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:-$LUSTRE}
52 export MULTIOP=${MULTIOP:-multiop}
54 . $LUSTRE/tests/test-framework.sh
57 # use small MDS + OST size to speed formatting time
58 # do not use too small MDSSIZE/OSTSIZE, which affect the default jouranl size
61 . ${CONFIG:=$LUSTRE/tests/cfg/$NAME.sh}
63 if ! combined_mgs_mds; then
64 # bug number for skipped test: 23954
65 ALWAYS_EXCEPT="$ALWAYS_EXCEPT 24b"
68 # STORED_MDSSIZE is used in test_18
69 if [ -n "$MDSSIZE" ]; then
70 STORED_MDSSIZE=$MDSSIZE
73 # pass "-E lazy_itable_init" to mke2fs to speed up the formatting time
74 if [[ "$LDISKFS_MKFS_OPTS" != *lazy_itable_init* ]]; then
75 LDISKFS_MKFS_OPTS=$(csa_add "$LDISKFS_MKFS_OPTS" -E lazy_itable_init)
78 [ $(facet_fstype $SINGLEMDS) = "zfs" ] &&
79 # bug number for skipped test: LU-2778 LU-2059
80 ALWAYS_EXCEPT="$ALWAYS_EXCEPT 57b 50h"
85 require_dsh_mds || exit 0
86 require_dsh_ost || exit 0
88 [ "$SLOW" = "no" ] && EXCEPT_SLOW="30a 31 45"
94 # The MGS must be started before the OSTs for a new fs, so start
95 # and stop to generate the startup logs.
98 wait_osc_import_state mds ost FULL
103 reformat_and_config() {
105 if ! combined_mgs_mds ; then
111 writeconf_or_reformat() {
112 # There are at most 2 OSTs for write_conf test
113 # who knows if/where $TUNEFS is installed?
114 # Better reformat if it fails...
115 writeconf_all $MDSCOUNT 2 ||
116 { echo "tunefs failed, reformatting instead" &&
117 reformat_and_config && return 0; }
127 start mgs $MGSDEV $MGS_MOUNT_OPTS
133 local dev=$(mdsdevname $num)
136 echo "start mds service on `facet_active_host $facet`"
137 start $facet ${dev} $MDS_MOUNT_OPTS $@ || return 94
143 local dev=$(mdsdevname $num)
146 echo "stop mds service on `facet_active_host $facet`"
147 # These tests all use non-failover stop
148 stop $facet -f || return 97
154 for num in $(seq $MDSCOUNT); do
155 start_mdt $num $@ || return 94
160 if ! combined_mgs_mds ; then
168 for num in $(seq $MDSCOUNT); do
169 stop_mdt $num || return 97
174 echo "stop mgs service on `facet_active_host mgs`"
175 # These tests all use non-failover stop
176 stop mgs -f || return 97
180 echo "start ost1 service on `facet_active_host ost1`"
181 start ost1 `ostdevname 1` $OST_MOUNT_OPTS $@ || return 95
185 echo "stop ost1 service on `facet_active_host ost1`"
186 # These tests all use non-failover stop
187 stop ost1 -f || return 98
191 echo "start ost2 service on `facet_active_host ost2`"
192 start ost2 `ostdevname 2` $OST_MOUNT_OPTS $@ || return 92
196 echo "stop ost2 service on `facet_active_host ost2`"
197 # These tests all use non-failover stop
198 stop ost2 -f || return 93
203 echo "mount $FSNAME on ${MOUNTPATH}....."
204 zconf_mount `hostname` $MOUNTPATH || return 96
208 local mountopt="-o remount,$1"
210 echo "remount '$1' lustre on ${MOUNTPATH}....."
211 zconf_mount `hostname` $MOUNTPATH "$mountopt" || return 96
216 echo "umount lustre on ${MOUNTPATH}....."
217 zconf_umount `hostname` $MOUNTPATH || return 97
220 manual_umount_client(){
223 echo "manual umount lustre on ${MOUNT}...."
224 do_facet client "umount -d ${FORCE} $MOUNT"
230 start_mds || error "MDT start failed"
231 start_ost || error "OST start failed"
232 mount_client $MOUNT || error "client start failed"
233 client_up || error "client_up failed"
237 if ! combined_mgs_mds ; then
246 unload_modules_conf () {
247 if combined_mgs_mds || ! local_mode; then
248 unload_modules || return 1
253 stop_ost || return 202
254 stop_mds || return 201
255 unload_modules_conf || return 203
259 umount_client $MOUNT || return 200
260 cleanup_nocli || return $?
264 do_facet client "cp /etc/passwd $DIR/a" || return 71
265 do_facet client "rm $DIR/a" || return 72
266 # make sure lustre is actually mounted (touch will block,
267 # but grep won't, so do it after)
268 do_facet client "grep $MOUNT' ' /proc/mounts > /dev/null" || return 73
269 echo "setup single mount lustre success"
273 do_facet client "touch $DIR/a" || return 71
274 do_facet client "rm $DIR/a" || return 72
275 do_facet client "touch $DIR2/a" || return 73
276 do_facet client "rm $DIR2/a" || return 74
277 echo "setup double mount lustre success"
282 if [ "$ONLY" == "setup" ]; then
287 if [ "$ONLY" == "cleanup" ]; then
294 #create single point mountpoint
300 check_mount || return 41
303 run_test 0 "single mount setup"
306 start_mds || error "MDT start failed"
308 echo "start ost second time..."
309 start_ost && error "2nd OST start should fail"
310 mount_client $MOUNT || error "client start failed"
311 check_mount || return 42
314 run_test 1 "start up ost twice (should return errors)"
318 echo "start mds second time.."
319 start_mdt 1 && error "2nd MDT start should fail"
322 check_mount || return 43
325 run_test 2 "start up mds twice (should return err)"
329 #mount.lustre returns an error if already in mtab
330 mount_client $MOUNT && error "2nd client mount should fail"
331 check_mount || return 44
334 run_test 3 "mount client twice (should return err)"
338 touch $DIR/$tfile || return 85
342 # ok for ost to fail shutdown
343 if [ 202 -ne $eno ]; then
348 run_test 4 "force cleanup ost, then cleanup"
350 test_5a() { # was test_5
352 touch $DIR/$tfile || return 1
353 fuser -m -v $MOUNT && echo "$MOUNT is in use by user space process."
355 stop_mds -f || return 2
357 # cleanup may return an error from the failed
358 # disconnects; for now I'll consider this successful
359 # if all the modules have unloaded.
363 echo "killing umount"
364 kill -TERM $UMOUNT_PID
365 echo "waiting for umount to finish"
367 if grep " $MOUNT " /proc/mounts; then
368 echo "test 5: /proc/mounts after failed umount"
372 echo "killing umount"
373 kill -TERM $UMOUNT_PID
374 echo "waiting for umount to finish"
376 grep " $MOUNT " /proc/mounts && echo "test 5: /proc/mounts after second umount" && return 11
380 # stop_mds is a no-op here, and should not fail
381 cleanup_nocli || return $?
382 # df may have lingering entry
384 # mtab may have lingering entry
388 while [ "$WAIT" -ne "$MAX_WAIT" ]; do
390 grep -q $MOUNT" " /etc/mtab || break
391 echo "Waiting /etc/mtab updated ... "
392 WAIT=$(( WAIT + sleep))
394 [ "$WAIT" -eq "$MAX_WAIT" ] && error "/etc/mtab is not updated in $WAIT secs"
395 echo "/etc/mtab updated in $WAIT secs"
397 run_test 5a "force cleanup mds, then cleanup"
405 grep " $MOUNT " /etc/mtab && \
406 error false "unexpected entry in mtab before mount" && return 10
410 if ! combined_mgs_mds ; then
411 trap cleanup_5b EXIT ERR
416 [ -d $MOUNT ] || mkdir -p $MOUNT
417 mount_client $MOUNT && rc=1
418 grep " $MOUNT " /etc/mtab && \
419 error "$MOUNT entry in mtab after failed mount" && rc=11
421 # stop_mds is a no-op here, and should not fail
422 cleanup_nocli || rc=$?
423 if ! combined_mgs_mds ; then
428 run_test 5b "Try to start a client with no MGS (should return errs)"
431 grep " $MOUNT " /etc/mtab && \
432 error false "unexpected entry in mtab before mount" && return 10
437 [ -d $MOUNT ] || mkdir -p $MOUNT
438 local oldfs="${FSNAME}"
439 FSNAME="wrong.${FSNAME}"
440 mount_client $MOUNT || :
442 grep " $MOUNT " /etc/mtab && \
443 error "$MOUNT entry in mtab after failed mount" && rc=11
445 cleanup_nocli || rc=$?
448 run_test 5c "cleanup after failed mount (bug 2712) (should return errs)"
451 grep " $MOUNT " /etc/mtab && \
452 error false "unexpected entry in mtab before mount" && return 10
454 [ "$(facet_fstype ost1)" = "zfs" ] &&
455 skip "LU-2059: no local config for ZFS OSTs" && return
461 mount_client $MOUNT || rc=1
463 grep " $MOUNT " /etc/mtab && \
464 error "$MOUNT entry in mtab after unmount" && rc=11
467 run_test 5d "mount with ost down"
470 grep " $MOUNT " /etc/mtab && \
471 error false "unexpected entry in mtab before mount" && return 10
477 #define OBD_FAIL_PTLRPC_DELAY_SEND 0x506
478 do_facet client "lctl set_param fail_loc=0x80000506"
479 mount_client $MOUNT || echo "mount failed (not fatal)"
481 grep " $MOUNT " /etc/mtab && \
482 error "$MOUNT entry in mtab after unmount" && rc=11
485 run_test 5e "delayed connect, don't crash (bug 10268)"
488 if combined_mgs_mds ; then
489 skip "combined mgs and mds"
493 grep " $MOUNT " /etc/mtab && \
494 error false "unexpected entry in mtab before mount" && return 10
498 [ -d $MOUNT ] || mkdir -p $MOUNT
499 mount_client $MOUNT &
501 echo client_mount pid is $pid
505 if ! ps -f -p $pid >/dev/null; then
508 grep " $MOUNT " /etc/mtab && echo "test 5f: mtab after mount"
509 error "mount returns $rc, expected to hang"
518 # mount should succeed after start mds
521 [ $rc -eq 0 ] || error "mount returned $rc"
522 grep " $MOUNT " /etc/mtab && echo "test 5f: mtab after mount"
526 run_test 5f "mds down, cleanup after failed mount (bug 2712)"
531 mount_client ${MOUNT} || return 87
532 touch $DIR/a || return 86
535 run_test 6 "manual umount, then mount again"
540 cleanup_nocli || return $?
542 run_test 7 "manual umount, then cleanup"
547 check_mount2 || return 45
548 umount_client $MOUNT2
551 run_test 8 "double mount setup"
556 do_facet ost1 lctl set_param debug=\'inode trace\' || return 1
557 do_facet ost1 lctl set_param subsystem_debug=\'mds ost\' || return 1
559 CHECK_PTLDEBUG="`do_facet ost1 lctl get_param -n debug`"
560 if [ "$CHECK_PTLDEBUG" ] && { \
561 [ "$CHECK_PTLDEBUG" = "trace inode warning error emerg console" ] ||
562 [ "$CHECK_PTLDEBUG" = "trace inode" ]; }; then
563 echo "lnet.debug success"
565 echo "lnet.debug: want 'trace inode', have '$CHECK_PTLDEBUG'"
568 CHECK_SUBSYS="`do_facet ost1 lctl get_param -n subsystem_debug`"
569 if [ "$CHECK_SUBSYS" ] && [ "$CHECK_SUBSYS" = "mds ost" ]; then
570 echo "lnet.subsystem_debug success"
572 echo "lnet.subsystem_debug: want 'mds ost', have '$CHECK_SUBSYS'"
575 stop_ost || return $?
577 run_test 9 "test ptldebug and subsystem for mkfs"
585 do_facet $facet "test -b $dev" || rc=1
586 if [[ "$size" ]]; then
587 local in=$(do_facet $facet "dd if=$dev of=/dev/null bs=1k count=1 skip=$size 2>&1" |\
588 awk '($3 == "in") { print $1 }')
589 [[ $in = "1+0" ]] || rc=1
595 # Test 16 was to "verify that lustre will correct the mode of OBJECTS".
596 # But with new MDS stack we don't care about the mode of local objects
597 # anymore, so this test is removed. See bug 22944 for more details.
601 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
602 skip "Only applicable to ldiskfs-based MDTs"
607 check_mount || return 41
610 echo "Remove mds config log"
611 if ! combined_mgs_mds ; then
615 do_facet mgs "$DEBUGFS -w -R 'unlink CONFIGS/$FSNAME-MDT0000' \
616 $(mgsdevname) || return \$?" || return $?
618 if ! combined_mgs_mds ; then
623 start_mds && return 42
626 run_test 17 "Verify failed mds_postsetup won't fail assertion (2936) (should return errs)"
629 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
630 skip "Only applicable to ldiskfs-based MDTs"
634 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
639 # check if current MDSSIZE is large enough
640 [ $MDSSIZE -ge $MIN ] && OK=1 && myMDSSIZE=$MDSSIZE && \
641 log "use MDSSIZE=$MDSSIZE"
643 # check if the global config has a large enough MDSSIZE
644 [ -z "$OK" -a ! -z "$STORED_MDSSIZE" ] && [ $STORED_MDSSIZE -ge $MIN ] && \
645 OK=1 && myMDSSIZE=$STORED_MDSSIZE && \
646 log "use STORED_MDSSIZE=$STORED_MDSSIZE"
648 # check if the block device is large enough
649 is_blkdev $SINGLEMDS $MDSDEV $MIN
650 local large_enough=$?
651 if [ -n "$OK" ]; then
652 [ $large_enough -ne 0 ] && OK=""
654 [ $large_enough -eq 0 ] && OK=1 && myMDSSIZE=$MIN &&
655 log "use device $MDSDEV with MIN=$MIN"
658 # check if a loopback device has enough space for fs metadata (5%)
660 if [ -z "$OK" ]; then
661 local SPACE=$(do_facet $SINGLEMDS "[ -f $MDSDEV -o ! -e $MDSDEV ] && df -P \\\$(dirname $MDSDEV)" |
662 awk '($1 != "Filesystem") {print $4}')
663 ! [ -z "$SPACE" ] && [ $SPACE -gt $((MIN / 20)) ] && \
664 OK=1 && myMDSSIZE=$MIN && \
665 log "use file $MDSDEV with MIN=$MIN"
668 [ -z "$OK" ] && skip_env "$MDSDEV too small for ${MIN}kB MDS" && return
671 echo "mount mds with large journal..."
673 local OLD_MDSSIZE=$MDSSIZE
677 echo "mount lustre system..."
679 check_mount || return 41
681 echo "check journal size..."
