[ ! -f "$SGPDDSURVEY" ] && export SGPDDSURVEY=$(which sgpdd-survey)
export MCREATE=${MCREATE:-mcreate}
export MULTIOP=${MULTIOP:-multiop}
+ export MMAP_CAT=${MMAP_CAT:-mmap_cat}
+ export STATX=${STATX:-statx}
# Ubuntu, at least, has a truncate command in /usr/bin
# so fully path our truncate command.
export TRUNCATE=${TRUNCATE:-$LUSTRE/tests/truncate}
[ ! -f "$LCTL" ] && export LCTL=$(which lctl)
export LFS=${LFS:-"$LUSTRE/utils/lfs"}
[ ! -f "$LFS" ] && export LFS=$(which lfs)
- SETSTRIPE=${SETSTRIPE:-"$LFS setstripe"}
- GETSTRIPE=${GETSTRIPE:-"$LFS getstripe"}
export PERM_CMD=${PERM_CMD:-"$LCTL conf_param"}
. ${CONFIG:=$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() {
# 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 << 16) | ($2 << 8) | $3))"
+ 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/")
# 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 $lver
}
# Report the Lustre numeric build version code for the supplied facet.
# if there is more than 4 CPU cores, libcfs should create multiple CPU
# partitions. So we just force libcfs to create 2 partitions for
# system with 2 or 4 cores
+ local saved_opts="$MODOPTS_LIBCFS"
if [ $ncpus -le 4 ] && [ $ncpus -gt 1 ]; then
# force to enable multiple CPU partitions
echo "Force libcfs to create 2 CPU partitions"
load_module ../libcfs/libcfs/libcfs
# Prevent local MODOPTS_LIBCFS being passed as part of environment
# variable to remote nodes
- unset MODOPTS_LIBCFS
+ MODOPTS_LIBCFS=$saved_opts
set_default_debug
load_module ../lnet/lnet/lnet
load_module fid/fid
load_module lmv/lmv
load_module osc/osc
- load_module mdc/mdc
load_module lov/lov
+ load_module mdc/mdc
load_module mgc/mgc
load_module obdecho/obdecho
if ! client_only; then
}
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() {
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
return
fi
- case $LUSTRE in
- /usr/lib/lustre/* | /usr/lib64/lustre/* | /usr/lib/lustre | \
- /usr/lib64/lustre )
- from_build_tree=false
- ;;
- *)
- from_build_tree=true
- ;;
- esac
-
if ! module_loaded ptlrpc_gss; then
load_module ptlrpc/gss/ptlrpc_gss
module_loaded ptlrpc_gss ||
start_gss_daemons || error_exit "start gss daemon failed! rc=$?"
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}
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."
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 $(all_nodes)) "mount \
OST_MOUNT_OPTS=$(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
+ 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"
fi
}
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"
$SK_PATH/$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 $(all_nodes)) "while grep -q \
}
set_default_debug () {
- local debug=${1:-"$PTLDEBUG"}
- local subsys=${2:-"$SUBSYSTEM"}
- local debug_size=${3:-$DEBUG_SIZE}
+ local debug=${1:-"$PTLDEBUG"}
+ local subsys=${2:-"$SUBSYSTEM"}
+ local debug_size=${3:-$DEBUG_SIZE}
- [ -n "$debug" ] && lctl set_param debug="$debug" >/dev/null
- [ -n "$subsys" ] && lctl set_param subsystem_debug="${subsys# }" >/dev/null
+ [ -n "$debug" ] && lctl set_param debug="$debug" >/dev/null
+ [ -n "$subsys" ] && lctl set_param subsystem_debug="${subsys# }" >/dev/null
