set -e
+ONLY=${ONLY:-"$*"}
+
+# These tests don't apply to mountconf
+MOUNTCONFSKIP="10 11 12 13 13b 14 15"
+# bug number for skipped test: 13739
+HEAD_EXCEPT=" 32a 32b "
+
+# bug number for skipped test: 10510 12743
+ALWAYS_EXCEPT=" $CONF_SANITY_EXCEPT $MOUNTCONFSKIP $HEAD_EXCEPT 23 36"
+# UPDATE THE COMMENT ABOVE WITH BUG NUMBERS WHEN CHANGING ALWAYS_EXCEPT!
+
+#
+[ "$SLOW" = "no" ] && EXCEPT_SLOW="0 1 2 3 6 7 15 18 24b 25 30 31 32 33 34a "
+
SRCDIR=`dirname $0`
PATH=$PWD/$SRCDIR:$SRCDIR:$SRCDIR/../utils:$PATH
+PTLDEBUG=${PTLDEBUG:--1}
+SAVE_PWD=$PWD
LUSTRE=${LUSTRE:-`dirname $0`/..}
RLUSTRE=${RLUSTRE:-$LUSTRE}
MOUNTLUSTRE=${MOUNTLUSTRE:-/sbin/mount.lustre}
+MKFSLUSTRE=${MKFSLUSTRE:-/usr/sbin/mkfs.lustre}
+HOSTNAME=`hostname`
. $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/local.sh}
-gen_config() {
- rm -f $XMLCONFIG
-
- add_mds mds1 --dev $MDSDEV --size $MDSSIZE
- add_lov lov1 mds1 --stripe_sz $STRIPE_BYTES\
- --stripe_cnt $STRIPES_PER_OBJ --stripe_pattern 0
- add_ost ost --lov lov1 --dev $OSTDEV --size $OSTSIZE
- add_client client mds1 --lov lov1 --path $MOUNT
+reformat() {
+ formatall
}
-gen_second_config() {
- rm -f $XMLCONFIG
+writeconf() {
+ local facet=mds
+ shift
+ stop ${facet} -f
+ rm -f ${facet}active
+ # who knows if/where $TUNEFS is installed? Better reformat if it fails...
+ do_facet ${facet} "$TUNEFS --writeconf $MDSDEV" || echo "tunefs failed, reformatting instead" && reformat
+}
- add_mds mds2 --dev $MDSDEV --size $MDSSIZE
- add_lov lov2 mds2 --stripe_sz $STRIPE_BYTES\
- --stripe_cnt $STRIPES_PER_OBJ --stripe_pattern 0
- add_ost ost2 --lov lov2 --dev $OSTDEV --size $OSTSIZE
- add_client client --mds mds2 --lov lov2 --path $MOUNT2
+gen_config() {
+ reformat
+ # The MGS must be started before the OSTs for a new fs, so start
+ # and stop to generate the startup logs.
+ start_mds
+ start_ost
+ sleep 5
+ stop_ost
+ stop_mds
}
start_mds() {
- echo "start mds1 service on `facet_active_host mds1`"
- start mds1 --reformat $MDSLCONFARGS || return 94
- start_lsvcgssd || return 501
+ echo "start mds service on `facet_active_host mds`"
+ start mds $MDSDEV $MDS_MOUNT_OPTS || return 94
}
stop_mds() {
- echo "stop mds1 service on `facet_active_host mds1`"
- stop mds1 $@ || return 97
- stop_lsvcgssd
+ echo "stop mds service on `facet_active_host mds`"
+ # These tests all use non-failover stop
+ stop mds -f || return 97
}
start_ost() {
- echo "start ost service on `facet_active_host ost`"
- start ost --reformat $OSTLCONFARGS || return 95
+ echo "start ost1 service on `facet_active_host ost1`"
+ start ost1 `ostdevname 1` $OST_MOUNT_OPTS || return 95
}
stop_ost() {
- echo "stop ost service on `facet_active_host ost`"
- stop ost $@ || return 98
+ echo "stop ost1 service on `facet_active_host ost1`"
+ # These tests all use non-failover stop
+ stop ost1 -f || return 98
+}
+
+start_ost2() {
+ echo "start ost2 service on `facet_active_host ost2`"
+ start ost2 `ostdevname 2` $OST_MOUNT_OPTS || return 92
+}
+
+stop_ost2() {
+ echo "stop ost2 service on `facet_active_host ost2`"
+ # These tests all use non-failover stop
+ stop ost2 -f || return 93
}
mount_client() {
local MOUNTPATH=$1
- start_lgssd || return 502
- echo "mount lustre on ${MOUNTPATH}....."
+ echo "mount $FSNAME on ${MOUNTPATH}....."
+ zconf_mount `hostname` $MOUNTPATH || return 96
+}
+
+remount_client() {
+ local SAVEMOUNTOPT=$MOUNTOPT
+ MOUNTOPT="remount,$1"
+ local MOUNTPATH=$2
+ echo "remount '$1' lustre on ${MOUNTPATH}....."
zconf_mount `hostname` $MOUNTPATH || return 96
+ MOUNTOPT=$SAVEMOUNTOPT
}
umount_client() {
local MOUNTPATH=$1
echo "umount lustre on ${MOUNTPATH}....."
- zconf_umount `hostname` $MOUNTPATH || return 97
- stop_lgssd
+ zconf_umount `hostname` $MOUNTPATH || return 97
}
manual_umount_client(){
- echo "manual umount lustre on ${MOUNTPATH}...."
- do_facet client "umount $MOUNT"
- stop_lgssd
+ echo "manual umount lustre on ${MOUNT}...."
+ do_facet client "umount -d $MOUNT"
}
setup() {
mount_client $MOUNT
}
-cleanup() {
- umount_client $MOUNT || return 200
+cleanup_nocli() {
stop_mds || return 201
stop_ost || return 202
- # catch case where these return just fine, but modules are still not unloaded
- /sbin/lsmod | grep -q portals
- if [ 1 -ne $? ]; then
- echo "modules still loaded..."
- return 203
- fi
+ unload_modules || return 203
+}
+
+cleanup() {
+ umount_client $MOUNT || return 200
+ cleanup_nocli || return $?
}
check_mount() {
- do_facet client "touch $DIR/a" || return 71
- do_facet client "rm $DIR/a" || return 72
+ do_facet client "cp /etc/passwd $DIR/a" || return 71
+ do_facet client "rm $DIR/a" || return 72
+ # make sure lustre is actually mounted (touch will block,
+ # but grep won't, so do it after)
+ do_facet client "grep $MOUNT' ' /proc/mounts > /dev/null" || return 73
echo "setup single mount lustre success"
}
build_test_filter
+if [ "$ONLY" == "setup" ]; then
+ setup
+ exit
+fi
+
+if [ "$ONLY" == "cleanup" ]; then
+ cleanup
+ exit
+fi
+
#create single point mountpoint
gen_config
-start_krb5_kdc || exit 1
+init_krb5_env
test_0() {
- start_ost
- start_mds
- mount_client $MOUNT
+ setup
check_mount || return 41
cleanup || return $?
}
test_1() {
start_ost
echo "start ost second time..."
- start ost --reformat $OSTLCONFARGS
- start_mds
- mount_client $MOUNT
+ setup
check_mount || return 42
cleanup || return $?
}
-run_test 1 "start up ost twice"
+run_test 1 "start up ost twice (should return errors)"
test_2() {
start_ost
start_mds
echo "start mds second time.."
- start mds1 --reformat $MDSLCONFARGS
-
+ start_mds
mount_client $MOUNT
check_mount || return 43
cleanup || return $?
}
-run_test 2 "start up mds twice"
+run_test 2 "start up mds twice (should return err)"
test_3() {
setup
- mount_client $MOUNT
-
+ #mount.lustre returns an error if already in mtab
+ mount_client $MOUNT && return $?
check_mount || return 44
-
- umount_client $MOUNT
- cleanup || return $?
+ cleanup || return $?
}
-run_test 3 "mount client twice"
+run_test 3 "mount client twice (should return err)"
test_4() {
setup
touch $DIR/$tfile || return 85
- stop_ost --force
+ stop_ost -f
cleanup
eno=$?