682 local FOUNDSIZE=$(do_facet $SINGLEMDS "$DEBUGFS -c -R 'stat <8>' $MDSDEV" | awk '/Size: / { print $NF; exit;}')
683 if [ $FOUNDSIZE -gt $((32 * 1024 * 1024)) ]; then
684 log "Success: mkfs creates large journals. Size: $((FOUNDSIZE >> 20))M"
686 error "expected journal size > 32M, found $((FOUNDSIZE >> 20))M"
694 run_test 18 "check mkfs creates large journals"
697 start_mds || return 1
698 stop_mds -f || return 2
700 run_test 19a "start/stop MDS without OSTs"
703 [ "$(facet_fstype ost1)" = "zfs" ] &&
704 skip "LU-2059: no local config for ZFS OSTs" && return
706 start_ost || return 1
707 stop_ost -f || return 2
709 run_test 19b "start/stop OSTs without MDS"
712 # first format the ost/mdt
716 check_mount || return 43
718 remount_client ro $MOUNT || return 44
719 touch $DIR/$tfile && echo "$DIR/$tfile created incorrectly" && return 45
720 [ -e $DIR/$tfile ] && echo "$DIR/$tfile exists incorrectly" && return 46
721 remount_client rw $MOUNT || return 47
723 [ ! -f $DIR/$tfile ] && echo "$DIR/$tfile missing" && return 48
724 MCNT=`grep -c $MOUNT /etc/mtab`
725 [ "$MCNT" -ne 1 ] && echo "$MOUNT in /etc/mtab $MCNT times" && return 49
730 run_test 20 "remount ro,rw mounts work and doesn't break /etc/mtab"
735 wait_osc_import_state mds ost FULL
739 run_test 21a "start mds before ost, stop ost first"
742 [ "$(facet_fstype ost1)" = "zfs" ] &&
743 skip "LU-2059: no local config for ZFS OSTs" && return
747 wait_osc_import_state mds ost FULL
751 run_test 21b "start ost before mds, stop mds first"
757 wait_osc_import_state mds ost2 FULL
761 #writeconf to remove all ost2 traces for subsequent tests
762 writeconf_or_reformat
764 run_test 21c "start mds between two osts, stop mds last"
767 if combined_mgs_mds ; then
768 skip "need separate mgs device" && return 0
778 wait_osc_import_state mds ost2 FULL
784 #writeconf to remove all ost2 traces for subsequent tests
785 writeconf_or_reformat
788 run_test 21d "start mgs then ost and then mds"
795 echo Client mount with ost in logs, but none running
797 # wait until mds connected to ost and open client connection
798 for num in $(seq 1 $MDSCOUNT); do
799 wait_osc_import_state mds${num} ost FULL
803 # check_mount will block trying to contact ost
804 mcreate $DIR/$tfile || return 40
805 rm -f $DIR/$tfile || return 42
809 echo Client mount with a running ost
812 # if gss enabled, wait full time to let connection from
813 # mds to ost be established, due to the mismatch between
814 # initial connect timeout and gss context negotiation timeout.
815 # This perhaps could be remove after AT landed.
816 echo "sleep $((TIMEOUT + TIMEOUT + TIMEOUT))s"
817 sleep $((TIMEOUT + TIMEOUT + TIMEOUT))
820 for num in $(seq 1 $MDSCOUNT); do
821 wait_osc_import_state mds${num} ost FULL
823 wait_osc_import_state client ost FULL
824 check_mount || return 41
829 run_test 22 "start a client before osts (should return errs)"
831 test_23a() { # was test_23
835 # force down client so that recovering mds waits for reconnect
836 local running=$(grep -c $MOUNT /proc/mounts) || true
837 if [ $running -ne 0 ]; then
838 echo "Stopping client $MOUNT (opts: -f)"
842 # enter recovery on mds
844 # try to start a new client
845 mount_client $MOUNT &
847 MOUNT_PID=$(ps -ef | grep "t lustre" | grep -v grep | awk '{print $2}')
848 MOUNT_LUSTRE_PID=`ps -ef | grep mount.lustre | grep -v grep | awk '{print $2}'`
849 echo mount pid is ${MOUNT_PID}, mount.lustre pid is ${MOUNT_LUSTRE_PID}
851 ps --ppid $MOUNT_LUSTRE_PID
852 echo "waiting for mount to finish"
854 # "ctrl-c" sends SIGINT but it usually (in script) does not work on child process
855 # SIGTERM works but it does not spread to offspring processses
856 kill -s TERM $MOUNT_PID
857 kill -s TERM $MOUNT_LUSTRE_PID
858 # we can not wait $MOUNT_PID because it is not a child of this shell
864 while [ "$WAIT" -lt "$MAX_WAIT" ]; do
866 PID1=$(ps -ef | awk '{print $2}' | grep -w $MOUNT_PID)
867 PID2=$(ps -ef | awk '{print $2}' | grep -w $MOUNT_LUSTRE_PID)
870 [ -z "$PID1" -a -z "$PID2" ] && break
871 echo "waiting for mount to finish ... "
872 WAIT=$(( WAIT + sleep))
874 if [ "$WAIT" -eq "$MAX_WAIT" ]; then
875 error "MOUNT_PID $MOUNT_PID and "\
876 "MOUNT_LUSTRE_PID $MOUNT_LUSTRE_PID still not killed in $WAIT secs"
882 run_test 23a "interrupt client during recovery mount delay"
887 test_23b() { # was test_23
890 # Simulate -EINTR during mount OBD_FAIL_LDLM_CLOSE_THREAD
891 lctl set_param fail_loc=0x80000313
895 run_test 23b "Simulate -EINTR during mount"
897 fs2mds_HOST=$mds_HOST
898 fs2ost_HOST=$ost_HOST
900 MDSDEV1_2=$fs2mds_DEV
901 OSTDEV1_2=$fs2ost_DEV
902 OSTDEV2_2=$fs3ost_DEV
906 echo "umount $MOUNT2 ..."
907 umount $MOUNT2 || true
908 echo "stopping fs2mds ..."
909 stop fs2mds -f || true
910 echo "stopping fs2ost ..."
911 stop fs2ost -f || true
915 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
917 if [ -z "$fs2ost_DEV" -o -z "$fs2mds_DEV" ]; then
918 is_blkdev $SINGLEMDS $MDSDEV && \
919 skip_env "mixed loopback and real device not working" && return
922 [ -n "$ost1_HOST" ] && fs2ost_HOST=$ost1_HOST
924 local fs2mdsdev=$(mdsdevname 1_2)
925 local fs2ostdev=$(ostdevname 1_2)
926 local fs2mdsvdev=$(mdsvdevname 1_2)
927 local fs2ostvdev=$(ostvdevname 1_2)
929 # test 8-char fsname as well
930 local FSNAME2=test1234
932 add fs2mds $(mkfs_opts mds1 ${fs2mdsdev} ) --nomgs --mgsnode=$MGSNID \
933 --fsname=${FSNAME2} --reformat $fs2mdsdev $fs2mdsvdev || exit 10
935 add fs2ost $(mkfs_opts ost1 ${fs2ostdev}) --fsname=${FSNAME2} \
936 --reformat $fs2ostdev $fs2ostvdev || exit 10
939 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS && trap cleanup_24a EXIT INT
940 start fs2ost $fs2ostdev $OST_MOUNT_OPTS
942 mount -t lustre $MGSNID:/${FSNAME2} $MOUNT2 || return 1
944 check_mount || return 2
945 # files written on 1 should not show up on 2
946 cp /etc/passwd $DIR/$tfile
948 [ -e $MOUNT2/$tfile ] && error "File bleed" && return 7
951 cp /etc/passwd $MOUNT2/b || return 3
952 rm $MOUNT2/b || return 4
953 # 2 is actually mounted
954 grep $MOUNT2' ' /proc/mounts > /dev/null || return 5
956 facet_failover fs2mds
957 facet_failover fs2ost
960 # the MDS must remain up until last MDT
962 MDS=$(do_facet $SINGLEMDS "lctl get_param -n devices" | awk '($3 ~ "mdt" && $4 ~ "MDT") { print $4 }' | head -1)
963 [ -z "$MDS" ] && error "No MDT" && return 8
965 cleanup_nocli || return 6
967 run_test 24a "Multiple MDTs on a single node"
970 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
972 if [ -z "$fs2mds_DEV" ]; then
973 local dev=${SINGLEMDS}_dev
975 is_blkdev $SINGLEMDS $MDSDEV && \
976 skip_env "mixed loopback and real device not working" && return
979 local fs2mdsdev=$(mdsdevname 1_2)
980 local fs2mdsvdev=$(mdsvdevname 1_2)
982 add fs2mds $(mkfs_opts mds1 ${fs2mdsdev} ) --mgs --fsname=${FSNAME}2 \
983 --reformat $fs2mdsdev $fs2mdsvdev || exit 10
985 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS && return 2
988 run_test 24b "Multiple MGSs on a single node (should return err)"
992 check_mount || return 2
993 local MODULES=$($LCTL modules | awk '{ print $2 }')
994 rmmod $MODULES 2>/dev/null || true
997 run_test 25 "Verify modules are referenced"
1001 # we need modules before mount for sysctl, so make sure...
1002 do_facet $SINGLEMDS "lsmod | grep -q lustre || modprobe lustre"
1003 #define OBD_FAIL_MDS_FS_SETUP 0x135
1004 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000135"
1005 start_mds && echo MDS started && return 1
1006 lctl get_param -n devices
1007 DEVS=$(lctl get_param -n devices | egrep -v MG | wc -l)
1008 [ $DEVS -gt 0 ] && return 2
1009 # start mds to drop writeconf setting
1010 start_mds || return 3
1011 stop_mds || return 4
1012 unload_modules_conf || return $?
1014 run_test 26 "MDT startup failure cleans LOV (should return errs)"
1017 [ "$(facet_fstype ost1)" = "zfs" ] &&
1018 skip "LU-2059: no local config for ZFS OSTs" && return
1020 start_ost || return 1
1021 start_mds || return 2
1022 echo "Requeue thread should have started: "
1023 ps -e | grep ll_cfg_requeue
1024 set_conf_param_and_check ost1 \
1025 "lctl get_param -n obdfilter.$FSNAME-OST0000.client_cache_seconds" \
1026 "$FSNAME-OST0000.ost.client_cache_seconds" || return 3
1029 run_test 27a "Reacquire MGS lock if OST started first"
1034 local device=$(do_facet $SINGLEMDS "lctl get_param -n devices" |
1035 awk '($3 ~ "mdt" && $4 ~ "MDT0000") { print $4 }')
1037 facet_failover $SINGLEMDS
1038 set_conf_param_and_check $SINGLEMDS \
1039 "lctl get_param -n mdt.$device.identity_acquire_expire" \
1040 "$device.mdt.identity_acquire_expire" || return 3
1041 set_conf_param_and_check client \
1042 "lctl get_param -n mdc.$device-mdc-*.max_rpcs_in_flight"\
1043 "$device.mdc.max_rpcs_in_flight" || return 4
1047 run_test 27b "Reacquire MGS lock after failover"
1051 TEST="lctl get_param -n llite.$FSNAME-*.max_read_ahead_whole_mb"
1052 PARAM="$FSNAME.llite.max_read_ahead_whole_mb"
1054 FINAL=$(($ORIG + 1))
1055 set_conf_param_and_check client "$TEST" "$PARAM" $FINAL || return 3
1056 FINAL=$(($FINAL + 1))
1057 set_conf_param_and_check client "$TEST" "$PARAM" $FINAL || return 4
1058 umount_client $MOUNT || return 200
1061 if [ $RESULT -ne $FINAL ]; then
1062 echo "New config not seen: wanted $FINAL got $RESULT"
1065 echo "New config success: got $RESULT"
1067 set_conf_param_and_check client "$TEST" "$PARAM" $ORIG || return 5
1070 run_test 28 "permanent parameter setting"
1073 [ "$OSTCOUNT" -lt "2" ] && skip_env "$OSTCOUNT < 2, skipping" && return
1074 setup > /dev/null 2>&1
1078 local PARAM="$FSNAME-OST0001.osc.active"
1079 local PROC_ACT="osc.$FSNAME-OST0001-osc-[^M]*.active"
1080 local PROC_UUID="osc.$FSNAME-OST0001-osc-[^M]*.ost_server_uuid"
1082 ACTV=$(lctl get_param -n $PROC_ACT)
1084 set_conf_param_and_check client \
1085 "lctl get_param -n $PROC_ACT" "$PARAM" $DEAC || return 2
1086 # also check ost_server_uuid status
1087 RESULT=$(lctl get_param -n $PROC_UUID | grep DEACTIV)
1088 if [ -z "$RESULT" ]; then
1089 echo "Live client not deactivated: $(lctl get_param -n $PROC_UUID)"
1092 echo "Live client success: got $RESULT"
1096 for num in $(seq $MDSCOUNT); do
1097 local mdtosc=$(get_mdtosc_proc_path mds${num} $FSNAME-OST0001)
1098 local MPROC="osc.$mdtosc.active"
1103 RESULT=$(do_facet mds${num} " lctl get_param -n $MPROC")
1104 [ ${PIPESTATUS[0]} = 0 ] || error "Can't read $MPROC"
1105 if [ $RESULT -eq $DEAC ]; then
1106 echo -n "MDT deactivated also after"
1107 echo "$WAIT sec (got $RESULT)"
1111 if [ $WAIT -eq $MAX ]; then
1112 echo -n "MDT not deactivated: wanted $DEAC"
1116 echo "Waiting $(($MAX - $WAIT))secs for MDT deactivated"
1119 # test new client starts deactivated
1120 umount_client $MOUNT || return 200
1122 RESULT=$(lctl get_param -n $PROC_UUID | grep DEACTIV | grep NEW)
1123 if [ -z "$RESULT" ]; then
1124 echo "New client not deactivated from start: $(lctl get_param -n $PROC_UUID)"
1127 echo "New client success: got $RESULT"
1130 # make sure it reactivates
1131 set_conf_param_and_check client \
1132 "lctl get_param -n $PROC_ACT" "$PARAM" $ACTV || return 6
1134 umount_client $MOUNT
1137 #writeconf to remove all ost2 traces for subsequent tests
1138 writeconf_or_reformat
1140 run_test 29 "permanently remove an OST"
1145 echo Big config llog
1146 TEST="lctl get_param -n llite.$FSNAME-*.max_read_ahead_whole_mb"
1148 LIST=(1 2 3 4 5 4 3 2 1 2 3 4 5 4 3 2 1 2 3 4 5)
1149 for i in ${LIST[@]}; do
1150 set_conf_param_and_check client "$TEST" \
1151 "$FSNAME.llite.max_read_ahead_whole_mb" $i || return 3
1153 # make sure client restart still works
1154 umount_client $MOUNT
1155 mount_client $MOUNT || return 4
1156 [ "$($TEST)" -ne "$i" ] && error "Param didn't stick across restart $($TEST) != $i"
1159 echo Erase parameter setting
1160 do_facet mgs "$LCTL conf_param -d $FSNAME.llite.max_read_ahead_whole_mb" || return 6
1161 umount_client $MOUNT
1162 mount_client $MOUNT || return 6
1164 echo "deleted (default) value=$FINAL, orig=$ORIG"
1165 # assumes this parameter started at the default value
1166 [ "$FINAL" -eq "$ORIG" ] || fail "Deleted value=$FINAL, orig=$ORIG"
1170 run_test 30a "Big config llog and conf_param deletion"
1175 # Make a fake nid. Use the OST nid, and add 20 to the least significant
1176 # numerical part of it. Hopefully that's not already a failover address for
1178 OSTNID=$(do_facet ost1 "$LCTL get_param nis" | tail -1 | awk '{print $1}')
1179 ORIGVAL=$(echo $OSTNID | egrep -oi "[0-9]*@")
1180 NEWVAL=$((($(echo $ORIGVAL | egrep -oi "[0-9]*") + 20) % 256))
1181 NEW=$(echo $OSTNID | sed "s/$ORIGVAL/$NEWVAL@/")
1182 echo "Using fake nid $NEW"
1184 TEST="$LCTL get_param -n osc.$FSNAME-OST0000-osc-[^M]*.import | grep failover_nids | sed -n 's/.*\($NEW\).*/\1/p'"
1185 set_conf_param_and_check client "$TEST" \
1186 "$FSNAME-OST0000.failover.node" $NEW ||
1187 error "didn't add failover nid $NEW"
1188 NIDS=$($LCTL get_param -n osc.$FSNAME-OST0000-osc-[^M]*.import | grep failover_nids)
1190 NIDCOUNT=$(($(echo "$NIDS" | wc -w) - 1))
1191 echo "should have 2 failover nids: $NIDCOUNT"
1192 [ $NIDCOUNT -eq 2 ] || error "Failover nid not added"
1193 do_facet mgs "$LCTL conf_param -d $FSNAME-OST0000.failover.node" || error "conf_param delete failed"
1194 umount_client $MOUNT
1195 mount_client $MOUNT || return 3
1197 NIDS=$($LCTL get_param -n osc.$FSNAME-OST0000-osc-[^M]*.import | grep failover_nids)
1199 NIDCOUNT=$(($(echo "$NIDS" | wc -w) - 1))
1200 echo "only 1 final nid should remain: $NIDCOUNT"
1201 [ $NIDCOUNT -eq 1 ] || error "Failover nids not removed"
1205 run_test 30b "Remove failover nids"
1207 test_31() { # bug 10734
1208 # ipaddr must not exist
1209 mount -t lustre 4.3.2.1@tcp:/lustre $MOUNT || true
1212 run_test 31 "Connect to non-existent node (shouldn't crash)"
1216 T32_BLIMIT=20480 # Kbytes
1220 # This is not really a test but a tool to create new disk
1221 # image tarballs for the upgrade tests.
