3 # FIXME - there is no reason to use all of these different
4 # return codes, espcially when most of them are mapped to something
5 # else anyway. The combination of test number and return code
6 # figure out what failed.
12 # bug number for skipped test: LU-2828
13 ALWAYS_EXCEPT="$CONF_SANITY_EXCEPT 59 64"
14 # UPDATE THE COMMENT ABOVE WITH BUG NUMBERS WHEN CHANGING ALWAYS_EXCEPT!
18 if [ -r /etc/SuSE-release ]
20 local vers=`grep VERSION /etc/SuSE-release | awk '{print $3}'`
21 local patchlev=`grep PATCHLEVEL /etc/SuSE-release \
23 if [ $vers -eq 11 ] && [ $patchlev -eq 2 ]
31 if is_sles11; then # LU-2181
32 ALWAYS_EXCEPT="$ALWAYS_EXCEPT 23a 34b"
35 if [ "$FAILURE_MODE" = "HARD" ]; then
36 CONFIG_EXCEPTIONS="24a " && \
37 echo "Except the tests: $CONFIG_EXCEPTIONS for FAILURE_MODE=$FAILURE_MODE, bug 23573" && \
38 ALWAYS_EXCEPT="$ALWAYS_EXCEPT $CONFIG_EXCEPTIONS"
41 # bug number for skipped test:
42 # a tool to create lustre filesystem images
43 ALWAYS_EXCEPT="32newtarball $ALWAYS_EXCEPT"
46 PATH=$PWD/$SRCDIR:$SRCDIR:$SRCDIR/../utils:$PATH
48 PTLDEBUG=${PTLDEBUG:--1}
50 LUSTRE=${LUSTRE:-`dirname $0`/..}
51 RLUSTRE=${RLUSTRE:-$LUSTRE}
52 LUSTRE_TESTS_API_DIR=${LUSTRE_TESTS_API_DIR:-${LUSTRE}/tests/clientapi}
53 export MULTIOP=${MULTIOP:-multiop}
55 . $LUSTRE/tests/test-framework.sh
58 # use small MDS + OST size to speed formatting time
59 # do not use too small MDSSIZE/OSTSIZE, which affect the default jouranl size
62 . ${CONFIG:=$LUSTRE/tests/cfg/$NAME.sh}
64 if ! combined_mgs_mds; then
65 # bug number for skipped test: 23954
66 ALWAYS_EXCEPT="$ALWAYS_EXCEPT 24b"
69 # STORED_MDSSIZE is used in test_18
70 if [ -n "$MDSSIZE" ]; then
71 STORED_MDSSIZE=$MDSSIZE
74 # pass "-E lazy_itable_init" to mke2fs to speed up the formatting time
75 if [[ "$LDISKFS_MKFS_OPTS" != *lazy_itable_init* ]]; then
76 LDISKFS_MKFS_OPTS=$(csa_add "$LDISKFS_MKFS_OPTS" -E lazy_itable_init)
79 [ $(facet_fstype $SINGLEMDS) = "zfs" ] &&
80 # bug number for skipped test: LU-2778 LU-4444
81 ALWAYS_EXCEPT="$ALWAYS_EXCEPT 57b 69"
86 require_dsh_mds || exit 0
87 require_dsh_ost || exit 0
89 [ "$SLOW" = "no" ] && EXCEPT_SLOW="30a 31 45 69"
95 # The MGS must be started before the OSTs for a new fs, so start
96 # and stop to generate the startup logs.
99 wait_osc_import_state mds ost FULL
104 reformat_and_config() {
106 if ! combined_mgs_mds ; then
112 writeconf_or_reformat() {
113 # There are at most 2 OSTs for write_conf test
114 # who knows if/where $TUNEFS is installed?
115 # Better reformat if it fails...
116 writeconf_all $MDSCOUNT 2 ||
117 { echo "tunefs failed, reformatting instead" &&
118 reformat_and_config && return 0; }
128 start mgs $MGSDEV $MGS_MOUNT_OPTS
134 local dev=$(mdsdevname $num)
137 echo "start mds service on `facet_active_host $facet`"
138 start $facet ${dev} $MDS_MOUNT_OPTS $@ || return 94
144 local dev=$(mdsdevname $num)
147 echo "stop mds service on `facet_active_host $facet`"
148 # These tests all use non-failover stop
149 stop $facet -f || return 97
155 for num in $(seq $MDSCOUNT); do
156 start_mdt $num $@ || return 94
161 if ! combined_mgs_mds ; then
169 for num in $(seq $MDSCOUNT); do
170 stop_mdt $num || return 97
175 echo "stop mgs service on `facet_active_host mgs`"
176 # These tests all use non-failover stop
177 stop mgs -f || return 97
181 echo "start ost1 service on `facet_active_host ost1`"
182 start ost1 `ostdevname 1` $OST_MOUNT_OPTS $@ || return 95
186 echo "stop ost1 service on `facet_active_host ost1`"
187 # These tests all use non-failover stop
188 stop ost1 -f || return 98
192 echo "start ost2 service on `facet_active_host ost2`"
193 start ost2 `ostdevname 2` $OST_MOUNT_OPTS $@ || return 92
197 echo "stop ost2 service on `facet_active_host ost2`"
198 # These tests all use non-failover stop
199 stop ost2 -f || return 93
204 echo "mount $FSNAME on ${MOUNTPATH}....."
205 zconf_mount `hostname` $MOUNTPATH || return 96
209 local mountopt="-o remount,$1"
211 echo "remount '$1' lustre on ${MOUNTPATH}....."
212 zconf_mount `hostname` $MOUNTPATH "$mountopt" || return 96
217 echo "umount lustre on ${MOUNTPATH}....."
218 zconf_umount `hostname` $MOUNTPATH || return 97
221 manual_umount_client(){
224 echo "manual umount lustre on ${MOUNT}...."
225 do_facet client "umount -d ${FORCE} $MOUNT"
231 start_mds || error "MDT start failed"
232 start_ost || error "OST start failed"
233 mount_client $MOUNT || error "client start failed"
234 client_up || error "client_up failed"
238 if ! combined_mgs_mds ; then
247 unload_modules_conf () {
248 if combined_mgs_mds || ! local_mode; then
249 unload_modules || return 1
254 stop_ost || return 202
255 stop_mds || return 201
256 unload_modules_conf || return 203
260 umount_client $MOUNT || return 200
261 cleanup_nocli || return $?
265 do_facet client "cp /etc/passwd $DIR/a" || return 71
266 do_facet client "rm $DIR/a" || return 72
267 # make sure lustre is actually mounted (touch will block,
268 # but grep won't, so do it after)
269 do_facet client "grep $MOUNT' ' /proc/mounts > /dev/null" || return 73
270 echo "setup single mount lustre success"
274 do_facet client "touch $DIR/a" || return 71
275 do_facet client "rm $DIR/a" || return 72
276 do_facet client "touch $DIR2/a" || return 73
277 do_facet client "rm $DIR2/a" || return 74
278 echo "setup double mount lustre success"
283 if [ "$ONLY" == "setup" ]; then
288 if [ "$ONLY" == "cleanup" ]; then
295 #create single point mountpoint
301 check_mount || return 41
304 run_test 0 "single mount setup"
307 start_mds || error "MDT start failed"
309 echo "start ost second time..."
310 start_ost && error "2nd OST start should fail"
311 mount_client $MOUNT || error "client start failed"
312 check_mount || return 42
315 run_test 1 "start up ost twice (should return errors)"
319 echo "start mds second time.."
320 start_mdt 1 && error "2nd MDT start should fail"
323 check_mount || return 43
326 run_test 2 "start up mds twice (should return err)"
330 #mount.lustre returns an error if already in mtab
331 mount_client $MOUNT && error "2nd client mount should fail"
332 check_mount || return 44
335 run_test 3 "mount client twice (should return err)"
339 touch $DIR/$tfile || return 85
343 # ok for ost to fail shutdown
344 if [ 202 -ne $eno ]; then
349 run_test 4 "force cleanup ost, then cleanup"
351 test_5a() { # was test_5
353 touch $DIR/$tfile || return 1
354 fuser -m -v $MOUNT && echo "$MOUNT is in use by user space process."
356 stop_mds -f || return 2
358 # cleanup may return an error from the failed
359 # disconnects; for now I'll consider this successful
360 # if all the modules have unloaded.
364 echo "killing umount"
365 kill -TERM $UMOUNT_PID
366 echo "waiting for umount to finish"
368 if grep " $MOUNT " /proc/mounts; then
369 echo "test 5: /proc/mounts after failed umount"
373 echo "killing umount"
374 kill -TERM $UMOUNT_PID
375 echo "waiting for umount to finish"
377 grep " $MOUNT " /proc/mounts && echo "test 5: /proc/mounts after second umount" && return 11
381 # stop_mds is a no-op here, and should not fail
382 cleanup_nocli || return $?
383 # df may have lingering entry
385 # mtab may have lingering entry
389 while [ "$WAIT" -ne "$MAX_WAIT" ]; do
391 grep -q $MOUNT" " /etc/mtab || break
392 echo "Waiting /etc/mtab updated ... "
393 WAIT=$(( WAIT + sleep))
395 [ "$WAIT" -eq "$MAX_WAIT" ] && error "/etc/mtab is not updated in $WAIT secs"
396 echo "/etc/mtab updated in $WAIT secs"
398 run_test 5a "force cleanup mds, then cleanup"
406 grep " $MOUNT " /etc/mtab && \
407 error false "unexpected entry in mtab before mount" && return 10
411 if ! combined_mgs_mds ; then
412 trap cleanup_5b EXIT ERR
417 [ -d $MOUNT ] || mkdir -p $MOUNT
418 mount_client $MOUNT && rc=1
419 grep " $MOUNT " /etc/mtab && \
420 error "$MOUNT entry in mtab after failed mount" && rc=11
422 # stop_mds is a no-op here, and should not fail
423 cleanup_nocli || rc=$?
424 if ! combined_mgs_mds ; then
429 run_test 5b "Try to start a client with no MGS (should return errs)"
432 grep " $MOUNT " /etc/mtab && \
433 error false "unexpected entry in mtab before mount" && return 10
438 [ -d $MOUNT ] || mkdir -p $MOUNT
439 local oldfs="${FSNAME}"
440 FSNAME="wrong.${FSNAME}"
441 mount_client $MOUNT || :
443 grep " $MOUNT " /etc/mtab && \
444 error "$MOUNT entry in mtab after failed mount" && rc=11
446 cleanup_nocli || rc=$?
449 run_test 5c "cleanup after failed mount (bug 2712) (should return errs)"
452 grep " $MOUNT " /etc/mtab && \
453 error false "unexpected entry in mtab before mount" && return 10
459 mount_client $MOUNT || rc=1
461 grep " $MOUNT " /etc/mtab && \
462 error "$MOUNT entry in mtab after unmount" && rc=11
465 run_test 5d "mount with ost down"
468 grep " $MOUNT " /etc/mtab && \
469 error false "unexpected entry in mtab before mount" && return 10
475 #define OBD_FAIL_PTLRPC_DELAY_SEND 0x506
476 do_facet client "lctl set_param fail_loc=0x80000506"
477 mount_client $MOUNT || echo "mount failed (not fatal)"
479 grep " $MOUNT " /etc/mtab && \
480 error "$MOUNT entry in mtab after unmount" && rc=11
483 run_test 5e "delayed connect, don't crash (bug 10268)"
486 if combined_mgs_mds ; then
487 skip "combined mgs and mds"
491 grep " $MOUNT " /etc/mtab && \
492 error false "unexpected entry in mtab before mount" && return 10
496 [ -d $MOUNT ] || mkdir -p $MOUNT
497 mount_client $MOUNT &
499 echo client_mount pid is $pid
503 if ! ps -f -p $pid >/dev/null; then
506 grep " $MOUNT " /etc/mtab && echo "test 5f: mtab after mount"
507 error "mount returns $rc, expected to hang"
516 # mount should succeed after start mds
519 [ $rc -eq 0 ] || error "mount returned $rc"
520 grep " $MOUNT " /etc/mtab && echo "test 5f: mtab after mount"
524 run_test 5f "mds down, cleanup after failed mount (bug 2712)"
529 mount_client ${MOUNT} || return 87
530 touch $DIR/a || return 86
533 run_test 6 "manual umount, then mount again"
538 cleanup_nocli || return $?
540 run_test 7 "manual umount, then cleanup"
545 check_mount2 || return 45
546 umount_client $MOUNT2
549 run_test 8 "double mount setup"
554 do_facet ost1 lctl set_param debug=\'inode trace\' || return 1
555 do_facet ost1 lctl set_param subsystem_debug=\'mds ost\' || return 1
557 CHECK_PTLDEBUG="`do_facet ost1 lctl get_param -n debug`"
558 if [ "$CHECK_PTLDEBUG" ] && { \
559 [ "$CHECK_PTLDEBUG" = "trace inode warning error emerg console" ] ||
560 [ "$CHECK_PTLDEBUG" = "trace inode" ]; }; then
561 echo "lnet.debug success"
563 echo "lnet.debug: want 'trace inode', have '$CHECK_PTLDEBUG'"
566 CHECK_SUBSYS="`do_facet ost1 lctl get_param -n subsystem_debug`"
567 if [ "$CHECK_SUBSYS" ] && [ "$CHECK_SUBSYS" = "mds ost" ]; then
568 echo "lnet.subsystem_debug success"
570 echo "lnet.subsystem_debug: want 'mds ost', have '$CHECK_SUBSYS'"
573 stop_ost || return $?
575 run_test 9 "test ptldebug and subsystem for mkfs"
583 do_facet $facet "test -b $dev" || rc=1
584 if [[ "$size" ]]; then
585 local in=$(do_facet $facet "dd if=$dev of=/dev/null bs=1k count=1 skip=$size 2>&1" |\
586 awk '($3 == "in") { print $1 }')
587 [[ $in = "1+0" ]] || rc=1
593 # Test 16 was to "verify that lustre will correct the mode of OBJECTS".
594 # But with new MDS stack we don't care about the mode of local objects
595 # anymore, so this test is removed. See bug 22944 for more details.
599 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
600 skip "Only applicable to ldiskfs-based MDTs"
605 check_mount || return 41
608 echo "Remove mds config log"
609 if ! combined_mgs_mds ; then
613 do_facet mgs "$DEBUGFS -w -R 'unlink CONFIGS/$FSNAME-MDT0000' \
614 $(mgsdevname) || return \$?" || return $?
616 if ! combined_mgs_mds ; then
621 start_mds && return 42
624 run_test 17 "Verify failed mds_postsetup won't fail assertion (2936) (should return errs)"
627 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
628 skip "Only applicable to ldiskfs-based MDTs"
632 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
637 # check if current MDSSIZE is large enough
638 [ $MDSSIZE -ge $MIN ] && OK=1 && myMDSSIZE=$MDSSIZE && \
639 log "use MDSSIZE=$MDSSIZE"
641 # check if the global config has a large enough MDSSIZE
642 [ -z "$OK" -a ! -z "$STORED_MDSSIZE" ] && [ $STORED_MDSSIZE -ge $MIN ] && \
643 OK=1 && myMDSSIZE=$STORED_MDSSIZE && \
644 log "use STORED_MDSSIZE=$STORED_MDSSIZE"
646 # check if the block device is large enough
647 is_blkdev $SINGLEMDS $MDSDEV $MIN
648 local large_enough=$?
649 if [ -n "$OK" ]; then
650 [ $large_enough -ne 0 ] && OK=""
652 [ $large_enough -eq 0 ] && OK=1 && myMDSSIZE=$MIN &&
653 log "use device $MDSDEV with MIN=$MIN"
656 # check if a loopback device has enough space for fs metadata (5%)
658 if [ -z "$OK" ]; then
659 local SPACE=$(do_facet $SINGLEMDS "[ -f $MDSDEV -o ! -e $MDSDEV ] && df -P \\\$(dirname $MDSDEV)" |
660 awk '($1 != "Filesystem") {print $4}')
661 ! [ -z "$SPACE" ] && [ $SPACE -gt $((MIN / 20)) ] && \
662 OK=1 && myMDSSIZE=$MIN && \
663 log "use file $MDSDEV with MIN=$MIN"
666 [ -z "$OK" ] && skip_env "$MDSDEV too small for ${MIN}kB MDS" && return
669 echo "mount mds with large journal..."
671 local OLD_MDSSIZE=$MDSSIZE
675 echo "mount lustre system..."
677 check_mount || return 41
679 echo "check journal size..."
680 local FOUNDSIZE=$(do_facet $SINGLEMDS "$DEBUGFS -c -R 'stat <8>' $MDSDEV" | awk '/Size: / { print $NF; exit;}')
681 if [ $FOUNDSIZE -gt $((32 * 1024 * 1024)) ]; then
682 log "Success: mkfs creates large journals. Size: $((FOUNDSIZE >> 20))M"
684 error "expected journal size > 32M, found $((FOUNDSIZE >> 20))M"
692 run_test 18 "check mkfs creates large journals"
695 start_mds || return 1
696 stop_mds -f || return 2
698 run_test 19a "start/stop MDS without OSTs"
701 start_ost || return 1
702 stop_ost -f || return 2
704 run_test 19b "start/stop OSTs without MDS"
707 # first format the ost/mdt
711 check_mount || return 43
713 remount_client ro $MOUNT || return 44
714 touch $DIR/$tfile && echo "$DIR/$tfile created incorrectly" && return 45
715 [ -e $DIR/$tfile ] && echo "$DIR/$tfile exists incorrectly" && return 46
716 remount_client rw $MOUNT || return 47
718 [ ! -f $DIR/$tfile ] && echo "$DIR/$tfile missing" && return 48
719 MCNT=`grep -c $MOUNT /etc/mtab`
720 [ "$MCNT" -ne 1 ] && echo "$MOUNT in /etc/mtab $MCNT times" && return 49
725 run_test 20 "remount ro,rw mounts work and doesn't break /etc/mtab"
730 wait_osc_import_state mds ost FULL
734 run_test 21a "start mds before ost, stop ost first"
739 wait_osc_import_state mds ost FULL
743 run_test 21b "start ost before mds, stop mds first"
749 wait_osc_import_state mds ost2 FULL
753 #writeconf to remove all ost2 traces for subsequent tests
754 writeconf_or_reformat
756 run_test 21c "start mds between two osts, stop mds last"
759 if combined_mgs_mds ; then
760 skip "need separate mgs device" && return 0
770 wait_osc_import_state mds ost2 FULL
776 #writeconf to remove all ost2 traces for subsequent tests
777 writeconf_or_reformat
780 run_test 21d "start mgs then ost and then mds"
785 echo Client mount with ost in logs, but none running
787 # wait until mds connected to ost and open client connection
788 wait_osc_import_state mds ost FULL
791 # check_mount will block trying to contact ost
792 mcreate $DIR/$tfile || return 40
793 rm -f $DIR/$tfile || return 42
797 echo Client mount with a running ost
800 # if gss enabled, wait full time to let connection from
801 # mds to ost be established, due to the mismatch between
802 # initial connect timeout and gss context negotiation timeout.
803 # This perhaps could be remove after AT landed.
