -v grep | grep -q multiop && error "multiop still running"
-    unset TESTNAME
-    unset tdir
-    umask $SAVE_UMASK
-    return 0
+	local testnum=$1
+	local message=$2
+	export tfile=f${testnum}.${TESTSUITE}
+	export tdir=d${testnum}.${TESTSUITE}
+	export TESTNAME=test_$testnum
+	local SAVE_UMASK=`umask`
+	umask 0022
+
+	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 $?"
+	check_catastrophe || error "LBUG/LASSERT detected"
+	if [ "$PARALLEL" != "yes" ]; then
+		ps auxww | grep -v grep | grep -q multiop &&
+					error "multiop still running"
+	fi
+	unset TESTNAME
+	unset tdir
+	unset tfile
+	umask $SAVE_UMASK
+	return 0
 }
 
 #
@@ -4148,6 +4726,7 @@ run_one_logged() {
 	local name=${TESTSUITE}.test_${1}.test_log.$(hostname -s).log
 	local test_log=$LOGDIR/$name
 	rm -rf $LOGDIR/err
+	rm -rf $LOGDIR/ignore
 	rm -rf $LOGDIR/skip
 	local SAVE_UMASK=`umask`
 	umask 0022
@@ -4157,7 +4736,7 @@ run_one_logged() {
 	(run_one $1 "$2") 2>&1 | tee -i $test_log
 	local RC=${PIPESTATUS[0]}
 
-	[ $RC -ne 0 ] && [ ! -f $LOGDIR/err ] && \
+	[ $RC -ne 0 ] && [ ! -f $LOGDIR/err ] &&
 		echo "test_$1 returned $RC" | tee $LOGDIR/err
 
 	duration=$((`date +%s` - $BEFORE))
@@ -4165,11 +4744,17 @@ run_one_logged() {
 
 	if [[ -f $LOGDIR/err ]]; then
 		TEST_ERROR=$(cat $LOGDIR/err)
+	elif [[ -f $LOGDIR/ignore ]]; then
+		TEST_ERROR=$(cat $LOGDIR/ignore)
 	elif [[ -f $LOGDIR/skip ]]; then
 		TEST_ERROR=$(cat $LOGDIR/skip)
 	fi
 	log_sub_test_end $TEST_STATUS $duration "$RC" "$TEST_ERROR"
 
+	if [[ "$TEST_STATUS" != "SKIP" ]] && [[ -f $TF_SKIP ]]; then
+		rm -f $TF_SKIP
+	fi
+
 	if [ -f $LOGDIR/err ]; then
 		$FAIL_ON_ERROR && exit $RC
 	fi
@@ -4264,9 +4849,45 @@ mdtuuid_from_index()
     $LFS mdts $2 | sed -ne "/^$1: /s/.* \(.*\) .*$/\1/p"
 }
 
+# Description:
+#   Return unique identifier for given hostname
+host_id() {
+	local host_name=$1
+	echo $host_name | md5sum | cut -d' ' -f1
+}
+
+# Description:
+#   Returns list of ip addresses for each interface
+local_addr_list() {
+	ip addr | awk '/inet\ / {print $2}' | awk -F\/ '{print $1}'
+}
+
+is_local_addr() {
+	local addr=$1
+	# Cache address list to avoid mutiple execution of local_addr_list
+	LOCAL_ADDR_LIST=${LOCAL_ADDR_LIST:-$(local_addr_list)}
+	local i
+	for i in $LOCAL_ADDR_LIST ; do
+		[[ "$i" == "$addr" ]] && return 0
+	done
+	return 1
+}
+
+local_node() {
+	local host_name=$1
+	local is_local="IS_LOCAL_$(host_id $host_name)"
+	if [ -z "${!is_local-}" ] ; then
+		eval $is_local=0
+		local host_ip=$($LUSTRE/tests/resolveip $host_name)
+		is_local_addr "$host_ip" && eval $is_local=1
+	fi
+	[[ "${!is_local}" == "1" ]]
+}
+
 remote_node () {
-    local node=$1
-    [ "$node" != "$(hostname)" ]
+	local node=$1
+	local_node $node && return 1
+	return 0
 }
 
 remote_mds ()
@@ -4326,64 +4947,126 @@ local_mode ()
         $(single_local_node $(comma_list $(nodes_list)))
 }
 
