Whamcloud - gitweb
LU-13352 tests: fix subtest duration for ONLY_REPEAT
[fs/lustre-release.git] / lustre / tests / test-framework.sh
index 66da8b5..d994b5a 100755 (executable)
@@ -429,6 +429,11 @@ init_test_env() {
        . ${CONFIG:=$LUSTRE/tests/cfg/$NAME.sh}
        get_lustre_env
 
+       # use localrecov to enable recovery for local clients, LU-12722
+       [[ $MDS1_VERSION -lt $(version_code 2.13.52) ]] ||
+               export MDS_MOUNT_OPTS=${MDS_MOUNT_OPTS:-"-o localrecov"}
+       [[ $OST1_VERSION -lt $(version_code 2.13.52) ]] ||
+               export OST_MOUNT_OPTS=${OST_MOUNT_OPTS:-"-o localrecov"}
 }
 
 check_cpt_number() {
@@ -451,7 +456,7 @@ version_code() {
        # split arguments like "1.8.6-wc3" into "1", "8", "6", "wc3"
        eval set -- $(tr "[:punct:]" " " <<< $*)
 
-       echo -n "$((($1 << 16) | ($2 << 8) | $3))"
+       echo -n $(((${1:-0} << 16) | (${2:-0} << 8) | ${3:-0}))
 }
 
 export LINUX_VERSION=$(uname -r | sed -e "s/\([0-9]*\.[0-9]*\.[0-9]*\).*/\1/")
@@ -488,14 +493,31 @@ export LINUX_VERSION_CODE=$(version_code ${LINUX_VERSION//\./ })
 # output: prints version string to stdout in (up to 4) dotted-decimal values
 lustre_build_version() {
        local facet=${1:-client}
-       local ver=$(do_facet $facet "$LCTL get_param -n version 2>/dev/null ||
-                               $LCTL lustre_build_version 2>/dev/null ||
-                               $LCTL --version 2>/dev/null | cut -d' ' -f2")
+       local facet_version=${facet}_VERSION
+
+       # if the global variable is already set, then use that
+       [ -n "${!facet_version}" ] && echo ${!facet_version} && return
+
+       # this is the currently-running version of the kernel modules
+       local ver=$(do_facet $facet "$LCTL get_param -n version 2>/dev/null")
+       # we mostly test 2.10+ systems, only try others if the above fails
+       if [ -z "$ver" ]; then
+               ver=$(do_facet $facet "$LCTL lustre_build_version 2>/dev/null")
+       fi
+       if [ -z "$ver" ]; then
+               ver=$(do_facet $facet "$LCTL --version 2>/dev/null" |
+                     cut -d' ' -f2)
+       fi
        local lver=$(egrep -i "lustre: |version: " <<<"$ver" | head -n 1)
        [ -n "$lver" ] && ver="$lver"
 
-       sed -e 's/[^:]*: //' -e 's/^v//' -e 's/[ -].*//' -e 's/_/./g' <<<$ver |
-               cut -d. -f1-4
+       lver=$(sed -e 's/[^:]*: //' -e 's/^v//' -e 's/[ -].*//' <<<$ver |
+              tr _ . | cut -d. -f1-4)
+
+       # save in global variable for the future
+       export $facet_version=$lver
+
+       echo $lver
 }
 
 # Report the Lustre numeric build version code for the supplied facet.
@@ -972,6 +994,22 @@ add_sk_mntflag() {
        echo -n $mt_opts
 }
 
+from_build_tree() {
+       local from_tree
+
+       case $LUSTRE in
+       /usr/lib/lustre/* | /usr/lib64/lustre/* | /usr/lib/lustre | \
+       /usr/lib64/lustre )
+               from_tree=false
+               ;;
+       *)
+               from_tree=true
+               ;;
+       esac
+
+       [ $from_tree = true ]
+}
+
 init_gss() {
        if $SHARED_KEY; then
                GSS=true
@@ -982,16 +1020,6 @@ init_gss() {
                return
        fi
 
-       case $LUSTRE in
-       /usr/lib/lustre/* | /usr/lib64/lustre/* | /usr/lib/lustre | \
-       /usr/lib64/lustre )
-               from_build_tree=false
-               ;;
-       *)
-               from_build_tree=true
-               ;;
-       esac
-
        if ! module_loaded ptlrpc_gss; then
                load_module ptlrpc/gss/ptlrpc_gss
                module_loaded ptlrpc_gss ||
@@ -1010,7 +1038,7 @@ init_gss() {
                SK_NO_KEY=false
                local lgssc_conf_file="/etc/request-key.d/lgssc.conf"
 