1223 # Disk image tarballs should be created on single-node
1224 # clusters by running this test with default configurations
1225 # plus a few mandatory environment settings that are verified
1226 # at the beginning of the test.
1228 test_32newtarball() {
1232 local tmp=$TMP/t32_image_create
1234 if [ $FSNAME != t32fs -o $MDSCOUNT -ne 1 -o \
1235 \( -z "$MDSDEV" -a -z "$MDSDEV1" \) -o $OSTCOUNT -ne 1 -o \
1236 -z "$OSTDEV1" ]; then
1237 error "Needs FSNAME=t32fs MDSCOUNT=1 MDSDEV1=<nonexistent_file>" \
1238 "(or MDSDEV, in the case of b1_8) OSTCOUNT=1" \
1239 "OSTDEV1=<nonexistent_file>"
1243 echo "Found stale $tmp"
1248 tar cf - -C $src . | tar xf - -C $tmp/src
1249 dd if=/dev/zero of=$tmp/src/t32_qf_old bs=1M \
1250 count=$(($T32_BLIMIT / 1024 / 2))
1251 chown $T32_QID.$T32_QID $tmp/src/t32_qf_old
1257 [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.3.50) ] &&
1258 $LFS quotacheck -ug /mnt/$FSNAME
1259 $LFS setquota -u $T32_QID -b 0 -B $T32_BLIMIT -i 0 -I $T32_ILIMIT \
1262 tar cf - -C $tmp/src . | tar xf - -C /mnt/$FSNAME
1269 ls -Rni --time-style=+%s >$tmp/img/list
1270 find . ! -name .lustre -type f -exec sha1sum {} \; |
1271 sort -k 2 >$tmp/img/sha1sums
1273 $LCTL get_param -n version | head -n 1 |
1274 sed -e 's/^lustre: *//' >$tmp/img/commit
1276 [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.3.50) ] &&
1277 $LFS quotaon -ug /mnt/$FSNAME
1278 $LFS quota -u $T32_QID -v /mnt/$FSNAME
1279 $LFS quota -v -u $T32_QID /mnt/$FSNAME |
1280 awk 'BEGIN { num='1' } { if ($1 == "'/mnt/$FSNAME'") \
1281 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1282 | tr -d "*" > $tmp/img/bspace
1283 $LFS quota -v -u $T32_QID /mnt/$FSNAME |
1284 awk 'BEGIN { num='5' } { if ($1 == "'/mnt/$FSNAME'") \
1285 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1286 | tr -d "*" > $tmp/img/ispace
1291 find -type f -exec sha1sum {} \; | sort -k 2 >$tmp/sha1sums.src
1294 if ! diff -u $tmp/sha1sums.src $tmp/img/sha1sums; then
1295 echo "Data verification failed"
1298 uname -r >$tmp/img/kernel
1299 uname -m >$tmp/img/arch
1301 mv ${MDSDEV1:-$MDSDEV} $tmp/img
1302 mv $OSTDEV1 $tmp/img
1304 version=$(sed -e 's/\(^[0-9]\+\.[0-9]\+\)\(.*$\)/\1/' $tmp/img/commit |
1305 sed -e 's/\./_/g') # E.g., "1.8.7" -> "1_8"
1308 tar cjvf $dst/disk$version-$(facet_fstype $SINGLEMDS).tar.bz2 -S *
1313 #run_test 32newtarball "Create a new test_32 disk image tarball for this version"
1316 # The list of applicable tarballs is returned via the caller's
1317 # variable "tarballs".
1320 local node=$(facet_active_host $SINGLEMDS)
1321 local r="do_node $node"
1323 if [ "$CLIENTONLY" ]; then
1324 skip "Client-only testing"
1328 if ! $r which $TUNEFS; then
1329 skip_env "tunefs.lustre required on $node"
1333 if [ -n "$($LCTL list_nids | grep -v '\(tcp\|lo\)[[:digit:]]*$')" ]; then
1334 skip "LU-2200: Test cannot run over Infiniband"
1338 local IMGTYPE=$(facet_fstype $SINGLEMDS)
1340 tarballs=$($r find $RLUSTRE/tests -maxdepth 1 -name \'disk*-$IMGTYPE.tar.bz2\')
1342 if [ -z "$tarballs" ]; then
1343 skip "No applicable tarballs found"
1348 t32_test_cleanup() {
1352 if $shall_cleanup_lustre; then
1353 umount $tmp/mnt/lustre || rc=$?
1355 if $shall_cleanup_mdt; then
1356 $r umount -d $tmp/mnt/mdt || rc=$?
1358 if $shall_cleanup_mdt1; then
1359 $r umount -d $tmp/mnt/mdt1 || rc=$?
1361 if $shall_cleanup_ost; then
1362 $r umount -d $tmp/mnt/ost || rc=$?
1370 t32_bits_per_long() {
1372 # Yes, this is not meant to be perfect.
1382 t32_reload_modules() {
1384 local all_removed=false
1387 while ((i < 20)); do
1388 echo "Unloading modules on $node: Attempt $i"
1389 do_rpc_nodes $node $LUSTRE_RMMOD $(facet_fstype $SINGLEMDS) &&
1391 do_rpc_nodes $node check_mem_leak || return 1
1392 if $all_removed; then
1393 do_rpc_nodes $node load_modules
1399 echo "Unloading modules on $node: Given up"
1403 t32_wait_til_devices_gone() {
1408 echo wait for devices to go
1409 while ((i < 20)); do
1410 devices=$(do_rpc_nodes $node $LCTL device_list | wc -l)
1411 ((devices == 0)) && return 0
1415 echo "waiting for devices on $node: Given up"
1419 t32_verify_quota() {
1423 local fstype=$(facet_fstype $SINGLEMDS)
1427 $LFS quota -u $T32_QID -v $mnt
1429 qval=$($LFS quota -v -u $T32_QID $mnt |
1430 awk 'BEGIN { num='1' } { if ($1 == "'$mnt'") \
1431 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1433 [ $qval -eq $img_bspace ] || {
1434 echo "bspace, act:$qval, exp:$img_bspace"
1438 qval=$($LFS quota -v -u $T32_QID $mnt |
1439 awk 'BEGIN { num='5' } { if ($1 == "'$mnt'") \
1440 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1442 [ $qval -eq $img_ispace ] || {
1443 echo "ispace, act:$qval, exp:$img_ispace"
1447 qval=$($LFS quota -v -u $T32_QID $mnt |
1448 awk 'BEGIN { num='3' } { if ($1 == "'$mnt'") \
1449 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1451 [ $qval -eq $T32_BLIMIT ] || {
1452 echo "blimit, act:$qval, exp:$T32_BLIMIT"
1456 qval=$($LFS quota -v -u $T32_QID $mnt |
1457 awk 'BEGIN { num='7' } { if ($1 == "'$mnt'") \
1458 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1460 [ $qval -eq $T32_ILIMIT ] || {
1461 echo "ilimit, act:$qval, exp:$T32_ILIMIT"
1465 do_node $node $LCTL conf_param $fsname.quota.mdt=ug
1466 cmd="$LCTL get_param -n osd-$fstype.$fsname-MDT0000"
1467 cmd=$cmd.quota_slave.enabled
1468 wait_update $node "$cmd" "ug" || {
1469 echo "Enable mdt quota failed"
1473 do_node $node $LCTL conf_param $fsname.quota.ost=ug
1474 cmd="$LCTL get_param -n osd-$fstype.$fsname-OST0000"
1475 cmd=$cmd.quota_slave.enabled
1476 wait_update $node "$cmd" "ug" || {
1477 echo "Enable ost quota failed"
1482 runas -u $T32_QID -g $T32_QID dd if=/dev/zero of=$mnt/t32_qf_new \
1483 bs=1M count=$(($T32_BLIMIT / 1024)) oflag=sync && {
1484 echo "Write succeed, but expect -EDQUOT"
1487 rm -f $mnt/t32_qf_new
1489 runas -u $T32_QID -g $T32_QID createmany -m $mnt/t32_qf_ \
1491 echo "Create succeed, but expect -EDQUOT"
1494 unlinkmany $mnt/t32_qf_ $T32_ILIMIT
1502 local dne_upgrade=${dne_upgrade:-"no"}
1503 local shall_cleanup_mdt=false
1504 local shall_cleanup_mdt1=false
1505 local shall_cleanup_ost=false
1506 local shall_cleanup_lustre=false
1507 local node=$(facet_active_host $SINGLEMDS)
1508 local r="do_node $node"
1509 local node2=$(facet_active_host mds2)
1517 local nid=$($r $LCTL list_nids | head -1)
1523 local fstype=$(facet_fstype $SINGLEMDS)
1525 trap 'trap - RETURN; t32_test_cleanup' RETURN
1527 mkdir -p $tmp/mnt/lustre
1528 $r mkdir -p $tmp/mnt/{mdt,ost}
1529 $r tar xjvf $tarball -S -C $tmp || {
1530 error_noexit "Unpacking the disk image tarball"
1533 img_commit=$($r cat $tmp/commit)
1534 img_kernel=$($r cat $tmp/kernel)
1535 img_arch=$($r cat $tmp/arch)
1536 img_bspace=$($r cat $tmp/bspace)
1537 img_ispace=$($r cat $tmp/ispace)
1538 echo "Upgrading from $(basename $tarball), created with:"
1539 echo " Commit: $img_commit"
1540 echo " Kernel: $img_kernel"
1541 echo " Arch: $img_arch"
1543 $r $LCTL set_param debug="$PTLDEBUG"
1545 $r $TUNEFS --dryrun $tmp/mdt || {
1546 error_noexit "tunefs.lustre before mounting the MDT"
1549 if [ "$writeconf" ]; then
1550 mopts=loop,writeconf
1551 if [ $fstype == "ldiskfs" ]; then
1552 $r $TUNEFS --quota $tmp/mdt || {
1553 error_noexit "Enable mdt quota feature"
1558 mopts=loop,exclude=$fsname-OST0000
1561 t32_wait_til_devices_gone $node
1563 $r mount -t lustre -o $mopts $tmp/mdt $tmp/mnt/mdt || {
1564 error_noexit "Mounting the MDT"
1567 shall_cleanup_mdt=true
1569 if [ "$dne_upgrade" != "no" ]; then
1570 local fs2mdsdev=$(mdsdevname 1_2)
1571 local fs2mdsvdev=$(mdsvdevname 1_2)
1573 echo "mkfs new MDT on ${fs2mdsdev}...."
1574 if [ $(facet_fstype mds1) == ldiskfs ]; then
1575 mkfsoptions="--mkfsoptions=\\\"-J size=8\\\""
1578 add fs2mds $(mkfs_opts mds2 $fs2mdsdev $fsname) --reformat \
1579 $mkfsoptions $fs2mdsdev $fs2mdsvdev > /dev/null || {
1580 error_noexit "Mkfs new MDT failed"
1584 $r $TUNEFS --dryrun $fs2mdsdev || {
1585 error_noexit "tunefs.lustre before mounting the MDT"
1589 echo "mount new MDT....$fs2mdsdev"
1590 $r mkdir -p $tmp/mnt/mdt1
1591 $r mount -t lustre -o $mopts $fs2mdsdev $tmp/mnt/mdt1 || {
1592 error_noexit "mount mdt1 failed"
1595 shall_cleanup_mdt1=true
1598 uuid=$($r $LCTL get_param -n mdt.$fsname-MDT0000.uuid) || {
1599 error_noexit "Getting MDT UUID"
1602 if [ "$uuid" != $fsname-MDT0000_UUID ]; then
1603 error_noexit "Unexpected MDT UUID: \"$uuid\""
1607 $r $TUNEFS --dryrun $tmp/ost || {
1608 error_noexit "tunefs.lustre before mounting the OST"
1611 if [ "$writeconf" ]; then
1612 mopts=loop,mgsnode=$nid,$writeconf
1613 if [ $fstype == "ldiskfs" ]; then
1614 $r $TUNEFS --quota $tmp/ost || {
1615 error_noexit "Enable ost quota feature"
1620 mopts=loop,mgsnode=$nid
1622 $r mount -t lustre -o $mopts $tmp/ost $tmp/mnt/ost || {
1623 error_noexit "Mounting the OST"
1626 shall_cleanup_ost=true
1628 uuid=$($r $LCTL get_param -n obdfilter.$fsname-OST0000.uuid) || {
1629 error_noexit "Getting OST UUID"
1632 if [ "$uuid" != $fsname-OST0000_UUID ]; then
1633 error_noexit "Unexpected OST UUID: \"$uuid\""
1637 $r $LCTL conf_param $fsname-OST0000.osc.max_dirty_mb=15 || {
1638 error_noexit "Setting \"max_dirty_mb\""
1641 $r $LCTL conf_param $fsname-OST0000.failover.node=$nid || {
1642 error_noexit "Setting OST \"failover.node\""
1645 $r $LCTL conf_param $fsname-MDT0000.mdc.max_rpcs_in_flight=9 || {
1646 error_noexit "Setting \"max_rpcs_in_flight\""
1649 $r $LCTL conf_param $fsname-MDT0000.failover.node=$nid || {
1650 error_noexit "Setting MDT \"failover.node\""
1653 $r $LCTL pool_new $fsname.interop || {
1654 error_noexit "Setting \"interop\""
1657 $r $LCTL conf_param $fsname-MDT0000.lov.stripesize=4M || {
1658 error_noexit "Setting \"lov.stripesize\""
1662 if [ "$dne_upgrade" != "no" ]; then
1663 $r $LCTL conf_param \
1664 $fsname-MDT0001.mdc.max_rpcs_in_flight=9 || {
1665 error_noexit "Setting MDT1 \"max_rpcs_in_flight\""
1668 $r $LCTL conf_param $fsname-MDT0001.failover.node=$nid || {
1669 error_noexit "Setting MDT1 \"failover.node\""
1672 $r $LCTL conf_param $fsname-MDT0001.lov.stripesize=4M || {
1673 error_noexit "Setting MDT1 \"lov.stripesize\""
1679 if [ "$writeconf" ]; then
1680 mount -t lustre $nid:/$fsname $tmp/mnt/lustre || {
1681 error_noexit "Mounting the client"
1684 shall_cleanup_lustre=true
1685 $LCTL set_param debug="$PTLDEBUG"
1687 t32_verify_quota $node $fsname $tmp/mnt/lustre || {
1688 error_noexit "verify quota failed"
1692 if [ "$dne_upgrade" != "no" ]; then
1693 $LFS mkdir -i 1 $tmp/mnt/lustre/remote_dir || {
1694 error_noexit "set remote dir failed"
1698 pushd $tmp/mnt/lustre
1699 tar -cf - . --exclude=./remote_dir |
1700 tar -xvf - -C remote_dir 1>/dev/null || {
1701 error_noexit "cp to remote dir failed"
1707 dd if=/dev/zero of=$tmp/mnt/lustre/tmp_file bs=10k count=10 || {
1708 error_noexit "dd failed"
1711 rm -rf $tmp/mnt/lustre/tmp_file || {
1712 error_noexit "rm failed"
1716 if $r test -f $tmp/sha1sums; then
1717 # LU-2393 - do both sorts on same node to ensure locale
1719 $r cat $tmp/sha1sums | sort -k 2 >$tmp/sha1sums.orig
1720 if [ "$dne_upgrade" != "no" ]; then
1721 pushd $tmp/mnt/lustre/remote_dir
1723 pushd $tmp/mnt/lustre
1726 find ! -name .lustre -type f -exec sha1sum {} \; |
1727 sort -k 2 >$tmp/sha1sums || {
1728 error_noexit "sha1sum"
1732 if ! diff -ub $tmp/sha1sums.orig $tmp/sha1sums; then
1733 error_noexit "sha1sum verification failed"
1737 echo "sha1sum verification skipped"
1740 if [ "$dne_upgrade" != "no" ]; then
1741 rm -rf $tmp/mnt/lustre/remote_dir || {
1742 error_noexit "remove remote dir failed"
1747 if $r test -f $tmp/list; then
1749 # There is not a Test Framework API to copy files to or
1750 # from a remote node.