-mdts_nodes () {
-    local MDSNODES
-    local NODES_sort
-    for num in `seq $MDSCOUNT`; do
-        MDSNODES="$MDSNODES $(facet_host mds$num)"
-    done
-    NODES_sort=$(for i in $MDSNODES; do echo $i; done | sort -u)
-
-    echo $NODES_sort
-}
-
 remote_servers () {
     remote_ost && remote_mds
 }
 
+# Get the active nodes for facets.
 facets_nodes () {
-    local facets=$1
-    local nodes
-    local NODES_sort
+	local facets=$1
+	local facet
+	local nodes
+	local nodes_sort
+	local i
 
-    for facet in ${facets//,/ }; do
-        if [ "$FAILURE_MODE" = HARD ]; then
-            nodes="$nodes $(facet_active_host $facet)"
-        else
-            nodes="$nodes $(facet_host $facet)"
-        fi
-    done
-    NODES_sort=$(for i in $nodes; do echo $i; done | sort -u)
+	for facet in ${facets//,/ }; do
+		nodes="$nodes $(facet_active_host $facet)"
+	done
+
+	nodes_sort=$(for i in $nodes; do echo $i; done | sort -u)
+	echo -n $nodes_sort
+}
 
-    echo $NODES_sort
+# Get all of the active MDS nodes.
+mdts_nodes () {
+	echo -n $(facets_nodes $(get_facets MDS))
 }
 
+# Get all of the active OSS nodes.
 osts_nodes () {
-    local facets=$(get_facets OST)
-    local nodes=$(facets_nodes $facets)
+	echo -n $(facets_nodes $(get_facets OST))
+}
 
-    echo $nodes
+# Get all of the active AGT (HSM agent) nodes.
+agts_nodes () {
+	echo -n $(facets_nodes $(get_facets AGT))
 }
 
+# Get all of the client nodes and active server nodes.
 nodes_list () {
-    # FIXME. We need a list of clients
-    local myNODES=$HOSTNAME
-    local myNODES_sort
-
-    # CLIENTS (if specified) contains the local client
-    [ -n "$CLIENTS" ] && myNODES=${CLIENTS//,/ }
+	local nodes=$HOSTNAME
+	local nodes_sort
+	local i
 
-    if [ "$PDSH" -a "$PDSH" != "no_dsh" ]; then
-        myNODES="$myNODES $(facets_nodes $(get_facets))"
-    fi
+	# CLIENTS (if specified) contains the local client
+	[ -n "$CLIENTS" ] && nodes=${CLIENTS//,/ }
 
-    myNODES_sort=$(for i in $myNODES; do echo $i; done | sort -u)
+	if [ "$PDSH" -a "$PDSH" != "no_dsh" ]; then
+		nodes="$nodes $(facets_nodes $(get_facets))"
+	fi
 
-    echo $myNODES_sort
+	nodes_sort=$(for i in $nodes; do echo $i; done | sort -u)
+	echo -n $nodes_sort
 }
 
+# Get all of the remote client nodes and remote active server nodes.
 remote_nodes_list () {
-	echo $(nodes_list) | sed -re "s/\<$HOSTNAME\>//g"
+	echo -n $(nodes_list) | sed -re "s/\<$HOSTNAME\>//g"
+}
+
+# Get all of the MDS nodes, including active and passive nodes.
+all_mdts_nodes () {
+	local host
+	local failover_host
+	local nodes
+	local nodes_sort
+	local i
+
+	for i in $(seq $MDSCOUNT); do
+		host=mds${i}_HOST
+		failover_host=mds${i}failover_HOST
+		nodes="$nodes ${!host} ${!failover_host}"
+	done
+
+	nodes_sort=$(for i in $nodes; do echo $i; done | sort -u)
+	echo -n $nodes_sort
+}
+
+# Get all of the OSS nodes, including active and passive nodes.
+all_osts_nodes () {
+	local host
+	local failover_host
+	local nodes
+	local nodes_sort
+	local i
+
+	for i in $(seq $OSTCOUNT); do
+		host=ost${i}_HOST
+		failover_host=ost${i}failover_HOST
+		nodes="$nodes ${!host} ${!failover_host}"
+	done
+
+	nodes_sort=$(for i in $nodes; do echo $i; done | sort -u)
+	echo -n $nodes_sort
+}
+
+# Get all of the server nodes, including active and passive nodes.
+all_server_nodes () {
+	local nodes
+	local nodes_sort
+	local i
+
+	nodes="$mgs_HOST $mgsfailover_HOST $(all_mdts_nodes) $(all_osts_nodes)"
+
+	nodes_sort=$(for i in $nodes; do echo $i; done | sort -u)
+	echo -n $nodes_sort
+}
+
+# Get all of the client and server nodes, including active and passive nodes.
+all_nodes () {
+	local nodes=$HOSTNAME
+	local nodes_sort
+	local i
+
+	# CLIENTS (if specified) contains the local client
+	[ -n "$CLIENTS" ] && nodes=${CLIENTS//,/ }
+
+	if [ "$PDSH" -a "$PDSH" != "no_dsh" ]; then
+		nodes="$nodes $(all_server_nodes)"
+	fi
+
+	nodes_sort=$(for i in $nodes; do echo $i; done | sort -u)
+	echo -n $nodes_sort
 }
 
