3 # FIXME - there is no reason to use all of these different return codes,
4 # espcially when most of them are mapped to something else anyway.
5 # The tests should use error() to describe the failure more clearly,
6 # and reduce the need to look into the tests to see what failed.
12 # bug number for skipped test: LU-2828
13 ALWAYS_EXCEPT="$CONF_SANITY_EXCEPT 59 64"
14 # UPDATE THE COMMENT ABOVE WITH BUG NUMBERS WHEN CHANGING ALWAYS_EXCEPT!
18 if [ -r /etc/SuSE-release ]
20 local vers=`grep VERSION /etc/SuSE-release | awk '{print $3}'`
21 local patchlev=`grep PATCHLEVEL /etc/SuSE-release \
23 if [ $vers -eq 11 ] && [ $patchlev -eq 2 ]
31 if is_sles11; then # LU-2181
32 ALWAYS_EXCEPT="$ALWAYS_EXCEPT 23a 34b"
35 if [ "$FAILURE_MODE" = "HARD" ]; then
36 CONFIG_EXCEPTIONS="24a " && \
37 echo "Except the tests: $CONFIG_EXCEPTIONS for FAILURE_MODE=$FAILURE_MODE, bug 23573" && \
38 ALWAYS_EXCEPT="$ALWAYS_EXCEPT $CONFIG_EXCEPTIONS"
41 # bug number for skipped test:
42 # a tool to create lustre filesystem images
43 ALWAYS_EXCEPT="32newtarball $ALWAYS_EXCEPT"
46 PATH=$PWD/$SRCDIR:$SRCDIR:$SRCDIR/../utils:$PATH
48 PTLDEBUG=${PTLDEBUG:--1}
50 LUSTRE=${LUSTRE:-`dirname $0`/..}
51 RLUSTRE=${RLUSTRE:-$LUSTRE}
52 LUSTRE_TESTS_API_DIR=${LUSTRE_TESTS_API_DIR:-${LUSTRE}/tests/clientapi}
53 export MULTIOP=${MULTIOP:-multiop}
55 . $LUSTRE/tests/test-framework.sh
57 . ${CONFIG:=$LUSTRE/tests/cfg/$NAME.sh}
59 # use small MDS + OST size to speed formatting time
60 # do not use too small MDSSIZE/OSTSIZE, which affect the default jouranl size
61 # STORED_MDSSIZE is used in test_18
62 STORED_MDSSIZE=$MDSSIZE
63 STORED_OSTSIZE=$OSTSIZE
67 if ! combined_mgs_mds; then
68 # bug number for skipped test: 23954
69 ALWAYS_EXCEPT="$ALWAYS_EXCEPT 24b"
72 # pass "-E lazy_itable_init" to mke2fs to speed up the formatting time
73 if [[ "$LDISKFS_MKFS_OPTS" != *lazy_itable_init* ]]; then
74 LDISKFS_MKFS_OPTS=$(csa_add "$LDISKFS_MKFS_OPTS" -E lazy_itable_init)
77 [ $(facet_fstype $SINGLEMDS) = "zfs" ] &&
78 # bug number for skipped test: LU-2778 LU-2059 LU-4444
79 ALWAYS_EXCEPT="$ALWAYS_EXCEPT 57b 50h 69"
84 require_dsh_mds || exit 0
85 require_dsh_ost || exit 0
87 [ "$SLOW" = "no" ] && EXCEPT_SLOW="30a 31 45 69"
93 # The MGS must be started before the OSTs for a new fs, so start
94 # and stop to generate the startup logs.
97 wait_osc_import_state mds ost FULL
102 reformat_and_config() {
104 if ! combined_mgs_mds ; then
110 writeconf_or_reformat() {
111 # There are at most 2 OSTs for write_conf test
112 # who knows if/where $TUNEFS is installed?
113 # Better reformat if it fails...
114 writeconf_all $MDSCOUNT 2 ||
115 { echo "tunefs failed, reformatting instead" &&
116 reformat_and_config && return 0; }
126 start mgs $(mgsdevname) $MGS_MOUNT_OPTS
132 local dev=$(mdsdevname $num)
135 echo "start mds service on `facet_active_host $facet`"
136 start $facet ${dev} $MDS_MOUNT_OPTS $@ || return 94
142 local dev=$(mdsdevname $num)
145 echo "stop mds service on `facet_active_host $facet`"
146 # These tests all use non-failover stop
147 stop $facet -f || return 97
153 for num in $(seq $MDSCOUNT); do
154 start_mdt $num $@ || return 94
159 if ! combined_mgs_mds ; then
167 for num in $(seq $MDSCOUNT); do
168 stop_mdt $num || return 97
173 echo "stop mgs service on `facet_active_host mgs`"
174 # These tests all use non-failover stop
175 stop mgs -f || return 97
179 echo "start ost1 service on `facet_active_host ost1`"
180 start ost1 `ostdevname 1` $OST_MOUNT_OPTS $@ || return 95
184 echo "stop ost1 service on `facet_active_host ost1`"
185 # These tests all use non-failover stop
186 stop ost1 -f || return 98
190 echo "start ost2 service on `facet_active_host ost2`"
191 start ost2 `ostdevname 2` $OST_MOUNT_OPTS $@ || return 92
195 echo "stop ost2 service on `facet_active_host ost2`"
196 # These tests all use non-failover stop
197 stop ost2 -f || return 93
202 echo "mount $FSNAME on ${MOUNTPATH}....."
