X-Git-Url: https://git.whamcloud.com/?p=fs%2Flustre-release.git;a=blobdiff_plain;f=lustre%2Ftests%2Ftest-framework.sh;h=0165fee28aed461a7d871aec1e5294be6b2812c8;hp=61e909ecfa26d2bc7a928e95e8fe9f372d927340;hb=8f6d85eebc312b64d8e8a35b0be3ae137a50a45c;hpb=82a1b0453c8dccf63d23f3c96f1b2355524b952c

diff --git a/lustre/tests/test-framework.sh b/lustre/tests/test-framework.sh
index 61e909e..0165fee 100644
--- a/lustre/tests/test-framework.sh
+++ b/lustre/tests/test-framework.sh
@@ -15,12 +15,36 @@ export GSS=false
 export GSS_KRB5=false
 export GSS_PIPEFS=false
 export IDENTITY_UPCALL=default
+export QUOTA_AUTO=1
 
 #export PDSH="pdsh -S -Rssh -w"
 
 # 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:-${LUSTRE}/tests/cfg}
+
+EXCEPT_LIST_FILE=${EXCEPT_LIST_FILE:-${LUSTRE_TESTS_CFG_DIR}/tests-to-skip.sh}
+
+if [ -f "$EXCEPT_LIST_FILE" ]; then
+    echo "Reading test skip list from $EXCEPT_LIST_FILE"
+    cat $EXCEPT_LIST_FILE
+    . $EXCEPT_LIST_FILE
+fi
+
+if [[ -x $LUSTRE/utils/llobdstat ]]; then
+        LLOBDSTAT=$LUSTRE/utils/llobdstat
+elif [[ -x /usr/bin/llobdstat ]]; then
+        LLOBDSTAT=/usr/bin/llobdstat
+else
+        # Good luck
+        LLOBDSTAT=llobdstat
+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=""
@@ -43,17 +67,17 @@ usage() {
 
 print_summary () {
     trap 0
-    [ "$TESTSUITE" == "lfscktest" ] && return 0
+    [ "$TESTSUITE" == "lfsck" ] && 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)"
     echo "------------------------------------------------------------------------------------"
-    for O in $TESTSUITE_LIST; do
+    for O in $DEFAULT_SUITES; do
         local skipped=""
         local slow=""
-        local o=$(echo $O | tr "[:upper:]" "[:lower:]")
+        O=$(echo $O  | tr "-" "_" | tr "[:lower:]" "[:upper:]")
+        local o=$(echo $O  | tr "[:upper:]" "[:lower:]")
         o=${o//_/-}
-        o=${o//tyn/tyN}
         local log=${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')
@@ -63,7 +87,8 @@ print_summary () {
 
     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:
@@ -76,7 +101,8 @@ print_summary () {
         fi
     done
 
-    for O in $TESTSUITE_LIST; do
+    for O in $DEFAULT_SUITES; do
+        O=$(echo $O  | tr "-" "_" | tr "[:lower:]" "[:upper:]")
         [ "${!O}" = "done" -o "${!O}" = "no" ] || \
             printf "$form" "UNFINISHED" "$O" ""
     done
@@ -98,6 +124,10 @@ init_test_env() {
     #[ -d /r ] && export ROOT=${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:$LUSTRE/utils
@@ -106,16 +136,19 @@ init_test_env() {
         export PATH=$PATH:$LUSTRE/utils/gss
     fi
     if ! echo $PATH | grep -q $LUSTRE/tests; then
-	export PATH=$PATH:$LUSTRE/tests
+        export PATH=$PATH:$LUSTRE/tests
     fi
+    export LST=${LST:-"$LUSTRE/../lnet/utils/lst"}
+    [ ! -f "$LST" ] && export LST=$(which lst)
     export MDSRATE=${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:$LUSTRE/tests/mpi
     fi
+    export RSYNC_RSH=${RSYNC_RSH:-rsh}
     export LCTL=${LCTL:-"$LUSTRE/utils/lctl"}
     [ ! -f "$LCTL" ] && export LCTL=$(which lctl)
     export LFS=${LFS:-"$LUSTRE/utils/lfs"}
@@ -155,7 +188,7 @@ 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
             ;;
@@ -204,21 +237,69 @@ case `uname -r` in
     *) EXT=".ko"; USE_QUOTA=yes;;
 esac
 