 init_clients_lists () {
@@ -4464,10 +5147,11 @@ generate_machine_file() {
 }
 
 get_stripe () {
-    local file=$1/stripe
-    touch $file
-    $LFS getstripe -v $file || error
-    rm -f $file
+	local file=$1/stripe
+
+	touch $file
+	$LFS getstripe -v $file || error "getstripe $file failed"
+	rm -f $file
 }
 
 setstripe_nfsserver () {
@@ -4684,21 +5368,33 @@ calc_osc_kbytes () {
         $LCTL get_param -n osc.*[oO][sS][cC][-_][0-9a-f]*.$1 | calc_sum
 }
 
-# save_lustre_params(node, parameter_mask)
-# generate a stream of formatted strings (<node> <param name>=<param value>)
+# save_lustre_params(comma separated facet list, parameter_mask)
+# generate a stream of formatted strings (<facet> <param name>=<param value>)
 save_lustre_params() {
-        local s
-        do_nodesv $1 "lctl get_param $2 | while read s; do echo \\\$s; done"
+	local facets=$1
+	local facet
+	local nodes
+	local node
+
+	for facet in ${facets//,/ }; do
+		node=$(facet_active_host $facet)
+		[[ *\ $node\ * = " $nodes " ]] && continue
+		nodes="$nodes $node"
+
+		do_node $node "$LCTL get_param $2 |
+			while read s; do echo $facet \\\$s; done"
+	done
 }
 
 # restore lustre parameters from input stream, produces by save_lustre_params
 restore_lustre_params() {
-        local node
-        local name
-        local val
-        while IFS=" =" read node name val; do
-                do_node ${node//:/} "lctl set_param -n $name $val"
-        done
+	local facet
+	local name
+	local val
+
+	while IFS=" =" read facet name val; do
+		do_facet $facet "$LCTL set_param -n $name $val"
+	done
 }
 
 check_catastrophe() {
@@ -4708,9 +5404,17 @@ check_catastrophe() {
 
 	[ -z "$rnodes" ] && return 0
 
-	do_nodes "$rnodes" "rc=\\\$([ -f $C ] && echo \\\$(< $C) || echo 0);
+	local data
+	data=$(do_nodes "$rnodes" "rc=\\\$([ -f $C ] &&
+		echo \\\$(< $C) || echo 0);
 		if [ \\\$rc -ne 0 ]; then echo \\\$(hostname): \\\$rc; fi
-		exit \\\$rc;"
+		exit \\\$rc")
+	local rc=$?
+	if [ -n "$data" ]; then
+	    echo $data
+	    return $rc
+	fi
+	return 0
 }
 
 # CMD: determine mds index where directory inode presents
@@ -4741,10 +5445,11 @@ get_mds_dir () {
 }
 
 mdsrate_cleanup () {
-    if [ -d $4 ]; then
-        mpi_run -np $1 -machinefile $2 ${MDSRATE} --unlink --nfiles $3 --dir $4 --filefmt $5 $6
-        rmdir $4
-    fi
+	if [ -d $4 ]; then
+		mpi_run ${MACHINEFILE_OPTION} $2 -np $1 ${MDSRATE} --unlink \
+			--nfiles $3 --dir $4 --filefmt $5 $6
+		rmdir $4
+	fi
 }
 
 delayed_recovery_enabled () {
@@ -4754,7 +5459,7 @@ delayed_recovery_enabled () {
 