804 echo "sleep $((TIMEOUT + TIMEOUT + TIMEOUT))s"
805 sleep $((TIMEOUT + TIMEOUT + TIMEOUT))
808 wait_osc_import_state mds ost FULL
809 wait_osc_import_state client ost FULL
810 check_mount || return 41
815 run_test 22 "start a client before osts (should return errs)"
817 test_23a() { # was test_23
821 # force down client so that recovering mds waits for reconnect
822 local running=$(grep -c $MOUNT /proc/mounts) || true
823 if [ $running -ne 0 ]; then
824 echo "Stopping client $MOUNT (opts: -f)"
828 # enter recovery on mds
830 # try to start a new client
831 mount_client $MOUNT &
833 MOUNT_PID=$(ps -ef | grep "t lustre" | grep -v grep | awk '{print $2}')
834 MOUNT_LUSTRE_PID=`ps -ef | grep mount.lustre | grep -v grep | awk '{print $2}'`
835 echo mount pid is ${MOUNT_PID}, mount.lustre pid is ${MOUNT_LUSTRE_PID}
837 ps --ppid $MOUNT_LUSTRE_PID
838 echo "waiting for mount to finish"
840 # "ctrl-c" sends SIGINT but it usually (in script) does not work on child process
841 # SIGTERM works but it does not spread to offspring processses
842 kill -s TERM $MOUNT_PID
843 kill -s TERM $MOUNT_LUSTRE_PID
844 # we can not wait $MOUNT_PID because it is not a child of this shell
850 while [ "$WAIT" -lt "$MAX_WAIT" ]; do
852 PID1=$(ps -ef | awk '{print $2}' | grep -w $MOUNT_PID)
853 PID2=$(ps -ef | awk '{print $2}' | grep -w $MOUNT_LUSTRE_PID)
856 [ -z "$PID1" -a -z "$PID2" ] && break
857 echo "waiting for mount to finish ... "
858 WAIT=$(( WAIT + sleep))
860 if [ "$WAIT" -eq "$MAX_WAIT" ]; then
861 error "MOUNT_PID $MOUNT_PID and "\
862 "MOUNT_LUSTRE_PID $MOUNT_LUSTRE_PID still not killed in $WAIT secs"
865 stop_mds || error "stopping MDSes failed"
866 stop_ost || error "stopping OSSes failed"
868 run_test 23a "interrupt client during recovery mount delay"
873 test_23b() { # was test_23
876 # Simulate -EINTR during mount OBD_FAIL_LDLM_CLOSE_THREAD
877 lctl set_param fail_loc=0x80000313
881 run_test 23b "Simulate -EINTR during mount"
883 fs2mds_HOST=$mds_HOST
884 fs2ost_HOST=$ost_HOST
886 MDSDEV1_2=$fs2mds_DEV
887 OSTDEV1_2=$fs2ost_DEV
888 OSTDEV2_2=$fs3ost_DEV
892 echo "umount $MOUNT2 ..."
893 umount $MOUNT2 || true
894 echo "stopping fs2mds ..."
895 stop fs2mds -f || true
896 echo "stopping fs2ost ..."
897 stop fs2ost -f || true
901 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
903 if [ -z "$fs2ost_DEV" -o -z "$fs2mds_DEV" ]; then
904 is_blkdev $SINGLEMDS $MDSDEV && \
905 skip_env "mixed loopback and real device not working" && return
908 [ -n "$ost1_HOST" ] && fs2ost_HOST=$ost1_HOST
910 local fs2mdsdev=$(mdsdevname 1_2)
911 local fs2ostdev=$(ostdevname 1_2)
912 local fs2mdsvdev=$(mdsvdevname 1_2)
913 local fs2ostvdev=$(ostvdevname 1_2)
915 # test 8-char fsname as well
916 local FSNAME2=test1234
918 add fs2mds $(mkfs_opts mds1 ${fs2mdsdev} ) --nomgs --mgsnode=$MGSNID \
919 --fsname=${FSNAME2} --reformat $fs2mdsdev $fs2mdsvdev || exit 10
921 add fs2ost $(mkfs_opts ost1 ${fs2ostdev}) --fsname=${FSNAME2} \
922 --reformat $fs2ostdev $fs2ostvdev || exit 10
925 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS && trap cleanup_24a EXIT INT
926 start fs2ost $fs2ostdev $OST_MOUNT_OPTS
928 $MOUNT_CMD $MGSNID:/${FSNAME2} $MOUNT2 || return 1
930 check_mount || return 2
931 # files written on 1 should not show up on 2
932 cp /etc/passwd $DIR/$tfile
934 [ -e $MOUNT2/$tfile ] && error "File bleed" && return 7
937 cp /etc/passwd $MOUNT2/b || return 3
938 rm $MOUNT2/b || return 4
939 # 2 is actually mounted
940 grep $MOUNT2' ' /proc/mounts > /dev/null || return 5
942 facet_failover fs2mds
943 facet_failover fs2ost
946 # the MDS must remain up until last MDT
948 MDS=$(do_facet $SINGLEMDS "lctl get_param -n devices" | awk '($3 ~ "mdt" && $4 ~ "MDT") { print $4 }' | head -1)
949 [ -z "$MDS" ] && error "No MDT" && return 8
951 cleanup_nocli || return 6
953 run_test 24a "Multiple MDTs on a single node"
956 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
958 if [ -z "$fs2mds_DEV" ]; then
959 local dev=${SINGLEMDS}_dev
961 is_blkdev $SINGLEMDS $MDSDEV && \
962 skip_env "mixed loopback and real device not working" && return
965 local fs2mdsdev=$(mdsdevname 1_2)
966 local fs2mdsvdev=$(mdsvdevname 1_2)
968 add fs2mds $(mkfs_opts mds1 ${fs2mdsdev} ) --mgs --fsname=${FSNAME}2 \
969 --reformat $fs2mdsdev $fs2mdsvdev || exit 10
971 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS && return 2
974 run_test 24b "Multiple MGSs on a single node (should return err)"
978 check_mount || return 2
979 local MODULES=$($LCTL modules | awk '{ print $2 }')
980 rmmod $MODULES 2>/dev/null || true
983 run_test 25 "Verify modules are referenced"
987 # we need modules before mount for sysctl, so make sure...
988 do_facet $SINGLEMDS "lsmod | grep -q lustre || modprobe lustre"
989 #define OBD_FAIL_MDS_FS_SETUP 0x135
990 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000135"
991 start_mds && echo MDS started && return 1
992 lctl get_param -n devices
993 DEVS=$(lctl get_param -n devices | egrep -v MG | wc -l)
994 [ $DEVS -gt 0 ] && return 2
995 # start mds to drop writeconf setting
996 start_mds || return 3
998 unload_modules_conf || return $?
1000 run_test 26 "MDT startup failure cleans LOV (should return errs)"
1003 start_ost || return 1
1004 start_mds || return 2
1005 echo "Requeue thread should have started: "
1006 ps -e | grep ll_cfg_requeue
1007 set_conf_param_and_check ost1 \
1008 "lctl get_param -n obdfilter.$FSNAME-OST0000.client_cache_seconds" \
1009 "$FSNAME-OST0000.ost.client_cache_seconds" || return 3
1012 run_test 27a "Reacquire MGS lock if OST started first"
1017 local device=$(do_facet $SINGLEMDS "lctl get_param -n devices" |
1018 awk '($3 ~ "mdt" && $4 ~ "MDT0000") { print $4 }')
1020 facet_failover $SINGLEMDS
1021 set_conf_param_and_check $SINGLEMDS \
1022 "lctl get_param -n mdt.$device.identity_acquire_expire" \
1023 "$device.mdt.identity_acquire_expire" || return 3
1024 set_conf_param_and_check client \
1025 "lctl get_param -n mdc.$device-mdc-*.max_rpcs_in_flight"\
1026 "$device.mdc.max_rpcs_in_flight" || return 4
1030 run_test 27b "Reacquire MGS lock after failover"
1034 TEST="lctl get_param -n llite.$FSNAME-*.max_read_ahead_whole_mb"
1035 PARAM="$FSNAME.llite.max_read_ahead_whole_mb"
1037 FINAL=$(($ORIG + 1))
1038 set_conf_param_and_check client "$TEST" "$PARAM" $FINAL || return 3
1039 FINAL=$(($FINAL + 1))
1040 set_conf_param_and_check client "$TEST" "$PARAM" $FINAL || return 4
1041 umount_client $MOUNT || return 200
1044 if [ $RESULT -ne $FINAL ]; then
1045 echo "New config not seen: wanted $FINAL got $RESULT"
1048 echo "New config success: got $RESULT"
1050 set_conf_param_and_check client "$TEST" "$PARAM" $ORIG || return 5
1053 run_test 28 "permanent parameter setting"
1055 test_28a() { # LU-4221
1056 [ $(lustre_version_code ost1) -eq $(version_code 2.5.0) -o \
1057 $(lustre_version_code ost1) -ge $(version_code 2.5.52) ] ||
1058 { skip "Need OST version >= 2.5.52 or = 2.5.0" &&
1060 [ "$(facet_fstype ost1)" = "zfs" ] &&
1061 skip "LU-4221: no such proc params for ZFS OSTs" && return
1068 local device="$FSNAME-OST0000"
1072 # In this test we will set three kinds of proc parameters with
1074 # 1. the ones moved from the OFD to the OSD, and only their
1075 # symlinks kept in obdfilter
1076 # 2. non-symlink ones in the OFD
1077 # 3. non-symlink ones in the OSD
1080 # prepare a symlink parameter in the OFD
1081 name="writethrough_cache_enable"
1082 param="$device.ost.$name"
1083 cmd="$LCTL get_param -n obdfilter.$device.$name"
1085 # conf_param the symlink parameter in the OFD
1086 old=$(do_facet ost1 $cmd)
1087 new=$(((old + 1) % 2))
1088 set_conf_param_and_check ost1 "$cmd" "$param" $new ||
1089 error "lctl conf_param $device.ost.$param=$new failed"
1091 # conf_param the target parameter in the OSD
1092 param="$device.osd.$name"
1093 cmd="$LCTL get_param -n osd-*.$device.$name"
1094 set_conf_param_and_check ost1 "$cmd" "$param" $old ||
1095 error "lctl conf_param $device.osd.$param=$old failed"
1098 # prepare a non-symlink parameter in the OFD
1099 name="client_cache_seconds"
1100 param="$device.ost.$name"
1101 cmd="$LCTL get_param -n obdfilter.$device.$name"
1103 # conf_param the parameter in the OFD
1104 old=$(do_facet ost1 $cmd)
1106 set_conf_param_and_check ost1 "$cmd" "$param" $new ||
1107 error "lctl conf_param $device.ost.$param=$new failed"
1108 set_conf_param_and_check ost1 "$cmd" "$param" $old ||
1109 error "lctl conf_param $device.ost.$param=$old failed"
1112 # prepare a non-symlink parameter in the OSD
1114 param="$device.osd.$name"
1115 cmd="$LCTL get_param -n osd-*.$device.$name"
1117 # conf_param the parameter in the OSD
1118 old=$(do_facet ost1 $cmd)
1119 new=$(((old + 1) % 2))
1120 set_conf_param_and_check ost1 "$cmd" "$param" $new ||
1121 error "lctl conf_param $device.osd.$param=$new failed"
1122 set_conf_param_and_check ost1 "$cmd" "$param" $old ||
1123 error "lctl conf_param $device.osd.$param=$old failed"
1127 run_test 28a "set symlink parameters permanently with conf_param"
1130 [ "$OSTCOUNT" -lt "2" ] && skip_env "$OSTCOUNT < 2, skipping" && return
1131 setup > /dev/null 2>&1
1135 local PARAM="$FSNAME-OST0001.osc.active"
1136 local PROC_ACT="osc.$FSNAME-OST0001-osc-[^M]*.active"
1137 local PROC_UUID="osc.$FSNAME-OST0001-osc-[^M]*.ost_server_uuid"
1139 ACTV=$(lctl get_param -n $PROC_ACT)
1141 set_conf_param_and_check client \
1142 "lctl get_param -n $PROC_ACT" "$PARAM" $DEAC || return 2
1143 # also check ost_server_uuid status
1144 RESULT=$(lctl get_param -n $PROC_UUID | grep DEACTIV)
1145 if [ -z "$RESULT" ]; then
1146 echo "Live client not deactivated: $(lctl get_param -n $PROC_UUID)"
1149 echo "Live client success: got $RESULT"
1153 for num in $(seq $MDSCOUNT); do
1154 local mdtosc=$(get_mdtosc_proc_path mds${num} $FSNAME-OST0001)
1155 local MPROC="osc.$mdtosc.active"
1160 RESULT=$(do_facet mds${num} " lctl get_param -n $MPROC")
1161 [ ${PIPESTATUS[0]} = 0 ] || error "Can't read $MPROC"
1162 if [ $RESULT -eq $DEAC ]; then
1163 echo -n "MDT deactivated also after"
1164 echo "$WAIT sec (got $RESULT)"
1168 if [ $WAIT -eq $MAX ]; then
1169 echo -n "MDT not deactivated: wanted $DEAC"
1173 echo "Waiting $(($MAX - $WAIT))secs for MDT deactivated"
1176 # test new client starts deactivated
1177 umount_client $MOUNT || return 200
1179 RESULT=$(lctl get_param -n $PROC_UUID | grep DEACTIV | grep NEW)
1180 if [ -z "$RESULT" ]; then
1181 echo "New client not deactivated from start: $(lctl get_param -n $PROC_UUID)"
1184 echo "New client success: got $RESULT"
1187 # make sure it reactivates
1188 set_conf_param_and_check client \
1189 "lctl get_param -n $PROC_ACT" "$PARAM" $ACTV || return 6
1191 umount_client $MOUNT
1194 #writeconf to remove all ost2 traces for subsequent tests
1195 writeconf_or_reformat
1197 run_test 29 "permanently remove an OST"
1202 echo Big config llog
1203 TEST="lctl get_param -n llite.$FSNAME-*.max_read_ahead_whole_mb"
1205 LIST=(1 2 3 4 5 4 3 2 1 2 3 4 5 4 3 2 1 2 3 4 5)
1206 for i in ${LIST[@]}; do
1207 set_conf_param_and_check client "$TEST" \
1208 "$FSNAME.llite.max_read_ahead_whole_mb" $i || return 3
1210 # make sure client restart still works
1211 umount_client $MOUNT
1212 mount_client $MOUNT || return 4
1213 [ "$($TEST)" -ne "$i" ] && error "Param didn't stick across restart $($TEST) != $i"
1216 echo Erase parameter setting
1217 do_facet mgs "$LCTL conf_param -d $FSNAME.llite.max_read_ahead_whole_mb" || return 6
1218 umount_client $MOUNT
1219 mount_client $MOUNT || return 6
1221 echo "deleted (default) value=$FINAL, orig=$ORIG"
1222 # assumes this parameter started at the default value
1223 [ "$FINAL" -eq "$ORIG" ] || fail "Deleted value=$FINAL, orig=$ORIG"
1227 run_test 30a "Big config llog and conf_param deletion"
1232 # Make a fake nid. Use the OST nid, and add 20 to the least significant
1233 # numerical part of it. Hopefully that's not already a failover address for
1235 OSTNID=$(do_facet ost1 "$LCTL get_param nis" | tail -1 | awk '{print $1}')
1236 ORIGVAL=$(echo $OSTNID | egrep -oi "[0-9]*@")
1237 NEWVAL=$((($(echo $ORIGVAL | egrep -oi "[0-9]*") + 20) % 256))
1238 NEW=$(echo $OSTNID | sed "s/$ORIGVAL/$NEWVAL@/")
1239 echo "Using fake nid $NEW"
1241 TEST="$LCTL get_param -n osc.$FSNAME-OST0000-osc-[^M]*.import |
1242 grep failover_nids | sed -n 's/.*\($NEW\).*/\1/p'"
1243 set_conf_param_and_check client "$TEST" \
1244 "$FSNAME-OST0000.failover.node" $NEW ||
1245 error "didn't add failover nid $NEW"
1246 NIDS=$($LCTL get_param -n osc.$FSNAME-OST0000-osc-[^M]*.import |
1249 NIDCOUNT=$(($(echo "$NIDS" | wc -w) - 3))
1250 echo "should have 2 failover nids: $NIDCOUNT"
1251 [ $NIDCOUNT -eq 2 ] || error "Failover nid not added"
1252 do_facet mgs "$LCTL conf_param -d $FSNAME-OST0000.failover.node" ||
1253 error "conf_param delete failed"
1254 umount_client $MOUNT
1255 mount_client $MOUNT || return 3
1257 NIDS=$($LCTL get_param -n osc.$FSNAME-OST0000-osc-[^M]*.import |
1260 NIDCOUNT=$(($(echo "$NIDS" | wc -w) - 3))
1261 echo "only 1 final nid should remain: $NIDCOUNT"
1262 [ $NIDCOUNT -eq 1 ] || error "Failover nids not removed"
1266 run_test 30b "Remove failover nids"
1268 test_31() { # bug 10734
1269 # ipaddr must not exist
1270 $MOUNT_CMD 4.3.2.1@tcp:/lustre $MOUNT || true
1273 run_test 31 "Connect to non-existent node (shouldn't crash)"
1277 T32_BLIMIT=20480 # Kbytes
1281 # This is not really a test but a tool to create new disk
1282 # image tarballs for the upgrade tests.
1284 # Disk image tarballs should be created on single-node
1285 # clusters by running this test with default configurations
1286 # plus a few mandatory environment settings that are verified
1287 # at the beginning of the test.
