# Return a numeric version code based on a version string. The version
# code is useful for comparison two version strings to see which is newer.
version_code() {
- # split arguments like "1.8.6-wc3" into "1", "8", "6", "wc3"
- eval set -- $(tr "[:punct:]" " " <<< $*)
+ # split arguments like "1.8.6-wc3" into "1", "8", "6", "3"
+ eval set -- $(tr "[:punct:][a-z]" " " <<< $*)
- echo -n $(((${1:-0} << 16) | (${2:-0} << 8) | ${3:-0}))
+ echo -n $(((${1:-0}<<24) | (${2:-0}<<16) | (${3:-0}<<8) | (${4:-0})))
}
export LINUX_VERSION=$(uname -r | sed -e "s/\([0-9]*\.[0-9]*\.[0-9]*\).*/\1/")
}
check_mem_leak () {
- LEAK_LUSTRE=$(dmesg | tail -n 30 | grep "obd_memory.*leaked" || true)
- LEAK_PORTALS=$(dmesg | tail -n 20 | grep "Portals memory leaked" || true)
- if [ "$LEAK_LUSTRE" -o "$LEAK_PORTALS" ]; then
- echo "$LEAK_LUSTRE" 1>&2
- echo "$LEAK_PORTALS" 1>&2
- mv $TMP/debug $TMP/debug-leak.`date +%s` || true
- echo "Memory leaks detected"
- [ -n "$IGNORE_LEAK" ] && { echo "ignoring leaks" && return 0; } || true
- return 1
- fi
+ LEAK_LUSTRE=$(dmesg | tail -n 30 | grep "obd_memory.*leaked" || true)
+ LEAK_PORTALS=$(dmesg | tail -n 20 | egrep -i "libcfs.*memory leaked" || true)
+ if [ "$LEAK_LUSTRE" -o "$LEAK_PORTALS" ]; then
+ echo "$LEAK_LUSTRE" 1>&2
+ echo "$LEAK_PORTALS" 1>&2
+ mv $TMP/debug $TMP/debug-leak.`date +%s` || true
+ echo "Memory leaks detected"
+ [ -n "$IGNORE_LEAK" ] && { echo "ignoring leaks" && return 0; } || true
+ return 1
+ fi
}
unload_modules() {
start_gss_daemons || error_exit "start gss daemon failed! rc=$?"
fi
+ if $GSS_SK && ! $SK_NO_KEY; then
+ echo "Loading basic SSK keys on all servers"
+ do_nodes $(comma_list $(all_server_nodes)) \
+ "lgss_sk -t server -l $SK_PATH/$FSNAME.key || true"
+ do_nodes $(comma_list $(all_server_nodes)) \
+ "keyctl show | grep lustre | cut -c1-11 |
+ sed -e 's/ //g;' |
+ xargs -IX keyctl setperm X 0x3f3f3f3f"
+ fi
+
if $GSS_SK && $SK_NO_KEY; then
local numclients=${1:-$CLIENTCOUNT}
local clients=${CLIENTS:-$HOSTNAME}
set_default_debug_nodes $node "$debug" "$subsys" $debug_size
}
+set_params_nodes () {
+ [[ $# -ge 2 ]] || return 0
+
+ local nodes=$1
+ shift
+ do_nodes $nodes $LCTL set_param $@
+}
+
+set_params_clients () {
+ local clients=${1:-$CLIENTS}
+ local params=${2:-$CLIENT_LCTL_SETPARAM_PARAM}
+
+ [[ -n $params ]] || return 0
+ set_params_nodes $clients $params
+}
+
set_hostid () {
local hostid=${1:-$(hostid)}
fi
set_default_debug_nodes $client
+ set_params_clients $client
return 0
}
do_nodes $clients "mount | grep $mnt' '"
set_default_debug_nodes $clients
+ set_params_clients $clients
return 0
}
reboot_facet() {
local facet=$1
+ local node=$(facet_active_host $facet)
+
if [ "$FAILURE_MODE" = HARD ]; then
- reboot_node $(facet_active_host $facet)
+ boot_node $node
else
sleep 10
fi
}
boot_node() {
- local node=$1
- if [ "$FAILURE_MODE" = HARD ]; then
- reboot_node $node
- wait_for_host $node
- fi
+ local node=$1
+
+ if [ "$FAILURE_MODE" = HARD ]; then
+ reboot_node $node
+ wait_for_host $node
+ if $LOAD_MODULES_REMOTE; then
+ echo "loading modules on $node: $facet"
+ do_rpc_nodes $node load_modules_local
+ fi
+ fi
}
facets_hosts () {
}
fail_nodf() {
- local facet=$1
- facet_failover $facet
+ local facet=$1
+
+ facet_failover $facet
}
fail_abort() {
local facet=$1
+ local abort_type=${2:-"abort_recovery"}
+
stop $facet
change_active $facet
wait_for_facet $facet
- mount_facet $facet -o abort_recovery
+ mount_facet $facet -o $abort_type
clients_up || echo "first stat failed: $?"
clients_up || error "post-failover stat: $?"