# ok for ost to fail shutdown
test_5() {
setup
touch $DIR/$tfile || return 1
- stop_mds --force || return 2
+ stop_mds -f || return 2
# cleanup may return an error from the failed
# disconnects; for now I'll consider this successful
# if all the modules have unloaded.
-
- # as MDS is down, umount without -f may cause blocking
- # and this test will never finish. Blocking is possible
- # as umount may want to cancel locks with RPC's and these
- # RPC's will wait forever, as pinger thread will try to
- # recover failed import endlessly.
- #
- # Thus, main point is: nobody should expect umount finish
- # quickly and cleanly without -f flag when MDS or OST is
- # down for sure. --umka
- umount -f $MOUNT &
+ umount -d $MOUNT &
UMOUNT_PID=$!
- sleep 2
+ sleep 6
echo "killing umount"
kill -TERM $UMOUNT_PID
echo "waiting for umount to finish"
wait $UMOUNT_PID
- stop_lgssd
+ if grep " $MOUNT " /etc/mtab; then
+ echo "test 5: mtab after failed umount"
+ umount $MOUNT &
+ UMOUNT_PID=$!
+ sleep 2
+ echo "killing umount"
+ 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
+ fi
- # cleanup client modules
- $LCONF --force --cleanup --nosetup --node client_facet $XMLCONFIG > /dev/null
-
+ manual_umount_client
# stop_mds is a no-op here, and should not fail
- stop_mds || return 4
-
- # this should have --force flag specified, as umount -f
- # will skip disconnect phase and thus OST will have one
- # extra refcount what will cause class_cleanup() failure
- # if --force is not specified. --umka
- stop_ost --force || return 5
-
- lsmod | grep -q portals && return 6
- return 0
+ cleanup_nocli || return $?
+ # df may have lingering entry
+ manual_umount_client
+ # mtab may have lingering entry
+ grep -v $MOUNT" " /etc/mtab > $TMP/mtabtemp
+ mv $TMP/mtabtemp /etc/mtab
}
-run_test 5 "force cleanup mds, then cleanup --force"
+run_test 5 "force cleanup mds, then cleanup"
test_5b() {
start_ost
- start_mds
- stop_mds
-
[ -d $MOUNT ] || mkdir -p $MOUNT
- $LCONF --nosetup --node client_facet $XMLCONFIG > /dev/null
- start_lgssd || return 1
- llmount -o nettype=$NETTYPE $mds_HOST://mds_svc/client_facet $MOUNT && exit 2
-
- # cleanup client modules
- $LCONF --cleanup --nosetup --node client_facet $XMLCONFIG > /dev/null
- stop_lgssd
-
+ grep " $MOUNT " /etc/mtab && echo "test 5b: mtab before mount" && return 10
+ mount_client $MOUNT && return 1
+ grep " $MOUNT " /etc/mtab && echo "test 5b: mtab after failed mount" && return 11
+ umount_client $MOUNT
# stop_mds is a no-op here, and should not fail
- stop_mds || return 3
- stop_ost || return 4
-
- lsmod | grep -q portals && return 5
+ cleanup_nocli || return $?
return 0
-
}
-run_test 5b "mds down, cleanup after failed mount (bug 2712)"
+run_test 5b "mds down, cleanup after failed mount (bug 2712) (should return errs)"
test_5c() {
start_ost
start_mds
-
[ -d $MOUNT ] || mkdir -p $MOUNT
- $LCONF --nosetup --node client_facet $XMLCONFIG > /dev/null
- start_lgssd || return 1
- llmount -o nettype=$NETTYPE $mds_HOST://wrong_mds_svc/client_facet $MOUNT && return 2
-
- # cleanup client modules
- $LCONF --cleanup --nosetup --node client_facet $XMLCONFIG > /dev/null
- stop_lgssd
-
- stop_mds || return 3
- stop_ost || return 4
-
- lsmod | grep -q portals && return 5
- return 0
-
+ grep " $MOUNT " /etc/mtab && echo "test 5c: mtab before mount" && return 10
+ local oldfs="${FSNAME}"
+ FSNAME="wrong.${FSNAME}"
+ mount_client $MOUNT || :
+ FSNAME=${oldfs}
+ grep " $MOUNT " /etc/mtab && echo "test 5c: mtab after failed mount" && return 11
+ umount_client $MOUNT
+ cleanup_nocli || return $?
}
-run_test 5c "cleanup after failed mount (bug 2712)"
+run_test 5c "cleanup after failed mount (bug 2712) (should return errs)"
test_5d() {
- start_ost
- start_mds
- stop_ost --force
-
- [ -d $MOUNT ] || mkdir -p $MOUNT
- $LCONF --nosetup --node client_facet $XMLCONFIG > /dev/null
- start_lgssd || return 1
- llmount $mds_HOST://mds1_svc/client_facet $MOUNT || return 1
-
- umount $MOUNT || return 2
- # cleanup client modules
- $LCONF --cleanup --nosetup --node client_facet $XMLCONFIG > /dev/null
- stop_lgssd
-
- stop_mds || return 3
+ start_ost
+ start_mds
+ stop_ost -f
+ grep " $MOUNT " /etc/mtab && echo "test 5d: mtab before mount" && return 10
+ mount_client $MOUNT || return 1
+ cleanup || return $?
+ grep " $MOUNT " /etc/mtab && echo "test 5d: mtab after unmount" && return 11
+ return 0
+}
+run_test 5d "mount with ost down"
- lsmod | grep -q portals && return 4
- return 0
+test_5e() {
+ start_ost
+ start_mds
+#define OBD_FAIL_PTLRPC_DELAY_SEND 0x506
+ do_facet client "sysctl -w lustre.fail_loc=0x80000506"
+ grep " $MOUNT " /etc/mtab && echo "test 5e: mtab before mount" && return 10
+ mount_client $MOUNT || echo "mount failed (not fatal)"
+ cleanup || return $?
+ grep " $MOUNT " /etc/mtab && echo "test 5e: mtab after unmount" && return 11
+ return 0
}
-run_test 5d "ost down, don't crash during mount attempt"
+run_test 5e "delayed connect, don't crash (bug 10268)"
test_6() {
setup
test_7() {
setup
manual_umount_client
- cleanup || return $?
+ cleanup_nocli || return $?
}
run_test 7 "manual umount, then cleanup"
test_8() {
- start_ost
- start_mds
-
- mount_client $MOUNT
+ setup
mount_client $MOUNT2
-
check_mount2 || return 45
- umount $MOUNT
umount_client $MOUNT2
-
- stop_mds
- stop_ost
+ cleanup || return $?
}
run_test 8 "double mount setup"
test_9() {
- # backup the old values of PTLDEBUG and SUBSYSTEM
- OLDPTLDEBUG=$PTLDEBUG
- OLDSUBSYSTEM=$SUBSYSTEM
-
- # generate new configuration file with lmc --ptldebug and --subsystem
- PTLDEBUG="trace"
- SUBSYSTEM="mdc"
- gen_config
-
- # check the result of lmc --ptldebug/subsystem
start_ost
- start_mds
- mount_client $MOUNT
- CHECK_PTLDEBUG="`cat /proc/sys/portals/debug`"
- if [ $CHECK_PTLDEBUG = "1" ]; then
- echo "lmc --debug success"
- else
- echo "lmc --debug: want 1, have $CHECK_PTLDEBUG"
- return 1
- fi
- CHECK_SUBSYSTEM="`cat /proc/sys/portals/subsystem_debug`"
- if [ $CHECK_SUBSYSTEM = "2" ]; then
- echo "lmc --subsystem success"
- else
- echo "lmc --subsystem: want 2, have $CHECK_SUBSYSTEM"
- return 1
- fi
- check_mount || return 41
- cleanup || return $?