- [ -n "$debug_size" ] && set_debug_size $debug_size > /dev/null
+ [ -n "$debug_size" ] && set_debug_size $debug_size > /dev/null
}
set_default_debug_nodes () {
local nodes="$1"
+ local debug="${2:-"$PTLDEBUG"}"
+ local subsys="${3:-"$SUBSYSTEM"}"
+ local debug_size="${4:-$DEBUG_SIZE}"
if [[ ,$nodes, = *,$HOSTNAME,* ]]; then
nodes=$(exclude_items_from_list "$nodes" "$HOSTNAME")
set_default_debug
fi
- do_rpc_nodes "$nodes" set_default_debug \
- \\\"$PTLDEBUG\\\" \\\"$SUBSYSTEM\\\" $DEBUG_SIZE || true
+ [[ -z "$nodes" ]] ||
+ do_rpc_nodes "$nodes" set_default_debug \
+ \\\"$debug\\\" \\\"$subsys\\\" $debug_size || true
}
set_default_debug_facet () {
- local facet=$1
- local node=$(facet_active_host $facet)
- [ -z "$node" ] && echo "No host defined for facet $facet" && exit 1
+ local facet=$1
+ local debug="${2:-"$PTLDEBUG"}"
+ local subsys="${3:-"$SUBSYSTEM"}"
+ local debug_size="${4:-$DEBUG_SIZE}"
+ local node=$(facet_active_host $facet)
- set_default_debug_nodes $node
+ [ -n "$node" ] || error "No host defined for facet $facet"
+
+ 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 () {
ost_dev_status() {
local ost_idx=$1
local mnt_pnt=${2:-$MOUNT}
+ local opts=$3
local ost_uuid
ost_uuid=$(ostuuid_from_index $ost_idx $mnt_pnt)
- lfs_df $mnt_pnt | awk '/'$ost_uuid'/ { print $7 }'
+ lfs_df $opts $mnt_pnt | awk '/'$ost_uuid'/ { print $7 }'
}
setup_quota(){
exit 1
fi
+ if $GSS_SK; then
+ # update mount option with skpath
+ opts=$(add_sk_mntflag $opts)
+ fi
+
echo "Starting client: $client: $flags $opts $device $mnt"
do_node $client mkdir -p $mnt
if [ -n "$FILESET" -a -z "$SKIP_FILESET" ];then
fi
set_default_debug_nodes $client
+ set_params_clients $client
return 0
}
fi
}
+# Mount the file system on the MDS
+mount_mds_client() {
+ local mds_HOST=${SINGLEMDS}_HOST
+ echo $mds_HOST
+ 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 "rmdir $MOUNT2"
+}
+
# nodes is comma list
sanity_mount_check_nodes () {
local nodes=$1
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 $node load_modules_local
+ fi
+ fi
}
facets_hosts () {
}
start_client_loads () {
- local -a clients=(${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
return 0
}
-wait_update () {
+##
+# wait for a command to return the expected result
+#
+# This will run @check on @node repeatedly until the output matches @expect
+# based on the supplied condition, or until @max_wait seconds have elapsed,
+# whichever comes first. @cond may be one of the normal bash operators,
+# "-gt", "-ge", "-eq", "-le", "-lt", "==", "!=", or "=~", and must be quoted
+# in the caller to avoid unintentional evaluation by the shell in the caller.
+#
+# If @max_wait is not specified, the condition will be checked for up to 90s.
+#
+# If --verbose is passed as the first argument, the result is printed on each
+# value change, otherwise it is only printed after every 10s interval.
+#
+# Using wait_update_cond() or related helper function is preferable to adding
+# a "long enough" wait for some state to change in the background, since
+# "long enough" may be too short due to tunables, system config, or running in
+# a VM, and must by necessity wait too long for most cases or risk failure.