203 zconf_mount `hostname` $MOUNTPATH || return 96
207 local mountopt="-o remount,$1"
209 echo "remount '$1' lustre on ${MOUNTPATH}....."
210 zconf_mount `hostname` $MOUNTPATH "$mountopt" || return 96
215 echo "umount lustre on ${MOUNTPATH}....."
216 zconf_umount `hostname` $MOUNTPATH || return 97
219 manual_umount_client(){
222 echo "manual umount lustre on ${MOUNT}...."
223 do_facet client "umount -d ${FORCE} $MOUNT"
229 start_mds || error "MDT start failed"
230 start_ost || error "OST start failed"
231 mount_client $MOUNT || error "client start failed"
232 client_up || error "client_up failed"
236 if ! combined_mgs_mds ; then
245 unload_modules_conf () {
246 if combined_mgs_mds || ! local_mode; then
247 unload_modules || return 1
252 stop_ost || return 202
253 stop_mds || return 201
254 unload_modules_conf || return 203
258 umount_client $MOUNT || return 200
259 cleanup_nocli || return $?
263 do_facet client "cp /etc/passwd $DIR/a" || return 71
264 do_facet client "rm $DIR/a" || return 72
265 # make sure lustre is actually mounted (touch will block,
266 # but grep won't, so do it after)
267 do_facet client "grep $MOUNT' ' /proc/mounts > /dev/null" || return 73
268 echo "setup single mount lustre success"
272 do_facet client "touch $DIR/a" || return 71
273 do_facet client "rm $DIR/a" || return 72
274 do_facet client "touch $DIR2/a" || return 73
275 do_facet client "rm $DIR2/a" || return 74
276 echo "setup double mount lustre success"
281 if [ "$ONLY" == "setup" ]; then
286 if [ "$ONLY" == "cleanup" ]; then
293 #create single point mountpoint
299 check_mount || return 41
302 run_test 0 "single mount setup"
305 start_mds || error "MDT start failed"
307 echo "start ost second time..."
308 start_ost && error "2nd OST start should fail"
309 mount_client $MOUNT || error "client start failed"
310 check_mount || return 42
313 run_test 1 "start up ost twice (should return errors)"
317 echo "start mds second time.."
318 start_mdt 1 && error "2nd MDT start should fail"
321 check_mount || return 43
324 run_test 2 "start up mds twice (should return err)"
328 #mount.lustre returns an error if already in mtab
329 mount_client $MOUNT && error "2nd client mount should fail"
330 check_mount || return 44
333 run_test 3 "mount client twice (should return err)"
337 touch $DIR/$tfile || return 85
341 # ok for ost to fail shutdown
342 if [ 202 -ne $eno ]; then
347 run_test 4 "force cleanup ost, then cleanup"
349 test_5a() { # was test_5
351 touch $DIR/$tfile || return 1
352 fuser -m -v $MOUNT && echo "$MOUNT is in use by user space process."
354 stop_mds -f || return 2
356 # cleanup may return an error from the failed
357 # disconnects; for now I'll consider this successful
358 # if all the modules have unloaded.
362 echo "killing umount"
363 kill -TERM $UMOUNT_PID
364 echo "waiting for umount to finish"
366 if grep " $MOUNT " /proc/mounts; then
367 echo "test 5: /proc/mounts after failed umount"
371 echo "killing umount"
372 kill -TERM $UMOUNT_PID
373 echo "waiting for umount to finish"
375 grep " $MOUNT " /proc/mounts && echo "test 5: /proc/mounts after second umount" && return 11
379 # stop_mds is a no-op here, and should not fail
380 cleanup_nocli || return $?
381 # df may have lingering entry
383 # mtab may have lingering entry
387 while [ "$WAIT" -ne "$MAX_WAIT" ]; do
389 grep -q $MOUNT" " /etc/mtab || break
390 echo "Waiting /etc/mtab updated ... "
391 WAIT=$(( WAIT + sleep))
393 [ "$WAIT" -eq "$MAX_WAIT" ] && error "/etc/mtab is not updated in $WAIT secs"
394 echo "/etc/mtab updated in $WAIT secs"
396 run_test 5a "force cleanup mds, then cleanup"
404 grep " $MOUNT " /etc/mtab && \
405 error false "unexpected entry in mtab before mount" && return 10
409 if ! combined_mgs_mds ; then
410 trap cleanup_5b EXIT ERR
415 [ -d $MOUNT ] || mkdir -p $MOUNT
416 mount_client $MOUNT && rc=1
417 grep " $MOUNT " /etc/mtab && \
418 error "$MOUNT entry in mtab after failed mount" && rc=11
420 # stop_mds is a no-op here, and should not fail
421 cleanup_nocli || rc=$?
