X-Git-Url: https://git.whamcloud.com/?p=fs%2Flustre-release.git;a=blobdiff_plain;f=lustre%2Ftests%2Ftest-framework.sh;h=bc8de5c9c14e319b76e2e10f4bd0a89e1853b1f2;hp=42e51a1ff483179b956c82d0e37679ef37a7360a;hb=d375299cacf275047936cdb67be78370751111a1;hpb=d753c060099e12655344fc461e4b6f7142b3b5a1

diff --git a/lustre/tests/test-framework.sh b/lustre/tests/test-framework.sh
index 42e51a1..bc8de5c 100644
--- a/lustre/tests/test-framework.sh
+++ b/lustre/tests/test-framework.sh
@@ -1,15 +1,14 @@
 #!/bin/bash
 # vim:expandtab:shiftwidth=4:softtabstop=4:tabstop=4:
 
-trap 'print_summary && echo "test-framework exiting on error"' ERR
+trap 'print_summary && touch $TF_FAIL && \
+    echo "test-framework exiting on error"' ERR
 set -e
 #set -x
 
-
 export REFORMAT=${REFORMAT:-""}
 export WRITECONF=${WRITECONF:-""}
 export VERBOSE=false
-export GMNALNID=${GMNALNID:-/usr/sbin/gmlndnid}
 export CATASTROPHE=${CATASTROPHE:-/proc/sys/lnet/catastrophe}
 export GSS=false
 export GSS_KRB5=false
@@ -22,6 +21,7 @@ export QUOTA_AUTO=1
 # function used by scripts run on remote nodes
 LUSTRE=${LUSTRE:-$(cd $(dirname $0)/..; echo $PWD)}
 . $LUSTRE/tests/functions.sh
+. $LUSTRE/tests/yaml.sh
 
 LUSTRE_TESTS_CFG_DIR=${LUSTRE_TESTS_CFG_DIR:-${LUSTRE}/tests/cfg}
 
@@ -33,6 +33,9 @@ if [ -f "$EXCEPT_LIST_FILE" ]; then
     . $EXCEPT_LIST_FILE
 fi
 
+[ -z "$MODPROBECONF" -a -f /etc/modprobe.conf ] && MODPROBECONF=/etc/modprobe.conf
+[ -z "$MODPROBECONF" -a -f /etc/modprobe.d/Lustre ] && MODPROBECONF=/etc/modprobe.d/Lustre
+
 assert_DIR () {
     local failed=""
     [[ $DIR/ = $MOUNT/* ]] || \
@@ -54,43 +57,51 @@ usage() {
 
 print_summary () {
     trap 0
-    [ "$TESTSUITE" == "lfscktest" ] && return 0
-    [ -n "$ONLY" ] && echo "WARNING: ONLY is set to ${ONLY}."
-    local form="%-13s %-17s %s\n"
-    printf "$form" "status" "script" "skipped tests E(xcluded) S(low)"
+    [ "$TESTSUITE" == "lfsck" ] && return 0
+    [ -n "$ONLY" ] && echo "WARNING: ONLY is set to $(echo $ONLY)"
+    local details
+    local form="%-13s %-17s %-9s %s %s\n"
+    printf "$form" "status" "script" "Total(sec)" "E(xcluded) S(low)"
     echo "------------------------------------------------------------------------------------"
-    for O in $TESTSUITE_LIST; do
-        local skipped=""
-        local slow=""
-        local o=$(echo $O | tr "[:upper:]" "[:lower:]")
-        o=${o//_/-}
-        o=${o//tyn/tyN}
+    for O in $DEFAULT_SUITES; do
+        O=$(echo $O  | tr "-" "_" | tr "[:lower:]" "[:upper:]")
+        [ "${!O}" = "no" ] && continue || true
+        local o=$(echo $O  | tr "[:upper:]_" "[:lower:]-")
         local log=${TMP}/${o}.log
-        [ -f $log ] && skipped=$(grep excluded $log | awk '{ printf " %s", $3 }' | sed 's/test_//g')
-        [ -f $log ] && slow=$(grep SLOW $log | awk '{ printf " %s", $3 }' | sed 's/test_//g')
-        [ "${!O}" = "done" ] && \
-            printf "$form" "Done" "$O" "E=$skipped" && \
-            [ -n "$slow" ] && printf "$form" "-" "-" "S=$slow"
-
+        if is_sanity_benchmark $o; then
+            log=${TMP}/sanity-benchmark.log
+        fi
+        local slow=
+        local skipped=
+        local total=
+        local status=Unfinished
+        if [ -f $log ]; then
+            skipped=$(grep excluded $log | awk '{ printf " %s", $3 }' | sed 's/test_//g')
+            slow=$(egrep "^PASS|^FAIL" $log | tr -d "("| sed s/s\)$//g | sort -nr -k 3  | head -5 |  awk '{ print $2":"$3"s" }')
+            total=$(grep duration $log | awk '{ print $2}')
+            if [ "${!O}" = "done" ]; then
+                status=Done
+            fi
+            if $DDETAILS; then
+                local durations=$(egrep "^PASS|^FAIL" $log |  tr -d "("| sed s/s\)$//g | awk '{ print $2":"$3"|" }')
+                details=$(printf "%s\n%s %s %s\n" "$details" "DDETAILS" "$O" "$(echo $durations)")
+            fi
+        fi
+        printf "$form" $status "$O" "${total}" "E=$skipped"
+        printf "$form" "-" "-" "-" "S=$(echo $slow)"
     done
 
-    for O in $TESTSUITE_LIST; do
+    for O in $DEFAULT_SUITES; do
+        O=$(echo $O  | tr "-" "_" | tr "[:lower:]" "[:upper:]")
         if [ "${!O}" = "no" ]; then
-            # FIXME.
-            # only for those tests suits which are run directly from acc-sm script:
-            # bonnie, iozone, etc.
-            if [ -f "$TESTSUITELOG" ] && grep FAIL $TESTSUITELOG | grep -q ' '$O  ; then
-               printf "$form" "UNFINISHED" "$O" ""  
-            else
-               printf "$form" "Skipped" "$O" ""
-            fi
+            printf "$form" "Skipped" "$O" ""
         fi
     done
 
-    for O in $TESTSUITE_LIST; do
-        [ "${!O}" = "done" -o "${!O}" = "no" ] || \
-            printf "$form" "UNFINISHED" "$O" ""
-    done
+    # print the detailed tests durations if DDETAILS=true
+    if $DDETAILS; then
+        echo "$details"
+    fi
 }
 
 init_test_env() {
@@ -105,10 +116,29 @@ init_test_env() {
     export E2LABEL=${E2LABEL:-e2label}
     export DUMPE2FS=${DUMPE2FS:-dumpe2fs}
     export E2FSCK=${E2FSCK:-e2fsck}
+    export LFSCK_BIN=${LFSCK_BIN:-lfsck}
+    export LFSCK_ALWAYS=${LFSCK_ALWAYS:-"no"} # check filesystem after each test suit
+    export SKIP_LFSCK=${SKIP_LFSCK:-"yes"} # bug 13698, change to "no" when fixed
+    export SHARED_DIRECTORY=${SHARED_DIRECTORY:-"/tmp"}
+    export FSCK_MAX_ERR=4   # File system errors left uncorrected
+    if [ "$SKIP_LFSCK" == "no" ]; then
+        if [ ! -x `which $LFSCK_BIN` ]; then
+            log "$($E2FSCK -V)"
+            error_exit "$E2FSCK does not support lfsck"
+        fi
 
+        export MDSDB=${MDSDB:-$SHARED_DIRECTORY/mdsdb}
+        export OSTDB=${OSTDB:-$SHARED_DIRECTORY/ostdb}
+        export MDSDB_OPT="--mdsdb $MDSDB"
+        export OSTDB_OPT="--ostdb $OSTDB-\$ostidx"
+    fi
     #[ -d /r ] && export ROOT=${ROOT:-/r}
     export TMP=${TMP:-$ROOT/tmp}
     export TESTSUITELOG=${TMP}/${TESTSUITE}.log
+    if [[ -z $LOGDIRSET ]]; then
+        export LOGDIR=${LOGDIR:-${TMP}/test_logs/}/$(date +%s)
+        export LOGDIRSET=true
+    fi
     export HOSTNAME=${HOSTNAME:-`hostname`}
     if ! echo $PATH | grep -q $LUSTRE/utils; then
         export PATH=$PATH:$LUSTRE/utils
@@ -117,12 +147,22 @@ init_test_env() {
         export PATH=$PATH:$LUSTRE/utils/gss
     fi
     if ! echo $PATH | grep -q $LUSTRE/tests; then
-	export PATH=$PATH:$LUSTRE/tests
+        export PATH=$LUSTRE/tests:$PATH
     fi
+    if ! echo $PATH | grep -q $LUSTRE/../lustre-iokit/sgpdd-survey; then
+        export PATH=$PATH:$LUSTRE/../lustre-iokit/sgpdd-survey
+    fi
+    export LST=${LST:-"$LUSTRE/../lnet/utils/lst"}
+    [ ! -f "$LST" ] && export LST=$(which lst)
+    export SGPDDSURVEY=${SGPDDSURVEY:-"$LUSTRE/../lustre-iokit/sgpdd-survey/sgpdd-survey")}
+    [ ! -f "$SGPDDSURVEY" ] && export SGPDDSURVEY=$(which sgpdd-survey)
+    # Ubuntu, at least, has a truncate command in /usr/bin
+    # so fully path our truncate command.
+    export TRUNCATE=${TRUNCATE:-$LUSTRE/tests/truncate}
     export MDSRATE=${MDSRATE:-"$LUSTRE/tests/mpi/mdsrate"}
     [ ! -f "$MDSRATE" ] && export MDSRATE=$(which mdsrate 2> /dev/null)
     if ! echo $PATH | grep -q $LUSTRE/tests/racer; then
-        export PATH=$PATH:$LUSTRE/tests/racer
+        export PATH=$LUSTRE/tests/racer:$PATH:
     fi
     if ! echo $PATH | grep -q $LUSTRE/tests/mpi; then
         export PATH=$PATH:$LUSTRE/tests/mpi
@@ -140,6 +180,8 @@ init_test_env() {
             export L_GETIDENTITY=NONE
         fi
     fi
+    export LL_DECODE_FILTER_FID=${LL_DECODE_FILTER_FID:-"$LUSTRE/utils/ll_decode_filter_fid"}
+    [ ! -f "$LL_DECODE_FILTER_FID" ] && export LL_DECODE_FILTER_FID=$(which ll_decode_filter_fid)
     export MKFS=${MKFS:-"$LUSTRE/utils/mkfs.lustre"}
     [ ! -f "$MKFS" ] && export MKFS=$(which mkfs.lustre)
     export TUNEFS=${TUNEFS:-"$LUSTRE/utils/tunefs.lustre"}
@@ -167,7 +209,7 @@ init_test_env() {
         xkrb5*)
             echo "Using GSS/krb5 ptlrpc security flavor"
             which lgss_keyring > /dev/null 2>&1 || \
-                error "built with gss disabled! SEC=$SEC"
+                error_exit "built with gss disabled! SEC=$SEC"
             GSS=true
             GSS_KRB5=true
             ;;
@@ -188,10 +230,13 @@ init_test_env() {
     export RPWD=${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
 
+    export SHUTDOWN_ATTEMPTS=${SHUTDOWN_ATTEMPTS:-3}
+
     # command line
 
     while getopts "rvwf:" opt $*; do
@@ -207,6 +252,8 @@ 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
 }
@@ -221,7 +268,16 @@ module_loaded () {
    /sbin/lsmod | grep -q $1
 }
 