+#
+# usage: wait_update_cond [--verbose] node check cond expect [max_wait]
+wait_update_cond() {
local verbose=false
- if [[ "$1" == "--verbose" ]]; then
- shift
- verbose=true
- fi
+ [[ "$1" == "--verbose" ]] && verbose=true && shift
local node=$1
- local TEST=$2
- local FINAL=$3
- local MAX=${4:-90}
- local RESULT
- local PREV_RESULT
- local WAIT=0
+ local check="$2"
+ local cond="$3"
+ local expect="$4"
+ local max_wait=${5:-90}
+ local result
+ local prev_result
+ local waited=0
+ local begin=$SECONDS
local sleep=1
local print=10
- PREV_RESULT=$(do_node $node "$TEST")
- while [ true ]; do
- RESULT=$(do_node $node "$TEST")
- if [[ "$RESULT" == "$FINAL" ]]; then
- [[ -z "$RESULT" || $WAIT -le $sleep ]] ||
- echo "Updated after ${WAIT}s: wanted '$FINAL'"\
- "got '$RESULT'"
+ while (( $waited <= $max_wait )); do
+ result=$(do_node $node "$check")
+
+ eval [[ "'$result'" $cond "'$expect'" ]]
+ if [[ $? == 0 ]]; then
+ [[ -z "$result" || $waited -le $sleep ]] ||
+ echo "Updated after ${waited}s: want '$expect' got '$result'"
return 0
fi
- if [[ $verbose && "$RESULT" != "$PREV_RESULT" ]]; then
- echo "Changed after ${WAIT}s: from '$PREV_RESULT'"\
- "to '$RESULT'"
- PREV_RESULT=$RESULT
+ if $verbose && [[ "$result" != "$prev_result" ]]; then
+ [[ -n "$prev_result" ]] &&
+ echo "Changed after ${waited}s: from '$prev_result' to '$result'"
+ prev_result="$result"
fi
- [[ $WAIT -ge $MAX ]] && break
- [[ $((WAIT % print)) -eq 0 ]] &&
- echo "Waiting $((MAX - WAIT)) secs for update"
- WAIT=$((WAIT + sleep))
+ (( $waited % $print == 0 )) &&
+ echo "Waiting $((max_wait - waited))s for '$expect'"
sleep $sleep
+ waited=$((SECONDS - begin))
done
- echo "Update not seen after ${MAX}s: wanted '$FINAL' got '$RESULT'"
+ echo "Update not seen after ${max_wait}s: want '$expect' got '$result'"
return 3
}
+# usage: wait_update [--verbose] node check expect [max_wait]
+wait_update() {
+ local verbose=
+ [ "$1" = "--verbose" ] && verbose="$1" && shift
+
+ local node="$1"
+ local check="$2"
+ local expect="$3"
+ local max_wait=$4
+
+ wait_update_cond $verbose $node "$check" "==" "$expect" $max_wait
+}
+
+# usage: wait_update_facet_cond [--verbose] facet check cond expect [max_wait]
+wait_update_facet_cond() {
+ local verbose=
+ [ "$1" = "--verbose" ] && verbose="$1" && shift
+
+ local node=$(facet_active_host $1)
+ local check="$2"
+ local cond="$3"
+ local expect="$4"
+ local max_wait=$5
+
+ wait_update_cond $verbose $node "$check" "$cond" "$expect" $max_wait
+}
+
+# usage: wait_update_facet [--verbose] facet check expect [max_wait]
wait_update_facet() {
local verbose=
[ "$1" = "--verbose" ] && verbose="$1" && shift
- local facet=$1
- shift
- wait_update $verbose $(facet_active_host $facet) "$@"
+ local node=$(facet_active_host $1)
+ local check="$2"
+ local expect="$3"
+ local max_wait=$4
+
+ wait_update_cond $verbose $node "$check" "==" "$expect" $max_wait
}
sync_all_data() {
}
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() {
}
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: $?"