- # the new PTLDEBUG/SUBSYSTEM used for lconf --ptldebug/subsystem
- PTLDEBUG="inode+trace"
- SUBSYSTEM="mds+ost"
+ do_facet ost1 sysctl lnet.debug=\'inode trace\' || return 1
+ do_facet ost1 sysctl lnet.subsystem_debug=\'mds ost\' || return 1
- # check lconf --ptldebug/subsystem overriding lmc --ptldebug/subsystem
- start_ost
- start_mds
- CHECK_PTLDEBUG="`do_facet mds1 cat /proc/sys/portals/debug`"
- if [ $CHECK_PTLDEBUG = "3" ]; then
- echo "lconf --debug success"
+ CHECK_PTLDEBUG="`do_facet ost1 sysctl -n lnet.debug`"
+ if [ "$CHECK_PTLDEBUG" ] && [ "$CHECK_PTLDEBUG" = "trace inode" ];then
+ echo "lnet.debug success"
else
- echo "lconf --debug: want 3, have $CHECK_PTLDEBUG"
+ echo "lnet.debug: want 'trace inode', have '$CHECK_PTLDEBUG'"
return 1
fi
- CHECK_SUBSYSTEM="`do_facet mds1 cat /proc/sys/portals/subsystem_debug`"
- if [ $CHECK_SUBSYSTEM = "20" ]; then
- echo "lconf --subsystem success"
+ CHECK_SUBSYS="`do_facet ost1 sysctl -n lnet.subsystem_debug`"
+ if [ "$CHECK_SUBSYS" ] && [ "$CHECK_SUBSYS" = "mds ost" ]; then
+ echo "lnet.subsystem_debug success"
else
- echo "lconf --subsystem: want 20, have $CHECK_SUBSYSTEM"
+ echo "lnet.subsystem_debug: want 'mds ost', have '$CHECK_SUBSYS'"
return 1
fi
- mount_client $MOUNT
- check_mount || return 41
- cleanup || return $?
-
- # resume the old configuration
- PTLDEBUG=$OLDPTLDEBUG
- SUBSYSTEM=$OLDSUBSYSTEM
- gen_config
+ stop_ost || return $?
}
-run_test 9 "test --ptldebug and --subsystem for lmc and lconf"
+run_test 9 "test ptldebug and subsystem for mkfs"
test_10() {
echo "generate configuration with the same name for node and mds"
OLDXMLCONFIG=$XMLCONFIG
XMLCONFIG="broken.xml"
[ -f "$XMLCONFIG" ] && rm -f $XMLCONFIG
- facet="mds1"
+ facet="mds"
rm -f ${facet}active
add_facet $facet
echo "the name for node and mds is the same"
add_ost ost --lov lov1 --dev $OSTDEV --size $OSTSIZE
facet="client"
add_facet $facet --lustre_upcall $UPCALL
- do_lmc --add mtpt --node ${facet}_facet --mds mds1_facet \
+ do_lmc --add mtpt --node ${facet}_facet --mds mds_facet \
--lov lov1 --path $MOUNT
echo "mount lustre"
XMLCONFIG="conf11.xml"
[ -f "$XMLCONFIG" ] && rm -f $XMLCONFIG
- add_mds mds1 --dev $MDSDEV --size $MDSSIZE
+ add_mds mds --dev $MDSDEV --size $MDSSIZE
add_ost ost --dev $OSTDEV --size $OSTSIZE
- add_client client mds1 --path $MOUNT --ost ost_svc || return $?
+ add_client client mds --path $MOUNT --ost ost_svc || return $?
echo "Default lov config success!"
[ -f "$XMLCONFIG" ] && rm -f $XMLCONFIG
- add_mds mds1 --dev $MDSDEV --size $MDSSIZE
+ add_mds mds --dev $MDSDEV --size $MDSSIZE
add_ost ost --dev $OSTDEV --size $OSTSIZE
- add_client client mds1 --path $MOUNT && return $?
+ add_client client mds --path $MOUNT && return $?
echo "--add mtpt with neither --lov nor --ost will return error"
echo ""
# test double quote
[ -f "$XMLCONFIG" ] && rm -f $XMLCONFIG
[ -f "$BATCHFILE" ] && rm -f $BATCHFILE
- echo "--add net --node localhost --nid localhost.localdomain --nettype tcp" > $BATCHFILE
- echo "--add mds --node localhost --mds mds1 --mkfsoptions \"-I 128\"" >> $BATCHFILE
+ echo "--add net --node $HOSTNAME --nid $HOSTNAME --nettype tcp" > $BATCHFILE
+ echo "--add mds --node $HOSTNAME --mds mds1 --mkfsoptions \"-I 128\"" >> $BATCHFILE
# --mkfsoptions "-I 128"
do_lmc -m $XMLCONFIG --batch $BATCHFILE || return $?
if [ `sed -n '/>-I 128</p' $XMLCONFIG | wc -l` -eq 1 ]; then
fi
rm -f $XMLCONFIG
rm -f $BATCHFILE
- echo "--add net --node localhost --nid localhost.localdomain --nettype tcp" > $BATCHFILE
- echo "--add mds --node localhost --mds mds1 --mkfsoptions \"-I 128" >> $BATCHFILE
+ echo "--add net --node $HOSTNAME --nid $HOSTNAME --nettype tcp" > $BATCHFILE
+ echo "--add mds --node $HOSTNAME --mds mds1 --mkfsoptions \"-I 128" >> $BATCHFILE
# --mkfsoptions "-I 128
do_lmc -m $XMLCONFIG --batch $BATCHFILE && return $?
echo "unmatched double quote should return error"
# test single quote
rm -f $BATCHFILE
- echo "--add net --node localhost --nid localhost.localdomain --nettype tcp" > $BATCHFILE
- echo "--add mds --node localhost --mds mds1 --mkfsoptions '-I 128'" >> $BATCHFILE
+ echo "--add net --node $HOSTNAME --nid $HOSTNAME --nettype tcp" > $BATCHFILE
+ echo "--add mds --node $HOSTNAME --mds mds1 --mkfsoptions '-I 128'" >> $BATCHFILE
# --mkfsoptions '-I 128'
do_lmc -m $XMLCONFIG --batch $BATCHFILE || return $?
if [ `sed -n '/>-I 128</p' $XMLCONFIG | wc -l` -eq 1 ]; then
fi
rm -f $XMLCONFIG
rm -f $BATCHFILE
- echo "--add net --node localhost --nid localhost.localdomain --nettype tcp" > $BATCHFILE
- echo "--add mds --node localhost --mds mds1 --mkfsoptions '-I 128" >> $BATCHFILE
+ echo "--add net --node $HOSTNAME --nid $HOSTNAME --nettype tcp" > $BATCHFILE
+ echo "--add mds --node $HOSTNAME --mds mds1 --mkfsoptions '-I 128" >> $BATCHFILE
# --mkfsoptions '-I 128
do_lmc -m $XMLCONFIG --batch $BATCHFILE && return $?
echo "unmatched single quote should return error"
# test backslash
rm -f $BATCHFILE
- echo "--add net --node localhost --nid localhost.localdomain --nettype tcp" > $BATCHFILE
- echo "--add mds --node localhost --mds mds1 --mkfsoptions \-\I\ \128" >> $BATCHFILE
+ echo "--add net --node $HOSTNAME --nid $HOSTNAME --nettype tcp" > $BATCHFILE
+ echo "--add mds --node $HOSTNAME --mds mds1 --mkfsoptions \-\I\ \128" >> $BATCHFILE
# --mkfsoptions \-\I\ \128
do_lmc -m $XMLCONFIG --batch $BATCHFILE || return $?
if [ `sed -n '/>-I 128</p' $XMLCONFIG | wc -l` -eq 1 ]; then
fi
rm -f $XMLCONFIG
rm -f $BATCHFILE
- echo "--add net --node localhost --nid localhost.localdomain --nettype tcp" > $BATCHFILE
- echo "--add mds --node localhost --mds mds1 --mkfsoptions -I\ 128\\" >> $BATCHFILE
+ echo "--add net --node $HOSTNAME --nid $HOSTNAME --nettype tcp" > $BATCHFILE
+ echo "--add mds --node $HOSTNAME --mds mds1 --mkfsoptions -I\ 128\\" >> $BATCHFILE
# --mkfsoptions -I\ 128\
do_lmc -m $XMLCONFIG --batch $BATCHFILE && return $?
echo "backslash followed by nothing should return error"
test_13() {
OLDXMLCONFIG=$XMLCONFIG
XMLCONFIG="conf13-1.xml"
- SECONDXMLCONFIG="conf13-2.xml"
# check long uuid will be truncated properly and uniquely
echo "To generate XML configuration file(with long ost name): $XMLCONFIG"
[ -f "$XMLCONFIG" ] && rm -f $XMLCONFIG
- do_lmc --add net --node localhost --nid localhost.localdomain --nettype tcp
- do_lmc --add mds --node localhost --mds mds1_name_longer_than_31characters
- do_lmc --add mds --node localhost --mds mds2_name_longer_than_31characters
+ do_lmc --add net --node $HOSTNAME --nid $HOSTNAME --nettype tcp
+ do_lmc --add mds --node $HOSTNAME --mds mds1_name_longer_than_31characters
+ do_lmc --add mds --node $HOSTNAME --mds mds2_name_longer_than_31characters
if [ ! -f "$XMLCONFIG" ]; then
echo "Error:no file $XMLCONFIG created!"
