[ -z "$MODPROBECONF" -a -f /etc/modprobe.conf ] &&
MODPROBECONF=/etc/modprobe.conf
+sanitize_parameters() {
+ for i in DIR DIR1 DIR2 MOUNT MOUNT1 MOUNT2
+ do
+ local path=${!i}
+ if [ -d "$path" ]; then
+ eval export $i=$(echo $path | sed -r 's/\/+$//g')
+ fi
+ done
+}
assert_DIR () {
- local failed=""
- [[ $DIR/ = $MOUNT/* ]] || \
- { failed=1 && echo "DIR=$DIR not in $MOUNT. Aborting."; }
- [[ $DIR1/ = $MOUNT1/* ]] || \
- { failed=1 && echo "DIR1=$DIR1 not in $MOUNT1. Aborting."; }
- [[ $DIR2/ = $MOUNT2/* ]] || \
- { failed=1 && echo "DIR2=$DIR2 not in $MOUNT2. Aborting"; }
+ local failed=""
+ [[ $DIR/ = $MOUNT/* ]] ||
+ { failed=1 && echo "DIR=$DIR not in $MOUNT. Aborting."; }
+ [[ $DIR1/ = $MOUNT1/* ]] ||
+ { failed=1 && echo "DIR1=$DIR1 not in $MOUNT1. Aborting."; }
+ [[ $DIR2/ = $MOUNT2/* ]] ||
+ { failed=1 && echo "DIR2=$DIR2 not in $MOUNT2. Aborting"; }
- [ -n "$failed" ] && exit 99 || true
+ [ -n "$failed" ] && exit 99 || true
}
usage() {
export TEST_FAILED=false
export FAIL_ON_SKIP_ENV=${FAIL_ON_SKIP_ENV:-false}
export RPC_MODE=${RPC_MODE:-false}
+ export DO_CLEANUP=${DO_CLEANUP:-true}
export MKE2FS=$MKE2FS
if [ -z "$MKE2FS" ]; then
export DIR2
export SAVE_PWD=${SAVE_PWD:-$LUSTRE/tests}
export AT_MAX_PATH
+ export LDEV=${LDEV:-"$LUSTRE/scripts/ldev"}
+ [ ! -f "$LDEV" ] && export LDEV=$(which ldev 2> /dev/null)
if [ "$ACCEPTOR_PORT" ]; then
export PORT_OPT="--port $ACCEPTOR_PORT"
export SHUTDOWN_ATTEMPTS=${SHUTDOWN_ATTEMPTS:-3}
export OSD_TRACK_DECLARES_LBUG=${OSD_TRACK_DECLARES_LBUG:-"yes"}
- # command line
+ # command line
- while getopts "rvwf:" opt $*; do
- case $opt in
- f) CONFIG=$OPTARG;;
- r) REFORMAT=--reformat;;
- v) VERBOSE=true;;
- w) WRITECONF=writeconf;;
- \?) usage;;
- esac
- done
+ while getopts "rvwf:" opt $*; do
+ case $opt in
+ f) CONFIG=$OPTARG;;
+ r) REFORMAT=yes;;
+ v) VERBOSE=true;;
+ w) WRITECONF=writeconf;;
+ \?) usage;;
+ esac
+ done
- shift $((OPTIND - 1))
- ONLY=${ONLY:-$*}
+ shift $((OPTIND - 1))
+ ONLY=${ONLY:-$*}
# print the durations of each test if "true"
DDETAILS=${DDETAILS:-false}
lustre_build_version() {
local facet=${1:-client}
- # lustre: 2.6.52
- # kernel: patchless_client
- # build: v2_6_92_0-gadb3ee4-2.6.32-431.29.2.el6_lustre.x86_64
+ # lustre: 2.8.52
local VER=$(do_facet $facet $LCTL get_param -n version 2> /dev/null |
awk '/lustre: / { print $2 }')
# lctl 2.6.50
load_module lov/lov
load_module mgc/mgc
load_module obdecho/obdecho
- if ! client_only; then
- SYMLIST=/proc/kallsyms
- grep -q crc16 $SYMLIST || { modprobe crc16 2>/dev/null || true; }
- grep -q -w jbd $SYMLIST || { modprobe jbd 2>/dev/null || true; }
- grep -q -w jbd2 $SYMLIST || { modprobe jbd2 2>/dev/null || true; }