422 if ! combined_mgs_mds ; then
427 run_test 5b "Try to start a client with no MGS (should return errs)"
430 grep " $MOUNT " /etc/mtab && \
431 error false "unexpected entry in mtab before mount" && return 10
436 [ -d $MOUNT ] || mkdir -p $MOUNT
437 local oldfs="${FSNAME}"
438 FSNAME="wrong.${FSNAME}"
439 mount_client $MOUNT || :
441 grep " $MOUNT " /etc/mtab && \
442 error "$MOUNT entry in mtab after failed mount" && rc=11
444 cleanup_nocli || rc=$?
447 run_test 5c "cleanup after failed mount (bug 2712) (should return errs)"
450 grep " $MOUNT " /etc/mtab && \
451 error false "unexpected entry in mtab before mount" && return 10
453 [ "$(facet_fstype ost1)" = "zfs" ] &&
454 skip "LU-2059: no local config for ZFS OSTs" && return
460 mount_client $MOUNT || rc=1
462 grep " $MOUNT " /etc/mtab && \
463 error "$MOUNT entry in mtab after unmount" && rc=11
466 run_test 5d "mount with ost down"
469 grep " $MOUNT " /etc/mtab && \
470 error false "unexpected entry in mtab before mount" && return 10
476 #define OBD_FAIL_PTLRPC_DELAY_SEND 0x506
477 do_facet client "lctl set_param fail_loc=0x80000506"
478 mount_client $MOUNT || echo "mount failed (not fatal)"
480 grep " $MOUNT " /etc/mtab && \
481 error "$MOUNT entry in mtab after unmount" && rc=11
484 run_test 5e "delayed connect, don't crash (bug 10268)"
487 if combined_mgs_mds ; then
488 skip "combined mgs and mds"
492 grep " $MOUNT " /etc/mtab && \
493 error false "unexpected entry in mtab before mount" && return 10
497 [ -d $MOUNT ] || mkdir -p $MOUNT
498 mount_client $MOUNT &
500 echo client_mount pid is $pid
504 if ! ps -f -p $pid >/dev/null; then
507 grep " $MOUNT " /etc/mtab && echo "test 5f: mtab after mount"
508 error "mount returns $rc, expected to hang"
517 # mount should succeed after start mds
520 [ $rc -eq 0 ] || error "mount returned $rc"
521 grep " $MOUNT " /etc/mtab && echo "test 5f: mtab after mount"
525 run_test 5f "mds down, cleanup after failed mount (bug 2712)"
530 mount_client ${MOUNT} || return 87
531 touch $DIR/a || return 86
534 run_test 6 "manual umount, then mount again"
539 cleanup_nocli || return $?
541 run_test 7 "manual umount, then cleanup"
546 check_mount2 || return 45
547 umount_client $MOUNT2
550 run_test 8 "double mount setup"
555 do_facet ost1 lctl set_param debug=\'inode trace\' || return 1
556 do_facet ost1 lctl set_param subsystem_debug=\'mds ost\' || return 1
558 CHECK_PTLDEBUG="`do_facet ost1 lctl get_param -n debug`"
559 if [ "$CHECK_PTLDEBUG" ] && { \
560 [ "$CHECK_PTLDEBUG" = "trace inode warning error emerg console" ] ||
561 [ "$CHECK_PTLDEBUG" = "trace inode" ]; }; then
562 echo "lnet.debug success"
564 echo "lnet.debug: want 'trace inode', have '$CHECK_PTLDEBUG'"
567 CHECK_SUBSYS="`do_facet ost1 lctl get_param -n subsystem_debug`"
568 if [ "$CHECK_SUBSYS" ] && [ "$CHECK_SUBSYS" = "mds ost" ]; then
569 echo "lnet.subsystem_debug success"
571 echo "lnet.subsystem_debug: want 'mds ost', have '$CHECK_SUBSYS'"
574 stop_ost || return $?
576 run_test 9 "test ptldebug and subsystem for mkfs"
584 do_facet $facet "test -b $dev" || rc=1
585 if [[ "$size" ]]; then
586 local in=$(do_facet $facet "dd if=$dev of=/dev/null bs=1k count=1 skip=$size 2>&1" |\
587 awk '($3 == "in") { print $1 }')
588 [[ $in = "1+0" ]] || rc=1
594 # Test 16 was to "verify that lustre will correct the mode of OBJECTS".
595 # But with new MDS stack we don't care about the mode of local objects
596 # anymore, so this test is removed. See bug 22944 for more details.
600 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
601 skip "Only applicable to ldiskfs-based MDTs"
606 check_mount || return 41
609 echo "Remove mds config log"
610 if ! combined_mgs_mds ; then
614 do_facet mgs "$DEBUGFS -w -R 'unlink CONFIGS/$FSNAME-MDT0000' \
615 $(mgsdevname) || return \$?" || return $?