-               if $from_build_tree; then
+               if from_build_tree; then
                        mkdir -p $SK_OM_PATH
                        if grep -q request-key /proc/mounts > /dev/null; then
                                echo "SSK: Request key already mounted."
@@ -1029,7 +1057,7 @@ init_gss() {
                cat $lgssc_conf_file
 
                if ! local_mode; then
-                       if $from_build_tree; then
+                       if from_build_tree; then
                                do_nodes $(comma_list $(all_nodes)) "mkdir -p \
                                        $SK_OM_PATH"
                                do_nodes $(comma_list $(all_nodes)) "mount \
@@ -1109,11 +1137,19 @@ init_gss() {
                OST_MOUNT_OPTS=$(add_sk_mntflag $OST_MOUNT_OPTS)
                MOUNT_OPTS=$(add_sk_mntflag $MOUNT_OPTS)
                SEC=$SK_FLAVOR
+               if [ -z "$LGSS_KEYRING_DEBUG" ]; then
+                       LGSS_KEYRING_DEBUG=4
+               fi
        fi
 
-       if [ -n "$LGSS_KEYRING_DEBUG" ]; then
+       if [ -n "$LGSS_KEYRING_DEBUG" ] && \
+              ( local_mode || from_build_tree ); then
+               lctl set_param -n \
+                    sptlrpc.gss.lgss_keyring.debug_level=$LGSS_KEYRING_DEBUG
+       elif [ -n "$LGSS_KEYRING_DEBUG" ]; then
+               do_nodes $(comma_list $(all_nodes)) "modprobe ptlrpc_gss && \
                lctl set_param -n \
-                       sptlrpc.gss.lgss_keyring.debug_level=$LGSS_KEYRING_DEBUG
+                  sptlrpc.gss.lgss_keyring.debug_level=$LGSS_KEYRING_DEBUG"
        fi
 }
 
@@ -1126,16 +1162,6 @@ cleanup_gss() {
 
 cleanup_sk() {
        if $GSS_SK; then
-               case $LUSTRE in
-               /usr/lib/lustre/* | /usr/lib64/lustre/* | /usr/lib/lustre | \
-               /usr/lib64/lustre )
-                       from_build_tree=false
-                       ;;
-               *)
-                       from_build_tree=true
-                       ;;
-               esac
-
                if $SK_S2S; then
                        do_node $(mgs_node) "$LCTL nodemap_del $SK_S2SNM"
                        do_node $(mgs_node) "$LCTL nodemap_del $SK_S2SNMCLI"
@@ -1148,7 +1174,7 @@ cleanup_sk() {
                        $SK_PATH/$FSNAME*.key $SK_PATH/nodemap/$FSNAME*.key"
                do_nodes $(comma_list $(all_nodes)) "keyctl show | \
                  awk '/lustre/ { print \\\$1 }' | xargs -IX keyctl unlink X"
-               if $from_build_tree; then
+               if from_build_tree; then
                        # Remove the mount and clean up the files we added to
                        # SK_PATH
                        do_nodes $(comma_list $(all_nodes)) "while grep -q \
@@ -2331,6 +2357,22 @@ zconf_umount() {
        fi
 }
 
+# Mount the file system on the MDS
+mount_mds_client() {
+       local mds_HOST=${SINGLEMDS}_HOST
+       echo $mds_HOST
+       do_facet $SINGLEMDS "mkdir -p $MOUNT2"
+       zconf_mount $mds1_HOST $MOUNT2 $MOUNT_OPTS ||
+               error "unable to mount $MOUNT2 on MDS"
+}
+
+# Unmount the file system on the MDS
+umount_mds_client() {
+       local mds_HOST=${SINGLEMDS}_HOST
+       zconf_umount $mds1_HOST $MOUNT2
+       do_facet $SINGLEMDS "rm -rf $MOUNT2"
+}
+
 # nodes is comma list
 sanity_mount_check_nodes () {
     local nodes=$1
@@ -2735,16 +2777,15 @@ start_client_load() {
 }
 
 start_client_loads () {
-    local -a clients=(${1//,/ })
-    local numloads=${#CLIENT_LOADS[@]}
-    local testnum
+       local -a clients=(${1//,/ })
+       local numloads=${#CLIENT_LOADS[@]}
 