+ if ! client_only; then
+ SYMLIST=/proc/kallsyms
+ grep -q crc16 $SYMLIST ||
+ { modprobe crc16 2>/dev/null || true; }
+ grep -q -w jbd2 $SYMLIST ||
+ { modprobe jbd2 2>/dev/null || true; }
load_module lfsck/lfsck
- [ "$LQUOTA" != "no" ] && load_module quota/lquota $LQUOTAOPTS
+ [ "$LQUOTA" != "no" ] &&
+ load_module quota/lquota $LQUOTAOPTS
if [[ $(node_fstypes $HOSTNAME) == *zfs* ]]; then
modprobe zfs
load_module osd-zfs/osd_zfs
if [[ $(node_fstypes $HOSTNAME) == *ldiskfs* ]]; then
grep -q exportfs_decode_fh $SYMLIST ||
{ modprobe exportfs 2> /dev/null || true; }
+ grep -q -w mbcache $SYMLIST ||
+ { modprobe mbcache 2>/dev/null || true; }
load_module ../ldiskfs/ldiskfs
load_module osd-ldiskfs/osd_ldiskfs
fi
load_module osp/osp
load_module ofd/ofd
load_module osp/osp
- fi
+ fi
load_module llite/lustre
llite_lloop_enabled && load_module llite/llite_lloop
facet_number() {
local facet=$1
- if [ $facet == mgs ]; then
+ if [ $facet == mgs ] || [ $facet == client ]; then
return 1
fi
local name=$3
do_nodes $nodes "$LCTL get_param -n obdfilter.$device.$name \
- osd-*.$device.$name 2>&1" | grep -v 'Found no match'
+ osd-*.$device.$name 2>&1" | grep -v 'error:'
}
set_osd_param() {
local value=$4
do_nodes $nodes "$LCTL set_param -n obdfilter.$device.$name=$value \
- osd-*.$device.$name=$value 2>&1" | grep -v 'Found no match'
+ osd-*.$device.$name=$value 2>&1" | grep -v 'error:'
}
set_debug_size () {
set_default_debug_nodes $node
}
+set_hostid () {
+ local hostid=${1:-$(hostid)}
+
+ if [ ! -s /etc/hostid ]; then
+ printf $(echo -n $hostid |
+ sed 's/\(..\)\(..\)\(..\)\(..\)/\\x\4\\x\3\\x\2\\x\1/') >/etc/hostid
+ fi
+}
+
# Facet functions
mount_facets () {
local facets=${1:-$(get_facets)}
local opt=${facet}_opt
local mntpt=$(facet_mntpt $facet)
local opts="${!opt} $@"
+ local fstype=$(facet_fstype $facet)
+ local devicelabel
module_loaded lustre || load_modules
import_zpool $facet || return ${PIPESTATUS[0]}
fi
+ case $fstype in
+ ldiskfs)
+ devicelabel=$(do_facet ${facet} "$E2LABEL ${!dev}");;
+ zfs)
+ devicelabel=$(do_facet ${facet} "$ZFS get -H -o value \
+ lustre:svname ${!dev}");;
+ *)
+ error "unknown fstype!";;
+ esac
+
echo "Starting ${facet}: $opts ${!dev} $mntpt"
# for testing LU-482 error handling in mount_facets() and test_0a()
if [ -f $TMP/test-lu482-trigger ]; then
return $RC
fi
+ health=$(do_facet ${facet} "$LCTL get_param -n health_check")
+ if [[ "$health" != "healthy" ]]; then
+ error "$facet is in a unhealthy state"
+ fi
+
set_default_debug_facet $facet
if [[ $facet == mds* ]]; then
if [[ $opts =~ .*nosvc.* ]]; then
echo "Start ${!dev} without service"
else