 ########################
 
-convert_facet2label() { 
+convert_facet2label() {
     local facet=$1
 
     if [ x$facet = xost ]; then
@@ -4765,13 +5470,13 @@ convert_facet2label() {
 
     if [ -n ${!varsvc} ]; then
         echo ${!varsvc}
-    else  
+    else
         error "No lablel for $facet!"
     fi
 }
 
 get_clientosc_proc_path() {
-    echo "${1}-osc-[^M]*"
+    echo "${1}-osc-*"
 }
 
 get_lustre_version () {
@@ -4835,7 +5540,7 @@ _wait_import_state () {
     local CONN_STATE
     local i=0
 
-    CONN_STATE=$($LCTL get_param -n $CONN_PROC 2>/dev/null | cut -f2)
+	CONN_STATE=$($LCTL get_param -n $CONN_PROC 2>/dev/null | cut -f2 | uniq)
     while [ "${CONN_STATE}" != "${expected}" ]; do
         if [ "${expected}" == "DISCONN" ]; then
             # for disconn we can check after proc entry is removed
@@ -4848,7 +5553,8 @@ _wait_import_state () {
             error "can't put import for $CONN_PROC into ${expected} state after $i sec, have ${CONN_STATE}" && \
             return 1
         sleep 1
-        CONN_STATE=$($LCTL get_param -n $CONN_PROC 2>/dev/null | cut -f2)
+	# Add uniq for multi-mount case
+	CONN_STATE=$($LCTL get_param -n $CONN_PROC 2>/dev/null | cut -f2 | uniq)
         i=$(($i + 1))
     done
 
@@ -4867,6 +5573,14 @@ wait_import_state() {
     done
 }
 
+wait_import_state_mount() {
+	if ! is_mounted $MOUNT && ! is_mounted $MOUNT2; then
+		return 0
+	fi
+
+	wait_import_state $*
+}
+
 # One client request could be timed out because server was not ready
 # when request was sent by client.
 # The request timeout calculation details :
@@ -4907,20 +5621,42 @@ request_timeout () {
     echo $(( init_connect_timeout + at_min ))
 }
 
-wait_osc_import_state() {
+_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.${ost}.ost_server_uuid"
+	local param="osc.${ost}.ost_server_uuid"
+	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 ! do_rpc_nodes "$(facet_host $facet)" \
-			_wait_import_state $expected $param $maxtime; then
+	#During setup time, the osc might not be setup, it need wait
+	#until list_param can return valid value. And also if there
+	#are mulitple osc entries we should list all of them before
+	#go to wait.
+	local params=$($LCTL list_param $param 2>/dev/null || true)
+	while [ -z "$params" ]; do
+		if [ $i -ge $maxtime ]; then
+			echo "can't get $param by list_param in $maxtime secs"
+			if [[ $facet != client* ]]; then
+				echo "Go with $param directly"
+				params=$param
+				break
+			else
+				return 1
+			fi
+		fi
+		sleep 1
+		i=$((i + 1))
+		params=$($LCTL list_param $param 2>/dev/null || true)
+	done
+
+	if ! do_rpc_nodes "$(facet_active_host $facet)" \
+			wait_import_state $expected "$params" $maxtime; then
 		error "import is not in ${expected} state"
 		return 1
 	fi
@@ -4928,6 +5664,21 @@ wait_osc_import_state() {
 	return 0
 }
 
+wait_osc_import_state() {
+	local facet=$1
+	local ost_facet=$2
+	local expected=$3
+	local num
+
+	if [[ $facet = mds ]]; then
+		for num in $(seq $MDSCOUNT); do
+			_wait_osc_import_state mds$num "$ost_facet" "$expected"
+		done
+	else
+		_wait_osc_import_state "$facet" "$ost_facet" "$expected"
+	fi
+}
+
 get_clientmdc_proc_path() {
     echo "${1}-mdc-*"
 }
@@ -4939,7 +5690,9 @@ do_rpc_nodes () {
 	[ -z "$list" ] && return 0
 