+# Load a module on the system where this is running.
+#
+# Synopsis: load_module module_name [module arguments for insmod/modprobe]
+#
+# If module arguments are not given but MODOPTS_<MODULE> is set, then its value
+# will be used as the arguments.  Otherwise arguments will be obtained from
+# /etc/modprobe.conf, from /etc/modprobe.d/Lustre, or else none will be used.
+#
 load_module() {
+    local optvar
     EXT=".ko"
     module=$1
     shift
@@ -229,15 +285,47 @@ load_module() {
 
     module_loaded ${BASE} && return
 
-    if [ -f ${LUSTRE}/${module}${EXT} ]; then
-        insmod ${LUSTRE}/${module}${EXT} $@
+    # If no module arguments were passed, get them from $MODOPTS_<MODULE>, else from
+    # modprobe.conf
+    if [ $# -eq 0 ]; then
+        # $MODOPTS_<MODULE>; we could use associative arrays, but that's not in
+        # Bash until 4.x, so we resort to eval.
+        optvar="MODOPTS_$(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 -- $(grep "^options\\s*\<${module}\>" $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
+        fi
+    fi
+
+    [ $# -gt 0 ] && echo "${module} options: '$*'"
+
+    # Note that insmod will ignore anything in modprobe.conf, which is why we're
+    # passing options on the command-line.
+    if [ "$BASE" == "lnet_selftest" ] && \
+            [ -f ${LUSTRE}/../lnet/selftest/${module}${EXT} ]; then
+        insmod ${LUSTRE}/../lnet/selftest/${module}${EXT}
+    elif [ -f ${LUSTRE}/${module}${EXT} ]; then
+        insmod ${LUSTRE}/${module}${EXT} "$@"
     else
         # must be testing a "make install" or "rpm" installation
         # note failed to load ptlrpc_gss is considered not fatal
         if [ "$BASE" == "ptlrpc_gss" ]; then
-            modprobe $BASE $@ 2>/dev/null || echo "gss/krb5 is not supported"
+            modprobe $BASE "$@" 2>/dev/null || echo "gss/krb5 is not supported"
         else
-            modprobe $BASE $@
+            modprobe $BASE "$@"
         fi
     fi
 }
@@ -245,34 +333,22 @@ load_module() {
 load_modules_local() {
     if [ -n "$MODPROBE" ]; then
         # use modprobe
-    return 0
-    fi
-    if [ "$HAVE_MODULES" = true ]; then
-    # we already loaded
+        echo "Using modprobe to load modules"
         return 0
     fi
-    HAVE_MODULES=true
 
     echo Loading modules from $LUSTRE
     load_module ../libcfs/libcfs/libcfs
     [ "$PTLDEBUG" ] && lctl set_param debug="$PTLDEBUG"
     [ "$SUBSYSTEM" ] && lctl set_param subsystem_debug="${SUBSYSTEM# }"
-    local MODPROBECONF=
-    [ -f /etc/modprobe.conf ] && MODPROBECONF=/etc/modprobe.conf
-    [ ! "$MODPROBECONF" -a -d /etc/modprobe.d ] && MODPROBECONF=/etc/modprobe.d/Lustre
-    [ -z "$LNETOPTS" -a "$MODPROBECONF" ] && \
-        LNETOPTS=$(awk '/^options lnet/ { print $0}' $MODPROBECONF | sed 's/^options lnet //g')
-    echo $LNETOPTS | grep -q "accept=all"  || LNETOPTS="$LNETOPTS accept=all";
-    echo "lnet options: '$LNETOPTS'"
-    # note that insmod will ignore anything in modprobe.conf
-    load_module ../lnet/lnet/lnet $LNETOPTS
+    load_module ../lnet/lnet/lnet
     LNETLND=${LNETLND:-"socklnd/ksocklnd"}
     load_module ../lnet/klnds/$LNETLND
     load_module lvfs/lvfs
     load_module obdclass/obdclass
     load_module ptlrpc/ptlrpc
     load_module ptlrpc/gss/ptlrpc_gss
-    [ "$USE_QUOTA" = "yes" -a "$LQUOTA" != "no" ] && load_module quota/lquota
+    [ "$USE_QUOTA" = "yes" -a "$LQUOTA" != "no" ] && load_module quota/lquota $LQUOTAOPTS
     load_module fld/fld
     load_module fid/fid
     load_module lmv/lmv
@@ -282,7 +358,8 @@ load_modules_local() {
     load_module mgc/mgc
     if ! client_only; then
         grep -q crc16 /proc/kallsyms || { modprobe crc16 2>/dev/null || true; }
-        grep -q jbd /proc/kallsyms || { modprobe jbd 2>/dev/null || true; }
+        grep -q -w jbd /proc/kallsyms || { modprobe jbd 2>/dev/null || true; }
+        grep -q -w jbd2 /proc/kallsyms || { modprobe jbd2 2>/dev/null || true; }
         [ "$FSTYPE" = "ldiskfs" ] && load_module ../ldiskfs/ldiskfs/ldiskfs
         load_module mgs/mgs
         load_module mds/mds
@@ -290,7 +367,7 @@ load_modules_local() {
         load_module mdt/mdt
         load_module lvfs/fsfilt_$FSTYPE
         load_module cmm/cmm
-        load_module osd/osd
+        load_module osd-ldiskfs/osd_ldiskfs
         load_module ost/ost
         load_module obdfilter/obdfilter
     fi
@@ -338,19 +415,30 @@ unload_modules() {
 
     if $LOAD_MODULES_REMOTE ; then
         local list=$(comma_list $(remote_nodes_list))
-        echo unloading modules on $list
-        do_rpc_nodes $list $LUSTRE_RMMOD $FSTYPE
-        do_rpc_nodes $list check_mem_leak
+        if [ ! -z $list ]; then
+            echo unloading modules on $list
+            do_rpc_nodes $list $LUSTRE_RMMOD $FSTYPE
+            do_rpc_nodes $list check_mem_leak
+        fi
     fi
 
-    HAVE_MODULES=false
-
     check_mem_leak || return 254
 
     echo "modules unloaded."
     return 0
 }
 
+check_gss_daemon_nodes() {
+    local list=$1
+    dname=$2
+
+    do_nodesv $list "num=\\\$(ps -o cmd -C $dname | grep $dname | wc -l);
+if [ \\\"\\\$num\\\" -ne 1 ]; then
+    echo \\\$num instance of $dname;
+    exit 1;
+fi; "
+}
+
 check_gss_daemon_facet() {
     facet=$1
     dname=$2
@@ -364,27 +452,41 @@ check_gss_daemon_facet() {
 }
 
 send_sigint() {
-    local facet=$1
+    local list=$1
     shift
-    do_facet $facet "killall -2 $@ 2>/dev/null || true"
+    echo Stopping $@ on $list
+    do_nodes $list "killall -2 $@ 2>/dev/null || true"
 }
 
+# start gss daemons on all nodes, or
+# "daemon" on "list" if set
 start_gss_daemons() {
-    # starting on MDT
-    for num in `seq $MDSCOUNT`; do
-        do_facet mds$num "$LSVCGSSD -v"
-        if $GSS_PIPEFS; then
-            do_facet mds$num "$LGSSD -v"
-        fi
-    done
-    # starting on OSTs
-    for num in `seq $OSTCOUNT`; do
-        do_facet ost$num "$LSVCGSSD -v"
-    done
-    # starting on client
-    # FIXME: is "client" the right facet name?
+    local list=$1
+    local daemon=$2
+
+    if [ "$list" ] && [ "$daemon" ] ; then
+        echo "Starting gss daemon on nodes: $list"
+        do_nodes $list "$daemon" || return 8
+        return 0
+    fi
+
+    local list=$(comma_list $(mdts_nodes))
+
+    echo "Starting gss daemon on mds: $list"
+    do_nodes $list "$LSVCGSSD -v" || return 1
+    if $GSS_PIPEFS; then
+        do_nodes $list "$LGSSD -v" || return 2
+    fi
+
+    list=$(comma_list $(osts_nodes))
+    echo "Starting gss daemon on ost: $list"
+    do_nodes $list "$LSVCGSSD -v" || return 3
+    # starting on clients
+
+    local clients=${CLIENTS:-`hostname`}
     if $GSS_PIPEFS; then
-        do_facet client "$LGSSD -v"
+        echo "Starting $LGSSD on clients $clients "
+        do_nodes $clients  "$LGSSD -v" || return 4
     fi
 
     # wait daemons entering "stable" status
@@ -393,33 +495,37 @@ start_gss_daemons() {
     #
     # check daemons are running
     #
-    for num in `seq $MDSCOUNT`; do
-        check_gss_daemon_facet mds$num lsvcgssd
-        if $GSS_PIPEFS; then
-            check_gss_daemon_facet mds$num lgssd
-        fi
-    done
-    for num in `seq $OSTCOUNT`; do
-        check_gss_daemon_facet ost$num lsvcgssd
-    done
+    list=$(comma_list $(mdts_nodes) $(osts_nodes))
+    check_gss_daemon_nodes $list lsvcgssd || return 5
+    if $GSS_PIPEFS; then
+        list=$(comma_list $(mdts_nodes))
+        check_gss_daemon_nodes $list lgssd || return 6
+    fi
     if $GSS_PIPEFS; then
-        check_gss_daemon_facet client lgssd
+        check_gss_daemon_nodes $clients lgssd || return 7
     fi
 }
 
 stop_gss_daemons() {
-    for num in `seq $MDSCOUNT`; do
-        send_sigint mds$num lsvcgssd lgssd
-    done
-    for num in `seq $OSTCOUNT`; do
-        send_sigint ost$num lsvcgssd
-    done
-    send_sigint client lgssd
+    local list=$(comma_list $(mdts_nodes))
+    
+    send_sigint $list lsvcgssd lgssd
+
+    list=$(comma_list $(osts_nodes))
+    send_sigint $list lsvcgssd
+
+    list=${CLIENTS:-`hostname`}
+    send_sigint $list lgssd
 }
 
 init_gss() {
     if $GSS; then
-        start_gss_daemons
+        if ! module_loaded ptlrpc_gss; then
+            load_module ptlrpc/gss/ptlrpc_gss
+            module_loaded ptlrpc_gss ||
+                error_exit "init_gss : GSS=$GSS, but gss/krb5 is not supported!"
+        fi
+        start_gss_daemons || error_exit "start gss daemon failed! rc=$?"
 
         if [ -n "$LGSS_KEYRING_DEBUG" ]; then
             echo $LGSS_KEYRING_DEBUG > /proc/fs/lustre/sptlrpc/gss/lgss_keyring/debug_level
@@ -449,16 +555,27 @@ ostdevlabel() {
 }
 
 # Facet functions
+mount_facets () {
+    local facets=${1:-$(get_facets)}
+    local facet
+
+    for facet in ${facets//,/ }; do
+        mount_facet $facet || error "Restart of $facet failed!"
+    done
+}
+
 mount_facet() {
     local facet=$1
     shift
     local dev=$(facet_active $facet)_dev
     local opt=${facet}_opt
-    echo "Starting ${facet}: ${!opt} $@ ${!dev} ${MOUNT%/*}/${facet}"
-    do_facet ${facet} mount -t lustre ${!opt} $@ ${!dev} ${MOUNT%/*}/${facet}
+    local mntpt=$(facet_mntpt $facet)
+
+    echo "Starting ${facet}: ${!opt} $@ ${!dev} $mntpt"
+    do_facet ${facet} "mkdir -p $mntpt; mount -t lustre ${!opt} $@ ${!dev} $mntpt"
     RC=${PIPESTATUS[0]}
     if [ $RC -ne 0 ]; then
-        echo "mount -t lustre $@ ${!dev} ${MOUNT%/*}/${facet}"
+        echo "mount -t lustre $@ ${!dev} $mntpt"
         echo "Start of ${!dev} on ${facet} failed ${RC}"
     else
         do_facet ${facet} "lctl set_param debug=\\\"$PTLDEBUG\\\"; \
@@ -490,7 +607,9 @@ start() {
         eval export ${facet}failover_dev=$device
     fi
 
-    do_facet ${facet} mkdir -p ${MOUNT%/*}/${facet}
+    local mntpt=$(facet_mntpt $facet)
+    do_facet ${facet} mkdir -p $mntpt
+    eval export ${facet}_MOUNT=$mntpt
     mount_facet ${facet}
     RC=$?
     return $RC
@@ -500,13 +619,14 @@ stop() {
     local running
     local facet=$1
     shift
-    HOST=`facet_active_host $facet`
+    local HOST=`facet_active_host $facet`
     [ -z $HOST ] && echo stop: no host for $facet && return 0
 
-    running=$(do_facet ${facet} "grep -c ${MOUNT%/*}/${facet}' ' /proc/mounts") || true
+    local mntpt=$(facet_mntpt $facet)
+    running=$(do_facet ${facet} "grep -c $mntpt' ' /proc/mounts") || true
     if [ ${running} -ne 0 ]; then
-        echo "Stopping ${MOUNT%/*}/${facet} (opts:$@)"
-        do_facet ${facet} umount -d $@ ${MOUNT%/*}/${facet}
+        echo "Stopping $mntpt (opts:$@)"
+        do_facet ${facet} umount -d $@ $mntpt
     fi
 
