set -e
#set -x
+export LANG=en_US
export EJOURNAL=${EJOURNAL:-""}
export REFORMAT=${REFORMAT:-""}
export WRITECONF=${WRITECONF:-""}
fi
fi
+ export RESIZE2FS=$RESIZE2FS
+ if [ -z "$RESIZE2FS" ]; then
+ if which resizefs.ldiskfs >/dev/null 2>&1; then
+ export RESIZE2FS=resizefs.ldiskfs
+ else
+ export RESIZE2FS=resize2fs
+ fi
+ fi
+
export LFSCK_BIN=${LFSCK_BIN:-lfsck}
export LFSCK_ALWAYS=${LFSCK_ALWAYS:-"no"} # check fs after each test suite
export FSCK_MAX_ERR=4 # File system errors left uncorrected
export RLUSTRE=${RLUSTRE:-$LUSTRE}
export RPWD=${RPWD:-$PWD}
export I_MOUNTED=${I_MOUNTED:-"no"}
- if [ ! -f /lib/modules/$(uname -r)/kernel/fs/lustre/mdt.ko -a \
- ! -f /lib/modules/$(uname -r)/updates/kernel/fs/lustre/mdt.ko -a \
- ! -f `dirname $0`/../mdt/mdt.ko ]; then
- export CLIENTMODSONLY=yes
- fi
+ if [ ! -f /lib/modules/$(uname -r)/kernel/fs/lustre/mdt.ko -a \
+ ! -f /lib/modules/$(uname -r)/updates/kernel/fs/lustre/mdt.ko -a \
+ ! -f /lib/modules/$(uname -r)/extra/kernel/fs/lustre/mdt.ko -a \
+ ! -f $LUSTRE/mdt/mdt.ko ]; then
+ export CLIENTMODSONLY=yes
+ fi
- export SHUTDOWN_ATTEMPTS=${SHUTDOWN_ATTEMPTS:-3}
+ export SHUTDOWN_ATTEMPTS=${SHUTDOWN_ATTEMPTS:-3}
export OSD_TRACK_DECLARES_LBUG=${OSD_TRACK_DECLARES_LBUG:-"yes"}
# command line
load_module ../lnet/lnet/lnet
LNETLND=${LNETLND:-"socklnd/ksocklnd"}
load_module ../lnet/klnds/$LNETLND
- load_module lvfs/lvfs
load_module obdclass/obdclass
load_module ptlrpc/ptlrpc
load_module ptlrpc/gss/ptlrpc_gss
modprobe zfs
load_module osd-zfs/osd_zfs
fi
- load_module mgs/mgs
- load_module mdd/mdd
if [[ $(node_fstypes $HOSTNAME) == *ldiskfs* ]]; then
- #
- # This block shall be moved up beside osd-zfs as soon
- # as osd-ldiskfs stops using mdd symbols.
- #
grep -q exportfs_decode_fh $SYMLIST ||
{ modprobe exportfs 2> /dev/null || true; }
- load_module ../ldiskfs/ldiskfs/ldiskfs
- load_module lvfs/fsfilt_ldiskfs
+ 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
load_module ost/ost
load_module lod/lod
fi
}
+facet_svc() {
+ local facet=$1
+ local var=${facet}_svc
+
+ echo -n ${!var}
+}
+
facet_type() {
local facet=$1
}
#
+# Create ZFS storage pool.
+#
+create_zpool() {
+ local facet=$1
+ local poolname=$2
+ local vdev=$3
+ shift 3
+ local opts=${@:-"-o cachefile=none"}
+
+ do_facet $facet "$ZPOOL list -H $poolname >/dev/null 2>&1 ||
+ $ZPOOL create -f $opts $poolname $vdev"
+}
+
+#
+# Create ZFS file system.
+#
+create_zfs() {
+ local facet=$1
+ local dataset=$2
+ shift 2
+ local opts=${@:-"-o mountpoint=legacy"}
+
+ do_facet $facet "$ZFS list -H $dataset >/dev/null 2>&1 ||
+ $ZFS create $opts $dataset"
+}
+
+#
# Export ZFS storage pool.
# Before exporting the pool, all datasets within the pool should be unmounted.
#
}
#
+# Destroy ZFS storage pool.
+# Destroy the given pool and free up any devices for other use. This command
+# tries to unmount any active datasets before destroying the pool.
+# -f Force any active datasets contained within the pool to be unmounted.
+#
+destroy_zpool() {
+ local facet=$1
+ local poolname=${2:-$(zpool_name $facet)}
+
+ if [[ -n "$poolname" ]]; then
+ do_facet $facet "! $ZPOOL list -H $poolname >/dev/null 2>&1 ||
+ $ZPOOL destroy -f $poolname"
+ fi
+}
+
+#
# Import ZFS storage pool.
# Force importing, even if the pool appears to be potentially active.
#
$LFS df $* | sed -e 's/filesystem /filesystem_/'
}
+# Get free inodes on the MDT specified by mdt index, free indoes on
+# the whole filesystem will be returned when index == -1.