1752 # LU-2393 - do both sorts on same node to ensure locale
1754 $r cat $tmp/list | sort -k 6 >$tmp/list.orig
1755 pushd $tmp/mnt/lustre
1756 ls -Rni --time-style=+%s | sort -k 6 >$tmp/list || {
1762 # 32-bit and 64-bit clients use different algorithms to
1763 # convert FIDs into inode numbers. Hence, remove the inode
1764 # numbers from the lists, if the original list was created
1765 # on an architecture with different number of bits per
1768 if [ $(t32_bits_per_long $(uname -m)) != \
1769 $(t32_bits_per_long $img_arch) ]; then
1770 echo "Different number of bits per \"long\" from the disk image"
1771 for list in list.orig list; do
1772 sed -i -e 's/^[0-9]\+[ \t]\+//' $tmp/$list
1775 if ! diff -ub $tmp/list.orig $tmp/list; then
1776 error_noexit "list verification failed"
1780 echo "list verification skipped"
1784 # When adding new data verification tests, please check for
1785 # the presence of the required reference files first, like
1786 # the "sha1sums" and "list" tests above, to avoid the need to
1787 # regenerate every image for each test addition.
1790 nrpcs_orig=$($LCTL get_param \
1791 -n mdc.*MDT0000*.max_rpcs_in_flight) || {
1792 error_noexit "Getting \"max_rpcs_in_flight\""
1795 nrpcs=$((nrpcs_orig + 5))
1796 $r $LCTL conf_param $fsname-MDT0000.mdc.max_rpcs_in_flight=$nrpcs || {
1797 error_noexit "Changing \"max_rpcs_in_flight\""
1800 wait_update $HOSTNAME "$LCTL get_param \
1801 -n mdc.*MDT0000*.max_rpcs_in_flight" $nrpcs || {
1802 error_noexit "Verifying \"max_rpcs_in_flight\""
1806 umount $tmp/mnt/lustre || {
1807 error_noexit "Unmounting the client"
1810 shall_cleanup_lustre=false
1812 if [ "$dne_upgrade" != "no" ]; then
1813 $r umount -d $tmp/mnt/mdt1 || {
1814 error_noexit "Unmounting the MDT2"
1817 shall_cleanup_mdt1=false
1820 $r umount -d $tmp/mnt/mdt || {
1821 error_noexit "Unmounting the MDT"
1824 shall_cleanup_mdt=false
1826 $r umount -d $tmp/mnt/ost || {
1827 error_noexit "Unmounting the OST"
1830 shall_cleanup_ost=false
1832 t32_reload_modules $node || {
1833 error_noexit "Reloading modules"
1837 # mount a second time to make sure we didnt leave upgrade flag on
1838 $r $TUNEFS --dryrun $tmp/mdt || {
1839 error_noexit "tunefs.lustre before remounting the MDT"
1842 $r mount -t lustre -o loop,exclude=$fsname-OST0000 $tmp/mdt \
1844 error_noexit "Remounting the MDT"
1847 shall_cleanup_mdt=true
1857 for tarball in $tarballs; do
1858 t32_test $tarball || rc=$?
1862 run_test 32a "Upgrade (not live)"
1870 for tarball in $tarballs; do
1871 t32_test $tarball writeconf || rc=$?
1875 run_test 32b "Upgrade with writeconf"
1882 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
1884 for tarball in $tarballs; do
1885 dne_upgrade=yes t32_test $tarball writeconf || rc=$?
1889 run_test 32c "dne upgrade test"
1891 test_33a() { # bug 12333, was test_33
1893 local FSNAME2=test-123
1894 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
1897 [ -n "$ost1_HOST" ] && fs2ost_HOST=$ost1_HOST
1899 if [ -z "$fs2ost_DEV" -o -z "$fs2mds_DEV" ]; then
1900 local dev=${SINGLEMDS}_dev
1901 local MDSDEV=${!dev}
1902 is_blkdev $SINGLEMDS $MDSDEV && \
1903 skip_env "mixed loopback and real device not working" && return
1906 local fs2mdsdev=$(mdsdevname 1_2)
1907 local fs2ostdev=$(ostdevname 1_2)
1908 local fs2mdsvdev=$(mdsvdevname 1_2)
1909 local fs2ostvdev=$(ostvdevname 1_2)
1911 if [ $(facet_fstype mds1) == ldiskfs ]; then
1912 mkfsoptions="--mkfsoptions=\\\"-J size=8\\\"" # See bug 17931.
1915 add fs2mds $(mkfs_opts mds1 ${fs2mdsdev}) --mgs --fsname=${FSNAME2} \
1916 --reformat $mkfsoptions $fs2mdsdev $fs2mdsvdev || exit 10
1917 add fs2ost $(mkfs_opts ost1 ${fs2ostdev}) --mgsnode=$MGSNID \
1918 --fsname=${FSNAME2} --index=8191 --reformat $fs2ostdev \
1919 $fs2ostvdev || exit 10
1921 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS && trap cleanup_24a EXIT INT
1922 start fs2ost $fs2ostdev $OST_MOUNT_OPTS
1923 do_facet $SINGLEMDS "$LCTL conf_param $FSNAME2.sys.timeout=200" || rc=1
1925 mount -t lustre $MGSNID:/${FSNAME2} $MOUNT2 || rc=2
1928 cp /etc/hosts $MOUNT2/ || rc=3
1929 $LFS getstripe $MOUNT2/hosts
1934 cleanup_nocli || rc=6
1937 run_test 33a "Mount ost with a large index number"
1939 test_33b() { # was test_34
1942 do_facet client dd if=/dev/zero of=$MOUNT/24 bs=1024k count=1
1943 # Drop lock cancelation reply during umount
1944 #define OBD_FAIL_LDLM_CANCEL_NET 0x304
1945 do_facet client lctl set_param fail_loc=0x80000304
1946 #lctl set_param debug=-1
1947 umount_client $MOUNT
1950 run_test 33b "Drop cancel during umount"
1954 do_facet client "sh runmultiop_bg_pause $DIR/file O_c"
1955 manual_umount_client
1957 do_facet client killall -USR1 multiop
1958 if [ $rc -eq 0 ]; then
1959 error "umount not fail!"
1964 run_test 34a "umount with opened file should be fail"
1969 touch $DIR/$tfile || return 1
1970 stop_mds --force || return 2
1972 manual_umount_client --force
1974 if [ $rc -ne 0 ]; then
1975 error "mtab after failed umount - rc $rc"
1981 run_test 34b "force umount with failed mds should be normal"
1985 touch $DIR/$tfile || return 1
1986 stop_ost --force || return 2
1988 manual_umount_client --force
1990 if [ $rc -ne 0 ]; then
1991 error "mtab after failed umount - rc $rc"
1997 run_test 34c "force umount with failed ost should be normal"
1999 test_35a() { # bug 12459
2002 DBG_SAVE="`lctl get_param -n debug`"
2003 lctl set_param debug="ha"
2005 log "Set up a fake failnode for the MDS"
2007 local device=$(do_facet $SINGLEMDS "lctl get_param -n devices" | awk '($3 ~ "mdt" && $4 ~ "MDT") { print $4 }' | head -1)
2008 do_facet mgs "$LCTL conf_param ${device}.failover.node=" \
2009 "$(h2$NETTYPE $FAKENID)" || return 4
2011 log "Wait for RECONNECT_INTERVAL seconds (10s)"
2014 MSG="conf-sanity.sh test_35a `date +%F%kh%Mm%Ss`"
2017 log "Stopping the MDT: $device"
2018 stop_mdt 1 || return 5
2020 df $MOUNT > /dev/null 2>&1 &
2022 log "Restarting the MDT: $device"
2023 start_mdt 1 || return 6
2024 log "Wait for df ($DFPID) ... "
2027 lctl set_param debug="$DBG_SAVE"
2029 # retrieve from the log the first server that the client tried to
2030 # contact after the connection loss
2031 $LCTL dk $TMP/lustre-log-$TESTNAME.log
2032 NEXTCONN=`awk "/${MSG}/ {start = 1;}
2033 /import_select_connection.*$device-mdc.* using connection/ {
2035 if (\\\$NF ~ /$FAKENID/)
2041 }" $TMP/lustre-log-$TESTNAME.log`
2042 [ "$NEXTCONN" != "0" ] && log "The client didn't try to reconnect to the last active server (tried ${NEXTCONN} instead)" && return 7
2044 # remove nid settings
2045 writeconf_or_reformat
2047 run_test 35a "Reconnect to the last active server first"
2049 test_35b() { # bug 18674
2050 remote_mds || { skip "local MDS" && return 0; }
2054 $LCTL set_param debug="ha"
2056 MSG="conf-sanity.sh test_35b `date +%F%kh%Mm%Ss`"
2059 log "Set up a fake failnode for the MDS"
2061 local device=$(do_facet $SINGLEMDS "$LCTL get_param -n devices" | \
2062 awk '($3 ~ "mdt" && $4 ~ "MDT") { print $4 }' | head -1)
2063 do_facet mgs "$LCTL conf_param ${device}.failover.node=" \
2064 "$(h2$NETTYPE $FAKENID)" || return 1
2066 local at_max_saved=0
2067 # adaptive timeouts may prevent seeing the issue
2068 if at_is_enabled; then
2069 at_max_saved=$(at_max_get mds)
2070 at_max_set 0 mds client
2073 mkdir -p $MOUNT/$tdir
2075 log "Injecting EBUSY on MDS"
2076 # Setting OBD_FAIL_MDS_RESEND=0x136
2077 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0x80000136" || return 2
2079 $LCTL set_param mdc.${FSNAME}*.stats=clear
2081 log "Creating a test file and stat it"
2082 touch $MOUNT/$tdir/$tfile
2083 stat $MOUNT/$tdir/$tfile
2085 log "Stop injecting EBUSY on MDS"
2086 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0" || return 3
2087 rm -f $MOUNT/$tdir/$tfile
2090 # restore adaptive timeout
2091 [ $at_max_saved -ne 0 ] && at_max_set $at_max_saved mds client
2093 $LCTL dk $TMP/lustre-log-$TESTNAME.log
2095 CONNCNT=`$LCTL get_param mdc.${FSNAME}*.stats | awk '/mds_connect/{print $2}'`
2097 # retrieve from the log if the client has ever tried to
2098 # contact the fake server after the loss of connection
2099 FAILCONN=`awk "BEGIN {ret = 0;}
2100 /import_select_connection.*${FSNAME}-MDT0000-mdc.* using connection/ {
2102 if (\\\$NF ~ /$FAKENID/) {
2107 END {print ret}" $TMP/lustre-log-$TESTNAME.log`
2109 [ "$FAILCONN" == "0" ] && \
2110 log "ERROR: The client reconnection has not been triggered" && \
2112 [ "$FAILCONN" == "2" ] && \
2113 log "ERROR: The client tried to reconnect to the failover server while the primary was busy" && \
2117 # When OBD_FAIL_MDS_RESEND is hit, we sleep for 2 * obd_timeout
2118 # Reconnects are supposed to be rate limited to one every 5s
2119 [ $CONNCNT -gt $((2 * $TIMEOUT / 5 + 1)) ] && \
2120 log "ERROR: Too many reconnects $CONNCNT" && \
2124 # remove nid settings
2125 writeconf_or_reformat
2127 run_test 35b "Continue reconnection retries, if the active server is busy"
2130 [ $OSTCOUNT -lt 2 ] && skip_env "skipping test for single OST" && return
2132 [ "$ost_HOST" = "`hostname`" -o "$ost1_HOST" = "`hostname`" ] || \
2133 { skip "remote OST" && return 0; }
2136 local FSNAME2=test1234
2137 local fs3ost_HOST=$ost_HOST
2138 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
2140 [ -n "$ost1_HOST" ] && fs2ost_HOST=$ost1_HOST && fs3ost_HOST=$ost1_HOST
2142 if [ -z "$fs2ost_DEV" -o -z "$fs2mds_DEV" -o -z "$fs3ost_DEV" ]; then
2143 is_blkdev $SINGLEMDS $MDSDEV && \
2144 skip_env "mixed loopback and real device not working" && return
2147 local fs2mdsdev=$(mdsdevname 1_2)
2148 local fs2ostdev=$(ostdevname 1_2)
2149 local fs3ostdev=$(ostdevname 2_2)
2150 local fs2mdsvdev=$(mdsvdevname 1_2)
2151 local fs2ostvdev=$(ostvdevname 1_2)
2152 local fs3ostvdev=$(ostvdevname 2_2)
2154 add fs2mds $(mkfs_opts mds1 ${fs2mdsdev}) --mgs --fsname=${FSNAME2} \
2155 --reformat $fs2mdsdev $fs2mdsvdev || exit 10
2156 # XXX after we support non 4K disk blocksize in ldiskfs, specify a
2157 # different one than the default value here.
