X-Git-Url: https://git.whamcloud.com/?p=fs%2Flustre-release.git;a=blobdiff_plain;f=lustre%2Ftests%2Ftest-framework.sh;h=04f9599afd29affedb127ddd7008460240eb4e0e;hp=033e32ac614351054f966d680149082368d88f7a;hb=c03ef784aa9b899da4461c6dd30223121a248cf4;hpb=7e0aeeec0bbff30956ce862546a1c71e322af5a9

diff --git a/lustre/tests/test-framework.sh b/lustre/tests/test-framework.sh
index 033e32a..04f9599 100755
--- a/lustre/tests/test-framework.sh
+++ b/lustre/tests/test-framework.sh
@@ -293,12 +293,17 @@ init_test_env() {
 	fi
 	export RSYNC_RSH=${RSYNC_RSH:-rsh}
 
+	export LNETCTL=${LNETCTL:-"$LUSTRE/../lnet/utils/lnetctl"}
+	[ ! -f "$LNETCTL" ] && export LNETCTL=$(which lnetctl 2> /dev/null)
 	export LCTL=${LCTL:-"$LUSTRE/utils/lctl"}
 	[ ! -f "$LCTL" ] && export LCTL=$(which lctl)
 	export LFS=${LFS:-"$LUSTRE/utils/lfs"}
 	[ ! -f "$LFS" ] && export LFS=$(which lfs)
+	export KSOCKLND_CONFIG=${KSOCKLND_CONFIG:-"$LUSTRE/scripts/ksocklnd-config"}
+	[ ! -f "$KSOCKLND_CONFIG" ] &&
+		export KSOCKLND_CONFIG=$(which ksocklnd-config 2> /dev/null)
 
-	export PERM_CMD=${PERM_CMD:-"$LCTL conf_param"}
+	export PERM_CMD=$(echo ${PERM_CMD:-"$LCTL conf_param"})
 
 	export L_GETIDENTITY=${L_GETIDENTITY:-"$LUSTRE/utils/l_getidentity"}
 	if [ ! -f "$L_GETIDENTITY" ]; then
@@ -426,14 +431,20 @@ init_test_env() {
 
 	# Constants used in more than one test script
 	export LOV_MAX_STRIPE_COUNT=2000
+	export DELETE_OLD_POOLS=${DELETE_OLD_POOLS:-false}
+	export KEEP_POOLS=${KEEP_POOLS:-false}
+	export PARALLEL=${PARALLEL:-"no"}
 
 	export MACHINEFILE=${MACHINEFILE:-$TMP/$(basename $0 .sh).machines}
 	. ${CONFIG:=$LUSTRE/tests/cfg/$NAME.sh}
 	get_lustre_env
 
 	# use localrecov to enable recovery for local clients, LU-12722
-	[[ $MDS1_VERSION -lt $(version_code 2.13.52) ]] ||
+	[[ $MDS1_VERSION -lt $(version_code 2.13.52) ]] || {
 		export MDS_MOUNT_OPTS=${MDS_MOUNT_OPTS:-"-o localrecov"}
+		export MGS_MOUNT_OPTS=${MGS_MOUNT_OPTS:-"-o localrecov"}
+	}
+
 	[[ $OST1_VERSION -lt $(version_code 2.13.52) ]] ||
 		export OST_MOUNT_OPTS=${OST_MOUNT_OPTS:-"-o localrecov"}
 }
@@ -531,6 +542,17 @@ module_loaded () {
 	/sbin/lsmod | grep -q "^\<$1\>"
 }
 
+check_lfs_df_ret_val() {
+	# Ignore only EOPNOTSUPP (which is 95; Operation not supported) error
+	# returned by 'lfs df' for valid dentry but not a lustrefs.
+	#
+	# 'lfs df' historically always returned success(0) instead of
+	# EOPNOTSUPP. This function for compatibility reason, ignores and
+	# masquerades EOPNOTSUPP as success.
+	[[ $1 -eq 95 ]] && return 0
+	return $1
+}
+
 PRLFS=false
 lustre_insmod() {
 	local module=$1
@@ -655,29 +677,20 @@ load_module() {
 	fi
 }
 
-load_modules_local() {
-	if [ -n "$MODPROBE" ]; then
-		# use modprobe
-		echo "Using modprobe to load modules"
-		return 0
-	fi
-
-	# Create special udev test rules on every node
-	if [ -f $LUSTRE/lustre/conf/99-lustre.rules ]; then {
-		sed -e 's|/usr/sbin/lctl|$LCTL|g' $LUSTRE/lustre/conf/99-lustre.rules > /etc/udev/rules.d/99-lustre-test.rules
-	} else {
-		echo "SUBSYSTEM==\"lustre\", ACTION==\"change\", ENV{PARAM}==\"?*\", RUN+=\"$LCTL set_param '\$env{PARAM}=\$env{SETTING}'\"" > /etc/udev/rules.d/99-lustre-test.rules
-	} fi
-	udevadm control --reload-rules
-	udevadm trigger
+do_lnetctl() {
+	$LCTL mark "$LNETCTL $*"
+	echo "$LNETCTL $*"
+	$LNETCTL "$@"
+}
 
+load_lnet() {
 	# For kmemleak-enabled kernels we need clear all past state
 	# that obviously has nothing to do with this Lustre run
 	# Disable automatic memory scanning to avoid perf hit.
 	if [ -f /sys/kernel/debug/kmemleak ] ; then
-		echo scan=off > /sys/kernel/debug/kmemleak
-		echo scan > /sys/kernel/debug/kmemleak
-		echo clear > /sys/kernel/debug/kmemleak
+		echo scan=off > /sys/kernel/debug/kmemleak || true
+		echo scan > /sys/kernel/debug/kmemleak || true
+		echo clear > /sys/kernel/debug/kmemleak || true
 	fi
 
 	echo Loading modules from $LUSTRE
@@ -714,24 +727,56 @@ load_modules_local() {
 	load_module ../libcfs/libcfs/libcfs
 	# Prevent local MODOPTS_LIBCFS being passed as part of environment
 	# variable to remote nodes
-	MODOPTS_LIBCFS=$saved_opts
+	unset MODOPTS_LIBCFS
 
-	set_default_debug
-	load_module ../lnet/lnet/lnet
+	set_default_debug "neterror net nettrace malloc"
+	if [ "$1" = "config_on_load=1" ]; then
+		load_module ../lnet/lnet/lnet
+	else
+		load_module ../lnet/lnet/lnet "$@"
+	fi
 
 	LNDPATH=${LNDPATH:-"../lnet/klnds"}
 	if [ -z "$LNETLND" ]; then
 		case $NETTYPE in
-		o2ib*)  LNETLND="o2iblnd/ko2iblnd" ;;
-		tcp*)   LNETLND="socklnd/ksocklnd" ;;
-		*)      local lnd="${NETTYPE%%[0-9]}lnd"
+		o2ib*)	LNETLND="o2iblnd/ko2iblnd" ;;
+		tcp*)	LNETLND="socklnd/ksocklnd" ;;
+		*)	local lnd="${NETTYPE%%[0-9]}lnd"
 			[ -f "$LNDPATH/$lnd/k$lnd.ko" ] &&
 				LNETLND="$lnd/k$lnd" ||
 				LNETLND="socklnd/ksocklnd"
 		esac
 	fi
 	load_module ../lnet/klnds/$LNETLND
+
+	if [ "$1" = "config_on_load=1" ]; then
+		do_lnetctl lnet configure --all ||
+			return $?
+	fi
+}
+
+load_modules_local() {
+	if [ -n "$MODPROBE" ]; then
+		# use modprobe
+		echo "Using modprobe to load modules"
+		return 0
+	fi
+
+	# Create special udev test rules on every node
+	if [ -f $LUSTRE/lustre/conf/99-lustre.rules ]; then {
+		sed -e 's|/usr/sbin/lctl|$LCTL|g' $LUSTRE/lustre/conf/99-lustre.rules > /etc/udev/rules.d/99-lustre-test.rules
+	} else {
+		echo "SUBSYSTEM==\"lustre\", ACTION==\"change\", ENV{PARAM}==\"?*\", RUN+=\"$LCTL set_param '\$env{PARAM}=\$env{SETTING}'\"" > /etc/udev/rules.d/99-lustre-test.rules
+	} fi
+	udevadm control --reload-rules
+	udevadm trigger
+
+	load_lnet
+
 	load_module obdclass/obdclass
+	if ! client_only; then
+		MODOPTS_PTLRPC=${MODOPTS_PTLRPC:-"lbug_on_grant_miscount=1"}
+	fi
 	load_module ptlrpc/ptlrpc
 	load_module ptlrpc/gss/ptlrpc_gss
 	load_module fld/fld
@@ -795,12 +840,24 @@ load_modules_local() {
 }
 
 load_modules () {
+	local facets
+	local facet
+	local failover
 	load_modules_local
 	# bug 19124
 	# load modules on remote nodes optionally
 	# lustre-tests have to be installed on these nodes
 	if $LOAD_MODULES_REMOTE; then
 		local list=$(comma_list $(remote_nodes_list))
+
+		# include failover nodes in case they are not in the list yet
+		facets=$(get_facets)
+		for facet in ${facets//,/ }; do
+			failover=$(facet_failover_host $facet)
+			[ -n "$list" ] && [[ ! "$list" =~ "$failover" ]] &&
+				list="$list,$failover"
+		done
+
 		if [ -n "$list" ]; then
 			echo "loading modules on: '$list'"
 			do_rpc_nodes "$list" load_modules_local
@@ -821,9 +878,7 @@ check_mem_leak () {
 	fi
 }
 
-unload_modules() {
-	wait_exit_ST client # bug 12845
-
+unload_modules_local() {
 	$LUSTRE_RMMOD ldiskfs || return 2
 
 	[ -f /etc/udev/rules.d/99-lustre-test.rules ] &&
@@ -831,15 +886,23 @@ unload_modules() {
 	udevadm control --reload-rules
 	udevadm trigger
 
+	check_mem_leak || return 254
+
+	return 0
+}
+
+unload_modules() {
+	local rc=0
+
+	wait_exit_ST client # bug 12845
+
+	unload_modules_local || rc=$?
+
 	if $LOAD_MODULES_REMOTE; then
 		local list=$(comma_list $(remote_nodes_list))
 		if [ -n "$list" ]; then
 			echo "unloading modules on: '$list'"
-			do_rpc_nodes "$list" $LUSTRE_RMMOD ldiskfs
-			do_rpc_nodes "$list" check_mem_leak
-			do_rpc_nodes "$list" "rm -f /etc/udev/rules.d/99-lustre-test.rules"
-			do_rpc_nodes "$list" "udevadm control --reload-rules"
-			do_rpc_nodes "$list" "udevadm trigger"
+			do_rpc_nodes "$list" unload_modules_local
 		fi
 	fi
 
@@ -850,10 +913,9 @@ unload_modules() {
 			rm -f $sbin_mount
 	fi
 
-	check_mem_leak || return 254
+	[[ $rc -eq 0 ]] && echo "modules unloaded."
 