}
local nodes=$1
local net=${2:-"."}
- do_nodes $nodes "$LCTL list_nids | grep $net | cut -f 1 -d @"
+ do_nodes $nodes "$LCTL list_nids | grep -w $net | cut -f 1 -d @"
}
h2name_or_ip() {
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 $facet) == zfs ]]; then
var=${type}_FS_MKFS_OPTS
fs_mkfs_opts+=${!var:+" ${!var}"}
+ [[ "$QUOTA_TYPE" =~ "p" ]] && fs_mkfs_opts+=" -O project"
+
[ $fstype == ldiskfs ] && fs_mkfs_opts=$(squash_opt $fs_mkfs_opts)
if [ -n "${fs_mkfs_opts## }" ]; then
# $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
}
}
is_mounted () {
- local mntpt=$1
- [ -z $mntpt ] && return 1
- local mounted=$(mounted_lustre_filesystems)
+ local mntpt=$1
+ [ -z $mntpt ] && return 1
+ local mounted=$(mounted_lustre_filesystems)
- echo $mounted' ' | grep -w -q $mntpt' '
+ echo $mounted' ' | grep -w -q $mntpt' '
}
is_empty_dir() {
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 $OSD_TRACK_DECLARES_LBUG) == 'yes' ]; then
local facets=""
fi
fi
+ if [ -n "$fs_STRIPEPARAMS" ]; then
+ setstripe_getstripe $MOUNT $fs_STRIPEPARAMS
+ fi
if $GSS_SK; then
set_flavor_all null
elif $GSS; then
local log=$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]}
[ "$host" = "$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() {
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=$lru_size
}
flock_is_enabled()
}
start_full_debug_logging() {
- debugsave
- debug_size_save
+ debugsave
+ debug_size_save
- local FULLDEBUG=-1
- local DEBUG_SIZE=150
+ local fulldebug=-1
+ local debug_size=150
+ local nodes=$(comma_list $(nodes_list))
- do_nodes $(comma_list $(nodes_list)) "$LCTL set_param debug_mb=$DEBUG_SIZE"
- do_nodes $(comma_list $(nodes_list)) "$LCTL set_param debug=$FULLDEBUG;"
+ do_nodes $nodes "$LCTL set_param debug=$fulldebug debug_mb=$debug_size"
}
stop_full_debug_logging() {
- debug_size_restore
- debugrestore
+ debug_size_restore
+ debugrestore
}
# prints bash call stack
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 $status
}
[[ -n "$TESTSUITELOG" ]] &&
echo "$TESTSUITE: SKIP: $TESTNAME $@" >> $TESTSUITELOG || true
+ unset TESTNAME
}
skip() {
fi
done
- [ "$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() {
if [[ $1 = [a-z]* ]]; then
echo $1
else
- echo ${1%%[a-z]*}
+ echo ${1%%[a-zA-Z]*}
fi
}
export LAST_SKIPPED=
export ALWAYS_SKIPPED=
#
-# Main entry into test-framework. This is called with the name and
-# description of a test. The name is used to find the function to run
+# Main entry into test-framework. This is called with the number and
+# description of a test. The number is used to find the function to run
# the test using "test_$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 "$HONOR_EXCEPT" ]]; then
LAST_SKIPPED="y"
[ -n "$LAST_SKIPPED" ] &&
echo "" && LAST_SKIPPED=
ALWAYS_SKIPPED=
- run_one_logged $1 "$2"
+ run_one_logged $testnum "$testmsg"
return $?
fi
skip_noexit "$skip_message"
return 0
else
- run_one_logged $1 "$2"
+ run_one_logged $testnum "$testmsg"
return $?
fi
-
}
log() {
#
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
$SETUP
fi
- banner "test $testnum: $message"
+ banner "test $testnum: $testmsg"
test_${testnum} || error "test_$testnum failed with $?"
cd $SAVE_PWD
reset_fail_loc
check_node_health
check_dmesg_for_errors || error "Error in dmesg detected"
if [ "$PARALLEL" != "yes" ]; then
- ps auxww | grep -v grep | grep -q multiop &&
+ ps auxww | grep -v grep | grep -q "multiop " &&
error "multiop still running"
fi
- unset TESTNAME
- unset tdir
- unset tfile
umask $SAVE_UMASK
$CLEANUP
return 0
# - test result is saved to data file
#
run_one_logged() {
- local BEFORE=$(date +%s)
- local TEST_ERROR
- local name=${TESTSUITE}.test_${1}.test_log.$(hostname -s).log
- local test_log=$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 $LOGDIR/skip
+ local before=$SECONDS
+ local testnum=$1
+ local testmsg=$2
+ export tfile=f${testnum}.${TESTSUITE}
+ export tdir=d${testnum}.${TESTSUITE}
+ local test_log=$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
+ 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
}
(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)
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
}
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"
}
do_and_time () {
- local cmd=$1
- local rc
-
- SECONDS=0
- eval '$cmd'
+ local cmd="$1"
+ local start
+ local rc
- [ ${PIPESTATUS[0]} -eq 0 ] || rc=1
+ start=$SECONDS
+ eval '$cmd'
+ [ ${PIPESTATUS[0]} -eq 0 ] || rc=1
- echo $SECONDS
- return $rc
+ echo $((SECONDS - start))