-    for ((nodenum=0; nodenum < ${#clients[@]}; nodenum++ )); do
-        testnum=$((nodenum % numloads))
-        start_client_load ${clients[nodenum]} ${CLIENT_LOADS[testnum]}
-    done
-    # bug 22169: wait the background threads to start
-    sleep 2
+       for ((nodenum=0; nodenum < ${#clients[@]}; nodenum++ )); do
+               local load=$((nodenum % numloads))
+               start_client_load ${clients[nodenum]} ${CLIENT_LOADS[load]}
+       done
+       # bug 22169: wait the background threads to start
+       sleep 2
 }
 
 # only for remote client
@@ -3198,13 +3239,13 @@ wait_delete_completed_mds() {
 }
 
 wait_for_host() {
-    local hostlist=$1
+       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
+       # 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_facet() {
@@ -5193,6 +5234,10 @@ init_param_vars () {
                        # $LFS quotaoff -ug $MOUNT > /dev/null 2>&1
                fi
        fi
+
+       (( MDS1_VERSION <= $(version_code 2.13.52) )) ||
+               do_nodes $(comma_list $(mdts_nodes)) \
+                       "$LCTL set_param lod.*.mdt_hash=crush"
        return 0
 }
 
@@ -5376,6 +5421,9 @@ check_and_setup_lustre() {
                fi
        fi
 
+       if [ -n "$fs_STRIPEPARAMS" ]; then
+               setstripe_getstripe $MOUNT $fs_STRIPEPARAMS
+       fi
        if $GSS_SK; then
                set_flavor_all null
        elif $GSS; then
@@ -5659,13 +5707,10 @@ check_network() {
 
        [ "$host" = "$HOSTNAME" ] && return 0
 
-       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!"
+       if ! wait_for_function --quiet "ping -c 1 -w 3 $host" $max $sleep; then
+               echo "$(date +'%H:%M:%S (%s)') waited for $host network ${max}s"
                exit 1
        fi
-
-       echo "$(date +'%H:%M:%S (%s)') network interface is UP"
 }
 
 no_dsh() {
@@ -6176,6 +6221,7 @@ skip_noexit() {
 
        [[ -n "$TESTSUITELOG" ]] &&
                echo "$TESTSUITE: SKIP: $TESTNAME $@" >> $TESTSUITELOG || true
+       unset TESTNAME
 }
 
 skip() {
@@ -6196,22 +6242,22 @@ build_test_filter() {
                fi
        done
 
-    [ "$EXCEPT$ALWAYS_EXCEPT" ] && \
-        log "excepting tests: `echo $EXCEPT $ALWAYS_EXCEPT`"
-    [ "$EXCEPT_SLOW" ] && \
-        log "skipping tests SLOW=no: `echo $EXCEPT_SLOW`"
-    for E in $EXCEPT; do
-        eval EXCEPT_${E}=true
-    done
-    for E in $ALWAYS_EXCEPT; do
-        eval EXCEPT_ALWAYS_${E}=true
-    done
-    for E in $EXCEPT_SLOW; do
-        eval EXCEPT_SLOW_${E}=true
-    done
-    for G in $GRANT_CHECK_LIST; do
-        eval GCHECK_ONLY_${G}=true
-        done
+       [ "$EXCEPT$ALWAYS_EXCEPT" ] &&
+               log "excepting tests: `echo $EXCEPT $ALWAYS_EXCEPT`"
+       [ "$EXCEPT_SLOW" ] &&
+               log "skipping tests SLOW=no: `echo $EXCEPT_SLOW`"
+       for E in $EXCEPT; do
+               eval EXCEPT_${E}=true
+       done
+       for E in $ALWAYS_EXCEPT; do
+               eval EXCEPT_ALWAYS_${E}=true
+       done
+       for E in $EXCEPT_SLOW; do
+               eval EXCEPT_SLOW_${E}=true
+       done
+       for G in $GRANT_CHECK_LIST; do
+               eval GCHECK_ONLY_${G}=true
+       done
 }
 
 basetest() {
@@ -6226,59 +6272,63 @@ basetest() {
 export LAST_SKIPPED=
 export ALWAYS_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
+# Main entry into test-framework. This is called with the number and
+# description of a test. The number is used to find the function to run
 # the test using "test_$name".
 #
 # This supports a variety of methods of specifying specific test to
-# run or not run.  These need to be documented...
+# run or not run:
+# - ONLY= env variable with space-separated list of test numbers to run
+# - EXCEPT= env variable with space-separated list of test numbers to exclude
 #
 run_test() {
        assert_DIR
-       export base=$(basetest $1)
-       TESTNAME=test_$1
+       local testnum=$1
+       local testmsg=$2
+       export base=$(basetest $testnum)
+       export TESTNAME=test_$testnum
        LAST_SKIPPED=
        ALWAYS_SKIPPED=
 