1289 test_32newtarball() {
1293 local tmp=$TMP/t32_image_create
1295 if [ $FSNAME != t32fs -o $MDSCOUNT -ne 1 -o \
1296 \( -z "$MDSDEV" -a -z "$MDSDEV1" \) -o $OSTCOUNT -ne 1 -o \
1297 -z "$OSTDEV1" ]; then
1298 error "Needs FSNAME=t32fs MDSCOUNT=1 MDSDEV1=<nonexistent_file>" \
1299 "(or MDSDEV, in the case of b1_8) OSTCOUNT=1" \
1300 "OSTDEV1=<nonexistent_file>"
1304 echo "Found stale $tmp"
1309 tar cf - -C $src . | tar xf - -C $tmp/src
1310 dd if=/dev/zero of=$tmp/src/t32_qf_old bs=1M \
1311 count=$(($T32_BLIMIT / 1024 / 2))
1312 chown $T32_QID.$T32_QID $tmp/src/t32_qf_old
1314 # format ost with comma-separated NIDs to verify LU-4460
1315 local failnid="$(h2$NETTYPE 1.2.3.4),$(h2$NETTYPE 4.3.2.1)"
1316 MGSNID="$MGSNID,$MGSNID" OSTOPT="--failnode=$failnid" formatall
1320 [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.3.50) ] &&
1321 $LFS quotacheck -ug /mnt/$FSNAME
1322 $LFS setquota -u $T32_QID -b 0 -B $T32_BLIMIT -i 0 -I $T32_ILIMIT \
1325 tar cf - -C $tmp/src . | tar xf - -C /mnt/$FSNAME
1332 ls -Rni --time-style=+%s >$tmp/img/list
1333 find . ! -name .lustre -type f -exec sha1sum {} \; |
1334 sort -k 2 >$tmp/img/sha1sums
1336 $LCTL get_param -n version | head -n 1 |
1337 sed -e 's/^lustre: *//' >$tmp/img/commit
1339 [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.3.50) ] &&
1340 $LFS quotaon -ug /mnt/$FSNAME
1341 $LFS quota -u $T32_QID -v /mnt/$FSNAME
1342 $LFS quota -v -u $T32_QID /mnt/$FSNAME |
1343 awk 'BEGIN { num='1' } { if ($1 == "'/mnt/$FSNAME'") \
1344 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1345 | tr -d "*" > $tmp/img/bspace
1346 $LFS quota -v -u $T32_QID /mnt/$FSNAME |
1347 awk 'BEGIN { num='5' } { if ($1 == "'/mnt/$FSNAME'") \
1348 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1349 | tr -d "*" > $tmp/img/ispace
1354 find -type f -exec sha1sum {} \; | sort -k 2 >$tmp/sha1sums.src
1357 if ! diff -u $tmp/sha1sums.src $tmp/img/sha1sums; then
1358 echo "Data verification failed"
1361 uname -r >$tmp/img/kernel
1362 uname -m >$tmp/img/arch
1364 mv ${MDSDEV1:-$MDSDEV} $tmp/img
1365 mv $OSTDEV1 $tmp/img
1367 version=$(sed -e 's/\(^[0-9]\+\.[0-9]\+\)\(.*$\)/\1/' $tmp/img/commit |
1368 sed -e 's/\./_/g') # E.g., "1.8.7" -> "1_8"
1371 tar cjvf $dst/disk$version-$(facet_fstype $SINGLEMDS).tar.bz2 -S *
1376 #run_test 32newtarball "Create a new test_32 disk image tarball for this version"
1379 # The list of applicable tarballs is returned via the caller's
1380 # variable "tarballs".
1383 local node=$(facet_active_host $SINGLEMDS)
1384 local r="do_node $node"
1386 if [ "$CLIENTONLY" ]; then
1387 skip "Client-only testing"
1391 if ! $r which $TUNEFS; then
1392 skip_env "tunefs.lustre required on $node"
1396 local IMGTYPE=$(facet_fstype $SINGLEMDS)
1398 tarballs=$($r find $RLUSTRE/tests -maxdepth 1 -name \'disk*-$IMGTYPE.tar.bz2\')
1400 if [ -z "$tarballs" ]; then
1401 skip "No applicable tarballs found"
1406 t32_test_cleanup() {
1408 local fstype=$(facet_fstype $SINGLEMDS)
1411 if $shall_cleanup_lustre; then
1412 umount $tmp/mnt/lustre || rc=$?
1414 if $shall_cleanup_mdt; then
1415 $r umount -d $tmp/mnt/mdt || rc=$?
1417 if $shall_cleanup_mdt1; then
1418 $r umount -d $tmp/mnt/mdt1 || rc=$?
1420 if $shall_cleanup_ost; then
1421 $r umount -d $tmp/mnt/ost || rc=$?
1426 if [ $fstype == "zfs" ]; then
1427 $r $ZPOOL destroy t32fs-mdt1 || rc=$?
1428 $r $ZPOOL destroy t32fs-ost1 || rc=$?
1433 t32_bits_per_long() {
1435 # Yes, this is not meant to be perfect.
1445 t32_reload_modules() {
1447 local all_removed=false
1450 while ((i < 20)); do
1451 echo "Unloading modules on $node: Attempt $i"
1452 do_rpc_nodes $node $LUSTRE_RMMOD $(facet_fstype $SINGLEMDS) &&
1454 do_rpc_nodes $node check_mem_leak || return 1
1455 if $all_removed; then
1456 do_rpc_nodes $node load_modules
1462 echo "Unloading modules on $node: Given up"
1466 t32_wait_til_devices_gone() {
1472 echo wait for devices to go
1473 while ((i < 20)); do
1474 devices=$(do_rpc_nodes $node $LCTL device_list | wc -l)
1475 loops=$(do_rpc_nodes $node losetup -a | grep -c t32)
1476 ((devices == 0 && loops == 0)) && return 0
1480 echo "waiting for dev on $node: dev $devices loop $loops given up"
1481 do_rpc_nodes $node "losetup -a"
1482 do_rpc_nodes $node "$LCTL devices_list"
1486 t32_verify_quota() {
1490 local fstype=$(facet_fstype $SINGLEMDS)
1494 $LFS quota -u $T32_QID -v $mnt
1496 qval=$($LFS quota -v -u $T32_QID $mnt |
1497 awk 'BEGIN { num='1' } { if ($1 == "'$mnt'") \
1498 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1500 [ $qval -eq $img_bspace ] || {
1501 echo "bspace, act:$qval, exp:$img_bspace"
1505 qval=$($LFS quota -v -u $T32_QID $mnt |
1506 awk 'BEGIN { num='5' } { if ($1 == "'$mnt'") \
1507 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1509 [ $qval -eq $img_ispace ] || {
1510 echo "ispace, act:$qval, exp:$img_ispace"
1514 qval=$($LFS quota -v -u $T32_QID $mnt |
1515 awk 'BEGIN { num='3' } { if ($1 == "'$mnt'") \
1516 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1518 [ $qval -eq $T32_BLIMIT ] || {
1519 echo "blimit, act:$qval, exp:$T32_BLIMIT"
1523 qval=$($LFS quota -v -u $T32_QID $mnt |
1524 awk 'BEGIN { num='7' } { if ($1 == "'$mnt'") \
1525 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1527 [ $qval -eq $T32_ILIMIT ] || {
1528 echo "ilimit, act:$qval, exp:$T32_ILIMIT"
1532 do_node $node $LCTL conf_param $fsname.quota.mdt=ug
1533 cmd="$LCTL get_param -n osd-$fstype.$fsname-MDT0000"
1534 cmd=$cmd.quota_slave.enabled
1535 wait_update $node "$cmd" "ug" || {
1536 echo "Enable mdt quota failed"
1540 do_node $node $LCTL conf_param $fsname.quota.ost=ug
1541 cmd="$LCTL get_param -n osd-$fstype.$fsname-OST0000"
1542 cmd=$cmd.quota_slave.enabled
1543 wait_update $node "$cmd" "ug" || {
1544 echo "Enable ost quota failed"
1549 runas -u $T32_QID -g $T32_QID dd if=/dev/zero of=$mnt/t32_qf_new \
1550 bs=1M count=$(($T32_BLIMIT / 1024)) oflag=sync && {
1551 echo "Write succeed, but expect -EDQUOT"
1554 rm -f $mnt/t32_qf_new
1556 runas -u $T32_QID -g $T32_QID createmany -m $mnt/t32_qf_ \
1558 echo "Create succeed, but expect -EDQUOT"
1561 unlinkmany $mnt/t32_qf_ $T32_ILIMIT
1569 local dne_upgrade=${dne_upgrade:-"no"}
1570 local ff_convert=${ff_convert:-"no"}
1571 local shall_cleanup_mdt=false
1572 local shall_cleanup_mdt1=false
1573 local shall_cleanup_ost=false
1574 local shall_cleanup_lustre=false
1575 local node=$(facet_active_host $SINGLEMDS)
1576 local r="do_node $node"
1577 local node2=$(facet_active_host mds2)
1585 local nid=$($r $LCTL list_nids | head -1)
1591 local fstype=$(facet_fstype $SINGLEMDS)
1592 local mdt_dev=$tmp/mdt
1593 local ost_dev=$tmp/ost
1595 trap 'trap - RETURN; t32_test_cleanup' RETURN
1597 mkdir -p $tmp/mnt/lustre
1598 $r mkdir -p $tmp/mnt/{mdt,ost}
1599 $r tar xjvf $tarball -S -C $tmp || {
1600 error_noexit "Unpacking the disk image tarball"
1603 img_commit=$($r cat $tmp/commit)
1604 img_kernel=$($r cat $tmp/kernel)
1605 img_arch=$($r cat $tmp/arch)
1606 img_bspace=$($r cat $tmp/bspace)
1607 img_ispace=$($r cat $tmp/ispace)
1608 echo "Upgrading from $(basename $tarball), created with:"
1609 echo " Commit: $img_commit"
1610 echo " Kernel: $img_kernel"
1611 echo " Arch: $img_arch"
1612 echo "OST version: $(get_lustre_version ost1)"
1614 # The convertion can be made only when both of the following
1615 # conditions are satisfied:
1616 # - client img version < 2.3.64
1617 # - ost server version >= 2.5
1618 [ $(version_code $img_commit) -ge $(version_code 2.3.64) -o \
1619 $(lustre_version_code ost1) -lt $(version_code 2.5.0) ] &&
1622 if [ $fstype == "zfs" ]; then
1624 $r $ZPOOL import -f -d $tmp t32fs-mdt1
1625 $r $ZPOOL import -f -d $tmp t32fs-ost1
1626 mdt_dev=t32fs-mdt1/mdt1
1627 ost_dev=t32fs-ost1/ost1
1628 wait_update_facet $SINGLEMDS "$ZPOOL list |
1629 awk '/^t32fs-mdt1/ { print \\\$1 }'" "t32fs-mdt1" || {
1630 error_noexit "import zfs pool failed"
1635 $r $LCTL set_param debug="$PTLDEBUG"
1637 $r $TUNEFS --dryrun $mdt_dev || {
1639 error_noexit "tunefs.lustre before mounting the MDT"
1642 if [ "$writeconf" ]; then
1644 if [ $fstype == "ldiskfs" ]; then
1646 $r $TUNEFS --quota $mdt_dev || {
1648 error_noexit "Enable mdt quota feature"
1653 if [ -n "$($LCTL list_nids | grep -v '\(tcp\|lo\)[[:digit:]]*$')" ]; then
1654 [[ $(lustre_version_code mgs) -ge $(version_code 2.3.59) ]] ||
1655 { skip "LU-2200: Cannot run over Inifiniband w/o lctl replace_nids "
1656 "(Need MGS version at least 2.3.59)"; return 0; }
1658 local osthost=$(facet_active_host ost1)
1659 local ostnid=$(do_node $osthost $LCTL list_nids | head -1)
1662 if [ $fstype == "ldiskfs" ]; then
1665 $r $MOUNT_CMD -o $mopts $mdt_dev $tmp/mnt/mdt
1666 $r lctl replace_nids $fsname-OST0000 $ostnid
1667 $r lctl replace_nids $fsname-MDT0000 $nid
1668 $r umount -d $tmp/mnt/mdt
1671 mopts=exclude=$fsname-OST0000
1672 if [ $fstype == "ldiskfs" ]; then
1677 t32_wait_til_devices_gone $node
1679 $r $MOUNT_CMD -o $mopts $mdt_dev $tmp/mnt/mdt || {
1681 error_noexit "Mounting the MDT"
1684 shall_cleanup_mdt=true
1686 if [ "$dne_upgrade" != "no" ]; then
1687 local fs2mdsdev=$(mdsdevname 1_2)
1688 local fs2mdsvdev=$(mdsvdevname 1_2)
1690 echo "mkfs new MDT on ${fs2mdsdev}...."
1691 if [ $(facet_fstype mds1) == ldiskfs ]; then
1692 mkfsoptions="--mkfsoptions=\\\"-J size=8\\\""
1695 add fs2mds $(mkfs_opts mds2 $fs2mdsdev $fsname) --reformat \
1696 $mkfsoptions $fs2mdsdev $fs2mdsvdev > /dev/null || {
1697 error_noexit "Mkfs new MDT failed"
1701 $r $TUNEFS --dryrun $fs2mdsdev || {
1702 error_noexit "tunefs.lustre before mounting the MDT"
1706 echo "mount new MDT....$fs2mdsdev"
1707 $r mkdir -p $tmp/mnt/mdt1
1708 $r $MOUNT_CMD -o $mopts $fs2mdsdev $tmp/mnt/mdt1 || {
1709 error_noexit "mount mdt1 failed"
1712 shall_cleanup_mdt1=true
1715 uuid=$($r $LCTL get_param -n mdt.$fsname-MDT0000.uuid) || {
1716 error_noexit "Getting MDT UUID"
1719 if [ "$uuid" != $fsname-MDT0000_UUID ]; then
1720 error_noexit "Unexpected MDT UUID: \"$uuid\""
1724 $r $TUNEFS --dryrun $ost_dev || {
1725 error_noexit "tunefs.lustre before mounting the OST"
1728 if [ "$writeconf" ]; then
1729 mopts=mgsnode=$nid,$writeconf
1730 if [ $fstype == "ldiskfs" ]; then
1732 $r $TUNEFS --quota $ost_dev || {
1734 error_noexit "Enable ost quota feature"
1740 if [ $fstype == "ldiskfs" ]; then
1744 $r $MOUNT_CMD -o $mopts $ost_dev $tmp/mnt/ost || {
1745 error_noexit "Mounting the OST"
1748 shall_cleanup_ost=true
1750 uuid=$($r $LCTL get_param -n obdfilter.$fsname-OST0000.uuid) || {
1751 error_noexit "Getting OST UUID"
1754 if [ "$uuid" != $fsname-OST0000_UUID ]; then
1755 error_noexit "Unexpected OST UUID: \"$uuid\""
1759 $r $LCTL conf_param $fsname-OST0000.osc.max_dirty_mb=15 || {
1760 error_noexit "Setting \"max_dirty_mb\""
1763 $r $LCTL conf_param $fsname-OST0000.failover.node=$nid || {
1764 error_noexit "Setting OST \"failover.node\""
1767 $r $LCTL conf_param $fsname-MDT0000.mdc.max_rpcs_in_flight=9 || {
1768 error_noexit "Setting \"max_rpcs_in_flight\""
1771 $r $LCTL conf_param $fsname-MDT0000.failover.node=$nid || {
1772 error_noexit "Setting MDT \"failover.node\""
1775 $r $LCTL pool_new $fsname.interop || {
1776 error_noexit "Setting \"interop\""
1779 $r $LCTL conf_param $fsname-MDT0000.lov.stripesize=4M || {
1780 error_noexit "Setting \"lov.stripesize\""
1784 if [ "$ff_convert" != "no" -a $(facet_fstype ost1) == "ldiskfs" ]; then
1785 $r $LCTL lfsck_start -M $fsname-OST0000 || {
1786 error_noexit "Start OI scrub on OST0"
1790 # The oi_scrub should be on ost1, but for test_32(),
1791 # all on the SINGLEMDS.
1792 wait_update_facet $SINGLEMDS "$LCTL get_param -n \
1793 osd-ldiskfs.$fsname-OST0000.oi_scrub |
1794 awk '/^status/ { print \\\$2 }'" "completed" 30 || {
1795 error_noexit "Failed to get the expected 'completed'"
1799 local UPDATED=$($r $LCTL get_param -n \
1800 osd-ldiskfs.$fsname-OST0000.oi_scrub |
1801 awk '/^updated/ { print $2 }')
1802 [ $UPDATED -ge 1 ] || {
1803 error_noexit "Only $UPDATED objects have been converted"
1808 if [ "$dne_upgrade" != "no" ]; then
1809 $r $LCTL conf_param \
1810 $fsname-MDT0001.mdc.max_rpcs_in_flight=9 || {
1811 error_noexit "Setting MDT1 \"max_rpcs_in_flight\""
1814 $r $LCTL conf_param $fsname-MDT0001.failover.node=$nid || {
1815 error_noexit "Setting MDT1 \"failover.node\""
1818 $r $LCTL conf_param $fsname-MDT0001.lov.stripesize=4M || {
1819 error_noexit "Setting MDT1 \"lov.stripesize\""
1825 if [ "$writeconf" ]; then
1826 $MOUNT_CMD $nid:/$fsname $tmp/mnt/lustre || {
1827 error_noexit "Mounting the client"
1830 shall_cleanup_lustre=true
1831 $LCTL set_param debug="$PTLDEBUG"
1833 t32_verify_quota $node $fsname $tmp/mnt/lustre || {
1834 error_noexit "verify quota failed"
1838 if [ "$dne_upgrade" != "no" ]; then
1839 $LFS mkdir -i 1 $tmp/mnt/lustre/remote_dir || {
1840 error_noexit "set remote dir failed"
1844 pushd $tmp/mnt/lustre
1845 tar -cf - . --exclude=./remote_dir |
1846 tar -xvf - -C remote_dir 1>/dev/null || {
1847 error_noexit "cp to remote dir failed"
1853 dd if=/dev/zero of=$tmp/mnt/lustre/tmp_file bs=10k count=10 || {
1854 error_noexit "dd failed"
1857 rm -rf $tmp/mnt/lustre/tmp_file || {
1858 error_noexit "rm failed"
1862 if $r test -f $tmp/sha1sums; then
1863 # LU-2393 - do both sorts on same node to ensure locale
1865 $r cat $tmp/sha1sums | sort -k 2 >$tmp/sha1sums.orig
1866 if [ "$dne_upgrade" != "no" ]; then
1867 pushd $tmp/mnt/lustre/remote_dir
1869 pushd $tmp/mnt/lustre
1872 find ! -name .lustre -type f -exec sha1sum {} \; |
1873 sort -k 2 >$tmp/sha1sums || {
1874 error_noexit "sha1sum"
1878 if ! diff -ub $tmp/sha1sums.orig $tmp/sha1sums; then
1879 error_noexit "sha1sum verification failed"
1883 echo "sha1sum verification skipped"
1886 if [ "$dne_upgrade" != "no" ]; then
1887 rm -rf $tmp/mnt/lustre/remote_dir || {
1888 error_noexit "remove remote dir failed"
1893 if $r test -f $tmp/list; then
1895 # There is not a Test Framework API to copy files to or
1896 # from a remote node.
1898 # LU-2393 - do both sorts on same node to ensure locale
1900 $r cat $tmp/list | sort -k 6 >$tmp/list.orig
1901 pushd $tmp/mnt/lustre
1902 ls -Rni --time-style=+%s | sort -k 6 >$tmp/list || {
1908 # 32-bit and 64-bit clients use different algorithms to
1909 # convert FIDs into inode numbers. Hence, remove the inode
1910 # numbers from the lists, if the original list was created
1911 # on an architecture with different number of bits per
1914 if [ $(t32_bits_per_long $(uname -m)) != \
1915 $(t32_bits_per_long $img_arch) ]; then
1916 echo "Different number of bits per \"long\" from the disk image"
1917 for list in list.orig list; do
1918 sed -i -e 's/^[0-9]\+[ \t]\+//' $tmp/$list
1921 if ! diff -ub $tmp/list.orig $tmp/list; then
1922 error_noexit "list verification failed"
1926 echo "list verification skipped"
1930 # When adding new data verification tests, please check for
1931 # the presence of the required reference files first, like
1932 # the "sha1sums" and "list" tests above, to avoid the need to
1933 # regenerate every image for each test addition.