     # umount should block, but we should wait for unrelated obd's
@@ -600,13 +720,10 @@ setup_quota(){
 }
 
 zconf_mount() {
-    local OPTIONS
     local client=$1
     local mnt=$2
-    # Only supply -o to mount if we have options
-    if [ -n "$MOUNTOPT" ]; then
-        OPTIONS="-o $MOUNTOPT"
-    fi
+    local OPTIONS=${3:-$MOUNTOPT}
+
     local device=$MGSNID:/$FSNAME
     if [ -z "$mnt" -o -z "$FSNAME" ]; then
         echo Bad zconf mount command: opt=$OPTIONS dev=$device mnt=$mnt
@@ -682,12 +799,12 @@ sanity_mount_check_servers () {
     # FIXME: modify get_facets to display all facets wo params
     local facets="$(get_facets OST),$(get_facets MDS),mgs"
     local node
-    local mnt
+    local mntpt
     local facet
     for facet in ${facets//,/ }; do
         node=$(facet_host ${facet})
-        mnt=${MOUNT%/*}/${facet}
-        sanity_mount_check_nodes $node $mnt ||
+        mntpt=$(facet_mntpt $facet)
+        sanity_mount_check_nodes $node $mntpt ||
             { error "server $node environments are insane!"; return 1; }
     done
 }
@@ -715,10 +832,6 @@ zconf_mount_clients() {
     local mnt=$2
     local OPTIONS=${3:-$MOUNTOPT}
 
-    # Only supply -o to mount if we have options
-    if [ "$OPTIONS" ]; then
-        OPTIONS="-o $OPTIONS"
-    fi
     local device=$MGSNID:/$FSNAME
     if [ -z "$mnt" -o -z "$FSNAME" ]; then
         echo Bad zconf mount command: opt=$OPTIONS dev=$device mnt=$mnt
@@ -773,14 +886,20 @@ fi
 fi"
 }
 
-shudown_node_hard () {
+shutdown_node () {
+    local node=$1
+    echo + $POWER_DOWN $node
+    $POWER_DOWN $node
+}
+
+shutdown_node_hard () {
     local host=$1
-    local attempts=3
+    local attempts=$SHUTDOWN_ATTEMPTS
 
     for i in $(seq $attempts) ; do
-        $POWER_DOWN $host
+        shutdown_node $host
         sleep 1
-        ping -w 3 -c 1 $host > /dev/null 2>&1 || return 0
+        wait_for_function --quiet "! ping -w 3 -c 1 $host" 5 1 && return 0
         echo "waiting for $host to fail attempts=$attempts"
         [ $i -lt $attempts ] || \
             { echo "$host still pingable after power down! attempts=$attempts" && return 1; } 
@@ -793,25 +912,77 @@ shutdown_client() {
     local attempts=3
 
     if [ "$FAILURE_MODE" = HARD ]; then
-        shudown_node_hard $client 
+        shutdown_node_hard $client
     else
        zconf_umount_clients $client $mnt -f
     fi
 }
 
+facets_on_host () {
+    local host=$1
+    local facets="$(get_facets OST),$(get_facets MDS)"
+    local affected
+
+    combined_mgs_mds || facets="$facets,mgs"
+
+    for facet in ${facets//,/ }; do
+        if [ $(facet_active_host $facet) == $host ]; then
+           affected="$affected $facet"
+        fi
+    done
+
+    echo $(comma_list $affected)
+}
+
+facet_up () {
+    local facet=$1
+    local host=${2:-$(facet_host $facet)}
+
+    local label=$(convert_facet2label $facet)
+    do_node $host lctl dl | awk '{print $4}' | grep -q $label
+}
+
+facets_up_on_host () {
+    local host=$1
+    local facets=$(facets_on_host $host)
+    local affected_up
+
+    for facet in ${facets//,/ }; do
+        if $(facet_up $facet $host); then
+            affected_up="$affected_up $facet"
+        fi
+    done
+
+    echo $(comma_list $affected_up)
+}
+
 shutdown_facet() {
     local facet=$1
+
     if [ "$FAILURE_MODE" = HARD ]; then
-        shudown_node_hard $(facet_active_host $facet)
-    elif [ "$FAILURE_MODE" = SOFT ]; then
+        shutdown_node_hard $(facet_active_host $facet)
+    else
         stop $facet
     fi
 }
 
+reboot_node() {
+    local node=$1
+    echo + $POWER_UP $node
+    $POWER_UP $node
+}
+
+remount_facet() {
+    local facet=$1
+
+    stop $facet
+    mount_facet $facet
+}
+
 reboot_facet() {
-    facet=$1
+    local facet=$1
     if [ "$FAILURE_MODE" = HARD ]; then
-        $POWER_UP `facet_active_host $facet`
+        reboot_node $(facet_active_host $facet)
     else
         sleep 10
     fi
@@ -820,27 +991,43 @@ reboot_facet() {
 boot_node() {
     local node=$1
     if [ "$FAILURE_MODE" = HARD ]; then
-       $POWER_UP $node
+       reboot_node $node
        wait_for_host $node
     fi
 }
 
-# recovery-scale functions
+facets_hosts () {
+    local facets=$1
+    local hosts
+
+    for facet in ${facets//,/ }; do
+        hosts=$(expand_list $hosts $(facet_host $facet) )
+    done
+
+    echo $hosts
+}
+
+_check_progs_installed () {
+    local progs=$@
+    local rc=0
+
+    for prog in $progs; do
+        if ! [ "$(which $prog)"  -o  "${!prog}" ]; then
+           echo $prog missing on $(hostname)
+           rc=1
+        fi
+    done
+    return $rc
+}
+
 check_progs_installed () {
-    local clients=$1
+    local nodes=$1
     shift
-    local progs=$@
 
-    do_nodes $clients "PATH=:$PATH; status=true;
-for prog in $progs; do
-    if ! [ \\\"\\\$(which \\\$prog)\\\"  -o  \\\"\\\${!prog}\\\" ]; then
-       echo \\\$prog missing on \\\$(hostname);
-       status=false;
-    fi
-done;
-eval \\\$status"
+    do_rpc_nodes $nodes _check_progs_installed $@
 }
 
+# recovery-scale functions
 client_var_name() {
     echo __$(echo $1 | tr '-' 'X')
 }
@@ -872,6 +1059,8 @@ start_client_loads () {
         testnum=$((nodenum % numloads))
         start_client_load ${clients[nodenum]} ${CLIENT_LOADS[testnum]}
     done
+    # bug 22169: wait the background threads to start
+    sleep 2
 }
 
 # only for remote client 
@@ -1042,42 +1231,68 @@ wait_delete_completed () {
 }
 
 wait_for_host() {
-    local host=$1
-    check_network "$host" 900
-    while ! do_node $host hostname  > /dev/null; do sleep 5; done
+    local hostlist=$1
+
+    # we can use "for" here because we are waiting the slowest
+    for host in ${hostlist//,/ }; do
+        check_network "$host" 900
+    done
+    while ! do_nodes $hostlist hostname  > /dev/null; do sleep 5; done
 }
 
-wait_for() {
-    local facet=$1
-    local host=`facet_active_host $facet`
-    wait_for_host $host
+wait_for_facet() {
+    local facetlist=$1
+    local hostlist
+
+    for facet in ${facetlist//,/ }; do
+        hostlist=$(expand_list $hostlist $(facet_active_host $facet))
+    done
+    wait_for_host $hostlist
 }
 
-wait_recovery_complete () {
-    local facet=$1
+_wait_recovery_complete () {
+    local param=$1
 
     # Use default policy if $2 is not passed by caller.
-    #define OBD_RECOVERY_TIMEOUT (obd_timeout * 5 / 2)
-    # as we are in process of changing obd_timeout in different ways
-    # let's set MAX longer than that
-    local MAX=${2:-$(( TIMEOUT * 4 ))}
+    local MAX=${2:-$(max_recovery_time)}
 
-    local var_svc=${facet}_svc
-    local procfile="*.${!var_svc}.recovery_status"
     local WAIT=0
     local STATUS=
 
     while [ $WAIT -lt $MAX ]; do
-        STATUS=$(do_facet $facet lctl get_param -n $procfile | grep status)
-        [[ $STATUS = "status: COMPLETE" ]] && return 0
+        STATUS=$(lctl get_param -n $param | grep status)
+        echo $param $STATUS
+        [[ $STATUS = "status: COMPLETE" || $STATUS = "status: INACTIVE" ]] && return 0
         sleep 5
         WAIT=$((WAIT + 5))
-        echo "Waiting $((MAX - WAIT)) secs for $facet recovery done. $STATUS"
+        echo "Waiting $((MAX - WAIT)) secs for $param recovery done. $STATUS"
     done
-    echo "$facet recovery not done in $MAX sec. $STATUS"
+    echo "$param recovery not done in $MAX sec. $STATUS"
     return 1
 }
 
+wait_recovery_complete () {
+    local facet=$1
+
+    # with an assumption that at_max is the same on all nodes
+    local MAX=${2:-$(max_recovery_time)}
+
+    local facets=$facet
+    if [ "$FAILURE_MODE" = HARD ]; then
+        facets=$(facets_on_host $(facet_active_host $facet))
+    fi
+    echo affected facets: $facets
+
+    # we can use "for" here because we are waiting the slowest
+    for facet in ${facets//,/ }; do
+        local var_svc=${facet}_svc
+        local param="*.${!var_svc}.recovery_status"
+
+        local host=$(facet_active_host $facet)
+        do_rpc_nodes $host _wait_recovery_complete $param $MAX
+    done
+}
+
 wait_mds_ost_sync () {
     # just because recovery is done doesn't mean we've finished
     # orphan cleanup. Wait for llogs to get synchronized.
@@ -1089,6 +1304,7 @@ wait_mds_ost_sync () {
         local -a sync=($(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
@@ -1220,18 +1436,45 @@ client_reconnect() {
     rm $MOUNT/recon
 }
 
+affected_facets () {
+    local facet=$1
+
+    local host=$(facet_active_host $facet)
+    local affected=$facet
+
+    if [ "$FAILURE_MODE" = HARD ]; then
+        affected=$(facets_up_on_host $host)
+    fi
+    echo $affected
+}
+
 facet_failover() {
     local facet=$1
     local sleep_time=$2
-    echo "Failing $facet on node `facet_active_host $facet`"
+    local host=$(facet_active_host $facet)
+
+    echo "Failing $facet on node $host"
+
+    local affected=$(affected_facets $facet)
+
     shutdown_facet $facet
+
+    echo affected facets: $affected
+
     [ -n "$sleep_time" ] && sleep $sleep_time
+
     reboot_facet $facet
-    change_active $facet
-    local TO=`facet_active_host $facet`
-    echo "Failover $facet to $TO"
-    wait_for $facet
-    mount_facet $facet || error "Restart of $facet failed"
+
+    change_active $affected
+
+    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
 }
 
 obd_name() {
@@ -1242,6 +1485,11 @@ replay_barrier() {
     local facet=$1
     do_facet $facet sync
     df $MOUNT
+
+    # make sure there will be no seq change
+    local clients=${CLIENTS:-$HOSTNAME}
+    do_nodes $clients "f=${MOUNT}/fsa-\\\$(hostname); mcreate \\\$f; rm \\\$f"
+
     local svc=${facet}_svc
     do_facet $facet $LCTL --device %${!svc} notransno
     do_facet $facet $LCTL --device %${!svc} readonly
@@ -1304,13 +1552,6 @@ do_lmc() {
     exit 1
 }
 
-h2gm () {
-    if [ "$1" = "client" -o "$1" = "'*'" ]; then echo \'*\'; else
-        ID=`$PDSH $1 $GMNALNID -l | cut -d\  -f2`
-        echo $ID"@gm"
-    fi
-}
-
 h2name_or_ip() {
     if [ "$1" = "client" -o "$1" = "'*'" ]; then echo \'*\'; else
         echo $1"@$2"
@@ -1346,11 +1587,6 @@ h2elan() {
 }
 declare -fx h2elan
 
