}
init_test_env() {
- export LUSTRE=`absolute_path $LUSTRE`
- export TESTSUITE=`basename $0 .sh`
- export TEST_FAILED=false
- export FAIL_ON_SKIP_ENV=${FAIL_ON_SKIP_ENV:-false}
-
- export MKE2FS=${MKE2FS:-mke2fs}
- export DEBUGFS=${DEBUGFS:-debugfs}
- export TUNE2FS=${TUNE2FS:-tune2fs}
- export E2LABEL=${E2LABEL:-e2label}
- export DUMPE2FS=${DUMPE2FS:-dumpe2fs}
- export E2FSCK=${E2FSCK:-e2fsck}
- export LFSCK_BIN=${LFSCK_BIN:-lfsck}
+ export LUSTRE=$(absolute_path $LUSTRE)
+ export TESTSUITE=$(basename $0 .sh)
+ export TEST_FAILED=false
+ export FAIL_ON_SKIP_ENV=${FAIL_ON_SKIP_ENV:-false}
+ export RPC_MODE=${RPC_MODE:-false}
+
+ export MKE2FS=$MKE2FS
+ if [ -z "$MKE2FS" ]; then
+ if which mkfs.ldiskfs >/dev/null 2>&1; then
+ export MKE2FS=mkfs.ldiskfs
+ else
+ export MKE2FS=mke2fs
+ fi
+ fi
+
+ export DEBUGFS=$DEBUGFS
+ if [ -z "$DEBUGFS" ]; then
+ if which debugfs.ldiskfs >/dev/null 2>&1; then
+ export DEBUGFS=debugfs.ldiskfs
+ else
+ export DEBUGFS=debugfs
+ fi
+ fi
+
+ export TUNE2FS=$TUNE2FS
+ if [ -z "$TUNE2FS" ]; then
+ if which tunefs.ldiskfs >/dev/null 2>&1; then
+ export TUNE2FS=tunefs.ldiskfs
+ else
+ export TUNE2FS=tune2fs
+ fi
+ fi
+
+ export E2LABEL=$E2LABEL
+ if [ -z "$E2LABEL" ]; then
+ if which label.ldiskfs >/dev/null 2>&1; then
+ export E2LABEL=label.ldiskfs
+ else
+ export E2LABEL=e2label
+ fi
+ fi
+ export DUMPE2FS=$DUMPE2FS
+ if [ -z "$DUMPE2FS" ]; then
+ if which dumpfs.ldiskfs >/dev/null 2>&1; then
+ export DUMPE2FS=dumpfs.ldiskfs
+ else
+ export DUMPE2FS=dumpe2fs
+ fi
+ fi
+
+ export E2FSCK=$E2FSCK
+ if [ -z "$E2FSCK" ]; then
+ if which fsck.ldiskfs >/dev/null 2>&1; then
+ export E2FSCK=fsck.ldiskfs
+ else
+ export E2FSCK=e2fsck
+ fi
+ fi
+
+ export LFSCK_BIN=${LFSCK_BIN:-lfsck}
export LFSCK_ALWAYS=${LFSCK_ALWAYS:-"no"} # check fs after each test suite
export FSCK_MAX_ERR=4 # File system errors left uncorrected
fi
fi
export LL_DECODE_FILTER_FID=${LL_DECODE_FILTER_FID:-"$LUSTRE/utils/ll_decode_filter_fid"}
- [ ! -f "$LL_DECODE_FILTER_FID" ] && export LL_DECODE_FILTER_FID=$(which ll_decode_filter_fid)
+ [ ! -f "$LL_DECODE_FILTER_FID" ] && export LL_DECODE_FILTER_FID="ll_decode_filter_fid"
export MKFS=${MKFS:-"$LUSTRE/utils/mkfs.lustre"}
- [ ! -f "$MKFS" ] && export MKFS=$(which mkfs.lustre)
+ [ ! -f "$MKFS" ] && export MKFS="mkfs.lustre"
export TUNEFS=${TUNEFS:-"$LUSTRE/utils/tunefs.lustre"}
- [ ! -f "$TUNEFS" ] && export TUNEFS=$(which tunefs.lustre)
+ [ ! -f "$TUNEFS" ] && export TUNEFS="tunefs.lustre"
export CHECKSTAT="${CHECKSTAT:-"checkstat -v"} "
export LUSTRE_RMMOD=${LUSTRE_RMMOD:-$LUSTRE/scripts/lustre_rmmod}
[ ! -f "$LUSTRE_RMMOD" ] &&
IDENTITY_UPCALL=false
;;
esac
+
export LOAD_MODULES_REMOTE=${LOAD_MODULES_REMOTE:-false}
# Paths on remote nodes, if different
export RLUSTRE=${RLUSTRE:-$LUSTRE}
export RPWD=${RPWD:-$PWD}
export I_MOUNTED=${I_MOUNTED:-"no"}
- if [ ! -f /lib/modules/$(uname -r)/kernel/fs/lustre/mds.ko -a \
- ! -f /lib/modules/$(uname -r)/updates/kernel/fs/lustre/mds.ko -a \
- ! -f `dirname $0`/../mds/mds.ko ]; then
+ if [ ! -f /lib/modules/$(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
shift $((OPTIND - 1))
ONLY=${ONLY:-$*}
- # print the durations of each test if "true"
- DDETAILS=${DDETAILS:-false}
- [ "$TESTSUITELOG" ] && rm -f $TESTSUITELOG || true
- rm -f $TMP/*active
+ # print the durations of each test if "true"
+ DDETAILS=${DDETAILS:-false}
+ [ "$TESTSUITELOG" ] && rm -f $TESTSUITELOG || true
+ if ! $RPC_MODE; then
+ rm -f $TMP/*active
+ fi
}
check_cpt_number() {
# split arguments like "1.8.6-wc3" into "1", "8", "6", "wc3"
eval set -- $(tr "[:punct:]" " " <<< $*)
- echo -n $((($1 << 16) | ($2 << 8) | $3))
+ echo -n "$((($1 << 16) | ($2 << 8) | $3))"
}
export LINUX_VERSION=$(uname -r | sed -e "s/[-.]/ /3" -e "s/ .*//")
optvar="MODOPTS_$(basename $module | tr a-z A-Z)"
eval set -- \$$optvar
if [ $# -eq 0 -a -n "$MODPROBECONF" ]; then
- # Nothing in $MODOPTS_<MODULE>; try modprobe.conf
- set -- $(grep -P "^options\\s+${BASE}" $MODPROBECONF)
- # Get rid of "options $module"
- (($# > 0)) && shift 2
-
- # Ensure we have accept=all for lnet
- if [ $(basename $module) = lnet ]; then
- # OK, this is a bit wordy...
- local arg accept_all_present=false
- for arg in "$@"; do
- [ "$arg" = accept=all ] && accept_all_present=true
- done
- $accept_all_present || set -- "$@" accept=all
- fi
+ # Nothing in $MODOPTS_<MODULE>; try modprobe.conf
+ local opt
+ opt=$(awk -v var="^options $BASE" '$0 ~ var \
+ {gsub("'"options $BASE"'",""); print}' $MODPROBECONF)
+ set -- $(echo -n $opt)
+
+ # Ensure we have accept=all for lnet
+ if [ $(basename $module) = lnet ]; then
+ # OK, this is a bit wordy...