-	echo "modules unloaded."
-	return 0
+	return $rc
 }
 
 fs_log_size() {
@@ -861,7 +923,7 @@ fs_log_size() {
 	local size=0
 
 	case $(facet_fstype $facet) in
-		ldiskfs) size=50;; # largest seen is 44, leave some headroom
+		ldiskfs) size=72;; # largest seen is 64, leave some headroom
 		# grant_block_size is in bytes, allow at least 2x max blocksize
 		zfs)     size=$(lctl get_param osc.$FSNAME*.import |
 				awk '/grant_block_size:/ {print $2/512; exit;}')
@@ -1311,20 +1373,6 @@ devicelabel() {
 	echo -n $label
 }
 
-mdsdevlabel() {
-	local num=$1
-	local device=$(mdsdevname $num)
-	local label=$(devicelabel mds$num ${device} | grep -v "CMD: ")
-	echo -n $label
-}
-
-ostdevlabel() {
-	local num=$1
-	local device=$(ostdevname $num)
-	local label=$(devicelabel ost$num ${device} | grep -v "CMD: ")
-	echo -n $label
-}
-
 #
 # Get the device of a facet.
 #
@@ -1557,8 +1605,7 @@ get_osd_param() {
 	local device=${2:-$FSNAME-OST*}
 	local name=$3
 
-	do_nodes $nodes "$LCTL get_param -n obdfilter.$device.$name \
-		osd-*.$device.$name 2>&1" | grep -v 'error:'
+	do_nodes $nodes "$LCTL get_param -n osd-*.$device.$name"
 }
 
 set_osd_param() {
@@ -1567,8 +1614,7 @@ set_osd_param() {
 	local name=$3
 	local value=$4
 
-	do_nodes $nodes "$LCTL set_param -n obdfilter.$device.$name=$value \
-		osd-*.$device.$name=$value 2>&1" | grep -v 'error:'
+	do_nodes $nodes "$LCTL set_param -n osd-*.$device.$name=$value"
 }
 
 set_debug_size () {
@@ -1656,9 +1702,17 @@ set_hostid () {
 mount_facets () {
 	local facets=${1:-$(get_facets)}
 	local facet
+	local -a mountpids
+	local total=0
+	local ret=0
 
 	for facet in ${facets//,/ }; do
-		mount_facet $facet
+		mount_facet $facet &
+		mountpids[total]=$!
+		total=$((total+1))
+	done
+	for ((index=0; index<$total; index++)); do
+		wait ${mountpids[index]}
 		local RC=$?
 		[ $RC -eq 0 ] && continue
 
@@ -1668,8 +1722,9 @@ mount_facets () {
 		else
 			error "Restart of $facet failed!"
 		fi
-		return $RC
+		ret=$RC
 	done
+	return $ret
 }
 
 #
@@ -2037,6 +2092,9 @@ mount_facet() {
 	local devicelabel
 	local dm_dev=${!dev}
 
+	[[ $dev == "mgsfailover_dev" ]] && combined_mgs_mds &&
+		dev=mds1failover_dev
+
 	module_loaded lustre || load_modules
 
 	case $fstype in
@@ -2129,11 +2187,17 @@ start() {
 	eval export ${dev_alias}_dev=${device}
 	eval export ${facet}_opt=\"$@\"
 
+	combined_mgs_mds && [[ ${dev_alias} == mds1 ]] &&
+		eval export mgs_dev=${device}
+
 	local varname=${dev_alias}failover_dev
 	if [ -n "${!varname}" ] ; then
 		eval export ${dev_alias}failover_dev=${!varname}
 	else
 		eval export ${dev_alias}failover_dev=$device
+		combined_mgs_mds && [[ ${dev_alias} == mds1 ]] &&
+			eval export mgsfailover_dev=${device}
+
 	fi
 
 	local mntpt=$(facet_mntpt $facet)
@@ -2196,7 +2260,7 @@ restore_quota() {
 	if [ "$old_MDT_QUOTA_TYPE" ]; then
 		if [[ $PERM_CMD == *"set_param -P"* ]]; then
 			do_facet mgs $PERM_CMD \
-				osd-*.$FSNAME-MDT*.quota_slave.enable = \
+				osd-*.$FSNAME-MDT*.quota_slave.enabled = \
 				$old_MDT_QUOTA_TYPE
 		else
 			do_facet mgs $PERM_CMD \
@@ -2206,7 +2270,7 @@ restore_quota() {
 	if [ "$old_OST_QUOTA_TYPE" ]; then
 		if [[ $PERM_CMD == *"set_param -P"* ]]; then
 			do_facet mgs $PERM_CMD \
-				osd-*.$FSNAME-OST*.quota_slave.enable = \
+				osd-*.$FSNAME-OST*.quota_slave.enabled = \
 				$old_OST_QUOTA_TYPE
 		else
 			do_facet mgs $LCTL conf_param \
@@ -2220,6 +2284,7 @@ restore_quota() {
 # This will allow fixing the "lfs df" summary line in the future.
 lfs_df() {
 	$LFS df $* | sed -e 's/filesystem /filesystem_/'
+	check_lfs_df_ret_val $?
 }
 
 # Get free inodes on the MDT specified by mdt index, free indoes on
@@ -2267,9 +2332,9 @@ setup_quota(){
 
 	if [[ $PERM_CMD == *"set_param -P"* ]]; then
 		do_facet mgs $PERM_CMD \
-			osd-*.$FSNAME-MDT*.quota_slave.enable=$QUOTA_TYPE
+			osd-*.$FSNAME-MDT*.quota_slave.enabled=$QUOTA_TYPE
 		do_facet mgs $PERM_CMD \
-			osd-*.$FSNAME-OST*.quota_slave.enable=$QUOTA_TYPE
+			osd-*.$FSNAME-OST*.quota_slave.enabled=$QUOTA_TYPE
 	else
 		do_facet mgs $PERM_CMD $FSNAME.quota.mdt=$QUOTA_TYPE ||
 			error "set mdt quota type failed"
@@ -2390,16 +2455,16 @@ zconf_umount() {
 
 # Mount the file system on the MDS
 mount_mds_client() {
-	local mds_HOST=${SINGLEMDS}_HOST
-	echo $mds_HOST
-	zconf_mount $mds1_HOST $MOUNT2 $MOUNT_OPTS ||
-		error "unable to mount $MOUNT2 on MDS"
+	local host=$(facet_active_host $SINGLEMDS)
+	echo $host
+	zconf_mount $host $MOUNT2 $MOUNT_OPTS ||
+		error "unable to mount $MOUNT2 on $host"
 }
 
 # Unmount the file system on the MDS
 umount_mds_client() {
-	local mds_HOST=${SINGLEMDS}_HOST
-	zconf_umount $mds1_HOST $MOUNT2
+	local host=$(facet_active_host $SINGLEMDS)
+	zconf_umount $host $MOUNT2
 	do_facet $SINGLEMDS "rmdir $MOUNT2"
 }
 
@@ -2717,11 +2782,12 @@ remount_facet() {
 reboot_facet() {
 	local facet=$1
 	local node=$(facet_active_host $facet)
+	local sleep_time=${2:-10}
 
 	if [ "$FAILURE_MODE" = HARD ]; then
 		boot_node $node
 	else
-		sleep 10
+		sleep $sleep_time
 	fi
 }
 
@@ -2989,35 +3055,6 @@ stop_client_loads() {
 }
 # End recovery-scale functions
 
-# verify that lustre actually cleaned up properly
-cleanup_check() {
-	VAR=$(lctl get_param -n catastrophe 2>&1)
-	if [ $? = 0 ] ; then
-		if [ $VAR != 0 ]; then
-			error "LBUG/LASSERT detected"
-		fi
-	fi
-	BUSY=$(dmesg | grep -i destruct || true)
-	if [ -n "$BUSY" ]; then
-		echo "$BUSY" 1>&2
-		[ -e $TMP/debug ] && mv $TMP/debug $TMP/debug-busy.$(date +%s)
-		exit 205
-	fi
-
-	check_mem_leak || exit 204
-
-	[[ $($LCTL dl 2>/dev/null | wc -l) -gt 0 ]] && $LCTL dl &&
-		echo "$TESTSUITE: lustre didn't clean up..." 1>&2 &&
-		return 202 || true
-
-	if module_loaded lnet || module_loaded libcfs; then
-		echo "$TESTSUITE: modules still loaded..." 1>&2
-		/sbin/lsmod 1>&2
-		return 203
-	fi
-	return 0
-}
-
 ##
 # wait for a command to return the expected result
 #
@@ -3032,15 +3069,21 @@ cleanup_check() {
 # If --verbose is passed as the first argument, the result is printed on each
 # value change, otherwise it is only printed after every 10s interval.
 #
+# If --quiet is passed as the first/second argument, the do_node() command
+# will not print the remote command before executing it each time.
+#
 # Using wait_update_cond() or related helper function is preferable to adding
 # a "long enough" wait for some state to change in the background, since
 # "long enough" may be too short due to tunables, system config, or running in
 # a VM, and must by necessity wait too long for most cases or risk failure.
 #
-# usage: wait_update_cond [--verbose] node check cond expect [max_wait]
+# usage: wait_update_cond [--verbose] [--quiet] node check cond expect [max_wait]
 wait_update_cond() {
-	local verbose=false
-	[[ "$1" == "--verbose" ]] && verbose=true && shift
+	local verbose
+	local quiet
+
+	[[ "$1" == "--verbose" ]] && verbose="$1" && shift
+	[[ "$1" == "--quiet" || "$1" == "-q" ]] && quiet="$1" && shift
 
 	local node=$1
 	local check="$2"
@@ -3055,7 +3098,7 @@ wait_update_cond() {
 	local print=10
 
 	while (( $waited <= $max_wait )); do
-		result=$(do_node $node "$check")
+		result=$(do_node $quiet $node "$check")
 
 		eval [[ "'$result'" $cond "'$expect'" ]]
 		if [[ $? == 0 ]]; then
@@ -3063,7 +3106,7 @@ wait_update_cond() {
 				echo "Updated after ${waited}s: want '$expect' got '$result'"
 			return 0
 		fi
-		if $verbose && [[ "$result" != "$prev_result" ]]; then
+		if [[ -n "$verbose" && "$result" != "$prev_result" ]]; then
 			[[ -n "$prev_result" ]] &&
 				echo "Changed after ${waited}s: from '$prev_result' to '$result'"
 			prev_result="$result"
@@ -3077,23 +3120,29 @@ wait_update_cond() {
 	return 3
 }
 
-# usage: wait_update [--verbose] node check expect [max_wait]
+# usage: wait_update [--verbose] [--quiet] node check expect [max_wait]
 wait_update() {
-	local verbose=
-	[ "$1" = "--verbose" ] && verbose="$1" && shift
+	local verbose
+	local quiet
+
+	[[ "$1" == "--verbose" ]] && verbose="$1" && shift
+	[[ "$1" == "--quiet" || "$1" == "-q" ]] && quiet="$1" && shift
 
 	local node="$1"
 	local check="$2"
 	local expect="$3"
 	local max_wait=$4
 
-	wait_update_cond $verbose $node "$check" "==" "$expect" $max_wait
+	wait_update_cond $verbose $quiet $node "$check" "==" "$expect" $max_wait
 }
 
 # usage: wait_update_facet_cond [--verbose] facet check cond expect [max_wait]
 wait_update_facet_cond() {
-	local verbose=
-	[ "$1" = "--verbose" ] && verbose="$1" && shift
+	local verbose
+	local quiet
+
+	[[ "$1" == "--verbose" ]] && verbose="$1" && shift
+	[[ "$1" == "--quiet" || "$1" == "-q" ]] && quiet="$1" && shift
 
 	local node=$(facet_active_host $1)
 	local check="$2"
@@ -3101,20 +3150,23 @@ wait_update_facet_cond() {
 	local expect="$4"
 	local max_wait=$5
 
-	wait_update_cond $verbose $node "$check" "$cond" "$expect" $max_wait
+	wait_update_cond $verbose $quiet $node "$check" "$cond" "$expect" $max_wait
 }
 
 # usage: wait_update_facet [--verbose] facet check expect [max_wait]
 wait_update_facet() {
-	local verbose=
-	[ "$1" = "--verbose" ] && verbose="$1" && shift
+	local verbose
+	local quiet
+
+	[[ "$1" == "--verbose" ]] && verbose="$1" && shift
+	[[ "$1" == "--quiet" || "$1" == "-q" ]] && quiet="$1" && shift
 
 	local node=$(facet_active_host $1)
 	local check="$2"
 	local expect="$3"
 	local max_wait=$4
 
-	wait_update_cond $verbose $node "$check" "==" "$expect" $max_wait
+	wait_update_cond $verbose $quiet $node "$check" "==" "$expect" $max_wait
 }
 
 sync_all_data() {
@@ -3378,8 +3430,11 @@ wait_recovery_complete () {
     fi
     echo affected facets: $facets
 
-	# we can use "for" here because we are waiting the slowest
-	for facet in ${facets//,/ }; do
+	facets=${facets//,/ }
+	# We can use "for" here because we are waiting the slowest.
+	# The mgs not having the recovery_status proc entry, exclude it
+	# from the facet list.
+	for facet in ${facets//mgs/ }; do
 		local var_svc=${facet}_svc
 		local param="*.${!var_svc}.recovery_status"
 
@@ -3450,14 +3505,15 @@ wait_osts_up() {
 }
 
 wait_destroy_complete () {
-	echo "Waiting for local destroys to complete"
+	echo "Waiting for MDT destroys to complete"
 	# MAX value shouldn't be big as this mean server responsiveness
 	# never increase this just to make test pass but investigate
 	# why it takes so long time
-	local MAX=5
+	local MAX=${1:-5}
 	local WAIT=0
+	local list=$(comma_list $(mdts_nodes))
 	while [ $WAIT -lt $MAX ]; do
-		local -a RPCs=($($LCTL get_param -n osc.*.destroys_in_flight))
+		local -a RPCs=($(do_nodes $list $LCTL get_param -n osp.*.destroys_in_flight))
 		local con=1
 		local i
 
@@ -3472,7 +3528,7 @@ wait_destroy_complete () {
 		echo "Waiting ${WAIT}s for local destroys to complete"
 		WAIT=$((WAIT + 1))
 	done
-	echo "Local destroys weren't done in $MAX sec."
+	echo "MDT destroys weren't done in $MAX sec."
 	return 1
 }
 