2158 add fs2ost $(mkfs_opts ost1 ${fs2ostdev}) --mgsnode=$MGSNID \
2159 --fsname=${FSNAME2} --reformat $fs2ostdev $fs2ostvdev || exit 10
2160 add fs3ost $(mkfs_opts ost1 ${fs3ostdev}) --mgsnode=$MGSNID \
2161 --fsname=${FSNAME2} --reformat $fs3ostdev $fs3ostvdev || exit 10
2163 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS
2164 start fs2ost $fs2ostdev $OST_MOUNT_OPTS
2165 start fs3ost $fs3ostdev $OST_MOUNT_OPTS
2167 mount -t lustre $MGSNID:/${FSNAME2} $MOUNT2 || return 1
2169 sleep 5 # until 11778 fixed
2171 dd if=/dev/zero of=$MOUNT2/$tfile bs=1M count=7 || return 2
2173 BKTOTAL=`lctl get_param -n obdfilter.*.kbytestotal | awk 'BEGIN{total=0}; {total+=$1}; END{print total}'`
2174 BKFREE=`lctl get_param -n obdfilter.*.kbytesfree | awk 'BEGIN{free=0}; {free+=$1}; END{print free}'`
2175 BKAVAIL=`lctl get_param -n obdfilter.*.kbytesavail | awk 'BEGIN{avail=0}; {avail+=$1}; END{print avail}'`
2176 STRING=`df -P $MOUNT2 | tail -n 1 | awk '{print $2","$3","$4}'`
2177 DFTOTAL=`echo $STRING | cut -d, -f1`
2178 DFUSED=`echo $STRING | cut -d, -f2`
2179 DFAVAIL=`echo $STRING | cut -d, -f3`
2180 DFFREE=$(($DFTOTAL - $DFUSED))
2182 ALLOWANCE=$((64 * $OSTCOUNT))
2184 if [ $DFTOTAL -lt $(($BKTOTAL - $ALLOWANCE)) ] ||
2185 [ $DFTOTAL -gt $(($BKTOTAL + $ALLOWANCE)) ] ; then
2186 echo "**** FAIL: df total($DFTOTAL) mismatch OST total($BKTOTAL)"
2189 if [ $DFFREE -lt $(($BKFREE - $ALLOWANCE)) ] ||
2190 [ $DFFREE -gt $(($BKFREE + $ALLOWANCE)) ] ; then
2191 echo "**** FAIL: df free($DFFREE) mismatch OST free($BKFREE)"
2194 if [ $DFAVAIL -lt $(($BKAVAIL - $ALLOWANCE)) ] ||
2195 [ $DFAVAIL -gt $(($BKAVAIL + $ALLOWANCE)) ] ; then
2196 echo "**** FAIL: df avail($DFAVAIL) mismatch OST avail($BKAVAIL)"
2201 stop fs3ost -f || return 200
2202 stop fs2ost -f || return 201
2203 stop fs2mds -f || return 202
2204 unload_modules_conf || return 203
2207 run_test 36 "df report consistency on OSTs with different block size"
2210 local mntpt=$(facet_mntpt $SINGLEMDS)
2211 local mdsdev=$(mdsdevname ${SINGLEMDS//mds/})
2212 local mdsdev_sym="$TMP/sym_mdt.img"
2213 local opts=$MDS_MOUNT_OPTS
2216 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
2217 skip "Currently only applicable to ldiskfs-based MDTs"
2221 echo "MDS : $mdsdev"
2222 echo "SYMLINK : $mdsdev_sym"
2223 do_facet $SINGLEMDS rm -f $mdsdev_sym
2225 do_facet $SINGLEMDS ln -s $mdsdev $mdsdev_sym
2227 echo "mount symlink device - $mdsdev_sym"
2229 if ! do_facet $SINGLEMDS test -b $mdsdev; then
2230 opts=$(csa_add "$opts" -o loop)
2232 mount_op=$(do_facet $SINGLEMDS mount -v -t lustre $opts \
2233 $mdsdev_sym $mntpt 2>&1)
2236 echo mount_op=$mount_op
2238 do_facet $SINGLEMDS "umount -d $mntpt && rm -f $mdsdev_sym"
2240 if $(echo $mount_op | grep -q "unable to set tunable"); then
2241 error "set tunables failed for symlink device"
2244 [ $rc -eq 0 ] || error "mount symlink $mdsdev_sym failed! rc=$rc"
2248 run_test 37 "verify set tunables works for symlink device"
2250 test_38() { # bug 14222
2251 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
2252 skip "Only applicable to ldiskfs-based MDTs"
2260 FILES=`find $SRC -type f -mtime +1 | head -n $COUNT`
2261 log "copying $(echo $FILES | wc -w) files to $DIR/$tdir"
2263 tar cf - $FILES | tar xf - -C $DIR/$tdir || \
2264 error "copying $SRC to $DIR/$tdir"
2266 umount_client $MOUNT
2268 log "rename lov_objid file on MDS"
2269 rm -f $TMP/lov_objid.orig
2271 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
2272 do_facet $SINGLEMDS "$DEBUGFS -c -R \\\"dump lov_objid $TMP/lov_objid.orig\\\" $MDSDEV"
2273 do_facet $SINGLEMDS "$DEBUGFS -w -R \\\"rm lov_objid\\\" $MDSDEV"
2275 do_facet $SINGLEMDS "od -Ax -td8 $TMP/lov_objid.orig"
2276 # check create in mds_lov_connect
2280 [ $V ] && log "verifying $DIR/$tdir/$f"
2281 diff -q $f $DIR/$tdir/$f || ERROR=y
2283 do_facet $SINGLEMDS "$DEBUGFS -c -R \\\"dump lov_objid $TMP/lov_objid.new\\\" $MDSDEV"
2284 do_facet $SINGLEMDS "od -Ax -td8 $TMP/lov_objid.new"
2285 [ "$ERROR" = "y" ] && error "old and new files are different after connect" || true
2287 # check it's updates in sync
2288 umount_client $MOUNT
2291 do_facet $SINGLEMDS dd if=/dev/zero of=$TMP/lov_objid.clear bs=4096 count=1
2292 do_facet $SINGLEMDS "$DEBUGFS -w -R \\\"rm lov_objid\\\" $MDSDEV"
2293 do_facet $SINGLEMDS "$DEBUGFS -w -R \\\"write $TMP/lov_objid.clear lov_objid\\\" $MDSDEV "
2298 [ $V ] && log "verifying $DIR/$tdir/$f"
2299 diff -q $f $DIR/$tdir/$f || ERROR=y
2301 do_facet $SINGLEMDS "$DEBUGFS -c -R \\\"dump lov_objid $TMP/lov_objid.new1\\\" $MDSDEV"
2302 do_facet $SINGLEMDS "od -Ax -td8 $TMP/lov_objid.new1"
2303 umount_client $MOUNT
2305 [ "$ERROR" = "y" ] && error "old and new files are different after sync" || true
2307 log "files compared the same"
2310 run_test 38 "MDS recreates missing lov_objid file from OST data"
2316 perl $SRCDIR/leak_finder.pl $TMP/debug 2>&1 | egrep '*** Leak:' &&
2317 error "memory leak detected" || true
2319 run_test 39 "leak_finder recognizes both LUSTRE and LNET malloc messages"
2321 test_40() { # bug 15759
2323 #define OBD_FAIL_TGT_TOOMANY_THREADS 0x706
2324 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0x80000706"
2328 run_test 40 "race during service thread startup"
2330 test_41a() { #bug 14134
2331 if [ $(facet_fstype $SINGLEMDS) == ldiskfs ] &&
2332 ! do_facet $SINGLEMDS test -b $(mdsdevname 1); then
2333 skip "Loop devices does not work with nosvc option"
2338 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
2340 start $SINGLEMDS $MDSDEV $MDS_MOUNT_OPTS -o nosvc -n
2341 start ost1 `ostdevname 1` $OST_MOUNT_OPTS
2342 start $SINGLEMDS $MDSDEV $MDS_MOUNT_OPTS -o nomgs,force
2344 mount_client $MOUNT || return 1
2347 echo "blah blah" > $MOUNT/$tfile
2350 umount_client $MOUNT
2351 stop ost1 -f || return 201
2352 stop_mds -f || return 202
2353 stop_mds -f || return 203
2354 unload_modules_conf || return 204
2357 run_test 41a "mount mds with --nosvc and --nomgs"
2360 if [ $(facet_fstype $SINGLEMDS) == ldiskfs ] &&
2361 ! do_facet $SINGLEMDS test -b $(mdsdevname 1); then
2362 skip "Loop devices does not work with nosvc option"
2366 ! combined_mgs_mds && skip "needs combined mgs device" && return 0
2370 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
2372 start $SINGLEMDS $MDSDEV $MDS_MOUNT_OPTS -o nosvc -n
2374 start $SINGLEMDS $MDSDEV $MDS_MOUNT_OPTS -o nomgs,force
2376 mount_client $MOUNT || return 1
2379 echo "blah blah" > $MOUNT/$tfile
2380 cat $MOUNT/$tfile || return 200
2382 umount_client $MOUNT
2383 stop_ost || return 201
2384 stop_mds -f || return 202
2385 stop_mds -f || return 203
2388 run_test 41b "mount mds with --nosvc and --nomgs on first mount"
2390 test_42() { #bug 14693
2392 check_mount || error "client was not mounted"
2394 do_facet mgs $LCTL conf_param $FSNAME.llite.some_wrong_param=10
2395 umount_client $MOUNT ||
2396 error "unmounting client failed with invalid llite param"
2397 mount_client $MOUNT ||
2398 error "mounting client failed with invalid llite param"
2400 do_facet mgs $LCTL conf_param $FSNAME.sys.some_wrong_param=20
2401 cleanup || error "stopping $FSNAME failed with invalid sys param"
2404 check_mount || "client was not mounted with invalid sys param"
2405 cleanup || error "stopping $FSNAME failed with invalid sys param"
2408 run_test 42 "allow client/server mount/unmount with invalid config param"
2411 [ $UID -ne 0 -o $RUNAS_ID -eq 0 ] && skip_env "run as root"
2413 chmod ugo+x $DIR || error "chmod 0 failed"
2414 set_conf_param_and_check mds \
2415 "lctl get_param -n mdt.$FSNAME-MDT0000.root_squash" \
2416 "$FSNAME.mdt.root_squash" \
2418 set_conf_param_and_check mds \
2419 "lctl get_param -n mdt.$FSNAME-MDT0000.nosquash_nids" \
2420 "$FSNAME.mdt.nosquash_nids" \
2424 # create set of test files
2426 echo "111" > $DIR/$tfile-userfile || error "write 1 failed"
2427 chmod go-rw $DIR/$tfile-userfile || error "chmod 1 failed"
2428 chown $RUNAS_ID.$RUNAS_ID $DIR/$tfile-userfile || error "chown failed"
2430 echo "222" > $DIR/$tfile-rootfile || error "write 2 failed"
2431 chmod go-rw $DIR/$tfile-rootfile || error "chmod 2 faield"
2433 mkdir $DIR/$tdir-rootdir -p || error "mkdir failed"
2434 chmod go-rwx $DIR/$tdir-rootdir || error "chmod 3 failed"
2435 touch $DIR/$tdir-rootdir/tfile-1 || error "touch failed"
2438 # check root_squash:
2439 # set root squash UID:GID to RUNAS_ID
2440 # root should be able to access only files owned by RUNAS_ID
2442 set_conf_param_and_check mds \
2443 "lctl get_param -n mdt.$FSNAME-MDT0000.root_squash" \
2444 "$FSNAME.mdt.root_squash" \
2445 "$RUNAS_ID:$RUNAS_ID"
2447 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-userfile)
2448 dd if=$DIR/$tfile-userfile 1>/dev/null 2>/dev/null || \
2449 error "$ST: root read permission is denied"
2450 echo "$ST: root read permission is granted - ok"
2453 dd conv=notrunc if=$DIR/$tfile-userfile 1>/dev/null 2>/dev/null || \
2454 error "$ST: root write permission is denied"
2455 echo "$ST: root write permission is granted - ok"
2457 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-rootfile)
2458 dd if=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null && \
2459 error "$ST: root read permission is granted"
2460 echo "$ST: root read permission is denied - ok"
2463 dd conv=notrunc of=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null && \
2464 error "$ST: root write permission is granted"
2465 echo "$ST: root write permission is denied - ok"
2467 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tdir-rootdir)
2468 rm $DIR/$tdir-rootdir/tfile-1 1>/dev/null 2>/dev/null && \
2469 error "$ST: root unlink permission is granted"
2470 echo "$ST: root unlink permission is denied - ok"
2472 touch $DIR/tdir-rootdir/tfile-2 1>/dev/null 2>/dev/null && \
2473 error "$ST: root create permission is granted"
2474 echo "$ST: root create permission is denied - ok"
2477 # check nosquash_nids:
2478 # put client's NID into nosquash_nids list,
2479 # root should be able to access root file after that
2481 local NIDLIST=$(lctl list_nids all | tr '\n' ' ')
2482 NIDLIST="2@elan $NIDLIST 192.168.0.[2,10]@tcp"
2483 NIDLIST=$(echo $NIDLIST | tr -s ' ' ' ')
2484 set_conf_param_and_check mds \
2485 "lctl get_param -n mdt.$FSNAME-MDT0000.nosquash_nids" \
2486 "$FSNAME-MDTall.mdt.nosquash_nids" \
2489 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-rootfile)
2490 dd if=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null || \
2491 error "$ST: root read permission is denied"
2492 echo "$ST: root read permission is granted - ok"
2495 dd conv=notrunc of=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null || \
2496 error "$ST: root write permission is denied"
2497 echo "$ST: root write permission is granted - ok"
2499 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tdir-rootdir)
2500 rm $DIR/$tdir-rootdir/tfile-1 || \
2501 error "$ST: root unlink permission is denied"
2502 echo "$ST: root unlink permission is granted - ok"
2503 touch $DIR/$tdir-rootdir/tfile-2 || \
2504 error "$ST: root create permission is denied"
2505 echo "$ST: root create permission is granted - ok"
2509 run_test 43 "check root_squash and nosquash_nids"
2511 umount_client $MOUNT
2516 check_mount || return 2
2517 UUID=$($LCTL get_param llite.${FSNAME}*.uuid | cut -d= -f2)
2519 UUIDS=$(do_facet $SINGLEMDS "$LCTL get_param mdt.${FSNAME}*.exports.*.uuid")
2520 for VAL in $UUIDS; do
2521 NID=$(echo $VAL | cut -d= -f1)
2522 CLUUID=$(echo $VAL | cut -d= -f2)
2523 [ "$UUID" = "$CLUUID" ] && STATS_FOUND=yes && break
2525 [ "$STATS_FOUND" = "no" ] && error "stats not found for client"
2529 run_test 44 "mounted client proc entry exists"
2533 check_mount || return 2
2538 #define OBD_FAIL_PTLRPC_LONG_UNLINK 0x50f
2539 do_facet client "lctl set_param fail_loc=0x50f"
2542 manual_umount_client --force || return 3
2543 do_facet client "lctl set_param fail_loc=0x0"
2545 mount_client $MOUNT || return 4
2549 run_test 45 "long unlink handling in ptlrpcd"
2556 umount_client $MOUNT2 || rc=$?
2557 umount_client $MOUNT || rc=$?
2558 while [ $count -gt 0 ]; do
2559 stop ost${count} -f || rc=$?
2563 cleanup_nocli || rc=$?
2564 #writeconf to remove all ost2 traces for subsequent tests
2565 writeconf_or_reformat
2570 echo "Testing with $OSTCOUNT OSTs"
2572 start_mds || return 1
2573 #first client should see only one ost
2574 start_ost || return 2
2575 wait_osc_import_state mds ost FULL
2577 mount_client $MOUNT || return 3
2578 trap "cleanup_46a $OSTCOUNT" EXIT ERR
2581 for (( i=2; i<=$OSTCOUNT; i++ )); do
2582 start ost$i `ostdevname $i` $OST_MOUNT_OPTS || return $((i+2))
2585 # wait until osts in sync
2586 for (( i=2; i<=$OSTCOUNT; i++ )); do
2587 wait_osc_import_state mds ost$i FULL
2588 wait_osc_import_state client ost$i FULL
2591 #second client see all ost's
2593 mount_client $MOUNT2 || return 8
2594 $LFS setstripe -c -1 $MOUNT2 || return 9
2595 $LFS getstripe $MOUNT2 || return 10
2597 echo "ok" > $MOUNT2/widestripe
2598 $LFS getstripe $MOUNT2/widestripe || return 11
2599 # fill acl buffer for avoid expand lsm to them
2600 awk -F : '{if (FNR < 25) { print "u:"$1":rwx" }}' /etc/passwd | while read acl; do
2601 setfacl -m $acl $MOUNT2/widestripe
2605 stat $MOUNT/widestripe || return 12
2607 cleanup_46a $OSTCOUNT || { echo "cleanup_46a failed!" && return 13; }
2610 run_test 46a "handle ost additional - wide striped file"
2615 check_mount || return 2
2616 $LCTL set_param ldlm.namespaces.$FSNAME-*-*-*.lru_size=100
2620 for ns in $($LCTL get_param ldlm.namespaces.$FSNAME-*-*-*.lru_size); do
2621 if echo $ns | grep "MDT[[:digit:]]*"; then
2624 lrs=$(echo $ns | sed 's/.*lru_size=//')
2625 lru_size[count]=$lrs
2630 facet_failover $SINGLEMDS
2631 client_up || return 3
2634 for ns in $($LCTL get_param ldlm.namespaces.$FSNAME-*-*-*.lru_size); do
2635 if echo $ns | grep "MDT[[:digit:]]*"; then
2638 lrs=$(echo $ns | sed 's/.*lru_size=//')
2639 if ! test "$lrs" -eq "${lru_size[count]}"; then
2640 n=$(echo $ns | sed -e 's/ldlm.namespaces.//' -e 's/.lru_size=.*//')
2641 error "$n has lost lru_size: $lrs vs. ${lru_size[count]}"
2649 run_test 47 "server restart does not make client loss lru_resize settings"
2654 # reformat after this test is needed - if test will failed
2655 # we will have unkillable file at FS
2659 test_48() { # bug 17636
2662 check_mount || return 2
2664 $LFS setstripe -c -1 $MOUNT || return 9
2665 $LFS getstripe $MOUNT || return 10
2667 echo "ok" > $MOUNT/widestripe
2668 $LFS getstripe $MOUNT/widestripe || return 11
2670 trap cleanup_48 EXIT ERR
2672 # fill acl buffer for avoid expand lsm to them
2673 getent passwd | awk -F : '{ print "u:"$1":rwx" }' | while read acl; do
2674 setfacl -m $acl $MOUNT/widestripe
2677 stat $MOUNT/widestripe || return 12
2682 run_test 48 "too many acls on file"
2684 # check PARAM_SYS_LDLM_TIMEOUT option of MKFS.LUSTRE
2685 test_49() { # bug 17710
2686 local timeout_orig=$TIMEOUT
2687 local ldlm_timeout_orig=$LDLM_TIMEOUT
2688 local LOCAL_TIMEOUT=20
2690 LDLM_TIMEOUT=$LOCAL_TIMEOUT
2691 TIMEOUT=$LOCAL_TIMEOUT
2695 check_mount || return 1
2697 echo "check ldlm_timout..."