+mdt_free_inodes() {
+ local index=$1
+ local free_inodes
+ local mdt_uuid
+
+ if [ $index -eq -1 ]; then
+ mdt_uuid="summary"
+ else
+ mdt_uuid=$(mdtuuid_from_index $index)
+ fi
+
+ free_inodes=$(lfs_df -i $MOUNT | grep $mdt_uuid | awk '{print $4}')
+ echo $free_inodes
+}
+
setup_quota(){
if [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.3.50) ]; then
setup_quota_old $1
TESTNAME=$TESTNAME \
DBENCH_LIB=$DBENCH_LIB \
DBENCH_SRC=$DBENCH_SRC \
+CLIENT_COUNT=$((CLIENTCOUNT - 1)) \
LFS=$LFS \
run_${load}.sh" &
local ppid=$!
}
wait_update () {
- local node=$1
- local TEST=$2
- local FINAL=$3
- local MAX=${4:-90}
-
- local RESULT
- local WAIT=0
- local sleep=1
- local print=10
- while [ true ]; do
- RESULT=$(do_node $node "$TEST")
- if [ "$RESULT" == "$FINAL" ]; then
- [ -z "$RESULT" -o $WAIT -le $sleep ] ||
- echo "Updated after ${WAIT}s: wanted '$FINAL' got '$RESULT'"
- return 0
- fi
- [ $WAIT -ge $MAX ] && break
- [ $((WAIT % print)) -eq 0 ] &&
- echo "Waiting $((MAX - WAIT)) secs for update"
- WAIT=$((WAIT + sleep))
- sleep $sleep
- done
- echo "Update not seen after ${MAX}s: wanted '$FINAL' got '$RESULT'"
- return 3
+ local verbose=false
+ if [[ "$1" == "--verbose" ]]; then
+ shift
+ verbose=true
+ fi
+
+ local node=$1
+ local TEST=$2
+ local FINAL=$3
+ local MAX=${4:-90}
+ local RESULT
+ local PREV_RESULT
+ local WAIT=0
+ 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'"
+ return 0
+ fi
+ if [[ $verbose && "$RESULT" != "$PREV_RESULT" ]]; then
+ echo "Changed after ${WAIT}s: from '$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))
+ sleep $sleep
+ done
+ echo "Update not seen after ${MAX}s: wanted '$FINAL' got '$RESULT'"
+ return 3
}
wait_update_facet() {
echo "Waiting for orphan cleanup..."
# MAX value includes time needed for MDS-OST reconnection
local MAX=$(( TIMEOUT * 2 ))
+ local WAIT_TIMEOUT=${1:-$MAX}
local WAIT=0
local new_wait=true
local list=$(comma_list $(mdts_nodes))
list=$(comma_list $(osts_nodes))
cmd="$LCTL get_param -n obdfilter.*.mds_sync"
fi
- while [ $WAIT -lt $MAX ]; do
+
+ echo "wait $WAIT_TIMEOUT secs maximumly for $list mds-ost sync done."
+ while [ $WAIT -lt $WAIT_TIMEOUT ]; do
local -a sync=($(do_nodes $list "$cmd"))
local con=1
local i
done
sleep 2 # increase waiting time and cover statfs cache
[ ${con} -eq 1 ] && return 0
- echo "Waiting $WAIT secs for $facet mds-ost sync done."
+ echo "Waiting $WAIT secs for $list $i mds-ost sync done."
WAIT=$((WAIT + 2))
done
- echo "$facet recovery not done in $MAX sec. $STATUS"
+
+ # show which nodes are not finished.
+ do_nodes $list "$cmd"
+ echo "$facet recovery node $i not done in $WAIT_TIMEOUT sec. $STATUS"
return 1
}
$LCTL mark "local REPLAY BARRIER on ${!svc}"
}
+#
+# Get Lustre client uuid for a given Lustre mount point.
+#
+get_client_uuid() {
+ local mntpnt=${1:-$MOUNT}
+
+ local name=$($LFS getname $mntpnt | cut -d' ' -f1)
+ local uuid=$($LCTL get_param -n llite.$name.uuid)
+
+ echo -n $uuid
+}
+
mds_evict_client() {
- UUID=`lctl get_param -n mdc.${mds1_svc}-mdc-*.uuid`
- do_facet mds1 "lctl set_param -n mdt.${mds1_svc}.evict_client $UUID"
+ local mntpnt=${1:-$MOUNT}
+ local uuid=$(get_client_uuid $mntpnt)
+
+ do_facet $SINGLEMDS \
+ "$LCTL set_param -n mdt.${mds1_svc}.evict_client $uuid"
}
ost_evict_client() {
- UUID=`lctl get_param -n devices| grep ${ost1_svc}-osc- | egrep -v 'MDT' | awk '{print $5}'`
- do_facet ost1 "lctl set_param -n obdfilter.${ost1_svc}.evict_client $UUID"
+ local mntpnt=${1:-$MOUNT}
+ local uuid=$(get_client_uuid $mntpnt)
+
+ do_facet ost1 \
+ "$LCTL set_param -n obdfilter.${ost1_svc}.evict_client $uuid"
}
fail() {
}
h2name_or_ip() {
- if [ "$1" = "client" -o "$1" = "'*'" ]; then echo \'*\'; else
- echo $1"@$2"
- fi
+ if [ "$1" = "'*'" ]; then echo \'*\'; else
+ echo $1"@$2"
+ fi
}
h2ptl() {
- if [ "$1" = "client" -o "$1" = "'*'" ]; then echo \'*\'; else
- ID=`xtprocadmin -n $1 2>/dev/null | egrep -v 'NID' | awk '{print $1}'`
- if [ -z "$ID" ]; then
- echo "Could not get a ptl id for $1..."