1936 nrpcs_orig=$($LCTL get_param \
1937 -n mdc.*MDT0000*.max_rpcs_in_flight) || {
1938 error_noexit "Getting \"max_rpcs_in_flight\""
1941 nrpcs=$((nrpcs_orig + 5))
1942 $r $LCTL conf_param $fsname-MDT0000.mdc.max_rpcs_in_flight=$nrpcs || {
1943 error_noexit "Changing \"max_rpcs_in_flight\""
1946 wait_update $HOSTNAME "$LCTL get_param \
1947 -n mdc.*MDT0000*.max_rpcs_in_flight" $nrpcs || {
1948 error_noexit "Verifying \"max_rpcs_in_flight\""
1952 umount $tmp/mnt/lustre || {
1953 error_noexit "Unmounting the client"
1956 shall_cleanup_lustre=false
1958 if [ "$dne_upgrade" != "no" ]; then
1959 $r umount -d $tmp/mnt/mdt1 || {
1960 error_noexit "Unmounting the MDT2"
1963 shall_cleanup_mdt1=false
1966 $r umount -d $tmp/mnt/mdt || {
1967 error_noexit "Unmounting the MDT"
1970 shall_cleanup_mdt=false
1972 $r umount -d $tmp/mnt/ost || {
1973 error_noexit "Unmounting the OST"
1976 shall_cleanup_ost=false
1978 t32_reload_modules $node || {
1979 error_noexit "Reloading modules"
1983 # mount a second time to make sure we didnt leave upgrade flag on
1984 $r $TUNEFS --dryrun $mdt_dev || {
1986 error_noexit "tunefs.lustre before remounting the MDT"
1990 mopts=exclude=$fsname-OST0000
1991 if [ $fstype == "ldiskfs" ]; then
1994 $r $MOUNT_CMD -o $mopts $mdt_dev $tmp/mnt/mdt || {
1995 error_noexit "Remounting the MDT"
1998 shall_cleanup_mdt=true
2008 for tarball in $tarballs; do
2009 t32_test $tarball || let "rc += $?"
2013 run_test 32a "Upgrade (not live)"
2021 for tarball in $tarballs; do
2022 t32_test $tarball writeconf || let "rc += $?"
2026 run_test 32b "Upgrade with writeconf"
2033 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
2035 for tarball in $tarballs; do
2036 dne_upgrade=yes t32_test $tarball writeconf || rc=$?
2040 run_test 32c "dne upgrade test"
2048 for tarball in $tarballs; do
2049 ff_convert=yes t32_test $tarball || rc=$?
2053 run_test 32d "convert ff test"
2055 test_33a() { # bug 12333, was test_33
2057 local FSNAME2=test-123
2058 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
2061 [ -n "$ost1_HOST" ] && fs2ost_HOST=$ost1_HOST
2063 if [ -z "$fs2ost_DEV" -o -z "$fs2mds_DEV" ]; then
2064 local dev=${SINGLEMDS}_dev
2065 local MDSDEV=${!dev}
2066 is_blkdev $SINGLEMDS $MDSDEV && \
2067 skip_env "mixed loopback and real device not working" && return
2070 local fs2mdsdev=$(mdsdevname 1_2)
2071 local fs2ostdev=$(ostdevname 1_2)
2072 local fs2mdsvdev=$(mdsvdevname 1_2)
2073 local fs2ostvdev=$(ostvdevname 1_2)
2075 if [ $(facet_fstype mds1) == ldiskfs ]; then
2076 mkfsoptions="--mkfsoptions=\\\"-J size=8\\\"" # See bug 17931.
2079 add fs2mds $(mkfs_opts mds1 ${fs2mdsdev}) --mgs --fsname=${FSNAME2} \
2080 --reformat $mkfsoptions $fs2mdsdev $fs2mdsvdev || exit 10
2081 add fs2ost $(mkfs_opts ost1 ${fs2ostdev}) --mgsnode=$MGSNID \
2082 --fsname=${FSNAME2} --index=8191 --reformat $fs2ostdev \
2083 $fs2ostvdev || exit 10
2085 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS && trap cleanup_24a EXIT INT
2086 start fs2ost $fs2ostdev $OST_MOUNT_OPTS
2087 do_facet $SINGLEMDS "$LCTL conf_param $FSNAME2.sys.timeout=200" || rc=1
2089 $MOUNT_CMD $MGSNID:/${FSNAME2} $MOUNT2 || rc=2
2092 cp /etc/hosts $MOUNT2/ || rc=3
2093 $LFS getstripe $MOUNT2/hosts
2098 cleanup_nocli || rc=6
2101 run_test 33a "Mount ost with a large index number"
2103 test_33b() { # was test_34
2106 do_facet client dd if=/dev/zero of=$MOUNT/24 bs=1024k count=1
2107 # Drop lock cancelation reply during umount
2108 #define OBD_FAIL_LDLM_CANCEL_NET 0x304
2109 do_facet client lctl set_param fail_loc=0x80000304
2110 #lctl set_param debug=-1
2111 umount_client $MOUNT
2114 run_test 33b "Drop cancel during umount"
2118 do_facet client "sh runmultiop_bg_pause $DIR/file O_c"
2119 manual_umount_client
2121 do_facet client killall -USR1 multiop
2122 if [ $rc -eq 0 ]; then
2123 error "umount not fail!"
2128 run_test 34a "umount with opened file should be fail"
2133 touch $DIR/$tfile || return 1
2134 stop_mds --force || return 2
2136 manual_umount_client --force
2138 if [ $rc -ne 0 ]; then
2139 error "mtab after failed umount - rc $rc"
2145 run_test 34b "force umount with failed mds should be normal"
2149 touch $DIR/$tfile || return 1
2150 stop_ost --force || return 2
2152 manual_umount_client --force
2154 if [ $rc -ne 0 ]; then
2155 error "mtab after failed umount - rc $rc"
2161 run_test 34c "force umount with failed ost should be normal"
2163 test_35a() { # bug 12459
2166 DBG_SAVE="`lctl get_param -n debug`"
2167 lctl set_param debug="ha"
2169 log "Set up a fake failnode for the MDS"
2171 local device=$(do_facet $SINGLEMDS "lctl get_param -n devices" |
2172 awk '($3 ~ "mdt" && $4 ~ "MDT") { print $4 }' | head -1)
2173 do_facet mgs "$LCTL conf_param \
2174 ${device}.failover.node=$(h2$NETTYPE $FAKENID)" || return 4
2176 log "Wait for RECONNECT_INTERVAL seconds (10s)"
2179 MSG="conf-sanity.sh test_35a `date +%F%kh%Mm%Ss`"
2182 log "Stopping the MDT: $device"
2183 stop_mdt 1 || return 5
2185 df $MOUNT > /dev/null 2>&1 &
2187 log "Restarting the MDT: $device"
2188 start_mdt 1 || return 6
2189 log "Wait for df ($DFPID) ... "
2192 lctl set_param debug="$DBG_SAVE"
2194 # retrieve from the log the first server that the client tried to
2195 # contact after the connection loss
2196 $LCTL dk $TMP/lustre-log-$TESTNAME.log
2197 NEXTCONN=`awk "/${MSG}/ {start = 1;}
2198 /import_select_connection.*$device-mdc.* using connection/ {
2200 if (\\\$NF ~ /$FAKENID/)
2206 }" $TMP/lustre-log-$TESTNAME.log`
2207 [ "$NEXTCONN" != "0" ] && log "The client didn't try to reconnect to the last active server (tried ${NEXTCONN} instead)" && return 7
2209 # remove nid settings
2210 writeconf_or_reformat
2212 run_test 35a "Reconnect to the last active server first"
2214 test_35b() { # bug 18674
2215 remote_mds || { skip "local MDS" && return 0; }
2219 $LCTL set_param debug="ha"
2221 MSG="conf-sanity.sh test_35b `date +%F%kh%Mm%Ss`"
2224 log "Set up a fake failnode for the MDS"
2226 local device=$(do_facet $SINGLEMDS "$LCTL get_param -n devices" |
2227 awk '($3 ~ "mdt" && $4 ~ "MDT") { print $4 }' | head -1)
2228 do_facet mgs "$LCTL conf_param \
2229 ${device}.failover.node=$(h2$NETTYPE $FAKENID)" || return 1
2231 local at_max_saved=0
2232 # adaptive timeouts may prevent seeing the issue
2233 if at_is_enabled; then
2234 at_max_saved=$(at_max_get mds)
2235 at_max_set 0 mds client
2238 mkdir -p $MOUNT/$tdir
2240 log "Injecting EBUSY on MDS"
2241 # Setting OBD_FAIL_MDS_RESEND=0x136
2242 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0x80000136" || return 2
2244 $LCTL set_param mdc.${FSNAME}*.stats=clear
2246 log "Creating a test file and stat it"
2247 touch $MOUNT/$tdir/$tfile
2248 stat $MOUNT/$tdir/$tfile
2250 log "Stop injecting EBUSY on MDS"
2251 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0" || return 3
2252 rm -f $MOUNT/$tdir/$tfile
2255 # restore adaptive timeout
2256 [ $at_max_saved -ne 0 ] && at_max_set $at_max_saved mds client
2258 $LCTL dk $TMP/lustre-log-$TESTNAME.log
2260 CONNCNT=`$LCTL get_param mdc.${FSNAME}*.stats | awk '/mds_connect/{print $2}'`
2262 # retrieve from the log if the client has ever tried to
2263 # contact the fake server after the loss of connection
2264 FAILCONN=`awk "BEGIN {ret = 0;}
2265 /import_select_connection.*${FSNAME}-MDT0000-mdc.* using connection/ {
2267 if (\\\$NF ~ /$FAKENID/) {
2272 END {print ret}" $TMP/lustre-log-$TESTNAME.log`
2274 [ "$FAILCONN" == "0" ] && \
2275 log "ERROR: The client reconnection has not been triggered" && \
2277 [ "$FAILCONN" == "2" ] && \
2278 log "ERROR: The client tried to reconnect to the failover server while the primary was busy" && \
2282 # When OBD_FAIL_MDS_RESEND is hit, we sleep for 2 * obd_timeout
2283 # Reconnects are supposed to be rate limited to one every 5s
2284 [ $CONNCNT -gt $((2 * $TIMEOUT / 5 + 1)) ] && \
2285 log "ERROR: Too many reconnects $CONNCNT" && \
2289 # remove nid settings
2290 writeconf_or_reformat
2292 run_test 35b "Continue reconnection retries, if the active server is busy"
2295 [ $OSTCOUNT -lt 2 ] && skip_env "skipping test for single OST" && return
2297 [ "$ost_HOST" = "`hostname`" -o "$ost1_HOST" = "`hostname`" ] || \
2298 { skip "remote OST" && return 0; }
2301 local FSNAME2=test1234
2302 local fs3ost_HOST=$ost_HOST
2303 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
2305 [ -n "$ost1_HOST" ] && fs2ost_HOST=$ost1_HOST && fs3ost_HOST=$ost1_HOST
2307 if [ -z "$fs2ost_DEV" -o -z "$fs2mds_DEV" -o -z "$fs3ost_DEV" ]; then
2308 is_blkdev $SINGLEMDS $MDSDEV && \
2309 skip_env "mixed loopback and real device not working" && return
2312 local fs2mdsdev=$(mdsdevname 1_2)
2313 local fs2ostdev=$(ostdevname 1_2)
2314 local fs3ostdev=$(ostdevname 2_2)
2315 local fs2mdsvdev=$(mdsvdevname 1_2)
2316 local fs2ostvdev=$(ostvdevname 1_2)
2317 local fs3ostvdev=$(ostvdevname 2_2)
2319 add fs2mds $(mkfs_opts mds1 ${fs2mdsdev}) --mgs --fsname=${FSNAME2} \
2320 --reformat $fs2mdsdev $fs2mdsvdev || exit 10
2321 # XXX after we support non 4K disk blocksize in ldiskfs, specify a
2322 # different one than the default value here.
2323 add fs2ost $(mkfs_opts ost1 ${fs2ostdev}) --mgsnode=$MGSNID \
2324 --fsname=${FSNAME2} --reformat $fs2ostdev $fs2ostvdev || exit 10
2325 add fs3ost $(mkfs_opts ost1 ${fs3ostdev}) --mgsnode=$MGSNID \
2326 --fsname=${FSNAME2} --reformat $fs3ostdev $fs3ostvdev || exit 10
2328 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS
2329 start fs2ost $fs2ostdev $OST_MOUNT_OPTS
2330 start fs3ost $fs3ostdev $OST_MOUNT_OPTS
2332 $MOUNT_CMD $MGSNID:/${FSNAME2} $MOUNT2 || return 1
2334 sleep 5 # until 11778 fixed
2336 dd if=/dev/zero of=$MOUNT2/$tfile bs=1M count=7 || return 2
2338 BKTOTAL=`lctl get_param -n obdfilter.*.kbytestotal | awk 'BEGIN{total=0}; {total+=$1}; END{print total}'`
2339 BKFREE=`lctl get_param -n obdfilter.*.kbytesfree | awk 'BEGIN{free=0}; {free+=$1}; END{print free}'`
2340 BKAVAIL=`lctl get_param -n obdfilter.*.kbytesavail | awk 'BEGIN{avail=0}; {avail+=$1}; END{print avail}'`
2341 STRING=`df -P $MOUNT2 | tail -n 1 | awk '{print $2","$3","$4}'`
2342 DFTOTAL=`echo $STRING | cut -d, -f1`
2343 DFUSED=`echo $STRING | cut -d, -f2`
2344 DFAVAIL=`echo $STRING | cut -d, -f3`
2345 DFFREE=$(($DFTOTAL - $DFUSED))
2347 ALLOWANCE=$((64 * $OSTCOUNT))
2349 if [ $DFTOTAL -lt $(($BKTOTAL - $ALLOWANCE)) ] ||
2350 [ $DFTOTAL -gt $(($BKTOTAL + $ALLOWANCE)) ] ; then
2351 echo "**** FAIL: df total($DFTOTAL) mismatch OST total($BKTOTAL)"
2354 if [ $DFFREE -lt $(($BKFREE - $ALLOWANCE)) ] ||
2355 [ $DFFREE -gt $(($BKFREE + $ALLOWANCE)) ] ; then
2356 echo "**** FAIL: df free($DFFREE) mismatch OST free($BKFREE)"
2359 if [ $DFAVAIL -lt $(($BKAVAIL - $ALLOWANCE)) ] ||
2360 [ $DFAVAIL -gt $(($BKAVAIL + $ALLOWANCE)) ] ; then
2361 echo "**** FAIL: df avail($DFAVAIL) mismatch OST avail($BKAVAIL)"
2366 stop fs3ost -f || return 200
2367 stop fs2ost -f || return 201
2368 stop fs2mds -f || return 202
2369 unload_modules_conf || return 203
2372 run_test 36 "df report consistency on OSTs with different block size"
2375 local mntpt=$(facet_mntpt $SINGLEMDS)
2376 local mdsdev=$(mdsdevname ${SINGLEMDS//mds/})
2377 local mdsdev_sym="$TMP/sym_mdt.img"
2378 local opts=$MDS_MOUNT_OPTS
2381 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
2382 skip "Currently only applicable to ldiskfs-based MDTs"
2386 echo "MDS : $mdsdev"
2387 echo "SYMLINK : $mdsdev_sym"
2388 do_facet $SINGLEMDS rm -f $mdsdev_sym
2390 do_facet $SINGLEMDS ln -s $mdsdev $mdsdev_sym
2392 echo "mount symlink device - $mdsdev_sym"
2394 if ! do_facet $SINGLEMDS test -b $mdsdev; then
2395 opts=$(csa_add "$opts" -o loop)
2397 mount_op=$(do_facet $SINGLEMDS mount -v -t lustre $opts \
2398 $mdsdev_sym $mntpt 2>&1)
2401 echo mount_op=$mount_op
2403 do_facet $SINGLEMDS "umount -d $mntpt && rm -f $mdsdev_sym"
2405 if $(echo $mount_op | grep -q "unable to set tunable"); then
2406 error "set tunables failed for symlink device"
2409 [ $rc -eq 0 ] || error "mount symlink $mdsdev_sym failed! rc=$rc"
2413 run_test 37 "verify set tunables works for symlink device"
2415 test_38() { # bug 14222
2416 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
2417 skip "Only applicable to ldiskfs-based MDTs"
2425 FILES=`find $SRC -type f -mtime +1 | head -n $COUNT`
2426 log "copying $(echo $FILES | wc -w) files to $DIR/$tdir"
2428 tar cf - $FILES | tar xf - -C $DIR/$tdir || \
2429 error "copying $SRC to $DIR/$tdir"
2431 umount_client $MOUNT
2433 log "rename lov_objid file on MDS"
2434 rm -f $TMP/lov_objid.orig
2436 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
2437 do_facet $SINGLEMDS "$DEBUGFS -c -R \\\"dump lov_objid $TMP/lov_objid.orig\\\" $MDSDEV"
2438 do_facet $SINGLEMDS "$DEBUGFS -w -R \\\"rm lov_objid\\\" $MDSDEV"
2440 do_facet $SINGLEMDS "od -Ax -td8 $TMP/lov_objid.orig"
2441 # check create in mds_lov_connect
2445 [ $V ] && log "verifying $DIR/$tdir/$f"
2446 diff -q $f $DIR/$tdir/$f || ERROR=y
2448 do_facet $SINGLEMDS "$DEBUGFS -c -R \\\"dump lov_objid $TMP/lov_objid.new\\\" $MDSDEV"
2449 do_facet $SINGLEMDS "od -Ax -td8 $TMP/lov_objid.new"
2450 [ "$ERROR" = "y" ] && error "old and new files are different after connect" || true
2452 # check it's updates in sync
2453 umount_client $MOUNT
2456 do_facet $SINGLEMDS dd if=/dev/zero of=$TMP/lov_objid.clear bs=4096 count=1
2457 do_facet $SINGLEMDS "$DEBUGFS -w -R \\\"rm lov_objid\\\" $MDSDEV"
2458 do_facet $SINGLEMDS "$DEBUGFS -w -R \\\"write $TMP/lov_objid.clear lov_objid\\\" $MDSDEV "
2463 [ $V ] && log "verifying $DIR/$tdir/$f"
2464 diff -q $f $DIR/$tdir/$f || ERROR=y
2466 do_facet $SINGLEMDS "$DEBUGFS -c -R \\\"dump lov_objid $TMP/lov_objid.new1\\\" $MDSDEV"
2467 do_facet $SINGLEMDS "od -Ax -td8 $TMP/lov_objid.new1"
2468 umount_client $MOUNT
2470 [ "$ERROR" = "y" ] && error "old and new files are different after sync" || true
2472 log "files compared the same"
2475 run_test 38 "MDS recreates missing lov_objid file from OST data"
2481 perl $SRCDIR/leak_finder.pl $TMP/debug 2>&1 | egrep '*** Leak:' &&
2482 error "memory leak detected" || true
2484 run_test 39 "leak_finder recognizes both LUSTRE and LNET malloc messages"
2486 test_40() { # bug 15759
2488 #define OBD_FAIL_TGT_TOOMANY_THREADS 0x706
2489 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0x80000706"
2493 run_test 40 "race during service thread startup"
2495 test_41a() { #bug 14134
2496 if [ $(facet_fstype $SINGLEMDS) == ldiskfs ] &&
2497 ! do_facet $SINGLEMDS test -b $(mdsdevname 1); then
2498 skip "Loop devices does not work with nosvc option"
2503 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
2505 start $SINGLEMDS $MDSDEV $MDS_MOUNT_OPTS -o nosvc -n
2506 start ost1 `ostdevname 1` $OST_MOUNT_OPTS
2507 start $SINGLEMDS $MDSDEV $MDS_MOUNT_OPTS -o nomgs,force
2509 mount_client $MOUNT || return 1
2512 echo "blah blah" > $MOUNT/$tfile
2515 umount_client $MOUNT
2516 stop ost1 -f || return 201
2517 stop_mds -f || return 202
2518 stop_mds -f || return 203
2519 unload_modules_conf || return 204
2522 run_test 41a "mount mds with --nosvc and --nomgs"
2525 if [ $(facet_fstype $SINGLEMDS) == ldiskfs ] &&
2526 ! do_facet $SINGLEMDS test -b $(mdsdevname 1); then
2527 skip "Loop devices does not work with nosvc option"
2531 ! combined_mgs_mds && skip "needs combined mgs device" && return 0
2535 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
2537 start $SINGLEMDS $MDSDEV $MDS_MOUNT_OPTS -o nosvc -n
2539 start $SINGLEMDS $MDSDEV $MDS_MOUNT_OPTS -o nomgs,force
2541 mount_client $MOUNT || return 1
2544 echo "blah blah" > $MOUNT/$tfile
2545 cat $MOUNT/$tfile || return 200
2547 umount_client $MOUNT
2548 stop_ost || return 201
2549 stop_mds -f || return 202
2550 stop_mds -f || return 203
2553 run_test 41b "mount mds with --nosvc and --nomgs on first mount"
2556 local server_version=$(lustre_version_code $SINGLEMDS)
2558 [[ $server_version -ge $(version_code 2.6.52) ]] ||
2559 [[ $server_version -ge $(version_code 2.5.26) &&
2560 $server_version -lt $(version_code 2.5.50) ]] ||
2561 [[ $server_version -ge $(version_code 2.5.3) &&
2562 $server_version -lt $(version_code 2.5.11) ]] ||
2563 { skip "Need MDS version 2.5.4+ or 2.5.26+ or 2.6.52+"; return; }
2566 # MDT concurent start
2567 #define OBD_FAIL_TGT_DELAY_CONNECT 0x703
2568 do_facet $SINGLEMDS "lctl set_param fail_loc=0x703"
2569 start mds1 $(mdsdevname 1) $MDS_MOUNT_OPTS &
2572 do_facet $SINGLEMDS "lctl set_param fail_loc=0x0"
2573 start mds1 $(mdsdevname 1) $MDS_MOUNT_OPTS &
2579 if [ $rc == 0 ] && [ $rc2 == 114 ]; then
2580 echo "1st MDT start succeed"
2581 echo "2nd MDT start failed with EALREADY"
2582 elif [ $rc2 == 0 ] && [ $rc == 114 ]; then
2583 echo "1st MDT start failed with EALREADY"
2584 echo "2nd MDT start succeed"
2587 error "unexpected concurent MDT mounts result, rc=$rc rc2=$rc2"
2590 # OST concurent start
2591 #define OBD_FAIL_TGT_DELAY_CONNECT 0x703
2592 do_facet ost1 "lctl set_param fail_loc=0x703"
2593 start ost1 $(ostdevname 1) $OST_MOUNT_OPTS &
2596 do_facet ost1 "lctl set_param fail_loc=0x0"
2597 start ost1 $(ostdevname 1) $OST_MOUNT_OPTS &
2603 if [ $rc == 0 ] && [ $rc2 == 114 ]; then
2604 echo "1st OST start succeed"
2605 echo "2nd OST start failed with EALREADY"
2606 elif [ $rc2 == 0 ] && [ $rc == 114 ]; then
2607 echo "1st OST start failed with EALREADY"
2608 echo "2nd OST start succeed"
2612 error "unexpected concurent OST mounts result, rc=$rc rc2=$rc2"
2618 # verify everything ok
2623 error "MDT(s) start failed"
2631 error "OST(s) start failed"
2639 error "client start failed"
2646 error "client mount failed"
2650 run_test 41c "concurent mounts of MDT/OST should all fail but one"
2652 test_42() { #bug 14693
2654 check_mount || error "client was not mounted"
2656 do_facet mgs $LCTL conf_param $FSNAME.llite.some_wrong_param=10
2657 umount_client $MOUNT ||
2658 error "unmounting client failed with invalid llite param"
2659 mount_client $MOUNT ||
2660 error "mounting client failed with invalid llite param"
2662 do_facet mgs $LCTL conf_param $FSNAME.sys.some_wrong_param=20
2663 cleanup || error "stopping $FSNAME failed with invalid sys param"
2666 check_mount || "client was not mounted with invalid sys param"
2667 cleanup || error "stopping $FSNAME failed with invalid sys param"
2670 run_test 42 "allow client/server mount/unmount with invalid config param"
2673 local server_version=$(lustre_version_code $SINGLEMDS)
2675 [[ $server_version -ge $(version_code 2.5.59) ]] ||
2676 [[ $server_version -ge $(version_code 2.5.3) &&
2677 $server_version -lt $(version_code 2.5.11) ]] ||
2678 { skip "Need MDS version 2.5.4+ or 2.5.59+"; return; }
2680 [ $UID -ne 0 -o $RUNAS_ID -eq 0 ] && skip_env "run as root"
2683 USER1=$(cat /etc/passwd | grep :$ID1:$ID1: | cut -d: -f1)
2684 [ -z "$USER1" ] && skip_env "missing user with uid=$ID1 gid=$ID1" &&
2688 chmod ugo+x $DIR || error "chmod 0 failed"
2689 set_conf_param_and_check mds \
2690 "lctl get_param -n mdt.$FSNAME-MDT0000.root_squash" \
2691 "$FSNAME.mdt.root_squash" \
2693 wait_update $HOSTNAME \
2694 "lctl get_param -n llite.${FSNAME}*.root_squash" \
2696 error "check llite root_squash failed!"