@@ -3537,12 +3593,15 @@ wait_remote_prog () {
 
 lfs_df_check() {
 	local clients=${1:-$CLIENTS}
+	local rc=0
 
 	if [ -z "$clients" ]; then
-		$LFS df $MOUNT
+		$LFS df $MOUNT > /dev/null || rc=$?
 	else
-		$PDSH $clients "$LFS df $MOUNT" > /dev/null
+		$PDSH $clients "$LFS df $MOUNT" > /dev/null || rc=$?
 	fi
+
+	check_lfs_df_ret_val $rc
 }
 
 clients_up() {
@@ -3551,6 +3610,21 @@ clients_up() {
 	lfs_df_check
 }
 
+all_mds_up() {
+	(( MDSCOUNT == 1 )) && return
+
+	# wait so that statfs data on MDT expire
+	local delay=$(do_facet mds1 $LCTL \
+		get_param -n osp.*MDT*MDT0000.maxage | sort -n | tail -1)
+
+	[ -n "$delay" ] || error "fail to get maxage"
+	sleep $delay
+	local nodes=$(comma_list $(mdts_nodes))
+	# initiate statfs RPC, all to all MDTs
+	do_nodes $nodes $LCTL get_param -N osp.*MDT*MDT*.filesfree >&/dev/null
+	do_nodes $nodes $LCTL get_param -N osp.*MDT*MDT*.filesfree >&/dev/null
+}
+
 client_up() {
 	# usually checked on particular client or locally
 	sleep 1
@@ -3639,11 +3713,13 @@ facet_failover() {
 	$E2FSCK_ON_MDT0 && (run_e2fsck $(facet_active_host $SINGLEMDS) \
 		$(mdsdevname 1) "-n" || error "Running e2fsck")
 
+	local -a mountpids
+
 	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 $sleep_time
 
 		change_active ${affecteds[index]}
 
@@ -3656,11 +3732,20 @@ facet_failover() {
 		if ! combined_mgs_mds &&
 			list_member ${affecteds[index]} mgs; then
 			mount_facet mgs || error "Restart of mgs failed"
+			affecteds[index]=$(exclude_items_from_list \
+				${affecteds[index]} mgs)
+		fi
+		if [ -n "${affecteds[index]}" ]; then
+			echo mount facets: ${affecteds[index]}
+			mount_facets ${affecteds[index]} &
+			mountpids[index]=$!
+		fi
+	done
+	for ((index=0; index<$total; index++)); do
+		if [ -n "${affecteds[index]}" ]; then
+			wait ${mountpids[index]}
 		fi
-		# FIXME; has to be changed to mount all facets concurrently
-		affected=$(exclude_items_from_list ${affecteds[index]} mgs)
-		echo mount facets: ${affecteds[index]}
-		mount_facets ${affecteds[index]}
+
 		if $GSS_SK; then
 			do_nodes $(comma_list $(all_nodes)) \
 				"keyctl show | grep lustre | cut -c1-11 |
@@ -3670,10 +3755,6 @@ facet_failover() {
 	done
 }
 
-obd_name() {
-    local facet=$1
-}
-
 replay_barrier() {
 	local facet=$1
 	do_facet $facet "sync; sync; sync"
@@ -3788,6 +3869,7 @@ fail_abort() {
 	mount_facet $facet -o $abort_type
 	clients_up || echo "first stat failed: $?"
 	clients_up || error "post-failover stat: $?"
+	all_mds_up
 }
 
 host_nids_address() {
@@ -3947,7 +4029,14 @@ facet_host() {
 		elif [ "${facet:0:3}" == "mdt" -o \
 			"${facet:0:3}" == "mds" -o \
 			"${facet:0:3}" == "mgs" ]; then
-			eval export ${facet}_HOST=${mds_HOST}
+			local temp
+			if [ "${facet}" == "mgsfailover" ] &&
+			   [ -n "$mds1failover_HOST" ]; then
+				temp=$mds1failover_HOST
+			else
+				temp=${mds_HOST}
+			fi
+			eval export ${facet}_HOST=$temp
 		fi
 	fi
 	echo -n ${!varname}
@@ -3963,6 +4052,13 @@ facet_failover_host() {
 		return
 	fi
 
+	if combined_mgs_mds && [ $facet == "mgs" ] &&
+		[ -z $mds1failover_HOST ]; then
+		temp=mds1failover_HOST
+		echo ${!temp}
+		return
+	fi
+
 	if [ "${facet:0:3}" == "mdt" -o "${facet:0:3}" == "mds" -o \
 	     "${facet:0:3}" == "mgs" ]; then
 
@@ -4020,8 +4116,6 @@ change_active() {
     local facetlist=$1
     local facet
 
-    facetlist=$(exclude_items_from_list $facetlist mgs)
-
     for facet in ${facetlist//,/ }; do
     local failover=${facet}failover
     local host=`facet_host $failover`
@@ -4044,26 +4138,27 @@ change_active() {
 }
 
 do_node() {
-    local verbose=false
-    # do not stripe off hostname if verbose, bug 19215
-    if [ x$1 = x--verbose ]; then
-        shift
-        verbose=true
-    fi
+	local verbose
+	local quiet
 
-    local HOST=$1
-    shift
-    local myPDSH=$PDSH
-    if [ "$HOST" = "$HOSTNAME" ]; then
-        myPDSH="no_dsh"
-    elif [ -z "$myPDSH" -o "$myPDSH" = "no_dsh" ]; then
-        echo "cannot run remote command on $HOST with $myPDSH"
-        return 128
-    fi
-    if $VERBOSE; then
-        echo "CMD: $HOST $@" >&2
-        $myPDSH $HOST "$LCTL mark \"$@\"" > /dev/null 2>&1 || :
-    fi
+	# do not strip off hostname if verbose, b=19215
+	[[ "$1" == "--verbose" ]] && verbose="$1" && shift
+	[[ "$1" == "--quiet" || "$1" == "-q" ]] && quiet="$1" && shift
+
+	local HOST=$1
+	shift
+	local myPDSH=$PDSH
+
+	if [ "$HOST" = "$HOSTNAME" ]; then
+		myPDSH="no_dsh"
+	elif [ -z "$myPDSH" -o "$myPDSH" = "no_dsh" ]; then
+		echo "cannot run remote command on $HOST with $myPDSH"
+		return 128
+	fi
+	if $VERBOSE && [[ -z "$quiet" ]]; then
+		echo "CMD: $HOST $@" >&2
+		$myPDSH $HOST "$LCTL mark \"$@\"" > /dev/null 2>&1 || :
+	fi
 
 	if [[ "$myPDSH" == "rsh" ]] ||
 	   [[ "$myPDSH" == *pdsh* && "$myPDSH" != *-S* ]]; then
@@ -4074,31 +4169,35 @@ do_node() {
 		eval $myPDSH $HOST "(PATH=\$PATH:$RLUSTRE/utils:$RLUSTRE/tests;
 				     PATH=\$PATH:/sbin:/usr/sbin;
 				     cd $RPWD;
-				     LUSTRE=\"$RLUSTRE\" sh -c \"$@\") ||
+				     LUSTRE=\"$RLUSTRE\" bash -c \"$@\") ||
 				     echo command failed >$command_status"
 		[[ -n "$($myPDSH $HOST cat $command_status)" ]] && return 1 ||
 			return 0
 	fi
 
-    if $verbose ; then
-        # print HOSTNAME for myPDSH="no_dsh"
-        if [[ $myPDSH = no_dsh ]]; then
-            $myPDSH $HOST "(PATH=\$PATH:$RLUSTRE/utils:$RLUSTRE/tests:/sbin:/usr/sbin; cd $RPWD; LUSTRE=\"$RLUSTRE\" sh -c \"$@\")" | sed -e "s/^/${HOSTNAME}: /"
-        else
-            $myPDSH $HOST "(PATH=\$PATH:$RLUSTRE/utils:$RLUSTRE/tests:/sbin:/usr/sbin; cd $RPWD; LUSTRE=\"$RLUSTRE\" sh -c \"$@\")"
-        fi
-    else
-        $myPDSH $HOST "(PATH=\$PATH:$RLUSTRE/utils:$RLUSTRE/tests:/sbin:/usr/sbin; cd $RPWD; LUSTRE=\"$RLUSTRE\" sh -c \"$@\")" | sed "s/^${HOST}: //"
-    fi
-    return ${PIPESTATUS[0]}
-}
-
-do_nodev() {
-    do_node --verbose "$@"
+	if [[ -n "$verbose" ]]; then
+		# print HOSTNAME for myPDSH="no_dsh"
+		if [[ $myPDSH = no_dsh ]]; then
+			$myPDSH $HOST \
+			"(PATH=\$PATH:$RLUSTRE/utils:$RLUSTRE/tests:/sbin:/usr/sbin;\
+			cd $RPWD; LUSTRE=\"$RLUSTRE\" bash -c \"$@\")" |
+			sed -e "s/^/${HOSTNAME}: /"
+		else
+			$myPDSH $HOST \
+			"(PATH=\$PATH:$RLUSTRE/utils:$RLUSTRE/tests:/sbin:/usr/sbin;\
+			cd $RPWD; LUSTRE=\"$RLUSTRE\" bash -c \"$@\")"
+		fi
+	else
+		$myPDSH $HOST \
+		"(PATH=\$PATH:$RLUSTRE/utils:$RLUSTRE/tests:/sbin:/usr/sbin;\
+		cd $RPWD; LUSTRE=\"$RLUSTRE\" bash -c \"$@\")" |
+		sed "s/^${HOST}: //"
+	fi
+	return ${PIPESTATUS[0]}
 }
 
 single_local_node () {
-   [ "$1" = "$HOSTNAME" ]
+	[ "$1" = "$HOSTNAME" ]
 }
 
 # Outputs environment variable assignments that should be passed to remote nodes
@@ -4144,45 +4243,42 @@ get_env_vars() {
 }
 
 do_nodes() {
-    local verbose=false
-    # do not stripe off hostname if verbose, bug 19215
-    if [ x$1 = x--verbose ]; then
-        shift
-        verbose=true
-    fi
+	local verbose
+	local quiet
 
-    local rnodes=$1
-    shift
+	# do not strip off hostname if verbose, b=19215
+	[[ "$1" == "--verbose" ]] && verbose="$1" && shift
+	[[ "$1" == "--quiet" || "$1" == "-q" ]] && quiet="$1" && shift
 
-    if single_local_node $rnodes; then
-        if $verbose; then
-           do_nodev $rnodes "$@"
-        else
-           do_node $rnodes "$@"
-        fi
-        return $?
-    fi
+	local rnodes=$1
+	shift
 
-    # This is part from do_node
-    local myPDSH=$PDSH
+	if single_local_node $rnodes; then
+		do_node $verbose $quiet $rnodes "$@"
+		return $?
+	fi
 
-    [ -z "$myPDSH" -o "$myPDSH" = "no_dsh" -o "$myPDSH" = "rsh" ] && \
-        echo "cannot run remote command on $rnodes with $myPDSH" && return 128
+	# This is part from do_node
+	local myPDSH=$PDSH
 
-    export FANOUT=$(get_node_count "${rnodes//,/ }")
-    if $VERBOSE; then
-        echo "CMD: $rnodes $@" >&2
-        $myPDSH $rnodes "$LCTL mark \"$@\"" > /dev/null 2>&1 || :
-    fi
+	[ -z "$myPDSH" -o "$myPDSH" = "no_dsh" -o "$myPDSH" = "rsh" ] &&
+		echo "cannot run remote command on $rnodes with $myPDSH" &&
+		return 128
 
-    # do not replace anything from pdsh output if -N is used
-    # -N     Disable hostname: prefix on lines of output.
-    if $verbose || [[ $myPDSH = *-N* ]]; then
-        $myPDSH $rnodes "(PATH=\$PATH:$RLUSTRE/utils:$RLUSTRE/tests:/sbin:/usr/sbin; cd $RPWD; LUSTRE=\"$RLUSTRE\" $(get_env_vars) sh -c \"$@\")"
-    else
-        $myPDSH $rnodes "(PATH=\$PATH:$RLUSTRE/utils:$RLUSTRE/tests:/sbin:/usr/sbin; cd $RPWD; LUSTRE=\"$RLUSTRE\" $(get_env_vars) sh -c \"$@\")" | sed -re "s/^[^:]*: //g"
-    fi
-    return ${PIPESTATUS[0]}
+	export FANOUT=$(get_node_count "${rnodes//,/ }")
+	if $VERBOSE && [[ -z "$quiet" ]]; then
+		echo "CMD: $rnodes $@" >&2
+		$myPDSH $rnodes "$LCTL mark \"$@\"" > /dev/null 2>&1 || :
+	fi
+
+	# do not replace anything from pdsh output if -N is used
+	# -N     Disable hostname: prefix on lines of output.
+	if [[ -n "$verbose" || $myPDSH = *-N* ]]; then
+		$myPDSH $rnodes "(PATH=\$PATH:$RLUSTRE/utils:$RLUSTRE/tests:/sbin:/usr/sbin; cd $RPWD; LUSTRE=\"$RLUSTRE\" $(get_env_vars) bash -c \"$@\")"
+	else
+		$myPDSH $rnodes "(PATH=\$PATH:$RLUSTRE/utils:$RLUSTRE/tests:/sbin:/usr/sbin; cd $RPWD; LUSTRE=\"$RLUSTRE\" $(get_env_vars) bash -c \"$@\")" | sed -re "s/^[^:]*: //g"
+	fi
+	return ${PIPESTATUS[0]}
 }
 