-h2openib() {
-    h2name_or_ip "$1" "openib"
-}
-declare -fx h2openib
-
 h2o2ib() {
     h2name_or_ip "$1" "o2ib"
 }
@@ -1396,10 +1632,16 @@ facet_active_host() {
 }
 
 change_active() {
-    local facet=$1
+    local facetlist=$1
+    local facet
+
+    facetlist=$(exclude_items_from_list $facetlist mgs)
+
+    for facet in ${facetlist//,/ }; do
     local failover=${facet}failover
-    host=`facet_host $failover`
+    local host=`facet_host $failover`
     [ -z "$host" ] && return
+
     local curactive=`facet_active $facet`
     if [ -z "${curactive}" -o "$curactive" == "$failover" ] ; then
         eval export ${facet}active=$facet
@@ -1409,6 +1651,9 @@ change_active() {
     # save the active host for this facet
     local activevar=${facet}active
     echo "$activevar=${!activevar}" > $TMP/$activevar
+    local TO=`facet_active_host $facet`
+    echo "Failover $facet to $TO"
+    done
 }
 
 do_node() {
@@ -1435,32 +1680,47 @@ do_node() {
 
     if [ "$myPDSH" = "rsh" ]; then
 # we need this because rsh does not return exit code of an executed command
-	local command_status="$TMP/cs"
-	rsh $HOST ":> $command_status"
-	rsh $HOST "(PATH=\$PATH:$RLUSTRE/utils:$RLUSTRE/tests:/sbin:/usr/sbin;
-		    cd $RPWD; sh -c \"$@\") ||
-		    echo command failed >$command_status"
-	[ -n "$($myPDSH $HOST cat $command_status)" ] && return 1 || true
+        local command_status="$TMP/cs"
+        rsh $HOST ":> $command_status"
+        rsh $HOST "(PATH=\$PATH:$RLUSTRE/utils:$RLUSTRE/tests:/sbin:/usr/sbin;
+                    cd $RPWD; LUSTRE=\"$RLUSTRE\" sh -c \"$@\") ||
+                    echo command failed >$command_status"
+        [ -n "$($myPDSH $HOST cat $command_status)" ] && return 1 || true
         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 $RPWD; sh -c \"$@\")" | sed -e "s/^/${HOSTNAME}: /"
+            $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 $RPWD; sh -c \"$@\")"
+            $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 $RPWD; sh -c \"$@\")" | sed "s/^${HOST}: //"
+        $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 "$@"
+}
+
 single_local_node () {
    [ "$1" = "$HOSTNAME" ]
 }
 
+# Outputs environment variable assignments that should be passed to remote nodes
+get_env_vars() {
+    local var
+    local value
+
+    for var in ${!MODOPTS_*}; do
+        value=${!var}
+        echo "${var}=\"$value\""
+    done
+}
+
 do_nodes() {
     local verbose=false
     # do not stripe off hostname if verbose, bug 19215
@@ -1472,11 +1732,11 @@ do_nodes() {
     local rnodes=$1
     shift
 
-    if $(single_local_node $rnodes); then
+    if single_local_node $rnodes; then
         if $verbose; then
-           do_node --verbose $rnodes $@
+           do_nodev $rnodes "$@"
         else
-           do_node $rnodes $@
+           do_node $rnodes "$@"
         fi
         return $?
     fi
@@ -1492,10 +1752,12 @@ do_nodes() {
         $myPDSH $rnodes $LCTL mark "$@" > /dev/null 2>&1 || :
     fi
 
-    if $verbose ; then
-        $myPDSH $rnodes "(PATH=\$PATH:$RLUSTRE/utils:$RLUSTRE/tests:/sbin:/usr/sbin; cd $RPWD; sh -c \"$@\")"
+    # 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 $RPWD; sh -c \"$@\")" | sed -re "s/\w+:\s//g"
+        $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]}
 }
@@ -1508,6 +1770,10 @@ do_facet() {
     do_node $HOST "$@"
 }
 
+do_nodesv() {
+    do_nodes --verbose "$@"
+}
+
 add() {
     local facet=$1
     shift
@@ -1533,6 +1799,14 @@ mdsdevname() {
     echo -n $DEVPTR
 }
 
+facet_mntpt () {
+    local facet=$1
+    local var=${facet}_MOUNT
+    eval mntpt=${!var:-${MOUNT%/*}/$facet}
+
+    echo -n $mntpt
+}
+
 ########
 ## MountConf setup
 
@@ -1564,61 +1838,116 @@ stopall() {
         rm -f $TMP/ost${num}active
     done
 
+    if ! combined_mgs_mds ; then
+        stop mgs
+    fi
+
     return 0
 }
 
+cleanup_echo_devs () {
+    local devs=$($LCTL dl | grep echo | awk '{print $4}')
+
+    for dev in $devs; do
+        $LCTL --device $dev cleanup
+        $LCTL --device $dev detach
+    done
+}
+
 cleanupall() {
     nfs_client_mode && return
 
     stopall $*
+    cleanup_echo_devs
+
     unload_modules
     cleanup_gss
 }
 
-mdsmkfsopts()
-{
-    local nr=$1
-    test $nr = 1 && echo -n $MDS_MKFS_OPTS || echo -n $MDSn_MKFS_OPTS
+combined_mgs_mds () {
+    [[ $MDSDEV1 = $MGSDEV ]] && [[ $mds1_HOST = $mgs_HOST ]]
+}
+
+mkfs_opts () {
+    local facet=$1
+
+    local tgt=$(echo $facet | tr -d [:digit:] | tr "[:lower:]" "[:upper:]")
+    local optvar=${tgt}_MKFS_OPTS
+    local opt=${!optvar}
+
+    # FIXME: ! combo  mgs/mds + mgsfailover is not supported yet
+    [[ $facet = mgs ]] && echo $opt && return
+
+    # 1.
+    # --failnode options 
+    local var=${facet}failover_HOST
+    if [ x"${!var}" != x ] && [ x"${!var}" != x$(facet_host $facet) ] ; then
+        local failnode=$(h2$NETTYPE ${!var})
+        failnode="--failnode=$failnode"
+        # options does not contain
+        # or contains wrong --failnode=
+        if [[ $opt != *${failnode}* ]]; then
+            opt=$(echo $opt | sed 's/--failnode=.* / /')
+            opt="$opt $failnode"
+        fi
+    fi
+
+    # 2.
+    # --mgsnode options
+    # no additional mkfs mds "--mgsnode" option for this configuration
+    if [[ $facet = mds ]] && combined_mgs_mds; then
+        echo $opt
+        return
+    fi
+
+    # additional mkfs "--mgsnode"
+    local mgsnode="--mgsnode=$MGSNID"
+    opt=${opt//$mgsnode }
+    for nid in ${MGSNID//:/ }; do
+        local mgsnode="--mgsnode=$nid"
+        # options does not contain
+        # --mgsnode=$nid
+        if [[ $opt != *${mgsnode}" "* ]]; then
+            opt="$opt --mgsnode=$nid"
+        fi
+    done
+
+    echo $opt
 }
 
 formatall() {
     if [ "$IAMDIR" == "yes" ]; then
         MDS_MKFS_OPTS="$MDS_MKFS_OPTS --iam-dir"
-        MDSn_MKFS_OPTS="$MDSn_MKFS_OPTS --iam-dir"
     fi
 
     [ "$FSTYPE" ] && FSTYPE_OPT="--backfstype $FSTYPE"
 
-    if [ ! -z $SEC ]; then
-        MDS_MKFS_OPTS="$MDS_MKFS_OPTS --param srpc.flavor.default=$SEC"
-        MDSn_MKFS_OPTS="$MDSn_MKFS_OPTS --param srpc.flavor.default=$SEC"
-        OST_MKFS_OPTS="$OST_MKFS_OPTS --param srpc.flavor.default=$SEC"
-    fi
-
     stopall
     # We need ldiskfs here, may as well load them all
     load_modules
     [ "$CLIENTONLY" ] && return
     echo Formatting mgs, mds, osts
-    if [[ $MDSDEV1 != $MGSDEV ]] || [[ $mds1_HOST != $mgs_HOST ]]; then
-        add mgs $mgs_MKFS_OPTS $FSTYPE_OPT --reformat $MGSDEV || exit 10
+    if ! combined_mgs_mds ; then
+        add mgs $(mkfs_opts mgs) $FSTYPE_OPT --reformat $MGSDEV || exit 10
     fi
 
     for num in `seq $MDSCOUNT`; do
         echo "Format mds$num: $(mdsdevname $num)"
         if $VERBOSE; then
-            add mds$num `mdsmkfsopts $num` $FSTYPE_OPT --reformat `mdsdevname $num` || exit 9
+            add mds$num $(mkfs_opts mds) $FSTYPE_OPT --reformat $(mdsdevname $num) || exit 10
         else
-            add mds$num `mdsmkfsopts $num` $FSTYPE_OPT --reformat `mdsdevname $num` > /dev/null || exit 9
+            add mds$num $(mkfs_opts mds) $FSTYPE_OPT --reformat $(mdsdevname $num) > /dev/null || exit 10
         fi
     done
 
+    # the ost-s could have different OST_MKFS_OPTS
+    # because of different failnode-s
     for num in `seq $OSTCOUNT`; do
         echo "Format ost$num: $(ostdevname $num)"
         if $VERBOSE; then
-            add ost$num $OST_MKFS_OPTS --reformat `ostdevname $num` || exit 10
+            add ost$num $(mkfs_opts ost${num}) $FSTYPE_OPT --reformat `ostdevname $num` || exit 10
         else
-            add ost$num $OST_MKFS_OPTS --reformat `ostdevname $num` > /dev/null || exit 10
+            add ost$num $(mkfs_opts ost${num}) $FSTYPE_OPT --reformat `ostdevname $num` > /dev/null || exit 10
         fi
     done
 }
@@ -1653,7 +1982,7 @@ switch_identity() {
         do_facet mds$num "lctl set_param -n mdt.$MDT.identity_upcall \"NONE\""
     fi
 
-    do_facet mds$num "lctl set_param -n mdt/$MDT/identity_flush \"-1\""
+    do_facet mds$num "lctl set_param -n mdt/$MDT/identity_flush=-1"
 
     if [ $old = "NONE" ]; then
         return 1
@@ -1664,8 +1993,8 @@ switch_identity() {
 
 remount_client()
 {
-	zconf_umount `hostname` $1 || error "umount failed"
-	zconf_mount `hostname` $1 || error "mount failed"
+        zconf_umount `hostname` $1 || error "umount failed"
+        zconf_mount `hostname` $1 || error "mount failed"
 }
 
 writeconf_facet () {
@@ -1694,13 +2023,13 @@ setupall() {
         error "environments are insane!"
 
     load_modules
-    init_gss
+
     if [ -z "$CLIENTONLY" ]; then
         echo Setup mgs, mdt, osts
         echo $WRITECONF | grep -q "writeconf" && \
             writeconf_all
-        if [[ $mds1_HOST != $mgs_HOST ]] || [[ $MDSDEV1 != $MGSDEV ]]; then
-            start mgs $MGSDEV $mgs_MOUNT_OPTS
+        if ! combined_mgs_mds ; then
+            start mgs $MGSDEV $MGS_MOUNT_OPTS
         fi
 
         for num in `seq $MDSCOUNT`; do
@@ -1714,9 +2043,9 @@ setupall() {
                 eval mds${num}failover_HOST=$(facet_host mds$num)
             fi
 
-	    if [ $IDENTITY_UPCALL != "default" ]; then
+            if [ $IDENTITY_UPCALL != "default" ]; then
                 switch_identity $num $IDENTITY_UPCALL
-	    fi
+            fi
         done
         for num in `seq $OSTCOUNT`; do
             DEVNAME=$(ostdevname $num)
@@ -1731,16 +2060,21 @@ setupall() {
 
         done
     fi
+
+    init_gss
+
     # wait a while to allow sptlrpc configuration be propogated to targets,
     # only needed when mounting new target devices.
-    $GSS && sleep 10
+    if $GSS; then
+        sleep 10
+    fi
 
     [ "$DAEMONFILE" ] && $LCTL debug_daemon start $DAEMONFILE $DAEMONSIZE
     mount_client $MOUNT
     [ -n "$CLIENTS" ] && zconf_mount_clients $CLIENTS $MOUNT
 
     if [ "$MOUNT_2" ]; then
-	mount_client $MOUNT2
+        mount_client $MOUNT2
         [ -n "$CLIENTS" ] && zconf_mount_clients $CLIENTS $MOUNT2
     fi
 
@@ -1752,6 +2086,7 @@ setupall() {
     # by a context negotiation rpc with $TIMEOUT.
     # FIXME better by monitoring import status.
     if $GSS; then
+        set_flavor_all $SEC
         sleep $((TIMEOUT + 5))
     else
         sleep 5
@@ -1759,7 +2094,7 @@ setupall() {
 }
 
 mounted_lustre_filesystems() {
-	awk '($3 ~ "lustre" && $1 ~ ":") { print $2 }' /proc/mounts
+        awk '($3 ~ "lustre" && $1 ~ ":") { print $2 }' /proc/mounts
 }
 
 init_facet_vars () {
@@ -1791,6 +2126,17 @@ init_facet_vars () {
     else
         eval export ${facet}failover_dev=$device
     fi
+
+    # get mount point of already mounted device
+    # is facet_dev is already mounted then use the real
+    #  mount point of this facet; otherwise use $(facet_mntpt $facet)
+    # i.e. ${facet}_MOUNT if specified by user or default
+    local mntpt=$(do_facet ${facet} cat /proc/mounts | \
+            awk '"'${!dev}'" == $1 && $3 == "lustre" { print $2 }')
+    if [ -z $mntpt ]; then
+        mntpt=$(facet_mntpt $facet)
+    fi
+    eval export ${facet}_MOUNT=$mntpt
 }
 
 init_facets_vars () {
@@ -1803,6 +2149,8 @@ init_facets_vars () {
         done
     fi
 