}
stop ${facet} -f
rm -f $TMP/${facet}active
[[ $facet = mds1 ]] && combined_mgs_mds && rm -f $TMP/mgsactive
+
+ # make sure in-tree ldiskfs is loaded before mkfs
+ if local_mode && [[ $(node_fstypes $HOSTNAME) == *ldiskfs* ]]; then
+ load_module ../ldiskfs/ldiskfs
+ fi
+
do_facet ${facet} $MKFS $* || return ${PIPESTATUS[0]}
if [[ $(facet_fstype $facet) == zfs ]]; then
export I_MOUNTED2=yes
fi
- if $do_check; then
- # 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
+ if $do_check; then
+ # 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
- set_default_debug_nodes $(comma_list $(nodes_list))
- fi
+ set_default_debug_nodes $(comma_list $(nodes_list))
+ set_params_clients
+ fi
if [ -z "$CLIENTONLY" -a $(lower $OSD_TRACK_DECLARES_LBUG) == 'yes' ]; then
local facets=""
local log=$TMP/e2fsck.log
local rc=0
+ # turn on pfsck if it is supported
+ do_node $node $E2FSCK -h 2>&1 | grep -qw -- -m && cmd+=" -m8"
echo $cmd
do_node $node $cmd 2>&1 | tee $log
rc=${PIPESTATUS[0]}
default_lru_size()
{
- NR_CPU=$(grep -c "processor" /proc/cpuinfo)
- DEFAULT_LRU_SIZE=$((100 * NR_CPU))
- echo "$DEFAULT_LRU_SIZE"
+ local nr_cpu=$(grep -c "processor" /proc/cpuinfo)
+
+ echo $((100 * nr_cpu))
}
lru_resize_enable()
lru_resize_disable()
{
- lctl set_param ldlm.namespaces.*$1*.lru_size $(default_lru_size)
+ local dev=${1}
+ local lru_size=${2:-$(default_lru_size)}
+
+ $LCTL set_param ldlm.namespaces.*$dev*.lru_size=$lru_size
}
flock_is_enabled()
exit_status () {
local status=0
- local log=$TESTSUITELOG
+ local logs="$TESTSUITELOG $1"
+
+ for log in $logs; do
+ if [ -f "$log" ]; then
+ grep -qw FAIL $log && status=1
+ fi
+ done
- [ -f "$log" ] && grep -qw FAIL $log && status=1
exit $status
}
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=$TESTSUITE.$TESTNAME.zfs_log
- local zfs_debug_log=$LOGDIR/$zfs_log_name
+ local test_log=$TESTLOG_PREFIX.$TESTNAME.test_log.$(hostname -s).log
+ local zfs_debug_log=$TESTLOG_PREFIX.$TESTNAME.zfs_log
local SAVE_UMASK=$(umask)
local rc=0
umask 0022
# remove temp files between repetitions to avoid test failures
[ -n "$append" -a -n "$DIR" -a -n "$tdir" -a -n "$tfile" ] &&
- rm -rf $DIR/$tdir* $DIR/$tfile*
+ rm -rvf $DIR/$tdir* $DIR/$tfile*
# loop around subshell so stack_trap EXIT triggers each time
(run_one $testnum "$testmsg") 2>&1 | tee -i $append $test_log
rc=${PIPESTATUS[0]}
}
check_runas_id_ret() {
- local myRC=0
- local myRUNAS_UID=$1
- local myRUNAS_GID=$2
- shift 2
- local myRUNAS=$@
- if [ -z "$myRUNAS" ]; then
- error_exit "myRUNAS command must be specified for check_runas_id"
- fi
- if $GSS_KRB5; then
- $myRUNAS krb5_login.sh || \
- error "Failed to refresh Kerberos V5 TGT for UID $myRUNAS_ID."
- fi
- mkdir $DIR/d0_runas_test
- chmod 0755 $DIR
- chown $myRUNAS_UID:$myRUNAS_GID $DIR/d0_runas_test
- $myRUNAS touch $DIR/d0_runas_test/f$$ || myRC=$?