 ##
@@ -4193,11 +4289,18 @@ do_nodes() {
 #
 # usage: do_facet $facet command [arg ...]
 do_facet() {
+	local verbose
+	local quiet
+
+	[[ "$1" == "--verbose" ]] && verbose="$1" && shift
+	[[ "$1" == "--quiet" || "$1" == "-q" ]] && quiet="$1" && shift
+
 	local facet=$1
 	shift
-	local HOST=$(facet_active_host $facet)
-	[ -z $HOST ] && echo "No host defined for facet ${facet}" && exit 1
-	do_node $HOST "$@"
+	local host=$(facet_active_host $facet)
+
+	[ -z "$host" ] && echo "No host defined for facet ${facet}" && exit 1
+	do_node $verbose $quiet $host "$@"
 }
 
 # Function: do_facet_random_file $FACET $FILE $SIZE
@@ -4220,7 +4323,7 @@ do_facet_create_file() {
 }
 
 do_nodesv() {
-    do_nodes --verbose "$@"
+	do_nodes --verbose "$@"
 }
 
 add() {
@@ -4230,6 +4333,12 @@ add() {
 	stop ${facet} -f
 	rm -f $TMP/${facet}active
 	[[ $facet = mds1 ]] && combined_mgs_mds && rm -f $TMP/mgsactive
+
+	# make sure in-tree ldiskfs is loaded before mkfs
+	if local_mode && [[ $(node_fstypes $HOSTNAME) == *ldiskfs* ]]; then
+		load_module ../ldiskfs/ldiskfs
+	fi
+
 	do_facet ${facet} $MKFS $* || return ${PIPESTATUS[0]}
 
 	if [[ $(facet_fstype $facet) == zfs ]]; then
@@ -4643,14 +4752,11 @@ mkfs_opts() {
 	opts+=${LDLM_TIMEOUT:+" --param=sys.ldlm_timeout=$LDLM_TIMEOUT"}
 
 	if [ $type == MDS ]; then
-		opts+=${MDSCAPA:+" --param-mdt.capa=$MDSCAPA"}
 		opts+=${DEF_STRIPE_SIZE:+" --param=lov.stripesize=$DEF_STRIPE_SIZE"}
 		opts+=${DEF_STRIPE_COUNT:+" --param=lov.stripecount=$DEF_STRIPE_COUNT"}
 		opts+=${L_GETIDENTITY:+" --param=mdt.identity_upcall=$L_GETIDENTITY"}
 
 		if [ $fstype == ldiskfs ]; then
-			fs_mkfs_opts+="-O ea_inode,large_dir"
-
 			var=${facet}_JRN
 			if [ -n "${!var}" ]; then
 				fs_mkfs_opts+=" -J device=${!var}"
@@ -4662,8 +4768,6 @@ mkfs_opts() {
 	fi
 
 	if [ $type == OST ]; then
-		opts+=${OSSCAPA:+" --param=ost.capa=$OSSCAPA"}
-
 		if [ $fstype == ldiskfs ]; then
 			var=${facet}_JRN
 			if [ -n "${!var}" ]; then
@@ -5131,7 +5235,14 @@ init_facet_vars () {
 
 	local varname=${facet}failover_HOST
 	if [ -z "${!varname}" ]; then
-		eval export $varname=$(facet_host $facet)
+		local temp
+		if combined_mgs_mds && [ $facet == "mgs" ] &&
+		   [ -n "$mds1failover_HOST" ]; then
+			temp=$mds1failover_HOST
+		else
+			temp=$(facet_host $facet)
+		fi
+		eval export $varname=$temp
 	fi
 
 	varname=${facet}_HOST
@@ -5164,12 +5275,12 @@ init_facets_vars () {
 
 	if ! remote_mds_nodsh; then
 		for num in $(seq $MDSCOUNT); do
-			DEVNAME=`mdsdevname $num`
+			DEVNAME=$(mdsdevname $num)
 			init_facet_vars mds$num $DEVNAME $MDS_MOUNT_OPTS
 		done
 	fi
 
-	combined_mgs_mds || init_facet_vars mgs $(mgsdevname) $MGS_MOUNT_OPTS
+	init_facet_vars mgs $(mgsdevname) $MGS_MOUNT_OPTS
 
 	if ! remote_ost_nodsh; then
 		for num in $(seq $OSTCOUNT); do
@@ -5298,6 +5409,11 @@ init_param_vars () {
 	TIMEOUT=$(do_facet $SINGLEMDS "lctl get_param -n timeout")
 	log "Using TIMEOUT=$TIMEOUT"
 
+	# tune down to speed up testing on (usually) small setups
+	local mgc_timeout=/sys/module/mgc/parameters/mgc_requeue_timeout_min
+	do_nodes $(comma_list $(nodes_list)) \
+		"[ -f $mgc_timeout ] && echo 1 > $mgc_timeout; exit 0"
+
 	osc_ensure_active $SINGLEMDS $TIMEOUT
 	osc_ensure_active client $TIMEOUT
 	$LCTL set_param osc.*.idle_timeout=debug
@@ -5328,8 +5444,7 @@ init_param_vars () {
 	fi
 
 	(( MDS1_VERSION <= $(version_code 2.13.52) )) ||
-		do_nodes $(comma_list $(mdts_nodes)) \
-			"$LCTL set_param lod.*.mdt_hash=crush"
+		do_facet mgs "$LCTL set_param -P lod.*.mdt_hash=crush"
 	return 0
 }
 
@@ -5417,25 +5532,103 @@ is_mounted () {
 	echo $mounted' ' | grep -w -q $mntpt' '
 }
 
-is_empty_dir() {
-	[ $(find $1 -maxdepth 1 -print | wc -l) = 1 ] && return 0
-	return 1
+create_pools () {
+	local pool=$1
+	local ostsn=${2:-$OSTCOUNT}
+	local npools=${FS_NPOOLS:-$((OSTCOUNT / ostsn))}
+	local n
+
+	echo ostsn=$ostsn npools=$npools
+	if [[ $ostsn -gt $OSTCOUNT ]];  then
+		echo "request to use $ostsn OSTs in the pool, \
+			using max available OSTCOUNT=$OSTCOUNT"
+		ostsn=$OSTCOUNT
+	fi
+	for (( n=0; n < $npools; n++ )); do
+		p=${pool}$n
+		if ! $DELETE_OLD_POOLS; then
+			log "request to not delete old pools: $FSNAME.$p exist?"
+			if ! check_pool_not_exist $FSNAME.$p; then
+				echo "Using existing $FSNAME.$p"
+				$LCTL pool_list $FSNAME.$p
+				continue
+			fi
+		fi
+		create_pool $FSNAME.$p $KEEP_POOLS ||
+			error "create_pool $FSNAME.$p failed"
+
+		local first=$(( (n * ostsn) % OSTCOUNT ))
+		local last=$(( (first + ostsn - 1) % OSTCOUNT ))
+		if [[ $first -le $last ]]; then
+			pool_add_targets $p $first $last ||
+				error "pool_add_targets $p $first $last failed"
+		else
+			pool_add_targets $p $first $(( OSTCOUNT - 1 )) ||
+				error "pool_add_targets $p $first \
+					$(( OSTCOUNT - 1 )) failed"
+			pool_add_targets $p 0 $last ||
+				error "pool_add_targets $p 0 $last failed"
+		fi
+	done
 }
 
-# empty lustre filesystem may have empty directories lost+found and .lustre
-is_empty_fs() {
-	# 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
+set_pools_quota () {
+	local u
+	local o
+	local p
+	local i
+	local j
+
+	[[ $ENABLE_QUOTA ]] || error "Required Pool Quotas: \
+		$POOLS_QUOTA_USERS_SET, but ENABLE_QUOTA not set!"
+
+	# POOLS_QUOTA_USERS_SET=
+	#              "quota15_1:20M          -- for all of the found pools
+	#               quota15_2:1G:gpool0
+	#               quota15_3              -- for global limit only
+	#               quota15_4:200M:gpool0
+	#               quota15_4:200M:gpool1"
+
+	declare -a pq_userset=(${POOLS_QUOTA_USERS_SET="mpiuser"})
+	declare -a pq_users
+	declare -A pq_limits
+
+	for ((i=0; i<${#pq_userset[@]}; i++)); do
+		u=${pq_userset[i]%%:*}
+		o=""
+		# user gets no pool limits if
+		# POOLS_QUOTA_USERS_SET does not specify it
+		[[ ${pq_userset[i]} =~ : ]] && o=${pq_userset[i]##$u:}
+		pq_limits[$u]+=" $o"
+	done
+	pq_users=(${!pq_limits[@]})
+
+	declare -a opts
+	local pool
+
+	for ((i=0; i<${#pq_users[@]}; i++)); do
+		u=${pq_users[i]}
+		# set to max limit (_u64)
+		$LFS setquota -u $u -B $((2**24 - 1))T $DIR
+		opts=(${pq_limits[$u]})
+		for ((j=0; j<${#opts[@]}; j++)); do
+			p=${opts[j]##*:}
+			o=${opts[j]%%:*}
+			# Set limit for all existing pools if
+			# no pool specified
+			if [ $p == $o ];  then
+				p=$(list_pool $FSNAME | sed "s/$FSNAME.//")
+				echo "No pool specified for $u,
+					set limit $o for all existing pools"
+			fi
+			for pool in $p; do
+				$LFS setquota -u $u -B $o --pool $pool $DIR ||
+                                        error "setquota -u $u -B $o \
+						--pool $pool failed"
+			done
+		done
+		$LFS quota -uv $u --pool  $DIR
+	done
 }
 
 check_and_setup_lustre() {
@@ -5523,6 +5716,16 @@ check_and_setup_lustre() {
 		set_flavor_all $SEC
 	fi
 
+	if $DELETE_OLD_POOLS; then
+		destroy_all_pools
+	fi
+	if [[ -n "$FS_POOL" ]]; then
+		create_pools $FS_POOL $FS_POOL_NOSTS
+	fi
+
+	if [[ -n "$POOLS_QUOTA_USERS_SET" ]]; then
+		set_pools_quota
+	fi
 	if [ "$ONLY" == "setup" ]; then
 		exit 0
 	fi
@@ -5553,51 +5756,6 @@ cleanup_and_setup_lustre() {
     check_and_setup_lustre
 }
 
-# Get all of the server target devices from a given server node and type.
-get_mnt_devs() {
-	local node=$1
-	local type=$2
-	local devs
-	local dev
-
-	if [ "$type" == ost ]; then
-		devs=$(get_osd_param $node "" mntdev)
-	else
-		devs=$(do_node $node $LCTL get_param -n osd-*.$FSNAME-M*.mntdev)
-	fi
-	for dev in $devs; do
-		case $dev in
-		*loop*) do_node $node "losetup $dev" | \
-				sed -e "s/.*(//" -e "s/).*//" ;;
-		*) echo $dev ;;
-		esac
-	done
-}
-
-# Get all of the server target devices.
-get_svr_devs() {
-	local node
-	local i
-
-	# Master MDS parameters used by lfsck
-	MDTNODE=$(facet_active_host $SINGLEMDS)
-	MDTDEV=$(echo $(get_mnt_devs $MDTNODE mdt) | awk '{print $1}')
-
-	# MDT devices
-	i=0
-	for node in $(mdts_nodes); do
-		MDTDEVS[i]=$(get_mnt_devs $node mdt)
-		i=$((i + 1))
-	done
-
-	# OST devices
-	i=0
-	for node in $(osts_nodes); do
-		OSTDEVS[i]=$(get_mnt_devs $node ost)
-		i=$((i + 1))
-	done
-}
-
 # Run e2fsck on MDT or OST device.
 run_e2fsck() {
 	local node=$1
@@ -5769,30 +5927,30 @@ check_and_cleanup_lustre() {
 # General functions
 
 wait_for_function () {
-    local quiet=""
-
-    # suppress fn both stderr and stdout
-    if [ "$1" = "--quiet" ]; then
-        shift
-        quiet=" > /dev/null 2>&1"
+	local quiet=""
 