+ return $rc
}
inodes_available () {
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
}
}
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 () {
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() {
# ldiskfs xattrs over one block in size. Allow both the historical
# Lustre feature name (large_xattr) and the upstream name (ea_inode).
large_xattr_enabled() {
- [[ $(facet_fstype $SINGLEMDS) == zfs ]] && return 0
+ [[ $(facet_fstype $SINGLEMDS) == zfs ]] && return 1
local mds_dev=$(mdsdevname ${SINGLEMDS//mds/})
# Get the maximum xattr size supported by the filesystem.
max_xattr_size() {
- local size
-
- if large_xattr_enabled; then
- size=$($LCTL get_param -n llite.*.max_easize)
- else
- local mds_dev=$(mdsdevname ${SINGLEMDS//mds/})
- local block_size=$(get_block_size $SINGLEMDS $mds_dev)
-
- # maximum xattr size = size of block - size of header -
- # size of 1 entry - 4 null bytes
- size=$((block_size - 32 - 32 - 4))
- fi
-
- echo $size
+ $LCTL get_param -n llite.*.max_easize
}
# Dump the value of the named xattr from a file.
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 ${devname}; then
local devname=$(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 "$opts" -o loop)
test_mkdir() {
local path
local p_option
+ local hash_type
+ local hash_name=("all_char" "fnv_1a_64" "crush")
local dirstripe_count=${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
[ $# -eq 1 ] || error "Only creating single directory is supported"
path="$*"
+ local parent=$(dirname $path)
if [ "$p_option" == "-p" ]; then
- local parent=$(dirname $path)
-
[ -d $path ] && return 0
if [ ! -d ${parent} ]; then
mkdir -p ${parent} ||
fi
fi
- if [ $MDSCOUNT -le 1 ]; then
+ if [ $MDSCOUNT -le 1 ] || ! is_lustre ${parent}; then
mkdir $path || error "mkdir '$path' failed"
else
local mdt_index
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))
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
}
local file=$1
local pool=$2
local tlist="$3"
- local res=$($GETSTRIPE $file | grep 0x | cut -f2)
+ local res=$($LFS getstripe $file | grep 0x | cut -f2)
for i in $res
do
for t in $tlist ; do
local tdir=$2
echo "Setting pool on directory $tdir"
- $SETSTRIPE -c 2 -p $pool $tdir && return 0
+ $LFS setstripe -c 2 -p $pool $tdir && return 0
error_noexit "Cannot set pool $pool to $tdir"
return 1
local tdir=$2
echo "Checking pool on directory $tdir"
- local res=$($GETSTRIPE --pool $tdir | sed "s/\s*$//")
+ local res=$($LFS getstripe --pool $tdir | sed "s/\s*$//")
[ "$res" = "$pool" ] && return 0
error_noexit "Pool on '$tdir' is '$res', not '$pool'"
for i in $(seq -w 1 $count)
do
local file=$tdir/spoo-$i
- $SETSTRIPE -p $pool $file
+ $LFS setstripe -p $pool $file
check_file_in_pool $file $pool "$tlist" || \
failed=$((failed + 1))
done
mkdir -p $tdir ||
{ error_noexit "unable to create $tdir"; return 1 ; }
local file="/..$tdir/$tfile-1"
- $SETSTRIPE -p $pool $file ||
+ $LFS setstripe -p $pool $file ||
{ error_noexit "unable to create $file" ; return 2 ; }
cd $tdir
- $SETSTRIPE -p $pool $tfile-2 || {
+ $LFS setstripe -p $pool $tfile-2 || {
error_noexit "unable to create $tfile-2 in $tdir"
return 3
}
pool_remove_first_target() {
echo "Removing first target from a pool"
+ pool_remove_target $1 -1
+}
+
+pool_remove_target() {
local pool=$1
+ local index=$2
local pname="lov.$FSNAME-*.pools.$pool"
- local t=$($LCTL get_param -n $pname | head -1)
+ if [ $index -eq -1 ]; then
+ local t=$($LCTL get_param -n $pname | head -1)
+ else
+ local t=$(printf "$FSNAME-OST%04x_UUID" $index)
+ fi
+
+ echo "Removing $t from $pool"
do_facet mgs $LCTL pool_remove $FSNAME.$pool $t
for mds_id in $(seq $MDSCOUNT); do
local mdt_id=$((mds_id-1))
return 2
}
# setstripe on an empty pool should fail
- $SETSTRIPE -p $pool $file 2>/dev/null && {
+ $LFS setstripe -p $pool $file 2>/dev/null && {
error_noexit "expected failure when creating file" \
"with empty pool"
return 3
return 1
}
# setstripe on an empty pool should fail
- $SETSTRIPE -p $pool $file 2>/dev/null && {
+ $LFS setstripe -p $pool $file 2>/dev/null && {
error_noexit "expected failure when creating file" \
"with missing pool"
return 2
[[ -z "$file" || -z "$expected" ]] &&
error "check_obdidx: invalid argument!"