+    combined_mgs_mds || init_facet_vars mgs $MGSDEV $MGS_MOUNT_OPTS
+
     remote_ost_nodsh && return
 
     for num in `seq $OSTCOUNT`; do
@@ -1813,12 +2161,11 @@ init_facets_vars () {
 
 osc_ensure_active () {
     local facet=$1
-    local type=$2
-    local timeout=$3
+    local timeout=$2
     local period=0
 
     while [ $period -lt $timeout ]; do
-        count=$(do_facet $facet "lctl dl | grep '${FSNAME}-OST.*-osc-${type}' | grep ' IN ' 2>/dev/null | wc -l")
+        count=$(do_facet $facet "lctl dl | grep ' IN osc ' 2>/dev/null | wc -l")
         if [ $count -eq 0 ]; then
             break
         fi
@@ -1831,11 +2178,6 @@ osc_ensure_active () {
     [ $period -lt $timeout ] || log "$count OST are inactive after $timeout seconds, give up"
 }
 
-som_check() {
-    SOM_ENABLED=$(do_facet $SINGLEMDS "$LCTL get_param mdt.*.som" | awk -F= ' {print $2}' | head -n 1)
-    echo $SOM_ENABLED
-}
-
 init_param_vars () {
     if ! remote_ost_nodsh && ! remote_mds_nodsh; then
         export MDSVER=$(do_facet $SINGLEMDS "lctl get_param version" | cut -d. -f1,2)
@@ -1848,13 +2190,9 @@ init_param_vars () {
 
     log "Using TIMEOUT=$TIMEOUT"
 
-    osc_ensure_active $SINGLEMDS M $TIMEOUT
-    osc_ensure_active client c $TIMEOUT
+    osc_ensure_active $SINGLEMDS $TIMEOUT
+    osc_ensure_active client $TIMEOUT
 
-    if [ x"$(som_check)" = x"enabled" ]; then
-        ENABLE_QUOTA=""
-        echo "disable quota temporary when SOM enabled"
-    fi
     if [ $QUOTA_AUTO -ne 0 ]; then
         if [ "$ENABLE_QUOTA" ]; then
             echo "enable quota as required"
@@ -1948,11 +2286,26 @@ check_timeout () {
 
 is_mounted () {
     local mntpt=$1
+    [ -z $mntpt ] && return 1
     local mounted=$(mounted_lustre_filesystems)
 
     echo $mounted' ' | grep -w -q $mntpt' '
 }
 
+is_empty_dir() {
+    [ $(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
+}
+
 check_and_setup_lustre() {
     nfs_client_mode && return
 
@@ -2013,6 +2366,10 @@ check_and_setup_lustre() {
             lctl set_param debug_mb=${DEBUG_SIZE};
             sync"
     fi
+
+    init_gss
+    set_flavor_all $SEC
+
     if [ "$ONLY" == "setup" ]; then
         exit 0
     fi
@@ -2037,13 +2394,127 @@ cleanup_and_setup_lustre() {
         lctl set_param debug=0 || true
         cleanupall
         if [ "$ONLY" == "cleanup" ]; then
-    	    exit 0
+            exit 0
         fi
     fi
     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 obd_type
+    local devs
+    local dev
+
+    case $type in
+    mdt) obd_type="osd" ;;
+    ost) obd_type="obdfilter" ;; # needs to be fixed when OST also uses an OSD
+    *) echo "invalid server type" && return 1 ;;
+    esac
+
+    devs=$(do_node $node "lctl get_param -n $obd_type*.*.mntdev")
+    for dev in $devs; do
+        case $dev in
+        *loop*) do_node $node "losetup $dev" | \
+                sed -e "s/.*(//" -e "s/).*//" ;;
+        *) echo $dev ;;
+        esac
+    done
+}
+
+# Get all of the server target devices.
+get_svr_devs() {
+    local i
+
+    # MDT device
+    MDTDEV=$(get_mnt_devs $(mdts_nodes) mdt)
+
+    # 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
+    local target_dev=$2
+    local ostidx=$3
+    local ostdb_opt=$4
+
+    df > /dev/null      # update statfs data on disk
+    local cmd="$E2FSCK -d -v -f -n $MDSDB_OPT $ostdb_opt $target_dev"
+    echo $cmd
+    local rc=0
+    do_node $node $cmd || rc=$?
+    [ $rc -le $FSCK_MAX_ERR ] || \
+        error "$cmd returned $rc, should be <= $FSCK_MAX_ERR"
+    return 0
+}
+
+# Run e2fsck on MDT and OST(s) to generate databases used for lfsck.
+generate_db() {
+    local i
+    local ostidx
+    local dev
+    local tmp_file
+
+    [ $MDSCOUNT -eq 1 ] || error "CMD is not supported"
+    tmp_file=$(mktemp -p $SHARED_DIRECTORY || 
+        error "fail to create file in $SHARED_DIRECTORY")
+
+    # make sure everything gets to the backing store
+    local list=$(comma_list $CLIENTS $(facet_host $SINGLEMDS) $(osts_nodes))
+    do_nodes $list "sync; sleep 2; sync"
+
+    do_nodes $list ls $tmp_file || \
+        error "$SHARED_DIRECTORY is not a shared directory"
+    rm $tmp_file
+
+    run_e2fsck $(mdts_nodes) $MDTDEV
+
+    i=0
+    ostidx=0
+    OSTDB_LIST=""
+    for node in $(osts_nodes); do
+        for dev in ${OSTDEVS[i]}; do
+            local ostdb_opt=`eval echo $OSTDB_OPT`
+            run_e2fsck $node $dev $ostidx "$ostdb_opt"
+            OSTDB_LIST="$OSTDB_LIST $OSTDB-$ostidx"
+            ostidx=$((ostidx + 1))
+        done
+        i=$((i + 1))
+    done
+}
+
+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"
+
+    rm -rvf $MDSDB* $OSTDB* || true
+    return 0
+}
+
 check_and_cleanup_lustre() {
+    if [ "$LFSCK_ALWAYS" = "yes" ]; then
+        get_svr_devs
+        generate_db
+        if [ "$SKIP_LFSCK" == "no" ]; then
+            run_lfsck
+        else
+            echo "skip lfsck"
+        fi
+    fi
+
     if is_mounted $MOUNT; then
         [ -n "$DIR" ] && rm -rf $DIR/[Rdfs][0-9]*
         [ "$ENABLE_QUOTA" ] && restore_quota_type || true
@@ -2066,28 +2537,45 @@ check_and_cleanup_lustre() {
 #######
 # General functions
 
-check_network() {
-    local NETWORK=0
-    local WAIT=0
-    local MAX=$2
-    while [ $NETWORK -eq 0 ]; do
-        if ping -c 1 -w 3 $1 > /dev/null; then
-            NETWORK=1
-        else
-            WAIT=$((WAIT + 5))
-            echo "waiting for $1, $((MAX - WAIT)) secs left"
-            sleep 5
-        fi
-        if [ $WAIT -gt $MAX ]; then
-            echo "Network not available"
-            exit 1
-        fi
+wait_for_function () {
+    local quiet=""
+
+    # 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 wait=0
+
+    while true; do
+
+        eval $fn $quiet && return 0
+
+        wait=$((wait + sleep))
+        [ $wait -lt $max ] || return 1
+        echo waiting $fn, $((max - wait)) secs left ...
+        sleep $sleep
     done
 }
-check_port() {
-    while( !($DSH2 $1 "netstat -tna | grep -q $2") ) ; do
-        sleep 9
-    done
+
+check_network() {
+    local host=$1
+    local max=$2
+    local sleep=${3:-5}
+
+    echo `date +"%H:%M:%S (%s)"` waiting for $host network $max secs ...
+    if ! wait_for_function --quiet "ping -c 1 -w 3 $host" $max $sleep ; then
+        echo "Network not available!"
+        exit 1
+    fi
+
+    echo `date +"%H:%M:%S (%s)"` network interface is UP
 }
 
 no_dsh() {
@@ -2101,6 +2589,12 @@ comma_list() {
     echo "$*" | tr -s " " "\n" | sort -b -u | tr "\n" " " | sed 's/ \([^$]\)/,\1/g'
 }
 
+list_member () {
+    local list=$1
+    local item=$2
+    echo $list | grep -qw $item
+}
+
 # list, excluded are the comma separated lists
 exclude_items_from_list () {
     local list=$1
@@ -2111,7 +2605,7 @@ exclude_items_from_list () {
     for item in ${excluded//,/ }; do
         list=$(echo " $list " | sed -re "s/\s+$item\s+/ /g")
     done
-    echo $(comma_list $list) 
+    echo $(comma_list $list)
 }
 
 # list, expand  are the comma separated lists
@@ -2149,13 +2643,23 @@ absolute_path() {
 }
 
 get_facets () {
-    local name=$(echo $1 | tr "[:upper:]" "[:lower:]")
-    local type=$(echo $1 | tr "[:lower:]" "[:upper:]")
+    local types=${1:-"OST MDS MGS"}
 
     local list=""
-    local count=${type}COUNT
-    for ((i=1; i<=${!count}; i++)) do
-        list="$list ${name}$i"
+
+    for entry in $types; do
+        local name=$(echo $entry | tr "[:upper:]" "[:lower:]")
+        local type=$(echo $entry | tr "[:lower:]" "[:upper:]")
+
+        case $type in
+                MGS ) list="$list $name";;
+            MDS|OST ) local count=${type}COUNT
+                       for ((i=1; i<=${!count}; i++)) do
+                          list="$list ${name}$i"
+                      done;;
+                  * ) error "Invalid facet type"
+                 exit 1;;
+        esac
     done
     echo $(comma_list $list)
 }
@@ -2178,9 +2682,9 @@ at_max_get() {
 
     # suppose that all ost-s has the same at_max set
     if [ $facet == "ost" ]; then
-	do_facet ost1 "lctl get_param -n at_max"
+        do_facet ost1 "lctl get_param -n at_max"
     else
-	do_facet $facet "lctl get_param -n at_max"
+        do_facet $facet "lctl get_param -n at_max"
     fi
 }
 
@@ -2192,15 +2696,15 @@ at_max_set() {
     for facet in $@; do
         if [ $facet == "ost" ]; then
             for i in `seq $OSTCOUNT`; do
-		do_facet ost$i "lctl set_param at_max=$at_max"
+                do_facet ost$i "lctl set_param at_max=$at_max"
 
             done
         elif [ $facet == "mds" ]; then
             for i in `seq $MDSCOUNT`; do
-		do_facet mds$i "lctl set_param at_max=$at_max"
+                do_facet mds$i "lctl set_param at_max=$at_max"
             done
         else
-	    do_facet $facet "lctl set_param at_max=$at_max"
+            do_facet $facet "lctl set_param at_max=$at_max"
         fi
     done
 }
@@ -2319,10 +2823,6 @@ pgcache_empty() {
             return 1
         fi
     done
-    if [[ $MDSDEV1 != $MGSDEV ]]; then
-        stop mgs 
-    fi
-
     return 0
 }
 
@@ -2368,7 +2868,6 @@ stop_full_debug_logging() {
 
 error_noexit() {
     local TYPE=${TYPE:-"FAIL"}
-    local ERRLOG
 
     local dump=true
     # do not dump logs if $1=false
@@ -2379,28 +2878,31 @@ error_noexit() {
 
     log " ${TESTSUITE} ${TESTNAME}: @@@@@@ ${TYPE}: $@ "
 
+    # We need to dump the logs on all nodes
     if $dump; then
-        ERRLOG=$TMP/lustre_${TESTSUITE}_${TESTNAME}.$(date +%s)
-        echo "Dumping lctl log to $ERRLOG"
-        # We need to dump the logs on all nodes
-        do_nodes $(comma_list $(nodes_list)) $NODE $LCTL dk $ERRLOG
+        gather_logs $(comma_list $(nodes_list))
     fi
+
     debugrestore
     [ "$TESTSUITELOG" ] && echo "$0: ${TYPE}: $TESTNAME $@" >> $TESTSUITELOG
-    TEST_FAILED=true
+    echo "$@" > $LOGDIR/err
+}
+
+exit_status () {
+    local status=0
+    local log=$TESTSUITELOG
+
+    [ -f "$log" ] && grep -q FAIL $log && status=1
+    exit $status
 }
 
 error() {
     error_noexit "$@"
-    if $FAIL_ON_ERROR; then
-        reset_fail_loc
-        exit 1
-    fi
+    exit 1
 }
 
 error_exit() {
-    error_noexit "$@"
-    exit 1
+    error "$@"
 }
 
 # use only if we are ignoring failures for this test, bugno required.
@@ -2442,7 +2944,7 @@ build_test_filter() {
     done
     for G in $GRANT_CHECK_LIST; do
         eval GCHECK_ONLY_${G}=true
-   	done
+        done
 }
 
 basetest() {
@@ -2455,6 +2957,14 @@ basetest() {
 
 # print a newline if the last test was skipped
 export LAST_SKIPPED=
+#
+# Main entry into test-framework. This is called with the name and
+# description of a test. The name is used to find the function to run
+# the test using "test_$name".
+#
+# This supports a variety of methods of specifying specific test to
+# run or not run.  These need to be documented...
+#
 run_test() {
     assert_DIR
 