2697 set_conf_param_and_check mds \
2698 "lctl get_param -n mdt.$FSNAME-MDT0000.nosquash_nids" \
2699 "$FSNAME.mdt.nosquash_nids" \
2701 wait_update $HOSTNAME \
2702 "lctl get_param -n llite.${FSNAME}*.nosquash_nids" \
2704 error "check llite nosquash_nids failed!"
2707 # create set of test files
2709 echo "111" > $DIR/$tfile-userfile || error "write 1 failed"
2710 chmod go-rw $DIR/$tfile-userfile || error "chmod 1 failed"
2711 chown $RUNAS_ID.$RUNAS_ID $DIR/$tfile-userfile || error "chown failed"
2713 echo "222" > $DIR/$tfile-rootfile || error "write 2 failed"
2714 chmod go-rw $DIR/$tfile-rootfile || error "chmod 2 faield"
2716 mkdir $DIR/$tdir-rootdir -p || error "mkdir failed"
2717 chmod go-rwx $DIR/$tdir-rootdir || error "chmod 3 failed"
2718 touch $DIR/$tdir-rootdir/tfile-1 || error "touch failed"
2720 echo "777" > $DIR/$tfile-user1file || error "write 7 failed"
2721 chmod go-rw $DIR/$tfile-user1file || error "chmod 7 failed"
2722 chown $ID1.$ID1 $DIR/$tfile-user1file || error "chown failed"
2725 # check root_squash:
2726 # set root squash UID:GID to RUNAS_ID
2727 # root should be able to access only files owned by RUNAS_ID
2729 set_conf_param_and_check mds \
2730 "lctl get_param -n mdt.$FSNAME-MDT0000.root_squash" \
2731 "$FSNAME.mdt.root_squash" \
2732 "$RUNAS_ID:$RUNAS_ID"
2733 wait_update $HOSTNAME \
2734 "lctl get_param -n llite.${FSNAME}*.root_squash" \
2735 "$RUNAS_ID:$RUNAS_ID" ||
2736 error "check llite root_squash failed!"
2738 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-userfile)
2739 dd if=$DIR/$tfile-userfile 1>/dev/null 2>/dev/null || \
2740 error "$ST: root read permission is denied"
2741 echo "$ST: root read permission is granted - ok"
2744 dd conv=notrunc of=$DIR/$tfile-userfile 1>/dev/null 2>/dev/null || \
2745 error "$ST: root write permission is denied"
2746 echo "$ST: root write permission is granted - ok"
2748 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-rootfile)
2749 dd if=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null && \
2750 error "$ST: root read permission is granted"
2751 echo "$ST: root read permission is denied - ok"
2754 dd conv=notrunc of=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null && \
2755 error "$ST: root write permission is granted"
2756 echo "$ST: root write permission is denied - ok"
2758 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tdir-rootdir)
2759 rm $DIR/$tdir-rootdir/tfile-1 1>/dev/null 2>/dev/null && \
2760 error "$ST: root unlink permission is granted"
2761 echo "$ST: root unlink permission is denied - ok"
2763 touch $DIR/tdir-rootdir/tfile-2 1>/dev/null 2>/dev/null && \
2764 error "$ST: root create permission is granted"
2765 echo "$ST: root create permission is denied - ok"
2769 # check root_squash is enforced independently
2770 # of client cache content
2772 # access file by USER1, keep access open
2773 # root should be denied access to user file
2775 runas -u $ID1 tail -f $DIR/$tfile-user1file 1>/dev/null 2>&1 &
2779 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-user1file)
2780 dd if=$DIR/$tfile-user1file 1>/dev/null 2>&1 &&
2781 { kill $pid; error "$ST: root read permission is granted"; }
2782 echo "$ST: root read permission is denied - ok"
2785 dd conv=notrunc of=$DIR/$tfile-user1file 1>/dev/null 2>&1 &&
2786 { kill $pid; error "$ST: root write permission is granted"; }
2787 echo "$ST: root write permission is denied - ok"
2793 # check nosquash_nids:
2794 # put client's NID into nosquash_nids list,
2795 # root should be able to access root file after that
2797 local NIDLIST=$(lctl list_nids all | tr '\n' ' ')
2798 NIDLIST="2@elan $NIDLIST 192.168.0.[2,10]@tcp"
2799 NIDLIST=$(echo $NIDLIST | tr -s ' ' ' ')
2800 set_conf_param_and_check mds \
2801 "lctl get_param -n mdt.$FSNAME-MDT0000.nosquash_nids" \
2802 "$FSNAME-MDTall.mdt.nosquash_nids" \
2804 wait_update $HOSTNAME \
2805 "lctl get_param -n llite.${FSNAME}*.nosquash_nids" \
2807 error "check llite nosquash_nids failed!"
2809 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-rootfile)
2810 dd if=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null || \
2811 error "$ST: root read permission is denied"
2812 echo "$ST: root read permission is granted - ok"
2815 dd conv=notrunc of=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null || \
2816 error "$ST: root write permission is denied"
2817 echo "$ST: root write permission is granted - ok"
2819 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tdir-rootdir)
2820 rm $DIR/$tdir-rootdir/tfile-1 || \
2821 error "$ST: root unlink permission is denied"
2822 echo "$ST: root unlink permission is granted - ok"
2823 touch $DIR/$tdir-rootdir/tfile-2 || \
2824 error "$ST: root create permission is denied"
2825 echo "$ST: root create permission is granted - ok"
2829 run_test 43 "check root_squash and nosquash_nids"
2831 umount_client $MOUNT
2836 check_mount || return 2
2837 UUID=$($LCTL get_param llite.${FSNAME}*.uuid | cut -d= -f2)
2839 UUIDS=$(do_facet $SINGLEMDS "$LCTL get_param mdt.${FSNAME}*.exports.*.uuid")
2840 for VAL in $UUIDS; do
2841 NID=$(echo $VAL | cut -d= -f1)
2842 CLUUID=$(echo $VAL | cut -d= -f2)
2843 [ "$UUID" = "$CLUUID" ] && STATS_FOUND=yes && break
2845 [ "$STATS_FOUND" = "no" ] && error "stats not found for client"
2849 run_test 44 "mounted client proc entry exists"
2853 check_mount || return 2
2858 #define OBD_FAIL_PTLRPC_LONG_UNLINK 0x50f
2859 do_facet client "lctl set_param fail_loc=0x50f"
2862 manual_umount_client --force || return 3
2863 do_facet client "lctl set_param fail_loc=0x0"
2865 mount_client $MOUNT || return 4
2869 run_test 45 "long unlink handling in ptlrpcd"
2876 umount_client $MOUNT2 || rc=$?
2877 umount_client $MOUNT || rc=$?
2878 while [ $count -gt 0 ]; do
2879 stop ost${count} -f || rc=$?
2883 cleanup_nocli || rc=$?
2884 #writeconf to remove all ost2 traces for subsequent tests
2885 writeconf_or_reformat
2890 echo "Testing with $OSTCOUNT OSTs"
2892 start_mds || return 1
2893 #first client should see only one ost
2894 start_ost || return 2
2895 wait_osc_import_state mds ost FULL
2897 mount_client $MOUNT || return 3
2898 trap "cleanup_46a $OSTCOUNT" EXIT ERR
2901 for (( i=2; i<=$OSTCOUNT; i++ )); do
2902 start ost$i `ostdevname $i` $OST_MOUNT_OPTS || return $((i+2))
2905 # wait until osts in sync
2906 for (( i=2; i<=$OSTCOUNT; i++ )); do
2907 wait_osc_import_state mds ost$i FULL
2908 wait_osc_import_state client ost$i FULL
2911 #second client see all ost's
2913 mount_client $MOUNT2 || return 8
2914 $LFS setstripe -c -1 $MOUNT2 || return 9
2915 $LFS getstripe $MOUNT2 || return 10
2917 echo "ok" > $MOUNT2/widestripe
2918 $LFS getstripe $MOUNT2/widestripe || return 11
2919 # fill acl buffer for avoid expand lsm to them
2920 awk -F : '{if (FNR < 25) { print "u:"$1":rwx" }}' /etc/passwd | while read acl; do
2921 setfacl -m $acl $MOUNT2/widestripe
2925 stat $MOUNT/widestripe || return 12
2927 cleanup_46a $OSTCOUNT || { echo "cleanup_46a failed!" && return 13; }
2930 run_test 46a "handle ost additional - wide striped file"
2935 check_mount || return 2
2936 $LCTL set_param ldlm.namespaces.$FSNAME-*-*-*.lru_size=100
2940 for ns in $($LCTL get_param ldlm.namespaces.$FSNAME-*-*-*.lru_size); do
2941 if echo $ns | grep "MDT[[:digit:]]*"; then
2944 lrs=$(echo $ns | sed 's/.*lru_size=//')
2945 lru_size[count]=$lrs
2950 facet_failover $SINGLEMDS
2951 client_up || return 3
2954 for ns in $($LCTL get_param ldlm.namespaces.$FSNAME-*-*-*.lru_size); do
2955 if echo $ns | grep "MDT[[:digit:]]*"; then
2958 lrs=$(echo $ns | sed 's/.*lru_size=//')
2959 if ! test "$lrs" -eq "${lru_size[count]}"; then
2960 n=$(echo $ns | sed -e 's/ldlm.namespaces.//' -e 's/.lru_size=.*//')
2961 error "$n has lost lru_size: $lrs vs. ${lru_size[count]}"
2969 run_test 47 "server restart does not make client loss lru_resize settings"
2974 # reformat after this test is needed - if test will failed
2975 # we will have unkillable file at FS
2979 test_48() { # bug 17636
2982 check_mount || return 2
2984 $LFS setstripe -c -1 $MOUNT || return 9
2985 $LFS getstripe $MOUNT || return 10
2987 echo "ok" > $MOUNT/widestripe
2988 $LFS getstripe $MOUNT/widestripe || return 11
2990 trap cleanup_48 EXIT ERR
2992 # fill acl buffer for avoid expand lsm to them
2993 getent passwd | awk -F : '{ print "u:"$1":rwx" }' | while read acl; do
2994 setfacl -m $acl $MOUNT/widestripe
2997 stat $MOUNT/widestripe || return 12
3002 run_test 48 "too many acls on file"
3004 # check PARAM_SYS_LDLM_TIMEOUT option of MKFS.LUSTRE
3005 test_49() { # bug 17710
3006 local timeout_orig=$TIMEOUT
3007 local ldlm_timeout_orig=$LDLM_TIMEOUT
3008 local LOCAL_TIMEOUT=20
3010 LDLM_TIMEOUT=$LOCAL_TIMEOUT
3011 TIMEOUT=$LOCAL_TIMEOUT
3015 check_mount || return 1
3017 echo "check ldlm_timout..."
3018 LDLM_MDS="`do_facet $SINGLEMDS lctl get_param -n ldlm_timeout`"
3019 LDLM_OST1="`do_facet ost1 lctl get_param -n ldlm_timeout`"
3020 LDLM_CLIENT="`do_facet client lctl get_param -n ldlm_timeout`"
3022 if [ $LDLM_MDS -ne $LDLM_OST1 ] || [ $LDLM_MDS -ne $LDLM_CLIENT ]; then
3023 error "Different LDLM_TIMEOUT:$LDLM_MDS $LDLM_OST1 $LDLM_CLIENT"
3026 if [ $LDLM_MDS -ne $((LOCAL_TIMEOUT / 3)) ]; then
3027 error "LDLM_TIMEOUT($LDLM_MDS) is not correct"
3030 umount_client $MOUNT
3031 stop_ost || return 2
3032 stop_mds || return 3
3034 LDLM_TIMEOUT=$((LOCAL_TIMEOUT - 1))
3038 check_mount || return 7
3040 LDLM_MDS="`do_facet $SINGLEMDS lctl get_param -n ldlm_timeout`"
3041 LDLM_OST1="`do_facet ost1 lctl get_param -n ldlm_timeout`"
3042 LDLM_CLIENT="`do_facet client lctl get_param -n ldlm_timeout`"
3044 if [ $LDLM_MDS -ne $LDLM_OST1 ] || [ $LDLM_MDS -ne $LDLM_CLIENT ]; then
3045 error "Different LDLM_TIMEOUT:$LDLM_MDS $LDLM_OST1 $LDLM_CLIENT"
3048 if [ $LDLM_MDS -ne $((LOCAL_TIMEOUT - 1)) ]; then
3049 error "LDLM_TIMEOUT($LDLM_MDS) is not correct"
3052 cleanup || return $?
3054 LDLM_TIMEOUT=$ldlm_timeout_orig
3055 TIMEOUT=$timeout_orig
3057 run_test 49 "check PARAM_SYS_LDLM_TIMEOUT option of MKFS.LUSTRE"
3060 # Test both statfs and lfs df and fail if either one fails
3061 multiop_bg_pause $1 f_
3064 killall -USR1 multiop
3065 [ $RC1 -ne 0 ] && log "lazystatfs multiop failed"
3066 wait $PID || { RC1=$?; log "multiop return error "; }
3073 if [ $RC2 -eq 0 ]; then
3075 log "lazystatfs df failed"
3079 [[ $RC1 -ne 0 || $RC2 -eq 0 ]] && RC=1
3085 lctl set_param llite.$FSNAME-*.lazystatfs=1
3088 lazystatfs $MOUNT || error "lazystatfs failed but no down servers"
3090 cleanup || return $?