+
+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
     BASE=`basename $module $EXT`
-    lsmod | grep -q ${BASE} || \
-      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 [ $module = lnet ]; then
+                # OK, this is a bit wordy...
+                local arg accept_all_present=false
+                for arg in "$@"; do
+                    [ "$arg" = accept=all ] && accept_all_present=true
+                done
+                $accept_all_present || set -- "$@" accept=all
+            fi
+        fi
+    fi
+
+    [ $# -gt 0 ] && echo "${module} options: '$*'"
+
+    module_loaded ${BASE} && return
+
+    # Note that insmod will ignore anything in modprobe.conf, which is why we're
+    # passing options on the command-line.
+    if [ "$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
 }
@@ -226,10 +307,12 @@ load_module() {
 load_modules_local() {
     if [ -n "$MODPROBE" ]; then
         # use modprobe
-    return 0
+        echo "Using modprobe to load modules"
+        return 0
     fi
     if [ "$HAVE_MODULES" = true ]; then
-    # we already loaded
+        # we already loaded
+        echo "Modules already loaded"
         return 0
     fi
     HAVE_MODULES=true
@@ -238,15 +321,7 @@ load_modules_local() {
     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
@@ -319,9 +394,11 @@ 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
@@ -332,6 +409,17 @@ unload_modules() {
     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
@@ -345,27 +433,41 @@ 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
@@ -374,33 +476,37 @@ 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 $GSS_PIPEFS; then
-        check_gss_daemon_facet client lgssd
+        list=$(comma_list $(mdts_nodes))
+        check_gss_daemon_nodes $list lgssd || return 6
+    fi
+    if $GSS_PIPEFS; then
+        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
@@ -538,9 +644,7 @@ restore_quota_type () {
 setup_quota(){
     local mntpt=$1
 
-    # We need:
-    # 1. run quotacheck only if quota is off
-    # 2. save the original quota_type params, restore them after testing
+    # We need save the original quota_type params, and restore them after testing
 
     # Suppose that quota type the same on mds and ost
     local quota_type=$(quota_type | grep MDT | cut -d "=" -f2)
@@ -549,6 +653,9 @@ setup_quota(){
     if [ "$quota_type" != "$QUOTA_TYPE" ]; then
         export old_QUOTA_TYPE=$quota_type
         quota_save_version $QUOTA_TYPE
+    else
+        qtype=$(tr -c -d "ug" <<< $QUOTA_TYPE)
+        $LFS quotacheck -$qtype $mntpt || error "quotacheck has failed for $type"
     fi
 
     local quota_usrs=$QUOTA_USERS
@@ -691,14 +798,13 @@ sanity_mount_check () {
 
 # mount clients if not mouted
 zconf_mount_clients() {
-    local OPTIONS
     local clients=$1
     local mnt=$2
-
+    local OPTIONS=${3:-$MOUNTOPT}
 
     # Only supply -o to mount if we have options
-    if [ -n "$MOUNTOPT" ]; then
-        OPTIONS="-o $MOUNTOPT"
+    if [ "$OPTIONS" ]; then
+        OPTIONS="-o $OPTIONS"
     fi
     local device=$MGSNID:/$FSNAME
     if [ -z "$mnt" -o -z "$FSNAME" ]; then
@@ -738,7 +844,7 @@ zconf_umount_clients() {
     echo "Stopping clients: $clients $mnt (opts:$force)"
     do_nodes $clients "running=\\\$(grep -c $mnt' ' /proc/mounts);
 if [ \\\$running -ne 0 ] ; then
-echo Stopping client \\\$(hostname) client $mnt opts:$force;
+echo Stopping client \\\$(hostname) $mnt opts:$force;
 lsof -t $mnt || need_kill=no;
 if [ "x$force" != "x" -a "x\\\$need_kill" != "xno" ]; then
     pids=\\\$(lsof -t $mnt | sort -u);
@@ -853,6 +959,8 @@ start_client_loads () {
         testnum=$((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 
@@ -966,7 +1074,7 @@ cleanup_check() {
     [ "`lctl dl 2> /dev/null | wc -l`" -gt 0 ] && lctl dl && \
         echo "$0: lustre didn't clean up..." 1>&2 && return 202 || true
 
-    if [ "`/sbin/lsmod 2>&1 | egrep 'lnet|libcfs'`" ]; then
+    if module_loaded lnet || module_loaded libcfs; then
         echo "$0: modules still loaded..." 1>&2
         /sbin/lsmod 1>&2
         return 203
@@ -1013,12 +1121,13 @@ wait_delete_completed () {
         sleep 1
         TOTAL=`lctl get_param -n osc.*.kbytesavail | \
                awk 'BEGIN{total=0}; {total+=$1}; END{print total}'`
-        [ "$TOTAL" -eq "$TOTALPREV" ] && break
+        [ "$TOTAL" -eq "$TOTALPREV" ] && return 0
         echo "Waiting delete completed ... prev: $TOTALPREV current: $TOTAL "
         TOTALPREV=$TOTAL
         WAIT=$(( WAIT + 1))
     done
-    echo "Delete completed."
+    echo "Delete is not completed in $MAX_WAIT sec"
+    return 1
 }
 
 wait_for_host() {
@@ -1036,12 +1145,12 @@ wait_for() {
 wait_recovery_complete () {
     local facet=$1
 