- exit 1
- fi
- echo $ID"@ptl"
- fi
+ if [ "$1" = "'*'" ]; then echo \'*\'; else
+ ID=`xtprocadmin -n $1 2>/dev/null | egrep -v 'NID' | \
+ awk '{print $1}'`
+ if [ -z "$ID" ]; then
+ echo "Could not get a ptl id for $1..."
+ exit 1
+ fi
+ echo $ID"@ptl"
+ fi
}
declare -fx h2ptl
h2tcp() {
- h2name_or_ip "$1" "tcp"
+ h2name_or_ip "$1" "tcp"
}
declare -fx h2tcp
h2elan() {
- if [ "$1" = "client" -o "$1" = "'*'" ]; then echo \'*\'; else
- if type __h2elan >/dev/null 2>&1; then
- ID=$(__h2elan $1)
- else
- ID=`echo $1 | sed 's/[^0-9]*//g'`
- fi
- echo $ID"@elan"
- fi
+ if [ "$1" = "'*'" ]; then echo \'*\'; else
+ if type __h2elan >/dev/null 2>&1; then
+ ID=$(__h2elan $1)
+ else
+ ID=`echo $1 | sed 's/[^0-9]*//g'`
+ fi
+ echo $ID"@elan"
+ fi
}
declare -fx h2elan
h2o2ib() {
- h2name_or_ip "$1" "o2ib"
+ h2name_or_ip "$1" "o2ib"
}
declare -fx h2o2ib
varname=${facet}_HOST
if [ -z "${!varname}" ]; then
if [ "${facet:0:3}" == "ost" ]; then
- eval export ${facet}_HOST=${ost_HOST}
+ local fh=${facet%failover}_HOST
+ eval export ${facet}_HOST=${!fh}
+ if [ -z "${!varname}" ]; then
+ eval export ${facet}_HOST=${ost_HOST}
+ fi
elif [ "${facet:0:3}" == "mdt" -o \
"${facet:0:3}" == "mds" -o \
"${facet:0:3}" == "mgs" ]; then
fi
}
+# Device formatted as ost
ostdevname() {
- num=$1
- DEVNAME=OSTDEV$num
+ local num=$1
+ local DEVNAME=OSTDEV$num
local fstype=$(facet_fstype ost$num)
#if $OSTDEVn isn't defined, default is $OSTDEVBASE + num
eval DEVPTR=${!DEVNAME:=${OSTDEVBASE}${num}};;
zfs )
- #dataset name is independent of vdev device names
- eval DEVPTR=${FSNAME}-ost${num}/ost${num};;
+ #try $OSTZFSDEVn - independent of vdev
+ DEVNAME=OSTZFSDEV$num
+ eval DEVPTR=${!DEVNAME:=${FSNAME}-ost${num}/ost${num}};;
* )
error "unknown fstype!";;
esac
echo -n $DEVPTR
}
+# Physical device location of data
ostvdevname() {
- num=$1
- DEVNAME=OSTDEV$num
+ local num=$1
+ local DEVNAME
+ local VDEVPTR
local fstype=$(facet_fstype ost$num)
# vdevs are not supported by ldiskfs
eval VDEVPTR="";;
zfs )
- #if $OSTDEVn isn't defined, default is $OSTDEVBASE + num
+ #if $OSTDEVn isn't defined, default is $OSTDEVBASE{n}
+ # Device formated by zfs
+ DEVNAME=OSTDEV$num
eval VDEVPTR=${!DEVNAME:=${OSTDEVBASE}${num}};;
* )
error "unknown fstype!";;
echo -n $VDEVPTR
}
+# Logical device formated for lustre
mdsdevname() {
- num=$1
- DEVNAME=MDSDEV$num
+ local num=$1
+ local DEVNAME=MDSDEV$num
local fstype=$(facet_fstype mds$num)
case $fstype in
ldiskfs )
- #if $MDSDEVn isn't defined, default is $MDSDEVBASE + num
+ #if $MDSDEVn isn't defined, default is $MDSDEVBASE{n}
eval DEVPTR=${!DEVNAME:=${MDSDEVBASE}${num}};;
zfs )
- #dataset name is independent of vdev device names
- eval DEVPTR=${FSNAME}-mdt${num}/mdt${num};;
+ # try $MDSZFSDEVn - independent of vdev
+ DEVNAME=MDSZFSDEV$num
+ eval DEVPTR=${!DEVNAME:=${FSNAME}-mdt${num}/mdt${num}};;
* )
error "unknown fstype!";;
esac
echo -n $DEVPTR
}
+# Physical location of data
mdsvdevname() {
- num=$1
- DEVNAME=MDSDEV$num
-
+ local VDEVPTR=""
+ local num=$1
local fstype=$(facet_fstype mds$num)
case $fstype in
# vdevs are not supported by ldiskfs
eval VDEVPTR="";;
zfs )
- #if $MDSDEVn isn't defined, default is $MDSDEVBASE + num
+ # if $MDSDEVn isn't defined, default is $MDSDEVBASE{n}
+ # Device formated by ZFS