return 1
| sed "s/ /\n\r/g" | awk -F"'" '/uuid=/{print $2}'`
FOUNDMDS2UUID=`awk -F"'" '/<mds .*uuid=/' $XMLCONFIG | sed -n '2p' \
| sed "s/ /\n\r/g" | awk -F"'" '/uuid=/{print $2}'`
+ [ -z "$FOUNDMDS1UUID" ] && echo "MDS1 UUID empty" && return 1
+ [ -z "$FOUNDMDS2UUID" ] && echo "MDS2 UUID empty" && return 1
if ([ $EXPECTEDMDS1UUID = $FOUNDMDS1UUID ] && [ $EXPECTEDMDS2UUID = $FOUNDMDS2UUID ]) || \
([ $EXPECTEDMDS1UUID = $FOUNDMDS2UUID ] && [ $EXPECTEDMDS2UUID = $FOUNDMDS1UUID ]); then
echo "Success:long uuid truncated successfully and being unique."
echo "but: found uuid for mds1 and mds2: $FOUNDMDS1UUID; $FOUNDMDS2UUID"
return 1
fi
+ rm -f $XMLCONFIG
+ XMLCONFIG=$OLDXMLCONFIG
+}
+run_test 13 "check new_uuid of lmc operating correctly"
+test_13b() {
+ OLDXMLCONFIG=$XMLCONFIG
+ XMLCONFIG="conf13-1.xml"
+ SECONDXMLCONFIG="conf13-2.xml"
# check multiple invocations for lmc generate same XML configuration file
rm -f $XMLCONFIG
echo "Generate the first XML configuration file"
gen_config
echo "mv $XMLCONFIG to $SECONDXMLCONFIG"
- mv $XMLCONFIG $SECONDXMLCONFIG || return $?
+ sed -e "s/mtime[^ ]*//" $XMLCONFIG > $SECONDXMLCONFIG || return $?
echo "Generate the second XML configuration file"
gen_config
- if [ `diff $XMLCONFIG $SECONDXMLCONFIG | wc -l` -eq 0 ]; then
+ # don't compare .xml mtime, it will always be different
+ if [ `sed -e "s/mtime[^ ]*//" $XMLCONFIG | diff - $SECONDXMLCONFIG | wc -l` -eq 0 ]; then
echo "Success:multiple invocations for lmc generate same XML file"
else
echo "Error: multiple invocations for lmc generate different XML file"
return 1
fi
- rm -f $XMLCONFIG
- rm -f $SECONDXMLCONFIG
+ rm -f $XMLCONFIG $SECONDXMLCONFIG
XMLCONFIG=$OLDXMLCONFIG
}
-run_test 13 "check new_uuid of lmc operating correctly"
+run_test 13b "check lmc generates consistent .xml file"
test_14() {
rm -f $XMLCONFIG
# create xml file with --mkfsoptions for ost
echo "create xml file with --mkfsoptions for ost"
- add_mds mds1 --dev $MDSDEV --size $MDSSIZE
- add_lov lov1 mds1 --stripe_sz $STRIPE_BYTES\
+ add_mds mds --dev $MDSDEV --size $MDSSIZE
+ add_lov lov1 mds --stripe_sz $STRIPE_BYTES\
--stripe_cnt $STRIPES_PER_OBJ --stripe_pattern 0
add_ost ost --lov lov1 --dev $OSTDEV --size $OSTSIZE \
--mkfsoptions "-Llabel_conf_14"
- add_client client mds1 --lov lov1 --path $MOUNT
+ add_client client mds --lov lov1 --path $MOUNT
FOUNDSTRING=`awk -F"<" '/<mkfsoptions>/{print $2}' $XMLCONFIG`
EXPECTEDSTRING="mkfsoptions>-Llabel_conf_14"
- if [ $EXPECTEDSTRING != $FOUNDSTRING ]; then
+ if [ "$EXPECTEDSTRING" != "$FOUNDSTRING" ]; then
echo "Error: expected: $EXPECTEDSTRING; found: $FOUNDSTRING"
return 1
fi
start_ost
start_mds
mount_client $MOUNT || return $?
- if [ -z "`dumpe2fs -h $OSTDEV | grep label_conf_14`" ]; then
+ if [ -z "`do_facet ost1 dumpe2fs -h $OSTDEV | grep label_conf_14`" ]; then
echo "Error: the mkoptions not applied to mke2fs of ost."
return 1
fi
fi
}
+# this only tests the kernel mount command, not anything about lustre.
test_15() {
+ MOUNTLUSTRE=${MOUNTLUSTRE:-/sbin/mount.lustre}
start_ost
start_mds
- echo "mount lustre on ${MOUNT} with $MOUNTLUSTRE....."
+
+ echo "mount lustre on ${MOUNT} without $MOUNTLUSTRE....."
if [ -f "$MOUNTLUSTRE" ]; then
echo "save $MOUNTLUSTRE to $MOUNTLUSTRE.sav"
- mv $MOUNTLUSTRE $MOUNTLUSTRE.sav
+ mv $MOUNTLUSTRE $MOUNTLUSTRE.sav && trap cleanup_15 EXIT INT
+ if [ -f $MOUNTLUSTRE ]; then
+ skip "$MOUNTLUSTRE cannot be moved, skipping test"
+ return 0
+ fi
fi
- [ -f "$MOUNTLUSTRE" ] && echo "can't move $MOUNTLUSTRE" && return 40
- trap cleanup_15 EXIT INT
- [ ! `cp $LUSTRE/utils/llmount $MOUNTLUSTRE` ] || return $?
- do_node `hostname` mkdir -p $MOUNT 2> /dev/null
- # load llite module on the client if it isn't in /lib/modules
- do_node `hostname` $LCONF --nosetup --node client_facet $XMLCONFIG
- do_node `hostname` mount -t lustre -o nettype=$NETTYPE \
- `facet_active_host mds1`:/mds1_svc/client_facet $MOUNT ||return $?
- echo "mount lustre on $MOUNT with $MOUNTLUSTRE: success"
- [ -d /r ] && $LCTL modules > /r/tmp/ogdb-`hostname`
- check_mount || return 41
- do_node `hostname` umount $MOUNT
- [ -f "$MOUNTLUSTRE" ] && rm -f $MOUNTLUSTRE
- echo "mount lustre on ${MOUNT} without $MOUNTLUSTRE....."
- do_node `hostname` mount -t lustre -o nettype=$NETTYPE \
- `facet_active_host mds1`:/mds1_svc/client_facet $MOUNT &&return $?
+ mount_client $MOUNT && error "mount succeeded" && return 1
echo "mount lustre on $MOUNT without $MOUNTLUSTRE failed as expected"
- cleanup || return $?
cleanup_15
+ cleanup || return $?
}
run_test 15 "zconf-mount without /sbin/mount.lustre (should return error)"
-is_digit() {
- local value=$1
- echo $value | grep -q "^[[:digit:]]*$"
- return $?
-}
-
+# LOGS/PENDING do not exist anymore since CMD3
test_16() {
- TMPMTPT="/mnt/conf16"
-
- if [ ! -f "$MDSDEV" ]; then
- echo "no $MDSDEV existing, so mount Lustre to create one"
- start_ost
- start_mds
- mount_client $MOUNT
+ TMPMTPT="${MOUNT%/*}/conf16"
+
+ 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"
- [ -d $TMPMTPT ] || mkdir -p $TMPMTPT
- mount -o loop -t ext3 $MDSDEV $TMPMTPT || return $?
- chmod 555 $TMPMTPT/OBJECTS || return $?
- chmod 555 $TMPMTPT/LOGS || return $?
- chmod 555 $TMPMTPT/PENDING || return $?
- umount $TMPMTPT || return $?
-
- echo "mount Lustre to change the mode of OBJECTS/LOGS/PENDING, then umount Lustre"
- start_ost
- start_mds
- mount_client $MOUNT
+
+ [ -f "$MDSDEV" ] && LOOPOPT="-o loop"
+
+ log "change the mode of $MDSDEV/OBJECTS to 555"
+ do_facet mds "mkdir -p $TMPMTPT &&
+ mount $LOOPOPT -t $FSTYPE $MDSDEV $TMPMTPT &&
+ chmod 555 $TMPMTPT/OBJECTS &&
+ umount $TMPMTPT" || return $?