617 if ! combined_mgs_mds ; then
622 start_mds && return 42
625 run_test 17 "Verify failed mds_postsetup won't fail assertion (2936) (should return errs)"
628 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
629 skip "Only applicable to ldiskfs-based MDTs"
633 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
638 # check if current MDSSIZE is large enough
639 [ $MDSSIZE -ge $MIN ] && OK=1 && myMDSSIZE=$MDSSIZE && \
640 log "use MDSSIZE=$MDSSIZE"
642 # check if the global config has a large enough MDSSIZE
643 [ -z "$OK" -a ! -z "$STORED_MDSSIZE" ] && [ $STORED_MDSSIZE -ge $MIN ] && \
644 OK=1 && myMDSSIZE=$STORED_MDSSIZE && \
645 log "use STORED_MDSSIZE=$STORED_MDSSIZE"
647 # check if the block device is large enough
648 is_blkdev $SINGLEMDS $MDSDEV $MIN
649 local large_enough=$?
650 if [ -n "$OK" ]; then
651 [ $large_enough -ne 0 ] && OK=""
653 [ $large_enough -eq 0 ] && OK=1 && myMDSSIZE=$MIN &&
654 log "use device $MDSDEV with MIN=$MIN"
657 # check if a loopback device has enough space for fs metadata (5%)
659 if [ -z "$OK" ]; then
660 local SPACE=$(do_facet $SINGLEMDS "[ -f $MDSDEV -o ! -e $MDSDEV ] && df -P \\\$(dirname $MDSDEV)" |
661 awk '($1 != "Filesystem") {print $4}')
662 ! [ -z "$SPACE" ] && [ $SPACE -gt $((MIN / 20)) ] && \
663 OK=1 && myMDSSIZE=$MIN && \
664 log "use file $MDSDEV with MIN=$MIN"
667 [ -z "$OK" ] && skip_env "$MDSDEV too small for ${MIN}kB MDS" && return
670 echo "mount mds with large journal..."
672 local OLD_MDSSIZE=$MDSSIZE
676 echo "mount lustre system..."
678 check_mount || return 41
680 echo "check journal size..."
681 local FOUNDSIZE=$(do_facet $SINGLEMDS "$DEBUGFS -c -R 'stat <8>' $MDSDEV" | awk '/Size: / { print $NF; exit;}')
682 if [ $FOUNDSIZE -gt $((32 * 1024 * 1024)) ]; then
683 log "Success: mkfs creates large journals. Size: $((FOUNDSIZE >> 20))M"
685 error "expected journal size > 32M, found $((FOUNDSIZE >> 20))M"
693 run_test 18 "check mkfs creates large journals"
696 start_mds || return 1
697 stop_mds -f || return 2
699 run_test 19a "start/stop MDS without OSTs"
702 [ "$(facet_fstype ost1)" = "zfs" ] &&
703 skip "LU-2059: no local config for ZFS OSTs" && return
705 start_ost || return 1
706 stop_ost -f || return 2
708 run_test 19b "start/stop OSTs without MDS"
711 # first format the ost/mdt
715 check_mount || return 43
717 remount_client ro $MOUNT || return 44
718 touch $DIR/$tfile && echo "$DIR/$tfile created incorrectly" && return 45
719 [ -e $DIR/$tfile ] && echo "$DIR/$tfile exists incorrectly" && return 46
720 remount_client rw $MOUNT || return 47
722 [ ! -f $DIR/$tfile ] && echo "$DIR/$tfile missing" && return 48
723 MCNT=`grep -c $MOUNT /etc/mtab`
724 [ "$MCNT" -ne 1 ] && echo "$MOUNT in /etc/mtab $MCNT times" && return 49
729 run_test 20 "remount ro,rw mounts work and doesn't break /etc/mtab"
734 wait_osc_import_state mds ost FULL
738 run_test 21a "start mds before ost, stop ost first"
741 [ "$(facet_fstype ost1)" = "zfs" ] &&
742 skip "LU-2059: no local config for ZFS OSTs" && return
746 wait_osc_import_state mds ost FULL
750 run_test 21b "start ost before mds, stop mds first"
756 wait_osc_import_state mds ost2 FULL
760 #writeconf to remove all ost2 traces for subsequent tests
761 writeconf_or_reformat
763 run_test 21c "start mds between two osts, stop mds last"
766 if combined_mgs_mds ; then
767 skip "need separate mgs device" && return 0
777 wait_osc_import_state mds ost2 FULL
783 #writeconf to remove all ost2 traces for subsequent tests
784 writeconf_or_reformat
787 run_test 21d "start mgs then ost and then mds"
792 echo Client mount with ost in logs, but none running
794 # wait until mds connected to ost and open client connection
795 wait_osc_import_state mds ost FULL
798 # check_mount will block trying to contact ost
799 mcreate $DIR/$tfile || return 40
800 rm -f $DIR/$tfile || return 42
804 echo Client mount with a running ost
807 # if gss enabled, wait full time to let connection from
808 # mds to ost be established, due to the mismatch between
809 # initial connect timeout and gss context negotiation timeout.