3092 run_test 50a "lazystatfs all servers available =========================="
3096 lctl set_param llite.$FSNAME-*.lazystatfs=1
3099 # Wait for client to detect down OST
3100 stop_ost || error "Unable to stop OST1"
3101 wait_osc_import_state mds ost DISCONN
3103 lazystatfs $MOUNT || error "lazystatfs should don't have returned EIO"
3105 umount_client $MOUNT || error "Unable to unmount client"
3106 stop_mds || error "Unable to stop MDS"
3108 run_test 50b "lazystatfs all servers down =========================="
3111 start_mds || error "Unable to start MDS"
3112 start_ost || error "Unable to start OST1"
3113 start_ost2 || error "Unable to start OST2"
3114 mount_client $MOUNT || error "Unable to mount client"
3115 lctl set_param llite.$FSNAME-*.lazystatfs=1
3118 # Wait for client to detect down OST
3119 stop_ost || error "Unable to stop OST1"
3120 wait_osc_import_state mds ost DISCONN
3121 lazystatfs $MOUNT || error "lazystatfs failed with one down server"
3123 umount_client $MOUNT || error "Unable to unmount client"
3124 stop_ost2 || error "Unable to stop OST2"
3125 stop_mds || error "Unable to stop MDS"
3126 #writeconf to remove all ost2 traces for subsequent tests
3127 writeconf_or_reformat
3129 run_test 50c "lazystatfs one server down =========================="
3132 start_mds || error "Unable to start MDS"
3133 start_ost || error "Unable to start OST1"
3134 start_ost2 || error "Unable to start OST2"
3135 mount_client $MOUNT || error "Unable to mount client"
3136 lctl set_param llite.$FSNAME-*.lazystatfs=1
3139 # Issue the statfs during the window where the client still
3140 # belives the OST to be available but it is in fact down.
3141 # No failure just a statfs which hangs for a timeout interval.
3142 stop_ost || error "Unable to stop OST1"
3143 lazystatfs $MOUNT || error "lazystatfs failed with one down server"
3145 umount_client $MOUNT || error "Unable to unmount client"
3146 stop_ost2 || error "Unable to stop OST2"
3147 stop_mds || error "Unable to stop MDS"
3148 #writeconf to remove all ost2 traces for subsequent tests
3149 writeconf_or_reformat
3151 run_test 50d "lazystatfs client/server conn race =========================="
3158 start_mds || return 1
3159 #first client should see only one ost
3160 start_ost || return 2
3161 wait_osc_import_state mds ost FULL
3163 # Wait for client to detect down OST
3164 stop_ost || error "Unable to stop OST1"
3165 wait_osc_import_state mds ost DISCONN
3167 mount_client $MOUNT || error "Unable to mount client"
3168 lctl set_param llite.$FSNAME-*.lazystatfs=0
3170 multiop_bg_pause $MOUNT _f
3174 if [ $RC1 -ne 0 ]; then
3175 log "multiop failed $RC1"
3178 sleep $(( $TIMEOUT+1 ))
3180 [ $? -ne 0 ] && error "process isn't sleep"
3181 start_ost || error "Unable to start OST1"
3182 wait $pid || error "statfs failed"
3185 umount_client $MOUNT || error "Unable to unmount client"
3186 stop_ost || error "Unable to stop OST1"
3187 stop_mds || error "Unable to stop MDS"
3189 run_test 50e "normal statfs all servers down =========================="
3194 CONN_PROC="osc.$FSNAME-OST0001-osc-[M]*.ost_server_uuid"
3196 start_mds || error "Unable to start mds"
3197 #first client should see only one ost
3198 start_ost || error "Unable to start OST1"
3199 wait_osc_import_state mds ost FULL
3201 start_ost2 || error "Unable to start OST2"
3202 wait_osc_import_state mds ost2 FULL
3204 # Wait for client to detect down OST
3205 stop_ost2 || error "Unable to stop OST2"
3207 wait_osc_import_state mds ost2 DISCONN
3208 mount_client $MOUNT || error "Unable to mount client"
3209 lctl set_param llite.$FSNAME-*.lazystatfs=0
3211 multiop_bg_pause $MOUNT _f
3215 if [ $RC1 -ne 0 ]; then
3216 log "lazystatfs multiop failed $RC1"
3219 sleep $(( $TIMEOUT+1 ))
3221 [ $? -ne 0 ] && error "process isn't sleep"
3222 start_ost2 || error "Unable to start OST2"
3223 wait $pid || error "statfs failed"
3224 stop_ost2 || error "Unable to stop OST2"
3227 umount_client $MOUNT || error "Unable to unmount client"
3228 stop_ost || error "Unable to stop OST1"
3229 stop_mds || error "Unable to stop MDS"
3230 #writeconf to remove all ost2 traces for subsequent tests
3231 writeconf_or_reformat
3233 run_test 50f "normal statfs one server in down =========================="
3236 [ "$OSTCOUNT" -lt "2" ] && skip_env "$OSTCOUNT < 2, skipping" && return
3238 start_ost2 || error "Unable to start OST2"
3239 wait_osc_import_state mds ost2 FULL
3240 wait_osc_import_state client ost2 FULL
3242 local PARAM="${FSNAME}-OST0001.osc.active"
3244 $LFS setstripe -c -1 $DIR/$tfile || error "Unable to lfs setstripe"
3245 do_facet mgs $LCTL conf_param $PARAM=0 || error "Unable to deactivate OST"
3247 umount_client $MOUNT || error "Unable to unmount client"
3248 mount_client $MOUNT || error "Unable to mount client"
3249 # This df should not cause a panic
3252 do_facet mgs $LCTL conf_param $PARAM=1 || error "Unable to activate OST"
3254 umount_client $MOUNT || error "Unable to unmount client"
3255 stop_ost2 || error "Unable to stop OST2"
3256 stop_ost || error "Unable to stop OST1"
3257 stop_mds || error "Unable to stop MDS"
3258 #writeconf to remove all ost2 traces for subsequent tests
3259 writeconf_or_reformat
3261 run_test 50g "deactivated OST should not cause panic====================="
3265 # prepare MDT/OST, make OSC inactive for OST1
3266 [ "$OSTCOUNT" -lt "2" ] && skip_env "$OSTCOUNT < 2, skipping" && return
3268 [ $(facet_fstype ost1) == zfs ] && import_zpool ost1
3269 do_facet ost1 "$TUNEFS --param osc.active=0 `ostdevname 1`" ||
3270 error "tunefs OST1 failed"
3271 start_mds || error "Unable to start MDT"
3272 start_ost || error "Unable to start OST1"
3273 start_ost2 || error "Unable to start OST2"
3274 mount_client $MOUNT || error "client start failed"
3278 # activatate OSC for OST1
3279 local TEST="$LCTL get_param -n osc.${FSNAME}-OST0000-osc-[!M]*.active"
3280 set_conf_param_and_check client \
3281 "$TEST" "${FSNAME}-OST0000.osc.active" 1 ||
3282 error "Unable to activate OST1"
3284 mkdir -p $DIR/$tdir/2
3285 $LFS setstripe -c -1 -i 0 $DIR/$tdir/2
3286 sleep 1 && echo "create a file after OST1 is activated"
3288 createmany -o $DIR/$tdir/2/$tfile-%d 1
3290 # check OSC import is working
3291 stat $DIR/$tdir/2/* >/dev/null 2>&1 ||
3292 error "some OSC imports are still not connected"
3295 umount_client $MOUNT || error "Unable to umount client"
3296 stop_ost2 || error "Unable to stop OST2"
3299 run_test 50h "LU-642: activate deactivated OST ==="
3302 local LOCAL_TIMEOUT=20
3306 check_mount || return 1
3309 $LFS setstripe -c -1 $MOUNT/d1
3310 #define OBD_FAIL_MDS_REINT_DELAY 0x142
3311 do_facet $SINGLEMDS "lctl set_param fail_loc=0x142"
3312 touch $MOUNT/d1/f1 &
3315 start_ost2 || return 2
3317 stop_ost2 || return 3
3319 #writeconf to remove all ost2 traces for subsequent tests
3320 writeconf_or_reformat
3322 run_test 51 "Verify that mdt_reint handles RMF_MDT_MD correctly when an OST is added"
3331 do_node $node mkdir -p $dest
3332 [ $? -eq 0 ] || { error "Unable to create directory"; return 1; }
3334 do_node $node 'tar cf - '$@' | tar xf - -C '$dest';
3335 [ \"\${PIPESTATUS[*]}\" = \"0 0\" ] || exit 1'
3336 [ $? -eq 0 ] || { error "Unable to tar files"; return 2; }
3338 do_node $node 'getfattr -d -m "[a-z]*\\." '$@' > '$xattrs
3339 [ $? -eq 0 ] || { error "Unable to read xattrs"; return 3; }
3349 local backup2=${TMP}/backup2
3351 do_node $node mkdir -p $backup2
3352 [ $? -eq 0 ] || { error "Unable to create directory"; return 1; }
3354 do_node $node 'tar cf - '$@' | tar xf - -C '$backup2';
3355 [ \"\${PIPESTATUS[*]}\" = \"0 0\" ] || exit 1'
3356 [ $? -eq 0 ] || { error "Unable to tar files to diff"; return 2; }
3358 do_node $node "diff -rq $backup $backup2"
3359 [ $? -eq 0 ] || { error "contents differ"; return 3; }
3361 local xattrs2=${TMP}/xattrs2
3362 do_node $node 'getfattr -d -m "[a-z]*\\." '$@' > '$xattrs2
3363 [ $? -eq 0 ] || { error "Unable to read xattrs to diff"; return 4; }
3365 do_node $node "diff $xattrs $xattrs2"
3366 [ $? -eq 0 ] || { error "xattrs differ"; return 5; }
3368 do_node $node "rm -rf $backup2 $xattrs2"
3369 [ $? -eq 0 ] || { error "Unable to delete temporary files"; return 6; }
3373 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
3374 skip "Only applicable to ldiskfs-based MDTs"
3379 [ $? -eq 0 ] || { error "Unable to start MDS"; return 1; }
3381 [ $? -eq 0 ] || { error "Unable to start OST1"; return 2; }
3383 [ $? -eq 0 ] || { error "Unable to mount client"; return 3; }
3386 local ost1mnt=$(facet_mntpt ost1)
3387 local ost1node=$(facet_active_host ost1)
3388 local ost1tmp=$TMP/conf52
3392 [ $? -eq 0 ] || { error "Unable to create tdir"; return 4; }
3393 touch $TMP/modified_first
3394 [ $? -eq 0 ] || { error "Unable to create temporary file"; return 5; }
3395 local mtime=$(stat -c %Y $TMP/modified_first)
3396 do_node $ost1node "mkdir -p $ost1tmp && touch -m -d @$mtime $ost1tmp/modified_first"
3398 [ $? -eq 0 ] || { error "Unable to create temporary file"; return 6; }
3401 $LFS setstripe -c -1 -S 1M $DIR/$tdir
3402 [ $? -eq 0 ] || { error "lfs setstripe failed"; return 7; }
3404 for (( i=0; i < nrfiles; i++ )); do
3405 multiop $DIR/$tdir/$tfile-$i Ow1048576w1048576w524288c
3406 [ $? -eq 0 ] || { error "multiop failed"; return 8; }
3412 echo backup files to $TMP/files
3413 local files=$(find $DIR/$tdir -type f -newer $TMP/modified_first)
3414 copy_files_xattrs `hostname` $TMP/files $TMP/file_xattrs $files
3415 [ $? -eq 0 ] || { error "Unable to copy files"; return 9; }
3417 umount_client $MOUNT
3418 [ $? -eq 0 ] || { error "Unable to umount client"; return 10; }
3420 [ $? -eq 0 ] || { error "Unable to stop ost1"; return 11; }
3422 echo mount ost1 as ldiskfs
3423 do_node $ost1node mkdir -p $ost1mnt
3424 [ $? -eq 0 ] || { error "Unable to create $ost1mnt"; return 23; }
3425 if ! do_node $ost1node test -b $ost1_dev; then
3428 do_node $ost1node mount -t $(facet_fstype ost1) $loop $ost1_dev \
3430 [ $? -eq 0 ] || { error "Unable to mount ost1 as ldiskfs"; return 12; }
3433 echo backup objects to $ost1tmp/objects
3434 local objects=$(do_node $ost1node 'find '$ost1mnt'/O/[0-9]* -type f'\
3435 '-size +0 -newer '$ost1tmp'/modified_first -regex ".*\/[0-9]+"')
3436 copy_files_xattrs $ost1node $ost1tmp/objects $ost1tmp/object_xattrs \
3438 [ $? -eq 0 ] || { error "Unable to copy objects"; return 13; }
3440 # move objects to lost+found
3441 do_node $ost1node 'mv '$objects' '${ost1mnt}'/lost+found'
3442 [ $? -eq 0 ] || { error "Unable to move objects"; return 14; }
3445 do_node $ost1node "ll_recover_lost_found_objs -d $ost1mnt/lost+found"
3446 [ $? -eq 0 ] || { error "ll_recover_lost_found_objs failed"; return 15; }
3448 # compare restored objects against saved ones
3449 diff_files_xattrs $ost1node $ost1tmp/objects $ost1tmp/object_xattrs $objects
3450 [ $? -eq 0 ] || { error "Unable to diff objects"; return 16; }
3452 do_node $ost1node "umount $ost1mnt"
3453 [ $? -eq 0 ] || { error "Unable to umount ost1 as ldiskfs"; return 17; }
3456 [ $? -eq 0 ] || { error "Unable to start ost1"; return 18; }
3458 [ $? -eq 0 ] || { error "Unable to mount client"; return 19; }
3461 diff_files_xattrs `hostname` $TMP/files $TMP/file_xattrs $files
3462 [ $? -eq 0 ] || { error "Unable to diff files"; return 20; }
3464 rm -rf $TMP/files $TMP/file_xattrs
3465 [ $? -eq 0 ] || { error "Unable to delete temporary files"; return 21; }
3466 do_node $ost1node "rm -rf $ost1tmp"
3467 [ $? -eq 0 ] || { error "Unable to delete temporary files"; return 22; }
3470 run_test 52 "check recovering objects from lost+found"
3472 # Checks threads_min/max/started for some service
3474 # Arguments: service name (OST or MDT), facet (e.g., ost1, $SINGLEMDS), and a
3475 # parameter pattern prefix like 'ost.*.ost'.
3488 local msg="Insane $modname thread counts"
3489 local ncpts=$(check_cpt_number $facet)
3493 check_mount || return 41
3495 # We need to expand $parampat, but it may match multiple parameters, so
3496 # we'll pick the first one
3497 if ! paramp=$(do_facet $facet "lctl get_param -N ${parampat}.threads_min"|head -1); then
3498 error "Couldn't expand ${parampat}.threads_min parameter name"
3502 # Remove the .threads_min part
3503 paramp=${paramp%.threads_min}
3505 # Check for sanity in defaults
3506 tmin=$(do_facet $facet "lctl get_param -n ${paramp}.threads_min" || echo 0)
3507 tmax=$(do_facet $facet "lctl get_param -n ${paramp}.threads_max" || echo 0)
3508 tstarted=$(do_facet $facet "lctl get_param -n ${paramp}.threads_started" || echo 0)
3509 lassert 23 "$msg (PDSH problems?)" '(($tstarted && $tmin && $tmax))' || return $?
3510 lassert 24 "$msg" '(($tstarted >= $tmin && $tstarted <= $tmax ))' || return $?
3511 nthrs=$(expr $tmax - $tmin)
3512 if [ $nthrs -lt $ncpts ]; then
3518 [ $tmin -eq $tmax -a $tmin -eq $tstarted ] &&
3519 skip_env "module parameter forced $facet thread count" &&
3520 tmin=3 && tmax=$((3 * tmax))
3522 # Check that we can change min/max
3523 do_facet $facet "lctl set_param ${paramp}.threads_min=$((tmin + nthrs))"
3524 do_facet $facet "lctl set_param ${paramp}.threads_max=$((tmax - nthrs))"
3525 tmin2=$(do_facet $facet "lctl get_param -n ${paramp}.threads_min" || echo 0)
3526 tmax2=$(do_facet $facet "lctl get_param -n ${paramp}.threads_max" || echo 0)
3527 lassert 25 "$msg" '(($tmin2 == ($tmin + $nthrs) && $tmax2 == ($tmax - $nthrs)))' || return $?
3529 # Check that we can set min/max to the same value
3530 tmin=$(do_facet $facet "lctl get_param -n ${paramp}.threads_min" || echo 0)
3531 do_facet $facet "lctl set_param ${paramp}.threads_max=$tmin"
3532 tmin2=$(do_facet $facet "lctl get_param -n ${paramp}.threads_min" || echo 0)
3533 tmax2=$(do_facet $facet "lctl get_param -n ${paramp}.threads_max" || echo 0)
3534 lassert 26 "$msg" '(($tmin2 == $tmin && $tmax2 == $tmin))' || return $?
3536 # Check that we can't set max < min
3537 do_facet $facet "lctl set_param ${paramp}.threads_max=$((tmin - 1))"
3538 tmin2=$(do_facet $facet "lctl get_param -n ${paramp}.threads_min" || echo 0)
3539 tmax2=$(do_facet $facet "lctl get_param -n ${paramp}.threads_max" || echo 0)
3540 lassert 27 "$msg" '(($tmin2 <= $tmax2))' || return $?
3542 # We need to ensure that we get the module options desired; to do this
3543 # we set LOAD_MODULES_REMOTE=true and we call setmodopts below.
3544 LOAD_MODULES_REMOTE=true
3547 local newvalue="${opts}=$(expr $basethr \* $ncpts)"
3548 setmodopts -a $modname "$newvalue" oldvalue
3552 check_mount || return 41
3554 # Restore previous setting of MODOPTS_*
3555 setmodopts $modname "$oldvalue"
3557 # Check that $opts took
3558 tmin=$(do_facet $facet "lctl get_param -n ${paramp}.threads_min")
3559 tmax=$(do_facet $facet "lctl get_param -n ${paramp}.threads_max")
3560 tstarted=$(do_facet $facet "lctl get_param -n ${paramp}.threads_started")
3561 lassert 28 "$msg" '(($tstarted >= $tmin && $tstarted <= $tmax ))' || return $?
3570 thread_sanity OST ost1 'ost.*.ost' 'oss_num_threads' '16'
3573 run_test 53a "check OSS thread count params"
3577 local mds=$(do_facet $SINGLEMDS "lctl get_param -N mds.*.*.threads_max \
3579 if [ -z "$mds" ]; then
3580 #running this on an old MDT
3581 thread_sanity MDT $SINGLEMDS 'mdt.*.*.' 'mdt_num_threads' 16
3583 thread_sanity MDT $SINGLEMDS 'mds.*.*.' 'mds_num_threads' 16
3587 run_test 53b "check MDS thread count params"
3590 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
3591 skip "Only applicable to ldiskfs-based MDTs"
3595 do_rpc_nodes $(facet_host ost1) run_llverdev $(ostdevname 1) -p
3596 [ $? -eq 0 ] || error "llverdev failed!"
3599 run_test 54a "test llverdev and partial verify of device"
3602 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
3603 skip "Only applicable to ldiskfs-based MDTs"
3608 run_llverfs $MOUNT -p
3609 [ $? -eq 0 ] || error "llverfs failed!"
