. $LUSTRE/tests/test-framework.sh
init_test_env $@
+# use small MDS + OST size to speed formatting time
+MDSSIZE=40000
+OSTSIZE=40000
. ${CONFIG:=$LUSTRE/tests/cfg/$NAME.sh}
reformat() {
stop ost2 -f || return 93
}
+start_client() {
+ echo "start client on `facet_active_host client`"
+ start client || return 99
+}
+
+stop_client() {
+ echo "stop client on `facet_active_host client`"
+ stop client || return 100
+}
+
mount_client() {
local MOUNTPATH=$1
echo "mount $FSNAME on ${MOUNTPATH}....."
}
manual_umount_client(){
+ local rc
+ local FORCE=$1
echo "manual umount lustre on ${MOUNT}...."
- do_facet client "umount -d $MOUNT"
+ do_facet client "umount -d ${FORCE} $MOUNT"
+ rc=$?
+ return $rc
}
setup() {
kill -TERM $UMOUNT_PID
echo "waiting for umount to finish"
wait $UMOUNT_PID
- if grep " $MOUNT " /etc/mtab; then
- echo "test 5: mtab after failed umount"
+ if grep " $MOUNT " /proc/mounts; then
+ echo "test 5: /proc/mounts after failed umount"
umount $MOUNT &
UMOUNT_PID=$!
sleep 2
kill -TERM $UMOUNT_PID
echo "waiting for umount to finish"
wait $UMOUNT_PID
- grep " $MOUNT " /etc/mtab && echo "test 5: mtab after second umount" && return 11
+ grep " $MOUNT " /proc/mounts && echo "test 5: /proc/mounts after second umount" && return 11
fi
manual_umount_client
test_16() {
TMPMTPT="${MOUNT%/*}/conf16"
- if [ ! -f "$MDSDEV" ]; then
- echo "no $MDSDEV existing, so mount Lustre to create one"
- setup
+ if [ ! -e "$MDSDEV" ]; then
+ log "no $MDSDEV existing, so mount Lustre to create one"
+ setup
check_mount || return 41
cleanup || return $?
fi
- echo "change the mode of $MDSDEV/OBJECTS,LOGS,PENDING to 555"
+ [ -f "$MDSDEV" ] && LOOPOPT="-o loop"
+
+ log "change the mode of $MDSDEV/OBJECTS,LOGS,PENDING to 555"
do_facet mds "mkdir -p $TMPMTPT &&
- mount -o loop -t ext3 $MDSDEV $TMPMTPT &&
+ mount $LOOPOPT -t $FSTYPE $MDSDEV $TMPMTPT &&
chmod 555 $TMPMTPT/{OBJECTS,LOGS,PENDING} &&
umount $TMPMTPT" || return $?
- echo "mount Lustre to change the mode of OBJECTS/LOGS/PENDING, then umount Lustre"
+ log "mount Lustre to change the mode of OBJECTS/LOGS/PENDING, then umount Lustre"
setup
check_mount || return 41
cleanup || return $?
- echo "read the mode of OBJECTS/LOGS/PENDING and check if they has been changed properly"
+ log "read the mode of OBJECTS/LOGS/PENDING and check if they has been changed properly"
EXPECTEDOBJECTSMODE=`do_facet mds "debugfs -R 'stat OBJECTS' $MDSDEV 2> /dev/null" | grep 'Mode: ' | sed -e "s/.*Mode: *//" -e "s/ *Flags:.*//"`
EXPECTEDLOGSMODE=`do_facet mds "debugfs -R 'stat LOGS' $MDSDEV 2> /dev/null" | grep 'Mode: ' | sed -e "s/.*Mode: *//" -e "s/ *Flags:.*//"`
EXPECTEDPENDINGMODE=`do_facet mds "debugfs -R 'stat PENDING' $MDSDEV 2> /dev/null" | grep 'Mode: ' | sed -e "s/.*Mode: *//" -e "s/ *Flags:.*//"`
if [ "$EXPECTEDOBJECTSMODE" = "0777" ]; then
- echo "Success:Lustre change the mode of OBJECTS correctly"
+ log "Success:Lustre change the mode of OBJECTS correctly"
else
- echo "Error: Lustre does not change mode of OBJECTS properly"
- return 1
+ error "Lustre does not change mode of OBJECTS properly"
fi
if [ "$EXPECTEDLOGSMODE" = "0777" ]; then
- echo "Success:Lustre change the mode of LOGS correctly"
+ log "Success:Lustre change the mode of LOGS correctly"
else
- echo "Error: Lustre does not change mode of LOGS properly"
- return 1
+ error "Lustre does not change mode of LOGS properly"
fi
if [ "$EXPECTEDPENDINGMODE" = "0777" ]; then
- echo "Success:Lustre change the mode of PENDING correctly"
+ log "Success:Lustre change the mode of PENDING correctly"
else
- echo "Error: Lustre does not change mode of PENDING properly"
- return 1
+ error "Lustre does not change mode of PENDING properly"
fi
}
run_test 16 "verify that lustre will correct the mode of OBJECTS/LOGS/PENDING"
test_17() {
- if [ ! -f "$MDSDEV" ]; then
+ if [ ! -e "$MDSDEV" ]; then
echo "no $MDSDEV existing, so mount Lustre to create one"
setup
check_mount || return 41
check_mount || return 41
echo "check journal size..."