-    # Use default policy if $2 is not passed by caller. 
+    # 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:-$(( TIMEOUT * 4 ))}
- 
+
     local var_svc=${facet}_svc
     local procfile="*.${!var_svc}.recovery_status"
     local WAIT=0
@@ -1058,6 +1167,57 @@ wait_recovery_complete () {
     return 1
 }
 
+wait_mds_ost_sync () {
+    # just because recovery is done doesn't mean we've finished
+    # orphan cleanup. Wait for llogs to get synchronized.
+    echo "Waiting for orphan cleanup..."
+    # MAX value includes time needed for MDS-OST reconnection
+    local MAX=$(( TIMEOUT * 2 ))
+    local WAIT=0
+    while [ $WAIT -lt $MAX ]; do
+        local -a sync=($(do_nodes $(comma_list $(osts_nodes)) \
+            "$LCTL get_param -n obdfilter.*.mds_sync"))
+        local con=1
+        for ((i=0; i<${#sync[@]}; i++)); do
+            [ ${sync[$i]} -eq 0 ] && continue
+            # there is a not finished MDS-OST synchronization
+            con=0
+            break;
+        done
+        sleep 2 # increase waiting time and cover statfs cache
+        [ ${con} -eq 1 ] && return 0
+        echo "Waiting $WAIT secs for $facet mds-ost sync done."
+        WAIT=$((WAIT + 2))
+    done
+    echo "$facet recovery not done in $MAX sec. $STATUS"
+    return 1
+}
+
+wait_destroy_complete () {
+    echo "Waiting for destroy to be done..."
+    # 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 WAIT=0
+    while [ $WAIT -lt $MAX ]; do
+        local -a RPCs=($($LCTL get_param -n osc.*.destroys_in_flight))
+        local con=1
+        for ((i=0; i<${#RPCs[@]}; i++)); do
+            [ ${RPCs[$i]} -eq 0 ] && continue
+            # there are still some destroy RPCs in flight
+            con=0
+            break;
+        done
+        sleep 1
+        [ ${con} -eq 1 ] && return 0 # done waiting
+        echo "Waiting $WAIT secs for destroys to be done."
+        WAIT=$((WAIT + 1))
+    done
+    echo "Destroys weren't done in $MAX sec."
+    return 1
+}
+
 wait_exit_ST () {
     local facet=$1
 
@@ -1111,15 +1271,31 @@ wait_remote_prog () {
     return $rc
 }
 
-client_df() {
+clients_up() {
     # not every config has many clients
+    sleep 1
     if [ ! -z "$CLIENTS" ]; then
-        $PDSH $CLIENTS "df $MOUNT" > /dev/null
+        $PDSH $CLIENTS "stat -f $MOUNT" > /dev/null
     else
-	df $MOUNT > /dev/null
+        stat -f $MOUNT > /dev/null
     fi
 }
 
+client_up() {
+    local client=$1
+    # usually checked on particular client or locally
+    sleep 1
+    if [ ! -z "$client" ]; then
+        $PDSH $client "stat -f $MOUNT" > /dev/null
+    else
+        stat -f $MOUNT > /dev/null
+    fi
+}
+
+client_evicted() {
+    ! client_up $1
+}
+
 client_reconnect() {
     uname -n >> $MOUNT/recon
     if [ -z "$CLIENTS" ]; then
@@ -1195,7 +1371,7 @@ ost_evict_client() {
 
 fail() {
     facet_failover $* || error "failover: $?"
-    client_df || error "post-failover df: $?"
+    clients_up || error "post-failover df: $?"
 }
 
 fail_nodf() {
@@ -1208,9 +1384,8 @@ fail_abort() {
     stop $facet
     change_active $facet
     mount_facet $facet -o abort_recovery
-    client_df || echo "first df failed: $?"
-    sleep 1
-    client_df || error "post-failover df: $?"
+    clients_up || echo "first df failed: $?"
+    clients_up || error "post-failover df: $?"
 }
 
 do_lmc() {
@@ -1349,32 +1524,47 @@ 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
@@ -1386,11 +1576,11 @@ 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
@@ -1407,9 +1597,9 @@ do_nodes() {
     fi
 
     if $verbose ; then
-        $myPDSH $rnodes "(PATH=\$PATH:$RLUSTRE/utils:$RLUSTRE/tests:/sbin:/usr/sbin; cd $RPWD; sh -c \"$@\")"
+        $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/\w+:\s//g"
     fi
     return ${PIPESTATUS[0]}
 }
@@ -1422,6 +1612,10 @@ do_facet() {
     do_node $HOST "$@"
 }
 
+do_nodesv() {
+    do_nodes --verbose "$@"
+}
+
 add() {
     local facet=$1
     shift
@@ -1478,6 +1672,10 @@ stopall() {
         rm -f $TMP/ost${num}active
     done
 