810 # This perhaps could be remove after AT landed.
811 echo "sleep $((TIMEOUT + TIMEOUT + TIMEOUT))s"
812 sleep $((TIMEOUT + TIMEOUT + TIMEOUT))
815 wait_osc_import_state mds ost FULL
816 wait_osc_import_state client ost FULL
817 check_mount || return 41
822 run_test 22 "start a client before osts (should return errs)"
824 test_23a() { # was test_23
828 # force down client so that recovering mds waits for reconnect
829 local running=$(grep -c $MOUNT /proc/mounts) || true
830 if [ $running -ne 0 ]; then
831 echo "Stopping client $MOUNT (opts: -f)"
835 # enter recovery on mds
837 # try to start a new client
838 mount_client $MOUNT &
840 MOUNT_PID=$(ps -ef | grep "t lustre" | grep -v grep | awk '{print $2}')
841 MOUNT_LUSTRE_PID=`ps -ef | grep mount.lustre | grep -v grep | awk '{print $2}'`
842 echo mount pid is ${MOUNT_PID}, mount.lustre pid is ${MOUNT_LUSTRE_PID}
844 ps --ppid $MOUNT_LUSTRE_PID
845 echo "waiting for mount to finish"
847 # "ctrl-c" sends SIGINT but it usually (in script) does not work on child process
848 # SIGTERM works but it does not spread to offspring processses
849 kill -s TERM $MOUNT_PID
850 kill -s TERM $MOUNT_LUSTRE_PID
851 # we can not wait $MOUNT_PID because it is not a child of this shell
857 while [ "$WAIT" -lt "$MAX_WAIT" ]; do
859 PID1=$(ps -ef | awk '{print $2}' | grep -w $MOUNT_PID)
860 PID2=$(ps -ef | awk '{print $2}' | grep -w $MOUNT_LUSTRE_PID)
863 [ -z "$PID1" -a -z "$PID2" ] && break
864 echo "waiting for mount to finish ... "
865 WAIT=$(( WAIT + sleep))
867 if [ "$WAIT" -eq "$MAX_WAIT" ]; then
868 error "MOUNT_PID $MOUNT_PID and "\
869 "MOUNT_LUSTRE_PID $MOUNT_LUSTRE_PID still not killed in $WAIT secs"
872 stop_mds || error "stopping MDSes failed"
873 stop_ost || error "stopping OSSes failed"
875 run_test 23a "interrupt client during recovery mount delay"
880 test_23b() { # was test_23
883 # Simulate -EINTR during mount OBD_FAIL_LDLM_CLOSE_THREAD
884 lctl set_param fail_loc=0x80000313
888 run_test 23b "Simulate -EINTR during mount"
890 fs2mds_HOST=$mds_HOST
891 fs2ost_HOST=$ost_HOST
893 MDSDEV1_2=$fs2mds_DEV
894 OSTDEV1_2=$fs2ost_DEV
895 OSTDEV2_2=$fs3ost_DEV
899 echo "umount $MOUNT2 ..."
900 umount $MOUNT2 || true
901 echo "stopping fs2mds ..."
902 stop fs2mds -f || true
903 echo "stopping fs2ost ..."
904 stop fs2ost -f || true
908 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
910 if [ -z "$fs2ost_DEV" -o -z "$fs2mds_DEV" ]; then
911 is_blkdev $SINGLEMDS $MDSDEV && \
912 skip_env "mixed loopback and real device not working" && return
915 [ -n "$ost1_HOST" ] && fs2ost_HOST=$ost1_HOST
917 local fs2mdsdev=$(mdsdevname 1_2)
918 local fs2ostdev=$(ostdevname 1_2)
919 local fs2mdsvdev=$(mdsvdevname 1_2)
920 local fs2ostvdev=$(ostvdevname 1_2)
922 # test 8-char fsname as well
923 local FSNAME2=test1234
925 add fs2mds $(mkfs_opts mds1 ${fs2mdsdev} ) --nomgs --mgsnode=$MGSNID \
926 --fsname=${FSNAME2} --reformat $fs2mdsdev $fs2mdsvdev || exit 10
928 add fs2ost $(mkfs_opts ost1 ${fs2ostdev}) --fsname=${FSNAME2} \
929 --reformat $fs2ostdev $fs2ostvdev || exit 10
932 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS && trap cleanup_fs2 EXIT INT
933 start fs2ost $fs2ostdev $OST_MOUNT_OPTS
935 mount -t lustre $MGSNID:/${FSNAME2} $MOUNT2 || return 1
937 check_mount || return 2
938 # files written on 1 should not show up on 2
939 cp /etc/passwd $DIR/$tfile
941 [ -e $MOUNT2/$tfile ] && error "File bleed" && return 7
944 cp /etc/passwd $MOUNT2/b || return 3
945 rm $MOUNT2/b || return 4
946 # 2 is actually mounted
947 grep $MOUNT2' ' /proc/mounts > /dev/null || return 5
949 facet_failover fs2mds
950 facet_failover fs2ost
953 # the MDS must remain up until last MDT