+ local DEVNAME=MDSDEV$num
eval VDEVPTR=${!DEVNAME:=${MDSDEVBASE}${num}};;
* )
error "unknown fstype!";;
fi;;
zfs )
if [ $(facet_host mgs) = $(facet_host mds1) ] &&
- ( [ -z "$MGSDEV" ] || [ $MGSDEV = $(mdsvdevname 1) ] ); then
+ ( [ -z "$MGSZFSDEV" ] &&
+ [ -z "$MGSDEV" -o "$MGSDEV" = $(mdsvdevname 1) ] ); then
DEVPTR=$(mdsdevname 1)
else
- DEVPTR=${FSNAME}-mgs/mgs
+ DEVPTR=${MGSZFSDEV:-${FSNAME}-mgs/mgs}
fi;;
* )
error "unknown fstype!";;
}
mgsvdevname() {
- local VDEVPTR
- DEVNAME=MGSDEV
+ local VDEVPTR=""
local fstype=$(facet_fstype mgs)
;;
zfs )
if [ $(facet_host mgs) = $(facet_host mds1) ] &&
- ( [ -z "$MGSDEV" ] || [ $MGSDEV = $(mdsvdevname 1) ] ); then
+ ( [ -z "$MGSDEV" ] &&
+ [ -z "$MGSZFSDEV" -o "$MGSZFSDEV" = $(mdsdevname 1) ]); then
VDEVPTR=$(mdsvdevname 1)
- else
+ elif [ -n "$MGSDEV" ]; then
VDEVPTR=$MGSDEV
fi;;
* )
echo -n $mntpt
}
+mount_ldiskfs() {
+ local facet=$1
+ local dev=$(facet_device $facet)
+ local mnt=$(facet_mntpt $facet)
+ local opts
+
+ if ! do_facet $facet test -b $dev; then
+ opts="-o loop"
+ fi
+ do_facet $facet mount -t ldiskfs $opts $dev $mnt
+}
+
+unmount_ldiskfs() {
+ local facet=$1
+ local dev=$(facet_device $facet)
+ local mnt=$(facet_mntpt $facet)
+
+ do_facet $facet umount -d $mnt
+}
+
+var_name() {
+ echo -n "$1" | tr -c '[:alnum:]\n' '_'
+}
+
+mount_zfs() {
+ local facet=$1
+ local ds=$(facet_device $facet)
+ local mnt=$(facet_mntpt $facet)
+ local canmnt
+ local mntpt
+
+ import_zpool $facet
+ canmnt=$(do_facet $facet $ZFS get -H -o value canmount $ds)
+ mntpt=$(do_facet $facet $ZFS get -H -o value mountpoint $ds)
+ do_facet $facet $ZFS set canmount=noauto $ds
+ #
+ # The "legacy" mount method is used here because "zfs unmount $mnt"
+ # calls stat(2) on $mnt/../*, which may include $MOUNT. If certain
+ # targets are not available at the time, the stat(2) on $MOUNT will
+ # hang.
+ #
+ do_facet $facet $ZFS set mountpoint=legacy $ds
+ do_facet $facet mount -t zfs $ds $mnt
+ eval export mz_$(var_name ${facet}_$ds)_canmount=$canmnt
+ eval export mz_$(var_name ${facet}_$ds)_mountpoint=$mntpt
+}
+
+unmount_zfs() {
+ local facet=$1
+ local ds=$(facet_device $facet)
+ local mnt=$(facet_mntpt $facet)
+ local var_mntpt=mz_$(var_name ${facet}_$ds)_mountpoint
+ local var_canmnt=mz_$(var_name ${facet}_$ds)_canmount
+ local mntpt=${!var_mntpt}
+ local canmnt=${!var_canmnt}
+
+ unset $var_mntpt
+ unset $var_canmnt
+ do_facet $facet umount $mnt
+ do_facet $facet $ZFS set mountpoint=$mntpt $ds
+ do_facet $facet $ZFS set canmount=$canmnt $ds
+ export_zpool $facet
+}
+
+mount_fstype() {
+ local facet=$1
+ local fstype=$(facet_fstype $facet)
+
+ mount_$fstype $facet
+}
+
+unmount_fstype() {
+ local facet=$1
+ local fstype=$(facet_fstype $facet)
+
+ unmount_$fstype $facet
+}
+
########
## MountConf setup
opts+=${L_GETIDENTITY:+" --param=mdt.identity_upcall=$L_GETIDENTITY"}
if [ $fstype == ldiskfs ]; then
+ # Check for wide striping
+ if [ $OSTCOUNT -gt 160 ]; then
+ MDSJOURNALSIZE=${MDSJOURNALSIZE:-4096}
+ fs_mkfs_opts+="-O large_xattr"
+ fi
+
fs_mkfs_opts+=${MDSJOURNALSIZE:+" -J size=$MDSJOURNALSIZE"}
if [ ! -z $EJOURNAL ]; then
fs_mkfs_opts+=${MDSJOURNALSIZE:+" device=$EJOURNAL"}
# empty lustre filesystem may have empty directories lost+found and .lustre
is_empty_fs() {