-    fi
+	# suppress fn both stderr and stdout
+	if [ "$1" = "--quiet" ]; then
+		shift
+		quiet=" > /dev/null 2>&1"
+	fi
 
-    local fn=$1
-    local max=${2:-900}
-    local sleep=${3:-5}
+	local fn=$1
+	local max=${2:-900}
+	local sleep=${3:-5}
 
-    local wait=0
+	local wait=0
 
-    while true; do
+	while true; do
 
-        eval $fn $quiet && return 0
+		eval $fn $quiet && return 0
 
-        wait=$((wait + sleep))
-        [ $wait -lt $max ] || return 1
-        echo waiting $fn, $((max - wait)) secs left ...
-        sleep $sleep
-    done
+		[ $wait -lt $max ] || return 1
+		echo waiting $fn, $((max - wait)) secs left ...
+		wait=$((wait + sleep))
+		[ $wait -gt $max ] && ((sleep -= wait - max))
+		sleep $sleep
+	done
 }
 
 check_network() {
@@ -5923,10 +6081,6 @@ at_max_get() {
     at_get $1 at_max
 }
 
-at_min_get() {
-	at_get $1 at_min
-}
-
 at_max_set() {
     local at_max=$1
     shift
@@ -6332,6 +6486,37 @@ skip() {
 	exit 0
 }
 
+#
+# For interop testing treate EOPNOTSUPP as success
+# and skip
+#
+skip_eopnotsupp() {
+	local retstr=$@
+
+	echo $retstr | awk -F'|' '{print $1}' |
+		grep -E unsupported\|"(Operation not supported)"
+	(( $? == 0 )) || error "$retstr"
+	skip $retstr
+}
+
+# Add a list of tests to ALWAYS_EXCEPT due to an issue.
+# Usage: always_except LU-4815 23 42q ...
+#
+function always_except() {
+	local issue="${1:-}" # single jira style issue ("LU-4815")
+	local test_num
+
+	shift
+
+	if ! [[ "$issue" =~ ^[[:upper:]]+-[[:digit:]]+$ ]]; then
+		error "always_except: invalid issue '$issue' for tests '$*'"
+	fi
+
+	for test_num in "$@"; do
+		ALWAYS_EXCEPT+=" $test_num"
+	done
+}
+
 build_test_filter() {
 	EXCEPT="$EXCEPT $(testslist_filter)"
 
@@ -6516,10 +6701,10 @@ 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 --quiet $(comma_list $(nodes_list)) \
+		"lctl set_param -n fail_loc=0 fail_val=0 2>/dev/null" || true
+	#echo done.
 }
 
 
@@ -6528,7 +6713,8 @@ reset_fail_loc () {
 # Also appends a timestamp and prepends the testsuite name.
 #
 
-EQUALS="===================================================================================================="
+# ======================================================== 15:06:12 (1624050372)
+EQUALS="========================================================"
 banner() {
     msg="== ${TESTSUITE} $*"
     last=${msg: -1:1}
@@ -6544,7 +6730,7 @@ check_dmesg_for_errors() {
 ldiskfs_check_descriptors: Checksum for group 0 failed\|\
 group descriptors corrupted"
 
-	res=$(do_nodes $(comma_list $(nodes_list)) "dmesg" | grep "$errors")
+	res=$(do_nodes -q $(comma_list $(nodes_list)) "dmesg" | grep "$errors")
 	[ -z "$res" ] && return 0
 	echo "Kernel error detected: $res"
 	return 1
@@ -6612,7 +6798,7 @@ run_one_logged() {
 
 		# remove temp files between repetitions to avoid test failures
 		[ -n "$append" -a -n "$DIR" -a -n "$tdir" -a -n "$tfile" ] &&
-			rm -rf $DIR/$tdir* $DIR/$tfile*
+			rm -rvf $DIR/$tdir* $DIR/$tfile*
 		# loop around subshell so stack_trap EXIT triggers each time
 		(run_one $testnum "$testmsg") 2>&1 | tee -i $append $test_log
 		rc=${PIPESTATUS[0]}
@@ -6667,10 +6853,6 @@ skip_logged(){
 	log_sub_test_end "SKIP" "0" "0" "$@"
 }
 
-canonical_path() {
-	(cd $(dirname $1); echo $PWD/$(basename $1))
-}
-
 grant_from_clients() {
 	local nodes="$1"
 
@@ -6709,7 +6891,7 @@ check_grant() {
 
 	# sync all the data and make sure no pending data on server
 	do_nodes $clients sync
-	clients_up # initiate all idling connections
+	do_nodes $clients $LFS df # initiate all idling connections
 
 	# get client grant
 	cli_grant=$(grant_from_clients $clients)
@@ -6756,7 +6938,7 @@ ostuuid_from_index()
 }
 
 ostname_from_index() {
-    local uuid=$(ostuuid_from_index $1)
+    local uuid=$(ostuuid_from_index $1 $2)
     echo ${uuid/_UUID/}
 }
 
@@ -6790,7 +6972,34 @@ host_id() {
 # Description:
 #   Returns list of ip addresses for each interface
 local_addr_list() {
-	ip addr | awk '/inet\ / {print $2}' | awk -F\/ '{print $1}'
+	ip addr | awk '/inet / {print $2}' | awk -F/ '{print $1}'
+}
+
+# Description:
+#   Returns list of interfaces configured for LNet
+lnet_if_list() {
+	local nids=( $($LCTL list_nids | xargs echo) )
+
+	[[ -z ${nids[@]} ]] &&
+		return 0
+
+	declare -a INTERFACES
+
+	for ((i = 0; i < ${#nids[@]}; i++)); do
+		ip=$(sed 's/^\(.*\)@.*$/\1/'<<<${nids[i]})
+		INTERFACES[i]=$(ip -o a s |
+				awk '$4 ~ /^'$ip'\//{print $2}')
+		INTERFACES=($(echo "${INTERFACES[@]}" | tr ' ' '\n' | uniq | tr '\n' ' '))
+		if [[ -z ${INTERFACES[i]} ]]; then
+			error "Can't determine interface name for NID ${nids[i]}"
+		elif [[ 1 -ne $(wc -w <<<${INTERFACES[i]}) ]]; then
+			error "Found $(wc -w <<<${INTERFACES[i]}) interfaces for NID ${nids[i]}. Expect 1"
+		fi
+	done
+
+	echo "${INTERFACES[@]}"
+
+	return 0
 }
 
 is_local_addr() {
@@ -6914,11 +7123,6 @@ osts_nodes () {
 	echo -n $(facets_nodes $(get_facets OST))
 }
 
-# Get all of the active AGT (HSM agent) nodes.
-agts_nodes () {
-	echo -n $(facets_nodes $(get_facets AGT))
-}
-
 # Get all of the client nodes and active server nodes.
 nodes_list () {
 	local nodes=$HOSTNAME
@@ -7016,20 +7220,23 @@ init_clients_lists () {
     # Sanity check: exclude the dup entries
     RCLIENTS=$(for i in ${rclients//,/ }; do echo $i; done | sort -u)
 
-    clients="$SINGLECLIENT $HOSTNAME $RCLIENTS"
+	export CLIENT1=${CLIENT1:-$HOSTNAME}
+	export SINGLECLIENT=$CLIENT1
+
+	clients="$SINGLECLIENT $HOSTNAME $RCLIENTS"
 
     # Sanity check: exclude the dup entries from CLIENTS
     # for those configs which has SINGLCLIENT set to local client
     clients=$(for i in $clients; do echo $i; done | sort -u)
 
-    CLIENTS=$(comma_list $clients)
+	export CLIENTS=$(comma_list $clients)
     local -a remoteclients=($RCLIENTS)
     for ((i=0; $i<${#remoteclients[@]}; i++)); do
             varname=CLIENT$((i + 2))
-            eval $varname=${remoteclients[i]}
+			eval export $varname=${remoteclients[i]}
     done
 
-    CLIENTCOUNT=$((${#remoteclients[@]} + 1))
+	export CLIENTCOUNT=$((${#remoteclients[@]} + 1))
 }
 
 get_random_entry () {
@@ -7058,12 +7265,6 @@ get_node_count() {
     echo $nodes | wc -w || true
 }
 
-mixed_ost_devs () {
-    local nodes=$(osts_nodes)
-    local osscount=$(get_node_count "$nodes")
-    [ ! "$OSTCOUNT" = "$osscount" ]
-}
-
 mixed_mdt_devs () {
     local nodes=$(mdts_nodes)
     local mdtcount=$(get_node_count "$nodes")
@@ -7088,23 +7289,6 @@ get_stripe () {
 	rm -f $file
 }
 
-setstripe_nfsserver () {
-	local dir=$1
-	local nfsexportdir=$2
-	shift
-	shift
-
-	local -a nfsexport=($(awk '"'$dir'" ~ $2 && $3 ~ "nfs" && $2 != "/" \
-		{ print $1 }' /proc/mounts | cut -f 1 -d :))
-
-	# check that only one nfs mounted
-	[[ -z $nfsexport ]] && echo "$dir is not nfs mounted" && return 1
-	(( ${#nfsexport[@]} == 1 )) ||
-		error "several nfs mounts found for $dir: ${nfsexport[@]} !"
-
-	do_nodev ${nfsexport[0]} lfs setstripe $nfsexportdir "$@"
-}
-
 # Check and add a test group.
 add_group() {
 	local group_id=$1
@@ -7163,35 +7347,46 @@ add_user() {
 }
 
 check_runas_id_ret() {
-    local myRC=0
-    local myRUNAS_UID=$1
-    local myRUNAS_GID=$2
-    shift 2
-    local myRUNAS=$@
-    if [ -z "$myRUNAS" ]; then
-        error_exit "myRUNAS command must be specified for check_runas_id"
-    fi
-    if $GSS_KRB5; then
-        $myRUNAS krb5_login.sh || \
-            error "Failed to refresh Kerberos V5 TGT for UID $myRUNAS_ID."
-    fi
-    mkdir $DIR/d0_runas_test
-    chmod 0755 $DIR
-    chown $myRUNAS_UID:$myRUNAS_GID $DIR/d0_runas_test
-    $myRUNAS touch $DIR/d0_runas_test/f$$ || myRC=$?
-    rm -rf $DIR/d0_runas_test
-    return $myRC
+	local myRC=0
+	local myRUNAS_UID=$1
+	local myRUNAS_GID=$2
+	shift 2
+	local myRUNAS=$@
+
+	if [ -z "$myRUNAS" ]; then
+		error_exit "check_runas_id_ret requires myRUNAS argument"
+	fi
+
+	$myRUNAS true ||
+		error "Unable to execute $myRUNAS"
+
+	id $myRUNAS_UID > /dev/null ||
+		error "Invalid RUNAS_ID $myRUNAS_UID. Please set RUNAS_ID to " \
+		      "some UID which exists on MDS and client or add user " \
+		      "$myRUNAS_UID:$myRUNAS_GID on these nodes."
+
+	if $GSS_KRB5; then
+		$myRUNAS krb5_login.sh ||
+			error "Failed to refresh krb5 TGT for UID $myRUNAS_ID."
+	fi
+	mkdir $DIR/d0_runas_test
+	chmod 0755 $DIR
+	chown $myRUNAS_UID:$myRUNAS_GID $DIR/d0_runas_test
+	$myRUNAS -u $myRUNAS_UID -g $myRUNAS_GID touch $DIR/d0_runas_test/f$$ ||
+		myRC=$?
+	rm -rf $DIR/d0_runas_test
+	return $myRC
 }
 
 check_runas_id() {
-    local myRUNAS_UID=$1
-    local myRUNAS_GID=$2
-    shift 2
-    local myRUNAS=$@
-    check_runas_id_ret $myRUNAS_UID $myRUNAS_GID $myRUNAS || \
-        error "unable to write to $DIR/d0_runas_test as UID $myRUNAS_UID.
-        Please set RUNAS_ID to some UID which exists on MDS and client or
-        add user $myRUNAS_UID:$myRUNAS_GID on these nodes."
+	local myRUNAS_UID=$1
+	local myRUNAS_GID=$2
+	shift 2
+	local myRUNAS=$@
+
+	check_runas_id_ret $myRUNAS_UID $myRUNAS_GID $myRUNAS || \
+		error "unable to write to $DIR/d0_runas_test as " \
+		      "UID $myRUNAS_UID."
 }
 
 # obtain the UID/GID for MPI_USER
@@ -7299,6 +7494,29 @@ calc_osc_kbytes () {
 	$LCTL get_param -n osc.*[oO][sS][cC][-_][0-9a-f]*.$1 | calc_sum
 }
 