955 MDS=$(do_facet $SINGLEMDS "lctl get_param -n devices" | awk '($3 ~ "mdt" && $4 ~ "MDT") { print $4 }' | head -1)
956 [ -z "$MDS" ] && error "No MDT" && return 8
958 cleanup_nocli || return 6
960 run_test 24a "Multiple MDTs on a single node"
963 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
965 if [ -z "$fs2mds_DEV" ]; then
966 local dev=${SINGLEMDS}_dev
968 is_blkdev $SINGLEMDS $MDSDEV && \
969 skip_env "mixed loopback and real device not working" && return
972 local fs2mdsdev=$(mdsdevname 1_2)
973 local fs2mdsvdev=$(mdsvdevname 1_2)
975 add fs2mds $(mkfs_opts mds1 ${fs2mdsdev} ) --mgs --fsname=${FSNAME}2 \
976 --reformat $fs2mdsdev $fs2mdsvdev || exit 10
978 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS && return 2
981 run_test 24b "Multiple MGSs on a single node (should return err)"
985 check_mount || return 2
986 local MODULES=$($LCTL modules | awk '{ print $2 }')
987 rmmod $MODULES 2>/dev/null || true
990 run_test 25 "Verify modules are referenced"
994 # we need modules before mount for sysctl, so make sure...
995 do_facet $SINGLEMDS "lsmod | grep -q lustre || modprobe lustre"
996 #define OBD_FAIL_MDS_FS_SETUP 0x135
997 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000135"
998 start_mds && echo MDS started && return 1
999 lctl get_param -n devices
1000 DEVS=$(lctl get_param -n devices | egrep -v MG | wc -l)
1001 [ $DEVS -gt 0 ] && return 2
1002 # start mds to drop writeconf setting
1003 start_mds || return 3
1004 stop_mds || return 4
1005 unload_modules_conf || return $?
1007 run_test 26 "MDT startup failure cleans LOV (should return errs)"
1010 [ "$(facet_fstype ost1)" = "zfs" ] &&
1011 skip "LU-2059: no local config for ZFS OSTs" && return
1013 start_ost || return 1
1014 start_mds || return 2
1015 echo "Requeue thread should have started: "
1016 ps -e | grep ll_cfg_requeue
1017 set_conf_param_and_check ost1 \
1018 "lctl get_param -n obdfilter.$FSNAME-OST0000.client_cache_seconds" \
1019 "$FSNAME-OST0000.ost.client_cache_seconds" || return 3
1022 run_test 27a "Reacquire MGS lock if OST started first"
1027 local device=$(do_facet $SINGLEMDS "lctl get_param -n devices" |
1028 awk '($3 ~ "mdt" && $4 ~ "MDT0000") { print $4 }')
1030 facet_failover $SINGLEMDS
1031 set_conf_param_and_check $SINGLEMDS \
1032 "lctl get_param -n mdt.$device.identity_acquire_expire" \
1033 "$device.mdt.identity_acquire_expire" || return 3
1034 set_conf_param_and_check client \
1035 "lctl get_param -n mdc.$device-mdc-*.max_rpcs_in_flight"\
1036 "$device.mdc.max_rpcs_in_flight" || return 4
1040 run_test 27b "Reacquire MGS lock after failover"
1044 TEST="lctl get_param -n llite.$FSNAME-*.max_read_ahead_whole_mb"
1045 PARAM="$FSNAME.llite.max_read_ahead_whole_mb"
1047 FINAL=$(($ORIG + 1))
1048 set_conf_param_and_check client "$TEST" "$PARAM" $FINAL || return 3
1049 FINAL=$(($FINAL + 1))
1050 set_conf_param_and_check client "$TEST" "$PARAM" $FINAL || return 4
1051 umount_client $MOUNT || return 200
1054 if [ $RESULT -ne $FINAL ]; then
1055 echo "New config not seen: wanted $FINAL got $RESULT"
1058 echo "New config success: got $RESULT"
1060 set_conf_param_and_check client "$TEST" "$PARAM" $ORIG || return 5
1063 run_test 28 "permanent parameter setting"
1065 test_28a() { # LU-4221
1066 [[ $(lustre_version_code ost1) -ge $(version_code 2.5.52) ]] ||
1067 { skip "Need OST version at least 2.5.52" && return 0; }
1068 [ "$(facet_fstype ost1)" = "zfs" ] &&
1069 skip "LU-4221: no such proc params for ZFS OSTs" && return
1076 local device="$FSNAME-OST0000"
1080 # In this test we will set three kinds of proc parameters with
1082 # 1. the ones moved from the OFD to the OSD, and only their
1083 # symlinks kept in obdfilter
1084 # 2. non-symlink ones in the OFD