+ # exclude .lustre & lost+found
[ $(find $1 -maxdepth 1 -name lost+found -o -name .lustre -prune -o \
-print | wc -l) = 1 ] || return 1
[ ! -d $1/lost+found ] || is_empty_dir $1/lost+found || return 1
- [ ! -d $1/.lustre ] || is_empty_dir $1/.lustre || return 1
+ if [ $(lustre_version_code $SINGLEMDS) -gt $(version_code 2.4.0) ]; then
+ # exclude .lustre/fid (LU-2780)
+ [ $(find $1/.lustre -maxdepth 1 -name fid -prune -o \
+ -print | wc -l) = 1 ] || return 1
+ else
+ [ ! -d $1/.lustre ] || is_empty_dir $1/.lustre || return 1
+ fi
return 0
}
fi
if [ $(lower $OSD_TRACK_DECLARES_LBUG) == 'yes' ] ; then
- local facets="$(get_facets OST),$(get_facets MDS),mgs"
+ local facets=""
+ [ "$(facet_fstype ost1)" = "ldiskfs" ] &&
+ facets="$(get_facets OST)"
+ [ "$(facet_fstype mds1)" = "ldiskfs" ] &&
+ facets="$facets,$(get_facets MDS)"
+ [ "$(facet_fstype mgs)" = "ldiskfs" ] &&
+ facets="$facets,mgs"
local nodes="$(facets_hosts ${facets})"
if [ -n "$nodes" ] ; then
do_nodes $nodes "$LCTL set_param \
}
cleanup_mount () {
- local clients=${CLIENTS:-$HOSTNAME}
- local mntpt=$1
+ local clients=${CLIENTS:-$HOSTNAME}
+ local mntpt=$1
- zconf_umount_clients $clients $mntpt
+ zconf_umount_clients $clients $mntpt
}
cleanup_and_setup_lustre() {
if [ "$type" == ost ]; then
devs=$(get_osd_param $node "" mntdev)
else
- devs=$(do_node $node \
- "lctl get_param -n osd-*.$FSNAME-M*.mntdev")
+ devs=$(do_node $node $LCTL get_param -n osd-*.$FSNAME-M*.mntdev)
fi
for dev in $devs; do
case $dev in
# Get all of the server target devices.
get_svr_devs() {
- local i
+ local node
+ local i
- # MDT device
- MDTDEV=$(get_mnt_devs $(mdts_nodes) mdt)
+ # Master MDS parameters used by lfsck
+ MDTNODE=$(facet_active_host $SINGLEMDS)
+ MDTDEV=$(echo $(get_mnt_devs $MDTNODE mdt) | awk '{print $1}')
- # OST devices
- i=0
- for node in $(osts_nodes); do
- OSTDEVS[i]=$(get_mnt_devs $node ost)
- i=$((i + 1))
- done
+ # MDT devices
+ i=0
+ for node in $(mdts_nodes); do
+ MDTDEVS[i]=$(get_mnt_devs $node mdt)
+ i=$((i + 1))
+ done
+
+ # OST devices
+ i=0
+ for node in $(osts_nodes); do
+ OSTDEVS[i]=$(get_mnt_devs $node ost)
+ i=$((i + 1))
+ done
}
# Run e2fsck on MDT or OST device.
run_e2fsck() {
- local node=$1
- local target_dev=$2
- local extra_opts=$3
+ local node=$1
+ local target_dev=$2
+ local extra_opts=$3
+ local cmd="$E2FSCK -d -v -t -t -f $extra_opts $target_dev"
+ local log=$TMP/e2fsck.log
+ local rc=0
- df > /dev/null # update statfs data on disk
- local cmd="$E2FSCK -d -v -t -t -f $extra_opts $target_dev"
- echo $cmd
- local rc=0
- do_node $node $cmd || rc=$?
- [ $rc -le $FSCK_MAX_ERR ] || \
- error "$cmd returned $rc, should be <= $FSCK_MAX_ERR"
- return 0
+ echo $cmd
+ do_node $node $cmd 2>&1 | tee $log
+ rc=${PIPESTATUS[0]}
+ if [ -n "$(grep "DNE mode isn't supported" $log)" ]; then
+ rm -f $log
+ if [ $MDSCOUNT -gt 1 ]; then
+ skip "DNE mode isn't supported!"
+ cleanupall
+ exit_status
+ else
+ error "It's not DNE mode."
+ fi
+ fi
+ rm -f $log
+
+ [ $rc -le $FSCK_MAX_ERR ] ||
+ error "$cmd returned $rc, should be <= $FSCK_MAX_ERR"
+
+ return 0
+}
+
+#
+# Run resize2fs on MDT or OST device.
+#
+run_resize2fs() {
+ local facet=$1
+ local device=$2
+ local size=$3
+ shift 3
+ local opts="$@"
+
+ do_facet $facet "$RESIZE2FS $opts $device $size"