- FOUNDJOURNALSIZE=`do_facet mds "debugfs -R 'stat <8>' $MDSDEV" | awk '/Size: / { print $NF; exit;}'`
- if [ "$FOUNDJOURNALSIZE" = "79691776" ]; then
- echo "Success:lconf creates large journals"
+ FOUNDSIZE=`do_facet mds "debugfs -c -R 'stat <8>' $MDSDEV" | awk '/Size: / { print $NF; exit;}'`
+ if [ $FOUNDSIZE -gt $((32 * 1024 * 1024)) ]; then
+ log "Success: mkfs creates large journals"
else
- echo "Error:lconf not create large journals correctly"
- echo "expected journal size: 79691776(76M), found journal size: $FOUNDJOURNALSIZE"
- return 1
+ error "expected journal size > 32M, found $((FOUNDSIZE >> 20))M"
fi
cleanup || return $?
MDSSIZE=$OLDMDSSIZE
gen_config
}
-run_test 18 "check lconf creates large journals"
+run_test 18 "check mkfs creates large journals"
test_19a() {
start_mds || return 1
test_24a() {
local fs2mds_HOST=$mds_HOST
+ local fs2ost_HOST=$ost_HOST
+
+ [ -z "$fs2ost_DEV" -o -z "$fs2mds_DEV" ] && [ -b "$MDSDEV" ] && \
+ log "mixed loopback and real device not working" && return
+
+ local fs2mdsdev=${fs2mds_DEV:-${MDSDEV}_2}
+ local fs2ostdev=${fs2ost_DEV:-$(ostdevname 1)_2}
+
# test 8-char fsname as well
local FSNAME2=test1234
- add fs2mds $MDS_MKFS_OPTS --fsname=${FSNAME2} --nomgs --mgsnode=$MGSNID --reformat ${MDSDEV}_2 || exit 10
+ add fs2mds $MDS_MKFS_OPTS --fsname=${FSNAME2} --nomgs --mgsnode=$MGSNID --reformat $fs2mdsdev || exit 10
- local fs2ost_HOST=$ost_HOST
- local fs2ostdev=$(ostdevname 1)_2
add fs2ost $OST_MKFS_OPTS --fsname=${FSNAME2} --reformat $fs2ostdev || exit 10
setup
- start fs2mds ${MDSDEV}_2 $MDS_MOUNT_OPTS
+ start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS
start fs2ost $fs2ostdev $OST_MOUNT_OPTS
mkdir -p $MOUNT2
mount -t lustre $MGSNID:/${FSNAME2} $MOUNT2 || return 1
umount_client $MOUNT
# the MDS must remain up until last MDT
stop_mds
- MDS=$(awk '($3 ~ "mdt" && $4 ~ "MDS") { print $4 }' $LPROC/devices)
+ MDS=$(do_facet mds "cat $LPROC/devices" | awk '($3 ~ "mdt" && $4 ~ "MDS") { print $4 }')
[ -z "$MDS" ] && error "No MDS" && return 8
umount $MOUNT2
stop fs2mds -f
test_24b() {
local fs2mds_HOST=$mds_HOST
- add fs2mds $MDS_MKFS_OPTS --fsname=${FSNAME}2 --mgs --reformat ${MDSDEV}_2 || exit 10
+ [ -z "$fs2mds_DEV" ] && [ -b "$MDSDEV" ] && \
+ log "mixed loopback and real device not working" && return
+
+ local fs2mdsdev=${fs2mds_DEV:-${MDSDEV}_2}
+
+ add fs2mds $MDS_MKFS_OPTS --fsname=${FSNAME}2 --mgs --reformat $fs2mdsdev || exit 10
setup
- start fs2mds ${MDSDEV}_2 $MDS_MOUNT_OPTS && return 2
+ start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS && return 2
cleanup || return 6
}
run_test 24b "Multiple MGSs on a single node (should return err)"