+    if ! combined_mgs_mds ; then
+        stop mgs
+    fi
+
     return 0
 }
 
@@ -1495,6 +1693,10 @@ mdsmkfsopts()
     test $nr = 1 && echo -n $MDS_MKFS_OPTS || echo -n $MDSn_MKFS_OPTS
 }
 
+combined_mgs_mds () {
+    [[ $MDSDEV1 = $MGSDEV ]] && [[ $mds1_HOST = $mgs_HOST ]]
+}
+
 formatall() {
     if [ "$IAMDIR" == "yes" ]; then
         MDS_MKFS_OPTS="$MDS_MKFS_OPTS --iam-dir"
@@ -1503,18 +1705,12 @@ formatall() {
 
     [ "$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
+    if ! combined_mgs_mds ; then
         add mgs $mgs_MKFS_OPTS $FSTYPE_OPT --reformat $MGSDEV || exit 10
     fi
 
@@ -1578,8 +1774,8 @@ 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 () {
@@ -1608,12 +1804,12 @@ 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
+        if ! combined_mgs_mds ; then
             start mgs $MGSDEV $mgs_MOUNT_OPTS
         fi
 
@@ -1628,9 +1824,9 @@ 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)
@@ -1645,16 +1841,21 @@ 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
 
@@ -1666,6 +1867,7 @@ 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
@@ -1673,7 +1875,7 @@ setupall() {
 }
 
 mounted_lustre_filesystems() {
-	awk '($3 ~ "lustre" && $1 ~ ":") { print $2 }' /proc/mounts
+        awk '($3 ~ "lustre" && $1 ~ ":") { print $2 }' /proc/mounts
 }
 
 init_facet_vars () {
@@ -1745,11 +1947,6 @@ 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)
@@ -1765,12 +1962,17 @@ init_param_vars () {
     osc_ensure_active $SINGLEMDS M $TIMEOUT
     osc_ensure_active client c $TIMEOUT
 
-    if [ x"$(som_check)" = x"enabled" ]; then
-        ENABLE_QUOTA=""
-    fi
-    if [ "$ENABLE_QUOTA" ]; then
-        setup_quota $MOUNT  || return 2
+    if [ $QUOTA_AUTO -ne 0 ]; then
+        if [ "$ENABLE_QUOTA" ]; then
+            echo "enable quota as required"
+            setup_quota $MOUNT || return 2
+        else
+            echo "disable quota as required"
+            $LFS quotaoff -ug $MOUNT > /dev/null 2>&1
+        fi
     fi
+
+    return 0
 }
 
 nfs_client_mode () {
@@ -1788,9 +1990,7 @@ nfs_client_mode () {
     return 1
 }
 
-check_config () {
-    nfs_client_mode && return
-
+check_config_client () {
     local mntpt=$1
 
     local mounted=$(mount | grep " $mntpt ")
@@ -1809,8 +2009,8 @@ check_config () {
         if [[ x$mgc != xMGC$MGSNID ]]; then
             if [ "$mgs_HOST" ]; then
                 local mgc_ip=$(ping -q -c1 -w1 $mgs_HOST | grep PING | awk '{print $3}' | sed -e "s/(//g" -e "s/)//g")
-                [[ x$mgc = xMGC$mgc_ip@$NETTYPE ]] ||
-                    error_exit "MGSNID=$MGSNID, mounted: $mounted, MGC : $mgc"
+#                [[ x$mgc = xMGC$mgc_ip@$NETTYPE ]] ||
+#                    error_exit "MGSNID=$MGSNID, mounted: $mounted, MGC : $mgc"
             fi
         fi
         return 0
@@ -1826,10 +2026,20 @@ check_config () {
     mgshost=$(echo $mgshost | awk -F: '{print $1}')
 
 #    if [ "$mgshost" != "$myMGS_host" ]; then
-#            error_exit "Bad config file: lustre is mounted with mgs $mgshost, but mgs_HOST=$mgs_HOST, NETTYPE=$NETTYPE
+#            log "Bad config file: lustre is mounted with mgs $mgshost, but mgs_HOST=$mgs_HOST, NETTYPE=$NETTYPE
 #                   Please use correct config or set mds_HOST correctly!"
 #    fi
 
+}
+
+check_config_clients () {
+    local clients=${CLIENTS:-$HOSTNAME}
+    local mntpt=$1
+
+    nfs_client_mode && return
+
+    do_rpc_nodes $clients check_config_client $mntpt
+
     sanity_mount_check ||
         error "environments are insane!"
 }
@@ -1843,37 +2053,65 @@ check_timeout () {
     fi
 }
 