2698 LDLM_MDS="`do_facet $SINGLEMDS lctl get_param -n ldlm_timeout`"
2699 LDLM_OST1="`do_facet ost1 lctl get_param -n ldlm_timeout`"
2700 LDLM_CLIENT="`do_facet client lctl get_param -n ldlm_timeout`"
2702 if [ $LDLM_MDS -ne $LDLM_OST1 ] || [ $LDLM_MDS -ne $LDLM_CLIENT ]; then
2703 error "Different LDLM_TIMEOUT:$LDLM_MDS $LDLM_OST1 $LDLM_CLIENT"
2706 if [ $LDLM_MDS -ne $((LOCAL_TIMEOUT / 3)) ]; then
2707 error "LDLM_TIMEOUT($LDLM_MDS) is not correct"
2710 umount_client $MOUNT
2711 stop_ost || return 2
2712 stop_mds || return 3
2714 LDLM_TIMEOUT=$((LOCAL_TIMEOUT - 1))
2718 check_mount || return 7
2720 LDLM_MDS="`do_facet $SINGLEMDS lctl get_param -n ldlm_timeout`"
2721 LDLM_OST1="`do_facet ost1 lctl get_param -n ldlm_timeout`"
2722 LDLM_CLIENT="`do_facet client lctl get_param -n ldlm_timeout`"
2724 if [ $LDLM_MDS -ne $LDLM_OST1 ] || [ $LDLM_MDS -ne $LDLM_CLIENT ]; then
2725 error "Different LDLM_TIMEOUT:$LDLM_MDS $LDLM_OST1 $LDLM_CLIENT"
2728 if [ $LDLM_MDS -ne $((LOCAL_TIMEOUT - 1)) ]; then
2729 error "LDLM_TIMEOUT($LDLM_MDS) is not correct"
2732 cleanup || return $?
2734 LDLM_TIMEOUT=$ldlm_timeout_orig
2735 TIMEOUT=$timeout_orig
2737 run_test 49 "check PARAM_SYS_LDLM_TIMEOUT option of MKFS.LUSTRE"
2740 # Test both statfs and lfs df and fail if either one fails
2741 multiop_bg_pause $1 f_
2744 killall -USR1 multiop
2745 [ $RC1 -ne 0 ] && log "lazystatfs multiop failed"
2746 wait $PID || { RC1=$?; log "multiop return error "; }
2753 if [ $RC2 -eq 0 ]; then
2755 log "lazystatfs df failed"
2759 [[ $RC1 -ne 0 || $RC2 -eq 0 ]] && RC=1
2765 lctl set_param llite.$FSNAME-*.lazystatfs=1
2768 lazystatfs $MOUNT || error "lazystatfs failed but no down servers"
2770 cleanup || return $?
2772 run_test 50a "lazystatfs all servers available =========================="
2776 lctl set_param llite.$FSNAME-*.lazystatfs=1
2779 # Wait for client to detect down OST
2780 stop_ost || error "Unable to stop OST1"
2781 wait_osc_import_state mds ost DISCONN
2783 lazystatfs $MOUNT || error "lazystatfs should don't have returned EIO"
2785 umount_client $MOUNT || error "Unable to unmount client"
2786 stop_mds || error "Unable to stop MDS"
2788 run_test 50b "lazystatfs all servers down =========================="
2791 start_mds || error "Unable to start MDS"
2792 start_ost || error "Unable to start OST1"
2793 start_ost2 || error "Unable to start OST2"
2794 mount_client $MOUNT || error "Unable to mount client"
2795 lctl set_param llite.$FSNAME-*.lazystatfs=1
2798 # Wait for client to detect down OST
2799 stop_ost || error "Unable to stop OST1"
2800 wait_osc_import_state mds ost DISCONN
2801 lazystatfs $MOUNT || error "lazystatfs failed with one down server"
2803 umount_client $MOUNT || error "Unable to unmount client"
2804 stop_ost2 || error "Unable to stop OST2"
2805 stop_mds || error "Unable to stop MDS"
2806 #writeconf to remove all ost2 traces for subsequent tests
2807 writeconf_or_reformat
2809 run_test 50c "lazystatfs one server down =========================="
2812 start_mds || error "Unable to start MDS"
2813 start_ost || error "Unable to start OST1"
2814 start_ost2 || error "Unable to start OST2"
2815 mount_client $MOUNT || error "Unable to mount client"
2816 lctl set_param llite.$FSNAME-*.lazystatfs=1
2819 # Issue the statfs during the window where the client still
2820 # belives the OST to be available but it is in fact down.
2821 # No failure just a statfs which hangs for a timeout interval.
2822 stop_ost || error "Unable to stop OST1"
2823 lazystatfs $MOUNT || error "lazystatfs failed with one down server"
2825 umount_client $MOUNT || error "Unable to unmount client"
2826 stop_ost2 || error "Unable to stop OST2"
2827 stop_mds || error "Unable to stop MDS"
2828 #writeconf to remove all ost2 traces for subsequent tests
2829 writeconf_or_reformat
2831 run_test 50d "lazystatfs client/server conn race =========================="
2838 start_mds || return 1
2839 #first client should see only one ost
2840 start_ost || return 2
2841 wait_osc_import_state mds ost FULL
2843 # Wait for client to detect down OST
2844 stop_ost || error "Unable to stop OST1"
2845 wait_osc_import_state mds ost DISCONN
2847 mount_client $MOUNT || error "Unable to mount client"
2848 lctl set_param llite.$FSNAME-*.lazystatfs=0
2850 multiop_bg_pause $MOUNT _f
2854 if [ $RC1 -ne 0 ]; then
2855 log "multiop failed $RC1"
2858 sleep $(( $TIMEOUT+1 ))
2860 [ $? -ne 0 ] && error "process isn't sleep"
2861 start_ost || error "Unable to start OST1"
2862 wait $pid || error "statfs failed"
2865 umount_client $MOUNT || error "Unable to unmount client"
2866 stop_ost || error "Unable to stop OST1"
2867 stop_mds || error "Unable to stop MDS"
2869 run_test 50e "normal statfs all servers down =========================="
2874 CONN_PROC="osc.$FSNAME-OST0001-osc-[M]*.ost_server_uuid"
2876 start_mds || error "Unable to start mds"
2877 #first client should see only one ost
2878 start_ost || error "Unable to start OST1"
2879 wait_osc_import_state mds ost FULL
2881 start_ost2 || error "Unable to start OST2"
2882 wait_osc_import_state mds ost2 FULL
2884 # Wait for client to detect down OST
2885 stop_ost2 || error "Unable to stop OST2"
2887 wait_osc_import_state mds ost2 DISCONN
2888 mount_client $MOUNT || error "Unable to mount client"
2889 lctl set_param llite.$FSNAME-*.lazystatfs=0
2891 multiop_bg_pause $MOUNT _f
2895 if [ $RC1 -ne 0 ]; then
2896 log "lazystatfs multiop failed $RC1"
2899 sleep $(( $TIMEOUT+1 ))
2901 [ $? -ne 0 ] && error "process isn't sleep"
2902 start_ost2 || error "Unable to start OST2"
2903 wait $pid || error "statfs failed"
2904 stop_ost2 || error "Unable to stop OST2"
2907 umount_client $MOUNT || error "Unable to unmount client"
2908 stop_ost || error "Unable to stop OST1"
2909 stop_mds || error "Unable to stop MDS"
2910 #writeconf to remove all ost2 traces for subsequent tests
2911 writeconf_or_reformat
2913 run_test 50f "normal statfs one server in down =========================="
2916 [ "$OSTCOUNT" -lt "2" ] && skip_env "$OSTCOUNT < 2, skipping" && return
2918 start_ost2 || error "Unable to start OST2"
2919 wait_osc_import_state mds ost2 FULL
2920 wait_osc_import_state client ost2 FULL
2922 local PARAM="${FSNAME}-OST0001.osc.active"
2924 $LFS setstripe -c -1 $DIR/$tfile || error "Unable to lfs setstripe"
2925 do_facet mgs $LCTL conf_param $PARAM=0 || error "Unable to deactivate OST"
2927 umount_client $MOUNT || error "Unable to unmount client"
2928 mount_client $MOUNT || error "Unable to mount client"
2929 # This df should not cause a panic
2932 do_facet mgs $LCTL conf_param $PARAM=1 || error "Unable to activate OST"
2934 umount_client $MOUNT || error "Unable to unmount client"
2935 stop_ost2 || error "Unable to stop OST2"
2936 stop_ost || error "Unable to stop OST1"
2937 stop_mds || error "Unable to stop MDS"
2938 #writeconf to remove all ost2 traces for subsequent tests
2939 writeconf_or_reformat
2941 run_test 50g "deactivated OST should not cause panic====================="
2945 # prepare MDT/OST, make OSC inactive for OST1
2946 [ "$OSTCOUNT" -lt "2" ] && skip_env "$OSTCOUNT < 2, skipping" && return
2947 do_facet ost1 "$TUNEFS --param osc.active=0 `ostdevname 1`" ||
2948 error "tunefs OST1 failed"
2949 start_mds || error "Unable to start MDT"
2950 start_ost || error "Unable to start OST1"
2951 start_ost2 || error "Unable to start OST2"
2952 mount_client $MOUNT || error "client start failed"
2956 # activatate OSC for OST1
2957 local TEST="$LCTL get_param -n osc.${FSNAME}-OST0000-osc-[!M]*.active"
2958 set_conf_param_and_check client \
2959 "$TEST" "${FSNAME}-OST0000.osc.active" 1 ||
2960 error "Unable to activate OST1"
2962 mkdir -p $DIR/$tdir/2
2963 $LFS setstripe -c -1 -i 0 $DIR/$tdir/2
2964 sleep 1 && echo "create a file after OST1 is activated"
2966 createmany -o $DIR/$tdir/2/$tfile-%d 1
2968 # check OSC import is working
2969 stat $DIR/$tdir/2/* >/dev/null 2>&1 ||
2970 error "some OSC imports are still not connected"
2973 umount_client $MOUNT || error "Unable to umount client"
2974 stop_ost2 || error "Unable to stop OST2"
2977 run_test 50h "LU-642: activate deactivated OST ==="
2980 local LOCAL_TIMEOUT=20
2984 check_mount || return 1
2987 $LFS setstripe -c -1 $MOUNT/d1
2988 #define OBD_FAIL_MDS_REINT_DELAY 0x142
2989 do_facet $SINGLEMDS "lctl set_param fail_loc=0x142"
2990 touch $MOUNT/d1/f1 &
2993 start_ost2 || return 2
2995 stop_ost2 || return 3
2997 #writeconf to remove all ost2 traces for subsequent tests
2998 writeconf_or_reformat
3000 run_test 51 "Verify that mdt_reint handles RMF_MDT_MD correctly when an OST is added"
3009 do_node $node mkdir -p $dest
3010 [ $? -eq 0 ] || { error "Unable to create directory"; return 1; }
3012 do_node $node 'tar cf - '$@' | tar xf - -C '$dest';
3013 [ \"\${PIPESTATUS[*]}\" = \"0 0\" ] || exit 1'
3014 [ $? -eq 0 ] || { error "Unable to tar files"; return 2; }
3016 do_node $node 'getfattr -d -m "[a-z]*\\." '$@' > '$xattrs
3017 [ $? -eq 0 ] || { error "Unable to read xattrs"; return 3; }
3027 local backup2=${TMP}/backup2
3029 do_node $node mkdir -p $backup2
3030 [ $? -eq 0 ] || { error "Unable to create directory"; return 1; }
3032 do_node $node 'tar cf - '$@' | tar xf - -C '$backup2';
3033 [ \"\${PIPESTATUS[*]}\" = \"0 0\" ] || exit 1'
3034 [ $? -eq 0 ] || { error "Unable to tar files to diff"; return 2; }
3036 do_node $node "diff -rq $backup $backup2"
3037 [ $? -eq 0 ] || { error "contents differ"; return 3; }
3039 local xattrs2=${TMP}/xattrs2
3040 do_node $node 'getfattr -d -m "[a-z]*\\." '$@' > '$xattrs2
3041 [ $? -eq 0 ] || { error "Unable to read xattrs to diff"; return 4; }
3043 do_node $node "diff $xattrs $xattrs2"
3044 [ $? -eq 0 ] || { error "xattrs differ"; return 5; }
3046 do_node $node "rm -rf $backup2 $xattrs2"
3047 [ $? -eq 0 ] || { error "Unable to delete temporary files"; return 6; }
3051 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
3052 skip "Only applicable to ldiskfs-based MDTs"
3057 [ $? -eq 0 ] || { error "Unable to start MDS"; return 1; }
3059 [ $? -eq 0 ] || { error "Unable to start OST1"; return 2; }
3061 [ $? -eq 0 ] || { error "Unable to mount client"; return 3; }
3064 local ost1mnt=$(facet_mntpt ost1)
3065 local ost1node=$(facet_active_host ost1)
3066 local ost1tmp=$TMP/conf52
3070 [ $? -eq 0 ] || { error "Unable to create tdir"; return 4; }
3071 touch $TMP/modified_first
3072 [ $? -eq 0 ] || { error "Unable to create temporary file"; return 5; }
3073 local mtime=$(stat -c %Y $TMP/modified_first)
3074 do_node $ost1node "mkdir -p $ost1tmp && touch -m -d @$mtime $ost1tmp/modified_first"
3076 [ $? -eq 0 ] || { error "Unable to create temporary file"; return 6; }
3079 $LFS setstripe -c -1 -S 1M $DIR/$tdir
3080 [ $? -eq 0 ] || { error "lfs setstripe failed"; return 7; }
3082 for (( i=0; i < nrfiles; i++ )); do
3083 multiop $DIR/$tdir/$tfile-$i Ow1048576w1048576w524288c
3084 [ $? -eq 0 ] || { error "multiop failed"; return 8; }
3090 echo backup files to $TMP/files
3091 local files=$(find $DIR/$tdir -type f -newer $TMP/modified_first)
3092 copy_files_xattrs `hostname` $TMP/files $TMP/file_xattrs $files
3093 [ $? -eq 0 ] || { error "Unable to copy files"; return 9; }
3095 umount_client $MOUNT
3096 [ $? -eq 0 ] || { error "Unable to umount client"; return 10; }
3098 [ $? -eq 0 ] || { error "Unable to stop ost1"; return 11; }
3100 echo mount ost1 as ldiskfs
3101 do_node $ost1node mkdir -p $ost1mnt
3102 [ $? -eq 0 ] || { error "Unable to create $ost1mnt"; return 23; }
3103 if ! do_node $ost1node test -b $ost1_dev; then
3106 do_node $ost1node mount -t $(facet_fstype ost1) $loop $ost1_dev \
3108 [ $? -eq 0 ] || { error "Unable to mount ost1 as ldiskfs"; return 12; }
3111 echo backup objects to $ost1tmp/objects
3112 local objects=$(do_node $ost1node 'find '$ost1mnt'/O/[0-9]* -type f'\
3113 '-size +0 -newer '$ost1tmp'/modified_first -regex ".*\/[0-9]+"')
3114 copy_files_xattrs $ost1node $ost1tmp/objects $ost1tmp/object_xattrs \
3116 [ $? -eq 0 ] || { error "Unable to copy objects"; return 13; }
3118 # move objects to lost+found
3119 do_node $ost1node 'mv '$objects' '${ost1mnt}'/lost+found'
3120 [ $? -eq 0 ] || { error "Unable to move objects"; return 14; }
3123 do_node $ost1node "ll_recover_lost_found_objs -d $ost1mnt/lost+found"
3124 [ $? -eq 0 ] || { error "ll_recover_lost_found_objs failed"; return 15; }
3126 # compare restored objects against saved ones
3127 diff_files_xattrs $ost1node $ost1tmp/objects $ost1tmp/object_xattrs $objects
3128 [ $? -eq 0 ] || { error "Unable to diff objects"; return 16; }
3130 do_node $ost1node "umount $ost1mnt"
3131 [ $? -eq 0 ] || { error "Unable to umount ost1 as ldiskfs"; return 17; }
3134 [ $? -eq 0 ] || { error "Unable to start ost1"; return 18; }
3136 [ $? -eq 0 ] || { error "Unable to mount client"; return 19; }
3139 diff_files_xattrs `hostname` $TMP/files $TMP/file_xattrs $files
3140 [ $? -eq 0 ] || { error "Unable to diff files"; return 20; }
3142 rm -rf $TMP/files $TMP/file_xattrs
3143 [ $? -eq 0 ] || { error "Unable to delete temporary files"; return 21; }
3144 do_node $ost1node "rm -rf $ost1tmp"
3145 [ $? -eq 0 ] || { error "Unable to delete temporary files"; return 22; }
3148 run_test 52 "check recovering objects from lost+found"
3150 # Checks threads_min/max/started for some service
3152 # Arguments: service name (OST or MDT), facet (e.g., ost1, $SINGLEMDS), and a
3153 # parameter pattern prefix like 'ost.*.ost'.