        # Check the EXCEPT, ALWAYS_EXCEPT and SLOW lists to see if we
        # need to skip the current test. If so, set the ALWAYS_SKIPPED flag.
-       local testname=EXCEPT_$1
-       local testname_base=EXCEPT_$base
-       if [ ${!testname}x != x ]; then
+       local isexcept=EXCEPT_$testnum
+       local isexcept_base=EXCEPT_$base
+       if [ ${!isexcept}x != x ]; then
                ALWAYS_SKIPPED="y"
-               skip_message="skipping excluded test $1"
-       elif [ ${!testname_base}x != x ]; then
+               skip_message="skipping excluded test $testnum"
+       elif [ ${!isexcept_base}x != x ]; then
                ALWAYS_SKIPPED="y"
-               skip_message="skipping excluded test $1 (base $base)"
+               skip_message="skipping excluded test $testnum (base $base)"
        fi
 
-       testname=EXCEPT_ALWAYS_$1
-       testname_base=EXCEPT_ALWAYS_$base
-       if [ ${!testname}x != x ]; then
+       isexcept=EXCEPT_ALWAYS_$testnum
+       isexcept_base=EXCEPT_ALWAYS_$base
+       if [ ${!isexcept}x != x ]; then
                ALWAYS_SKIPPED="y"
-               skip_message="skipping ALWAYS excluded test $1"
-       elif [ ${!testname_base}x != x ]; then
+               skip_message="skipping ALWAYS excluded test $testnum"
+       elif [ ${!isexcept_base}x != x ]; then
                ALWAYS_SKIPPED="y"
-               skip_message="skipping ALWAYS excluded test $1 (base $base)"
+               skip_message="skipping ALWAYS excluded test $testnum (base $base)"
        fi
 
-       testname=EXCEPT_SLOW_$1
-       testname_base=EXCEPT_SLOW_$base
-       if [ ${!testname}x != x ]; then
+       isexcept=EXCEPT_SLOW_$testnum
+       isexcept_base=EXCEPT_SLOW_$base
+       if [ ${!isexcept}x != x ]; then
                ALWAYS_SKIPPED="y"
-               skip_message="skipping SLOW test $1"
-       elif [ ${!testname_base}x != x ]; then
+               skip_message="skipping SLOW test $testnum"
+       elif [ ${!isexcept_base}x != x ]; then
                ALWAYS_SKIPPED="y"
-               skip_message="skipping SLOW test $1 (base $base)"
+               skip_message="skipping SLOW test $testnum (base $base)"
        fi
 
        # If there are tests on the ONLY list, check if the current test
        # is on that list and, if so, check if the test is to be skipped
        # and if we are supposed to honor the skip lists.
        if [ -n "$ONLY" ]; then
-               testname=ONLY_$1
-               testname_base=ONLY_$base
-               if [[ ${!testname}x != x || ${!testname_base}x != x ]]; then
+               local isonly=ONLY_$testnum
+               local isonly_base=ONLY_$base
+               if [[ ${!isonly}x != x || ${!isonly_base}x != x ]]; then
 
                        if [[ -n "$ALWAYS_SKIPPED" && -n "$HONOR_EXCEPT" ]]; then
                                LAST_SKIPPED="y"
@@ -6288,7 +6338,7 @@ run_test() {
                                [ -n "$LAST_SKIPPED" ] &&
                                        echo "" && LAST_SKIPPED=
                                ALWAYS_SKIPPED=
-                               run_one_logged $1 "$2"
+                               run_one_logged $testnum "$testmsg"
                                return $?
                        fi
 
@@ -6303,10 +6353,9 @@ run_test() {
                skip_noexit "$skip_message"
                return 0
        else
-               run_one_logged $1 "$2"
+               run_one_logged $testnum "$testmsg"
                return $?
        fi
-
 }
 
 log() {
@@ -6406,10 +6455,7 @@ group descriptors corrupted"
 #
 run_one() {
        local testnum=$1
-       local message=$2
-       export tfile=f${testnum}.${TESTSUITE}
-       export tdir=d${testnum}.${TESTSUITE}
-       export TESTNAME=test_$testnum
+       local testmsg="$2"
        local SAVE_UMASK=`umask`
        umask 0022
 