+is_mounted () {
+    local mntpt=$1
+    local mounted=$(mounted_lustre_filesystems)
+
+    echo $mounted' ' | grep -w -q $mntpt' '
+}
+
 check_and_setup_lustre() {
     nfs_client_mode && return
 
     local MOUNTED=$(mounted_lustre_filesystems)
 
     local do_check=true
-    # MOUNT is not mounted
-    if [ -z "$MOUNTED" ] || ! $(echo $MOUNTED | grep -w -q $MOUNT); then
+    # 1.
+    # both MOUNT and MOUNT2 are not mounted
+    if ! is_mounted $MOUNT && ! is_mounted $MOUNT2; then
         [ "$REFORMAT" ] && formatall
+        # setupall mounts both MOUNT and MOUNT2 (if MOUNT_2 is set)
         setupall
-        MOUNTED=$(mounted_lustre_filesystems | head -1)
-        [ -z "$MOUNTED" ] && error "NAME=$NAME not mounted"
+        is_mounted $MOUNT || error "NAME=$NAME not mounted"
         export I_MOUNTED=yes
         do_check=false
-
-    # MOUNT and MOUNT2 are mounted
-    elif $(echo $MOUNTED | grep -w -q $MOUNT2); then
-
-        # MOUNT2 is mounted,  MOUNT_2 is not set
-        if ! [ "$MOUNT_2" ]; then
-            zconf_umount `hostname` $MOUNT2
-            export I_UMOUNTED2=yes
-
-        # MOUNT2 is mounted, MOUNT_2 is set
-        else
-            check_config $MOUNT2
-        fi 
+    # 2.
+    # MOUNT2 is mounted
+    elif is_mounted $MOUNT2; then
+            # 3.
+            # MOUNT2 is mounted, while MOUNT_2 is not set
+            if ! [ "$MOUNT_2" ]; then
+                cleanup_mount $MOUNT2
+                export I_UMOUNTED2=yes
+
+            # 4.
+            # MOUNT2 is mounted, MOUNT_2 is set
+            else
+                # FIXME: what to do if check_config failed?
+                # i.e. if:
+                # 1) remote client has mounted other Lustre fs ?
+                # 2) it has insane env ?
+                # let's try umount MOUNT2 on all clients and mount it again:
+                if ! check_config_clients $MOUNT2; then
+                    cleanup_mount $MOUNT2
+                    restore_mount $MOUNT2
+                    export I_MOUNTED2=yes
+                fi
+            fi 
+
+    # 5.
+    # MOUNT is mounted MOUNT2 is not mounted
+    elif [ "$MOUNT_2" ]; then
+        restore_mount $MOUNT2
+        export I_MOUNTED2=yes
     fi
 
     if $do_check; then
-        check_config $MOUNT
+        # FIXME: what to do if check_config failed?
+        # i.e. if:
+        # 1) remote client has mounted other Lustre fs?
+        # 2) lustre is mounted on remote_clients atall ?
+        check_config_clients $MOUNT
         init_facets_vars
         init_param_vars
 
@@ -1882,35 +2120,58 @@ 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
 }
 
+restore_mount () {
+   local clients=${CLIENTS:-$HOSTNAME}
+   local mntpt=$1
+
+   zconf_mount_clients $clients $mntpt
+}
+
+cleanup_mount () {
+    local clients=${CLIENTS:-$HOSTNAME}
+    local mntpt=$1
+
+    zconf_umount_clients $clients $mntpt    
+}
+
 cleanup_and_setup_lustre() {
     if [ "$ONLY" == "cleanup" -o "`mount | grep $MOUNT`" ]; then
         lctl set_param debug=0 || true
         cleanupall
         if [ "$ONLY" == "cleanup" ]; then
-    	    exit 0
+            exit 0
         fi
     fi
     check_and_setup_lustre
 }
 
 check_and_cleanup_lustre() {
-    if [ "`mount | grep $MOUNT`" ]; then
+    if is_mounted $MOUNT; then
         [ -n "$DIR" ] && rm -rf $DIR/[Rdfs][0-9]*
         [ "$ENABLE_QUOTA" ] && restore_quota_type || true
     fi
+
     if [ "$I_UMOUNTED2" = "yes" ]; then
-        mount_client $MOUNT2 || error "restore MOUNT2 failed"
+        restore_mount $MOUNT2 || error "restore $MOUNT2 failed"
+    fi
+
+    if [ "$I_MOUNTED2" = "yes" ]; then
+        cleanup_mount $MOUNT2
     fi
 
     if [ "$I_MOUNTED" = "yes" ]; then
         cleanupall -f || error "cleanup failed"
+        unset I_MOUNTED
     fi
-    unset I_MOUNTED
 }
 
 #######
@@ -1982,9 +2243,9 @@ testslist_filter () {
     local start_at=$START_AT
     local stop_at=$STOP_AT
 
-    local var=${TESTSUITE}_START_AT
+    local var=${TESTSUITE//-/_}_START_AT
     [ x"${!var}" != x ] && start_at=${!var}
-    var=${TESTSUITE}_STOP_AT
+    var=${TESTSUITE//-/_}_STOP_AT
     [ x"${!var}" != x ] && stop_at=${!var}
 
     sed -n 's/^test_\([^ (]*\).*/\1/p' $script | \
@@ -2028,9 +2289,9 @@ 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
 }
 
