Whamcloud - gitweb
LU-5759 tests: use lfs getstripe -M instead of get_mds_num
[fs/lustre-release.git] / lustre / tests / test-framework.sh
index 8028991..13d02b9 100755 (executable)
@@ -370,7 +370,7 @@ module_loaded () {
 
 # Load a module on the system where this is running.
 #
-# Synopsis: load_module module_name [module arguments for insmod/modprobe]
+# usage: 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
@@ -416,22 +416,24 @@ load_module() {
 
     [ $# -gt 0 ] && echo "${module} options: '$*'"
 
-    # Note that insmod will ignore anything in modprobe.conf, which is why we're
-    # passing options on the command-line.
-    if [ "$BASE" == "lnet_selftest" ] && \
-            [ -f ${LUSTRE}/../lnet/selftest/${module}${EXT} ]; then
-        insmod ${LUSTRE}/../lnet/selftest/${module}${EXT}
-    elif [ -f ${LUSTRE}/${module}${EXT} ]; then
-        insmod ${LUSTRE}/${module}${EXT} "$@"
-    else
-        # must be testing a "make install" or "rpm" installation
-        # note failed to load ptlrpc_gss is considered not fatal
-        if [ "$BASE" == "ptlrpc_gss" ]; then
-            modprobe $BASE "$@" 2>/dev/null || echo "gss/krb5 is not supported"
-        else
-            modprobe $BASE "$@"
-        fi
-    fi
+       # 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}${EXT} ]]; then
+               [[ "$BASE" != "ptlrpc_gss" ]] || modprobe sunrpc
+               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"
+               else
+                       modprobe $BASE "$@"
+               fi
+       fi
 }
 
 llite_lloop_enabled() {
@@ -520,7 +522,6 @@ load_modules_local() {
                        load_module ../ldiskfs/ldiskfs
                        load_module osd-ldiskfs/osd_ldiskfs
                fi
-               load_module nodemap/nodemap
                load_module mgs/mgs
                load_module mdd/mdd
                load_module mdt/mdt
@@ -816,6 +817,25 @@ node_fstypes() {
        echo -n $fstypes
 }
 
+facet_index() {
+       local facet=$1
+       local num=$(facet_number $facet)
+       local index
+
+       if [[ $(facet_type $facet) = OST ]]; then
+               index=OSTINDEX${num}
+               if [[ -n "${!index}" ]]; then
+                       echo -n ${!index}
+                       return
+               fi
+
+               index=${OST_INDICES[num - 1]}
+       fi
+
+       [[ -n "$index" ]] || index=$((num - 1))
+       echo -n $index
+}
+
 devicelabel() {
        local facet=$1
        local dev=$2
@@ -2044,6 +2064,8 @@ sync_all_data() {
 
 wait_delete_completed_mds() {
        local MAX_WAIT=${1:-20}
+       # for ZFS, waiting more time for DMUs to be committed
+       local ZFS_WAIT=${2:-5}
        local mds2sync=""
        local stime=$(date +%s)
        local etime
@@ -2079,6 +2101,14 @@ wait_delete_completed_mds() {
                if [[ $changes -eq 0 ]]; then
                        etime=$(date +%s)
                        #echo "delete took $((etime - stime)) seconds"
+
+                       # the occupied disk space will be released
+                       # only after DMUs are committed
+                       if [[ $(facet_fstype $SINGLEMDS) == zfs ]]; then
+                               echo "sleep $ZFS_WAIT for ZFS OSD"
+                               sleep $ZFS_WAIT
+                       fi
+
                        return
                fi
                sleep 1
@@ -3258,7 +3288,7 @@ mkfs_opts() {
        local dev=$2
        local fsname=${3:-"$FSNAME"}
        local type=$(facet_type $facet)
-       local index=$(($(facet_number $facet) - 1))
+       local index=$(facet_index $facet)
        local fstype=$(facet_fstype $facet)
        local host=$(facet_host $facet)
        local opts
@@ -3344,6 +3374,24 @@ mkfs_opts() {
        echo -n "$opts"
 }
 