- obdidx=$(comma_list $($GETSTRIPE $file | grep -A $OSTCOUNT obdidx |
+ obdidx=$(comma_list $($LFS getstripe $file | grep -A $OSTCOUNT obdidx |
grep -v obdidx | awk '{print $1}' | xargs))
[[ $obdidx = $expected ]] ||
[[ -z "$file" || -z "$expected" ]] &&
error "check_start_ost_idx: invalid argument!"
- start_ost_idx=$($GETSTRIPE $file | grep -A 1 obdidx | grep -v obdidx |
- awk '{print $1}')
+ start_ost_idx=$($LFS 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" \
}
changelog_dump() {
+ local rc
+
for M in $(seq $MDSCOUNT); do
local facet=mds$M
local mdt="$(facet_svc $facet)"
-
- $LFS changelog $mdt | sed -e 's/^/'$mdt'./'
+ local output
+ local ret
+
+ output=$($LFS changelog $mdt)
+ ret=$?
+ if [ $ret -ne 0 ]; then
+ rc=${rc:-$ret}
+ elif [ -n "$output" ]; then
+ echo "$output" | sed -e 's/^/'$mdt'./'
+ fi
done
+
+ return ${rc:-0}
}
changelog_extract_field() {
is_project_quota_supported() {
$ENABLE_PROJECT_QUOTAS || return 1
- [ "$(facet_fstype $SINGLEMDS)" == "ldiskfs" ] &&
- [ $(lustre_version_code $SINGLEMDS) -gt \
- $(version_code 2.9.55) ] &&
- lfs --help | grep project >&/dev/null &&
- egrep -q "7." /etc/redhat-release && return 0
- if [ "$(facet_fstype $SINGLEMDS)" == "zfs" ]; then
- [ $(lustre_version_code $SINGLEMDS) -le \
- $(version_code 2.10.53) ] && return 1
+ [[ "$(facet_fstype $SINGLEMDS)" == "ldiskfs" &&
+ $(lustre_version_code $SINGLEMDS) -gt $(version_code 2.9.55) ]] &&
+ do_facet mds1 lfs --help |& grep -q project && return 0
- do_facet mds1 $ZPOOL upgrade -v |
- grep project_quota && return 0
- fi
+ [[ "$(facet_fstype $SINGLEMDS)" == "zfs" &&
+ $(lustre_version_code $SINGLEMDS) -gt $(version_code 2.10.53) ]] &&
+ do_facet mds1 $ZPOOL get all | grep -q project_quota && return 0
return 1
}
+# ZFS project quota enable/disable:
+# This feature will become active as soon as it is enabled and will never
+# return to being disabled. Each filesystem will be upgraded automatically
+# when remounted or when [a] new file is created under that filesystem. The
+# upgrade can also be triggered on filesystems via `zfs set version=current
+# <pool/fs>`. The upgrade process runs in the background and may take a
+# while to complete for the filesystems containing a large number of files.