@@ -2463,13 +2973,13 @@ run_test() {
         testname=ONLY_$1
         if [ ${!testname}x != x ]; then
             [ "$LAST_SKIPPED" ] && echo "" && LAST_SKIPPED=
-            run_one $1 "$2"
+            run_one_logged $1 "$2"
             return $?
         fi
         testname=ONLY_$base
         if [ ${!testname}x != x ]; then
             [ "$LAST_SKIPPED" ] && echo "" && LAST_SKIPPED=
-            run_one $1 "$2"
+            run_one_logged $1 "$2"
             return $?
         fi
         LAST_SKIPPED="y"
@@ -2502,18 +3012,13 @@ run_test() {
     fi
 
     LAST_SKIPPED=
-    run_one $1 "$2"
+    run_one_logged $1 "$2"
 
     return $?
 }
 
-EQUALS="======================================================================"
 equals_msg() {
-    msg="$@"
-
-    local suffixlen=$((${#EQUALS} - ${#msg}))
-    [ $suffixlen -lt 5 ] && suffixlen=5
-    log `echo $(printf '===== %s %.*s\n' "$msg" $suffixlen $EQUALS)`
+    banner "$*"
 }
 
 log() {
@@ -2534,16 +3039,27 @@ log() {
 }
 
 trace() {
-	log "STARTING: $*"
-	strace -o $TMP/$1.strace -ttt $*
-	RC=$?
-	log "FINISHED: $*: rc $RC"
-	return 1
+        log "STARTING: $*"
+        strace -o $TMP/$1.strace -ttt $*
+        RC=$?
+        log "FINISHED: $*: rc $RC"
+        return 1
+}
+
+complete () {
+    equals_msg $1 test complete, duration $2 sec
+    [ -f "$TESTSUITELOG" ] && egrep .FAIL $TESTSUITELOG || true
+    echo duration $2 >>$TESTSUITELOG
 }
 
 pass() {
-    $TEST_FAILED && echo -n "FAIL " || echo -n "PASS " 
-    echo $@
+    # Set TEST_STATUS here; will be used for logging the result
+    if [ -f $LOGDIR/err ]; then
+        TEST_STATUS="FAIL"
+    else
+        TEST_STATUS="PASS"
+    fi
+    echo "$TEST_STATUS $@" 2>&1 | tee -a $TESTSUITELOG
 }
 
 check_mds() {
@@ -2558,86 +3074,124 @@ reset_fail_loc () {
     echo done.
 }
 
+
+#
+# Log a message (on all nodes) padded with "=" before and after. 
+# Also appends a timestamp and prepends the testsuite name.
+# 
+
+EQUALS="===================================================================================================="
+banner() {
+    msg="== ${TESTSUITE} $*"
+    last=${msg: -1:1}
+    [[ $last != "=" && $last != " " ]] && msg="$msg "
+    msg=$(printf '%s%.*s'  "$msg"  $((${#EQUALS} - ${#msg})) $EQUALS )
+    # always include at least == after the message
+    log "$msg== $(date +"%H:%M:%S (%s)")"
+}
+
+#
+# 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() {
-    testnum=$1
-    message=$2
+    local testnum=$1
+    local message=$2
     tfile=f${testnum}
     export tdir=d0.${TESTSUITE}/d${base}
-
+    export TESTNAME=test_$testnum
     local SAVE_UMASK=`umask`
     umask 0022
 
-    local BEFORE=`date +%s`
-    echo
-    log "== test $testnum: $message == `date +%H:%M:%S` ($BEFORE)"
-    #check_mds
-    export TESTNAME=test_$testnum
-    TEST_FAILED=false
+    banner "test $testnum: $message"
     test_${testnum} || error "test_$testnum failed with $?"
-    #check_mds
     cd $SAVE_PWD
     reset_fail_loc
     check_grant ${testnum} || error "check_grant $testnum failed with $?"
     check_catastrophe || error "LBUG/LASSERT detected"
     ps auxww | grep -v grep | grep -q multiop && error "multiop still running"
-    pass "($((`date +%s` - $BEFORE))s)"
-    TEST_FAILED=false
     unset TESTNAME
     unset tdir
     umask $SAVE_UMASK
+    return 0
+}
+
+#
+# Wrapper around run_one to ensure:
+#  - test runs in subshell
+#  - output of test is saved to separate log file for error reporting
+#  - test result is saved to data file
+#
+run_one_logged() {
+    local BEFORE=`date +%s`
+    local TEST_ERROR
+    local name=${TESTSUITE}.test_${1}.test_log.$(hostname).log
+    local test_log=$LOGDIR/$name
+    rm -rf $LOGDIR/err
+
+    echo
+    log_sub_test_begin test_${1}
+    (run_one $1 "$2") 2>&1 | tee $test_log
+    local RC=${PIPESTATUS[0]}
+
+    [ $RC -ne 0 ] && [ ! -f $LOGDIR/err ] && \
+        echo "test_$1 returned $RC" | tee $LOGDIR/err
+
+    duration=$((`date +%s` - $BEFORE))
+    pass "$1" "(${duration}s)"
+    [ -f $LOGDIR/err ] && TEST_ERROR=$(cat $LOGDIR/err)
+    log_sub_test_end $TEST_STATUS $duration "$RC" "$TEST_ERROR"
+
+    if [ -f $LOGDIR/err ]; then
+        $FAIL_ON_ERROR && exit $RC
+    fi
+
+    return 0
 }
 
 canonical_path() {
     (cd `dirname $1`; echo $PWD/`basename $1`)
 }
 
-sync_clients() {
-    [ -d $DIR1 ] && cd $DIR1 && sync; sleep 1; sync
-    [ -d $DIR2 ] && cd $DIR2 && sync; sleep 1; sync
-	cd $SAVE_PWD
-}
 
 check_grant() {
     export base=`basetest $1`
     [ "$CHECK_GRANT" == "no" ] && return 0
 
-	testname=GCHECK_ONLY_${base}
+        testname=GCHECK_ONLY_${base}
         [ ${!testname}x == x ] && return 0
 
     echo -n "checking grant......"
-	cd $SAVE_PWD
-	# write some data to sync client lost_grant
-	rm -f $DIR1/${tfile}_check_grant_* 2>&1
-	for i in `seq $OSTCOUNT`; do
-		$LFS setstripe $DIR1/${tfile}_check_grant_$i -i $(($i -1)) -c 1
-		dd if=/dev/zero of=$DIR1/${tfile}_check_grant_$i bs=4k \
-					      count=1 > /dev/null 2>&1
-	done
-	# sync all the data and make sure no pending data on server
-	sync_clients
-	
-	#get client grant and server grant
-	client_grant=0
-    for d in `lctl get_param -n osc.*.cur_grant_bytes`; do
-		client_grant=$((client_grant + $d))
-	done
-	server_grant=0
-	for d in `lctl get_param -n obdfilter.*.tot_granted`; do
-		server_grant=$((server_grant + $d))
-	done
-
-	# cleanup the check_grant file
-	for i in `seq $OSTCOUNT`; do
-	        rm $DIR1/${tfile}_check_grant_$i
-	done
-
-	#check whether client grant == server grant
-	if [ $client_grant != $server_grant ]; then
-		echo "failed: client:${client_grant} server: ${server_grant}"
-		return 1
-	else
-		echo "pass"
-	fi
+
+        local clients=$CLIENTS
+        [ -z $clients ] && clients=$(hostname)
+
+    # sync all the data and make sure no pending data on server
+    do_nodes $clients sync
+
+    # get client grant
+    client_grant=`do_nodes $clients \
+                    "$LCTL get_param -n osc.${FSNAME}-*.cur_*grant_bytes" | \
+                    awk '{total += $1} END{print total}'`
+
+    # get server grant
+    server_grant=`do_nodes $(comma_list $(osts_nodes)) \
+                    "$LCTL get_param -n obdfilter.${FSNAME}-OST*.tot_granted" | \
+                    awk '{total += $1} END{print total}'`
+
+    # check whether client grant == server grant
+    if [ $client_grant -ne $server_grant ]; then
+        echo "failed: client:${client_grant} server: ${server_grant}."
+        do_nodes $(comma_list $(osts_nodes)) \
+                   "$LCTL get_param obdfilter.${FSNAME}-OST*.tot*"
+        do_nodes $clients "$LCTL get_param osc.${FSNAME}-*.cur_*_bytes"
+        return 1
+    else
+        echo "pass: client:${client_grant} server: ${server_grant}"
+    fi
+
 }
 
 ########################
@@ -2653,6 +3207,11 @@ osc_to_ost()
     echo $ost
 }
 
+ostuuid_from_index()
+{
+    $LFS osts $2 | awk '/^'$1'/ { print $2 }'
+}
+
 remote_node () {
     local node=$1
     [ "$node" != "$(hostname)" ]
@@ -2673,6 +3232,13 @@ remote_mds_nodsh()
     remote_mds && [ "$PDSH" = "no_dsh" -o -z "$PDSH" -o -z "$mds_HOST" ]
 }
 
+require_dsh_mds()
+{
+        remote_mds_nodsh && echo "SKIP: $TESTSUITE: remote MDS with nodsh" && \
+            MSKIPPED=1 && return 1
+        return 0
+}
+
 remote_ost ()
 {
     local node
@@ -2688,6 +3254,13 @@ remote_ost_nodsh()
     remote_ost && [ "$PDSH" = "no_dsh" -o -z "$PDSH" -o -z "$ost_HOST" ]
 }
 
+require_dsh_ost()
+{
+        remote_ost_nodsh && echo "SKIP: $TESTSUITE: remote OST with nodsh" && \
+            OSKIPPED=1 && return 1
+        return 0
+}
+
 remote_mgs_nodsh()
 {
     local MGS 
@@ -2695,6 +3268,12 @@ remote_mgs_nodsh()
     remote_node $MGS && [ "$PDSH" = "no_dsh" -o -z "$PDSH" -o -z "$ost_HOST" ]
 }
 
+local_mode ()
+{
+    remote_mds_nodsh || remote_ost_nodsh || \
+        $(single_local_node $(comma_list $(nodes_list)))
+}
+
 mdts_nodes () {
     local MDSNODES
     local NODES_sort
@@ -2710,19 +3289,30 @@ remote_servers () {
     remote_ost && remote_mds
 }
 
-osts_nodes () {
-    local OSTNODES=$(facet_host ost1)
+facets_nodes () {
+    local facets=$1
+    local nodes
     local NODES_sort
 
-    for num in `seq $OSTCOUNT`; do
-        local myOST=$(facet_host ost$num)
-        OSTNODES="$OSTNODES $myOST"
+    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 $OSTNODES; do echo $i; done | sort -u)
+    NODES_sort=$(for i in $nodes; do echo $i; done | sort -u)
 
     echo $NODES_sort
 }
 
+osts_nodes () {
+    local facets=$(get_facets OST)
+    local nodes=$(facets_nodes $facets)
+
+    echo $nodes
+}
+
 nodes_list () {
     # FIXME. We need a list of clients
     local myNODES=$HOSTNAME
@@ -2732,7 +3322,7 @@ nodes_list () {
     [ -n "$CLIENTS" ] && myNODES=${CLIENTS//,/ }
 
     if [ "$PDSH" -a "$PDSH" != "no_dsh" ]; then
-        myNODES="$myNODES $(osts_nodes) $(mdts_nodes)"
+        myNODES="$myNODES $(facets_nodes $(get_facets))"
     fi
 
     myNODES_sort=$(for i in $myNODES; do echo $i; done | sort -u)
@@ -2814,9 +3404,10 @@ mixed_mdt_devs () {
 generate_machine_file() {
     local nodes=${1//,/ }
     local machinefile=$2
-    rm -f $machinefile || error "can't rm $machinefile"
+    rm -f $machinefile
     for node in $nodes; do
-        echo $node >>$machinefile
+        echo $node >>$machinefile || \
+            { echo "can not generate machinefile $machinefile" && return 1; }
     done
 }
 