}
# verify a directory is shared among nodes.
check_shared_dir() {
local dir=$1
+ local list=${2:-$(comma_list $(nodes_list))}
[ -z "$dir" ] && return 1
- do_rpc_nodes "$(comma_list $(nodes_list))" check_logdir $dir
- check_write_access $dir || return 1
+ do_rpc_nodes "$list" check_logdir $dir
+ check_write_access $dir "$list" || return 1
return 0
}
# Run e2fsck on MDT and OST(s) to generate databases used for lfsck.
generate_db() {
- local i
- local ostidx
- local dev
+ local i
+ local ostidx
+ local dev
+ local node
[[ $(lustre_version_code $SINGLEMDS) -ne $(version_code 2.2.0) ]] ||
{ skip "Lustre 2.2.0 lacks the patch for LU-1255"; exit 0; }
- check_shared_dir $SHARED_DIRECTORY ||
- error "$SHARED_DIRECTORY isn't a shared directory"
+ check_shared_dir $SHARED_DIRECTORY ||
+ error "$SHARED_DIRECTORY isn't a shared directory"
- export MDSDB=$SHARED_DIRECTORY/mdsdb
- export OSTDB=$SHARED_DIRECTORY/ostdb
+ export MDSDB=$SHARED_DIRECTORY/mdsdb
+ export OSTDB=$SHARED_DIRECTORY/ostdb
- [ $MDSCOUNT -eq 1 ] || error "CMD is not supported"
-
- run_e2fsck $(mdts_nodes) $MDTDEV "-n --mdsdb $MDSDB"
+ # DNE is not supported, so when running e2fsck on a DNE filesystem,
+ # we only pass master MDS parameters.
+ run_e2fsck $MDTNODE $MDTDEV "-n --mdsdb $MDSDB"
i=0
ostidx=0
fi
#Run lfsck
echo $cmd
- do_node $node $cmd || rc=$?
+ do_node $client $cmd || rc=$?
#Umount if necessary
if ! $mounted; then
zconf_umount $client $MOUNT ||
case $type in
MGS ) list="$list $name";;
- MDS|OST ) local count=${type}COUNT
+ MDS|OST|AGT ) local count=${type}COUNT
for ((i=1; i<=${!count}; i++)) do
list="$list ${name}$i"
done;;
pause_bulk() {
#define OBD_FAIL_OST_BRW_PAUSE_BULK 0x214
- RC=0
- do_facet ost1 lctl set_param fail_loc=0x214
- do_facet client "$1" || RC=$?
- do_facet client "sync"
- do_facet ost1 lctl set_param fail_loc=0
- return $RC
+ RC=0
+
+ local timeout=${2:-0}
+ # default is (obd_timeout / 4) if unspecified
+ echo "timeout is $timeout/$2"
+ do_facet ost1 lctl set_param fail_val=$timeout fail_loc=0x80000214
+ do_facet client "$1" || RC=$?
+ do_facet client "sync"
+ do_facet ost1 lctl set_param fail_loc=0
+ return $RC
}
drop_ldlm_cancel() {
}
set_nodes_failloc () {
- do_nodes $(comma_list $1) lctl set_param fail_loc=$2
+ do_nodes $(comma_list $1) lctl set_param fail_val=0 fail_loc=$2
}
cancel_lru_locks() {
if [ -z "$*" ]; then
echo "error() without useful message, please fix" > $LOGDIR/err
else
- echo "$@" > $LOGDIR/err
+ if [[ `echo $TYPE | grep ^IGNORE` ]]; then
+ echo "$@" > $LOGDIR/ignore
+ else
+ echo "$@" > $LOGDIR/err
+ fi
fi
}
# use only if we are ignoring failures for this test, bugno required.
# (like ALWAYS_EXCEPT, but run the test and ignore the results.)
-# e.g. error_ignore 5494 "your message"
+# e.g. error_ignore bz5494 "your message" or
+# error_ignore LU-5494 "your message"
error_ignore() {
- local TYPE="IGNORE (bz$1)"
+ local TYPE="IGNORE ($1)"
shift
error_noexit "$@"
}
build_test_filter() {
EXCEPT="$EXCEPT $(testslist_filter)"
- [ "$ONLY" ] && log "only running test `echo $ONLY`"
- for O in $ONLY; do
- eval ONLY_${O}=true
- done
+ for O in $ONLY; do
+ if [[ $O = [0-9]*-[0-9]* ]]; then
+ for num in $(seq $(echo $O | tr '-' ' ')); do
+ eval ONLY_$num=true
+ done
+ else
+ eval ONLY_${O}=true
+ fi
+ done
+
[ "$EXCEPT$ALWAYS_EXCEPT" ] && \
log "excepting tests: `echo $EXCEPT $ALWAYS_EXCEPT`"
[ "$EXCEPT_SLOW" ] && \
}
#
-# Run a single test function and cleanup after it.
+# Run a single test function and cleanup after it.
#
# This function should be run in a subshell so the test func can
# exit() without stopping the whole script.
#
run_one() {
- local testnum=$1
- local message=$2
- tfile=f.${TESTSUITE}.${testnum}
- export tdir=d0.${TESTSUITE}/d${base}
- export TESTNAME=test_$testnum
- local SAVE_UMASK=`umask`
- umask 0022
-
- banner "test $testnum: $message"
- test_${testnum} || error "test_$testnum failed with $?"
- cd $SAVE_PWD
- reset_fail_loc
- check_grant ${testnum} || error "check_grant $testnum failed with $?"
- check_catastrophe || error "LBUG/LASSERT detected"
+ local testnum=$1
+ local message=$2
+ export tfile=f${testnum}.${TESTSUITE}
+ export tdir=d${testnum}.${TESTSUITE}
+ export TESTNAME=test_$testnum
+ local SAVE_UMASK=`umask`
+ umask 0022
+
+ banner "test $testnum: $message"
+ test_${testnum} || error "test_$testnum failed with $?"
+ cd $SAVE_PWD
+ reset_fail_loc
+ check_grant ${testnum} || error "check_grant $testnum failed with $?"