+ local arg accept_all_present=false
+
+ for arg in "$@"; do
+ [ "$arg" = accept=all ] && \
+ accept_all_present=true
+ done
+ $accept_all_present || set -- "$@" accept=all
+ fi
+ export $optvar="$*"
fi
fi
}
load_modules_local() {
- if [ "$USE_OFD" == yes ]; then
- if module_loaded obdfilter; then
- if ! $LUSTRE_RMMOD ldiskfs; then
- echo "$HOSTNAME may still be using obdfilter.ko"
- return 1
- fi
- fi
- else
- if module_loaded ofd; then
- if ! $LUSTRE_RMMOD ldiskfs; then
- echo "$HOSTNAME may still be using ofd.ko"
- return 1
- fi
- fi
- fi
-
if [ -n "$MODPROBE" ]; then
# use modprobe
echo "Using modprobe to load modules"
fi
fi
- local ncpts=0
# if there is only one CPU core, libcfs can only create one partition
# if there is more than 4 CPU cores, libcfs should create multiple CPU
# partitions. So we just force libcfs to create 2 partitions for
if [ $ncpus -le 4 ] && [ $ncpus -gt 1 ]; then
# force to enable multiple CPU partitions
echo "Force libcfs to create 2 CPU partitions"
- ncpts=2
+ MODOPTS_LIBCFS="cpu_npartitions=2 $MODOPTS_LIBCFS"
else
echo "libcfs will create CPU partition based on online CPUs"
fi
- load_module ../libcfs/libcfs/libcfs cpu_npartitions=$ncpts
+ load_module ../libcfs/libcfs/libcfs
[ "$PTLDEBUG" ] && lctl set_param debug="$PTLDEBUG"
[ "$SUBSYSTEM" ] && lctl set_param subsystem_debug="${SUBSYSTEM# }"
grep -q -w jbd2 $SYMLIST || { modprobe jbd2 2>/dev/null || true; }
[ "$LQUOTA" != "no" ] && load_module quota/lquota $LQUOTAOPTS
if [[ $(node_fstypes $HOSTNAME) == *zfs* ]]; then
+ modprobe zfs
load_module osd-zfs/osd_zfs
fi
load_module mgs/mgs
- load_module mds/mds
load_module mdd/mdd
if [[ $(node_fstypes $HOSTNAME) == *ldiskfs* ]]; then
#
load_module osd-ldiskfs/osd_ldiskfs
fi
load_module mdt/mdt
- load_module cmm/cmm
load_module ost/ost
load_module lod/lod
- if [ "$USE_OFD" == yes ]; then
- load_module ofd/ofd
- else
- load_module obdfilter/obdfilter
- fi
+ load_module osp/osp
+ load_module ofd/ofd
+ load_module osp/osp
fi
# bug 19124
# load modules on remote nodes optionally
# lustre-tests have to be installed on these nodes
- if $LOAD_MODULES_REMOTE ; then
+ if $LOAD_MODULES_REMOTE; then
local list=$(comma_list $(remote_nodes_list))
- echo loading modules on $list
- do_rpc_nodes $list load_modules_local
+ if [ -n "$list" ]; then
+ echo "loading modules on: '$list'"
+ do_rpc_nodes "$list" load_modules_local
+ fi
fi
}
$LUSTRE_RMMOD ldiskfs || return 2
- if $LOAD_MODULES_REMOTE ; then
+ if $LOAD_MODULES_REMOTE; then
local list=$(comma_list $(remote_nodes_list))
- if [ ! -z $list ]; then
- echo unloading modules on $list
- do_rpc_nodes $list $LUSTRE_RMMOD ldiskfs
- do_rpc_nodes $list check_mem_leak
+ if [ -n "$list" ]; then
+ echo "unloading modules on: '$list'"
+ do_rpc_nodes "$list" $LUSTRE_RMMOD ldiskfs
+ do_rpc_nodes "$list" check_mem_leak
fi
fi
return 0
}
+fs_log_size() {
+ local facet=${1:-$SINGLEMDS}
+ local fstype=$(facet_fstype $facet)
+ local size=0
+ case $fstype in
+ ldiskfs) size=50;; # largest seen is 44, leave some headroom
+ zfs) size=256;;
+ esac
+
+ echo -n $size
+}
+
check_gss_daemon_nodes() {
local list=$1
dname=$2
facet_type() {
local facet=$1
- echo -n $facet | sed -e 's/^fs[0-9]\+//' -e 's/[0-9]\+//' |
+ echo -n $facet | sed -e 's/^fs[0-9]\+//' -e 's/[0-9_]\+//' |
tr '[:lower:]' '[:upper:]'
}
return
fi
+ if [[ $facet == mgs ]] && combined_mgs_mds; then
+ facet_fstype mds1
+ return
+ fi
+
return 1
}
}
#
-# This and set_obdfilter_param() shall be used to access OSD parameters
+# This and set_osd_param() shall be used to access OSD parameters
# once existed under "obdfilter":
#
# mntdev
# read_cache_enable
# writethrough_cache_enable
#
-get_obdfilter_param() {
+get_osd_param() {
local nodes=$1
local device=${2:-$FSNAME-OST*}
local name=$3
osd-*.$device.$name 2>&1" | grep -v 'Found no match'
}
-set_obdfilter_param() {
+set_osd_param() {
local nodes=$1
local device=${2:-$FSNAME-OST*}
local name=$3
}
set_default_debug_nodes () {
- local nodes=$1
+ local nodes="$1"
- if [[ ,$nodes, = *,$HOSTNAME,* ]]; then
- nodes=$(exclude_items_from_list "$nodes" "$HOSTNAME")
- set_default_debug
- fi
+ if [[ ,$nodes, = *,$HOSTNAME,* ]]; then
+ nodes=$(exclude_items_from_list "$nodes" "$HOSTNAME")
+ set_default_debug
+ fi
- [[ -n $nodes ]] && do_rpc_nodes $nodes set_default_debug \
- \\\"$PTLDEBUG\\\" \\\"$SUBSYSTEM\\\" $DEBUG_SIZE || true
+ do_rpc_nodes "$nodes" set_default_debug \
+ \\\"$PTLDEBUG\\\" \\\"$SUBSYSTEM\\\" $DEBUG_SIZE || true
}
set_default_debug_facet () {
# XXX This function is kept for interoperability with old server (< 2.3.50),
# it should be removed whenever we drop the interoperability for such
# server.
-quota_type () {
- local fsname=${1:-$FSNAME}
- local rc=0
- do_facet mgs lctl get_param mdd.${fsname}-MDT*.quota_type || rc=$?
- do_nodes $(comma_list $(osts_nodes)) \
- lctl get_param obdfilter.${fsname}-OST*.quota_type || rc=$?
- return $rc
+quota_type() {
+ local fsname=${1:-$FSNAME}
+ local rc=0
+ do_facet $SINGLEMDS lctl get_param mdd.${fsname}-MDT*.quota_type ||
+ rc=$?
+ do_nodes $(comma_list $(osts_nodes)) \
+ lctl get_param obdfilter.${fsname}-OST*.quota_type || rc=$?
+ return $rc
}
# XXX This function is kept for interoperability with old server (< 2.3.50),
local quota_usrs=$QUOTA_USERS
# get_filesystem_size
- local disksz=$(lfs df $mntpt | grep "filesystem summary:" | awk '{print $3}')
+ local disksz=$(lfs_df $mntpt | grep "summary" | awk '{print $2}')
local blk_soft=$((disksz + 1024))
local blk_hard=$((blk_soft + blk_soft / 20)) # Go 5% over
- local Inodes=$(lfs df -i $mntpt | grep "filesystem summary:" | awk '{print $3}')
+ local Inodes=$(lfs_df -i $mntpt | grep "summary" | awk '{print $2}')
local i_soft=$Inodes
local i_hard=$((i_soft + i_soft / 20))
mdt_quota_type() {
local varsvc=${SINGLEMDS}_svc
do_facet $SINGLEMDS $LCTL get_param -n \
- osd-$FSTYPE.${!varsvc}.quota_slave.enabled
+ osd-$(facet_fstype $SINGLEMDS).${!varsvc}.quota_slave.enabled
}
# get ost quota type
# All OSTs should have same quota type
local varsvc=ost1_svc
do_facet ost1 $LCTL get_param -n \
- osd-$FSTYPE.${!varsvc}.quota_slave.enabled
+ osd-$(facet_fstype ost1).${!varsvc}.quota_slave.enabled
}
# restore old quota type settings
fi
}
+# Handle the case when there is a space in the lfs df
+# "filesystem summary" line the same as when there is no space.
+# This will allow fixing the "lfs df" summary line in the future.