- local fstype=$(facet_fstype $facet)
case $fstype in
ldiskfs)
esac
fi
+ # commit the device label change to disk
+ if [[ $devicelabel =~ (:[a-zA-Z]{3}[0-9]{4}) ]]; then
+ do_facet $facet "sync; sync; sync"
+ fi
+
+
label=$(devicelabel ${facet} ${!dev})
[ -z "$label" ] && echo no label for ${!dev} && exit 1
eval export ${facet}_svc=${label}
}
zconf_mount() {
- local client=$1
- local mnt=$2
- local opts=${3:-$MOUNT_OPTS}
- opts=${opts:+-o $opts}
- local flags=${4:-$MOUNT_FLAGS}
+ local client=$1
+ local mnt=$2
+ local opts=${3:-$MOUNT_OPTS}
+ opts=${opts:+-o $opts}
+ local flags=${4:-$MOUNT_FLAGS}
+
+ local device=$MGSNID:/$FSNAME$FILESET
+ if [ -z "$mnt" -o -z "$FSNAME" ]; then
+ echo "Bad mount command: opt=$flags $opts dev=$device " \
+ "mnt=$mnt"
+ exit 1
+ fi
+
+ echo "Starting client: $client: $flags $opts $device $mnt"
+ do_node $client mkdir -p $mnt
+ if [ -n "$FILESET" -a -z "$SKIP_FILESET" ];then
+ do_node $client $MOUNT_CMD $flags $opts $MGSNID:/$FSNAME \
+ $mnt || return 1
+ #disable FILESET if not supported
+ do_nodes $client lctl get_param -n \
+ mdc.$FSNAME-MDT0000*.import | grep -q subtree ||
+ device=$MGSNID:/$FSNAME
+ do_node $client mkdir -p $mnt/$FILESET
+ do_node $client "! grep -q $mnt' ' /proc/mounts ||
+ umount $mnt"
+ fi
+ do_node $client $MOUNT_CMD $flags $opts $device $mnt || return 1
+
+ set_default_debug_nodes $client
- local device=$MGSNID:/$FSNAME
- if [ -z "$mnt" -o -z "$FSNAME" ]; then
- echo Bad zconf mount command: opt=$flags $opts dev=$device mnt=$mnt
- exit 1
- fi
-
- echo "Starting client: $client: $flags $opts $device $mnt"
- do_node $client mkdir -p $mnt
- do_node $client $MOUNT_CMD $flags $opts $device $mnt || return 1
-
- set_default_debug_nodes $client
-
- return 0
+ return 0
}
zconf_umount() {
# mount clients if not mouted
zconf_mount_clients() {
- local clients=$1
- local mnt=$2
- local opts=${3:-$MOUNT_OPTS}
- opts=${opts:+-o $opts}
- local flags=${4:-$MOUNT_FLAGS}
-
- local device=$MGSNID:/$FSNAME
- if [ -z "$mnt" -o -z "$FSNAME" ]; then
- echo Bad zconf mount command: opt=$flags $opts dev=$device mnt=$mnt
- exit 1
- fi
-
- echo "Starting client $clients: $flags $opts $device $mnt"
-
- do_nodes $clients "
+ local clients=$1
+ local mnt=$2
+ local opts=${3:-$MOUNT_OPTS}
+ opts=${opts:+-o $opts}
+ local flags=${4:-$MOUNT_FLAGS}
+
+ local device=$MGSNID:/$FSNAME$FILESET
+ if [ -z "$mnt" -o -z "$FSNAME" ]; then
+ echo "Bad conf mount command: opt=$flags $opts dev=$device " \
+ "mnt=$mnt"
+ exit 1
+ fi
+
+ echo "Starting client $clients: $flags $opts $device $mnt"
+ if [ -n "$FILESET" -a ! -n "$SKIP_FILESET" ]; then
+ do_nodes $clients "! grep -q $mnt' ' /proc/mounts ||