enable_project_quota() {
is_project_quota_supported || return 0
- [ "$(facet_fstype $SINGLEMDS)" != "ldiskfs" ] && return 0
+ local zkeeper=${KEEP_ZPOOL}
+ stack_trap "KEEP_ZPOOL=$zkeeper" EXIT
+ KEEP_ZPOOL="true"
stopall || error "failed to stopall (1)"
- for num in $(seq $MDSCOUNT); do
- do_facet mds$num $TUNE2FS -O project $(mdsdevname $num) ||
- error "tune2fs $(mdsdevname $num) failed"
- done
+ local zfeat_en="feature@project_quota=enabled"
+ for facet in $(seq -f mds%g $MDSCOUNT) $(seq -f ost%g $OSTCOUNT); do
+ local facet_fstype=${facet:0:3}1_FSTYPE
+ local devname
- for num in $(seq $OSTCOUNT); do
- do_facet ost$num $TUNE2FS -O project $(ostdevname $num) ||
- error "tune2fs $(ostdevname $num) failed"
+ if [ "${!facet_fstype}" = "zfs" ]; then
+ devname=$(zpool_name ${facet})
+ do_facet ${facet} $ZPOOL set "$zfeat_en" $devname ||
+ error "$ZPOOL set $zfeat_en $devname"
+ else
+ [ ${facet:0:3} == "mds" ] &&
+ devname=$(mdsdevname ${facet:3}) ||
+ devname=$(ostdevname ${facet:3})
+ do_facet ${facet} $TUNE2FS -O project $devname ||
+ error "tune2fs $devname failed"
+ fi
done
+ KEEP_ZPOOL="${zkeeper}"
mount
setupall
}
disable_project_quota() {
is_project_quota_supported || return 0
- [ "$(facet_fstype $SINGLEMDS)" != "ldiskfs" ] && return 0
+ [ "$mds1_FSTYPE" != "ldiskfs" ] && return 0
stopall || error "failed to stopall (1)"
for num in $(seq $MDSCOUNT); do
export SINGLEAGT=${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=${HSMTOOL_EVENT_FIFO:=""}
export HSMTOOL_TESTDIR
- export HSMTOOL_BASE=$(basename "$HSMTOOL" | cut -f1 -d" ")
+
+ if ! [[ $HSMTOOL =~ hsmtool ]]; then
+ echo "HSMTOOL = '$HSMTOOL' does not contain 'hsmtool', GLWT" >&2
+ fi
HSM_ARCHIVE_NUMBER=2
done
}
-search_copytools() {
- local hosts=${1:-$(facet_active_host $SINGLEAGT)}
- do_nodesv $hosts "pgrep -x $HSMTOOL_BASE"
+pkill_copytools() {
+ local hosts="$1"
+ local signal="$2"
+
+ do_nodes "$hosts" "pkill --pidfile=$HSMTOOL_PID_FILE --signal=$signal hsmtool"
}
-kill_copytools() {
- local hosts=${1:-$(facet_active_host $SINGLEAGT)}
+copytool_continue() {
+ local agents=${1:-$(facet_active_host $SINGLEAGT)}
- echo "Killing existing copytools on $hosts"
- do_nodesv $hosts "killall -q $HSMTOOL_BASE" || true
+ pkill_copytools "$agents" CONT || return 0
+ echo "Copytool is continued on $agents"
}
-wait_copytools() {
+kill_copytools() {
local hosts=${1:-$(facet_active_host $SINGLEAGT)}
- local wait_timeout=200
- local wait_start=$SECONDS
- local wait_end=$((wait_start + wait_timeout))
- local sleep_time=100000 # 0.1 second
-
- while ((SECONDS < wait_end)); do
- if ! search_copytools $hosts; then
- echo "copytools stopped in $((SECONDS - wait_start))s"
- return 0
- fi
-
- echo "copytools still running on $hosts"
- usleep $sleep_time
- [ $sleep_time -lt 32000000 ] && # 3.2 seconds
- sleep_time=$(bc <<< "$sleep_time * 2")
- done
- # try to dump Copytool's stack
- do_nodesv $hosts "echo 1 >/proc/sys/kernel/sysrq ; " \
- "echo t >/proc/sysrq-trigger"
-
- echo "copytools failed to stop in ${wait_timeout}s"
-
- return 1
+ echo "Killing existing copytools on $hosts"
+ pkill_copytools "$hosts" TERM || return 0
+ copytool_continue "$hosts"
}
copytool_monitor_cleanup() {
__lhsmtool_setup()
{