@@ -6417,7 +6463,7 @@ run_one() {
                $SETUP
        fi
 
-       banner "test $testnum: $message"
+       banner "test $testnum: $testmsg"
        test_${testnum} || error "test_$testnum failed with $?"
        cd $SAVE_PWD
        reset_fail_loc
@@ -6428,9 +6474,6 @@ run_one() {
                ps auxww | grep -v grep | grep -q "multiop " &&
                                        error "multiop still running"
        fi
-       unset TESTNAME
-       unset tdir
-       unset tfile
        umask $SAVE_UMASK
        $CLEANUP
        return 0
@@ -6443,49 +6486,74 @@ run_one() {
 #  - test result is saved to data file
 #
 run_one_logged() {
-       local BEFORE=$(date +%s)
-       local TEST_ERROR
-       local name=${TESTSUITE}.test_${1}.test_log.$(hostname -s).log
+       local before=$SECONDS
+       local testnum=$1
+       local testmsg=$2
+       export tfile=f${testnum}.${TESTSUITE}
+       export tdir=d${testnum}.${TESTSUITE}
+       local name=$TESTSUITE.$TESTNAME.test_log.$(hostname -s).log
        local test_log=$LOGDIR/$name
-       local zfs_log_name=${TESTSUITE}.test_${1}.zfs_log
+       local zfs_log_name=$TESTSUITE.$TESTNAME.zfs_log
        local zfs_debug_log=$LOGDIR/$zfs_log_name
-       rm -rf $LOGDIR/err
-       rm -rf $LOGDIR/ignore
-       rm -rf $LOGDIR/skip
        local SAVE_UMASK=$(umask)
+       local rc=0
        umask 0022
 
+       rm -f $LOGDIR/err $LOGDIR/ignore $LOGDIR/skip
        echo
-       log_sub_test_begin test_${1}
-       (run_one $1 "$2") 2>&1 | tee -i $test_log
-       local RC=${PIPESTATUS[0]}
-
-       [ $RC -ne 0 ] && [ ! -f $LOGDIR/err ] &&
-               echo "test_$1 returned $RC" | tee $LOGDIR/err
-
-       duration=$(($(date +%s) - $BEFORE))
-       pass "$1" "(${duration}s)"
+       # if ${ONLY_$testnum} set, repeat $ONLY_REPEAT times, otherwise once
+       local isonly=ONLY_$testnum
+       local repeat=${!isonly:+$ONLY_REPEAT}
+
+       for testiter in $(seq ${repeat:-1}); do
+               local before_sub=$SECONDS
+               log_sub_test_begin $TESTNAME
+
+               # remove temp files between repetitions to avoid test failures
+               [ -n "$append" -a -n "$DIR" -a -n "$tdir" -a -n "$tfile" ] &&
+                       rm -rf $DIR/$tdir* $DIR/$tfile*
+               # loop around subshell so stack_trap EXIT triggers each time
+               (run_one $testnum "$testmsg") 2>&1 | tee -i $append $test_log
+               rc=${PIPESTATUS[0]}
+               local append=-a
+               local duration_sub=$((SECONDS - before_sub))
+               local test_error
+
+               [[ $rc != 0 && ! -f $LOGDIR/err ]] &&
+                       echo "$TESTNAME returned $rc" | tee $LOGDIR/err
+
+               if [[ -f $LOGDIR/err ]]; then
+                       test_error=$(cat $LOGDIR/err)
+                       TEST_STATUS="FAIL"
+               elif [[ -f $LOGDIR/ignore ]]; then
+                       test_error=$(cat $LOGDIR/ignore)
+               elif [[ -f $LOGDIR/skip ]]; then
+                       test_error=$(cat $LOGDIR/skip)
+                       TEST_STATUS="SKIP"
+               else
+                       TEST_STATUS="PASS"
+               fi
 
-       if [[ -f $LOGDIR/err ]]; then
-               TEST_ERROR=$(cat $LOGDIR/err)
-       elif [[ -f $LOGDIR/ignore ]]; then
-               TEST_ERROR=$(cat $LOGDIR/ignore)
-       elif [[ -f $LOGDIR/skip ]]; then
-               TEST_ERROR=$(cat $LOGDIR/skip)
-       fi
-       log_sub_test_end $TEST_STATUS $duration "$RC" "$TEST_ERROR"
+               pass "$testnum" "(${duration_sub}s)"
+               log_sub_test_end $TEST_STATUS $duration_sub "$rc" "$test_error"
+               [[ $rc != 0 ]] && break
+       done
 
-       if [[ "$TEST_STATUS" != "SKIP" ]] && [[ -f $TF_SKIP ]]; then
+       if [[ "$TEST_STATUS" != "SKIP" && -f $TF_SKIP ]]; then
                rm -f $TF_SKIP
        fi
 
        if [ -f $LOGDIR/err ]; then
                log_zfs_info "$zfs_debug_log"
-               $FAIL_ON_ERROR && exit $RC
+               $FAIL_ON_ERROR && exit $rc
        fi
 
        umask $SAVE_UMASK
 
+       unset TESTNAME
+       unset tdir
+       unset tfile
+
        return 0
 }
 