+ umount $mnt"
+ do_nodes $clients $MOUNT_CMD $flags $opts $MGSNID:/$FSNAME \
+ $mnt || return 1
+ #disable FILESET if not supported
+ do_nodes $clients lctl get_param -n \
+ mdc.$FSNAME-MDT0000*.import | grep -q subtree ||
+ device=$MGSNID:/$FSNAME
+ do_nodes $clients mkdir -p $mnt/$FILESET
+ do_nodes $clients "! grep -q $mnt' ' /proc/mounts ||
+ umount $mnt"
+ fi
+
+ do_nodes $clients "
running=\\\$(mount | grep -c $mnt' ');
rc=0;
if [ \\\$running -eq 0 ] ; then
fi;
exit \\\$rc" || return ${PIPESTATUS[0]}
- echo "Started clients $clients: "
- do_nodes $clients "mount | grep $mnt' '"
+ echo "Started clients $clients: "
+ do_nodes $clients "mount | grep $mnt' '"
- set_default_debug_nodes $clients
+ set_default_debug_nodes $clients
- return 0
+ return 0
}
zconf_umount_clients() {
}
client_reconnect_try() {
- uname -n >> $MOUNT/recon
- if [ -z "$CLIENTS" ]; then
- df $MOUNT; uname -n >> $MOUNT/recon
- else
- do_nodes $CLIENTS "df $MOUNT; uname -n >> $MOUNT/recon" > /dev/null
- fi
- echo Connected clients:
- cat $MOUNT/recon
- ls -l $MOUNT/recon > /dev/null
- rm $MOUNT/recon
+ local f=$MOUNT/recon
+
+ uname -n >> $f
+ if [ -z "$CLIENTS" ]; then
+ $LFS df $MOUNT; uname -n >> $f
+ else
+ do_nodes $CLIENTS "$LFS df $MOUNT; uname -n >> $f" > /dev/null
+ fi
+ echo "Connected clients: $(cat $f)"
+ ls -l $f > /dev/null
+ rm $f
}
client_reconnect() {
replay_barrier() {
local facet=$1
do_facet $facet "sync; sync; sync"
- df $MOUNT
+ $LFS df $MOUNT
# make sure there will be no seq change
local clients=${CLIENTS:-$HOSTNAME}
facet_failover $* || error "failover: $?"
wait_clients_import_state "$clients" "$facets" FULL
- clients_up || error "post-failover df: $?"
+ clients_up || error "post-failover stat: $?"
}
fail_nodf() {
change_active $facet
wait_for_facet $facet
mount_facet $facet -o abort_recovery
- clients_up || echo "first df failed: $?"
- clients_up || error "post-failover df: $?"
+ clients_up || echo "first stat failed: $?"
+ clients_up || error "post-failover stat: $?"
}
do_lmc() {
[[ $facet = mgs ]] && combined_mgs_mds && facet="mds1"
local var=${facet}_MOUNT
- eval mntpt=${!var:-${MOUNT%/*}/$facet}
+ eval mntpt=${!var:-${MOUNT}-$facet}
echo -n $mntpt
}
formatall() {
stopall
+ # Set hostid for ZFS/SPL zpool import protection
+ do_rpc_nodes "$(comma_list $(remote_nodes_list))" set_hostid
+
# We need ldiskfs here, may as well load them all
load_modules
[ "$CLIENTONLY" ] && return
do_facet mgs "$LCTL conf_param $PARAM='$FINAL'" ||
error "conf_param $PARAM failed"
- wait_update $(facet_host $myfacet) "$TEST" "$FINAL" ||
+ wait_update_facet $myfacet "$TEST" "$FINAL" ||
error "check $PARAM failed!"