+ check_catastrophe || error "LBUG/LASSERT detected"
if [ "$PARALLEL" != "yes" ]; then
ps auxww | grep -v grep | grep -q multiop &&
error "multiop still running"
fi
- unset TESTNAME
- unset tdir
- umask $SAVE_UMASK
- return 0
+ unset TESTNAME
+ unset tdir
+ unset tfile
+ umask $SAVE_UMASK
+ return 0
}
#
local name=${TESTSUITE}.test_${1}.test_log.$(hostname -s).log
local test_log=$LOGDIR/$name
rm -rf $LOGDIR/err
+ rm -rf $LOGDIR/ignore
rm -rf $LOGDIR/skip
local SAVE_UMASK=`umask`
umask 0022
(run_one $1 "$2") 2>&1 | tee -i $test_log
local RC=${PIPESTATUS[0]}
- [ $RC -ne 0 ] && [ ! -f $LOGDIR/err ] && \
+ [ $RC -ne 0 ] && [ ! -f $LOGDIR/err ] &&
echo "test_$1 returned $RC" | tee $LOGDIR/err
duration=$((`date +%s` - $BEFORE))
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" "$TEST_ERROR"
+ if [[ "$TEST_STATUS" != "SKIP" ]] && [[ -f $TF_SKIP ]]; then
+ rm -f $TF_SKIP
+ fi
+
if [ -f $LOGDIR/err ]; then
$FAIL_ON_ERROR && exit $RC
fi
echo -n $(facets_nodes $(get_facets OST))
}
+# Get all of the active AGT (HSM agent) nodes.
+agts_nodes () {
+ echo -n $(facets_nodes $(get_facets AGT))
+}
+
# Get all of the client nodes and active server nodes.
nodes_list () {
local nodes=$HOSTNAME
mdsrate_cleanup () {
if [ -d $4 ]; then
- mpi_run -np $1 ${MACHINEFILE_OPTION} $2 ${MDSRATE} --unlink \
+ mpi_run ${MACHINEFILE_OPTION} $2 -np $1 ${MDSRATE} --unlink \
--nfiles $3 --dir $4 --filefmt $5 $6
rmdir $4
fi
########################
-convert_facet2label() {
+convert_facet2label() {
local facet=$1
if [ x$facet = xost ]; then
if [ -n ${!varsvc} ]; then
echo ${!varsvc}
- else
+ else
error "No lablel for $facet!"
fi
}
}
_wait_osc_import_state() {
- local facet=$1
- local ost_facet=$2
- local expected=$3
- local ost=$(get_osc_import_name $facet $ost_facet)
-
+ local facet=$1
+ local ost_facet=$2
+ local expected=$3
+ local ost=$(get_osc_import_name $facet $ost_facet)
local param="osc.${ost}.ost_server_uuid"
+ local params=$param
local i=0
- # 1. wait the deadline of client 1st request (it could be skipped)
- # 2. wait the deadline of client 2nd request
- local maxtime=$(( 2 * $(request_timeout $facet)))
-
- #During setup time, the osc might not be setup, it need wait
- #until list_param can return valid value. And also if there
- #are mulitple osc entries we should list all of them before
- #go to wait.
- local params=$($LCTL list_param $param 2>/dev/null || true)
- while [ -z "$params" ]; do
- if [ $i -ge $maxtime ]; then
- echo "can't get $param by list_param in $maxtime secs"
- if [[ $facet != client* ]]; then
- echo "Go with $param directly"
- params=$param
- break
- else
+ # 1. wait the deadline of client 1st request (it could be skipped)
+ # 2. wait the deadline of client 2nd request
+ local maxtime=$(( 2 * $(request_timeout $facet)))
+
+ if [[ $facet == client* ]]; then
+ # During setup time, the osc might not be setup, it need wait
+ # until list_param can return valid value. And also if there
+ # are mulitple osc entries we should list all of them before
+ # go to wait.
+ params=$($LCTL list_param $param 2>/dev/null || true)
+ while [ -z "$params" ]; do
+ if [ $i -ge $maxtime ]; then
+ echo "can't get $param in $maxtime secs"
return 1
fi
- fi
- sleep 1
- i=$((i + 1))
- params=$($LCTL list_param $param 2>/dev/null || true)
- done
-
+ sleep 1
+ i=$((i + 1))
+ params=$($LCTL list_param $param 2>/dev/null || true)
+ done
+ fi
if ! do_rpc_nodes "$(facet_active_host $facet)" \
wait_import_state $expected "$params" $maxtime; then
error "import is not in ${expected} state"
}
check_write_access() {
- local dir=$1
- local node
- local file
+ local dir=$1
+ local list=${2:-$(comma_list $(nodes_list))}
+ local node
+ local file
- for node in $(nodes_list); do
- file=$dir/check_file.$(short_hostname $node)
- if [[ ! -f "$file" ]]; then
- # Logdir not accessible/writable from this node.
- return 1
- fi
- rm -f $file || return 1
- done
- return 0
+ for node in ${list//,/ }; do
+ file=$dir/check_file.$(short_nodename $node)
+ if [[ ! -f "$file" ]]; then
+ # Logdir not accessible/writable from this node.
+ return 1
+ fi
+ rm -f $file || return 1
+ done
+ return 0
}
init_logging() {
fi
mount -t $(facet_fstype $facet) $opts $ostdev $mntpt ||
return $?