+lfs_df() {
+ $LFS df $* | sed -e 's/filesystem /filesystem_/'
+}
+
setup_quota(){
if [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.3.50) ]; then
setup_quota_old $1
return
fi
- # XXX remove it once all quota code landed
- echo "skip quota setup"
- return
-
local mntpt=$1
# save old quota type & set new quota type
local quota_usrs=$QUOTA_USERS
# get_filesystem_size
- local disksz=$(lfs df $mntpt | grep "filesystem summary:" |
- awk '{print $3}')
+ local disksz=$(lfs_df $mntpt | grep "summary" | awk '{print $2}')
local blk_soft=$((disksz + 1024))
local blk_hard=$((blk_soft + blk_soft / 20)) # Go 5% over
- local inodes=$(lfs df -i $mntpt | grep "filesystem summary:" |
- awk '{print $3}')
+ local inodes=$(lfs_df -i $mntpt | grep "summary" | awk '{print $2}')
local i_soft=$inodes
local i_hard=$((i_soft + i_soft / 20))
}
check_progs_installed () {
- local nodes=$1
- shift
+ local nodes=$1
+ shift
- do_rpc_nodes $nodes _check_progs_installed $@
+ do_rpc_nodes "$nodes" _check_progs_installed $@
}
# recovery-scale functions
TESTNAME=$TESTNAME \
DBENCH_LIB=$DBENCH_LIB \
DBENCH_SRC=$DBENCH_SRC \
+LFS=$LFS \
run_${load}.sh" &
local ppid=$!
log "Started client load: ${load} on $client"
check_mem_leak || exit 204
- [ "`lctl dl 2> /dev/null | wc -l`" -gt 0 ] && lctl dl && \
- echo "$0: lustre didn't clean up..." 1>&2 && return 202 || true
+ [ "`lctl dl 2> /dev/null | wc -l`" -gt 0 ] && lctl dl &&
+ echo "$TESTSUITE: lustre didn't clean up..." 1>&2 &&
+ return 202 || true
- if module_loaded lnet || module_loaded libcfs; then
- echo "$0: modules still loaded..." 1>&2
- /sbin/lsmod 1>&2
- return 203
- fi
- return 0
+ if module_loaded lnet || module_loaded libcfs; then
+ echo "$TESTSUITE: modules still loaded..." 1>&2
+ /sbin/lsmod 1>&2
+ return 203
+ fi
+ return 0
}
wait_update () {
fi
echo affected facets: $facets
- # we can use "for" here because we are waiting the slowest
- for facet in ${facets//,/ }; do
- local var_svc=${facet}_svc
- local param="*.${!var_svc}.recovery_status"
+ # we can use "for" here because we are waiting the slowest
+ for facet in ${facets//,/ }; do
+ local var_svc=${facet}_svc
+ local param="*.${!var_svc}.recovery_status"
- local host=$(facet_active_host $facet)
- do_rpc_nodes $host _wait_recovery_complete $param $MAX
- done
+ local host=$(facet_active_host $facet)
+ do_rpc_nodes "$host" _wait_recovery_complete $param $MAX
+ done
}
wait_mds_ost_sync () {
- # just because recovery is done doesn't mean we've finished
- # orphan cleanup. Wait for llogs to get synchronized.
- echo "Waiting for orphan cleanup..."
- # MAX value includes time needed for MDS-OST reconnection
- local MAX=$(( TIMEOUT * 2 ))
- local WAIT=0
- while [ $WAIT -lt $MAX ]; do
- local -a sync=($(do_nodes $(comma_list $(osts_nodes)) \
- "$LCTL get_param -n obdfilter.*.mds_sync"))
- local con=1
- local i
- for ((i=0; i<${#sync[@]}; i++)); do
- [ ${sync[$i]} -eq 0 ] && continue
- # there is a not finished MDS-OST synchronization
- con=0
- break;
- done
- sleep 2 # increase waiting time and cover statfs cache
- [ ${con} -eq 1 ] && return 0
- echo "Waiting $WAIT secs for $facet mds-ost sync done."
- WAIT=$((WAIT + 2))
- done
- echo "$facet recovery not done in $MAX sec. $STATUS"
- return 1
+ # just because recovery is done doesn't mean we've finished
+ # orphan cleanup. Wait for llogs to get synchronized.
+ echo "Waiting for orphan cleanup..."
+ # MAX value includes time needed for MDS-OST reconnection
+ local MAX=$(( TIMEOUT * 2 ))
+ local WAIT=0
+ local new_wait=true
+ local list=$(comma_list $(mdts_nodes))
+ local cmd="$LCTL get_param -n osp.*osc*.old_sync_processed"
+ if ! do_facet $SINGLEMDS \
+ "$LCTL list_param osp.*osc*.old_sync_processed 2> /dev/null"
+ then
+ # old way, use mds_sync
+ new_wait=false
+ list=$(comma_list $(osts_nodes))
+ cmd="$LCTL get_param -n obdfilter.*.mds_sync"
+ fi
+ while [ $WAIT -lt $MAX ]; do
+ local -a sync=($(do_nodes $list "$cmd"))
+ local con=1
+ local i
+ for ((i=0; i<${#sync[@]}; i++)); do
+ if $new_wait; then
+ [ ${sync[$i]} -eq 1 ] && continue
+ else
+ [ ${sync[$i]} -eq 0 ] && continue
+ fi
+ # there is a not finished MDS-OST synchronization
+ con=0
+ break;
+ done
+ sleep 2 # increase waiting time and cover statfs cache
+ [ ${con} -eq 1 ] && return 0
+ echo "Waiting $WAIT secs for $facet mds-ost sync done."
+ WAIT=$((WAIT + 2))
+ done
+ echo "$facet recovery not done in $MAX sec. $STATUS"
+ return 1
}
wait_destroy_complete () {
}
facet_failover() {
- local facet=$1
- local sleep_time=$2
- local host=$(facet_active_host $facet)
-
- echo "Failing $facet on node $host"
-
- local affected=$(affected_facets $facet)
+ local facets=$1
+ local sleep_time=$2
+ local -a affecteds
+ local facet
+ local total=0
+ local index=0
+ local skip
+
+ #Because it will only get up facets, we need get affected
+ #facets before shutdown
+ #For HARD Failure mode, it needs make sure facets on the same
+ #HOST will only be shutdown and reboot once
+ for facet in ${facets//,/ }; do
+ local affected_facet
+ skip=0
+ #check whether facet has been included in other affected facets
+ for ((index=0; index<$total; index++)); do
+ [[ *,$facet,* == ,${affecteds[index]}, ]] && skip=1
+ done
- shutdown_facet $facet
+ if [ $skip -eq 0 ]; then
+ affecteds[$total]=$(affected_facets $facet)
+ total=$((total+1))
+ fi
+ done
- echo affected facets: $affected
+ for ((index=0; index<$total; index++)); do
+ facet=$(echo ${affecteds[index]} | tr -s " " | cut -d"," -f 1)
+ local host=$(facet_active_host $facet)
+ echo "Failing ${affecteds[index]} on $host"
+ shutdown_facet $facet
+ done
- [ -n "$sleep_time" ] && sleep $sleep_time
+ for ((index=0; index<$total; index++)); do