3166 local msg="Insane $modname thread counts"
3167 local ncpts=$(check_cpt_number $facet)
3171 check_mount || return 41
3173 # We need to expand $parampat, but it may match multiple parameters, so
3174 # we'll pick the first one
3175 if ! paramp=$(do_facet $facet "lctl get_param -N ${parampat}.threads_min"|head -1); then
3176 error "Couldn't expand ${parampat}.threads_min parameter name"
3180 # Remove the .threads_min part
3181 paramp=${paramp%.threads_min}
3183 # Check for sanity in defaults
3184 tmin=$(do_facet $facet "lctl get_param -n ${paramp}.threads_min" || echo 0)
3185 tmax=$(do_facet $facet "lctl get_param -n ${paramp}.threads_max" || echo 0)
3186 tstarted=$(do_facet $facet "lctl get_param -n ${paramp}.threads_started" || echo 0)
3187 lassert 23 "$msg (PDSH problems?)" '(($tstarted && $tmin && $tmax))' || return $?
3188 lassert 24 "$msg" '(($tstarted >= $tmin && $tstarted <= $tmax ))' || return $?
3189 nthrs=$(expr $tmax - $tmin)
3190 if [ $nthrs -lt $ncpts ]; then
3196 [ $tmin -eq $tmax -a $tmin -eq $tstarted ] &&
3197 skip_env "module parameter forced $facet thread count" &&
3198 tmin=3 && tmax=$((3 * tmax))
3200 # Check that we can change min/max
3201 do_facet $facet "lctl set_param ${paramp}.threads_min=$((tmin + nthrs))"
3202 do_facet $facet "lctl set_param ${paramp}.threads_max=$((tmax - nthrs))"
3203 tmin2=$(do_facet $facet "lctl get_param -n ${paramp}.threads_min" || echo 0)
3204 tmax2=$(do_facet $facet "lctl get_param -n ${paramp}.threads_max" || echo 0)
3205 lassert 25 "$msg" '(($tmin2 == ($tmin + $nthrs) && $tmax2 == ($tmax - $nthrs)))' || return $?
3207 # Check that we can set min/max to the same value
3208 tmin=$(do_facet $facet "lctl get_param -n ${paramp}.threads_min" || echo 0)
3209 do_facet $facet "lctl set_param ${paramp}.threads_max=$tmin"
3210 tmin2=$(do_facet $facet "lctl get_param -n ${paramp}.threads_min" || echo 0)
3211 tmax2=$(do_facet $facet "lctl get_param -n ${paramp}.threads_max" || echo 0)
3212 lassert 26 "$msg" '(($tmin2 == $tmin && $tmax2 == $tmin))' || return $?
3214 # Check that we can't set max < min
3215 do_facet $facet "lctl set_param ${paramp}.threads_max=$((tmin - 1))"
3216 tmin2=$(do_facet $facet "lctl get_param -n ${paramp}.threads_min" || echo 0)
3217 tmax2=$(do_facet $facet "lctl get_param -n ${paramp}.threads_max" || echo 0)
3218 lassert 27 "$msg" '(($tmin2 <= $tmax2))' || return $?
3220 # We need to ensure that we get the module options desired; to do this
3221 # we set LOAD_MODULES_REMOTE=true and we call setmodopts below.
3222 LOAD_MODULES_REMOTE=true
3225 local newvalue="${opts}=$(expr $basethr \* $ncpts)"
3226 setmodopts -a $modname "$newvalue" oldvalue
3230 check_mount || return 41
3232 # Restore previous setting of MODOPTS_*
3233 setmodopts $modname "$oldvalue"
3235 # Check that $opts took
3236 tmin=$(do_facet $facet "lctl get_param -n ${paramp}.threads_min")
3237 tmax=$(do_facet $facet "lctl get_param -n ${paramp}.threads_max")
3238 tstarted=$(do_facet $facet "lctl get_param -n ${paramp}.threads_started")
3239 lassert 28 "$msg" '(($tstarted == $tmin && $tstarted == $tmax ))' || return $?
3248 thread_sanity OST ost1 'ost.*.ost' 'oss_num_threads' '16'
3251 run_test 53a "check OSS thread count params"
3255 local mds=$(do_facet $SINGLEMDS "lctl get_param -N mds.*.*.threads_max \
3257 if [ -z "$mds" ]; then
3258 #running this on an old MDT
3259 thread_sanity MDT $SINGLEMDS 'mdt.*.*.' 'mdt_num_threads' 16
3261 thread_sanity MDT $SINGLEMDS 'mds.*.*.' 'mds_num_threads' 16
3265 run_test 53b "check MDS thread count params"
3268 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
3269 skip "Only applicable to ldiskfs-based MDTs"
3273 do_rpc_nodes $(facet_host ost1) run_llverdev $(ostdevname 1) -p
3274 [ $? -eq 0 ] || error "llverdev failed!"
3277 run_test 54a "test llverdev and partial verify of device"
3280 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
3281 skip "Only applicable to ldiskfs-based MDTs"
3286 run_llverfs $MOUNT -p
3287 [ $? -eq 0 ] || error "llverfs failed!"
3290 run_test 54b "test llverfs and partial verify of filesystem"
3294 local max_ost_index=$1
3295 echo -n $(((max_ost_index + 1) * 8))
3299 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
3300 skip "Only applicable to ldiskfs-based MDTs"
3304 local mdsdev=$(mdsdevname 1)
3305 local mdsvdev=$(mdsvdevname 1)
3309 add mds1 $(mkfs_opts mds1 ${mdsdev}) --reformat $mdsdev \
3311 add ost1 $(mkfs_opts ost1 $(ostdevname 1)) --index=$i \
3312 --reformat $(ostdevname 1) $(ostvdevname 1)
3318 echo checking size of lov_objid for ost index $i
3319 LOV_OBJID_SIZE=$(do_facet mds1 "$DEBUGFS -R 'stat lov_objid' $mdsdev 2>/dev/null" | grep ^User | awk '{print $6}')
3320 if [ "$LOV_OBJID_SIZE" != $(lov_objid_size $i) ]; then
3321 error "lov_objid size has to be $(lov_objid_size $i), not $LOV_OBJID_SIZE"
3323 echo ok, lov_objid size is correct: $LOV_OBJID_SIZE
3330 run_test 55 "check lov_objid size"
3333 local mds_journal_size_orig=$MDSJOURNALSIZE
3337 for num in $(seq 1 $MDSCOUNT); do
3338 add mds${num} $(mkfs_opts mds${num} $(mdsdevname $num)) \
3339 --reformat $(mdsdevname $num) $(mdsvdevname $num)
3341 add ost1 $(mkfs_opts ost1 $(ostdevname 1)) --index=1000 --reformat \
3342 $(ostdevname 1) $(ostvdevname 1)
3343 add ost2 $(mkfs_opts ost2 $(ostdevname 2)) --index=10000 --reformat \
3344 $(ostdevname 2) $(ostvdevname 2)
3348 start_ost2 || error "Unable to start second ost"
3349 mount_client $MOUNT || error "Unable to mount client"
3353 MDSJOURNALSIZE=$mds_journal_size_orig
3356 run_test 56 "check big indexes"
3358 test_57a() { # bug 22656
3359 local NID=$(do_facet ost1 "$LCTL get_param nis" | tail -1 | awk '{print $1}')
3360 writeconf_or_reformat
3361 do_facet ost1 "$TUNEFS --failnode=$NID `ostdevname 1`" || error "tunefs failed"
3363 start_ost && error "OST registration from failnode should fail"
3366 run_test 57a "initial registration from failnode should fail (should return errs)"
3369 local NID=$(do_facet ost1 "$LCTL get_param nis" | tail -1 | awk '{print $1}')
3370 writeconf_or_reformat
3371 do_facet ost1 "$TUNEFS --servicenode=$NID `ostdevname 1`" || error "tunefs failed"
3373 start_ost || error "OST registration from servicenode should not fail"
3376 run_test 57b "initial registration from servicenode should not fail"
3379 do_facet mgs $LCTL get_param mgs.MGS.live.$FSNAME | grep OST | wc -l
3382 test_58() { # bug 22658
3383 if [ $(facet_fstype mds) != ldiskfs ]; then
3384 skip "Only applicable to ldiskfs-based MDTs"
3389 createmany -o $DIR/$tdir/$tfile-%d 100
3390 # make sure that OSTs do not cancel llog cookies before we unmount the MDS
3391 #define OBD_FAIL_OBD_LOG_CANCEL_NET 0x601
3392 do_facet $SINGLEMDS "lctl set_param fail_loc=0x601"
3393 unlinkmany $DIR/$tdir/$tfile-%d 100
3396 local MNTDIR=$(facet_mntpt $SINGLEMDS)
3397 local devname=$(mdsdevname ${SINGLEMDS//mds/})
3399 if ! do_facet $SINGLEMDS "test -b $devname"; then
3403 # remove all files from the OBJECTS dir
3404 do_facet $SINGLEMDS "mount -t ldiskfs $opts $devname $MNTDIR"
3405 do_facet $SINGLEMDS "find $MNTDIR/O/1/d* -type f -delete"
3406 do_facet $SINGLEMDS "umount $MNTDIR"
3407 # restart MDS with missing llog files
3409 do_facet mds "lctl set_param fail_loc=0"
3412 run_test 58 "missing llog files must not prevent MDT from mounting"
3415 start_mgsmds >> /dev/null
3416 local C1=$(count_osts)
3417 if [ $C1 -eq 0 ]; then
3418 start_ost >> /dev/null
3422 echo "original ost count: $C1 (expect > 0)"
3423 [ $C1 -gt 0 ] || error "No OSTs in $FSNAME log"
3424 start_mgsmds -o writeconf >> /dev/null || error "MDT start failed"
3425 local C2=$(count_osts)
3426 echo "after mdt writeconf count: $C2 (expect 0)"
3427 [ $C2 -gt 0 ] && error "MDT writeconf should erase OST logs"
3428 echo "OST start without writeconf should fail:"
3429 start_ost >> /dev/null && error "OST start without writeconf didn't fail"
3430 echo "OST start with writeconf should succeed:"
3431 start_ost -o writeconf >> /dev/null || error "OST1 start failed"
3432 local C3=$(count_osts)
3433 echo "after ost writeconf count: $C3 (expect 1)"
3434 [ $C3 -eq 1 ] || error "new OST writeconf should add:"
3435 start_ost2 -o writeconf >> /dev/null || error "OST2 start failed"
3436 local C4=$(count_osts)
3437 echo "after ost2 writeconf count: $C4 (expect 2)"
3438 [ $C4 -eq 2 ] || error "OST2 writeconf should add log"
3439 stop_ost2 >> /dev/null
3440 cleanup_nocli >> /dev/null
3441 #writeconf to remove all ost2 traces for subsequent tests
3442 writeconf_or_reformat
3444 run_test 59 "writeconf mount option"
3446 test_60() { # LU-471
3449 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
3450 skip "Only applicable to ldiskfs-based MDTs"
3454 for num in $(seq $MDSCOUNT); do
3455 add mds${num} $(mkfs_opts mds${num} $(mdsdevname $num)) \
3456 --mkfsoptions='\" -E stride=64 -O ^uninit_bg\"' \
3457 --reformat $(mdsdevname $num) $(mdsvdevname $num) ||
3461 dump=$(do_facet $SINGLEMDS dumpe2fs $(mdsdevname 1))
3463 [ $rc -eq 0 ] || error "dumpe2fs $(mdsdevname 1) failed"
3465 # MDT default has dirdata feature
3466 echo $dump | grep dirdata > /dev/null || error "dirdata is not set"
3467 # we disable uninit_bg feature
3468 echo $dump | grep uninit_bg > /dev/null && error "uninit_bg is set"
3469 # we set stride extended options
3470 echo $dump | grep stride > /dev/null || error "stride is not set"
3473 run_test 60 "check mkfs.lustre --mkfsoptions -E -O options setting"
3476 local reformat=false
3478 [ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.1.53) ] ||
3479 { skip "Need MDS version at least 2.1.53"; return 0; }
3481 if [ $(facet_fstype $SINGLEMDS) == ldiskfs ] &&
3482 ! large_xattr_enabled; then
3484 LDISKFS_MKFS_OPTS+=" -O large_xattr"
3486 for num in $(seq $MDSCOUNT); do
3487 add mds${num} $(mkfs_opts mds$num $(mdsdevname $num)) \
3488 --reformat $(mdsdevname $num) $(mdsvdevname $num) ||
3489 error "add mds $num failed"
3493 setup_noconfig || error "setting up the filesystem failed"
3494 client_up || error "starting client failed"
3496 local file=$DIR/$tfile
3499 local large_value="$(generate_string $(max_xattr_size))"
3500 local small_value="bar"
3502 local name="trusted.big"
3503 log "save large xattr $name on $file"
3504 setfattr -n $name -v $large_value $file ||
3505 error "saving $name on $file failed"
3507 local new_value=$(get_xattr_value $name $file)
3508 [[ "$new_value" != "$large_value" ]] &&
3509 error "$name different after saving"
3511 log "shrink value of $name on $file"
3512 setfattr -n $name -v $small_value $file ||
3513 error "shrinking value of $name on $file failed"
3515 new_value=$(get_xattr_value $name $file)
3516 [[ "$new_value" != "$small_value" ]] &&
3517 error "$name different after shrinking"
3519 log "grow value of $name on $file"
3520 setfattr -n $name -v $large_value $file ||
3521 error "growing value of $name on $file failed"
3523 new_value=$(get_xattr_value $name $file)
3524 [[ "$new_value" != "$large_value" ]] &&
3525 error "$name different after growing"
3527 log "check value of $name on $file after remounting MDS"
3529 new_value=$(get_xattr_value $name $file)
3530 [[ "$new_value" != "$large_value" ]] &&
3531 error "$name different after remounting MDS"
3533 log "remove large xattr $name from $file"
3534 setfattr -x $name $file || error "removing $name from $file failed"
3539 LDISKFS_MKFS_OPTS=${LDISKFS_MKFS_OPTS% -O large_xattr}
3543 run_test 61 "large xattr"
3546 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
3547 skip "Only applicable to ldiskfs-based MDTs"
3552 local mdsdev=$(mdsdevname 1)
3553 local ostdev=$(ostdevname 1)
3555 [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.2.51) ]] ||
3556 { skip "Need MDS version at least 2.2.51"; return 0; }
3558 echo "disable journal for mds"
3559 do_facet mds tune2fs -O ^has_journal $mdsdev || error "tune2fs failed"
3560 start_mds && error "MDT start should fail"
3561 echo "disable journal for ost"
3562 do_facet ost1 tune2fs -O ^has_journal $ostdev || error "tune2fs failed"
3563 start_ost && error "OST start should fail"
3564 cleanup || return $?