- rm -rf $DIR/d0_runas_test
- return $myRC
+ local myRC=0
+ local myRUNAS_UID=$1
+ local myRUNAS_GID=$2
+ shift 2
+ local myRUNAS=$@
+
+ if [ -z "$myRUNAS" ]; then
+ error_exit "check_runas_id_ret requires myRUNAS argument"
+ fi
+
+ $myRUNAS true ||
+ error "Unable to execute $myRUNAS"
+
+ id $myRUNAS_UID > /dev/null ||
+ error "Invalid RUNAS_ID $myRUNAS_UID. Please set RUNAS_ID to " \
+ "some UID which exists on MDS and client or add user " \
+ "$myRUNAS_UID:$myRUNAS_GID on these nodes."
+
+ if $GSS_KRB5; then
+ $myRUNAS krb5_login.sh ||
+ error "Failed to refresh krb5 TGT for UID $myRUNAS_ID."
+ fi
+ mkdir $DIR/d0_runas_test
+ chmod 0755 $DIR
+ chown $myRUNAS_UID:$myRUNAS_GID $DIR/d0_runas_test
+ $myRUNAS -u $myRUNAS_UID -g $myRUNAS_GID touch $DIR/d0_runas_test/f$$ ||
+ myRC=$?
+ rm -rf $DIR/d0_runas_test
+ return $myRC
}
check_runas_id() {
- local myRUNAS_UID=$1
- local myRUNAS_GID=$2
- shift 2
- local myRUNAS=$@
- check_runas_id_ret $myRUNAS_UID $myRUNAS_GID $myRUNAS || \
- error "unable to write to $DIR/d0_runas_test as UID $myRUNAS_UID.
- Please set RUNAS_ID to some UID which exists on MDS and client or
- add user $myRUNAS_UID:$myRUNAS_GID on these nodes."
+ local myRUNAS_UID=$1
+ local myRUNAS_GID=$2
+ shift 2
+ local myRUNAS=$@
+
+ check_runas_id_ret $myRUNAS_UID $myRUNAS_GID $myRUNAS || \
+ error "unable to write to $DIR/d0_runas_test as " \
+ "UID $myRUNAS_UID."
}
# obtain the UID/GID for MPI_USER
fi
if [ ! -f $LOGDIR/shared ]; then
- do_nodes $list rsync -az "${prefix}.*.${suffix}" \
- $HOSTNAME:$LOGDIR
+ local remote_nodes=$(exclude_items_from_list $list $HOSTNAME)
+
+ for node in ${remote_nodes//,/ }; do
+ rsync -az -e ssh $node:${prefix}.'*'.${suffix} $LOGDIR &
+ done
fi
}
fi
local t=$(for i in $list; do printf "$FSNAME-OST%04x_UUID " $i; done)
+ local tg=$(for i in $list;
+ do printf -- "-e $FSNAME-OST%04x_UUID " $i; done)
+ local firstx=$(printf "%04x" $first)
+ local lastx=$(printf "%04x" $last)
+
do_facet mgs $LCTL pool_add \
- $FSNAME.$pool $FSNAME-OST[$first-$last/$step]
+ $FSNAME.$pool $FSNAME-OST[$firstx-$lastx/$step]
+ # ignore EEXIST(17)
+ if (( $? != 0 && $? != 17 )); then
+ error_noexit "pool_add $FSNAME-OST[$firstx-$lastx/$step] failed"
+ return 3
+ fi
# wait for OSTs to be added to the pool
for mds_id in $(seq $MDSCOUNT); do
local lodname=$FSNAME-MDT$(printf "%04x" $mdt_id)-mdtlov
wait_update_facet mds$mds_id \
"lctl get_param -n lod.$lodname.pools.$pool |
- sort -u | tr '\n' ' ' " "$t" || {
+ grep $tg | sort -u | tr '\n' ' '" "$t" || {
error_noexit "mds$mds_id: Add to pool failed"
- return 3
+ return 2
}
done
- wait_update $HOSTNAME "lctl get_param -n lov.$FSNAME-*.pools.$pool \
- | sort -u | tr '\n' ' ' " "$t" || {
+ wait_update $HOSTNAME "lctl get_param -n lov.$FSNAME-*.pools.$pool |
+ grep $tg | sort -u | tr '\n' ' ' " "$t" || {
error_noexit "Add to pool failed"
return 1
}
- local lfscount=$($LFS pool_list $FSNAME.$pool | grep -c "\-OST")
- local addcount=$(((last - first) / step + 1))
- [ $lfscount -eq $addcount ] || {
- error_noexit "lfs pool_list bad ost count" \
- "$lfscount != $addcount"
- return 2
- }
}
pool_set_dir() {
export SINGLEAGT=${SINGLEAGT:-agt1}
export HSMTOOL=${HSMTOOL:-"lhsmtool_posix"}
+ export HSMTOOL_PID_FILE=${HSMTOOL_PID_FILE:-"/var/run/lhsmtool_posix.pid"}