+
+ log "mount Lustre to change the mode of OBJECTS, 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"
- EXPECTEDOBJECTSMODE=`debugfs -R "stat OBJECTS" $MDSDEV 2> /dev/null | awk '/Mode: /{print $6}'`
- EXPECTEDLOGSMODE=`debugfs -R "stat LOGS" $MDSDEV 2> /dev/null | awk '/Mode: /{print $6}'`
- EXPECTEDPENDINGMODE=`debugfs -R "stat PENDING" $MDSDEV 2> /dev/null | awk '/Mode: /{print $6}'`
-
- # check if values are empty
- test "x$EXPECTEDOBJECTSMODE" = "x" && EXPECTEDOBJECTSMODE="<empty>"
- test "x$EXPECTEDLOGSMODE" = "x" && EXPECTEDLOGSMODE="<empty>"
- test "x$EXPECTEDPENDINGMODE" = "x" && EXPECTEDPENDINGMODE="<empty>"
-
- # check if values are valid digits
- is_digit $EXPECTEDOBJECTSMODE || {
- echo "Invalid OBJECTS mode obtained from debugfs: $EXPECTEDOBJECTSMODE"
- return 42
- }
-
- is_digit $EXPECTEDLOGSMODE || {
- echo "Invalid LOGS mode obtained from debugfs: $EXPECTEDLOGSMODE"
- return 42
- }
-
- is_digit $EXPECTEDPENDINGMODE || {
- echo "Invalid PINDING mode obtained from debugfs: $EXPECTEDPENDINGMODE"
- return 42
- }
-
- # check if values are those we expected
- if [ "x$EXPECTEDOBJECTSMODE" = "x0777" ]; then
- echo "Success: Lustre change the mode of OBJECTS correctly"
- else
- echo "Error: Lustre does not change the mode of OBJECTS properly"
- echo "Expected value: 0777, actual one: $EXPECTEDOBJECTSMODE"
- return 1
- fi
-
- if [ "x$EXPECTEDLOGSMODE" = "x0777" ]; then
- echo "Success: Lustre change the mode of LOGS correctly"
- else
- echo "Error: Lustre does not change the mode of LOGS properly"
- echo "Expected value: 0777, actual one: $EXPECTEDLOGSMODE"
- return 1
- fi
-
- if [ "x$EXPECTEDPENDINGMODE" = "x0777" ]; then
- echo "Success: Lustre change the mode of PENDING correctly"
+
+ log "read the mode of OBJECTS and check if they has been changed properly"
+ EXPECTEDOBJECTSMODE=`do_facet mds "debugfs -R 'stat OBJECTS' $MDSDEV 2> /dev/null" | grep 'Mode: ' | sed -e "s/.*Mode: *//" -e "s/ *Flags:.*//"`
+
+ if [ "$EXPECTEDOBJECTSMODE" = "0777" ]; then
+ log "Success:Lustre change the mode of OBJECTS correctly"
else
- echo "Error: Lustre does not change the mode of PENDING properly"
- echo "Expected value: 0777, actual one: $EXPECTEDPENDINGMODE"
- return 1
+ error "Lustre does not change mode of OBJECTS properly"
fi
}
-run_test 16 "verify that lustre will correct the mode of OBJECTS/LOGS/PENDING"
+run_test 16 "verify that lustre will correct the mode of OBJECTS"
test_17() {
- TMPMTPT="/mnt/conf17"
-
- if [ ! -f "$MDSDEV" ]; then
+ if [ ! -e "$MDSDEV" ]; then
echo "no $MDSDEV existing, so mount Lustre to create one"
- start_ost
- start_mds
- mount_client $MOUNT
+ setup
check_mount || return 41
cleanup || return $?
fi
echo "Remove mds config log"
- [ -d $TMPMTPT ] || mkdir -p $TMPMTPT
- mount -o loop -t ext3 $MDSDEV $TMPMTPT || return $?
- rm -f $TMPMTPT/LOGS/mds1_svc || return $?
- umount $TMPMTPT || return $?
+ do_facet mds "debugfs -w -R 'unlink CONFIGS/$FSNAME-MDT0000' $MDSDEV || return \$?" || return $?
start_ost
- start mds1 $MDSLCONFARGS && return 42
- cleanup || return $?
+ start_mds && return 42
+ gen_config
}
-run_test 17 "Verify failed mds_postsetup won't fail assertion (2936)"
+run_test 17 "Verify failed mds_postsetup won't fail assertion (2936) (should return errs)"
test_18() {
[ -f $MDSDEV ] && echo "remove $MDSDEV" && rm -f $MDSDEV
echo "mount mds with large journal..."
- OLDMDSSIZE=$MDSSIZE
- MDSSIZE=2000000
+ local myMDSSIZE=2000000
+ OLD_MDS_MKFS_OPTS=$MDS_MKFS_OPTS
+
+ MDS_MKFS_OPTS="--mgs --mdt --fsname=$FSNAME --device-size=$myMDSSIZE --param sys.timeout=$TIMEOUT $MDSOPT"
+
gen_config
-
echo "mount lustre system..."
- start_ost
- start_mds
- mount_client $MOUNT
+ setup
check_mount || return 41
-
+
echo "check journal size..."
- FOUNDJOURNALSIZE=`debugfs -R "stat <8>" $MDSDEV | awk '/Size: / { print $6; 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. Size: $((FOUNDSIZE >> 20))M"
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
+
+ MDS_MKFS_OPTS=$OLD_MDS_MKFS_OPTS
gen_config
}
-run_test 18 "check lconf creates large journals"
+run_test 18 "check mkfs creates large journals"
+
+test_19a() {
+ start_mds || return 1
+ stop_mds -f || return 2
+}
+run_test 19a "start/stop MDS without OSTs"
+
+test_19b() {
+ start_ost || return 1
+ stop_ost -f || return 2
+}
+run_test 19b "start/stop OSTs without MDS"
+
+test_20() {
+ # first format the ost/mdt
+ start_ost
+ start_mds
+ mount_client $MOUNT
+ check_mount || return 43
+ rm -f $DIR/$tfile
+ remount_client ro $MOUNT || return 44
+ touch $DIR/$tfile && echo "$DIR/$tfile created incorrectly" && return 45
+ [ -e $DIR/$tfile ] && echo "$DIR/$tfile exists incorrectly" && return 46
+ remount_client rw $MOUNT || return 47
+ touch $DIR/$tfile
+ [ ! -f $DIR/$tfile ] && echo "$DIR/$tfile missing" && return 48
+ MCNT=`grep -c $MOUNT /etc/mtab`
+ [ "$MCNT" -ne 1 ] && echo "$MOUNT in /etc/mtab $MCNT times" && return 49
+ umount_client $MOUNT
+ stop_mds
+ stop_ost
+}
+run_test 20 "remount ro,rw mounts work and doesn't break /etc/mtab"
+
+test_21a() {
+ start_mds
+ start_ost
+ stop_ost
+ stop_mds
+}
+run_test 21a "start mds before ost, stop ost first"
+
+test_21b() {
+ start_ost
+ start_mds
+ stop_mds
+ stop_ost
+}
+run_test 21b "start ost before mds, stop mds first"
+
+test_21c() {
+ start_ost
+ start_mds
+ start_ost2
+ stop_ost
+ stop_ost2
+ stop_mds
+}
+run_test 21c "start mds between two osts, stop mds last"
+
+test_22() {
+ #reformat to remove all logs
+ reformat
+ start_mds
+ echo Client mount before any osts are in the logs
+ mount_client $MOUNT
+ check_mount && return 41
+ pass
+
+ echo Client mount with ost in logs, but none running
+ start_ost
+ stop_ost
+ mount_client $MOUNT
+ # check_mount will block trying to contact ost
+ umount_client $MOUNT
+ pass
+
+ echo Client mount with a running ost
+ start_ost
+ mount_client $MOUNT
+ check_mount || return 41
+ pass
+
+ cleanup
+}
+run_test 22 "start a client before osts (should return errs)"
+
+test_23() {
+ setup
+ # fail mds
+ stop mds
+ # force down client so that recovering mds waits for reconnect
+ zconf_umount `hostname` $MOUNT -f
+ # enter recovery on mds
+ start_mds
+ # try to start a new client
+ mount_client $MOUNT &
+ MOUNT_PID=$!