+free_min_max () {
+	wait_delete_completed
+	AVAIL=($(lctl get_param -n osc.*[oO][sS][cC]-[^M]*.kbytesavail))
+	echo "OST kbytes available: ${AVAIL[*]}"
+	MAXV=${AVAIL[0]}
+	MAXI=0
+	MINV=${AVAIL[0]}
+	MINI=0
+	for ((i = 0; i < ${#AVAIL[@]}; i++)); do
+		#echo OST $i: ${AVAIL[i]}kb
+		if [[ ${AVAIL[i]} -gt $MAXV ]]; then
+			MAXV=${AVAIL[i]}
+			MAXI=$i
+		fi
+		if [[ ${AVAIL[i]} -lt $MINV ]]; then
+			MINV=${AVAIL[i]}
+			MINI=$i
+		fi
+	done
+	echo "Min free space: OST $MINI: $MINV"
+	echo "Max free space: OST $MAXI: $MAXV"
+}
+
 # save_lustre_params(comma separated facet list, parameter_mask)
 # generate a stream of formatted strings (<facet> <param name>=<param value>)
 save_lustre_params() {
@@ -7331,15 +7549,18 @@ restore_lustre_params() {
 
 check_node_health() {
 	local nodes=${1:-$(comma_list $(nodes_list))}
-
-	for node in ${nodes//,/ }; do
-		check_network "$node" 5
-		if [ $? -eq 0 ]; then
-			do_node $node "$LCTL get_param catastrophe 2>&1" |
-				grep -q "catastrophe=1" &&
-				error "$node:LBUG/LASSERT detected" || true
-		fi
-	done
+	local health=$TMP/node_health.$$
+
+	do_nodes -q $nodes "$LCTL get_param catastrophe 2>&1" | tee $health |
+		grep "catastrophe=1" && error "LBUG/LASSERT detected"
+	# Only check/report network health if get_param isn't reported, since
+	# *clearly* the network is working if get_param returned something.
+	if (( $(grep -c catastro $health) != $(wc -w <<< ${nodes//,/ }) )); then
+		for node in ${nodes//,/}; do
+			check_network $node 5
+		done
+	fi
+	rm -f $health
 }
 
 mdsrate_cleanup () {
@@ -7350,11 +7571,6 @@ mdsrate_cleanup () {
 	fi
 }
 
-delayed_recovery_enabled () {
-    local var=${SINGLEMDS}_svc
-    do_facet $SINGLEMDS lctl get_param -n mdd.${!var}.stale_export_age > /dev/null 2>&1
-}
-
 ########################
 
 convert_facet2label() {
@@ -7379,20 +7595,6 @@ get_clientosc_proc_path() {
 	echo "${1}-osc-[-0-9a-f]*"
 }
 
-# If the 2.0 MDS was mounted on 1.8 device, then the OSC and LOV names
-# used by MDT would not be changed.
-# mdt lov: fsname-mdtlov
-# mdt osc: fsname-OSTXXXX-osc
-mds_on_old_device() {
-    local mds=${1:-"$SINGLEMDS"}
-
-    if [ $(lustre_version_code $mds) -gt $(version_code 1.9.0) ]; then
-        do_facet $mds "lctl list_param osc.$FSNAME-OST*-osc \
-            > /dev/null 2>&1" && return 0
-    fi
-    return 1
-}
-
 get_mdtosc_proc_path() {
 	local mds_facet=$1
 	local ost_label=${2:-"*OST*"}
@@ -7423,48 +7625,49 @@ get_osc_import_name() {
 }
 
 _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
+	local expected="$1"
+	local CONN_PROC="$2"
+	local maxtime=${3:-$(max_recovery_time)}
+	local err_on_fail=${4:-1}
+	local CONN_STATE
+	local i=0
 
 	CONN_STATE=$($LCTL get_param -n $CONN_PROC 2>/dev/null | cut -f2 | uniq)
-    while ! echo "${CONN_STATE}" | egrep -q "^${expected}\$" ; do
-        if [ "${expected}" == "DISCONN" ]; then
-            # for disconn we can check after proc entry is removed
-            [ "x${CONN_STATE}" == "x" ] && return 0
-            #  with AT enabled, we can have connect request timeout near of
-            # reconnect timeout and test can't see real disconnect
-            [ "${CONN_STATE}" == "CONNECTING" ] && return 0
-        fi
-	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 $CONN_PROC 2>/dev/null | cut -f2 | uniq)
-        i=$(($i + 1))
-    done
+	while ! echo "${CONN_STATE}" | egrep -q "^${expected}\$" ; do
+		if [[ "${expected}" == "DISCONN" ]]; then
+			# for disconn we can check after proc entry is removed
+			[[ -z "${CONN_STATE}" ]] && return 0
+			# with AT, we can have connect request timeout near
+			# reconnect timeout and test can't see real disconnect
+			[[ "${CONN_STATE}" == "CONNECTING" ]] && return 0
+		fi
+		if (( $i >= $maxtime )); then
+			(( $err_on_fail != 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 $CONN_PROC 2>/dev/null |
+			     cut -f2 | uniq)
+		i=$((i + 1))
+	done
 
-    log "$CONN_PROC in ${CONN_STATE} state after $i sec"
-    return 0
+	log "$CONN_PROC in ${CONN_STATE} state after $i sec"
+	return 0
 }
 
 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 $error_on_failure || return
-    done
+	local expected="$1"
+	local params="$2"
+	local maxtime=${3:-$(max_recovery_time)}
+	local err_on_fail=${4:-1}
+	local param
+
+	for param in ${params//,/ }; do
+		_wait_import_state "$expected" "$param" $maxtime $err_on_fail ||
+		return
+	done
 }
 
 wait_import_state_mount() {
@@ -7472,7 +7675,7 @@ wait_import_state_mount() {
 		return 0
 	fi
 
-	wait_import_state $*
+	wait_import_state "$@"
 }
 
 # One client request could be timed out because server was not ready
@@ -7552,8 +7755,9 @@ _wait_osc_import_state() {
 		params=$param
 	fi
 
+	local plist=$(comma_list $params)
 	if ! do_rpc_nodes "$(facet_active_host $facet)" \
-			wait_import_state $expected "$params" $maxtime; then
+			wait_import_state $expected $plist $maxtime; then
 		error "$facet: import is not in $expected state after $maxtime"
 		return 1
 	fi
@@ -7608,8 +7812,9 @@ _wait_mgc_import_state() {
 			params=$($LCTL list_param $param 2>/dev/null || true)
 		done
 	fi
+	local plist=$(comma_list $params)
 	if ! do_rpc_nodes "$(facet_active_host $facet)" \
-			wait_import_state $expected "$params" $maxtime \
+			wait_import_state $expected $plist $maxtime \
 					  $error_on_failure; then
 		if [ $error_on_failure -ne 0 ]; then
 		    error "import is not in ${expected} state"
@@ -7656,6 +7861,10 @@ get_clientmgc_proc_path() {
 }
 
 do_rpc_nodes () {
+	local quiet
+
+	[[ "$1" == "--quiet" || "$1" == "-q" ]] && quiet="$1" && shift
+
 	local list=$1
 	shift
 
@@ -7665,15 +7874,14 @@ do_rpc_nodes () {
 	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} bash rpc.sh $@ "
+	do_nodes ${quiet:-"--verbose"} $list "${RPATH} NAME=${NAME} bash rpc.sh $@ "
 }
 
 wait_clients_import_state () {
-	local list=$1
-	local facet=$2
-	local expected=$3
-
-	local facets=$facet
+	local list="$1"
+	local facet="$2"
+	local expected="$3"
+	local facets="$facet"
 
 	if [ "$FAILURE_MODE" = HARD ]; then
 		facets=$(facets_on_host $(facet_active_host $facet))
@@ -7684,11 +7892,11 @@ wait_clients_import_state () {
 		local proc_path
 		case $facet in
 		ost* ) proc_path="osc.$(get_clientosc_proc_path \
-				  $label).ost_server_uuid" ;;
+					$label).ost_server_uuid" ;;
 		mds* ) proc_path="mdc.$(get_clientmdc_proc_path \
-				  $label).mds_server_uuid" ;;
+					$label).mds_server_uuid" ;;
 		mgs* ) proc_path="mgc.$(get_clientmgc_proc_path \
-				  $label).mgs_server_uuid" ;;
+					$label).mgs_server_uuid" ;;
 		*) error "unknown facet!" ;;
 		esac
 
@@ -7786,6 +7994,7 @@ check_pool_not_exist() {
 create_pool() {
 	local fsname=${1%%.*}
 	local poolname=${1##$fsname.}
+	local keep_pools=${2:-false}
 
 	stack_trap "destroy_test_pools $fsname" EXIT
 	do_facet mgs lctl pool_new $1
@@ -7804,7 +8013,7 @@ create_pool() {
 	wait_update $HOSTNAME "lctl get_param -n lov.$fsname-*.pools.$poolname \
 		2>/dev/null || echo foo" "" || error "pool_new failed $1"
 
-	add_pool_to_list $1
+	$keep_pools || add_pool_to_list $1
 	return $RC
 }
 
@@ -7822,10 +8031,18 @@ remove_pool_from_list () {
 	local poolname=${1##$fsname.}
 
 	local listvar=${fsname}_CREATED_POOLS
-	local temp=${listvar}=$(exclude_items_from_list ${!listvar} $poolname)
+	local temp=${listvar}=$(exclude_items_from_list "${!listvar}" $poolname)
 	eval export $temp
 }
 
+# cleanup all pools exist on $FSNAME
+destroy_all_pools () {
+	local i
+	for i in $(list_pool $FSNAME); do
+		destroy_pool $i
+	done
+}
+
 destroy_pool_int() {
 	local ost
 	local OSTS=$(list_pool $1)
@@ -7846,8 +8063,7 @@ destroy_pool() {
 
 	local RC
 
-	check_pool_not_exist $fsname.$poolname
-	[[ $? -eq 0 ]] && return 0
+	check_pool_not_exist $fsname.$poolname && return 0 || true
 
 	destroy_pool_int $fsname.$poolname
 	RC=$?
@@ -7941,8 +8157,11 @@ gather_logs () {
 	fi
 
 	if [ ! -f $LOGDIR/shared ]; then
-		do_nodes $list rsync -az "${prefix}.*.${suffix}" \
-			 $HOSTNAME:$LOGDIR
+		local remote_nodes=$(exclude_items_from_list $list $HOSTNAME)
+
+		for node in ${remote_nodes//,/ }; do
+			rsync -az -e ssh $node:${prefix}.'*'.${suffix} $LOGDIR &
+		done
 	fi
 }
 
@@ -8480,124 +8699,34 @@ log_sub_test_end() {
 
 run_llverdev()
 {
-        local dev=$1
-        local llverdev_opts=$2
-        local devname=$(basename $1)
-        local size=$(grep "$devname"$ /proc/partitions | awk '{print $3}')
-        # loop devices aren't in /proc/partitions
-        [ "x$size" == "x" ] && local size=$(ls -l $dev | awk '{print $5}')
+	local dev=$1; shift
+	local llverdev_opts="$*"
+	local devname=$(basename $dev)
+	local size=$(awk "/$devname$/ {print \$3}" /proc/partitions)
+	# loop devices aren't in /proc/partitions
+	[[ -z "$size" ]] && size=$(stat -c %s $dev)
 
-        size=$(($size / 1024 / 1024)) # Gb
+	local size_gb=$((size / 1024 / 1024)) # Gb
 
-        local partial_arg=""
-        # Run in partial (fast) mode if the size
-        # of a partition > 1 GB
-        [ $size -gt 1 ] && partial_arg="-p"
+	local partial_arg=""
+	# Run in partial (fast) mode if the size of a partition > 1 GB
+	(( $size == 0 || $size_gb > 1 )) && partial_arg="-p"
 
-        llverdev --force $partial_arg $llverdev_opts $dev
+	llverdev --force $partial_arg $llverdev_opts $dev
 }
 
 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 $2}' |sed 's/G//') #GB
-
-        # Run in partial (fast) mode if the size
-        # of a partition > 1 GB
-        [ "x$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
-}
+	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
 
-#Remove files from MDT
-remove_mdt_files() {
-	local facet=$1
-	local mdtdev=$2
-	shift 2
-	local files="$@"
-	local mntpt=$(facet_mntpt $facet)
-	local opts=$MDS_MOUNT_OPTS
+	# Run in partial (fast) mode if the size of a partition > 1 GB
+	[ "x$use_partial_arg" != "xno" ] && [ $size -gt 1 ] && partial_arg="-p"
 