@@ -2042,15 +2303,15 @@ 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
 }
@@ -2169,10 +2430,6 @@ pgcache_empty() {
             return 1
         fi
     done
-    if [[ $MDSDEV1 != $MGSDEV ]]; then
-        stop mgs 
-    fi
-
     return 0
 }
 
@@ -2218,8 +2475,6 @@ stop_full_debug_logging() {
 
 error_noexit() {
     local TYPE=${TYPE:-"FAIL"}
-    local ERRLOG
-    lctl set_param fail_loc=0 2>/dev/null || true
 
     local dump=true
     # do not dump logs if $1=false
@@ -2230,25 +2485,23 @@ 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
 }
 
 error() {
     error_noexit "$@"
-    $FAIL_ON_ERROR && exit 1 || true
+    exit 1
 }
 
 error_exit() {
-    error_noexit "$@"
-    exit 1
+    error "$@"
 }
 
 # use only if we are ignoring failures for this test, bugno required.
@@ -2290,15 +2543,27 @@ build_test_filter() {
     done
     for G in $GRANT_CHECK_LIST; do
         eval GCHECK_ONLY_${G}=true
-   	done
+        done
 }
 
 basetest() {
-    echo ${1%%[a-z]*}
+    if [[ $1 = [a-z]* ]]; then
+        echo $1
+    else
+        echo ${1%%[a-z]*}
+    fi
 }
 
 # 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
 
@@ -2307,13 +2572,13 @@ 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"
@@ -2346,23 +2611,18 @@ 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() {
     echo "$*"
-    lsmod | grep lnet > /dev/null || load_modules
+    module_loaded lnet || load_modules
 
     local MSG="$*"
     # Get rid of '
@@ -2378,16 +2638,21 @@ 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
 }
 
 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 " " $@
 }
 
 check_mds() {
@@ -2402,33 +2667,80 @@ 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
+    (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` - $BEFORE))
+    pass "(${duration}s)"
+    [ -f $LOGDIR/err ] && TEST_ERROR=$(cat $LOGDIR/err)
+    log_sub_test test_${1} $TEST_STATUS $duration "$RC" "$TEST_ERROR"
+
+    if [ -f $LOGDIR/err ]; then
+        $FAIL_ON_ERROR && exit $RC
+    fi
+
+    return 0
 }
 
 canonical_path() {
@@ -2438,50 +2750,50 @@ canonical_path() {
 sync_clients() {
     [ -d $DIR1 ] && cd $DIR1 && sync; sleep 1; sync
     [ -d $DIR2 ] && cd $DIR2 && sync; sleep 1; sync
-	cd $SAVE_PWD
+        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
+        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: ${server_grant}"
-		return 1
-	else
-		echo "pass"
-	fi
+                client_grant=$((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
 }
 
 ########################
@@ -2517,6 +2829,13 @@ remote_mds_nodsh()
     remote_mds && [ "$PDSH" = "no_dsh" -o -z "$PDSH" -o -z "$mds_HOST" ]
 }
 
+require_dsh_mds()
+{
+        remote_mds_nodsh && echo "SKIP: $TESTSUITE: remote MDS with nodsh" && \
+            MSKIPPED=1 && return 1
+        return 0
+}
+
 remote_ost ()
 {
     local node
@@ -2532,6 +2851,13 @@ remote_ost_nodsh()
     remote_ost && [ "$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 
@@ -2539,6 +2865,12 @@ 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 $(nodes_list)))
+}
+
 mdts_nodes () {
     local MDSNODES
     local NODES_sort
@@ -2658,9 +2990,10 @@ mixed_mdt_devs () {
 generate_machine_file() {
     local nodes=${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
 }
 
@@ -2671,6 +3004,17 @@ 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
@@ -2779,6 +3123,10 @@ 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
@@ -2816,7 +3164,7 @@ 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
@@ -2845,19 +3193,19 @@ 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
@@ -2865,12 +3213,13 @@ get_mds_dir () {
     local dir=$1
     local file=$dir/f0.get_mds_dir_tmpfile
 
+    mkdir -p $dir
     rm -f $file
     sleep 1
     local iused=$(lfs df -i $dir | grep MDT | awk '{print $3}')
     local -a oldused=($iused)
 