@@ -2827,6 +3418,17 @@ get_stripe () {
     rm -f $file
 }
 
+setstripe_nfsserver () {
+    local dir=$1
+
+    local nfsserver=$(awk '"'$dir'" ~ $2 && $3 ~ "nfs" && $2 != "/" \
+                { print $1 }' /proc/mounts | cut -f 1 -d : | head -1)
+
+    [ -z $nfsserver ] && echo "$dir is not nfs mounted" && return 1
+
+    do_nodev $nfsserver lfs setstripe "$@"
+}
+
 check_runas_id_ret() {
     local myRC=0
     local myRUNAS_UID=$1
@@ -2935,6 +3537,10 @@ inodes_available () {
     echo $IFree
 }
 
+mdsrate_inodes_available () {
+    echo $(($(inodes_available) - 1))
+}
+
 # reset llite stat counters
 clear_llite_stats(){
         lctl set_param -n llite.*.stats 0
@@ -2972,7 +3578,7 @@ calc_osc_kbytes () {
 # generate a stream of formatted strings (<node> <param name>=<param value>)
 save_lustre_params() {
         local s
-        do_nodes --verbose $1 "lctl get_param $2 | while read s; do echo \\\$s; done"
+        do_nodesv $1 "lctl get_param $2 | while read s; do echo \\\$s; done"
 }
 
 # restore lustre parameters from input stream, produces by save_lustre_params
@@ -3001,19 +3607,19 @@ exit \\\$rc;"
 # $2 file
 # $3 $RUNAS
 get_stripe_info() {
-	local tmp_file
+        local tmp_file
 
-	stripe_size=0
-	stripe_count=0
-	stripe_index=0
-	tmp_file=$(mktemp)
+        stripe_size=0
+        stripe_count=0
+        stripe_index=0
+        tmp_file=$(mktemp)
 
-	do_facet $1 $3 lfs getstripe -v $2 > $tmp_file
+        do_facet $1 $3 lfs getstripe -v $2 > $tmp_file
 
-	stripe_size=`awk '$1 ~ /size/ {print $2}' $tmp_file`
-	stripe_count=`awk '$1 ~ /count/ {print $2}' $tmp_file`
-	stripe_index=`awk '$1 ~ /stripe_offset/ {print $2}' $tmp_file`
-	rm -f $tmp_file
+        stripe_size=`awk '$1 ~ /size/ {print $2}' $tmp_file`
+        stripe_count=`awk '$1 ~ /count/ {print $2}' $tmp_file`
+        stripe_index=`awk '$1 ~ /stripe_offset/ {print $2}' $tmp_file`
+        rm -f $tmp_file
 }
 
 # CMD: determine mds index where directory inode presents
@@ -3076,32 +3682,56 @@ convert_facet2label() {
 get_clientosc_proc_path() {
     local ost=$1
 
-    echo "{$1}-osc-*"
+    echo "${1}-osc-*"
 }
 
 get_lustre_version () {
-    local node=${1:-"mds"}    
-    do_facet $node $LCTL get_param -n version |  awk '/^lustre:/ {print $2}'
+    local facet=${1:-"$SINGLEMDS"}    
+    do_facet $facet $LCTL get_param -n version |  awk '/^lustre:/ {print $2}'
 }
 
 get_mds_version_major () {
-    local version=$(get_lustre_version mds)
+    local facet=${1:-"$SINGLEMDS"}
+    local version=$(get_lustre_version $facet)
     echo $version | awk -F. '{print $1}'
 }
 
 get_mds_version_minor () {
-    local version=$(get_lustre_version mds)
+    local facet=${1:-"$SINGLEMDS"}
+    local version=$(get_lustre_version $facet)
     echo $version | awk -F. '{print $2}'
 }
 
+# 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"}
+    local major=$(get_mds_version_major $mds)
+    local minor=$(get_mds_version_minor $mds)
+
+    if [ $major -ge 2 ] || [ $major -eq 1 -a $minor -gt 8 ]; 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 ost=$1
-    local major=$(get_mds_version_major)
-    local minor=$(get_mds_version_minor)
-    if [ $major -le 1 -a $minor -le 8 ] ; then
-        echo "${ost}-osc"
+    local mds_facet=$1
+    local ost_label=${2:-"*OST*"}
+
+    [ "$mds_facet" = "mds" ] && mds_facet=$SINGLEMDS
+    local mdt_label=$(convert_facet2label $mds_facet)
+    local mdt_index=$(echo $mdt_label | sed -e 's/^.*-//')
+
+    local major=$(get_mds_version_major $mds_facet)
+    local minor=$(get_mds_version_minor $mds_facet)
+    if [ $major -le 1 -a $minor -le 8 ] || mds_on_old_device $mds_facet; then
+        echo "${ost_label}-osc"
     else
-        echo "${ost}-osc-MDT0000"
+        echo "${ost_label}-osc-${mdt_index}"
     fi
 }
 
@@ -3110,8 +3740,8 @@ get_osc_import_name() {
     local ost=$2
     local label=$(convert_facet2label $ost)
 
-    if [ "$facet" == "mds" ]; then
-        get_mdtosc_proc_path $label
+    if [ "${facet:0:3}" = "mds" ]; then
+        get_mdtosc_proc_path $facet $label
         return 0
     fi
 
@@ -3119,33 +3749,37 @@ get_osc_import_name() {
     return 0
 }
 
-wait_import_state () {
+_wait_import_state () {
     local expected=$1
     local CONN_PROC=$2
+    local maxtime=${3:-max_recovery_time}
     local CONN_STATE
     local i=0
 
     CONN_STATE=$($LCTL get_param -n $CONN_PROC 2>/dev/null | cut -f2)
     while [ "${CONN_STATE}" != "${expected}" ]; do
-        if [ "${expected}" == "DISCONN" ]; then
-            # for disconn we can check after proc entry is removed
-            [ "x${CONN_STATE}" == "x" ] && return 0
-            #  with AT we can have connect request timeout ~ reconnect timeout
-            # and test can't see real disconnect
-            [ "${CONN_STATE}" == "CONNECTING" ] && return 0
-        fi
-        # disconnect rpc should be wait not more obd_timeout
-        [ $i -ge $(($TIMEOUT * 3 / 2)) ] && \
-            error "can't put import for $CONN_PROC into ${expected} state" && return 1
+        [ $i -ge $maxtime ] && \
+            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 $CONN_PROC 2>/dev/null | cut -f2)
         i=$(($i + 1))
     done
 
-    log "$CONN_PROC now in ${CONN_STATE} state"
+    log "$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 param
+
+    for param in ${params//,/ }; do
+        _wait_import_state $state $param $maxtime || return
+    done
+}
 wait_osc_import_state() {
     local facet=$1
     local ost_facet=$2
@@ -3176,7 +3810,6 @@ wait_osc_import_state() {
     log "${ost_facet} now in ${CONN_STATE} state"
     return 0
 }
-
 get_clientmdc_proc_path() {
     echo "${1}-mdc-*"
 }
@@ -3185,15 +3818,23 @@ do_rpc_nodes () {
     local list=$1
     shift
 
-    do_nodes --verbose $list "PATH=$LUSTRE/tests/:$PATH sh rpc.sh $@ " 
+    # Add paths to lustre tests for 32 and 64 bit systems.
+    local RPATH="PATH=$RLUSTRE/tests:/usr/lib/lustre/tests:/usr/lib64/lustre/tests:$PATH"
+    do_nodesv $list "${RPATH} NAME=${NAME} sh rpc.sh $@ "
 }
 
 wait_clients_import_state () {
     local list=$1
     local facet=$2
     local expected=$3
-    shift
 
+    local facets=$facet
+
+    if [ "$FAILURE_MODE" = HARD ]; then
+        facets=$(facets_on_host $(facet_active_host $facet))
+    fi
+
+    for facet in ${facets//,/ }; do
     local label=$(convert_facet2label $facet)
     local proc_path
     case $facet in
@@ -3201,8 +3842,10 @@ wait_clients_import_state () {
         mds* ) proc_path="mdc.$(get_clientmdc_proc_path $label).mds_server_uuid" ;;
         *) error "unknown facet!" ;;
     esac
+    local params=$(expand_list $params $proc_path)
+    done
 
-    if ! do_rpc_nodes $list wait_import_state $expected $proc_path; then
+    if ! do_rpc_nodes $list wait_import_state $expected $params; then
         error "import is not in ${expected} state"
         return 1
     fi
@@ -3325,33 +3968,31 @@ gather_logs () {
     local list=$1
 
     local ts=$(date +%s)
-
-    # bug 20237, comment 11
-    # It would also be useful to provide the option
-    # of writing the file to an NFS directory so it doesn't need to be copied.
-    local tmp=$TMP
     local docp=true
-    [ -d "$SHARED_DIR_LOGS" ] && tmp=$SHARED_DIR_LOGS && docp=false
-
+    [ -f $LOGDIR/shared ] && docp=false
+ 
     # dump lustre logs, dmesg
-    do_nodes $list "log=$tmp/\\\$(hostname)-debug-$ts.log ;
-lctl dk \\\$log >/dev/null;
-log=$tmp/\\\$(hostname)-dmesg-$ts.log;
-dmesg > \\\$log; "
 
-    # FIXME: does it make sense to collect the logs for $ts only, but all
-    # TESTSUITE logs?
-    # rsync $TMP/*${TESTSUITE}* to gather the logs dumped by error fn
-    local logs=$TMP/'*'${TESTSUITE}'*'
-    if $docp; then
-        logs=$logs' '$tmp/'*'$ts'*'
+    prefix="$LOGDIR/${TESTSUITE}.${TESTNAME}"
+    suffix="$ts.log"
+    echo "Dumping lctl log to ${prefix}.*.${suffix}"
+
+    if [ "$CLIENTONLY" -o "$PDSH" == "no_dsh" ]; then
+        echo "Dumping logs only on local client."
+        $LCTL dk > ${prefix}.debug_log.$(hostname).${suffix}
+        dmesg > ${prefix}.dmesg.$(hostname).${suffix}
+        return
     fi
-    for node in ${list//,/ }; do
-        rsync -az $node:"$logs" $TMP
-    done
 
-    local archive=$TMP/${TESTSUITE}-$ts.tar.bz2
-    tar -jcf $archive $tmp/*$ts* $TMP/*${TESTSUITE}*
+    do_nodesv $list \
+        "$LCTL dk > ${prefix}.debug_log.\\\$(hostname).${suffix};
+         dmesg > ${prefix}.dmesg.\\\$(hostname).${suffix}"
+    if [ ! -f $LOGDIR/shared ]; then
+        do_nodes $list rsync -az "${prefix}.*.${suffix}" $HOSTNAME:$LOGDIR
+      fi
+
+    local archive=$LOGDIR/${TESTSUITE}-$ts.tar.bz2
+    tar -jcf $archive $LOGDIR/*$ts* $LOGDIR/*${TESTSUITE}*
 