@@ -6502,17 +6570,39 @@ canonical_path() {
        (cd $(dirname $1); echo $PWD/$(basename $1))
 }
 
+grant_from_clients() {
+       local nodes="$1"
+
+       # get client grant
+       do_nodes $nodes "$LCTL get_param -n osc.${FSNAME}-*.cur_*grant_bytes" |
+               calc_sum
+}
+
+grant_from_servers() {
+       local nodes="$1"
+
+       # get server grant
+       # which is tot_granted less grant_precreate
+       do_nodes $nodes "$LCTL get_param obdfilter.${FSNAME}-OST*.tot_granted" \
+               " obdfilter.${FSNAME}-OST*.tot_pending" \
+               " obdfilter.${FSNAME}-OST*.grant_precreate" |
+               tr '=' ' ' | awk '/tot_granted/{ total += $2 };
+                                 /tot_pending/{ total -= $2 };
+                                 /grant_precreate/{ total -= $2 };
+                                 END { printf("%0.0f", total) }'
+}
 
 check_grant() {
        export base=$(basetest $1)
        [ "$CHECK_GRANT" == "no" ] && return 0
 
-       testnamebase=GCHECK_ONLY_${base}
-       testname=GCHECK_ONLY_$1
-       [ ${!testnamebase}x == x -a ${!testname}x == x ] && return 0
+       local isonly_base=GCHECK_ONLY_${base}
+       local isonly=GCHECK_ONLY_$1
+       [ ${!isonly_base}x == x -a ${!isonly}x == x ] && return 0
 
        echo -n "checking grant......"
 
+       local osts=$(comma_list $(osts_nodes))
        local clients=$CLIENTS
        [ -z "$clients" ] && clients=$(hostname)
 
@@ -6521,29 +6611,28 @@ check_grant() {
        clients_up # initiate all idling connections
 
        # get client grant
-       client_grant=$(do_nodes $clients \
-               "$LCTL get_param -n osc.${FSNAME}-*.cur_*grant_bytes" |
-               awk '{ total += $1 } END { printf("%0.0f", total) }')
+       cli_grant=$(grant_from_clients $clients)
 
        # get server grant
        # which is tot_granted less grant_precreate
-       server_grant=$(do_nodes $(comma_list $(osts_nodes)) \
-               "$LCTL get_param "\
-               "obdfilter.${FSNAME}-OST*.{tot_granted,tot_pending,grant_precreate}" |
-               sed 's/=/ /'| awk '/tot_granted/{ total += $2 };
-                               /tot_pending/{ total -= $2 };
-                               /grant_precreate/{ total -= $2 };
-                               END { printf("%0.0f", total) }')
+       srv_grant=$(grant_from_servers $osts)
 
+       count=0
        # check whether client grant == server grant
-       if [[ $client_grant -ne $server_grant ]]; then
+       while [[ $cli_grant != $srv_grant && count++ -lt 30 ]]; do
+               echo "wait for client:$cli_grant == server:$srv_grant"
+               sleep 1
+               cli_grant=$(grant_from_clients $clients)
+               srv_grant=$(grant_from_servers $osts)
+       done
+       if [[ $cli_grant -ne $srv_grant ]]; then
                do_nodes $(comma_list $(osts_nodes)) \
                        "$LCTL get_param obdfilter.${FSNAME}-OST*.tot*" \
-                       "obdfilter.${FSNAME}-OST*.grant_*"
+                       "obdfilter.${FSNAME}-OST*.grant_*"
                do_nodes $clients "$LCTL get_param osc.${FSNAME}-*.cur_*_bytes"
-               error "failed: client:${client_grant} server: ${server_grant}."
+               error "failed grant check: client:$cli_grant server:$srv_grant"
        else
-               echo "pass: client:${client_grant} server: ${server_grant}"
+               echo "pass grant check: client:$cli_grant server:$srv_grant"
        fi
 }
 