+check_ost_indices() {
+       local index_count=${#OST_INDICES[@]}
+       [[ $index_count -eq 0 || $OSTCOUNT -le $index_count ]] && return 0
+
+       # OST count is greater than the index count in $OST_INDEX_LIST.
+       # We need check whether there are duplicate indices.
+       local i
+       local j
+       local index
+       for i in $(seq $((index_count + 1)) $OSTCOUNT); do
+               index=$(facet_index ost$i)
+               for j in $(seq 0 $((index_count - 1))); do
+                       [[ $index -ne ${OST_INDICES[j]} ]] ||
+                       error "ost$i has the same index $index as ost$((j+1))"
+               done
+       done
+}
+
 formatall() {
        local quiet
 
@@ -3370,6 +3418,8 @@ formatall() {
                        ${quiet:+>/dev/null} || exit 10
        done
 
+       export OST_INDICES=($(hostlist_expand "$OST_INDEX_LIST"))
+       check_ost_indices
        for num in $(seq $OSTCOUNT); do
                echo "Format ost$num: $(ostdevname $num)"
                add ost$num $(mkfs_opts ost$num $(ostdevname ${num})) \
@@ -4106,6 +4156,60 @@ run_lfsck() {
        return $rc
 }
 
+dump_file_contents() {
+       local nodes=$1
+       local dir=$2
+       local logname=$3
+       local node
+
+       if [ -z "$nodes" -o -z "$dir" -o -z "$logname" ]; then
+               error_noexit false \
+                       "Invalid parameters for dump_file_contents()"
+               return 1
+       fi
+       for node in ${nodes}; do
+               do_node $node "for i in \\\$(find $dir -type f); do
+                               echo ====\\\${i}=======================;
+                               cat \\\${i};
+                               done" >> ${logname}.${node}.log
+       done
+}
+
+dump_command_output() {
+       local nodes=$1
+       local cmd=$2
+       local logname=$3
+       local node
+
+       if [ -z "$nodes" -o -z "$cmd" -o -z "$logname" ]; then
+               error_noexit false \
+                       "Invalid parameters for dump_command_output()"
+               return 1
+       fi
+
+       for node in ${nodes}; do
+               do_node $node "echo ====${cmd}=======================;
+                               $cmd" >> ${logname}.${node}.log
+       done
+}
+
+log_zfs_info() {
+       local logname=$1
+
+       # dump file contents from /proc/spl in case of zfs test
+       if [ "$(facet_fstype ost1)" = "zfs" ]; then
+               dump_file_contents "$(osts_nodes)" "/proc/spl" "${logname}"
+               dump_command_output \
+                       "$(osts_nodes)" "zpool events -v" "${logname}"
+       fi
+
+       if [ "$(facet_fstype $SINGLEMDS)" = "zfs" ]; then
+               dump_file_contents "$(mdts_nodes)" "/proc/spl" "${logname}"
+               dump_command_output \
+                       "$(mdts_nodes)" "zpool events -v" "${logname}"
+       fi
+}
+
 check_and_cleanup_lustre() {
     if [ "$LFSCK_ALWAYS" = "yes" -a "$TESTSUITE" != "lfsck" ]; then
         get_svr_devs
@@ -4382,12 +4486,13 @@ drop_ldlm_cancel() {
 }
 
 drop_bl_callback_once() {
-       rc=0
+       local rc=0
        do_facet client lctl set_param ldlm.namespaces.*.early_lock_cancel=0
 #define OBD_FAIL_LDLM_BL_CALLBACK_NET                  0x305
        do_facet client lctl set_param fail_loc=0x80000305
        do_facet client "$@" || rc=$?
        do_facet client lctl set_param fail_loc=0
+       do_facet client lctl set_param fail_val=0
        do_facet client lctl set_param ldlm.namespaces.*.early_lock_cancel=1
        return $rc
 }
@@ -4399,6 +4504,7 @@ drop_bl_callback() {
        do_facet client lctl set_param fail_loc=0x305
        do_facet client "$@" || rc=$?
        do_facet client lctl set_param fail_loc=0
+       do_facet client lctl set_param fail_val=0
        do_facet client lctl set_param ldlm.namespaces.*.early_lock_cancel=1
        return $rc
 }
@@ -4542,7 +4648,7 @@ stop_full_debug_logging() {
 }
 