test_26() {
load_modules
# we need modules before mount for sysctl, so make sure...
- [ -z "$(lsmod | grep lustre)" ] && modprobe lustre
+ do_facet mds "lsmod | grep -q lustre || modprobe lustre"
#define OBD_FAIL_MDS_FS_SETUP 0x135
do_facet mds "sysctl -w lustre.fail_loc=0x80000135"
start_mds && echo MDS started && return 1
run_test 26 "MDT startup failure cleans LOV (should return errs)"
set_and_check() {
- local TEST=$1
- local PARAM=$2
- local ORIG=$($TEST)
- if [ $# -gt 2 ]; then
- local FINAL=$3
+ 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"
- $LCTL conf_param $PARAM=$FINAL
+ do_facet mds "$LCTL conf_param $PARAM=$FINAL" || error conf_param failed
local RESULT
local MAX=30
local WAIT=0
while [ 1 ]; do
sleep 5
- RESULT=$($TEST)
+ RESULT=$(do_facet $myfacet "$TEST")
if [ $RESULT -eq $FINAL ]; then
echo "Updated config after $WAIT sec (got $RESULT)"
break
start_mds || return 2
echo "Requeue thread should have started: "
ps -e | grep ll_cfg_requeue
- set_and_check "cat $LPROC/obdfilter/$FSNAME-OST0000/client_cache_seconds" "$FSNAME-OST0000.ost.client_cache_seconds" || return 3
+ set_and_check ost1 "cat $LPROC/obdfilter/$FSNAME-OST0000/client_cache_seconds" "$FSNAME-OST0000.ost.client_cache_seconds" || return 3
cleanup_nocli
}
run_test 27a "Reacquire MGS lock if OST started first"
test_27b() {
setup
facet_failover mds
- set_and_check "cat $LPROC/mds/$FSNAME-MDT0000/group_acquire_expire" "$FSNAME-MDT0000.mdt.group_acquire_expire" || return 3
- set_and_check "cat $LPROC/mdc/$FSNAME-MDT0000-mdc-*/max_rpcs_in_flight" "$FSNAME-MDT0000.mdc.max_rpcs_in_flight" || return 4
+ set_and_check mds "cat $LPROC/mds/$FSNAME-MDT0000/group_acquire_expire" "$FSNAME-MDT0000.mdt.group_acquire_expire" || return 3
+ set_and_check client "cat $LPROC/mdc/$FSNAME-MDT0000-mdc-*/max_rpcs_in_flight" "$FSNAME-MDT0000.mdc.max_rpcs_in_flight" || return 4
cleanup
}
run_test 27b "Reacquire MGS lock after failover"
ORIG=$($TEST)
declare -i FINAL
FINAL=$(($ORIG + 10))
- set_and_check "$TEST" "$FSNAME.llite.max_read_ahead_whole_mb" || return 3
- set_and_check "$TEST" "$FSNAME.llite.max_read_ahead_whole_mb" || return 3
+ set_and_check client "$TEST" "$FSNAME.llite.max_read_ahead_whole_mb" || return 3
+ set_and_check client "$TEST" "$FSNAME.llite.max_read_ahead_whole_mb" || return 3
umount_client $MOUNT || return 200
mount_client $MOUNT
RESULT=$($TEST)
fi
ACTV=$(cat $PROC_ACT)
DEAC=$((1 - $ACTV))
- set_and_check "cat $PROC_ACT" "$PARAM" $DEAC || return 2
+ set_and_check client "cat $PROC_ACT" "$PARAM" $DEAC || return 2
# also check ost_server_uuid status
RESULT=$(grep DEACTIV $PROC_UUID)
if [ -z "$RESULT" ]; then
# check MDT too
local MPROC="$LPROC/osc/$FSNAME-OST0001-osc/active"
- if [ -r $MPROC ]; then
- RESULT=$(cat $MPROC)
- if [ $RESULT -ne $DEAC ]; then
- echo "MDT not deactivated: $(cat $MPROC)"
- return 4
- fi