1085 # 3. non-symlink ones in the OSD
1088 # prepare a symlink parameter in the OFD
1089 name="writethrough_cache_enable"
1090 param="$device.ost.$name"
1091 cmd="$LCTL get_param -n obdfilter.$device.$name"
1093 # conf_param the symlink parameter in the OFD
1094 old=$(do_facet ost1 $cmd)
1095 new=$(((old + 1) % 2))
1096 set_conf_param_and_check ost1 "$cmd" "$param" $new ||
1097 error "lctl conf_param $device.ost.$param=$new failed"
1099 # conf_param the target parameter in the OSD
1100 param="$device.osd.$name"
1101 cmd="$LCTL get_param -n osd-*.$device.$name"
1102 set_conf_param_and_check ost1 "$cmd" "$param" $old ||
1103 error "lctl conf_param $device.osd.$param=$old failed"
1106 # prepare a non-symlink parameter in the OFD
1107 name="client_cache_seconds"
1108 param="$device.ost.$name"
1109 cmd="$LCTL get_param -n obdfilter.$device.$name"
1111 # conf_param the parameter in the OFD
1112 old=$(do_facet ost1 $cmd)
1114 set_conf_param_and_check ost1 "$cmd" "$param" $new ||
1115 error "lctl conf_param $device.ost.$param=$new failed"
1116 set_conf_param_and_check ost1 "$cmd" "$param" $old ||
1117 error "lctl conf_param $device.ost.$param=$old failed"
1120 # prepare a non-symlink parameter in the OSD
1121 name="lma_self_repair"
1122 param="$device.osd.$name"
1123 cmd="$LCTL get_param -n osd-*.$device.$name"
1125 # conf_param the parameter in the OSD
1126 old=$(do_facet ost1 $cmd)
1127 new=$(((old + 1) % 2))
1128 set_conf_param_and_check ost1 "$cmd" "$param" $new ||
1129 error "lctl conf_param $device.osd.$param=$new failed"
1130 set_conf_param_and_check ost1 "$cmd" "$param" $old ||
1131 error "lctl conf_param $device.osd.$param=$old failed"
1135 run_test 28a "set symlink parameters permanently with conf_param"
1138 [ "$OSTCOUNT" -lt "2" ] && skip_env "$OSTCOUNT < 2, skipping" && return
1139 setup > /dev/null 2>&1
1143 local PARAM="$FSNAME-OST0001.osc.active"
1144 local PROC_ACT="osc.$FSNAME-OST0001-osc-[^M]*.active"
1145 local PROC_UUID="osc.$FSNAME-OST0001-osc-[^M]*.ost_server_uuid"
1147 ACTV=$(lctl get_param -n $PROC_ACT)
1149 set_conf_param_and_check client \
1150 "lctl get_param -n $PROC_ACT" "$PARAM" $DEAC || return 2
1151 # also check ost_server_uuid status
1152 RESULT=$(lctl get_param -n $PROC_UUID | grep DEACTIV)
1153 if [ -z "$RESULT" ]; then
1154 echo "Live client not deactivated: $(lctl get_param -n $PROC_UUID)"
1157 echo "Live client success: got $RESULT"
1161 for num in $(seq $MDSCOUNT); do
1162 local mdtosc=$(get_mdtosc_proc_path mds${num} $FSNAME-OST0001)
1163 local MPROC="osc.$mdtosc.active"
1168 RESULT=$(do_facet mds${num} " lctl get_param -n $MPROC")
1169 [ ${PIPESTATUS[0]} = 0 ] || error "Can't read $MPROC"
1170 if [ $RESULT -eq $DEAC ]; then
1171 echo -n "MDT deactivated also after"
1172 echo "$WAIT sec (got $RESULT)"
1176 if [ $WAIT -eq $MAX ]; then
1177 echo -n "MDT not deactivated: wanted $DEAC"
1181 echo "Waiting $(($MAX - $WAIT))secs for MDT deactivated"
1184 # test new client starts deactivated
1185 umount_client $MOUNT || return 200
1187 RESULT=$(lctl get_param -n $PROC_UUID | grep DEACTIV | grep NEW)
1188 if [ -z "$RESULT" ]; then
1189 echo "New client not deactivated from start: $(lctl get_param -n $PROC_UUID)"
1192 echo "New client success: got $RESULT"
1195 # make sure it reactivates
1196 set_conf_param_and_check client \
1197 "lctl get_param -n $PROC_ACT" "$PARAM" $ACTV || return 6
1199 umount_client $MOUNT
1202 #writeconf to remove all ost2 traces for subsequent tests
1203 writeconf_or_reformat
1205 run_test 29 "permanently remove an OST"
1210 echo Big config llog
1211 TEST="lctl get_param -n llite.$FSNAME-*.max_read_ahead_whole_mb"
1213 LIST=(1 2 3 4 5 4 3 2 1 2 3 4 5 4 3 2 1 2 3 4 5)
1214 for i in ${LIST[@]}; do