 # prints bash call stack
-log_trace_dump() {
+print_stack_trace() {
        echo "  Trace dump:"
        for (( i=1; i < ${#BASH_LINENO[*]} ; i++ )) ; do
                local s=${BASH_SOURCE[$i]}
@@ -4568,7 +4674,7 @@ error_noexit() {
 
 
        log " ${TESTSUITE} ${TESTNAME}: @@@@@@ ${TYPE}: $@ "
-       log_trace_dump
+       print_stack_trace >&2
 
        mkdir -p $LOGDIR
        # We need to dump the logs on all nodes
@@ -4754,7 +4860,7 @@ run_test() {
 }
 
 log() {
-    echo "$*"
+       echo "$*" >&2
     module_loaded lnet || load_modules
 
     local MSG="$*"
@@ -4873,6 +4979,8 @@ run_one_logged() {
        local TEST_ERROR
        local name=${TESTSUITE}.test_${1}.test_log.$(hostname -s).log
        local test_log=$LOGDIR/$name
+       local zfs_log_name=${TESTSUITE}.test_${1}.zfs_log
+       local zfs_debug_log=$LOGDIR/$zfs_log_name
        rm -rf $LOGDIR/err
        rm -rf $LOGDIR/ignore
        rm -rf $LOGDIR/skip
@@ -4904,6 +5012,7 @@ run_one_logged() {
        fi
 
        if [ -f $LOGDIR/err ]; then
+               log_zfs_info "$zfs_debug_log"
                $FAIL_ON_ERROR && exit $RC
        fi
 
@@ -5546,53 +5655,15 @@ restore_lustre_params() {
 }
 
 check_catastrophe() {
-       local rnodes=${1:-$(comma_list $(remote_nodes_list))}
-       VAR=$(lctl get_param -n catastrophe 2>&1)
-       if [ $? = 0 ] ; then
-               if [ $VAR != 0 ]; then
-                       return 1
-               fi
-       fi
-
-       [ -z "$rnodes" ] && return 0
+       local nodes=${1:-$(comma_list $(nodes_list))}
 
-       local data
-       data=$(do_nodes "$rnodes" "rc=\\\$(lctl get_param -n catastrophe);
-               if [ \\\$rc -ne 0 ]; then echo \\\$(hostname): \\\$rc; fi
-               exit \\\$rc")
-       local rc=$?
-       if [ -n "$data" ]; then
-           echo $data
-           return $rc
-       fi
-       return 0
-}
-
-# CMD: determine mds index where directory inode presents
-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)
-
-    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)
-
-    local num=0
-    for ((i=0; i<${#newused[@]}; i++)); do
-         if [ ${oldused[$i]} -lt ${newused[$i]} ];  then
-             echo $(( i + 1 ))
-             rm -f $file
-             return 0
-         fi
-    done
-    error "mdt-s : inodes count OLD ${oldused[@]} NEW ${newused[@]}"
+       do_nodes $nodes "rc=0;
+val=\\\$($LCTL get_param -n catastrophe 2>&1);
+if [[ \\\$? -eq 0 && \\\$val -ne 0 ]]; then
+       echo \\\$(hostname -s): \\\$val;
+       rc=\\\$val;
+fi;
+exit \\\$rc"
 }
 
 mdsrate_cleanup () {
@@ -6051,22 +6122,23 @@ do_ls () {
     return $rc
 }
 
-# target_start_and_reset_recovery_timer()
-#        service_time = at_est2timeout(service_time);
-#        service_time += 2 * (CONNECTION_SWITCH_MAX + CONNECTION_SWITCH_INC +
-#                             INITIAL_CONNECT_TIMEOUT);
-# CONNECTION_SWITCH_MAX : min(25U, max(CONNECTION_SWITCH_MIN,obd_timeout))
-#define CONNECTION_SWITCH_INC 1
-#define INITIAL_CONNECT_TIMEOUT max(CONNECTION_SWITCH_MIN,obd_timeout/20)
-#define CONNECTION_SWITCH_MIN 5U
+# check_and_start_recovery_timer()
+#      service_time = at_est2timeout(service_time);
+#      service_time += 2 * INITIAL_CONNECT_TIMEOUT;
+#      service_time += 2 * (CONNECTION_SWITCH_MAX + CONNECTION_SWITCH_INC);
 
-max_recovery_time () {
-    local init_connect_timeout=$(( TIMEOUT / 20 ))
-    [[ $init_connect_timeout -ge 5 ]] || init_connect_timeout=5
+#define INITIAL_CONNECT_TIMEOUT max(CONNECTION_SWITCH_MIN, obd_timeout/20)
+#define CONNECTION_SWITCH_MAX min(50, max(CONNECTION_SWITCH_MIN, obd_timeout))
+#define CONNECTION_SWITCH_MIN 5
+#define CONNECTION_SWITCH_INC 5
+max_recovery_time() {
+       local init_connect_timeout=$((TIMEOUT / 20))
+       ((init_connect_timeout >= 5)) || init_connect_timeout=5
 