3612 run_test 54b "test llverfs and partial verify of filesystem"
3616 local max_ost_index=$1
3617 echo -n $(((max_ost_index + 1) * 8))
3621 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
3622 skip "Only applicable to ldiskfs-based MDTs"
3626 local mdsdev=$(mdsdevname 1)
3627 local mdsvdev=$(mdsvdevname 1)
3631 add mds1 $(mkfs_opts mds1 ${mdsdev}) --reformat $mdsdev \
3633 add ost1 $(mkfs_opts ost1 $(ostdevname 1)) --index=$i \
3634 --reformat $(ostdevname 1) $(ostvdevname 1)
3640 echo checking size of lov_objid for ost index $i
3641 LOV_OBJID_SIZE=$(do_facet mds1 "$DEBUGFS -R 'stat lov_objid' $mdsdev 2>/dev/null" | grep ^User | awk '{print $6}')
3642 if [ "$LOV_OBJID_SIZE" != $(lov_objid_size $i) ]; then
3643 error "lov_objid size has to be $(lov_objid_size $i), not $LOV_OBJID_SIZE"
3645 echo ok, lov_objid size is correct: $LOV_OBJID_SIZE
3652 run_test 55 "check lov_objid size"
3655 local mds_journal_size_orig=$MDSJOURNALSIZE
3660 for num in $(seq 1 $MDSCOUNT); do
3661 add mds${num} $(mkfs_opts mds${num} $(mdsdevname $num)) \
3662 --reformat $(mdsdevname $num) $(mdsvdevname $num)
3664 add ost1 $(mkfs_opts ost1 $(ostdevname 1)) --index=10000 --reformat \
3665 $(ostdevname 1) $(ostvdevname 1)
3666 add ost2 $(mkfs_opts ost2 $(ostdevname 2)) --index=1000 --reformat \
3667 $(ostdevname 2) $(ostvdevname 2)
3670 start_ost || error "Unable to start first ost (idx 10000)"
3671 start_ost2 || error "Unable to start second ost (idx 1000)"
3672 mount_client $MOUNT || error "Unable to mount client"
3675 wait_osc_import_state mds ost1 FULL
3676 wait_osc_import_state mds ost2 FULL
3677 $LFS setstripe --stripe-count=-1 $DIR/$tfile || error "Unable to create"
3678 n=$($LFS getstripe --stripe-count $DIR/$tfile)
3679 [ "$n" -eq 2 ] || error "Stripe count not two: $n"
3682 MDSJOURNALSIZE=$mds_journal_size_orig
3685 run_test 56 "check big OST indexes and out-of-index-order start"
3687 test_57a() { # bug 22656
3688 local NID=$(do_facet ost1 "$LCTL get_param nis" | tail -1 | awk '{print $1}')
3689 writeconf_or_reformat
3690 [ $(facet_fstype ost1) == zfs ] && import_zpool ost1
3691 do_facet ost1 "$TUNEFS --failnode=$NID `ostdevname 1`" || error "tunefs failed"
3693 start_ost && error "OST registration from failnode should fail"
3696 run_test 57a "initial registration from failnode should fail (should return errs)"
3699 local NID=$(do_facet ost1 "$LCTL get_param nis" | tail -1 | awk '{print $1}')
3700 writeconf_or_reformat
3701 [ $(facet_fstype ost1) == zfs ] && import_zpool ost1
3702 do_facet ost1 "$TUNEFS --servicenode=$NID `ostdevname 1`" || error "tunefs failed"
3704 start_ost || error "OST registration from servicenode should not fail"
3707 run_test 57b "initial registration from servicenode should not fail"
3710 do_facet mgs $LCTL get_param mgs.MGS.live.$FSNAME | grep OST | wc -l
3713 test_58() { # bug 22658
3714 if [ $(facet_fstype mds) != ldiskfs ]; then
3715 skip "Only applicable to ldiskfs-based MDTs"
3720 createmany -o $DIR/$tdir/$tfile-%d 100
3721 # make sure that OSTs do not cancel llog cookies before we unmount the MDS
3722 #define OBD_FAIL_OBD_LOG_CANCEL_NET 0x601
3723 do_facet $SINGLEMDS "lctl set_param fail_loc=0x601"
3724 unlinkmany $DIR/$tdir/$tfile-%d 100
3727 local MNTDIR=$(facet_mntpt $SINGLEMDS)
3728 local devname=$(mdsdevname ${SINGLEMDS//mds/})
3730 if ! do_facet $SINGLEMDS "test -b $devname"; then
3734 # remove all files from the OBJECTS dir
3735 do_facet $SINGLEMDS "mount -t ldiskfs $opts $devname $MNTDIR"
3736 do_facet $SINGLEMDS "find $MNTDIR/O/1/d* -type f -delete"
3737 do_facet $SINGLEMDS "umount -d $MNTDIR"
3738 # restart MDS with missing llog files
3740 do_facet mds "lctl set_param fail_loc=0"
3743 run_test 58 "missing llog files must not prevent MDT from mounting"
3746 start_mgsmds >> /dev/null
3747 local C1=$(count_osts)
3748 if [ $C1 -eq 0 ]; then
3749 start_ost >> /dev/null
3753 echo "original ost count: $C1 (expect > 0)"
3754 [ $C1 -gt 0 ] || error "No OSTs in $FSNAME log"
3755 start_mgsmds -o writeconf >> /dev/null || error "MDT start failed"
3756 local C2=$(count_osts)
3757 echo "after mdt writeconf count: $C2 (expect 0)"
3758 [ $C2 -gt 0 ] && error "MDT writeconf should erase OST logs"
3759 echo "OST start without writeconf should fail:"
3760 start_ost >> /dev/null && error "OST start without writeconf didn't fail"
3761 echo "OST start with writeconf should succeed:"
3762 start_ost -o writeconf >> /dev/null || error "OST1 start failed"
3763 local C3=$(count_osts)
3764 echo "after ost writeconf count: $C3 (expect 1)"
3765 [ $C3 -eq 1 ] || error "new OST writeconf should add:"
3766 start_ost2 -o writeconf >> /dev/null || error "OST2 start failed"
3767 local C4=$(count_osts)
3768 echo "after ost2 writeconf count: $C4 (expect 2)"
3769 [ $C4 -eq 2 ] || error "OST2 writeconf should add log"
3770 stop_ost2 >> /dev/null
3771 cleanup_nocli >> /dev/null
3772 #writeconf to remove all ost2 traces for subsequent tests
3773 writeconf_or_reformat
3775 run_test 59 "writeconf mount option"
3777 test_60() { # LU-471
3780 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
3781 skip "Only applicable to ldiskfs-based MDTs"
3785 for num in $(seq $MDSCOUNT); do
3786 add mds${num} $(mkfs_opts mds${num} $(mdsdevname $num)) \
3787 --mkfsoptions='\" -E stride=64 -O ^uninit_bg\"' \
3788 --reformat $(mdsdevname $num) $(mdsvdevname $num) ||
3792 dump=$(do_facet $SINGLEMDS dumpe2fs $(mdsdevname 1))
3794 [ $rc -eq 0 ] || error "dumpe2fs $(mdsdevname 1) failed"
3796 # MDT default has dirdata feature
3797 echo $dump | grep dirdata > /dev/null || error "dirdata is not set"
3798 # we disable uninit_bg feature
3799 echo $dump | grep uninit_bg > /dev/null && error "uninit_bg is set"
3800 # we set stride extended options
3801 echo $dump | grep stride > /dev/null || error "stride is not set"
3804 run_test 60 "check mkfs.lustre --mkfsoptions -E -O options setting"
3807 local reformat=false
3809 [ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.1.53) ] ||
3810 { skip "Need MDS version at least 2.1.53"; return 0; }
3812 if [ $(facet_fstype $SINGLEMDS) == ldiskfs ] &&
3813 ! large_xattr_enabled; then
3815 LDISKFS_MKFS_OPTS+=" -O large_xattr"
3817 for num in $(seq $MDSCOUNT); do
3818 add mds${num} $(mkfs_opts mds$num $(mdsdevname $num)) \
3819 --reformat $(mdsdevname $num) $(mdsvdevname $num) ||
3820 error "add mds $num failed"
3824 setup_noconfig || error "setting up the filesystem failed"
3825 client_up || error "starting client failed"
3827 local file=$DIR/$tfile
3830 local large_value="$(generate_string $(max_xattr_size))"
3831 local small_value="bar"
3833 local name="trusted.big"
3834 log "save large xattr $name on $file"
3835 setfattr -n $name -v $large_value $file ||
3836 error "saving $name on $file failed"
3838 local new_value=$(get_xattr_value $name $file)
3839 [[ "$new_value" != "$large_value" ]] &&
3840 error "$name different after saving"
3842 log "shrink value of $name on $file"
3843 setfattr -n $name -v $small_value $file ||
3844 error "shrinking value of $name on $file failed"
3846 new_value=$(get_xattr_value $name $file)
3847 [[ "$new_value" != "$small_value" ]] &&
3848 error "$name different after shrinking"
3850 log "grow value of $name on $file"
3851 setfattr -n $name -v $large_value $file ||
3852 error "growing value of $name on $file failed"
3854 new_value=$(get_xattr_value $name $file)
3855 [[ "$new_value" != "$large_value" ]] &&
3856 error "$name different after growing"
3858 log "check value of $name on $file after remounting MDS"
3860 new_value=$(get_xattr_value $name $file)
3861 [[ "$new_value" != "$large_value" ]] &&
3862 error "$name different after remounting MDS"
3864 log "remove large xattr $name from $file"
3865 setfattr -x $name $file || error "removing $name from $file failed"
3870 LDISKFS_MKFS_OPTS=${LDISKFS_MKFS_OPTS% -O large_xattr}
3874 run_test 61 "large xattr"
3877 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
3878 skip "Only applicable to ldiskfs-based MDTs"
3883 local mdsdev=$(mdsdevname 1)
3884 local ostdev=$(ostdevname 1)
3886 [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.2.51) ]] ||
3887 { skip "Need MDS version at least 2.2.51"; return 0; }
3889 echo "disable journal for mds"
3890 do_facet mds tune2fs -O ^has_journal $mdsdev || error "tune2fs failed"
3891 start_mds && error "MDT start should fail"
3892 echo "disable journal for ost"
3893 do_facet ost1 tune2fs -O ^has_journal $ostdev || error "tune2fs failed"
3894 start_ost && error "OST start should fail"
3895 cleanup || return $?
3898 run_test 62 "start with disabled journal"
3901 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
3902 skip "Only applicable to ldiskfs-based MDTs"
3906 local inode_slab=$(do_facet $SINGLEMDS \
3907 "awk '/ldiskfs_inode_cache/ { print \\\$5 }' /proc/slabinfo")
3908 if [ -z "$inode_slab" ]; then
3909 skip "ldiskfs module has not been loaded"
3913 echo "$inode_slab ldisk inodes per page"
3914 [ "$inode_slab" -ge "3" ] ||
3915 error "ldisk inode size is too big, $inode_slab objs per page"
3918 run_test 63 "Verify each page can at least hold 3 ldisk inodes"
3923 start_ost2 || error "Unable to start second ost"
3924 mount_client $MOUNT || error "Unable to mount client"
3925 stop_ost2 || error "Unable to stop second ost"
3927 $LFS df --lazy || error "lfs df failed"
3928 cleanup || return $?
3929 #writeconf to remove all ost2 traces for subsequent tests
3930 writeconf_or_reformat
3932 run_test 64 "check lfs df --lazy "
3934 test_65() { # LU-2237
3935 # Currently, the test is only valid for ldiskfs backend
3936 [ "$(facet_fstype $SINGLEMDS)" != "ldiskfs" ] &&
3937 skip "non-ldiskfs backend" && return
3939 local devname=$(mdsdevname ${SINGLEMDS//mds/})
3940 local brpt=$(facet_mntpt brpt)
3943 if ! do_facet $SINGLEMDS "test -b $devname"; then
3948 local obj=$(do_facet $SINGLEMDS \
3949 "$DEBUGFS -c -R \\\"stat last_rcvd\\\" $devname" |
3951 if [ -z "$obj" ]; then
3952 # The MDT may be just re-formatted, mount the MDT for the
3953 # first time to guarantee the "last_rcvd" file is there.
3954 start_mds || error "fail to mount the MDS for the first time"
3958 # remove the "last_rcvd" file
3959 do_facet $SINGLEMDS "mkdir -p $brpt"
3960 do_facet $SINGLEMDS \
3961 "mount -t $(facet_fstype $SINGLEMDS) $opts $devname $brpt"
3962 do_facet $SINGLEMDS "rm -f ${brpt}/last_rcvd"
3963 do_facet $SINGLEMDS "umount -d $brpt"
3965 # restart MDS, the "last_rcvd" file should be recreated.
3966 start_mds || error "fail to restart the MDS"
3968 obj=$(do_facet $SINGLEMDS \
3969 "$DEBUGFS -c -R \\\"stat last_rcvd\\\" $devname" | grep Inode)
3970 [ -n "$obj" ] || error "fail to re-create the last_rcvd"
3972 run_test 65 "re-create the lost last_rcvd file when server mount"
3975 [[ $(lustre_version_code mgs) -ge $(version_code 2.3.59) ]] ||
3976 { skip "Need MGS version at least 2.3.59"; return 0; }
3979 local OST1_NID=$(do_facet ost1 $LCTL list_nids | head -1)
3980 local MDS_NID=$(do_facet $SINGLEMDS $LCTL list_nids | head -1)
3982 echo "replace_nids should fail if MDS, OSTs and clients are UP"
3983 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 $OST1_NID &&
3984 error "replace_nids fail"
3986 umount_client $MOUNT || error "unmounting client failed"
3987 echo "replace_nids should fail if MDS and OSTs are UP"
3988 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 $OST1_NID &&
3989 error "replace_nids fail"
3992 echo "replace_nids should fail if MDS is UP"
3993 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 $OST1_NID &&
3994 error "replace_nids fail"
3996 stop_mds || error "stopping mds failed"
3998 if combined_mgs_mds; then
3999 start_mdt 1 "-o nosvc" ||
4000 error "starting mds with nosvc option failed"
4003 echo "command should accept two parameters"
4004 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 &&
4005 error "command should accept two params"
4007 echo "correct device name should be passed"
4008 do_facet mgs $LCTL replace_nids $FSNAME-WRONG0000 $OST1_NID &&
4009 error "wrong devname"
4011 echo "wrong nids list should not destroy the system"
4012 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 "wrong nids list" &&
4015 echo "replace OST nid"
4016 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 $OST1_NID ||
4017 error "replace nids failed"
4019 echo "command should accept two parameters"
4020 do_facet mgs $LCTL replace_nids $FSNAME-MDT0000 &&
4021 error "command should accept two params"
4023 echo "wrong nids list should not destroy the system"
4024 do_facet mgs $LCTL replace_nids $FSNAME-MDT0000 "wrong nids list" &&
4027 echo "replace MDS nid"
4028 do_facet mgs $LCTL replace_nids $FSNAME-MDT0000 $MDS_NID ||
4029 error "replace nids failed"
4031 if ! combined_mgs_mds ; then
4038 check_mount || error "error after nid replace"
4039 cleanup || error "cleanup failed"
4042 run_test 66 "replace nids"
4044 test_67() { #LU-2950
4045 local legacy="$TMP/legacy_lnet_config"
4046 local new="$TMP/new_routes_test"
4047 local out="$TMP/config_out_file"
4048 local verify="$TMP/conv_verify"
4049 local verify_conf="$TMP/conf_verify"
4051 # Create the legacy file that will be run through the
4052 # lustre_routes_conversion script
4053 cat <<- LEGACY_LNET_CONFIG > $legacy
4054 tcp1 23 192.168.213.1@tcp:1; tcp5 34 193.30.4.3@tcp:4;
4055 tcp2 54 10.1.3.2@tcp;
4056 tcp3 10.3.4.3@tcp:3;
4060 # Create the verification file to verify the output of
4061 # lustre_routes_conversion script against.
4062 cat <<- VERIFY_LNET_CONFIG > $verify
4063 tcp1: { gateway: 192.168.213.1@tcp, hop: 23, priority: 1 }
4064 tcp5: { gateway: 193.30.4.3@tcp, hop: 34, priority: 4 }
4065 tcp2: { gateway: 10.1.3.2@tcp, hop: 54 }
4066 tcp3: { gateway: 10.3.4.3@tcp, priority: 3 }
4067 tcp4: { gateway: 10.3.3.4@tcp }
4070 # Create the verification file to verify the output of
4071 # lustre_routes_config script against
4072 cat <<- VERIFY_LNET_CONFIG > $verify_conf
4073 lctl --net tcp1 add_route 192.168.213.1@tcp 23 1
4074 lctl --net tcp5 add_route 193.30.4.3@tcp 34 4
4075 lctl --net tcp2 add_route 10.1.3.2@tcp 54 4
4076 lctl --net tcp3 add_route 10.3.4.3@tcp 1 3
4077 lctl --net tcp4 add_route 10.3.3.4@tcp 1 3
4080 lustre_routes_conversion $legacy $new > /dev/null
4081 if [ -f $new ]; then
4082 # verify the conversion output
4083 cmp -s $new $verify > /dev/null
4084 if [ $? -eq 1 ]; then
4085 error "routes conversion failed"
4088 lustre_routes_config --dry-run --verbose $new > $out
4089 # check that the script succeeded
4090 cmp -s $out $verify_conf > /dev/null
4091 if [ $? -eq 1 ]; then
4092 error "routes config failed"
4095 error "routes conversion test failed"
4097 # remove generated files
4098 rm -f $new $legacy $verify $verify_conf $out
4100 run_test 67 "test routes conversion and configuration"
4108 [ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.4.53) ] ||
4109 { skip "Need MDS version at least 2.4.53"; return 0; }
4111 umount_client $MOUNT || error "umount client failed"
4113 start_mdt 1 || error "MDT start failed"
4116 # START-END - the sequences we'll be reserving
4117 START=$(do_facet $SINGLEMDS \
4118 lctl get_param -n seq.ctl*.space | awk -F'[[ ]' '{print $2}')
4119 END=$((START + (1 << 30)))
4120 do_facet $SINGLEMDS \
4121 lctl set_param seq.ctl*.fldb="[$START-$END\):0:mdt"
4123 # reset the sequences MDT0000 has already assigned
4124 do_facet $SINGLEMDS \
4125 lctl set_param seq.srv*MDT0000.space=clear
4127 # remount to let the client allocate new sequence
4128 mount_client $MOUNT || error "mount client failed"
4131 do_facet $SINGLEMDS \
4132 lctl get_param seq.srv*MDT0000.space
4133 $LFS path2fid $DIR/$tfile
4135 local old_ifs="$IFS"
4137 fid=($($LFS path2fid $DIR/$tfile))
4141 if [[ $seq < $END ]]; then
4142 error "used reserved sequence $seq?"
4144 cleanup || return $?