+ sleep 5
+ MOUNT_LUSTRE_PID=`ps -ef | grep mount.lustre | grep -v grep | awk '{print $2}'`
+ echo mount pid is ${MOUNT_PID}, mount.lustre pid is ${MOUNT_LUSTRE_PID}
+ ps --ppid $MOUNT_PID
+ ps --ppid $MOUNT_LUSTRE_PID
+ # FIXME why o why can't I kill these? Manual "ctrl-c" works...
+ kill -TERM $MOUNT_PID
+ echo "waiting for mount to finish"
+ ps -ef | grep mount
+ wait $MOUNT_PID
+
+ stop_mds
+ stop_ost
+}
+#this test isn't working yet
+#run_test 23 "interrupt client during recovery mount delay"
+
+test_24a() {
+ #set up fs1
+ gen_config
+ #set up fs2
+ local fs2mds_HOST=$mds_HOST
+ local fs2ost_HOST=$ost_HOST
+ [ -n "$ost1_HOST" ] && fs2ost_HOST=$ost1_HOST
+ if [ -z "$fs2ost_DEV" -o -z "$fs2mds_DEV" ]; then
+ do_facet $SINGLEMDS [ -b "$MDSDEV" ] && \
+ skip "mixed loopback and real device not working" && return
+ fi
+
+ 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 $fs2mdsdev || exit 10
+
+ add fs2ost $OST_MKFS_OPTS --fsname=${FSNAME2} --reformat $fs2ostdev || exit 10
+
+ setup
+ start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS
+ start fs2ost $fs2ostdev $OST_MOUNT_OPTS
+ mkdir -p $MOUNT2
+ mount -t lustre $MGSNID:/${FSNAME2} $MOUNT2 || return 1
+ # 1 still works
+ check_mount || return 2
+ # files written on 1 should not show up on 2
+ cp /etc/passwd $DIR/$tfile
+ sleep 10
+ [ -e $MOUNT2/$tfile ] && error "File bleed" && return 7
+ # 2 should work
+ sleep 5
+ cp /etc/passwd $MOUNT2/b || return 3
+ rm $MOUNT2/b || return 4
+ # 2 is actually mounted
+ grep $MOUNT2' ' /proc/mounts > /dev/null || return 5
+ # failover
+ facet_failover fs2mds
+ facet_failover fs2ost
+ df
+ umount_client $MOUNT
+ # the MDS must remain up until last MDT
+ stop_mds
+ MDS=$(do_facet $SINGLEMDS "cat $LPROC/devices" | awk '($3 ~ "mdt" && $4 ~ "MDT") { print $4 }')
+ [ -z "$MDS" ] && error "No MDT" && return 8
+ umount $MOUNT2
+ stop fs2mds -f
+ stop fs2ost -f
+ cleanup_nocli || return 6
+}
+run_test 24a "Multiple MDTs on a single node"
+
+test_24b() {
+ local fs2mds_HOST=$mds_HOST
+ if [ -z "$fs2mds_DEV" ]; then
+ do_facet $SINGLEMDS [ -b "$MDSDEV" ] && \
+ skip "mixed loopback and real device not working" && return
+ fi
+
+ local fs2mdsdev=${fs2mds_DEV:-${MDSDEV}_2}
+
+ add fs2mds $MDS_MKFS_OPTS --fsname=${FSNAME}2 --mgs --reformat $fs2mdsdev || exit 10
+ setup
+ start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS && return 2
+ cleanup || return 6
+}
+run_test 24b "Multiple MGSs on a single node (should return err)"
+
+test_25() {
+ setup
+ check_mount || return 2
+ local MODULES=$($LCTL modules | awk '{ print $2 }')
+ rmmod $MODULES 2>/dev/null || true
+ cleanup || return 6
+}
+run_test 25 "Verify modules are referenced"
+
+test_26() {
+ load_modules
+ # we need modules before mount for sysctl, so make sure...
+ 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
+ cat $LPROC/devices
+ DEVS=$(cat $LPROC/devices | wc -l)
+ [ $DEVS -gt 0 ] && return 2
+ unload_modules || return 203
+}
+run_test 26 "MDT startup failure cleans LOV (should return errs)"
+
+set_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 $FINAL"
+ do_facet mds "$LCTL conf_param $PARAM=$FINAL" || error conf_param failed
+ local RESULT
+ local MAX=90
+ local WAIT=0
+ while [ 1 ]; do
+ sleep 5
+ RESULT=$(do_facet $myfacet "$TEST")
+ if [ $RESULT -eq $FINAL ]; then
+ echo "Updated config after $WAIT sec (got $RESULT)"
+ break
+ fi
+ WAIT=$((WAIT + 5))
+ if [ $WAIT -eq $MAX ]; then
+ echo "Config update not seen: wanted $FINAL got $RESULT"
+ return 3
+ fi
+ echo "Waiting $(($MAX - $WAIT)) secs for config update"
+ done
+}
+
+test_27a() {
+ start_ost || return 1
+ start_mds || return 2
+ echo "Requeue thread should have started: "
+ ps -e | grep ll_cfg_requeue
+ 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 mds "cat $LPROC/mdt/$FSNAME-MDT0000/identity_acquire_expire" "$FSNAME-MDT0000.mdt.identity_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"
+
+test_28() {
+ setup
+ TEST="cat $LPROC/llite/$FSNAME-*/max_read_ahead_whole_mb"
+ ORIG=$($TEST)
+ declare -i FINAL
+ FINAL=$(($ORIG + 10))
+ 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)
+ if [ $RESULT -ne $FINAL ]; then
+ echo "New config not seen: wanted $FINAL got $RESULT"
+ return 4
+ else
+ echo "New config success: got $RESULT"
+ fi
+ cleanup
+}
+run_test 28 "permanent parameter setting"
+
+test_29() {
+ [ "$OSTCOUNT" -lt "2" ] && skip "$OSTCOUNT < 2, skipping" && return
+ setup > /dev/null 2>&1
+ start_ost2
+ sleep 10
+
+ local PARAM="$FSNAME-OST0001.osc.active"
+ local PROC_ACT="$LPROC/osc/$FSNAME-OST0001-osc-[^M]*/active"
+ local PROC_UUID="$LPROC/osc/$FSNAME-OST0001-osc-[^M]*/ost_server_uuid"
+ if [ ! -r $PROC_ACT ]; then
+ echo "Can't read $PROC_ACT"
+ ls $LPROC/osc/$FSNAME-*
+ return 1
+ fi
+ ACTV=$(cat $PROC_ACT)
+ DEAC=$((1 - $ACTV))
+ 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
+ echo "Live client not deactivated: $(cat $PROC_UUID)"
+ return 3
+ else
+ echo "Live client success: got $RESULT"
+ fi
+
+ # check MDT too
+ local MPROC="$LPROC/osc/$FSNAME-OST0001-osc-[M]*/active"
+ local MAX=30
+ local WAIT=0
+ while [ 1 ]; do
+ sleep 5
+ RESULT=`do_facet mds " [ -r $MPROC ] && cat $MPROC"`
+ [ ${PIPESTATUS[0]} = 0 ] || error "Can't read $MPROC"
+ if [ $RESULT -eq $DEAC ]; then
+ echo "MDT deactivated also after $WAIT sec (got $RESULT)"
+ break
+ fi
+ WAIT=$((WAIT + 5))
+ if [ $WAIT -eq $MAX ]; then
+ echo "MDT not deactivated: wanted $DEAC got $RESULT"
+ return 4
+ fi
+ echo "Waiting $(($MAX - $WAIT)) secs for MDT deactivated"
+ done
+
+ # test new client starts deactivated
+ umount_client $MOUNT || return 200
+ mount_client $MOUNT
+ RESULT=$(grep DEACTIV $PROC_UUID | grep NEW)
+ if [ -z "$RESULT" ]; then
+ echo "New client not deactivated from start: $(cat $PROC_UUID)"
+ return 5
+ else
+ echo "New client success: got $RESULT"
+ fi
+