 	# Add paths to lustre tests for 32 and 64 bit systems.
-	local RPATH="PATH=$RLUSTRE/tests:/usr/lib/lustre/tests:/usr/lib64/lustre/tests:$PATH"
+	local LIBPATH="/usr/lib/lustre/tests:/usr/lib64/lustre/tests:"
+	local TESTPATH="$RLUSTRE/tests:"
+	local RPATH="PATH=${TESTPATH}${LIBPATH}${PATH}:/sbin:/bin:/usr/sbin:"
 	do_nodesv $list "${RPATH} NAME=${NAME} sh rpc.sh $@ "
 }
 
@@ -4965,7 +5718,7 @@ wait_clients_import_state () {
     local params=$(expand_list $params $proc_path)
     done
 
-	if ! do_rpc_nodes "$list" wait_import_state $expected $params; then
+	if ! do_rpc_nodes "$list" wait_import_state_mount $expected $params; then
 		error "import is not in ${expected} state"
 		return 1
 	fi
@@ -5477,19 +6230,20 @@ check_logdir() {
 }
 
 check_write_access() {
-    local dir=$1
-    local node
-    local file
+	local dir=$1
+	local list=${2:-$(comma_list $(nodes_list))}
+	local node
+	local file
 
-    for node in $(nodes_list); do
-        file=$dir/check_file.$(short_hostname $node)
-        if [[ ! -f "$file" ]]; then
-            # Logdir not accessible/writable from this node.
-            return 1
-        fi
-        rm -f $file || return 1
-    done
-    return 0
+	for node in ${list//,/ }; do
+		file=$dir/check_file.$(short_nodename $node)
+		if [[ ! -f "$file" ]]; then
+			# Logdir not accessible/writable from this node.
+			return 1
+		fi
+		rm -f $file || return 1
+	done
+	return 0
 }
 
 init_logging() {
@@ -5557,16 +6311,44 @@ run_llverfs()
 {
         local dir=$1
         local llverfs_opts=$2
+        local use_partial_arg=$3
         local partial_arg=""
         local size=$(df -B G $dir |tail -n 1 |awk '{print $2}' |sed 's/G//') #GB
 
         # Run in partial (fast) mode if the size
         # of a partition > 1 GB
-        [ $size -gt 1 ] && partial_arg="-p"
+        [ "x$use_partial_arg" != "xno" ] && [ $size -gt 1 ] && partial_arg="-p"
 
         llverfs $partial_arg $llverfs_opts $dir
 }
 
+#Remove objects from OST
+remove_ost_objects() {
+	local facet=$1
+	local ostdev=$2
+	local group=$3
+	shift 3
+	local objids="$@"
+	local mntpt=$(facet_mntpt $facet)
+	local opts=$OST_MOUNT_OPTS
+	local i
+	local rc
+
+	echo "removing objects from $ostdev on $facet: $objids"
+	if ! test -b $ostdev; then
+		opts=$(csa_add "$opts" -o loop)
+	fi
+	mount -t $(facet_fstype $facet) $opts $ostdev $mntpt ||
+		return $?
+	rc=0
+	for i in $objids; do
+		rm $mntpt/O/$group/d$((i % 32))/$i || { rc=$?; break; }
+	done
+	umount -f $mntpt || return $?
+	return $rc
+}
+
+#Remove files from MDT
 remove_mdt_files() {
 	local facet=$1
 	local mdtdev=$2
@@ -5577,12 +6359,12 @@ remove_mdt_files() {
 
 	echo "removing files from $mdtdev on $facet: $files"
 	if [ $(facet_fstype $facet) == ldiskfs ] &&
-	   ! do_facet $facet test -b ${!dev}; then
+	   ! do_facet $facet test -b $mdtdev; then
 		opts=$(csa_add "$opts" -o loop)
 	fi
 	mount -t $(facet_fstype $facet) $opts $mdtdev $mntpt ||
 		return $?
-	rc=0;
+	rc=0
 	for f in $files; do
 		rm $mntpt/ROOT/$f || { rc=$?; break; }
 	done
@@ -5601,7 +6383,7 @@ duplicate_mdt_files() {
 	echo "duplicating files on $mdtdev on $facet: $files"
 	mkdir -p $mntpt || return $?
 	if [ $(facet_fstype $facet) == ldiskfs ] &&
-	   ! do_facet $facet test -b ${!dev}; then
+	   ! do_facet $facet test -b $mdtdev; then
 		opts=$(csa_add "$opts" -o loop)
 	fi
 	mount -t $(facet_fstype $facet) $opts $mdtdev $mntpt ||
@@ -5670,6 +6452,45 @@ min_ost_size () {
     $LCTL get_param -n osc.*.kbytesavail | sort -n | head -n1
 }
 