}
}
check_and_setup_lustre() {
- nfs_client_mode && return
+ sanitize_parameters
+ nfs_client_mode && return
cifs_client_mode && return
- local MOUNTED=$(mounted_lustre_filesystems)
-
- local do_check=true
- # 1.
- # both MOUNT and MOUNT2 are not mounted
- if ! is_mounted $MOUNT && ! is_mounted $MOUNT2; then
- [ "$REFORMAT" ] && formatall
- # setupall mounts both MOUNT and MOUNT2 (if MOUNT_2 is set)
- setupall
- is_mounted $MOUNT || error "NAME=$NAME not mounted"
- export I_MOUNTED=yes
- do_check=false
+ local MOUNTED=$(mounted_lustre_filesystems)
+
+ local do_check=true
+ # 1.
+ # both MOUNT and MOUNT2 are not mounted
+ if ! is_mounted $MOUNT && ! is_mounted $MOUNT2; then
+ [ "$REFORMAT" = "yes" ] && formatall
+ # setupall mounts both MOUNT and MOUNT2 (if MOUNT_2 is set)
+ setupall
+ is_mounted $MOUNT || error "NAME=$NAME not mounted"
+ export I_MOUNTED=yes
+ do_check=false
# 2.
# MOUNT2 is mounted
elif is_mounted $MOUNT2; then
fi
if is_mounted $MOUNT; then
- [ -n "$DIR" ] && rm -rf $DIR/[Rdfs][0-9]* ||
- error "remove sub-test dirs failed"
+ if $DO_CLEANUP; then
+ [ -n "$DIR" ] && rm -rf $DIR/[Rdfs][0-9]* ||
+ error "remove sub-test dirs failed"
+ else
+ echo "skip cleanup"
+ fi
[ "$ENABLE_QUOTA" ] && restore_quota || true
fi
}
set_nodes_failloc () {
- do_nodes $(comma_list $1) lctl set_param fail_val=0 fail_loc=$2
+ local fv=${3:-0}
+ do_nodes $(comma_list $1) lctl set_param fail_val=$fv fail_loc=$2
}
cancel_lru_locks() {
# prints bash call stack
print_stack_trace() {
+ local skip=${1:-1}
echo " Trace dump:"
- for (( i=1; i < ${#BASH_LINENO[*]} ; i++ )) ; do
- local s=${BASH_SOURCE[$i]}
- local l=${BASH_LINENO[$i-1]}
- local f=${FUNCNAME[$i]}
- echo " = $s:$l:$f()"
+ for (( i=$skip; i < ${#BASH_LINENO[*]} ; i++ )) ; do
+ local src=${BASH_SOURCE[$i]}
+ local lineno=${BASH_LINENO[$i-1]}
+ local funcname=${FUNCNAME[$i]}
+ echo " = $src:$lineno:$funcname()"
done
}
-##################################
-# Test interface
-##################################
-
-error_noexit() {
+report_error() {
local TYPE=${TYPE:-"FAIL"}
local dump=true
dump=false
fi
-
log " ${TESTSUITE} ${TESTNAME}: @@@@@@ ${TYPE}: $@ "
- print_stack_trace >&2
-
+ (print_stack_trace 2) >&2
mkdir -p $LOGDIR
# We need to dump the logs on all nodes
if $dump; then
echo "$@" > $LOGDIR/err
fi
fi
+
+ # cleanup the env for failed tests
+ reset_fail_loc
+}
+
+##################################
+# Test interface
+##################################
+
+error_noexit() {
+ report_error "$@"
}
exit_status () {
}
error() {
- error_noexit "$@"
+ report_error "$@"
exit 1
}
error_exit() {
- error "$@"
+ report_error "$@"
+ exit 1
}
# use only if we are ignoring failures for this test, bugno required.
error_ignore() {
local TYPE="IGNORE ($1)"
shift
- error_noexit "$@"
+ report_error "$@"
}
error_and_remount() {
- error_noexit "$@"
+ report_error "$@"
remount_client $MOUNT
exit 1
}
log "$msg== $(date +"%H:%M:%S (%s)")"
}
+check_dmesg_for_errors() {
+ local res
+ local errors="VFS: Busy inodes after unmount of\|\
+ldiskfs_check_descriptors: Checksum for group 0 failed\|\
+group descriptors corrupted"
+
+ res=$(do_nodes $(comma_list $(nodes_list)) "dmesg" | grep "$errors")
+ [ -z "$res" ] && return 0
+ echo "Kernel error detected: $res"