- rc=0;
+ rc=0
for i in $objids; do
rm $mntpt/O/$group/d$((i % 32))/$i || { rc=$?; break; }
done
fi
mount -t $(facet_fstype $facet) $opts $mdtdev $mntpt ||
return $?
- rc=0;
+ rc=0
for f in $files; do
rm $mntpt/ROOT/$f || { rc=$?; break; }
done
$LCTL get_param -n osc.*.kbytesavail | sort -n | head -n1
}
+#
+# Get the available size (KB) of a given obd target.
+#
+get_obd_size() {
+ local facet=$1
+ local obd=$2
+ local size
+
+ [[ $facet != client ]] || return 0
+
+ size=$(do_facet $facet $LCTL get_param -n *.$obd.kbytesavail | head -n1)
+ echo -n $size
+}
+
+#
+# Get the page size (bytes) on a given facet node.
+#
+get_page_size() {
+ local facet=$1
+ local size
+
+ size=$(do_facet $facet getconf PAGE_SIZE)
+ [[ ${PIPESTATUS[0]} = 0 && -n "$size" ]] || size=4096
+ echo -n $size
+}
+
+#
+# Get the block count of the filesystem.
+#
+get_block_count() {
+ local facet=$1
+ local device=$2
+ local count
+
+ count=$(do_facet $facet "$DUMPE2FS -h $device 2>&1" |
+ awk '/^Block count:/ {print $3}')
+ echo -n $count
+}
+
# Get the block size of the filesystem.
get_block_size() {
local facet=$1
local mds_dev=$(mdsdevname ${SINGLEMDS//mds/})
- do_facet $SINGLEMDS "$DUMPE2FS -h $mds_dev 2>&1 | grep -q large_xattr"
+ do_facet $SINGLEMDS "$DUMPE2FS -h $mds_dev 2>&1 |
+ grep -E -q '(ea_inode|large_xattr)'"
return ${PIPESTATUS[0]}
}
}
reformat_external_journal() {
+ local facet=$1
+
if [ ! -z ${EJOURNAL} ]; then
- local rcmd="do_facet ${SINGLEMDS}"
+ local rcmd="do_facet $facet"
- echo "reformat external journal on ${SINGLEMDS}:${EJOURNAL}"
+ echo "reformat external journal on $facet:${EJOURNAL}"
${rcmd} mke2fs -O journal_dev ${EJOURNAL} || return 1
fi
}
# MDT file-level backup/restore
mds_backup_restore() {
- local devname=$(mdsdevname ${SINGLEMDS//mds/})
+ local facet=$1
+ local igif=$2
+ local devname=$(mdsdevname $(facet_number $facet))
local mntpt=$(facet_mntpt brpt)
- local rcmd="do_facet ${SINGLEMDS}"
+ local rcmd="do_facet $facet"
local metaea=${TMP}/backup_restore.ea
local metadata=${TMP}/backup_restore.tgz
local opts=${MDS_MOUNT_OPTS}
- local svc=${SINGLEMDS}_svc
- local igif=$1
+ local svc=${facet}_svc
if ! ${rcmd} test -b ${devname}; then
opts=$(csa_add "$opts" -o loop)
fi
- echo "file-level backup/restore on ${SINGLEMDS}:${devname}"
+ echo "file-level backup/restore on $facet:${devname}"
# step 1: build mount point
${rcmd} mkdir -p $mntpt
# step 6: umount
${rcmd} umount -d $mntpt || return 4
# step 7: reformat external journal if needed
- reformat_external_journal || return 5
+ reformat_external_journal $facet || return 5
# step 8: reformat dev
echo "reformat new device"
- add ${SINGLEMDS} $(mkfs_opts ${SINGLEMDS} ${devname}) --backfstype \
- ldiskfs --reformat ${devname} $(mdsvdevname 1) > /dev/null ||
- exit 6
+ add $facet $(mkfs_opts $facet ${devname}) --backfstype ldiskfs \
+ --reformat ${devname} $(mdsvdevname $(facet_number $facet)) \
+ > /dev/null || exit 6
# step 9: mount dev
${rcmd} mount -t ldiskfs $opts $devname $mntpt || return 7
# step 10: restore metadata
# remove OI files
mds_remove_ois() {
- local devname=$(mdsdevname ${SINGLEMDS//mds/})
+ local facet=$1
+ local idx=$2
+ local devname=$(mdsdevname $(facet_number $facet))
local mntpt=$(facet_mntpt brpt)
- local rcmd="do_facet ${SINGLEMDS}"
- local idx=$1
+ local rcmd="do_facet $facet"
local opts=${MDS_MOUNT_OPTS}
if ! ${rcmd} test -b ${devname}; then
opts=$(csa_add "$opts" -o loop)
fi
- echo "remove OI files: idx=${idx}"
+ echo "removing OI files on $facet: idx=${idx}"
# step 1: build mount point
${rcmd} mkdir -p $mntpt
fi
return $rc
}
+
+# find the smallest and not in use file descriptor
+free_fd()
+{
+ local max_fd=$(ulimit -n)
+ local fd=3
+ while [[ $fd -le $max_fd && -e /proc/self/fd/$fd ]]; do
+ ((++fd))
+ done
+ [ $fd -lt $max_fd ] || error "finding free file descriptor failed"
+ echo $fd
+}