4146 run_test 68 "be able to reserve specific sequences in FLDB"
4149 local server_version=$(lustre_version_code $SINGLEMDS)
4151 [[ $server_version -lt $(version_code 2.4.2) ]] &&
4152 skip "Need MDS version at least 2.4.2" && return
4154 [[ $server_version -ge $(version_code 2.4.50) ]] &&
4155 [[ $server_version -lt $(version_code 2.5.0) ]] &&
4156 skip "Need MDS version at least 2.5.0" && return
4160 # use OST0000 since it probably has the most creations
4161 local OSTNAME=$(ostname_from_index 0)
4162 local mdtosc_proc1=$(get_mdtosc_proc_path mds1 $OSTNAME)
4163 local last_id=$(do_facet mds1 lctl get_param -n \
4164 osc.$mdtosc_proc1.prealloc_last_id)
4166 # Want to have OST LAST_ID over 1.5 * OST_MAX_PRECREATE to
4167 # verify that the LAST_ID recovery is working properly. If
4168 # not, then the OST will refuse to allow the MDS connect
4169 # because the LAST_ID value is too different from the MDS
4170 #define OST_MAX_PRECREATE=20000
4171 local num_create=$((20000 * 5))
4174 $LFS setstripe -i 0 $DIR/$tdir
4175 createmany -o $DIR/$tdir/$tfile- $num_create ||
4176 error "createmany: failed to create $num_create files: $?"
4177 # delete all of the files with objects on OST0 so the
4178 # filesystem is not inconsistent later on
4179 $LFS find $MOUNT --ost 0 | xargs rm
4181 stop_ost || error "OST0 stop failure"
4182 add ost1 $(mkfs_opts ost1 $(ostdevname 1)) --reformat --replace \
4183 $(ostdevname 1) $(ostvdevname 1) ||
4184 error "reformat and replace $ostdev failed"
4185 start_ost || error "OST0 restart failure"
4186 wait_osc_import_state mds ost FULL
4188 touch $DIR/$tdir/$tfile-last || error "create file after reformat"
4189 local idx=$($LFS getstripe -i $DIR/$tdir/$tfile-last)
4190 [ $idx -ne 0 ] && error "$DIR/$tdir/$tfile-last on $idx not 0" || true
4194 run_test 69 "replace an OST with the same index"
4197 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
4202 start_mdt 1 || error "MDT0 start fail"
4204 start_ost || error "OST0 start fail"
4206 start_mdt 2 || error "MDT1 start fail"
4208 mount_client $MOUNT || error "mount client fails"
4210 mkdir -p $DIR/$tdir || error "create dir fail"
4212 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
4213 error "create remote dir fail"
4215 rm -rf $DIR/$tdir || error "delete dir fail"
4216 cleanup || return $?
4218 run_test 70a "start MDT0, then OST, then MDT1"
4221 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
4224 start_ost || error "OST0 start fail"
4226 start_mdt 1 || error "MDT0 start fail"
4227 start_mdt 2 || error "MDT1 start fail"
4229 mount_client $MOUNT || error "mount client fails"
4231 mkdir -p $DIR/$tdir || error "create dir fail"
4233 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
4234 error "create remote dir fail"
4236 rm -rf $DIR/$tdir || error "delete dir fail"
4238 cleanup || return $?
4240 run_test 70b "start OST, MDT1, MDT0"
4243 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
4246 start_mdt 1 || error "MDT0 start fail"
4247 start_mdt 2 || error "MDT1 start fail"
4248 start_ost || error "OST0 start fail"
4250 mount_client $MOUNT || error "mount client fails"
4251 stop_mdt 1 || error "MDT1 start fail"
4253 local mdc_for_mdt1=$($LCTL dl | grep MDT0000-mdc | awk '{print $4}')
4254 echo "deactivate $mdc_for_mdt1"
4255 $LCTL --device $mdc_for_mdt1 deactivate || return 1
4257 mkdir -p $DIR/$tdir && error "mkdir succeed"
4259 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir &&
4260 error "create remote dir succeed"
4262 cleanup || return $?
4264 run_test 70c "stop MDT0, mkdir fail, create remote dir fail"
4267 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
4270 start_mdt 1 || error "MDT0 start fail"
4271 start_mdt 2 || error "MDT1 start fail"
4272 start_ost || error "OST0 start fail"
4274 mount_client $MOUNT || error "mount client fails"
4276 stop_mdt 2 || error "MDT1 start fail"
4278 local mdc_for_mdt2=$($LCTL dl | grep MDT0001-mdc |
4280 echo "deactivate $mdc_for_mdt2"
4281 $LCTL --device $mdc_for_mdt2 deactivate ||
4282 error "set $mdc_for_mdt2 deactivate failed"
4284 mkdir -p $DIR/$tdir || error "mkdir fail"
4285 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir &&
4286 error "create remote dir succeed"
4288 rm -rf $DIR/$tdir || error "delete dir fail"
4290 cleanup || return $?
4292 run_test 70d "stop MDT1, mkdir succeed, create remote dir fail"
4295 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
4296 if combined_mgs_mds; then
4297 skip "needs separate MGS/MDT" && return
4301 start_mdt 1 || error "MDT0 start fail"
4302 start_ost || error "OST0 start fail"
4303 start_mdt 2 || error "MDT1 start fail"
4304 start_ost2 || error "OST1 start fail"
4306 mount_client $MOUNT || error "mount client fails"
4308 mkdir -p $DIR/$tdir || error "mkdir fail"
4309 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
4310 error "create remote dir succeed"
4312 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
4313 rm -rf $DIR/$tdir || error "delete dir fail"
4315 umount_client $MOUNT
4316 stop_mdt 1 || error "MDT0 stop fail"
4317 stop_mdt 2 || error "MDT1 stop fail"
4318 stop_ost || error "OST0 stop fail"
4319 stop_ost2 || error "OST1 stop fail"
4321 run_test 71a "start MDT0 OST0, MDT1, OST1"
4324 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
4325 if combined_mgs_mds; then
4326 skip "needs separate MGS/MDT" && return
4330 start_mdt 2 || error "MDT1 start fail"
4331 start_ost || error "OST0 start fail"
4332 start_mdt 1 || error "MDT0 start fail"
4333 start_ost2 || error "OST1 start fail"
4335 mount_client $MOUNT || error "mount client fails"
4337 mkdir -p $DIR/$tdir || error "mkdir fail"
4338 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
4339 error "create remote dir succeed"
4341 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
4342 rm -rf $DIR/$tdir || error "delete dir fail"
4344 umount_client $MOUNT
4345 stop_mdt 1 || error "MDT0 stop fail"
4346 stop_mdt 2 || error "MDT1 stop fail"
4347 stop_ost || error "OST0 stop fail"
4348 stop_ost2 || error "OST1 stop fail"
4350 run_test 71b "start MDT1, OST0, MDT0, OST1"
4353 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
4354 if combined_mgs_mds; then
4355 skip "needs separate MGS/MDT" && return
4359 start_ost || error "OST0 start fail"
4360 start_ost2 || error "OST1 start fail"
4361 start_mdt 2 || error "MDT1 start fail"
4362 start_mdt 1 || error "MDT0 start fail"
4364 mount_client $MOUNT || error "mount client fails"
4366 mkdir -p $DIR/$tdir || error "mkdir fail"
4367 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
4368 error "create remote dir succeed"
4370 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
4371 rm -rf $DIR/$tdir || error "delete dir fail"
4373 umount_client $MOUNT
4374 stop_mdt 1 || error "MDT0 stop fail"
4375 stop_mdt 2 || error "MDT1 stop fail"
4376 stop_ost || error "OST0 stop fail"
4377 stop_ost2 || error "OST1 stop fail"
4380 run_test 71c "start OST0, OST1, MDT1, MDT0"
4383 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
4384 if combined_mgs_mds; then
4385 skip "needs separate MGS/MDT" && return
4389 start_ost || error "OST0 start fail"
4390 start_mdt 2 || error "MDT0 start fail"
4391 start_mdt 1 || error "MDT0 start fail"
4392 start_ost2 || error "OST1 start fail"
4394 mount_client $MOUNT || error "mount client fails"
4396 mkdir -p $DIR/$tdir || error "mkdir fail"
4397 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
4398 error "create remote dir succeed"
4400 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
4401 rm -rf $DIR/$tdir || error "delete dir fail"
4403 umount_client $MOUNT
4404 stop_mdt 1 || error "MDT0 stop fail"
4405 stop_mdt 2 || error "MDT1 stop fail"
4406 stop_ost || error "OST0 stop fail"
4407 stop_ost2 || error "OST1 stop fail"
4410 run_test 71d "start OST0, MDT1, MDT0, OST1"
4413 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
4414 if combined_mgs_mds; then
4415 skip "needs separate MGS/MDT" && return
4419 start_ost || error "OST0 start fail"
4420 start_mdt 2 || error "MDT1 start fail"
4421 start_ost2 || error "OST1 start fail"
4422 start_mdt 1 || error "MDT0 start fail"
4424 mount_client $MOUNT || error "mount client fails"
4426 mkdir -p $DIR/$tdir || error "mkdir fail"
4427 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
4428 error "create remote dir succeed"
4430 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
4431 rm -rf $DIR/$tdir || error "delete dir fail"
4433 umount_client $MOUNT
4434 stop_mdt 1 || error "MDT0 stop fail"
4435 stop_mdt 2 || error "MDT1 stop fail"
4436 stop_ost || error "OST0 stop fail"
4437 stop_ost2 || error "OST1 stop fail"
4440 run_test 71e "start OST0, MDT1, OST1, MDT0"
4442 test_72() { #LU-2634
4443 local mdsdev=$(mdsdevname 1)
4444 local ostdev=$(ostdevname 1)
4445 local cmd="$E2FSCK -fnvd $mdsdev"
4448 [ "$(facet_fstype $SINGLEMDS)" != "ldiskfs" ] &&
4449 skip "ldiskfs only test" && return
4451 #tune MDT with "-O extents"
4453 for num in $(seq $MDSCOUNT); do
4454 add mds${num} $(mkfs_opts mds$num $(mdsdevname $num)) \
4455 --reformat $(mdsdevname $num) $(mdsvdevname $num) ||
4456 error "add mds $num failed"
4457 $TUNE2FS -O extents $(mdsdevname $num)
4460 add ost1 $(mkfs_opts ost1 $ostdev) --reformat $ostdev ||
4461 error "add $ostdev failed"
4462 start_mgsmds || error "start mds failed"
4463 start_ost || error "start ost failed"
4464 mount_client $MOUNT || error "mount client failed"
4466 #create some short symlinks
4468 createmany -o $DIR/$tdir/$tfile-%d $fn
4469 echo "create $fn short symlinks"
4470 for i in $(seq -w 1 $fn); do
4471 ln -s $DIR/$tdir/$tfile-$i $MOUNT/$tfile-$i
4476 umount_client $MOUNT || error "umount client failed"
4477 stop_mds || error "stop mds failed"
4478 stop_ost || error "stop ost failed"
4481 run_e2fsck $(facet_active_host $SINGLEMDS) $mdsdev "-n"
4483 run_test 72 "test fast symlink with extents flag enabled"
4485 test_73() { #LU-3006
4487 [ $(facet_fstype ost1) == zfs ] && import_zpool ost1
4488 do_facet ost1 "$TUNEFS --failnode=1.2.3.4@$NETTYPE $(ostdevname 1)" ||
4489 error "1st tunefs failed"
4490 start_mgsmds || error "start mds failed"
4491 start_ost || error "start ost failed"
4492 mount_client $MOUNT || error "mount client failed"
4493 lctl get_param -n osc.*OST0000-osc-[^M]*.import | grep failover_nids |
4494 grep 1.2.3.4@$NETTYPE || error "failover nids haven't changed"
4495 umount_client $MOUNT || error "umount client failed"
4499 run_test 73 "failnode to update from mountdata properly"
4501 test_74() { # LU-1606
4502 for TESTPROG in $LUSTRE_TESTS_API_DIR/*.c; do
4503 gcc -Wall -Werror $LUSTRE_TESTS_API_DIR/simple_test.c \
4505 -L$LUSTRE/utils -llustreapi ||
4506 error "client api broken"
4508 cleanup || return $?
4510 run_test 74 "Lustre client api program can compile and link"
4512 test_75() { # LU-2374
4513 [[ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.4.1) ]] &&
4514 skip "Need MDS version at least 2.4.1" && return
4517 local opts_mds="$(mkfs_opts mds1 $(mdsdevname 1)) \
4518 --reformat $(mdsdevname 1) $(mdsvdevname 1)"
4519 local opts_ost="$(mkfs_opts ost1 $(ostdevname 1)) \
4520 --reformat $(ostdevname 1) $(ostvdevname 1)"
4522 #check with default parameters
4523 add mds1 $opts_mds || error "add mds1 failed for default params"
4524 add ost1 $opts_ost || error "add ost1 failed for default params"
4526 opts_mds=$(echo $opts_mds | sed -e "s/--mdt//")
4527 opts_mds=$(echo $opts_mds |
4528 sed -e "s/--index=$index/--index=$index --mdt/")
4529 opts_ost=$(echo $opts_ost | sed -e "s/--ost//")
4530 opts_ost=$(echo $opts_ost |
4531 sed -e "s/--index=$index/--index=$index --ost/")
4533 add mds1 $opts_mds || error "add mds1 failed for new params"
4534 add ost1 $opts_ost || error "add ost1 failed for new params"
4537 run_test 75 "The order of --index should be irrelevant"
4540 [[ $(lustre_version_code mgs) -ge $(version_code 2.4.52) ]] ||
4541 { skip "Need MDS version at least 2.4.52" && return 0; }
4543 local MDMB_PARAM="osc.*.max_dirty_mb"
4544 echo "Change MGS params"
4545 local MAX_DIRTY_MB=$($LCTL get_param -n $MDMB_PARAM |
4547 echo "max_dirty_mb: $MAX_DIRTY_MB"
4548 local NEW_MAX_DIRTY_MB=$((MAX_DIRTY_MB + MAX_DIRTY_MB))
4549 echo "new_max_dirty_mb: $NEW_MAX_DIRTY_MB"
4550 do_facet mgs $LCTL set_param -P $MDMB_PARAM=$NEW_MAX_DIRTY_MB
4551 wait_update $HOSTNAME "lctl get_param -n $MDMB_PARAM |
4552 head -1" $NEW_MAX_DIRTY_MB
4553 MAX_DIRTY_MB=$($LCTL get_param -n $MDMB_PARAM | head -1)
4554 echo "$MAX_DIRTY_MB"
4555 [ $MAX_DIRTY_MB = $NEW_MAX_DIRTY_MB ] ||
4556 error "error while apply max_dirty_mb"
4558 echo "Check the value is stored after remount"
4561 wait_update $HOSTNAME "lctl get_param -n $MDMB_PARAM |
4562 head -1" $NEW_MAX_DIRTY_MB
4563 MAX_DIRTY_MB=$($LCTL get_param -n $MDMB_PARAM | head -1)
4564 [ $MAX_DIRTY_MB = $NEW_MAX_DIRTY_MB ] ||
4565 error "max_dirty_mb is not saved after remount"
4567 echo "Change OST params"
4568 CLIENT_PARAM="obdfilter.*.client_cache_count"
4569 local CLIENT_CACHE_COUNT
4570 CLIENT_CACHE_COUNT=$(do_facet ost1 $LCTL get_param -n $CLIENT_PARAM |
4572 echo "client_cache_count: $CLIENT_CACHE_COUNT"
4573 NEW_CLIENT_CACHE_COUNT=$((CLIENT_CACHE_COUNT+CLIENT_CACHE_COUNT))
4574 echo "new_client_cache_count: $NEW_CLIENT_CACHE_COUNT"
4575 do_facet mgs $LCTL set_param -P $CLIENT_PARAM=$NEW_CLIENT_CACHE_COUNT
4576 wait_update $(facet_host ost1) "lctl get_param -n $CLIENT_PARAM |
4577 head -1" $NEW_CLIENT_CACHE_COUNT
4578 CLIENT_CACHE_COUNT=$(do_facet ost1 $LCTL get_param -n $CLIENT_PARAM |
4580 echo "$CLIENT_CACHE_COUNT"
4581 [ $CLIENT_CACHE_COUNT = $NEW_CLIENT_CACHE_COUNT ] ||
4582 error "error while apply client_cache_count"
4584 echo "Check the value is stored after remount"
4587 wait_update $(facet_host ost1) "lctl get_param -n $CLIENT_PARAM |
4588 head -1" $NEW_CLIENT_CACHE_COUNT
4589 CLIENT_CACHE_COUNT=$(do_facet ost1 $LCTL get_param -n $CLIENT_PARAM |
4591 echo "$CLIENT_CACHE_COUNT"
4592 [ $CLIENT_CACHE_COUNT = $NEW_CLIENT_CACHE_COUNT ] ||
4593 error "client_cache_count is not saved after remount"
4596 run_test 76a "set permanent params set_param -P"
4598 test_76b() { # LU-4783
4599 [[ $(lustre_version_code mgs) -ge $(version_code 2.5.57) ]] ||
4600 { skip "Need MGS version at least 2.5.57" && return 0; }
4603 do_facet mgs $LCTL get_param mgs.MGS.live.params ||
4604 error "start params log failed"
4607 run_test 76b "verify params log setup correctly"
4609 test_77() { # LU-3445
4610 local server_version=$(lustre_version_code $SINGLEMDS)
4612 [[ $server_version -ge $(version_code 2.2.60) ]] &&
4613 [[ $server_version -le $(version_code 2.4.0) ]] &&
4614 skip "Need MDS version < 2.2.60 or > 2.4.0" && return
4616 if [[ -z "$fs2ost_DEV" || -z "$fs2mds_DEV" ]]; then
4617 is_blkdev $SINGLEMDS $(mdsdevname ${SINGLEMDS//mds/}) &&
4618 skip_env "mixed loopback and real device not working" && return
4621 local fs2mdsdev=$(mdsdevname 1_2)
4622 local fs2ostdev=$(ostdevname 1_2)
4623 local fs2mdsvdev=$(mdsvdevname 1_2)
4624 local fs2ostvdev=$(ostvdevname 1_2)
4625 local fsname=test1234
4627 local failnid="$(h2$NETTYPE 1.2.3.4),$(h2$NETTYPE 4.3.2.1)"
4629 add fs2mds $(mkfs_opts mds1 $fs2mdsdev) --mgs --fsname=$fsname \
4630 --reformat $fs2mdsdev $fs2mdsvdev || error "add fs2mds failed"
4631 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS && trap cleanup_24a EXIT INT ||
4632 error "start fs2mds failed"
4634 mgsnid=$(do_facet fs2mds $LCTL list_nids | xargs | tr ' ' ,)
4635 [[ $mgsnid = *,* ]] || mgsnid+=",$mgsnid"
4637 add fs2ost $(mkfs_opts ost1 $fs2ostdev) --mgsnode=$mgsnid \
4638 --failnode=$failnid --fsname=$fsname \
4639 --reformat $fs2ostdev $fs2ostvdev ||
4640 error "add fs2ost failed"
4641 start fs2ost $fs2ostdev $OST_MOUNT_OPTS || error "start fs2ost failed"
4644 $MOUNT_CMD $mgsnid:/$fsname $MOUNT2 || error "mount $MOUNT2 failed"
4645 DIR=$MOUNT2 MOUNT=$MOUNT2 check_mount || error "check $MOUNT2 failed"
4648 run_test 77 "comma-separated MGS NIDs and failover node NIDs"
4650 if ! combined_mgs_mds ; then