+ facet=$(echo ${affecteds[index]} | tr -s " " | cut -d"," -f 1)
+ echo reboot facets: ${affecteds[index]}
- reboot_facet $facet
+ reboot_facet $facet
- change_active $affected
+ change_active ${affecteds[index]}
- wait_for_facet $affected
- # start mgs first if it is affected
- if ! combined_mgs_mds && list_member $affected mgs; then
- mount_facet mgs || error "Restart of mgs failed"
- fi
- # FIXME; has to be changed to mount all facets concurrently
- affected=$(exclude_items_from_list $affected mgs)
- mount_facets $affected
+ wait_for_facet ${affecteds[index]}
+ # start mgs first if it is affected
+ if ! combined_mgs_mds &&
+ list_member ${affecteds[index]} mgs; then
+ mount_facet mgs || error "Restart of mgs failed"
+ fi
+ # FIXME; has to be changed to mount all facets concurrently
+ affected=$(exclude_items_from_list ${affecteds[index]} mgs)
+ echo mount facets: ${affecteds[index]}
+ mount_facets ${affecteds[index]}
+ done
}
obd_name() {
}
replay_barrier() {
- local facet=$1
- do_facet $facet "sync; sync; sync"
- df $MOUNT
+ local facet=$1
+ do_facet $facet "sync; sync; sync"
+ df $MOUNT
- # make sure there will be no seq change
- local clients=${CLIENTS:-$HOSTNAME}
- do_nodes $clients "f=${MOUNT}/fsa-\\\$(hostname); mcreate \\\$f; rm \\\$f"
+ # make sure there will be no seq change
+ local clients=${CLIENTS:-$HOSTNAME}
+ local f=fsa-\\\$\(hostname\)
+ do_nodes $clients "mcreate $MOUNT/$f; rm $MOUNT/$f"
+ do_nodes $clients "if [ -d $MOUNT2 ]; then mcreate $MOUNT2/$f; rm $MOUNT2/$f; fi"
- local svc=${facet}_svc
- do_facet $facet $LCTL --device %${!svc} notransno
- do_facet $facet $LCTL --device %${!svc} readonly
- do_facet $facet $LCTL mark "$facet REPLAY BARRIER on ${!svc}"
- $LCTL mark "local REPLAY BARRIER on ${!svc}"
+ local svc=${facet}_svc
+ do_facet $facet $LCTL --device ${!svc} notransno
+ do_facet $facet $LCTL --device ${!svc} readonly
+ do_facet $facet $LCTL mark "$facet REPLAY BARRIER on ${!svc}"
+ $LCTL mark "local REPLAY BARRIER on ${!svc}"
}
replay_barrier_nodf() {
- local facet=$1 echo running=${running}
- do_facet $facet "sync; sync; sync"
- local svc=${facet}_svc
- echo Replay barrier on ${!svc}
- do_facet $facet $LCTL --device %${!svc} notransno
- do_facet $facet $LCTL --device %${!svc} readonly
- do_facet $facet $LCTL mark "$facet REPLAY BARRIER on ${!svc}"
- $LCTL mark "local REPLAY BARRIER on ${!svc}"
+ local facet=$1 echo running=${running}
+ do_facet $facet "sync; sync; sync"
+ local svc=${facet}_svc
+ echo Replay barrier on ${!svc}
+ do_facet $facet $LCTL --device ${!svc} notransno
+ do_facet $facet $LCTL --device ${!svc} readonly
+ do_facet $facet $LCTL mark "$facet REPLAY BARRIER on ${!svc}"
+ $LCTL mark "local REPLAY BARRIER on ${!svc}"
}
replay_barrier_nosync() {
- local facet=$1 echo running=${running}
- local svc=${facet}_svc
- echo Replay barrier on ${!svc}
- do_facet $facet $LCTL --device %${!svc} notransno
- do_facet $facet $LCTL --device %${!svc} readonly
- do_facet $facet $LCTL mark "$facet REPLAY BARRIER on ${!svc}"
- $LCTL mark "local REPLAY BARRIER on ${!svc}"
+ local facet=$1 echo running=${running}
+ local svc=${facet}_svc
+ echo Replay barrier on ${!svc}
+ do_facet $facet $LCTL --device ${!svc} notransno
+ do_facet $facet $LCTL --device ${!svc} readonly
+ do_facet $facet $LCTL mark "$facet REPLAY BARRIER on ${!svc}"
+ $LCTL mark "local REPLAY BARRIER on ${!svc}"
}
mds_evict_client() {
}
fail() {
- facet_failover $* || error "failover: $?"
- clients_up || error "post-failover df: $?"
+ local facets=$1
+ local clients=${CLIENTS:-$HOSTNAME}
+
+ facet_failover $* || error "failover: $?"
+ wait_clients_import_state "$clients" "$facets" FULL
+ clients_up || error "post-failover df: $?"
}
fail_nodf() {
}
facet_host() {
- local facet=$1
-
- [ "$facet" == client ] && echo -n $HOSTNAME && return
- varname=${facet}_HOST
- if [ -z "${!varname}" ]; then
- if [ "${facet:0:3}" == "ost" ]; then
- eval ${facet}_HOST=${ost_HOST}
- fi
- fi
- echo -n ${!varname}
+ local facet=$1
+ local varname
+
+ [ "$facet" == client ] && echo -n $HOSTNAME && return
+ varname=${facet}_HOST
+ if [ -z "${!varname}" ]; then
+ if [ "${facet:0:3}" == "ost" ]; then
+ eval export ${facet}_HOST=${ost_HOST}
+ elif [ "${facet:0:3}" == "mdt" -o \
+ "${facet:0:3}" == "mds" -o \
+ "${facet:0:3}" == "mgs" ]; then
+ eval export ${facet}_HOST=${mds_HOST}
+ fi
+ fi
+ echo -n ${!varname}
}
facet_failover_host() {
local facet=$1
- local var
+ local varname
var=${facet}failover_HOST
if [ -n "${!var}" ]; then
return
fi
+ if [ "${facet:0:3}" == "mdt" -o "${facet:0:3}" == "mds" -o \
+ "${facet:0:3}" == "mgs" ]; then
+
+ eval export ${facet}failover_host=${mds_HOST}
+ echo ${mds_HOST}
+ return
+ fi
+
if [[ $facet == ost* ]]; then
- var=ostfailover_HOST
- if [ -n "${!var}" ]; then
- echo ${!var}
- return
- fi
+ eval export ${facet}failover_host=${ost_HOST}
+ echo ${ost_HOST}
+ return
fi
}
echo -n " ${var}=\"$value\""
done
- echo -n " USE_OFD=$USE_OFD"
-
for facet in ${facets//,/ }; do
var=${facet}_FSTYPE
if [ -n "${!var}" ]; then
}
mgsdevname() {
- DEVNAME=MGSDEV
-
- local fstype=$(facet_fstype mds$num)
+ local DEVPTR
+ local fstype=$(facet_fstype mgs)
case $fstype in
- ldiskfs )
- #if $MGSDEV isn't defined, default is $MDSDEV1
- eval DEVPTR=${!DEVNAME:=${MDSDEV1}};;
- zfs )
- #dataset name is independent of vdev device names
- eval DEVPTR=${FSNAME}-mgs/mgs;;
- * )
- error "unknown fstype!";;
+ ldiskfs )
+ if [ $(facet_host mgs) = $(facet_host mds1) ] &&
+ ( [ -z "$MGSDEV" ] || [ $MGSDEV = $(mdsdevname 1) ] ); then
+ DEVPTR=$(mdsdevname 1)
+ else
+ DEVPTR=$MGSDEV
+ fi;;
+ zfs )
+ if [ $(facet_host mgs) = $(facet_host mds1) ] &&
+ ( [ -z "$MGSDEV" ] || [ $MGSDEV = $(mdsvdevname 1) ] ); then