- local cmd="$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 $facet libtool execute pkill -x '$HSMTOOL' || 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() {
local action=$1
shift
+ # Use default values
+ local facet=$SINGLEAGT
+ local mountpoint="${MOUNT2:-$MOUNT}"
+ local hsm_root="${hsm_root:-$(hsm_root "$facet")}"
+
# Parse arguments
local fail_on_error=true
local -a misc_options
case "$1" in
-f|--facet)
shift
- local facet="$1"
+ facet="$1"
;;
-m|--mountpoint)
shift
- local mountpoint="$1"
+ mountpoint="$1"
;;
-a|--archive-id)
shift
;;
-h|--hsm-root)
shift
- local hsm_root="$1"
+ hsm_root="$1"
;;
-b|--bwlimit)
shift
shift
done
- # Use default values if needed
- local facet=${facet:-$SINGLEAGT}
- local mountpoint="${mountpoint:-${MOUNT2:-$MOUNT}}"
- local hsm_root="${hsm_root:-$(hsm_root "$facet")}"
-
stack_trap "do_facet $facet rm -rf '$hsm_root'" EXIT
do_facet $facet mkdir -p "$hsm_root" ||
error "mkdir '$hsm_root' failed"
return $rc
}
-wait_result() {
- local facet=$1
- shift
- wait_update --verbose $(facet_active_host $facet) "$@"
-}
-
mdts_check_param() {
local key="$1"
local target="$2"
local timeout="$3"
local mdtno
+
for mdtno in $(seq 1 $MDSCOUNT); do
local idx=$(($mdtno - 1))
- wait_result mds${mdtno} \
+ wait_update_facet --verbose mds${mdtno} \
"$LCTL get_param -n $MDT_PREFIX${idx}.$key" "$target" \
$timeout ||
error "$key state is not '$target' on mds${mdtno}"
local cmd="$LCTL get_param -n ${MDT_PREFIX}${mdtidx}.hsm.actions"
cmd+=" | awk '/'$fid'.*action='$request'/ {print \\\$13}' | cut -f2 -d="
- wait_result $mds "$cmd" "$state" 200 ||
+ wait_update_facet --verbose $mds "$cmd" "$state" 200 ||
error "request on $fid is not $state on $mds"
}
error "No component starting at zero(!)"
}
-#TODO: This version is a placeholder, to be replaced before final commit
-SEL_VER="2.12.52"
+# version after which Self-Extending Layouts are available
+SEL_VER="2.12.55"
sel_layout_sanity() {
local file=$1
check_component_count $file $comp_cnt
}
+statx_supported() {
+ $STATX --quiet --version
+ return $?
+}
+
+#
+# wrappers for createmany and unlinkmany
+# to set debug=0 if number of creates is high enough
+# this is to speedup testing
+#
+function createmany() {
+ local count=${!#}
+
+ (( count > 100 )) && {
+ local saved_debug=$($LCTL get_param -n debug)
+ local list=$(comma_list $(all_nodes))
+
+ do_nodes $list $LCTL set_param debug=0
+ }
+ $LUSTRE/tests/createmany $*
+ local rc=$?
+ (( count > 100 )) &&
+ do_nodes $list "$LCTL set_param debug=\\\"$saved_debug\\\""
+ return $rc
+}
+
+function unlinkmany() {
+ local count=${!#}
+
+ (( count > 100 )) && {
+ local saved_debug=$($LCTL get_param -n debug)
+ local list=$(comma_list $(all_nodes))
+
+ do_nodes $list $LCTL set_param debug=0
+ }
+ $LUSTRE/tests/unlinkmany $*
+ local rc=$?
+ (( count > 100 )) &&
+ do_nodes $list "$LCTL set_param debug=\\\"$saved_debug\\\""
+ return $rc
+}
+
+function check_for_fallocate()
+{
+ [ "$ost1_FSTYPE" != ldiskfs ] && skip "non-ldiskfs backend"
+ 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" ] || skip "need at least 2.13.57 for fallocate"
+ stack_trap "do_nodes $osts $LCTL set_param $fa_mode=$old_mode"
+ do_nodes $osts $LCTL set_param $fa_mode=0 || error "set $fa_mode=0"
+}
+
git://git.whamcloud.com / fs / lustre-release.git / blobdiff