export HSMTOOL_VERBOSE=${HSMTOOL_VERBOSE:-""}
export HSMTOOL_UPDATE_INTERVAL=${HSMTOOL_UPDATE_INTERVAL:=""}
export HSMTOOL_EVENT_FIFO=${HSMTOOL_EVENT_FIFO:=""}
export HSMTOOL_TESTDIR
- # Copytools invoked from within the source tree are executed through a
- # libtool script and have a different process name
- from_build_tree && LIBTOOL_PREFIX=lt-
-
- export HSMTOOL_KILL=${LIBTOOL_PREFIX}${HSMTOOL}
-
- # pgrep(1) and pkill(1) limit process name matching to 15 characters
- export HSMTOOL_PKILL=${HSMTOOL_KILL:0:15}
+ if ! [[ $HSMTOOL =~ hsmtool ]]; then
+ echo "HSMTOOL = '$HSMTOOL' does not contain 'hsmtool', GLWT" >&2
+ fi
HSM_ARCHIVE_NUMBER=2
done
}
+pkill_copytools() {
+ local hosts="$1"
+ local signal="$2"
+
+ do_nodes "$hosts" "pkill --pidfile=$HSMTOOL_PID_FILE --signal=$signal hsmtool"
+}
+
copytool_continue() {
local agents=${1:-$(facet_active_host $SINGLEAGT)}
- do_nodesv $agents "pkill -CONT -x $HSMTOOL_PKILL" || return 0
+ pkill_copytools "$agents" CONT || return 0
echo "Copytool is continued on $agents"
}
local hosts=${1:-$(facet_active_host $SINGLEAGT)}
echo "Killing existing copytools on $hosts"
- do_nodesv $hosts "killall -q $HSMTOOL_KILL" || true
+ pkill_copytools "$hosts" TERM || return 0
copytool_continue "$hosts"
}
__lhsmtool_setup()
{
- local cmd="$HSMTOOL $HSMTOOL_VERBOSE --daemon --hsm-root \"$hsm_root\""
+ local host="$(facet_host "$facet")"
+ local cmd="$HSMTOOL $HSMTOOL_VERBOSE --daemon --pid-file=$HSMTOOL_PID_FILE --hsm-root \"$hsm_root\""
[ -n "$bandwidth" ] && cmd+=" --bandwidth $bandwidth"
[ -n "$archive_id" ] && cmd+=" --archive $archive_id"
[ ${#misc_options[@]} -gt 0 ] &&
cmd+=" $(IFS=" " echo "$@")"
cmd+=" \"$mountpoint\""
- echo "Starting copytool $facet on $(facet_host $facet)"
- stack_trap "do_facet $facet pkill -x '$HSMTOOL_PKILL' || true" EXIT
- do_facet $facet "$cmd < /dev/null > \"$(copytool_logfile $facet)\" 2>&1"
+ echo "Starting copytool '$facet' on '$host'"
+ stack_trap "pkill_copytools $host TERM || true" EXIT
+ do_node "$host" "$cmd < /dev/null > \"$(copytool_logfile $facet)\" 2>&1"
}
hsm_root() {
do_nodes $list "$LCTL set_param debug=\\\"$saved_debug\\\""
return $rc
}
+
+# Check if fallocate supported on OSTs, enable if unset, default mode=0
+# Optionally pass the OST fallocate mode (0=unwritten extents, 1=zero extents)
+function check_set_fallocate()
+{
+ local new_mode="$1"
+ local osts=$(comma_list $(osts_nodes))
+ local fa_mode="osd-ldiskfs.*.fallocate_zero_blocks"
+ local old_mode=$(do_facet ost1 $LCTL get_param -n $fa_mode 2>/dev/null|
+ head -n 1)
+
+ [[ -n "$old_mode" ]] || { echo "fallocate not supported"; return 1; }
+ [[ -z "$new_mode" && "$old_mode" != "-1" ]] &&
+ { echo "keep default fallocate mode: $old_mode"; return 0; }
+ [[ "$new_mode" && "$old_mode" == "$new_mode" ]] &&
+ { echo "keep current fallocate mode: $old_mode"; return 0; }
+
+ stack_trap "do_nodes $osts $LCTL set_param $fa_mode=$old_mode"
+ do_nodes $osts $LCTL set_param $fa_mode=${new_mode:-0} ||
+ error "set $fa_mode=$new_mode"
+}
+
+# Check if fallocate supported on OSTs, enable if unset, skip if unavailable
+function check_set_fallocate_or_skip()
+{
+ [ "$ost1_FSTYPE" != ldiskfs ] && skip "non-ldiskfs backend"
+ check_set_fallocate || skip "need at least 2.13.57 for fallocate"
+}
+