-    touch $file
+    openfile -f O_CREAT:O_LOV_DELAY_CREATE -m 0644 $file > /dev/null
     sleep 1
     iused=$(lfs df -i $dir | grep MDT | awk '{print $3}')
     local -a newused=($iused)
@@ -2887,7 +3236,10 @@ get_mds_dir () {
 }
 
 mdsrate_cleanup () {
-    mpi_run -np $1 -machinefile $2 ${MDSRATE} --unlink --nfiles $3 --dir $4 --filefmt $5 $6
+    if [ -d $4 ]; then
+        mpi_run -np $1 -machinefile $2 ${MDSRATE} --unlink --nfiles $3 --dir $4 --filefmt $5 $6
+        rmdir $4
+    fi
 }
 
 delayed_recovery_enabled () {
@@ -2960,12 +3312,13 @@ get_osc_import_name() {
 }
 
 wait_import_state () {
-    local expected=$1
-    local CONN_PROC=$2
+    local facet=$1
+    local expected=$2
+    local CONN_PROC=$3
     local CONN_STATE
     local i=0
 
-    CONN_STATE=$($LCTL get_param -n $CONN_PROC 2>/dev/null | cut -f2)
+    CONN_STATE=$(do_facet $facet $LCTL get_param -n $CONN_PROC | awk '/state/ {print $2}')
     while [ "${CONN_STATE}" != "${expected}" ]; do
         if [ "${expected}" == "DISCONN" ]; then
             # for disconn we can check after proc entry is removed
@@ -2978,7 +3331,7 @@ wait_import_state () {
         [ $i -ge $(($TIMEOUT * 3 / 2)) ] && \
             error "can't put import for $CONN_PROC into ${expected} state" && return 1
         sleep 1
-        CONN_STATE=$($LCTL get_param -n $CONN_PROC 2>/dev/null | cut -f2)
+        CONN_STATE=$(do_facet $facet $LCTL get_param -n $CONN_PROC | awk '/state/ {print $2}')
         i=$(($i + 1))
     done
 
@@ -2991,62 +3344,60 @@ wait_osc_import_state() {
     local ost_facet=$2
     local expected=$3
     local ost=$(get_osc_import_name $facet $ost_facet)
-    local CONN_PROC
-    local CONN_STATE
-    local i=0
+    local CONN_PROC="osc.${ost}.import"
 
-    CONN_PROC="osc.${ost}.ost_server_uuid"
-    CONN_STATE=$(do_facet $facet 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 ${ost}(${ost_facet}) into ${expected} state" && return 1
-        sleep 1
-        CONN_STATE=$(do_facet $facet lctl get_param -n $CONN_PROC 2>/dev/null | cut -f2)
-        i=$(($i + 1))
-    done
-
-    log "${ost_facet} now in ${CONN_STATE} state"
+    wait_import_state $facet $expected $CONN_PROC || return 1
     return 0
 }
 
 get_clientmdc_proc_path() {
-    echo "${1}-mdc-*"
+    local mdc=$(convert_facet2label $1)
+
+    echo "${mdc}-mdc-*"
+}
+
+wait_mdc_import_state() {
+    local facet=$1
+    local expected=$2
+    local mdc=$(get_clientmdc_proc_path $facet)
+    local CONN_PROC="mdc.${mdc}.import"
+
+    wait_import_state client $expected $CONN_PROC || return 1
+    return 0
 }
 
 do_rpc_nodes () {
     local list=$1
     shift
 
-    do_nodes --verbose $list "PATH=$LUSTRE/tests/:$PATH sh rpc.sh $@ " 
+    # Add paths to lustre tests for 32 and 64 bit systems.
+    local RPATH="$RLUSTRE/tests:/usr/lib/lustre/tests:/usr/lib64/lustre/tests:$PATH"
+    do_nodesv $list "PATH=$RPATH sh rpc.sh $@ "
 }
 
-wait_clients_import_state () {
-    local list=$1
-    local facet=$2
-    local expected=$3
+wait_client_import_state () {
+    local facet=$1
+    local expected=$2
     shift
 
-    local label=$(convert_facet2label $facet)
-    local proc_path
     case $facet in
-        ost* ) proc_path="osc.$(get_clientosc_proc_path $label).ost_server_uuid" ;;
-        mds* ) proc_path="mdc.$(get_clientmdc_proc_path $label).mds_server_uuid" ;;
-        *) error "unknown facet!" ;;
+        ost* ) wait_osc_import_state client $facet $expected || return 1;;
+        mds* ) wait_mdc_import_state $facet $expected || return 1 ;;
+           * ) error "unknown facet!"
+               return 1 ;;
     esac
+    return 0
+}
 