+ DEVPTR=$(mdsdevname 1)
+ else
+ DEVPTR=${FSNAME}-mgs/mgs
+ fi;;
+ * )
+ error "unknown fstype!";;
esac
- echo -n $DEVPTR
+ echo -n $DEVPTR
}
mgsvdevname() {
+ local VDEVPTR
DEVNAME=MGSDEV
- local fstype=$(facet_fstype mds$num)
+ local fstype=$(facet_fstype mgs)
case $fstype in
- ldiskfs )
- # vdevs are not supported by ldiskfs
- eval VDEVPTR="";;
- zfs )
- #if $MGSDEV isn't defined, default is $MGSDEV1
- eval VDEVPTR=${!DEVNAME:=${MDSDEV1}};;
- * )
- error "unknown fstype!";;
+ ldiskfs )
+ # vdevs are not supported by ldiskfs
+ ;;
+ zfs )
+ if [ $(facet_host mgs) = $(facet_host mds1) ] &&
+ ( [ -z "$MGSDEV" ] || [ $MGSDEV = $(mdsvdevname 1) ] ); then
+ VDEVPTR=$(mdsvdevname 1)
+ else
+ VDEVPTR=$MGSDEV
+ fi;;
+ * )
+ error "unknown fstype!";;
esac
echo -n $VDEVPTR
}
combined_mgs_mds () {
- [[ $MDSDEV1 = $MGSDEV ]] && [[ $mds1_HOST = $mgs_HOST ]]
+ [[ "$(mdsdevname 1)" = "$(mgsdevname)" ]] &&
+ [[ "$(facet_host mds1)" = "$(facet_host mgs)" ]]
}
lower() {
mkfs_opts() {
local facet=$1
+ local dev=$2
+ local fsname=${3:-"$FSNAME"}
local type=$(facet_type $facet)
local index=$(($(facet_number $facet) - 1))
local fstype=$(facet_fstype $facet)
+ local host=$(facet_host $facet)
local opts
local fs_mkfs_opts
local var
return 1
fi
- if [ $type == MGS ] || ( [ $type == MDS ] && combined_mgs_mds ); then
+ if [ $type == MGS ] || ( [ $type == MDS ] &&
+ [ "$dev" == $(mgsdevname) ] &&
+ [ "$host" == "$(facet_host mgs)" ] ); then
opts="--mgs"
else
opts="--mgsnode=$MGSNID"
fi
if [ $type != MGS ]; then
- opts+=" --fsname=$FSNAME --$(lower ${type/MDS/MDT}) --index=$index"
+ opts+=" --fsname=$fsname --$(lower ${type/MDS/MDT}) \
+ --index=$index"
fi
var=${facet}failover_HOST
opts+=${L_GETIDENTITY:+" --param=mdt.identity_upcall=$L_GETIDENTITY"}
if [ $fstype == ldiskfs ]; then
- opts+=${IAMDIR:+" --iam-dir"}
-
fs_mkfs_opts+=${MDSJOURNALSIZE:+" -J size=$MDSJOURNALSIZE"}
fs_mkfs_opts+=${MDSISIZE:+" -i $MDSISIZE"}
fi
echo Formatting mgs, mds, osts
if ! combined_mgs_mds ; then
echo "Format mgs: $(mgsdevname)"
- add mgs $(mkfs_opts mgs) --reformat $(mgsdevname) \
- $(mgsvdevname) ${quiet:+>/dev/null} || exit 10
- fi
+ add mgs $(mkfs_opts mgs $(mgsdevname)) --reformat \
+ $(mgsdevname) $(mgsvdevname) ${quiet:+>/dev/null} ||
+ exit 10
+ fi
- for num in `seq $MDSCOUNT`; do
- echo "Format mds$num: $(mdsdevname $num)"
- add mds$num $(mkfs_opts mds$num) --reformat \
- $(mdsdevname $num) $(mdsvdevname $num) \
+ for num in $(seq $MDSCOUNT); do
+ echo "Format mds$num: $(mdsdevname $num)"
+ add mds$num $(mkfs_opts mds$num $(mdsdevname ${num})) \
+ --reformat $(mdsdevname $num) $(mdsvdevname $num) \
${quiet:+>/dev/null} || exit 10
- done
+ done
- for num in `seq $OSTCOUNT`; do
- echo "Format ost$num: $(ostdevname $num)"
- add ost$num $(mkfs_opts ost$num) --reformat \
- $(ostdevname $num) $(ostvdevname ${num}) \
+ for num in $(seq $OSTCOUNT); do
+ echo "Format ost$num: $(ostdevname $num)"
+ add ost$num $(mkfs_opts ost$num $(ostdevname ${num})) \
+ --reformat $(ostdevname $num) $(ostvdevname ${num}) \
${quiet:+>/dev/null} || exit 10
- done
+ done
}
mount_client() {
zconf_mount `hostname` $1 || error "mount failed"
}
-writeconf_facet () {
- local facet=$1
- local dev=$2
+writeconf_facet() {
+ local facet=$1
+ local dev=$2
- do_facet $facet "$TUNEFS --writeconf $dev"
+ stop ${facet} -f
+ rm -f ${facet}active
+ do_facet ${facet} "$TUNEFS --quiet --writeconf $dev" || return 1
+ return 0
}
writeconf_all () {
- for num in `seq $MDSCOUNT`; do
- DEVNAME=$(mdsdevname $num)
- writeconf_facet mds$num $DEVNAME
- done
+ local mdt_count=${1:-$MDSCOUNT}
+ local ost_count=${2:-$OSTCOUNT}
+ local rc=0
- for num in `seq $OSTCOUNT`; do
- DEVNAME=$(ostdevname $num)
- writeconf_facet ost$num $DEVNAME
- done
+ for num in $(seq $mdt_count); do
+ DEVNAME=$(mdsdevname $num)
+ writeconf_facet mds$num $DEVNAME || rc=$?
+ done
+
+ for num in $(seq $ost_count); do
+ DEVNAME=$(ostdevname $num)
+ writeconf_facet ost$num $DEVNAME || rc=$?
+ done
+ return $rc
}
setupall() {
}
init_facet_vars () {
- [ "$CLIENTONLY" ] && return 0
- local facet=$1
- shift
- local device=$1
+ [ "$CLIENTONLY" ] && return 0
+ local facet=$1
+ shift
+ local device=$1
- shift
+ shift
- eval export ${facet}_dev=${device}
- eval export ${facet}_opt=\"$@\"
+ eval export ${facet}_dev=${device}
+ eval export ${facet}_opt=\"$@\"
- local dev=${facet}_dev
- local label=$(devicelabel ${facet} ${!dev})
- [ -z "$label" ] && echo no label for ${!dev} && exit 1
+ local dev=${facet}_dev
- eval export ${facet}_svc=${label}
+ # We need to loop for the label
+ # in case its not initialized yet.
+ for wait_time in {0,1,3,5,10}; do
- local varname=${facet}failover_HOST
- if [ -z "${!varname}" ]; then
- eval $varname=$(facet_host $facet)
- fi
+ if [ $wait_time -gt 0 ]; then
+ echo "${!dev} not yet initialized,"\
+ "waiting ${wait_time} seconds."
+ sleep $wait_time
+ fi
- # ${facet}failover_dev is set in cfg file
- varname=${facet}failover_dev
- if [ -n "${!varname}" ] ; then
- eval export ${facet}failover_dev=${!varname}
- else
- eval export ${facet}failover_dev=$device
- fi
+ local label=$(devicelabel ${facet} ${!dev})
- # get mount point of already mounted device
- # is facet_dev is already mounted then use the real
- # mount point of this facet; otherwise use $(facet_mntpt $facet)
- # i.e. ${facet}_MOUNT if specified by user or default
- local mntpt=$(do_facet ${facet} cat /proc/mounts | \
- awk '"'${!dev}'" == $1 && $3 == "lustre" { print $2 }')
- if [ -z $mntpt ]; then
- mntpt=$(facet_mntpt $facet)
- fi
- eval export ${facet}_MOUNT=$mntpt
+ # Check to make sure the label does
+ # not include ffff at the end of the label.
+ # This indicates it has not been initialized yet.