+#
+# Get the available size (KB) of a given obd target.
+#
+get_obd_size() {
+	local facet=$1
+	local obd=$2
+	local size
+
+	[[ $facet != client ]] || return 0
+
+	size=$(do_facet $facet $LCTL get_param -n *.$obd.kbytesavail | head -n1)
+	echo -n $size
+}
+
+#
+# Get the page size (bytes) on a given facet node.
+#
+get_page_size() {
+	local facet=$1
+	local size
+
+	size=$(do_facet $facet getconf PAGE_SIZE)
+	[[ ${PIPESTATUS[0]} = 0 && -n "$size" ]] || size=4096
+	echo -n $size
+}
+
+#
+# Get the block count of the filesystem.
+#
+get_block_count() {
+	local facet=$1
+	local device=$2
+	local count
+
+	count=$(do_facet $facet "$DUMPE2FS -h $device 2>&1" |
+		awk '/^Block count:/ {print $3}')
+	echo -n $count
+}
+
 # Get the block size of the filesystem.
 get_block_size() {
     local facet=$1
@@ -5683,10 +6504,13 @@ get_block_size() {
 
 # Check whether the "large_xattr" feature is enabled or not.
 large_xattr_enabled() {
-    local mds_dev=$(mdsdevname ${SINGLEMDS//mds/})
+	[[ $(facet_fstype $SINGLEMDS) == zfs ]] && return 0
 
-    do_facet $SINGLEMDS "$DUMPE2FS -h $mds_dev 2>&1 | grep -q large_xattr"
-    return ${PIPESTATUS[0]}
+	local mds_dev=$(mdsdevname ${SINGLEMDS//mds/})
+
+	do_facet $SINGLEMDS "$DUMPE2FS -h $mds_dev 2>&1 |
+		grep -E -q '(ea_inode|large_xattr)'"
+	return ${PIPESTATUS[0]}
 }
 
 # Get the maximum xattr size supported by the filesystem.
@@ -5724,28 +6548,33 @@ generate_string() {
 }
 
 reformat_external_journal() {
+	local facet=$1
+
 	if [ ! -z ${EJOURNAL} ]; then
-		local rcmd="do_facet ${SINGLEMDS}"
+		local rcmd="do_facet $facet"
 
-		echo "reformat external journal on ${SINGLEMDS}:${EJOURNAL}"
+		echo "reformat external journal on $facet:${EJOURNAL}"
 		${rcmd} mke2fs -O journal_dev ${EJOURNAL} || return 1
 	fi
 }
 
 # MDT file-level backup/restore
 mds_backup_restore() {
-	local devname=$(mdsdevname ${SINGLEMDS//mds/})
+	local facet=$1
+	local igif=$2
+	local devname=$(mdsdevname $(facet_number $facet))
 	local mntpt=$(facet_mntpt brpt)
-	local rcmd="do_facet ${SINGLEMDS}"
+	local rcmd="do_facet $facet"
 	local metaea=${TMP}/backup_restore.ea
 	local metadata=${TMP}/backup_restore.tgz
 	local opts=${MDS_MOUNT_OPTS}
+	local svc=${facet}_svc
 
 	if ! ${rcmd} test -b ${devname}; then
 		opts=$(csa_add "$opts" -o loop)
 	fi
 
-	echo "file-level backup/restore on ${SINGLEMDS}:${devname}"
+	echo "file-level backup/restore on $facet:${devname}"
 