     echo $archive
 }
@@ -3385,3 +4026,479 @@ do_ls () {
     return $rc
 }
 
+# target_start_and_reset_recovery_timer()
+#        service_time = at_est2timeout(service_time);
+#        service_time += 2 * (CONNECTION_SWITCH_MAX + CONNECTION_SWITCH_INC +
+#                             INITIAL_CONNECT_TIMEOUT);
+# CONNECTION_SWITCH_MAX : min(25U, max(CONNECTION_SWITCH_MIN,obd_timeout))
+#define CONNECTION_SWITCH_INC 1
+#define INITIAL_CONNECT_TIMEOUT max(CONNECTION_SWITCH_MIN,obd_timeout/20)
+#define CONNECTION_SWITCH_MIN 5U
+
+max_recovery_time () {
+    local init_connect_timeout=$(( TIMEOUT / 20 ))
+    [[ $init_connect_timeout > 5 ]] || init_connect_timeout=5 
+
+    local service_time=$(( $(at_max_get client) + $(( 2 * $(( 25 + 1  + init_connect_timeout)) )) ))
+
+    echo $service_time 
+}
+
+get_clients_mount_count () {
+    local clients=${CLIENTS:-`hostname`}
+
+    # we need to take into account the clients mounts and
+    # exclude mds/ost mounts if any;
+    do_nodes $clients cat /proc/mounts | grep lustre | grep $MOUNT | wc -l
+}
+
+# gss functions
+PROC_CLI="srpc_info"
+
+combination()
+{
+    local M=$1
+    local N=$2
+    local R=1
+
+    if [ $M -lt $N ]; then
+        R=0
+    else
+        N=$((N + 1))
+        while [ $N -lt $M ]; do
+            R=$((R * N))
+            N=$((N + 1))
+        done
+    fi
+
+    echo $R
+    return 0
+}
+
+calc_connection_cnt() {
+    local dir=$1
+
+    # MDT->MDT = 2 * C(M, 2)
+    # MDT->OST = M * O
+    # CLI->OST = C * O
+    # CLI->MDT = C * M
+    comb_m2=$(combination $MDSCOUNT 2)
+
+    local num_clients=$(get_clients_mount_count)
+
+    local cnt_mdt2mdt=$((comb_m2 * 2))
+    local cnt_mdt2ost=$((MDSCOUNT * OSTCOUNT))
+    local cnt_cli2ost=$((num_clients * OSTCOUNT))
+    local cnt_cli2mdt=$((num_clients * MDSCOUNT))
+    local cnt_all2ost=$((cnt_mdt2ost + cnt_cli2ost))
+    local cnt_all2mdt=$((cnt_mdt2mdt + cnt_cli2mdt))
+    local cnt_all2all=$((cnt_mdt2ost + cnt_mdt2mdt + cnt_cli2ost + cnt_cli2mdt))
+
+    local var=cnt_$dir
+    local res=${!var}
+
+    echo $res
+}
+
+set_rule()
+{
+    local tgt=$1
+    local net=$2
+    local dir=$3
+    local flavor=$4
+    local cmd="$tgt.srpc.flavor"
+
+    if [ $net == "any" ]; then
+        net="default"
+    fi
+    cmd="$cmd.$net"
+
+    if [ $dir != "any" ]; then
+        cmd="$cmd.$dir"
+    fi
+
+    cmd="$cmd=$flavor"
+    log "Setting sptlrpc rule: $cmd"
+    do_facet mgs "$LCTL conf_param $cmd"
+}
+
+count_flvr()
+{
+    local output=$1
+    local flavor=$2
+    local count=0
+
+    rpc_flvr=`echo $flavor | awk -F - '{ print $1 }'`
+    bulkspec=`echo $flavor | awk -F - '{ print $2 }'`
+
+    count=`echo "$output" | grep "rpc flavor" | grep $rpc_flvr | wc -l`
+
+    if [ "x$bulkspec" != "x" ]; then
+        algs=`echo $bulkspec | awk -F : '{ print $2 }'`
+
+        if [ "x$algs" != "x" ]; then
+            bulk_count=`echo "$output" | grep "bulk flavor" | grep $algs | wc -l`
+        else
+            bulk=`echo $bulkspec | awk -F : '{ print $1 }'`
+            if [ $bulk == "bulkn" ]; then
+                bulk_count=`echo "$output" | grep "bulk flavor" \
+                            | grep "null/null" | wc -l`
+            elif [ $bulk == "bulki" ]; then
+                bulk_count=`echo "$output" | grep "bulk flavor" \
+                            | grep "/null" | grep -v "null/" | wc -l`
+            else
+                bulk_count=`echo "$output" | grep "bulk flavor" \
+                            | grep -v "/null" | grep -v "null/" | wc -l`
+            fi
+        fi
+
+        [ $bulk_count -lt $count ] && count=$bulk_count
+    fi
+
+    echo $count
+}
+
+flvr_cnt_cli2mdt()
+{
+    local flavor=$1
+    local cnt
+
+    local clients=${CLIENTS:-`hostname`}
+
+    for c in ${clients//,/ }; do
+        output=`do_node $c lctl get_param -n mdc.*-MDT*-mdc-*.$PROC_CLI 2>/dev/null`
+        tmpcnt=`count_flvr "$output" $flavor`
+        cnt=$((cnt + tmpcnt))
+    done
+    echo $cnt
+}
+
+flvr_cnt_cli2ost()
+{
+    local flavor=$1
+    local cnt
+
+    local clients=${CLIENTS:-`hostname`}
+
+    for c in ${clients//,/ }; do
+        output=`do_node $c lctl get_param -n osc.*OST*-osc-[^M][^D][^T]*.$PROC_CLI 2>/dev/null`
+        tmpcnt=`count_flvr "$output" $flavor`
+        cnt=$((cnt + tmpcnt))
+    done
+    echo $cnt
+}
+
+flvr_cnt_mdt2mdt()
+{
+    local flavor=$1
+    local cnt=0
+
+    if [ $MDSCOUNT -le 1 ]; then
+        echo 0
+        return
+    fi
+
+    for num in `seq $MDSCOUNT`; do
+        output=`do_facet mds$num lctl get_param -n mdc.*-MDT*-mdc[0-9]*.$PROC_CLI 2>/dev/null`
+        tmpcnt=`count_flvr "$output" $flavor`
+        cnt=$((cnt + tmpcnt))
+    done
+    echo $cnt;
+}
+
+flvr_cnt_mdt2ost()
+{
+    local flavor=$1
+    local cnt=0
+    local mdtosc
+
+    for num in `seq $MDSCOUNT`; do
+        mdtosc=$(get_mdtosc_proc_path mds$num)
+        mdtosc=${mdtosc/-MDT*/-MDT\*}
+        output=$(do_facet mds$num lctl get_param -n \
+            osc.$mdtosc.$PROC_CLI 2>/dev/null)
+        tmpcnt=`count_flvr "$output" $flavor`
+        cnt=$((cnt + tmpcnt))
+    done
+    echo $cnt;
+}
+
+flvr_cnt_mgc2mgs()
+{
+    local flavor=$1
+
+    output=`do_facet client lctl get_param -n mgc.*.$PROC_CLI 2>/dev/null`
+    count_flvr "$output" $flavor
+}
+
+do_check_flavor()
+{
+    local dir=$1        # from to
+    local flavor=$2     # flavor expected
+    local res=0
+
+    if [ $dir == "cli2mdt" ]; then
+        res=`flvr_cnt_cli2mdt $flavor`
+    elif [ $dir == "cli2ost" ]; then
+        res=`flvr_cnt_cli2ost $flavor`
+    elif [ $dir == "mdt2mdt" ]; then
+        res=`flvr_cnt_mdt2mdt $flavor`
+    elif [ $dir == "mdt2ost" ]; then
+        res=`flvr_cnt_mdt2ost $flavor`
+    elif [ $dir == "all2ost" ]; then
+        res1=`flvr_cnt_mdt2ost $flavor`
+        res2=`flvr_cnt_cli2ost $flavor`
+        res=$((res1 + res2))
+    elif [ $dir == "all2mdt" ]; then
+        res1=`flvr_cnt_mdt2mdt $flavor`
+        res2=`flvr_cnt_cli2mdt $flavor`
+        res=$((res1 + res2))
+    elif [ $dir == "all2all" ]; then
+        res1=`flvr_cnt_mdt2ost $flavor`
+        res2=`flvr_cnt_cli2ost $flavor`
+        res3=`flvr_cnt_mdt2mdt $flavor`
+        res4=`flvr_cnt_cli2mdt $flavor`
+        res=$((res1 + res2 + res3 + res4))
+    fi
+
+    echo $res
+}
+
+wait_flavor()
+{
+    local dir=$1        # from to
+    local flavor=$2     # flavor expected
+    local expect=${3:-$(calc_connection_cnt $dir)}     # number expected
+
+    local res=0
+
+    for ((i=0;i<20;i++)); do
+        echo -n "checking $dir..."
+        res=$(do_check_flavor $dir $flavor)
+        echo "found $res/$expect $flavor connections"
+        [ $res -eq $expect ] && return 0
+        sleep 4
+    done
+
+    echo "Error checking $flavor of $dir: expect $expect, actual $res"
+    return 1
+}
+
+restore_to_default_flavor()
+{
+    local proc="mgs.MGS.live.$FSNAME"
+
+    echo "restoring to default flavor..."
+
+    nrule=`do_facet mgs lctl get_param -n $proc 2>/dev/null | grep ".srpc.flavor." | wc -l`
+
+    # remove all existing rules if any
+    if [ $nrule -ne 0 ]; then
+        echo "$nrule existing rules"
+        for rule in `do_facet mgs lctl get_param -n $proc 2>/dev/null | grep ".srpc.flavor."`; do
+            echo "remove rule: $rule"
+            spec=`echo $rule | awk -F = '{print $1}'`
+            do_facet mgs "$LCTL conf_param -d $spec"
+        done
+    fi
+
+    # verify no rules left
+    nrule=`do_facet mgs lctl get_param -n $proc 2>/dev/null | grep ".srpc.flavor." | wc -l`
+    [ $nrule -ne 0 ] && error "still $nrule rules left"
+
+    # wait for default flavor to be applied
+    # currently default flavor for all connections are 'null'
+    wait_flavor all2all null
+    echo "now at default flavor settings"
+}
+
+set_flavor_all()
+{
+    local flavor=${1:-null}
+
+    echo "setting all flavor to $flavor"
+
+    # FIXME need parameter to this fn
+    # and remove global vars
+    local cnt_all2all=$(calc_connection_cnt all2all)
+
+    local res=$(do_check_flavor all2all $flavor)
+    if [ $res -eq $cnt_all2all ]; then
+        echo "already have total $res $flavor connections"
+        return
+    fi
+
+    echo "found $res $flavor out of total $cnt_all2all connections"
+    restore_to_default_flavor
+
+    [[ $flavor = null ]] && return 0
+
+    set_rule $FSNAME any any $flavor
+    wait_flavor all2all $flavor
+}
+
+
+check_logdir() {
+    local dir=$1
+    # Checking for shared logdir
+    if [ ! -d $dir ]; then
+        # Not found. Create local logdir
+        mkdir -p $dir
+    else
+        touch $dir/node.$(hostname).yml
+    fi
+    return 0
+}
+
+check_write_access() {
+    local dir=$1
+    for node in $(nodes_list); do
+        if [ ! -f "$dir/node.${node}.yml" ]; then
+            # Logdir not accessible/writable from this node.
+            return 1
+        fi
+    done
+    return 0
+}
+
+init_logging() {
+    if [[ -n $YAML_LOG ]]; then
+        return
+    fi
+    export YAML_LOG=${LOGDIR}/results.yml
+    mkdir -p $LOGDIR
+    init_clients_lists
+
+    do_rpc_nodes $(comma_list $(nodes_list)) check_logdir $LOGDIR
+    if check_write_access $LOGDIR; then
+        touch $LOGDIR/shared
+        echo "Logging to shared log directory: $LOGDIR"
+    else
+        echo "Logging to local directory: $LOGDIR"
+    fi
+
+    yml_nodes_file $LOGDIR
+    yml_results_file >> $YAML_LOG
+}
+
+log_test() {
+    yml_log_test $1 >> $YAML_LOG
+}
+
+log_test_status() {
+     yml_log_test_status $@ >> $YAML_LOG
+}
+
+log_sub_test_begin() {
+    yml_log_sub_test_begin $@ >> $YAML_LOG
+}
+
+log_sub_test_end() {
+    yml_log_sub_test_end $@ >> $YAML_LOG
+}
+
+run_llverdev()
+{
+        local dev=$1
+        local devname=$(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 $5}')
+
+        size=$(($size / 1024 / 1024)) # Gb
+
+        local partial_arg=""
+        # Run in partial (fast) mode if the size
+        # of a partition > 10 GB
+        [ $size -gt 10 ] && partial_arg="-p"
+
+        llverdev --force $partial_arg $dev
+}
+
+remove_mdt_files() {
+    local facet=$1
+    local mdtdev=$2
+    shift 2
+    local files="$@"
+    local mntpt=${MOUNT%/*}/$facet
+
+    echo "removing files from $mdtdev on $facet: $files"
+    mount -t $FSTYPE $MDS_MOUNT_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=${MOUNT%/*}/$facet
+
+    echo "duplicating files on $mdtdev on $facet: $files"
+    mkdir -p $mntpt || return $?
+    mount -t $FSTYPE $MDS_MOUNT_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 $?
+    done
+    do_umount
+}
+
+run_sgpdd () {
+    local devs=${1//,/ }
+    shift
+    local params=$@
+    local rslt=$TMP/sgpdd_survey
+
+    # sgpdd-survey cleanups ${rslt}.* files
+
+    local cmd="rslt=$rslt $params scsidevs=\"$devs\" $SGPDDSURVEY"
+    echo + $cmd
+    eval $cmd
+    cat ${rslt}.detail
+}
+
+# returns the canonical name for an ldiskfs device
+ldiskfs_canon() {
+        local dev="$1"
+        local facet="$2"
+
+        do_facet $facet "dv=\\\$(lctl get_param -n $dev);
+if foo=\\\$(lvdisplay -c \\\$dv 2>/dev/null); then
+    echo dm-\\\${foo##*:};
+else
+    echo \\\$(basename \\\$dv);
+fi;"
+}
+
+is_sanity_benchmark() {
+    local benchmarks="dbench bonnie iozone fsx"
+    local suite=$1
+    for b in $benchmarks; do
+        if [ "$b" == "$suite" ]; then
+            return 0
+        fi
+    done
+    return 1
+}
+