+
+ if [[ $label =~ [f|F]{4}$ ]]; then
+ # label is not initialized, unset the result
+ # and either try again or fail
+ unset label
+ else
+ break
+ fi
+ done
+
+ [ -z "$label" ] && echo no label for ${!dev} && exit 1
+
+ eval export ${facet}_svc=${label}
+
+ local varname=${facet}failover_HOST
+ if [ -z "${!varname}" ]; then
+ eval export $varname=$(facet_host $facet)
+ fi
+
+ varname=${facet}_HOST
+ if [ -z "${!varname}" ]; then
+ eval export $varname=$(facet_host $facet)
+ fi
+
+ # ${facet}failover_dev is set in cfg file
+ varname=${facet}failover_dev
+ if [ -n "${!varname}" ] ; then
+ eval export ${facet}failover_dev=${!varname}
+ else
+ eval export ${facet}failover_dev=$device
+ fi
+
+ # get mount point of already mounted device
+ # is facet_dev is already mounted then use the real
+ # mount point of this facet; otherwise use $(facet_mntpt $facet)
+ # i.e. ${facet}_MOUNT if specified by user or default
+ local mntpt=$(do_facet ${facet} cat /proc/mounts | \
+ awk '"'${!dev}'" == $1 && $3 == "lustre" { print $2 }')
+ if [ -z $mntpt ]; then
+ mntpt=$(facet_mntpt $facet)
+ fi
+ eval export ${facet}_MOUNT=$mntpt
}
init_facets_vars () {
- local DEVNAME
+ local DEVNAME
- if ! remote_mds_nodsh; then
- for num in `seq $MDSCOUNT`; do
- DEVNAME=`mdsdevname $num`
- init_facet_vars mds$num $DEVNAME $MDS_MOUNT_OPTS
- done
- fi
+ if ! remote_mds_nodsh; then
+ for num in $(seq $MDSCOUNT); do
+ DEVNAME=`mdsdevname $num`
+ init_facet_vars mds$num $DEVNAME $MDS_MOUNT_OPTS
+ done
+ fi
combined_mgs_mds || init_facet_vars mgs $(mgsdevname) $MGS_MOUNT_OPTS
- remote_ost_nodsh && return
-
- for num in `seq $OSTCOUNT`; do
- DEVNAME=`ostdevname $num`
- init_facet_vars ost$num $DEVNAME $OST_MOUNT_OPTS
- done
+ if ! remote_ost_nodsh; then
+ for num in $(seq $OSTCOUNT); do
+ DEVNAME=$(ostdevname $num)
+ init_facet_vars ost$num $DEVNAME $OST_MOUNT_OPTS
+ done
+ fi
}
osc_ensure_active () {
[ $period -lt $timeout ] || log "$count OST are inactive after $timeout seconds, give up"
}
+set_conf_param_and_check() {
+ local myfacet=$1
+ local TEST=$2
+ local PARAM=$3
+ local ORIG=$(do_facet $myfacet "$TEST")
+ if [ $# -gt 3 ]; then
+ local FINAL=$4
+ else
+ local -i FINAL
+ FINAL=$((ORIG + 5))
+ fi
+ echo "Setting $PARAM from $ORIG to $FINAL"
+ do_facet mgs "$LCTL conf_param $PARAM='$FINAL'" ||
+ error "conf_param $PARAM failed"
+
+ wait_update $(facet_host $myfacet) "$TEST" "$FINAL" ||
+ error "check $PARAM failed!"
+}
+
init_param_vars () {
remote_mds_nodsh ||
TIMEOUT=$(do_facet $SINGLEMDS "lctl get_param -n timeout")
}
check_config_clients () {
- local clients=${CLIENTS:-$HOSTNAME}
- local mntpt=$1
+ local clients=${CLIENTS:-$HOSTNAME}
+ local mntpt=$1
- nfs_client_mode && return
+ nfs_client_mode && return
- do_rpc_nodes $clients check_config_client $mntpt
+ do_rpc_nodes "$clients" check_config_client $mntpt
- sanity_mount_check ||
- error "environments are insane!"
+ sanity_mount_check || error "environments are insane!"
}
check_timeout () {
}
is_empty_dir() {
- [ $(find $1 -maxdepth 1 -print | wc -l) = 1 ] && return 0
- return 1
+ [ $(find $1 -maxdepth 1 -print | wc -l) = 1 ] && return 0
+ return 1
}
# empty lustre filesystem may have empty directories lost+found and .lustre
is_empty_fs() {
- [ $(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 0
- [ ! -d $1/.lustre ] || is_empty_dir $1/.lustre && return 0
- return 1
+ [ $(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 $1/.lustre || return 1
+ return 0
}
check_and_setup_lustre() {
local dev
if [ "$type" == ost ]; then
- devs=$(get_obdfilter_param $node "" mntdev)
+ devs=$(get_osd_param $node "" mntdev)
else
devs=$(do_node $node \
"lctl get_param -n osd-*.$FSNAME-M*.mntdev")
local extra_opts=$3
df > /dev/null # update statfs data on disk
- local cmd="$E2FSCK -d -v -t -t -f -n $extra_opts $target_dev"
+ local cmd="$E2FSCK -d -v -t -t -f $extra_opts $target_dev"
echo $cmd
local rc=0
do_node $node $cmd || rc=$?
# verify a directory is shared among nodes.
check_shared_dir() {
- local dir=$1
+ local dir=$1
- [ -z "$dir" ] && return 1
- do_rpc_nodes $(comma_list $(nodes_list)) check_logdir $dir
- check_write_access $dir || return 1
- return 0
+ [ -z "$dir" ] && return 1
+ do_rpc_nodes "$(comma_list $(nodes_list))" check_logdir $dir
+ check_write_access $dir || return 1
+ return 0
}
# Run e2fsck on MDT and OST(s) to generate databases used for lfsck.
[ $MDSCOUNT -eq 1 ] || error "CMD is not supported"
- run_e2fsck $(mdts_nodes) $MDTDEV "--mdsdb $MDSDB"
+ run_e2fsck $(mdts_nodes) $MDTDEV "-n --mdsdb $MDSDB"
i=0
ostidx=0
OSTDB_LIST=""
for node in $(osts_nodes); do
for dev in ${OSTDEVS[i]}; do
- run_e2fsck $node $dev "--mdsdb $MDSDB --ostdb $OSTDB-$ostidx"
+ run_e2fsck $node $dev "-n --mdsdb $MDSDB --ostdb $OSTDB-$ostidx"
OSTDB_LIST="$OSTDB_LIST $OSTDB-$ostidx"
ostidx=$((ostidx + 1))
done
done
}
+# Run lfsck on server node if lfsck can't be found on client (LU-2571)
+run_lfsck_remote() {
+ local cmd="$LFSCK_BIN -c -l --mdsdb $MDSDB --ostdb $OSTDB_LIST $MOUNT"
+ local client=$1
+ local mounted=true
+ local rc=0
+
+ #Check if lustre is already mounted
+ do_rpc_nodes $client is_mounted $MOUNT || mounted=false
+ if ! $mounted; then
+ zconf_mount $client $MOUNT ||
+ error "failed to mount Lustre on $client"
+ fi
+ #Run lfsck
+ echo $cmd
+ do_node $node $cmd || rc=$?
+ #Umount if necessary
+ if ! $mounted; then
+ zconf_umount $client $MOUNT ||
+ error "failed to unmount Lustre on $client"
+ fi
+
+ [ $rc -le $FSCK_MAX_ERR ] ||
+ error "$cmd returned $rc, should be <= $FSCK_MAX_ERR"
+ echo "lfsck finished with rc=$rc"
+
+ return $rc
+}
+
run_lfsck() {
- local cmd="$LFSCK_BIN -c -l --mdsdb $MDSDB --ostdb $OSTDB_LIST $MOUNT"
- echo $cmd
- local rc=0
- eval $cmd || rc=$?
- [ $rc -le $FSCK_MAX_ERR ] || \
- error "$cmd returned $rc, should be <= $FSCK_MAX_ERR"
- echo "lfsck finished with rc=$rc"
+ local facets="client $SINGLEMDS"
+ local found=false
+ local facet
+ local node
+ local rc=0
- rm -rvf $MDSDB* $OSTDB* || true
- return 0
+ for facet in $facets; do
+ node=$(facet_active_host $facet)
+ if check_progs_installed $node $LFSCK_BIN; then
+ found=true
+ break
+ fi
+ done
+ ! $found && error "None of \"$facets\" supports lfsck"
+
+ run_lfsck_remote $node || rc=$?
+
+ rm -rvf $MDSDB* $OSTDB* || true
+ return $rc
}
check_and_cleanup_lustre() {
eval $@
}
+# Convert a space-delimited list to a comma-delimited list. If the input is
+# only whitespace, ensure the output is empty (i.e. "") so [ -n $list ] works
comma_list() {
- # the sed converts spaces to commas, but leaves the last space
- # alone, so the line doesn't end with a comma.