 	# step 1: build mount point
 	${rcmd} mkdir -p $mntpt
@@ -5753,6 +6582,10 @@ mds_backup_restore() {
 	${rcmd} rm -f $metaea $metadata
 	# step 3: mount dev
 	${rcmd} mount -t ldiskfs $opts $devname $mntpt || return 1
+	if [ ! -z $igif ]; then
+		# step 3.5: rm .lustre
+		${rcmd} rm -rf $mntpt/ROOT/.lustre || return 1
+	fi
 	# step 4: backup metaea
 	echo "backup EA"
 	${rcmd} "cd $mntpt && getfattr -R -d -m '.*' -P . > $metaea && cd -" ||
@@ -5763,11 +6596,12 @@ mds_backup_restore() {
 	# step 6: umount
 	${rcmd} umount -d $mntpt || return 4
 	# step 7: reformat external journal if needed
-	reformat_external_journal || return 5
+	reformat_external_journal $facet || return 5
 	# step 8: reformat dev
 	echo "reformat new device"
-	add ${SINGLEMDS} $(mkfs_opts ${SINGLEMDS}) --backfstype ldiskfs \
-		--reformat $devname > /dev/null || return 6
+	add $facet $(mkfs_opts $facet ${devname}) --backfstype ldiskfs \
+		--reformat ${devname} $(mdsvdevname $(facet_number $facet)) \
+		> /dev/null || exit 6
 	# step 9: mount dev
 	${rcmd} mount -t ldiskfs $opts $devname $mntpt || return 7
 	# step 10: restore metadata
@@ -5783,21 +6617,24 @@ mds_backup_restore() {
 	${rcmd} umount -d $mntpt || return 10
 	# step 14: cleanup tmp backup
 	${rcmd} rm -f $metaea $metadata
+	# step 15: reset device label - it's not virgin on
+	${rcmd} e2label $devname ${!svc}
 }
 
 # remove OI files
 mds_remove_ois() {
-	local devname=$(mdsdevname ${SINGLEMDS//mds/})
+	local facet=$1
+	local idx=$2
+	local devname=$(mdsdevname $(facet_number $facet))
 	local mntpt=$(facet_mntpt brpt)
-	local rcmd="do_facet ${SINGLEMDS}"
-	local idx=$1
+	local rcmd="do_facet $facet"
 	local opts=${MDS_MOUNT_OPTS}
 
 	if ! ${rcmd} test -b ${devname}; then
 		opts=$(csa_add "$opts" -o loop)
 	fi
 
-	echo "remove OI files: idx=${idx}"
+	echo "removing OI files on $facet: idx=${idx}"
 
 	# step 1: build mount point
 	${rcmd} mkdir -p $mntpt
@@ -5828,3 +6665,62 @@ generate_logname() {
 
 	echo "$TESTLOG_PREFIX.$TESTNAME.$logname.$(hostname -s).log"
 }
+
+# make directory on different MDTs
+test_mkdir() {
+	local option
+	local parent
+	local child
+	local path
+	local rc=0
+
+	case $# in
+		1) path=$1;;
+		2) option=$1
+		   path=$2;;
+		*) error "Only creating single directory is supported";;
+	esac
+
+	child=$(basename $path)
+	parent=$(dirname $path)
+
+	if [ "$option" == "-p" -a -d $parent/$child ]; then
+		return $rc
+	fi
+
+	if [ ! -d ${parent} ]; then
+		if [ "$option" == "-p" ]; then
+			mkdir -p ${parent}
+		else
+			return 1
+		fi
+	fi
+
+	if [ $MDSCOUNT -le 1 ]; then
+		mkdir $option $parent/$child || rc=$?
+	else
+		local mdt_idx=$($LFS getstripe -M $parent)
+		local test_num=$(echo $testnum | sed -e 's/[^0-9]*//g')
+
+		if [ "$mdt_idx" -ne 0 ]; then
+			mkdir $option $parent/$child || rc=$?
+		else
+			mdt_idx=$((test_num % MDSCOUNT))
+			echo "mkdir $mdt_idx for $parent/$child"
+			$LFS setdirstripe -i $mdt_idx $parent/$child || rc=$?
+		fi
+	fi
+	return $rc
+}
+
+# find the smallest and not in use file descriptor
+free_fd()
+{
+        local max_fd=$(ulimit -n)
+        local fd=3
+        while [[ $fd -le $max_fd && -e /proc/self/fd/$fd ]]; do
+                ((++fd))
+        done
+        [ $fd -lt $max_fd ] || error "finding free file descriptor failed"
+        echo $fd
+}