3567 run_test 62 "start with disabled journal"
3570 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
3571 skip "Only applicable to ldiskfs-based MDTs"
3575 local inode_slab=$(do_facet $SINGLEMDS \
3576 "awk '/ldiskfs_inode_cache/ { print \\\$5 }' /proc/slabinfo")
3577 if [ -z "$inode_slab" ]; then
3578 skip "ldiskfs module has not been loaded"
3582 echo "$inode_slab ldisk inodes per page"
3583 [ "$inode_slab" -ge "3" ] ||
3584 error "ldisk inode size is too big, $inode_slab objs per page"
3587 run_test 63 "Verify each page can at least hold 3 ldisk inodes"
3592 start_ost2 || error "Unable to start second ost"
3593 mount_client $MOUNT || error "Unable to mount client"
3594 stop_ost2 || error "Unable to stop second ost"
3596 $LFS df --lazy || error "lfs df failed"
3597 cleanup || return $?
3598 #writeconf to remove all ost2 traces for subsequent tests
3599 writeconf_or_reformat
3601 run_test 64 "check lfs df --lazy "
3603 test_65() { # LU-2237
3604 # Currently, the test is only valid for ldiskfs backend
3605 [ "$(facet_fstype $SINGLEMDS)" != "ldiskfs" ] &&
3606 skip "non-ldiskfs backend" && return
3608 local devname=$(mdsdevname ${SINGLEMDS//mds/})
3609 local brpt=$(facet_mntpt brpt)
3612 if ! do_facet $SINGLEMDS "test -b $devname"; then
3617 local obj=$(do_facet $SINGLEMDS \
3618 "$DEBUGFS -c -R \\\"stat last_rcvd\\\" $devname" |
3620 if [ -z "$obj" ]; then
3621 # The MDT may be just re-formatted, mount the MDT for the
3622 # first time to guarantee the "last_rcvd" file is there.
3623 start_mds || error "fail to mount the MDS for the first time"
3627 # remove the "last_rcvd" file
3628 do_facet $SINGLEMDS "mkdir -p $brpt"
3629 do_facet $SINGLEMDS \
3630 "mount -t $(facet_fstype $SINGLEMDS) $opts $devname $brpt"
3631 do_facet $SINGLEMDS "rm -f ${brpt}/last_rcvd"
3632 do_facet $SINGLEMDS "umount $brpt"
3634 # restart MDS, the "last_rcvd" file should be recreated.
3635 start_mds || error "fail to restart the MDS"
3637 obj=$(do_facet $SINGLEMDS \
3638 "$DEBUGFS -c -R \\\"stat last_rcvd\\\" $devname" | grep Inode)
3639 [ -n "$obj" ] || error "fail to re-create the last_rcvd"
3641 run_test 65 "re-create the lost last_rcvd file when server mount"
3644 [[ $(lustre_version_code mgs) -ge $(version_code 2.3.59) ]] ||
3645 { skip "Need MGS version at least 2.3.59"; return 0; }
3648 local OST1_NID=$(do_facet ost1 $LCTL list_nids | head -1)
3649 local MDS_NID=$(do_facet $SINGLEMDS $LCTL list_nids | head -1)
3651 echo "replace_nids should fail if MDS, OSTs and clients are UP"
3652 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 $OST1_NID &&
3653 error "replace_nids fail"
3655 umount_client $MOUNT || error "unmounting client failed"
3656 echo "replace_nids should fail if MDS and OSTs are UP"
3657 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 $OST1_NID &&
3658 error "replace_nids fail"
3661 echo "replace_nids should fail if MDS is UP"
3662 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 $OST1_NID &&
3663 error "replace_nids fail"
3665 stop_mds || error "stopping mds failed"
3667 if combined_mgs_mds; then
3668 start_mdt 1 "-o nosvc" ||
3669 error "starting mds with nosvc option failed"
3672 echo "command should accept two parameters"
3673 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 &&
3674 error "command should accept two params"
3676 echo "correct device name should be passed"
3677 do_facet mgs $LCTL replace_nids $FSNAME-WRONG0000 $OST1_NID &&
3678 error "wrong devname"
3680 echo "wrong nids list should not destroy the system"
3681 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 "wrong nids list" &&
3684 echo "replace OST nid"
3685 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 $OST1_NID ||
3686 error "replace nids failed"
3688 echo "command should accept two parameters"
3689 do_facet mgs $LCTL replace_nids $FSNAME-MDT0000 &&
3690 error "command should accept two params"
3692 echo "wrong nids list should not destroy the system"
3693 do_facet mgs $LCTL replace_nids $FSNAME-MDT0000 "wrong nids list" &&
3696 echo "replace MDS nid"
3697 do_facet mgs $LCTL replace_nids $FSNAME-MDT0000 $MDS_NID ||
3698 error "replace nids failed"
3700 if ! combined_mgs_mds ; then
3707 check_mount || error "error after nid replace"
3708 cleanup || error "cleanup failed"
3711 run_test 66 "replace nids"
3714 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
3717 start_mdt 1 || error "MDT0 start fail"
3719 start_ost || error "OST0 start fail"
3721 start_mdt 2 || error "MDT1 start fail"
3723 mount_client $MOUNT || error "mount client fails"
3725 mkdir -p $DIR/$tdir || error "create dir fail"
3727 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
3728 error "create remote dir fail"
3730 rm -rf $DIR/$tdir || error "delete dir fail"
3731 cleanup || return $?
3733 run_test 70a "start MDT0, then OST, then MDT1"
3736 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
3739 start_ost || error "OST0 start fail"
3741 start_mdt 1 || error "MDT0 start fail"
3742 start_mdt 2 || error "MDT1 start fail"
3744 mount_client $MOUNT || error "mount client fails"
3746 mkdir -p $DIR/$tdir || error "create dir fail"
3748 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
3749 error "create remote dir fail"
3751 rm -rf $DIR/$tdir || error "delete dir fail"
3753 cleanup || return $?
3755 run_test 70b "start OST, MDT1, MDT0"
3758 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
3761 start_mdt 1 || error "MDT0 start fail"
3762 start_mdt 2 || error "MDT1 start fail"
3763 start_ost || error "OST0 start fail"
3765 mount_client $MOUNT || error "mount client fails"
3766 stop_mdt 1 || error "MDT1 start fail"
3768 local mdc_for_mdt1=$($LCTL dl | grep MDT0000-mdc | awk '{print $4}')
3769 echo "deactivate $mdc_for_mdt1"
3770 $LCTL --device $mdc_for_mdt1 deactivate || return 1
3772 mkdir -p $DIR/$tdir && error "mkdir succeed"
3774 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir &&
3775 error "create remote dir succeed"
3777 cleanup || return $?
3779 run_test 70c "stop MDT0, mkdir fail, create remote dir fail"
3782 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
3785 start_mdt 1 || error "MDT0 start fail"
3786 start_mdt 2 || error "MDT1 start fail"
3787 start_ost || error "OST0 start fail"
3789 mount_client $MOUNT || error "mount client fails"
3791 stop_mdt 2 || error "MDT1 start fail"
3793 local mdc_for_mdt2=$($LCTL dl | grep MDT0001-mdc |
3795 echo "deactivate $mdc_for_mdt2"
3796 $LCTL --device $mdc_for_mdt2 deactivate ||
3797 error "set $mdc_for_mdt2 deactivate failed"
3799 mkdir -p $DIR/$tdir || error "mkdir fail"
3800 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir &&
3801 error "create remote dir succeed"
3803 rm -rf $DIR/$tdir || error "delete dir fail"
3805 cleanup || return $?
3807 run_test 70d "stop MDT1, mkdir succeed, create remote dir fail"
3810 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
3811 if combined_mgs_mds; then
3812 skip "needs separate MGS/MDT" && return
3816 start_mdt 1 || error "MDT0 start fail"
3817 start_ost || error "OST0 start fail"
3818 start_mdt 2 || error "MDT1 start fail"
3819 start_ost2 || error "OST1 start fail"
3821 mount_client $MOUNT || error "mount client fails"
3823 mkdir -p $DIR/$tdir || error "mkdir fail"
3824 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
3825 error "create remote dir succeed"
3827 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
3828 rm -rf $DIR/$tdir || error "delete dir fail"
3830 umount_client $MOUNT
3831 stop_mdt 1 || error "MDT0 stop fail"
3832 stop_mdt 2 || error "MDT1 stop fail"
3833 stop_ost || error "OST0 stop fail"
3834 stop_ost2 || error "OST1 stop fail"
3836 run_test 71a "start MDT0 OST0, MDT1, OST1"
3839 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
3840 if combined_mgs_mds; then
3841 skip "needs separate MGS/MDT" && return
3845 start_mdt 2 || error "MDT1 start fail"
3846 start_ost || error "OST0 start fail"
3847 start_mdt 1 || error "MDT0 start fail"
3848 start_ost2 || error "OST1 start fail"
3850 mount_client $MOUNT || error "mount client fails"
3852 mkdir -p $DIR/$tdir || error "mkdir fail"
3853 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
3854 error "create remote dir succeed"
3856 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
3857 rm -rf $DIR/$tdir || error "delete dir fail"
3859 umount_client $MOUNT
3860 stop_mdt 1 || error "MDT0 stop fail"
3861 stop_mdt 2 || error "MDT1 stop fail"
3862 stop_ost || error "OST0 stop fail"
3863 stop_ost2 || error "OST1 stop fail"
3865 run_test 71b "start MDT1, OST0, MDT0, OST1"
3868 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
3869 if combined_mgs_mds; then
3870 skip "needs separate MGS/MDT" && return
3874 start_ost || error "OST0 start fail"
3875 start_ost2 || error "OST1 start fail"
3876 start_mdt 2 || error "MDT1 start fail"
3877 start_mdt 1 || error "MDT0 start fail"
3879 mount_client $MOUNT || error "mount client fails"
3881 mkdir -p $DIR/$tdir || error "mkdir fail"
3882 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
3883 error "create remote dir succeed"
3885 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
3886 rm -rf $DIR/$tdir || error "delete dir fail"
3888 umount_client $MOUNT
3889 stop_mdt 1 || error "MDT0 stop fail"
3890 stop_mdt 2 || error "MDT1 stop fail"
3891 stop_ost || error "OST0 stop fail"
3892 stop_ost2 || error "OST1 stop fail"
3895 run_test 71c "start OST0, OST1, MDT1, MDT0"
3898 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
3899 if combined_mgs_mds; then
3900 skip "needs separate MGS/MDT" && return
3904 start_ost || error "OST0 start fail"
3905 start_mdt 2 || error "MDT0 start fail"
3906 start_mdt 1 || error "MDT0 start fail"
3907 start_ost2 || error "OST1 start fail"
3909 mount_client $MOUNT || error "mount client fails"
3911 mkdir -p $DIR/$tdir || error "mkdir fail"
3912 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
3913 error "create remote dir succeed"
3915 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
3916 rm -rf $DIR/$tdir || error "delete dir fail"
3918 umount_client $MOUNT
3919 stop_mdt 1 || error "MDT0 stop fail"
3920 stop_mdt 2 || error "MDT1 stop fail"
3921 stop_ost || error "OST0 stop fail"
3922 stop_ost2 || error "OST1 stop fail"
3925 run_test 71d "start OST0, MDT1, MDT0, OST1"
3928 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
3929 if combined_mgs_mds; then
3930 skip "needs separate MGS/MDT" && return
3934 start_ost || error "OST0 start fail"
3935 start_mdt 2 || error "MDT1 start fail"
3936 start_ost2 || error "OST1 start fail"
3937 start_mdt 1 || error "MDT0 start fail"
3939 mount_client $MOUNT || error "mount client fails"
3941 mkdir -p $DIR/$tdir || error "mkdir fail"
3942 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
3943 error "create remote dir succeed"
3945 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
3946 rm -rf $DIR/$tdir || error "delete dir fail"
3948 umount_client $MOUNT
3949 stop_mdt 1 || error "MDT0 stop fail"
3950 stop_mdt 2 || error "MDT1 stop fail"
3951 stop_ost || error "OST0 stop fail"
3952 stop_ost2 || error "OST1 stop fail"
3955 run_test 71e "start OST0, MDT1, OST1, MDT0"
3957 test_72() { #LU-2634
3958 local mdsdev=$(mdsdevname 1)
3959 local ostdev=$(ostdevname 1)
3960 local cmd="$E2FSCK -fnvd $mdsdev"
3963 [ "$(facet_fstype $SINGLEMDS)" != "ldiskfs" ] &&
3964 skip "ldiskfs only test" && return
3966 #tune MDT with "-O extents"
3968 for num in $(seq $MDSCOUNT); do
3969 add mds${num} $(mkfs_opts mds$num $(mdsdevname $num)) \
3970 --reformat $(mdsdevname $num) $(mdsvdevname $num) ||
3971 error "add mds $num failed"
3972 $TUNE2FS -O extents $(mdsdevname $num)
3975 add ost1 $(mkfs_opts ost1 $ostdev) --reformat $ostdev ||
3976 error "add $ostdev failed"
3977 start_mgsmds || error "start mds failed"
3978 start_ost || error "start ost failed"
3979 mount_client $MOUNT || error "mount client failed"
3981 #create some short symlinks
3983 createmany -o $DIR/$tdir/$tfile-%d $fn
3984 echo "create $fn short symlinks"
3985 for i in $(seq -w 1 $fn); do
3986 ln -s $DIR/$tdir/$tfile-$i $MOUNT/$tfile-$i
3991 umount_client $MOUNT || error "umount client failed"
3992 stop_mds || error "stop mds failed"
3993 stop_ost || error "stop ost failed"
3996 run_e2fsck $(facet_active_host $SINGLEMDS) $mdsdev "-n"
3998 run_test 72 "test fast symlink with extents flag enabled"
4000 if ! combined_mgs_mds ; then