- echo "$*" | tr -s " " "\n" | sort -b -u | tr "\n" " " | sed 's/ \([^$]\)/,\1/g'
+ # echo is used to convert newlines to spaces, since it doesn't
+ # introduce a trailing space as using "tr '\n' ' '" does
+ echo $(tr -s " " "\n" <<< $* | sort -b -u) | tr ' ' ','
}
list_member () {
return $RC
}
+drop_update_reply() {
+# OBD_FAIL_UPDATE_OBJ_NET_REP
+ local index=$1
+ shift 1
+ RC=0
+ do_facet mds${index} lctl set_param fail_loc=0x1701
+ do_facet client "$@" || RC=$?
+ do_facet mds${index} lctl set_param fail_loc=0
+ return $RC
+}
+
pause_bulk() {
#define OBD_FAIL_OST_BRW_PAUSE_BULK 0x214
RC=0
}
drop_ldlm_cancel() {
-#define OBD_FAIL_LDLM_CANCEL 0x304
+#define OBD_FAIL_LDLM_CANCEL_NET 0x304
local RC=0
local list=$(comma_list $(mdts_nodes) $(osts_nodes))
do_nodes $list lctl set_param fail_loc=0x304
}
drop_bl_callback() {
-#define OBD_FAIL_LDLM_BL_CALLBACK 0x305
- RC=0
- do_facet client lctl set_param fail_loc=0x305
- do_facet client "$@" || RC=$?
- do_facet client lctl set_param fail_loc=0
- return $RC
+#define OBD_FAIL_LDLM_BL_CALLBACK_NET 0x305
+ RC=0
+ do_facet client lctl set_param fail_loc=0x80000305
+ do_facet client "$@" || RC=$?
+ do_facet client lctl set_param fail_loc=0
+ return $RC
}
drop_ldlm_reply() {
gather_logs $(comma_list $(nodes_list))
fi
- debugrestore
- [ "$TESTSUITELOG" ] && echo "$0: ${TYPE}: $TESTNAME $@" >> $TESTSUITELOG
- echo "$@" > $LOGDIR/err
+ debugrestore
+ [ "$TESTSUITELOG" ] &&
+ echo "$TESTSUITE: $TYPE: $TESTNAME $@" >> $TESTSUITELOG
+ echo "$@" > $LOGDIR/err
}
exit_status () {
return $?
}
-equals_msg() {
- banner "$*"
-}
-
log() {
echo "$*"
module_loaded lnet || load_modules
}
complete () {
- equals_msg $1 test complete, duration $2 sec
+ local duration=$1
+
+ banner test complete, duration $duration sec
[ -f "$TESTSUITELOG" ] && egrep .FAIL $TESTSUITELOG || true
- echo duration $2 >>$TESTSUITELOG
+ echo duration $duration >>$TESTSUITELOG
}
pass() {
reset_fail_loc
check_grant ${testnum} || error "check_grant $testnum failed with $?"
check_catastrophe || error "LBUG/LASSERT detected"
- ps auxww | grep -v grep | grep -q multiop && error "multiop still running"
+ if [ "$PARALLEL" != "yes" ]; then
+ ps auxww | grep -v grep | grep -q multiop &&
+ error "multiop still running"
+ fi
unset TESTNAME
unset tdir
umask $SAVE_UMASK
$LFS mdts $2 | sed -ne "/^$1: /s/.* \(.*\) .*$/\1/p"
}
+# Description:
+# Return unique identifier for given hostname
+host_id() {
+ local host_name=$1
+ echo $host_name | md5sum | cut -d' ' -f1
+}
+
+# Description:
+# Returns list of ip addresses for each interface
+local_addr_list() {
+ ip addr | awk '/inet\ / {print $2}' | awk -F\/ '{print $1}'
+}
+
+is_local_addr() {
+ local addr=$1
+ # Cache address list to avoid mutiple execution of local_addr_list
+ LOCAL_ADDR_LIST=${LOCAL_ADDR_LIST:-$(local_addr_list)}
+ local i
+ for i in $LOCAL_ADDR_LIST ; do
+ [[ "$i" == "$addr" ]] && return 0
+ done
+ return 1
+}
+
+local_node() {
+ local host_name=$1
+ local is_local="IS_LOCAL_$(host_id $host_name)"
+ if [ -z "${!is_local-}" ] ; then
+ eval $is_local=0
+ local host_ip=$($LUSTRE/tests/resolveip $host_name)
+ is_local_addr "$host_ip" && eval $is_local=1
+ fi
+ [[ "${!is_local}" == "1" ]]
+}
+
remote_node () {
- local node=$1
- [ "$node" != "$(hostname)" ]
+ local node=$1
+ local_node $node && return 1
+ return 0
}
remote_mds ()
}
remote_nodes_list () {
- local rnodes=$(nodes_list)
- rnodes=$(echo " $rnodes " | sed -re "s/\s+$HOSTNAME\s+/ /g")
- echo $rnodes
+ echo $(nodes_list) | sed -re "s/\<$HOSTNAME\>//g"
}
init_clients_lists () {
}
check_catastrophe() {
- local rnodes=${1:-$(comma_list $(remote_nodes_list))}
- local C=$CATASTROPHE
- [ -f $C ] && [ $(cat $C) -ne 0 ] && return 1
-
- if [ $rnodes ]; then
- do_nodes $rnodes "rc=\\\$([ -f $C ] && echo \\\$(< $C) || echo 0);
-if [ \\\$rc -ne 0 ]; then echo \\\$(hostname): \\\$rc; fi
-exit \\\$rc;"
- fi
+ local rnodes=${1:-$(comma_list $(remote_nodes_list))}
+ local C=$CATASTROPHE
+ [ -f $C ] && [ $(cat $C) -ne 0 ] && return 1
+
+ [ -z "$rnodes" ] && return 0
+
+ local data
+ data=$(do_nodes "$rnodes" "rc=\\\$([ -f $C ] &&
+ echo \\\$(< $C) || echo 0);
+ if [ \\\$rc -ne 0 ]; then echo \\\$(hostname): \\\$rc; fi
+ exit \\\$rc")
+ local rc=$?
+ if [ -n "$data" ]; then
+ echo $data
+ return $rc
+ fi
+ return 0
}
# CMD: determine mds index where directory inode presents
}
mdsrate_cleanup () {
- if [ -d $4 ]; then
- mpi_run -np $1 -machinefile $2 ${MDSRATE} --unlink --nfiles $3 --dir $4 --filefmt $5 $6
- rmdir $4
- fi
+ if [ -d $4 ]; then
+ mpi_run -np $1 ${MACHINEFILE_OPTION} $2 ${MDSRATE} --unlink \
+ --nfiles $3 --dir $4 --filefmt $5 $6
+ rmdir $4
+ fi
}
delayed_recovery_enabled () {
}
get_clientosc_proc_path() {
- echo "${1}-osc-[^M]*"
+ echo "${1}-osc-*"
}
get_lustre_version () {
local CONN_STATE
local i=0
- CONN_STATE=$($LCTL get_param -n $CONN_PROC 2>/dev/null | cut -f2)
+ CONN_STATE=$($LCTL get_param -n $CONN_PROC 2>/dev/null | cut -f2 | uniq)
while [ "${CONN_STATE}" != "${expected}" ]; do
if [ "${expected}" == "DISCONN" ]; then
# for disconn we can check after proc entry is removed
error "can't put import for $CONN_PROC into ${expected} state after $i sec, have ${CONN_STATE}" && \
return 1
sleep 1
- CONN_STATE=$($LCTL get_param -n $CONN_PROC 2>/dev/null | cut -f2)
+ # Add uniq for multi-mount case
+ CONN_STATE=$($LCTL get_param -n $CONN_PROC 2>/dev/null | cut -f2 | uniq)
i=$(($i + 1))
done
done
}
+wait_import_state_mount() {
+ if ! is_mounted $MOUNT && ! is_mounted $MOUNT2; then
+ return 0
+ fi
+
+ wait_import_state $*
+}
+
# One client request could be timed out because server was not ready
# when request was sent by client.
# The request timeout calculation details :
local expected=$3
local ost=$(get_osc_import_name $facet $ost_facet)
- local param="osc.${ost}.ost_server_uuid"
+ local param="osc.${ost}.ost_server_uuid"
+ local i=0
# 1. wait the deadline of client 1st request (it could be skipped)
# 2. wait the deadline of client 2nd request
local maxtime=$(( 2 * $(request_timeout $facet)))
- if ! do_rpc_nodes $(facet_host $facet) \
- _wait_import_state $expected $param $maxtime; then
- error "import is not in ${expected} state"
- return 1
- fi
+ #During setup time, the osc might not be setup, it need wait
+ #until list_param can return valid value. And also if there
+ #are mulitple osc entries we should list all of them before
+ #go to wait.