+ return 1
+}
+
#
# Run a single test function and cleanup after it.
#
local SAVE_UMASK=`umask`
umask 0022
+ if ! grep -q $DIR /proc/mounts; then
+ $SETUP
+ fi
+
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 $?"
check_catastrophe || error "LBUG/LASSERT detected"
+ check_dmesg_for_errors || error "Error in dmesg detected"
if [ "$PARALLEL" != "yes" ]; then
ps auxww | grep -v grep | grep -q multiop &&
error "multiop still running"
unset tdir
unset tfile
umask $SAVE_UMASK
+ $CLEANUP
return 0
}
}
client_only () {
- [ "$CLIENTONLY" ] || [ "$CLIENTMODSONLY" = yes ]
-}
-
-is_patchless ()
-{
- lctl get_param version | grep -q patchless
+ [ "$CLIENTONLY" ] || [ "$CLIENTMODSONLY" = yes ]
}
check_versions () {
}
add_pool_to_list () {
- local fsname=${1%%.*}
- local poolname=${1##$fsname.}
+ local fsname=${1%%.*}
+ local poolname=${1##$fsname.}
- local listvar=${fsname}_CREATED_POOLS
- eval export ${listvar}=$(expand_list ${!listvar} $poolname)
+ local listvar=${fsname}_CREATED_POOLS
+ local temp=${listvar}=$(expand_list ${!listvar} $poolname)
+ eval export $temp
}
remove_pool_from_list () {
- local fsname=${1%%.*}
- local poolname=${1##$fsname.}
+ local fsname=${1%%.*}
+ local poolname=${1##$fsname.}
- local listvar=${fsname}_CREATED_POOLS
- eval export ${listvar}=$(exclude_items_from_list ${!listvar} $poolname)
+ local listvar=${fsname}_CREATED_POOLS
+ local temp=${listvar}=$(exclude_items_from_list ${!listvar} $poolname)
+ eval export $temp
}
destroy_pool_int() {
}
init_logging() {
- if [[ -n $YAML_LOG ]]; then
- return
- fi
- local SAVE_UMASK=`umask`
- umask 0000
-
- export YAML_LOG=${LOGDIR}/results.yml
- mkdir -p $LOGDIR
- init_clients_lists
-
- if [ ! -f $YAML_LOG ]; then # If the yaml log already exists then we will just append to it
- if check_shared_dir $LOGDIR; then
- touch $LOGDIR/shared
- echo "Logging to shared log directory: $LOGDIR"
- else
- echo "Logging to local directory: $LOGDIR"
- fi
+ [[ -n $YAML_LOG ]] && return
+ local save_umask=$(umask)
+ umask 0000
- yml_nodes_file $LOGDIR >> $YAML_LOG
- yml_results_file >> $YAML_LOG
- fi
+ export YAML_LOG=${LOGDIR}/results.yml
+ mkdir -p $LOGDIR
+ init_clients_lists
+
+ # If the yaml log already exists then we will just append to it
+ if [ ! -f $YAML_LOG ]; then
+ if check_shared_dir $LOGDIR; then
+ touch $LOGDIR/shared
+ echo "Logging to shared log directory: $LOGDIR"
+ else
+ echo "Logging to local directory: $LOGDIR"
+ fi
+
+ yml_nodes_file $LOGDIR >> $YAML_LOG
+ yml_results_file >> $YAML_LOG
+ fi
+
+ umask $save_umask
- 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_test() {