+wait_clients_import_state () {
+    local list=$1
+    shift
 
-    if ! do_rpc_nodes $list wait_import_state $expected $proc_path; then
-        error "import is not in ${expected} state"
+    if ! do_rpc_nodes $list wait_client_import_state "$@"; then
+        error "import is not in expected state"
         return 1
     fi
+    return 0
 }
 
 oos_full() {
@@ -3066,6 +3417,102 @@ oos_full() {
         return $OSCFULL
 }
 
+pool_list () {
+   do_facet mgs lctl pool_list $1
+}
+
+create_pool() {
+    local fsname=${1%%.*}
+    local poolname=${1##$fsname.}
+
+    do_facet mgs lctl pool_new $1
+    local RC=$?
+    # get param should return err unless pool is created
+    [[ $RC -ne 0 ]] && return $RC
+
+    wait_update $HOSTNAME "lctl get_param -n lov.$fsname-*.pools.$poolname \
+        2>/dev/null || echo foo" "" || RC=1
+    if [[ $RC -eq 0 ]]; then
+        add_pool_to_list $1
+    else
+        error "pool_new failed $1"
+    fi
+    return $RC
+}
+
+add_pool_to_list () {
+    local fsname=${1%%.*}
+    local poolname=${1##$fsname.}
+
+    local listvar=${fsname}_CREATED_POOLS
+    eval export ${listvar}=$(expand_list ${!listvar} $poolname)
+}
+
+remove_pool_from_list () {
+    local fsname=${1%%.*}
+    local poolname=${1##$fsname.}
+
+    local listvar=${fsname}_CREATED_POOLS
+    eval export ${listvar}=$(exclude_items_from_list ${!listvar} $poolname)
+}
+
+destroy_pool_int() {
+    local ost
+    local OSTS=$(do_facet $SINGLEMDS lctl pool_list $1 | \
+        awk '$1 !~ /^Pool:/ {print $1}')
+    for ost in $OSTS; do
+        do_facet mgs lctl pool_remove $1 $ost
+    done
+    do_facet mgs lctl pool_destroy $1
+}
+
+# <fsname>.<poolname> or <poolname>
+destroy_pool() {
+    local fsname=${1%%.*}
+    local poolname=${1##$fsname.}
+
+    [[ x$fsname = x$poolname ]] && fsname=$FSNAME
+
+    local RC
+
+    pool_list $fsname.$poolname || return $?
+
+    destroy_pool_int $fsname.$poolname
+    RC=$?
+    [[ $RC -ne 0 ]] && return $RC
+
+    wait_update $HOSTNAME "lctl get_param -n lov.$fsname-*.pools.$poolname \
+      2>/dev/null || echo foo" "foo" || RC=1
+
+    if [[ $RC -eq 0 ]]; then
+        remove_pool_from_list $fsname.$poolname
+    else
+        error "destroy pool failed $1"
+    fi
+    return $RC
+}
+
+destroy_pools () {
+    local fsname=${1:-$FSNAME}
+    local poolname
+    local listvar=${fsname}_CREATED_POOLS
+
+    pool_list $fsname
+
+    [ x${!listvar} = x ] && return 0
+
+    echo destroy the created pools: ${!listvar}
+    for poolname in ${!listvar//,/ }; do
+        destroy_pool $fsname.$poolname 
+    done
+}
+
+cleanup_pools () {
+    local fsname=${1:-$FSNAME}
+    trap 0
+    destroy_pools $fsname
+}
+
 gather_logs () {
     local list=$1
 
@@ -3076,27 +3523,30 @@ gather_logs () {
     # 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'*'
-    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}*
+    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
+
+    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
 }
@@ -3107,3 +3557,385 @@ cleanup_logs () {
     [ -n ${TESTSUITE} ] && do_nodes $list "rm -f $TMP/*${TESTSUITE}*" || true
 }
 
+do_ls () {
+    local mntpt_root=$1
+    local num_mntpts=$2
+    local dir=$3
+    local i
+    local cmd
+    local pids
+    local rc=0
+
+    for i in $(seq 0 $num_mntpts); do
+        cmd="ls -laf ${mntpt_root}$i/$dir"
+        echo + $cmd;
+        $cmd > /dev/null &
+        pids="$pids $!"
+    done
+    echo pids=$pids
+    for pid in $pids; do
+        wait $pid || rc=$?
+    done
+
+    return $rc
+}
+
+get_clients_mount_count () {
+    local clients=${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 -le $M ]; do
+            R=$((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
+
+    for num in `seq $MDSCOUNT`; do
+        output=`do_facet mds$num lctl get_param -n osc.*OST*-osc-MDT*.$PROC_CLI 2>/dev/null`
+        tmpcnt=`count_flvr "$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 $dir)}     # number expected
+
+    local res=0
+
+    for ((i=0;i<20;i++)); do
+        echo -n "checking..."
+        res=$(do_check_flavor $dir $flavor)
+        if [ $res -eq $expect ]; then
+            echo "found $res $flavor connections of $dir, OK"
+            return 0
+        else
+            echo "found $res $flavor connections of $dir, not ready ($expect)"
+            sleep 4
+        fi
+    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 $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_sub_test() {
+    yml_log_sub_test $@ >> $YAML_LOG
+}
+
+run_llverdev()
+{
+        local dev=$1
+        local devname=$(basename $1)
+        local size=$(grep "$devname"$ /proc/partitions | awk '{print $3}')
+        size=$(($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
+}