+ local params=$($LCTL list_param $param || true)
+ while [ -z "$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
+ return 1
+ fi
+ fi
+ sleep 1
+ i=$((i + 1))
+ params=$($LCTL list_param $param || true)
+ done
- return 0
+ if ! do_rpc_nodes "$(facet_host $facet)" \
+ wait_import_state $expected "$params" $maxtime; then
+ error "import is not in ${expected} state"
+ return 1
+ fi
+
+ return 0
}
get_clientmdc_proc_path() {
}
do_rpc_nodes () {
- local list=$1
- shift
+ local list=$1
+ shift
+
+ [ -z "$list" ] && return 0
- # Add paths to lustre tests for 32 and 64 bit systems.
- local RPATH="PATH=$RLUSTRE/tests:/usr/lib/lustre/tests:/usr/lib64/lustre/tests:$PATH"
- do_nodesv $list "${RPATH} NAME=${NAME} sh rpc.sh $@ "
+ # Add paths to lustre tests for 32 and 64 bit systems.
+ local RPATH="PATH=$RLUSTRE/tests:/usr/lib/lustre/tests:/usr/lib64/lustre/tests:$PATH"
+ do_nodesv $list "${RPATH} NAME=${NAME} sh rpc.sh $@ "
}
wait_clients_import_state () {
local params=$(expand_list $params $proc_path)
done
- if ! do_rpc_nodes $list wait_import_state $expected $params; then
- error "import is not in ${expected} state"
- return 1
- fi
+ if ! do_rpc_nodes "$list" wait_import_state_mount $expected $params; then
+ error "import is not in ${expected} state"
+ return 1
+ fi
}
oos_full() {
{
local dir=$1
local llverfs_opts=$2
+ local use_partial_arg=$3
local partial_arg=""
local size=$(df -B G $dir |tail -n 1 |awk '{print $2}' |sed 's/G//') #GB
# Run in partial (fast) mode if the size
# of a partition > 1 GB
- [ $size -gt 1 ] && partial_arg="-p"
+ [ "x$use_partial_arg" != "xno" ] && [ $size -gt 1 ] && partial_arg="-p"
llverfs $partial_arg $llverfs_opts $dir
}
echo "removing files from $mdtdev on $facet: $files"
if [ $(facet_fstype $facet) == ldiskfs ] &&
- ! do_facet $facet test -b ${!dev}; then
+ ! do_facet $facet test -b $mdtdev; then
opts=$(csa_add "$opts" -o loop)
fi
mount -t $(facet_fstype $facet) $opts $mdtdev $mntpt ||
echo "duplicating files on $mdtdev on $facet: $files"
mkdir -p $mntpt || return $?
if [ $(facet_fstype $facet) == ldiskfs ] &&
- ! do_facet $facet test -b ${!dev}; then
+ ! do_facet $facet test -b $mdtdev; then
opts=$(csa_add "$opts" -o loop)
fi
mount -t $(facet_fstype $facet) $opts $mdtdev $mntpt ||
# Check whether the "large_xattr" feature is enabled or not.
large_xattr_enabled() {
- local mds_dev=$(mdsdevname ${SINGLEMDS//mds/})
+ [[ $(facet_fstype $SINGLEMDS) == zfs ]] && return 0
- do_facet $SINGLEMDS "$DUMPE2FS -h $mds_dev 2>&1 | grep -q large_xattr"
- return ${PIPESTATUS[0]}
+ local mds_dev=$(mdsdevname ${SINGLEMDS//mds/})
+
+ do_facet $SINGLEMDS "$DUMPE2FS -h $mds_dev 2>&1 | grep -q large_xattr"
+ return ${PIPESTATUS[0]}
}
# Get the maximum xattr size supported by the filesystem.
local metaea=${TMP}/backup_restore.ea
local metadata=${TMP}/backup_restore.tgz
local opts=${MDS_MOUNT_OPTS}
+ local svc=${SINGLEMDS}_svc
+ local igif=$1
if ! ${rcmd} test -b ${devname}; then
opts=$(csa_add "$opts" -o loop)
${rcmd} rm -f $metaea $metadata
# step 3: mount dev
${rcmd} mount -t ldiskfs $opts $devname $mntpt || return 1
+ if [ ! -z $igif ]; then
+ # step 3.5: rm .lustre
+ ${rcmd} rm -rf $mntpt/ROOT/.lustre || return 1
+ fi
# step 4: backup metaea
echo "backup EA"
${rcmd} "cd $mntpt && getfattr -R -d -m '.*' -P . > $metaea && cd -" ||
${rcmd} umount -d $mntpt || return 10
# step 14: cleanup tmp backup
${rcmd} rm -f $metaea $metadata
+ # step 15: reset device label - it's not virgin on
+ ${rcmd} e2label $devname ${!svc}
}
# remove OI files
${rcmd} umount -d $mntpt || return 2
# OI files will be recreated when mounted as lustre next time.
}
+
+# generate maloo upload-able log file name
+# \param logname specify unique part of file name
+generate_logname() {
+ local logname=${1:-"default_logname"}
+
+ echo "$TESTLOG_PREFIX.$TESTNAME.$logname.$(hostname -s).log"
+}
+
+# mkdir directory on different MDTs
+test_mkdir() {
+ local option
+ local parent
+ local child
+ local path
+ local dir
+ local rc=0
+
+ if [ $# -eq 2 ]; then
+ option=$1
+ path=$2
+ else
+ path=$1
+ fi
+
+ child=${path##*/}
+ parent=${path%/*}
+
+ if [ "$parent" == "$child" ]; then
+ parent=$(pwd)
+ fi
+
+ if [ "$option" == "-p" -a -d ${parent}/${child} ]; then
+ return $rc
+ fi
+
+ # it needs to check whether there is further / in child
+ dir=$(echo $child | awk -F '/' '{print $2}')
+ if [ ! -z "$dir" ]; then
+ local subparent=$(echo $child | awk -F '/' '{ print $1 }')
+ parent=${parent}"/"${subparent}
+ child=$dir
+ fi
+
+ if [ ! -d ${parent} ]; then
+ if [ "$option" == "-p" ]; then
+ mkdir -p ${parent}
+ else
+ return 1
+ fi
+ fi
+
+ if [ $MDSCOUNT -le 1 ]; then
+ mkdir $option ${parent}/${child} || rc=$?
+ else
+ local mdt_idx=$($LFS getstripe -M $parent)
+
+ if [ "$mdt_idx" -ne 0 ]; then
+ mkdir $option ${parent}/${child} || rc=$?
+ return $rc
+ fi
+
+ local test_num=$(echo $testnum | sed -e 's/[^0-9]*//g')
+ local mdt_idx=$((test_num % MDSCOUNT))
+ echo "mkdir $mdt_idx for ${parent}/${child}"
+ $LFS setdirstripe -i $mdt_idx ${parent}/${child} || rc=$?
+ fi
+ return $rc
+}
git://git.whamcloud.com / fs / lustre-release.git / blobdiff