-	echo "removing files from $mdtdev on $facet: $files"
-	if [ $(facet_fstype $facet) == ldiskfs ] &&
-	   ! 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
-	for f in $files; do
-		rm $mntpt/ROOT/$f || { rc=$?; break; }
-	done
-	umount -f $mntpt || return $?
-	return $rc
-}
-
-duplicate_mdt_files() {
-	local facet=$1
-	local mdtdev=$2
-	shift 2
-	local files="$@"
-	local mntpt=$(facet_mntpt $facet)
-	local opts=$MDS_MOUNT_OPTS
-
-	echo "duplicating files on $mdtdev on $facet: $files"
-	mkdir -p $mntpt || return $?
-	if [ $(facet_fstype $facet) == ldiskfs ] &&
-	   ! do_facet $facet test -b $mdtdev; then
-		opts=$(csa_add "$opts" -o loop)
-	fi
-	mount -t $(facet_fstype $facet) $opts $mdtdev $mntpt ||
-		return $?
-
-    do_umount() {
-        trap 0
-        popd > /dev/null
-        rm $tmp
-        umount -f $mntpt
-    }
-    trap do_umount EXIT
-
-    tmp=$(mktemp $TMP/setfattr.XXXXXXXXXX)
-    pushd $mntpt/ROOT > /dev/null || return $?
-    rc=0
-    for f in $files; do
-        touch $f.bad || return $?
-        getfattr -n trusted.lov $f | sed "s#$f#&.bad#" > $tmp
-        rc=${PIPESTATUS[0]}
-        [ $rc -eq 0 ] || return $rc
-        setfattr --restore $tmp || return $?
-    done
-    do_umount
+	llverfs $partial_arg $llverfs_opts $dir
 }
 
 run_sgpdd () {
@@ -8687,22 +8816,11 @@ get_block_count() {
 	echo -n ${count:-0}
 }
 
-# Get the block size of the filesystem.
-get_block_size() {
-	local facet=$1
-	local device=$2
-	local size
-
-	[ -z "$CLIENTONLY" ] && size=$(do_facet $facet "$DUMPE2FS -h $device 2>&1" |
-		awk '/^Block size:/ {print $3}')
-	echo -n ${size:-0}
-}
-
 # Check whether the "ea_inode" feature is enabled or not, to allow
 # ldiskfs xattrs over one block in size.  Allow both the historical
 # Lustre feature name (large_xattr) and the upstream name (ea_inode).
 large_xattr_enabled() {
-	[[ $(facet_fstype $SINGLEMDS) == zfs ]] && return 1
+	[[ $(facet_fstype $SINGLEMDS) == zfs ]] && return 0
 
 	local mds_dev=$(mdsdevname ${SINGLEMDS//mds/})
 
@@ -8751,7 +8869,6 @@ mds_backup_restore() {
 	local devname=$(mdsdevname $(facet_number $facet))
 	local mntpt=$(facet_mntpt brpt)
 	local rcmd="do_facet $facet"
-	local metaea=${TMP}/backup_restore.ea
 	local metadata=${TMP}/backup_restore.tgz
 	local opts=${MDS_MOUNT_FS_OPTS}
 	local svc=${facet}_svc
@@ -8765,41 +8882,36 @@ mds_backup_restore() {
 	# step 1: build mount point
 	${rcmd} mkdir -p $mntpt
 	# step 2: cleanup old backup
-	${rcmd} rm -f $metaea $metadata
+	${rcmd} rm -f $metadata
 	# step 3: mount dev
-	${rcmd} mount -t ldiskfs $opts $devname $mntpt || return 1
+	${rcmd} mount -t ldiskfs $opts $devname $mntpt || return 3
 	if [ ! -z $igif ]; then
 		# step 3.5: rm .lustre
-		${rcmd} rm -rf $mntpt/ROOT/.lustre || return 1
+		${rcmd} rm -rf $mntpt/ROOT/.lustre || return 3
 	fi
-	# step 4: backup metaea
-	echo "backup EA"
-	${rcmd} "cd $mntpt && getfattr -R -d -m '.*' -P . > $metaea && cd -" ||
-		return 2
-	# step 5: backup metadata
+	# step 4: backup metadata
 	echo "backup data"
-	${rcmd} tar zcf $metadata -C $mntpt/ . > /dev/null 2>&1 || return 3
-	# step 6: umount
-	${rcmd} $UMOUNT $mntpt || return 4
-	# step 8: reformat dev
+	${rcmd} tar zcf $metadata --xattrs --xattrs-include="trusted.*" \
+		--sparse -C $mntpt/ . > /dev/null 2>&1 || return 4
+	# step 5: umount
+	${rcmd} $UMOUNT $mntpt || return 5
+	# step 6: reformat dev
 	echo "reformat new device"
 	format_mdt $(facet_number $facet)
-	# step 9: mount dev
+	# step 7: mount dev
 	${rcmd} mount -t ldiskfs $opts $devname $mntpt || return 7
-	# step 10: restore metadata
+	# step 8: restore metadata
 	echo "restore data"
-	${rcmd} tar zxfp $metadata -C $mntpt > /dev/null 2>&1 || return 8
-	# step 11: restore metaea
-	echo "restore EA"
-	${rcmd} "cd $mntpt && setfattr --restore=$metaea && cd - " || return 9
-	# step 12: remove recovery logs
+	${rcmd} tar zxfp $metadata --xattrs --xattrs-include="trusted.*" \
+		--sparse -C $mntpt > /dev/null 2>&1 || return 8
+	# step 9: remove recovery logs
 	echo "remove recovery logs"
 	${rcmd} rm -fv $mntpt/OBJECTS/* $mntpt/CATALOGS
-	# step 13: umount dev
+	# step 10: umount dev
 	${rcmd} $UMOUNT $mntpt || return 10
-	# step 14: cleanup tmp backup
+	# step 11: cleanup tmp backup
 	${rcmd} rm -f $metaea $metadata
-	# step 15: reset device label - it's not virgin on
+	# step 12: reset device label - it's not virgin on
 	${rcmd} e2label $devname ${!svc}
 }
 
@@ -8933,7 +9045,7 @@ check_mount_and_prep()
 	is_mounted $MOUNT || setupall
 
 	rm -rf $DIR/[df][0-9]* || error "Fail to cleanup the env!"
-	mkdir $DIR/$tdir || error "Fail to mkdir $DIR/$tdir."
+	mkdir_on_mdt0 $DIR/$tdir || error "Fail to mkdir $DIR/$tdir."
 	for idx in $(seq $MDSCOUNT); do
 		local name="MDT$(printf '%04x' $((idx - 1)))"
 		rm -rf $MOUNT/.lustre/lost+found/$name/*
@@ -8999,8 +9111,18 @@ pool_add_targets() {
 	fi
 
 	local t=$(for i in $list; do printf "$FSNAME-OST%04x_UUID " $i; done)
+	local tg=$(for i in $list;
+		do printf -- "-e $FSNAME-OST%04x_UUID " $i; done)
+	local firstx=$(printf "%04x" $first)
+	local lastx=$(printf "%04x" $last)
+
 	do_facet mgs $LCTL pool_add \
-			$FSNAME.$pool $FSNAME-OST[$first-$last/$step]
+		$FSNAME.$pool $FSNAME-OST[$firstx-$lastx/$step]
+	# ignore EEXIST(17)
+	if (( $? != 0 && $? != 17 )); then
+		error_noexit "pool_add $FSNAME-OST[$firstx-$lastx/$step] failed"
+		return 3
+	fi
 
 	# wait for OSTs to be added to the pool
 	for mds_id in $(seq $MDSCOUNT); do
@@ -9008,23 +9130,16 @@ pool_add_targets() {
 		local lodname=$FSNAME-MDT$(printf "%04x" $mdt_id)-mdtlov
 		wait_update_facet mds$mds_id \
 			"lctl get_param -n lod.$lodname.pools.$pool |
-				sort -u | tr '\n' ' ' " "$t" || {
+				grep $tg | sort -u | tr '\n' ' '" "$t" || {
 			error_noexit "mds$mds_id: Add to pool failed"
-			return 3
+			return 2
 		}
 	done
-	wait_update $HOSTNAME "lctl get_param -n lov.$FSNAME-*.pools.$pool \
-			| sort -u | tr '\n' ' ' " "$t" || {
+	wait_update $HOSTNAME "lctl get_param -n lov.$FSNAME-*.pools.$pool |
+			grep $tg | sort -u | tr '\n' ' ' " "$t" || {
 		error_noexit "Add to pool failed"
 		return 1
 	}
-	local lfscount=$($LFS pool_list $FSNAME.$pool | grep -c "\-OST")
-	local addcount=$(((last - first) / step + 1))
-	[ $lfscount -eq $addcount ] || {
-		error_noexit "lfs pool_list bad ost count" \
-						"$lfscount != $addcount"
-		return 2
-	}
 }
 
 pool_set_dir() {
@@ -9547,14 +9662,17 @@ check_clients_evicted() {
 	local rc=0
 
 	for osc in $oscs; do
-		((rc++))
 		echo "Check state for $osc"
 		local evicted=$(do_facet client $LCTL get_param osc.$osc.state |
-			tail -n 3 | awk -F"[ [,]" \
-			'/EVICTED ]$/ { if (mx<$5) {mx=$5;} } END { print mx }')
+			tail -n 5 | awk -F"[ ,]" \
+			'/EVICTED/ { if (mx<$4) { mx=$4; } } END { print mx }')
 		if (($? == 0)) && (($evicted > $before)); then
 			echo "$osc is evicted at $evicted"
-			((rc--))
+		else
+			((rc++))
+			echo "$osc was not evicted after $before:"
+			do_facet client $LCTL get_param osc.$osc.state |
+				tail -n 8
 		fi
 	done
 
@@ -9611,8 +9729,8 @@ changelog_register() {
 			error "$mdt: changelog_mask=+hsm failed: $?"
 
 		local cl_user
-		cl_user=$(do_facet $facet \
-				  $LCTL --device $mdt changelog_register -n) ||
+		cl_user=$(do_facet $facet $LCTL --device $mdt \
+			changelog_register -n $@) ||
 			error "$mdt: register changelog user failed: $?"
 		stack_trap "__changelog_deregister $facet $cl_user" EXIT
 
@@ -9699,6 +9817,7 @@ changelog_clear() {
 	# so reorder to get same order than in changelog_register()
 	local cl_facets=$(echo "${!CL_USERS[@]}" | tr " " "\n" | sort |
 			  tr "\n" " ")
+	local cl_user
 
 	for facet in $cl_facets; do
 		for cl_user in ${CL_USERS[$facet]}; do
@@ -10076,10 +10195,16 @@ init_agt_vars() {
 	export SINGLEAGT=${SINGLEAGT:-agt1}
 
 	export HSMTOOL=${HSMTOOL:-"lhsmtool_posix"}
+	export HSMTOOL_PID_FILE=${HSMTOOL_PID_FILE:-"/var/run/lhsmtool_posix.pid"}
 	export HSMTOOL_VERBOSE=${HSMTOOL_VERBOSE:-""}
 	export HSMTOOL_UPDATE_INTERVAL=${HSMTOOL_UPDATE_INTERVAL:=""}
 	export HSMTOOL_EVENT_FIFO=${HSMTOOL_EVENT_FIFO:=""}
 	export HSMTOOL_TESTDIR
+	export HSMTOOL_ARCHIVE_FORMAT=${HSMTOOL_ARCHIVE_FORMAT:-v2}
+
+	if ! [[ $HSMTOOL =~ hsmtool ]]; then
+		echo "HSMTOOL = '$HSMTOOL' does not contain 'hsmtool', GLWT" >&2
+	fi
 
 	HSM_ARCHIVE_NUMBER=2
 
@@ -10113,10 +10238,17 @@ get_mdt_devices() {
 	done
 }
 
+pkill_copytools() {
+	local hosts="$1"
+	local signal="$2"
+
+	do_nodes "$hosts" "pkill --pidfile=$HSMTOOL_PID_FILE --signal=$signal hsmtool"
+}
+
 copytool_continue() {
 	local agents=${1:-$(facet_active_host $SINGLEAGT)}
 
-	do_nodesv $agents "libtool execute pkill -CONT -x $HSMTOOL" || return 0
+	pkill_copytools "$agents" CONT || return 0
 	echo "Copytool is continued on $agents"
 }
 
@@ -10124,7 +10256,7 @@ kill_copytools() {
 	local hosts=${1:-$(facet_active_host $SINGLEAGT)}
 
 	echo "Killing existing copytools on $hosts"
-	do_nodesv $hosts "libtool execute killall -q $HSMTOOL" || true
+	pkill_copytools "$hosts" TERM || return 0
 	copytool_continue "$hosts"
 }
 
@@ -10160,28 +10292,29 @@ copytool_logfile()
 
 __lhsmtool_rebind()
 {
-	do_facet $facet $HSMTOOL -p "$hsm_root" --rebind "$@" "$mountpoint"
+	do_facet $facet $HSMTOOL "${hsmtool_options[@]}" --rebind "$@" "$mountpoint"
 }
 
 __lhsmtool_import()
 {
 	mkdir -p "$(dirname "$2")" ||
 		error "cannot create directory '$(dirname "$2")'"
-	do_facet $facet $HSMTOOL -p "$hsm_root" --import "$@" "$mountpoint"
+	do_facet $facet $HSMTOOL "${hsmtool_options[@]}" --import "$@" "$mountpoint"
 }
 