@@ -6570,6 +6659,16 @@ ostname_from_index() {
     echo ${uuid/_UUID/}
 }
 
+mdtname_from_index() {
+       local uuid=$(mdtuuid_from_index $1)
+       echo ${uuid/_UUID/}
+}
+
+mdssize_from_index () {
+       local mdt=$(mdtname_from_index $2)
+       $LFS df $1 | grep $mdt | awk '{ print $2 }'
+}
+
 index_from_ostuuid()
 {
     $LFS osts $2 | sed -ne "/${1}/s/\(.*\): .* .*$/\1/p"
@@ -7135,13 +7234,9 @@ check_node_health() {
        for node in ${nodes//,/ }; do
                check_network "$node" 5
                if [ $? -eq 0 ]; then
-                       do_node $node "rc=0;
-                       val=\\\$($LCTL get_param -n catastrophe 2>&1);
-                       if [[ \\\$? -eq 0 && \\\$val -ne 0 ]]; then
-                               echo \\\$(hostname -s): \\\$val;
-                               rc=\\\$val;
-                       fi;
-                       exit \\\$rc" || error "$node:LBUG/LASSERT detected"
+                       do_node $node "$LCTL get_param catastrophe 2>&1" |
+                               grep -q "catastrophe=1" &&
+                               error "$node:LBUG/LASSERT detected" || true
                fi
        done
 }
@@ -7692,39 +7787,62 @@ destroy_test_pools () {
 }
 
 gather_logs () {
-    local list=$1
+       local list=$1
 
-    local ts=$(date +%s)
-    local docp=true
+       local ts=$(date +%s)
+       local docp=true
 
-    if [[ ! -f "$YAML_LOG" ]]; then
-        # init_logging is not performed before gather_logs,
-        # so the $LOGDIR needs to be checked here
-        check_shared_dir $LOGDIR && touch $LOGDIR/shared
-    fi
+       if [[ ! -f "$YAML_LOG" ]]; then
+               # init_logging is not performed before gather_logs,
+               # so the $LOGDIR needs to be checked here
+               check_shared_dir $LOGDIR && touch $LOGDIR/shared
+       fi
 
-    [ -f $LOGDIR/shared ] && docp=false
+       [ -f $LOGDIR/shared ] && docp=false
 
-    # dump lustre logs, dmesg
+       # dump lustre logs, dmesg, and journal if GSS_SK=true
 
-    prefix="$TESTLOG_PREFIX.$TESTNAME"
-    suffix="$ts.log"
-    echo "Dumping lctl log to ${prefix}.*.${suffix}"
+       prefix="$TESTLOG_PREFIX.$TESTNAME"
+       suffix="$ts.log"
+       echo "Dumping lctl log to ${prefix}.*.${suffix}"
 
-    if [ -n "$CLIENTONLY" -o "$PDSH" == "no_dsh" ]; then
-        echo "Dumping logs only on local client."
-        $LCTL dk > ${prefix}.debug_log.$(hostname -s).${suffix}
-        dmesg > ${prefix}.dmesg.$(hostname -s).${suffix}
-        return
-    fi
+       if [ -n "$CLIENTONLY" -o "$PDSH" == "no_dsh" ]; then
+               echo "Dumping logs only on local client."
+               $LCTL dk > ${prefix}.debug_log.$(hostname -s).${suffix}
+               dmesg > ${prefix}.dmesg.$(hostname -s).${suffix}
+               [ "$SHARED_KEY" = true ] && find $SK_PATH -name '*.key' -exec \
+                       lgss_sk -r {} \; &> \
+                       ${prefix}.ssk_keys.$(hostname -s).${suffix}
+               [ "$SHARED_KEY" = true ] && lctl get_param 'nodemap.*.*' > \
+                       ${prefix}.nodemaps.$(hostname -s).${suffix}
+               [ "$GSS_SK" = true ] && keyctl show > \
+                       ${prefix}.keyring.$(hostname -s).${suffix}
+               [ "$GSS_SK" = true ] && journalctl -a > \
+                       ${prefix}.journal.$(hostname -s).${suffix}
+               return
+       fi
 
-    do_nodesv $list \
-        "$LCTL dk > ${prefix}.debug_log.\\\$(hostname -s).${suffix};
-         dmesg > ${prefix}.dmesg.\\\$(hostname -s).${suffix}"
+       do_nodesv $list \
+               "$LCTL dk > ${prefix}.debug_log.\\\$(hostname -s).${suffix};
+               dmesg > ${prefix}.dmesg.\\\$(hostname -s).${suffix}"
+       if [ "$SHARED_KEY" = true ]; then
+               do_nodesv $list "find $SK_PATH -name '*.key' -exec \
+                       lgss_sk -r {} \; &> \
+                       ${prefix}.ssk_keys.\\\$(hostname -s).${suffix}"
+               do_facet mds1 "lctl get_param 'nodemap.*.*' > \
+                       ${prefix}.nodemaps.\\\$(hostname -s).${suffix}"
+       fi
+       if [ "$GSS_SK" = true ]; then
+               do_nodesv $list "keyctl show > \
+                       ${prefix}.keyring.\\\$(hostname -s).${suffix}"
+               do_nodesv $list "journalctl -a > \
+                       ${prefix}.journal.\\\$(hostname -s).${suffix}"
+       fi
 