- echo "MDT deactivated also"
+ RESULT=`do_facet mds " [ -r $MPROC ] && cat $MPROC"`
+ [ ${PIPESTATUS[0]} = 0 ] || error "Can't read $MPROC"
+ if [ $RESULT -ne $DEAC ]; then
+ echo "MDT not deactivated: $RESULT"
+ return 4
+ else
+ echo "MDT deactivated: also"
fi
-
# test new client starts deactivated
umount_client $MOUNT || return 200
mount_client $MOUNT
fi
# make sure it reactivates
- set_and_check "cat $PROC_ACT" "$PARAM" $ACTV || return 6
+ set_and_check client "cat $PROC_ACT" "$PARAM" $ACTV || return 6
umount_client $MOUNT
stop_ost2
TEST="cat $LPROC/llite/$FSNAME-*/max_read_ahead_whole_mb"
ORIG=$($TEST)
for i in $(seq 1 20); do
- set_and_check "$TEST" "$FSNAME.llite.max_read_ahead_whole_mb" $i || return 3
+ set_and_check client "$TEST" "$FSNAME.llite.max_read_ahead_whole_mb" $i || return 3
done
# make sure client restart still works
umount_client $MOUNT
mount_client $MOUNT || return 4
[ "$($TEST)" -ne "$i" ] && return 5
- set_and_check "$TEST" "$FSNAME.llite.max_read_ahead_whole_mb" $ORIG || return 6
+ set_and_check client "$TEST" "$FSNAME.llite.max_read_ahead_whole_mb" $ORIG || return 6
cleanup
}
run_test 30 "Big config llog"
run_test 31 "Connect to non-existent node (shouldn't crash)"
test_32a() {
+ # XXX - make this run on client-only systems with real hardware on
+ # the OST and MDT
+ # there appears to be a lot of assumption here about loopback
+ # devices
+ # or maybe this test is just totally useless on a client-only system
[ -z "$TUNEFS" ] && echo "No tunefs" && return
[ ! -r disk1_4.zip ] && echo "Cant find disk1_4.zip, skipping" && return
unzip -o -j -d $TMP/$tdir disk1_4.zip || { echo "Cant unzip disk1_4, skipping" && return ; }
echo "ok."
cleanup_nocli
+ load_modules
# mount a second time to make sure we didnt leave upgrade flag on
$TUNEFS --dryrun $TMP/$tdir/mds || error "tunefs failed"
+ load_modules
start mds $TMP/$tdir/mds "-o loop,exclude=lustre-OST0000" || return 12
cleanup_nocli
run_test 32a "Upgrade from 1.4 (not live)"
test_32b() {
+ # XXX - make this run on client-only systems with real hardware on
+ # the OST and MDT
+ # there appears to be a lot of assumption here about loopback
+ # devices
+ # or maybe this test is just totally useless on a client-only system
[ -z "$TUNEFS" ] && echo "No tunefs" && return
[ ! -r disk1_4.zip ] && echo "Cant find disk1_4.zip, skipping" && return
unzip -o -j -d $TMP/$tdir disk1_4.zip || { echo "Cant unzip disk1_4, skipping" && return ; }
# a fully-functioning client
echo "Check client and old fs contents"
mount_client $MOUNT
- set_and_check "cat $LPROC/mdc/*/max_rpcs_in_flight" "lustre-MDT0000.mdc.max_rpcs_in_flight" || return 11
+ set_and_check client "cat $LPROC/mdc/*/max_rpcs_in_flight" "lustre-MDT0000.mdc.max_rpcs_in_flight" || return 11
[ "$(cksum $MOUNT/passwd | cut -d' ' -f 1,2)" == "2479747619 779" ] || return 12
echo "ok."