 __lhsmtool_setup()
 {
-	local cmd="$HSMTOOL $HSMTOOL_VERBOSE --daemon --hsm-root \"$hsm_root\""
+	local host="$(facet_host "$facet")"
+	local cmd="$HSMTOOL ${hsmtool_options[@]} --daemon --pid-file=$HSMTOOL_PID_FILE"
 	[ -n "$bandwidth" ] && cmd+=" --bandwidth $bandwidth"
 	[ -n "$archive_id" ] && cmd+=" --archive $archive_id"
-	[ ${#misc_options[@]} -gt 0 ] &&
-		cmd+=" $(IFS=" " echo "$@")"
-	cmd+=" \"$mountpoint\""
+#	[ ${#misc_options[@]} -gt 0 ] &&
+#		cmd+=" $(IFS=" " echo "$@")"
+	cmd+=" $@ \"$mountpoint\""
 
-	echo "Starting copytool $facet on $(facet_host $facet)"
-	stack_trap "do_facet $facet libtool execute pkill -x '$HSMTOOL' || true" EXIT
-	do_facet $facet "$cmd < /dev/null > \"$(copytool_logfile $facet)\" 2>&1"
+	echo "Starting copytool '$facet' on '$host' with cmdline '$cmd'"
+	stack_trap "pkill_copytools $host TERM || true" EXIT
+	do_node "$host" "$cmd < /dev/null > \"$(copytool_logfile $facet)\" 2>&1"
 }
 
 hsm_root() {
@@ -10214,7 +10347,17 @@ copytool()
 
 	# Parse arguments
 	local fail_on_error=true
-	local -a misc_options
+	local -a hsmtool_options=("--hsm-root=$hsm_root")
+	local -a action_options=()
+
+	if [[ -n "$HSMTOOL_ARCHIVE_FORMAT" ]]; then
+		hsmtool_options+=("--archive-format=$HSMTOOL_ARCHIVE_FORMAT")
+	fi
+
+	if [[ -n "$HSMTOOL_VERBOSE" ]]; then
+		hsmtool_options+=("$HSMTOOL_VERBOSE")
+	fi
+
 	while [ $# -gt 0 ]; do
 		case "$1" in
 		-f|--facet)
@@ -10242,7 +10385,7 @@ copytool()
 			;;
 		*)
 			# Uncommon(/copytool dependent) option
-			misc_options+=("$1")
+			action_options+=("$1")
 			;;
 		esac
 		shift
@@ -10258,7 +10401,7 @@ copytool()
 		;;
 	esac
 
-	__${copytool}_${action} "${misc_options[@]}"
+	__${copytool}_${action} "${action_options[@]}"
 	if [ $? -ne 0 ]; then
 		local error_msg
 
@@ -10268,8 +10411,8 @@ copytool()
 			error_msg="Failed to start copytool $facet on '$host'"
 			;;
 		import)
-			local src="${misc_options[0]}"
-			local dest="${misc_options[1]}"
+			local src="${action_options[0]}"
+			local dest="${action_options[1]}"
 			error_msg="Failed to import '$src' to '$dest'"
 			;;
 		rebind)
@@ -10511,32 +10654,244 @@ statx_supported() {
 #
 function createmany() {
 	local count=${!#}
+	local rc
 
-	(( count > 100 )) && {
-		local saved_debug=$($LCTL get_param -n debug)
-		local list=$(comma_list $(all_nodes))
-
-		do_nodes $list $LCTL set_param debug=0
-	}
+	if (( count > 100 )); then
+		debugsave
+		do_nodes $(comma_list $(all_nodes)) $LCTL set_param -n debug=0
+	fi
 	$LUSTRE/tests/createmany $*
-	local rc=$?
-	(( count > 100 )) &&
-		do_nodes $list "$LCTL set_param debug=\\\"$saved_debug\\\""
+	rc=$?
+	debugrestore
+
 	return $rc
 }
 
 function unlinkmany() {
 	local count=${!#}
+	local rc
 
-	(( count > 100 )) && {
-		local saved_debug=$($LCTL get_param -n debug)
-		local list=$(comma_list $(all_nodes))
-
-		do_nodes $list $LCTL set_param debug=0
-	}
+	if (( count > 100 )); then
+		debugsave
+		do_nodes $(comma_list $(all_nodes)) $LCTL set_param -n debug=0
+	fi
 	$LUSTRE/tests/unlinkmany $*
-	local rc=$?
-	(( count > 100 )) &&
-		do_nodes $list "$LCTL set_param debug=\\\"$saved_debug\\\""
+	rc=$?
+	debugrestore
+
 	return $rc
 }
+
+# Check if fallocate supported on OSTs, enable if unset, default mode=0
+# Optionally pass the OST fallocate mode (0=unwritten extents, 1=zero extents)
+function check_set_fallocate()
+{
+	local new_mode="$1"
+	local osts=$(comma_list $(osts_nodes))
+	local fa_mode="osd-ldiskfs.*.fallocate_zero_blocks"
+	local old_mode=$(do_facet ost1 $LCTL get_param -n $fa_mode 2>/dev/null|
+			 head -n 1)
+
+	[[ -n "$old_mode" ]] || { echo "fallocate not supported"; return 1; }
+	[[ -z "$new_mode" && "$old_mode" != "-1" ]] &&
+		{ echo "keep default fallocate mode: $old_mode"; return 0; }
+	[[ "$new_mode" && "$old_mode" == "$new_mode" ]] &&
+		{ echo "keep current fallocate mode: $old_mode"; return 0; }
+
+	stack_trap "do_nodes $osts $LCTL set_param $fa_mode=$old_mode"
+	do_nodes $osts $LCTL set_param $fa_mode=${new_mode:-0} ||
+		error "set $fa_mode=$new_mode"
+}
+
+# Check if fallocate supported on OSTs, enable if unset, skip if unavailable
+function check_set_fallocate_or_skip()
+{
+	[ "$ost1_FSTYPE" != ldiskfs ] && skip "non-ldiskfs backend"
+	check_set_fallocate || skip "need at least 2.13.57 for fallocate"
+}
+
+function disable_opencache()
+{
+	local state=$($LCTL get_param -n "llite.*.opencache_threshold_count" | head -1)
+
+	test -z "${saved_OPENCACHE_value}" &&
+					export saved_OPENCACHE_value="$state"
+
+	[[ "$state" = "off" ]] && return
+
+	$LCTL set_param -n "llite.*.opencache_threshold_count"=off
+}
+
+function set_opencache()
+{
+	local newvalue="$1"
+	local state=$($LCTL get_param -n "llite.*.opencache_threshold_count")
+
+	[[ -n "$newvalue" ]] || return
+
+	[[ -n "${saved_OPENCACHE_value}" ]] ||
+					export saved_OPENCACHE_value="$state"
+
+	$LCTL set_param -n "llite.*.opencache_threshold_count"=$newvalue
+}
+
+
+
+function restore_opencache()
+{
+	[[ -z "${saved_OPENCACHE_value}" ]] ||
+		$LCTL set_param -n "llite.*.opencache_threshold_count"=${saved_OPENCACHE_value}
+}
+
+# LU-13417: XXX lots of tests assume the directory to be created under MDT0,
+# created on MDT0, use this function to create directory on specific MDT
+# explicitly, and set default LMV to create subdirs on the same MDT too.
+mkdir_on_mdt() {
+	local mdt
+	local OPTIND=1
+
+	while getopts "i:" opt $*; do
+		case $opt in
+			i) mdt=$OPTARG;;
+		esac
+	done
+
+	shift $((OPTIND - 1))
+
+	$LFS mkdir -i $mdt -c 1 $*
+}
+
+mkdir_on_mdt0() {
+	mkdir_on_mdt -i0 $*
+}
+
+# Wait for nodemap synchronization
+wait_nm_sync() {
+	local nodemap_name=$1
+	local key=$2
+	local value=$3
+	local opt=$4
+	local proc_param
+	local is_active=$(do_facet mgs $LCTL get_param -n nodemap.active)
+	local max_retries=20
+	local is_sync
+	local out1=""
+	local out2
+	local mgs_ip=$(host_nids_address $mgs_HOST $NETTYPE | cut -d' ' -f1)
+	local i
+
+	if [ "$nodemap_name" == "active" ]; then
+		proc_param="active"
+	elif [ -z "$key" ]; then
+		proc_param=${nodemap_name}
+	else
+		proc_param="${nodemap_name}.${key}"
+	fi
+	if [ "$opt" == "inactive" ]; then
+		# check nm sync even if nodemap is not activated
+		is_active=1
+		opt=""
+	fi
+	(( is_active == 0 )) && [ "$proc_param" != "active" ] && return
+
+	if [ -z "$value" ]; then
+		out1=$(do_facet mgs $LCTL get_param $opt \
+			nodemap.${proc_param} 2>/dev/null)
+		echo "On MGS ${mgs_ip}, ${proc_param} = $out1"
+	else
+		out1=$value;
+	fi
+
+	# if servers run on the same node, it is impossible to tell if they get
+	# synced with the mgs, so just wait an arbitrary 10 seconds
+	if [ $(facet_active_host mgs) == $(facet_active_host mds) ] &&
+	   [ $(facet_active_host mgs) == $(facet_active_host ost1) ]; then
+		echo "waiting 10 secs for sync"
+		sleep 10
+		return
+	fi
+
+	# wait up to 10 seconds for other servers to sync with mgs
+	for i in $(seq 1 10); do
+		for node in $(all_server_nodes); do
+			local node_ip=$(host_nids_address $node $NETTYPE |
+					cut -d' ' -f1)
+
+			is_sync=true
+			if [ -z "$value" ]; then
+				[ $node_ip == $mgs_ip ] && continue
+			fi
+
+			out2=$(do_node $node_ip $LCTL get_param $opt \
+			       nodemap.$proc_param 2>/dev/null)
+			echo "On $node ${node_ip}, ${proc_param} = $out2"
+			[ "$out1" != "$out2" ] && is_sync=false && break
+		done
+		$is_sync && break
+		sleep 1
+	done
+	if ! $is_sync; then
+		echo MGS
+		echo $out1
+		echo OTHER - IP: $node_ip
+		echo $out2
+		error "mgs and $nodemap_name ${key} mismatch, $i attempts"
+	fi
+	echo "waited $((i - 1)) seconds for sync"
+}
+
+consume_precreations() {
+	local dir=$1
+	local mfacet=$2
+	local OSTIDX=$3
+	local extra=${4:-2}
+	local OST=$(ostname_from_index $OSTIDX $dir)
+
+	test_mkdir -p $dir/${OST}
+	$LFS setstripe -i $OSTIDX -c 1 ${dir}/${OST}
+
+	# on the mdt's osc
+	local mdtosc_proc=$(get_mdtosc_proc_path $mfacet $OST)
+	local last_id=$(do_facet $mfacet $LCTL get_param -n \
+			osp.$mdtosc_proc.prealloc_last_id)
+	local next_id=$(do_facet $mfacet $LCTL get_param -n \
+			osp.$mdtosc_proc.prealloc_next_id)
+	echo "Creating to objid $last_id on ost $OST..."
+	createmany -o $dir/${OST}/f $next_id $((last_id - next_id + extra))
+}
+
+__exhaust_precreations() {
+	local OSTIDX=$1
+	local FAILLOC=$2
+	local FAILIDX=${3:-$OSTIDX}
+	local ofacet=ost$((OSTIDX + 1))
+
+	mkdir_on_mdt0 $DIR/$tdir
+	local mdtidx=$($LFS getstripe -m $DIR/$tdir)
+	local mfacet=mds$((mdtidx + 1))
+	echo OSTIDX=$OSTIDX MDTIDX=$mdtidx
+
+	local mdtosc_proc=$(get_mdtosc_proc_path $mfacet)
+	do_facet $mfacet $LCTL get_param osp.$mdtosc_proc.prealloc*
+
+#define OBD_FAIL_OST_ENOSPC              0x215
+	do_facet $ofacet $LCTL set_param fail_val=$FAILIDX fail_loc=0x215
+
+	consume_precreations $DIR/$tdir $mfacet $OSTIDX
+
+	do_facet $mfacet $LCTL get_param osp.$mdtosc_proc.prealloc*
+	do_facet $ofacet $LCTL set_param fail_loc=$FAILLOC
+}
+
+exhaust_precreations() {
+	__exhaust_precreations $1 $2 $3
+	sleep_maxage
+}
+
+exhaust_all_precreations() {
+	local i
+	for (( i=0; i < OSTCOUNT; i++ )) ; do
+		__exhaust_precreations $i $1 -1
+	done
+	sleep_maxage
+}