-    local service_time=$(( $(at_max_get client) + $(( 2 * $(( 25 + 1  + init_connect_timeout)) )) ))
+       local service_time=$(($(at_max_get client) * 9 / 4 + 5))
+       service_time=$((service_time + 2 * (init_connect_timeout + 50 + 5)))
 
-    echo $service_time 
+       echo -n $service_time
 }
 
 get_clients_mount_count () {
@@ -6890,6 +6962,10 @@ check_mount_and_prep()
 
        rm -rf $DIR/[df][0-9]* || error "Fail to cleanup the env!"
        mkdir $DIR/$tdir || error "Fail to mkdir $DIR/$tdir."
+       for idx in $(seq $MDSCOUNT); do
+               local name="MDT$(printf '%04x' $((idx - 1)))"
+               rm -rf $MOUNT/.lustre/lost+found/$name/*
+       done
 }
 
 # calcule how many ost-objects to be created.
@@ -6900,11 +6976,10 @@ precreated_ost_obj_count()
        local mdt_name="MDT$(printf '%04x' $mdt_idx)"
        local ost_name="OST$(printf '%04x' $ost_idx)"
        local proc_path="${FSNAME}-${ost_name}-osc-${mdt_name}"
-       local last_id=$(do_facet mds${mdt_idx} lctl get_param -n \
+       local last_id=$(do_facet mds$((mdt_idx + 1)) lctl get_param -n \
                        osp.$proc_path.prealloc_last_id)
-       local next_id=$(do_facet mds${mdt_idx} lctl get_param -n \
+       local next_id=$(do_facet mds$((mdt_idx + 1)) lctl get_param -n \
                        osp.$proc_path.prealloc_next_id)
-
        echo $((last_id - next_id + 1))
 }
 
@@ -7151,3 +7226,61 @@ pool_remove() {
        error_noexit "Pool $FSNAME.$pool is not destroyed"
        return 3
 }
+
+# Get and check the actual stripe count of one file.
+# Usage: check_stripe_count <file> <expected_stripe_count>
+check_stripe_count() {
+       local file=$1
+       local expected=$2
+       local actual
+
+       [[ -z "$file" || -z "$expected" ]] &&
+               error "check_stripe_count: invalid argument"
+
+       local cmd="$GETSTRIPE -c $file"
+       actual=$($cmd) || error "$cmd failed"
+       actual=${actual%% *}
+
+       if [[ $actual -ne $expected ]]; then
+               [[ $expected -eq -1 ]] ||
+                       error "$cmd wrong: found $actual, expected $expected"
+               [[ $actual -eq $OSTCOUNT ]] ||
+                       error "$cmd wrong: found $actual, expected $OSTCOUNT"
+       fi
+}
+
+# Get and check the actual list of OST indices on one file.
+# Usage: check_obdidx <file> <expected_comma_separated_list_of_ost_indices>
+check_obdidx() {
+       local file=$1
+       local expected=$2
+       local obdidx
+
+       [[ -z "$file" || -z "$expected" ]] &&
+               error "check_obdidx: invalid argument!"
+
+       obdidx=$(comma_list $($GETSTRIPE $file | grep -A $OSTCOUNT obdidx |
+                             grep -v obdidx | awk '{print $1}' | xargs))
+
+       [[ $obdidx = $expected ]] ||
+               error "list of OST indices on $file is $obdidx," \
+                     "should be $expected"
+}
+
+# Get and check the actual OST index of the first stripe on one file.
+# Usage: check_start_ost_idx <file> <expected_start_ost_idx>
+check_start_ost_idx() {
+       local file=$1
+       local expected=$2
+       local start_ost_idx
+
+       [[ -z "$file" || -z "$expected" ]] &&
+               error "check_start_ost_idx: invalid argument!"
+
+       start_ost_idx=$($GETSTRIPE $file | grep -A 1 obdidx | grep -v obdidx |
+                       awk '{print $1}')
+
+       [[ $start_ost_idx = $expected ]] ||
+               error "OST index of the first stripe on $file is" \
+                     "$start_ost_idx, should be $expected"
+}