+ # make sure it reactivates
+ set_and_check client "cat $PROC_ACT" "$PARAM" $ACTV || return 6
+
+ umount_client $MOUNT
+ stop_ost2
+ cleanup_nocli
+ #writeconf to remove all ost2 traces for subsequent tests
+ writeconf
+ start_mds
+ start_ost
+ cleanup
+}
+run_test 29 "permanently remove an OST"
+
+test_30() {
+ setup
+
+ TEST="cat $LPROC/llite/$FSNAME-*/max_read_ahead_whole_mb"
+ ORIG=$($TEST)
+ for i in $(seq 1 20); do
+ set_and_check client "$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 client "$TEST" "$FSNAME.llite.max_read_ahead_whole_mb" $ORIG || return 6
+ cleanup
+}
+run_test 30 "Big config llog"
+
+test_31() { # bug 10734
+ # ipaddr must not exist
+ mount -t lustre 4.3.2.1@tcp:/lustre $MOUNT || true
+ cleanup
+}
+run_test 31 "Connect to non-existent node (shouldn't crash)"
+
+test_32a() {
+ # XXX - make this run on client-only systems with real hardware on
+ # the OST and MDT
+ # there appears to be a lot of assumption here about loopback
+ # devices
+ # or maybe this test is just totally useless on a client-only system
+ [ "$mds_HOST" = "`hostname`" ] || { skip "remote MDS" && return 0; }
+ [ "$ost_HOST" = "`hostname`" -o "$ost1_HOST" = "`hostname`" ] || \
+ { skip "remote OST" && return 0; }
+
+ [ -z "$TUNEFS" ] && skip "No tunefs" && return
+ local DISK1_4=$LUSTRE/tests/disk1_4.zip
+ [ ! -r $DISK1_4 ] && skip "Cant find $DISK1_4, skipping" && return
+ mkdir -p $TMP/$tdir
+ unzip -o -j -d $TMP/$tdir $DISK1_4 || { skip "Cant unzip $DISK1_4, skipping" && return ; }
+ load_modules
+ sysctl lnet.debug=$PTLDEBUG
+
+ $TUNEFS $TMP/$tdir/mds || error "tunefs failed"
+ # nids are wrong, so client wont work, but server should start
+ start mds $TMP/$tdir/mds "-o loop,exclude=lustre-OST0000" || return 3
+ local UUID=$(cat $LPROC/mds/lustre-MDT0000/uuid)
+ echo MDS uuid $UUID
+ [ "$UUID" == "mdsA_UUID" ] || error "UUID is wrong: $UUID"
+
+ $TUNEFS --mgsnode=`hostname` $TMP/$tdir/ost1 || error "tunefs failed"
+ start ost1 $TMP/$tdir/ost1 "-o loop" || return 5
+ UUID=$(cat $LPROC/obdfilter/lustre-OST0000/uuid)
+ echo OST uuid $UUID
+ [ "$UUID" == "ost1_UUID" ] || error "UUID is wrong: $UUID"
+
+ local NID=$($LCTL list_nids | head -1)
+
+ echo "OSC changes should return err:"
+ $LCTL conf_param lustre-OST0000.osc.max_dirty_mb=15 && return 7
+ $LCTL conf_param lustre-OST0000.failover.node=$NID && return 8
+ echo "ok."
+ echo "MDC changes should succeed:"
+ $LCTL conf_param lustre-MDT0000.mdc.max_rpcs_in_flight=9 || return 9
+ $LCTL conf_param lustre-MDT0000.failover.node=$NID || return 10
+ echo "ok."
+
+ # With a new good MDT failover nid, we should be able to mount a client
+ # (but it cant talk to OST)
+ local OLDMOUNTOPT=$MOUNTOPT
+ MOUNTOPT="exclude=lustre-OST0000"
+ mount_client $MOUNT
+ MOUNTOPT=$OLDMOUNTOPT
+ set_and_check client "cat $LPROC/mdc/*/max_rpcs_in_flight" "lustre-MDT0000.mdc.max_rpcs_in_flight" || return 11
+
+ zconf_umount `hostname` $MOUNT -f
+ cleanup_nocli
+
+ # mount a second time to make sure we didnt leave upgrade flag on
+ $TUNEFS --dryrun $TMP/$tdir/mds || error "tunefs failed"
+ start mds $TMP/$tdir/mds "-o loop,exclude=lustre-OST0000" || return 12
+ cleanup_nocli
+
+ [ -d $TMP/$tdir ] && rm -rf $TMP/$tdir
+}
+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
+ [ "$mds_HOST" = "`hostname`" ] || { skip "remote MDS" && return 0; }
+ [ "$ost_HOST" = "`hostname`" -o "$ost1_HOST" = "`hostname`" ] || \
+ { skip "remote OST" && return 0; }
+
+ [ -z "$TUNEFS" ] && skip "No tunefs" && return
+ local DISK1_4=$LUSTRE/tests/disk1_4.zip
+ [ ! -r $DISK1_4 ] && skip "Cant find $DISK1_4, skipping" && return
+ mkdir -p $TMP/$tdir
+ unzip -o -j -d $TMP/$tdir $DISK1_4 || { skip "Cant unzip $DISK1_4, skipping" && return ; }
+ load_modules
+ sysctl lnet.debug=$PTLDEBUG
+
+ # writeconf will cause servers to register with their current nids
+ $TUNEFS --writeconf $TMP/$tdir/mds || error "tunefs failed"
+ start mds $TMP/$tdir/mds "-o loop" || return 3
+ local UUID=$(cat $LPROC/mds/lustre-MDT0000/uuid)
+ echo MDS uuid $UUID
+ [ "$UUID" == "mdsA_UUID" ] || error "UUID is wrong: $UUID"
+
+ $TUNEFS --mgsnode=`hostname` $TMP/$tdir/ost1 || error "tunefs failed"
+ start ost1 $TMP/$tdir/ost1 "-o loop" || return 5
+ UUID=$(cat $LPROC/obdfilter/lustre-OST0000/uuid)
+ echo OST uuid $UUID
+ [ "$UUID" == "ost1_UUID" ] || error "UUID is wrong: $UUID"
+
+ echo "OSC changes should succeed:"
+ $LCTL conf_param lustre-OST0000.osc.max_dirty_mb=15 || return 7
+ $LCTL conf_param lustre-OST0000.failover.node=$NID || return 8
+ echo "ok."
+ echo "MDC changes should succeed:"
+ $LCTL conf_param lustre-MDT0000.mdc.max_rpcs_in_flight=9 || return 9
+ echo "ok."
+
+ # MDT and OST should have registered with new nids, so we should have
+ # a fully-functioning client
+ echo "Check client and old fs contents"
+ mount_client $MOUNT
+ set_and_check client "cat $LPROC/mdc/*/max_rpcs_in_flight" "${NEWNAME}-MDT0000.mdc.max_rpcs_in_flight" || return 11
+ [ "$(cksum $MOUNT/passwd | cut -d' ' -f 1,2)" == "2479747619 779" ] || return 12
+ echo "ok."
+
+ cleanup
+ [ -d $TMP/$tdir ] && rm -rf $TMP/$tdir
+}
+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
+ [ -n "$ost1_HOST" ] && fs2ost_HOST=$ost1_HOST
+
+ if [ -z "$fs2ost_DEV" -o -z "$fs2mds_DEV" ]; then
+ do_facet $SINGLEMDS [ -b "$MDSDEV" ] && \
+ skip "mixed loopback and real device not working" && return
+ fi
+
+ 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=$MGSNID --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 $fs2ostdev
+ cleanup_nocli || return 6
+}
+run_test 33 "Mount ost with a large index number"
+
+test_34() {
+ setup
+
+ do_facet client dd if=/dev/zero of=$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 34 "Drop cancel during umount"
+
+test_34a() {
+ setup
+ do_facet client multiop $DIR/file O_c &
+ sleep 0.500s
+ manual_umount_client
+ rc=$?