}
run_test 32b "Upgrade from 1.4 with writeconf"
+test_33() { # bug 12333
+ local FSNAME2=test1234
+ local fs2mds_HOST=$mds_HOST
+ local fs2ost_HOST=$ost_HOST
+
+ [ -z "$fs2ost_DEV" -o -z "$fs2mds_DEV" ] && [ -b "$MDSDEV" ] && \
+ log "mixed loopback and real device not working" && return
+
+ local fs2mdsdev=${fs2mds_DEV:-${MDSDEV}_2}
+ local fs2ostdev=${fs2ost_DEV:-$(ostdevname 1)_2}
+ add fs2mds $MDS_MKFS_OPTS --fsname=${FSNAME2} --reformat $fs2mdsdev || exit 10
+ add fs2ost $OST_MKFS_OPTS --fsname=${FSNAME2} --index=8191 --mgsnode=`hostname`@tcp --reformat $fs2ostdev || exit 10
+
+ start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS
+ start fs2ost $fs2ostdev $OST_MOUNT_OPTS
+ mkdir -p $MOUNT2
+ mount -t lustre $MGSNID:/${FSNAME2} $MOUNT2 || return 1
+ echo "ok."
+
+ umount -d $MOUNT2
+ stop fs2ost -f
+ stop fs2mds -f
+ rm -rf $MOUNT2 $fs2mdsdev $fs2ostdev
+ cleanup_nocli || return 6
+}
+run_test 33 "Mount ost with a large index number"
+
umount_client $MOUNT
cleanup_nocli
+test_33a() {
+ setup
+
+ do_facet client dd if=/dev/zero of=$MOUNT/24 bs=1024k count=1
+ # Drop lock cancelation reply during umount
+ #define OBD_FAIL_LDLM_CANCEL 0x304
+ do_facet client sysctl -w lustre.fail_loc=0x80000304
+ #sysctl -w lnet.debug=-1
+ umount_client $MOUNT
+ cleanup
+}
+run_test 33a "Drop cancel during umount"
+
+test_34a() {
+ setup
+ do_facet client multiop $DIR/file O_c &
+
+ manual_umount_client
+ rc=$?
+ do_facet client killall -USR1 multiop
+ if [ $rc -eq 0 ]; then
+ error "umount not fail!"
+ fi
+ sleep 1
+ cleanup
+}
+run_test 34a "umount with opened file should be fail"
+
+
+test_34b() {
+ setup
+ touch $DIR/$tfile || return 1
+ stop_mds --force || return 2
+
+ manual_umount_client --force
+ rc=$?
+ if [ $rc -ne 0 ]; then
+ error "mtab after failed umount - rc $rc"
+ fi
+
+ cleanup
+ return 0
+}
+run_test 34b "force umount with failed mds should be normal"
+
+test_34c() {
+ setup
+ touch $DIR/$tfile || return 1
+ stop_ost --force || return 2
+
+ manual_umount_client --force
+ rc=$?
+ if [ $rc -ne 0 ]; then
+ error "mtab after failed umount - rc $rc"
+ fi
+
+ cleanup
+ return 0
+}
+run_test 34c "force umount with failed mds should be normal"
+
+test_35() { # bug 12459
+ setup
+
+ debugsave
+ sysctl -w lnet.debug="ha"
+
+ log "Set up a fake failnode for the MDS"
+ FAKENID="127.0.0.2"
+ $LCTL conf_param ${FSNAME}-MDT0000.failover.node=$FAKENID || return 4
+
+ log "Wait for RECONNECT_INTERVAL seconds (10s)"
+ sleep 10
+
+ MSG="conf-sanity.sh test_33 `date +%F%kh%Mm%Ss`"
+ $LCTL clear
+ log "$MSG"
+ log "Stopping the MDT:"
+ stop_mds || return 5
+
+ df $MOUNT > /dev/null 2>&1 &
+ DFPID=$!
+ log "Restarting the MDT:"
+ start_mds || return 6
+ log "Wait for df ($DFPID) ... "
+ wait $DFPID
+ log "done"
+ debugrestore
+
+ # retrieve from the log the first server that the client tried to
+ # contact after the connection loss
+ $LCTL dk $TMP/lustre-log-$TESTNAME.log
+ NEXTCONN=`awk "/${MSG}/ {start = 1;}
+ /import_select_connection.*${FSNAME}-MDT0000-mdc.* using connection/ {
+ if (start) {
+ if (\\\$NF ~ /$FAKENID/)
+ print \\\$NF;
+ else
+ print 0;
+ exit;
+ }
+ }" $TMP/lustre-log-$TESTNAME.log`
+ [ "$NEXTCONN" != "0" ] && log "The client didn't try to reconnect to the last active server (tried ${NEXTCONN} instead)" && return 7
+ cleanup
+}
+run_test 35 "Reconnect to the last active server first"
+
equals_msg "Done"
echo "$0: completed"
git://git.whamcloud.com / fs / lustre-release.git / blobdiff