-    if [ ! -f $LOGDIR/shared ]; then
-        do_nodes $list rsync -az "${prefix}.*.${suffix}" $HOSTNAME:$LOGDIR
-    fi
+       if [ ! -f $LOGDIR/shared ]; then
+               do_nodes $list rsync -az "${prefix}.*.${suffix}" \
+                        $HOSTNAME:$LOGDIR
+       fi
 }
 
 do_ls () {
@@ -8237,13 +8355,10 @@ init_logging() {
 
        umask $save_umask
 
-       # If modules are not yet loaded then older "lctl lustre_build_version"
-       # will fail.  Use lctl build version instead.
-       log "Client: $($LCTL lustre_build_version)"
-       log "MDS: $(do_facet $SINGLEMDS $LCTL lustre_build_version 2>/dev/null||
-                   do_facet $SINGLEMDS $LCTL --version)"
-       log "OSS: $(do_facet ost1 $LCTL lustre_build_version 2> /dev/null ||
-                   do_facet ost1 $LCTL --version)"
+       # log actual client and server versions if needed for debugging
+       log "Client: $(lustre_build_version client)"
+       log "MDS: $(lustre_build_version mds1)"
+       log "OSS: $(lustre_build_version ost1)"
 }
 
 log_test() {
@@ -8537,7 +8652,7 @@ mds_backup_restore() {
        local rcmd="do_facet $facet"
        local metaea=${TMP}/backup_restore.ea
        local metadata=${TMP}/backup_restore.tgz
-       local opts=${MDS_MOUNT_OPTS}
+       local opts=${MDS_MOUNT_FS_OPTS}
        local svc=${facet}_svc
 
        if ! ${rcmd} test -b ${devname}; then
@@ -8594,7 +8709,7 @@ mds_remove_ois() {
        local devname=$(mdsdevname $(facet_number $facet))
        local mntpt=$(facet_mntpt brpt)
        local rcmd="do_facet $facet"
-       local opts=${MDS_MOUNT_OPTS}
+       local opts=${MDS_MOUNT_FS_OPTS}
 
        if ! ${rcmd} test -b ${devname}; then
                opts=$(csa_add "$opts" -o loop)
@@ -8636,16 +8751,19 @@ generate_logname() {
 test_mkdir() {
        local path
        local p_option
+       local hash_type
+       local hash_name=("all_char" "fnv_1a_64" "crush")
        local dirstripe_count=${DIRSTRIPE_COUNT:-"2"}
        local dirstripe_index=${DIRSTRIPE_INDEX:-$((base % $MDSCOUNT))}
        local OPTIND=1
 
-       while getopts "c:i:p" opt; do
+       while getopts "c:H:i:p" opt; do
                case $opt in
                        c) dirstripe_count=$OPTARG;;
+                       H) hash_type=$OPTARG;;
                        i) dirstripe_index=$OPTARG;;
                        p) p_option="-p";;
-                       \?) error "only support -i -c -p";;
+                       \?) error "only support -c -H -i -p";;
                esac
        done
 
@@ -8674,6 +8792,10 @@ test_mkdir() {
                        mdt_index=$dirstripe_index
                fi
 
+               # randomly choose hash type
+               [ -z "$hash_type" ] &&
+                       hash_type=${hash_name[$((RANDOM % ${#hash_name[@]}))]}
+
                if (($MDS1_VERSION >= $(version_code 2.8.0))); then
                        if [ $dirstripe_count -eq -1 ]; then
                                dirstripe_count=$((RANDOM % MDSCOUNT + 1))
@@ -8682,9 +8804,9 @@ test_mkdir() {
                        dirstripe_count=1
                fi
 
-               echo "striped dir -i$mdt_index -c$dirstripe_count $path"
-               $LFS mkdir -i$mdt_index -c$dirstripe_count $path ||
-                       error "mkdir -i $mdt_index -c$dirstripe_count $path failed"
+               echo "striped dir -i$mdt_index -c$dirstripe_count -H $hash_type $path"
+               $LFS mkdir -i$mdt_index -c$dirstripe_count -H $hash_type $path ||
+                       error "mkdir -i $mdt_index -c$dirstripe_count -H $hash_type $path failed"
        fi
 }