+ 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 ost should be normal"
+
+test_35() { # bug 12459
+ setup
+
+ DBG_SAVE="`sysctl -n lnet.debug`"
+ sysctl -w lnet.debug="ha"
+
+ log "Set up a fake failnode for the MDS"
+ FAKENID="127.0.0.2"
+ do_facet mds $LCTL conf_param ${FSNAME}-MDT0000.failover.node=$FAKENID || return 4
+
+ log "Wait for RECONNECT_INTERVAL seconds (10s)"
+ sleep 10
+
+ MSG="conf-sanity.sh test_35 `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 $DFPID
+ log "done"
+ sysctl -w lnet.debug="$DBG_SAVE"
+
+ # 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 ${NEXTCONN} instead)" && return 7
+ cleanup
+}
+run_test 35 "Reconnect to the last active server first"
+
+test_36() { # 12743
+ local rc
+ local FSNAME2=test1234
+ local fs2mds_HOST=$mds_HOST
+ local fs2ost_HOST=$ost_HOST
+ local fs3ost_HOST=$ost_HOST
+
+ [ -n "$ost1_HOST" ] && fs2ost_HOST=$ost1_HOST && fs3ost_HOST=$ost1_HOST
+ rc=0
+
+ if [ -z "$fs2ost_DEV" -o -z "$fs2mds_DEV" -o -z "$fs3ost_DEV" ]; then
+ do_facet $SINGLEMDS [ -b "$MDSDEV" ] && \
+ skip "mixed loopback and real device not working" && return
+ fi
+ [ $OSTCOUNT -lt 2 ] && skip "skipping test for single OST" && return
+
+ [ $(grep -c obdfilter $LPROC/devices) -eq 0 ] &&
+ skip "skipping test for remote OST" && return
+
+ local fs2mdsdev=${fs2mds_DEV:-${MDSDEV}_2}
+ local fs2ostdev=${fs2ost_DEV:-$(ostdevname 1)_2}
+ local fs3ostdev=${fs3ost_DEV:-$(ostdevname 2)_2}
+ add fs2mds $MDS_MKFS_OPTS --fsname=${FSNAME2} --reformat $fs2mdsdev || exit 10
+ add fs2ost $OST_MKFS_OPTS --mkfsoptions='-b1024' --fsname=${FSNAME2} --mgsnode=$MGSNID --reformat $fs2ostdev || exit 10
+ add fs3ost $OST_MKFS_OPTS --mkfsoptions='-b4096' --fsname=${FSNAME2} --mgsnode=$MGSNID --reformat $fs3ostdev || exit 10
+
+ start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS
+ start fs2ost $fs2ostdev $OST_MOUNT_OPTS
+ start fs3ost $fs3ostdev $OST_MOUNT_OPTS
+ mkdir -p $MOUNT2
+ mount -t lustre $MGSNID:/${FSNAME2} $MOUNT2 || return 1
+
+ sleep 5 # until 11778 fixed
+
+ dd if=/dev/zero of=$MOUNT2/$tfile bs=1M count=7 || return 2
+
+ BKTOTAL=`awk 'BEGIN{total=0}; {total+=$1}; END{print total}' \
+ $LPROC/obdfilter/*/kbytestotal`
+ BKFREE=`awk 'BEGIN{free=0}; {free+=$1}; END{print free}' \
+ $LPROC/obdfilter/*/kbytesfree`
+ BKAVAIL=`awk 'BEGIN{avail=0}; {avail+=$1}; END{print avail}' \
+ $LPROC/obdfilter/*/kbytesavail`
+ STRING=`df -P $MOUNT2 | tail -n 1 | awk '{print $2","$3","$4}'`
+ DFTOTAL=`echo $STRING | cut -d, -f1`
+ DFUSED=`echo $STRING | cut -d, -f2`
+ DFAVAIL=`echo $STRING | cut -d, -f3`
+ DFFREE=$(($DFTOTAL - $DFUSED))
+
+ ALLOWANCE=$((64 * $OSTCOUNT))
+
+ if [ $DFTOTAL -lt $(($BKTOTAL - $ALLOWANCE)) ] ||
+ [ $DFTOTAL -gt $(($BKTOTAL + $ALLOWANCE)) ] ; then
+ echo "**** FAIL: df total($DFTOTAL) mismatch OST total($BKTOTAL)"
+ rc=1
+ fi
+ if [ $DFFREE -lt $(($BKFREE - $ALLOWANCE)) ] ||
+ [ $DFFREE -gt $(($BKFREE + $ALLOWANCE)) ] ; then
+ echo "**** FAIL: df free($DFFREE) mismatch OST free($BKFREE)"
+ rc=2
+ fi
+ if [ $DFAVAIL -lt $(($BKAVAIL - $ALLOWANCE)) ] ||
+ [ $DFAVAIL -gt $(($BKAVAIL + $ALLOWANCE)) ] ; then
+ echo "**** FAIL: df avail($DFAVAIL) mismatch OST avail($BKAVAIL)"
+ rc=3
+ fi
+
+ umount -d $MOUNT2
+ stop fs3ost -f || return 200
+ stop fs2ost -f || return 201
+ stop fs2mds -f || return 202
+ rm -rf $MOUNT2 $fs2mdsdev $fs2ostdev $fs3ostdev
+ unload_modules || return 203
+ return $rc
+}
+run_test 36 "df report consistency on OSTs with different block size"
+
+test_37() {
+ LOCAL_MDSDEV="$TMP/mdt.img"
+ SYM_MDSDEV="$TMP/sym_mdt.img"
+
+ echo "MDS : $LOCAL_MDSDEV"
+ echo "SYMLINK : $SYM_MDSDEV"
+ rm -f $LOCAL_MDSDEV
+
+ touch $LOCAL_MDSDEV
+ mkfs.lustre --reformat --fsname=lustre --mdt --mgs --device-size=9000 $LOCAL_MDSDEV ||
+ error "mkfs.lustre $LOCAL_MDSDEV failed"
+ ln -s $LOCAL_MDSDEV $SYM_MDSDEV
+
+ echo "mount symlink device - $SYM_MDSDEV"
+
+ mount_op=`mount -v -t lustre -o loop $SYM_MDSDEV ${MOUNT%/*}/mds 2>&1 | grep "unable to set tunable"`
+ umount -d ${MOUNT%/*}/mds
+ rm -f $LOCAL_MDSDEV $SYM_MDSDEV
+
+ if [ -n "$mount_op" ]; then
+ error "**** FAIL: set tunables failed for symlink device"
+ fi
+ return 0
+}
+run_test 37 "verify set tunables works for symlink device"
+
+test_38() { # bug 14222
+ setup
+ # like runtests
+ COUNT=10
+ SRC="/etc /bin"
+ FILES=`find $SRC -type f -mtime +1 | head -n $COUNT`
+ log "copying $(echo $FILES | wc -w) files to $DIR/$tdir"
+ mkdir -p $DIR/$tdir
+ tar cf - $FILES | tar xf - -C $DIR/$tdir || \
+ error "copying $SRC to $DIR/$tdir"
+ sync
+ umount_client $MOUNT
+ stop_mds
+ log "rename lov_objid file on MDS"
+ rm -f $TMP/lov_objid.orig
+ do_facet mds "debugfs -w $MDSDEV" <<-EOF
+ dump lov_objid $TMP/lov_objid.orig
+ rm lov_objid
+ EOF
+ do_facet mds "od -Ax -td8 $TMP/lov_objid.orig"
+ # check create in mds_lov_connect
+ start_mds
+ mount_client $MOUNT
+ for f in $FILES; do
+ [ $V ] && log "verifying $DIR/$tdir/$f"
+ diff $f $DIR/$tdir/$f || ERROR=y
+ done
+ do_facet mds "debugfs -c $MDSDEV" <<-EOF
+ dump lov_objid $TMP/lov_objid.new
+ EOF
+ do_facet mds "od -Ax -td8 $TMP/lov_objid.new"
+ [ "$ERROR" = "y" ] && error "old and new files are different after connect" || true
+
+
+ # check it's updates in sync
+ umount_client $MOUNT
+ stop_mds
+ multiop $TMP/lov_objid.clear Ow4096c
+ do_facet mds "debugfs -w $MDSDEV" <<-EOF
+ write $TMP/lov_objid.clear lov_objid
+ EOF
+ start_mds
+ mount_client $MOUNT
+ for f in $FILES; do
+ [ $V ] && log "verifying $DIR/$tdir/$f"
+ diff $f $DIR/$tdir/$f || ERROR=y
+ done
+ do_facet mds "debugfs -c $MDSDEV" <<-EOF
+ dump lov_objid $TMP/lov_objid.new1
+ EOF
+ do_facet mds "od -Ax -td8 $TMP/lov_objid.new1"
+ umount_client $MOUNT
+ stop_mds
+ [ "$ERROR" = "y" ] && error "old and new files are different after sync" || true
+
+ log "files compared the same"
+ #cleanup
+}
+run_test 38 "MDS recreates missing lov_objid file from OST data"
+
+umount_client $MOUNT
+cleanup_nocli
+cleanup_krb5_env
-equals_msg "Done"
+equals_msg `basename $0`: test complete
+[ -f "$TESTSUITELOG" ] && cat $TESTSUITELOG || true
git://git.whamcloud.com / fs / lustre-release.git / blobdiff