1215 set_conf_param_and_check client "$TEST" \
1216 "$FSNAME.llite.max_read_ahead_whole_mb" $i || return 3
1218 # make sure client restart still works
1219 umount_client $MOUNT
1220 mount_client $MOUNT || return 4
1221 [ "$($TEST)" -ne "$i" ] && error "Param didn't stick across restart $($TEST) != $i"
1224 echo Erase parameter setting
1225 do_facet mgs "$LCTL conf_param -d $FSNAME.llite.max_read_ahead_whole_mb" || return 6
1226 umount_client $MOUNT
1227 mount_client $MOUNT || return 6
1229 echo "deleted (default) value=$FINAL, orig=$ORIG"
1230 # assumes this parameter started at the default value
1231 [ "$FINAL" -eq "$ORIG" ] || fail "Deleted value=$FINAL, orig=$ORIG"
1235 run_test 30a "Big config llog and conf_param deletion"
1240 # Make a fake nid. Use the OST nid, and add 20 to the least significant
1241 # numerical part of it. Hopefully that's not already a failover address for
1243 OSTNID=$(do_facet ost1 "$LCTL get_param nis" | tail -1 | awk '{print $1}')
1244 ORIGVAL=$(echo $OSTNID | egrep -oi "[0-9]*@")
1245 NEWVAL=$((($(echo $ORIGVAL | egrep -oi "[0-9]*") + 20) % 256))
1246 NEW=$(echo $OSTNID | sed "s/$ORIGVAL/$NEWVAL@/")
1247 echo "Using fake nid $NEW"
1249 TEST="$LCTL get_param -n osc.$FSNAME-OST0000-osc-[^M]*.import | grep failover_nids | sed -n 's/.*\($NEW\).*/\1/p'"
1250 set_conf_param_and_check client "$TEST" \
1251 "$FSNAME-OST0000.failover.node" $NEW ||
1252 error "didn't add failover nid $NEW"
1253 NIDS=$($LCTL get_param -n osc.$FSNAME-OST0000-osc-[^M]*.import | grep failover_nids)
1255 NIDCOUNT=$(($(echo "$NIDS" | wc -w) - 1))
1256 echo "should have 2 failover nids: $NIDCOUNT"
1257 [ $NIDCOUNT -eq 2 ] || error "Failover nid not added"
1258 do_facet mgs "$LCTL conf_param -d $FSNAME-OST0000.failover.node" || error "conf_param delete failed"
1259 umount_client $MOUNT
1260 mount_client $MOUNT || return 3
1262 NIDS=$($LCTL get_param -n osc.$FSNAME-OST0000-osc-[^M]*.import | grep failover_nids)
1264 NIDCOUNT=$(($(echo "$NIDS" | wc -w) - 1))
1265 echo "only 1 final nid should remain: $NIDCOUNT"
1266 [ $NIDCOUNT -eq 1 ] || error "Failover nids not removed"
1270 run_test 30b "Remove failover nids"
1272 test_31() { # bug 10734
1273 # ipaddr must not exist
1274 mount -t lustre 4.3.2.1@tcp:/lustre $MOUNT || true
1277 run_test 31 "Connect to non-existent node (shouldn't crash)"
1281 T32_BLIMIT=20480 # Kbytes
1285 # This is not really a test but a tool to create new disk
1286 # image tarballs for the upgrade tests.
1288 # Disk image tarballs should be created on single-node
1289 # clusters by running this test with default configurations
1290 # plus a few mandatory environment settings that are verified
1291 # at the beginning of the test.
1293 test_32newtarball() {
1297 local tmp=$TMP/t32_image_create
1299 if [ $FSNAME != t32fs -o $MDSCOUNT -ne 1 -o \
1300 \( -z "$MDSDEV" -a -z "$MDSDEV1" \) -o $OSTCOUNT -ne 1 -o \
1301 -z "$OSTDEV1" ]; then
1302 error "Needs FSNAME=t32fs MDSCOUNT=1 MDSDEV1=<nonexistent_file>" \
1303 "(or MDSDEV, in the case of b1_8) OSTCOUNT=1" \
1304 "OSTDEV1=<nonexistent_file>"
1308 echo "Found stale $tmp"
1313 tar cf - -C $src . | tar xf - -C $tmp/src
1314 dd if=/dev/zero of=$tmp/src/t32_qf_old bs=1M \
1315 count=$(($T32_BLIMIT / 1024 / 2))
1316 chown $T32_QID.$T32_QID $tmp/src/t32_qf_old
1322 [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.3.50) ] &&
1323 $LFS quotacheck -ug /mnt/$FSNAME
1324 $LFS setquota -u $T32_QID -b 0 -B $T32_BLIMIT -i 0 -I $T32_ILIMIT \
1327 tar cf - -C $tmp/src . | tar xf - -C /mnt/$FSNAME
1334 ls -Rni --time-style=+%s >$tmp/img/list
1335 find . ! -name .lustre -type f -exec sha1sum {} \; |
1336 sort -k 2 >$tmp/img/sha1sums
1338 $LCTL get_param -n version | head -n 1 |
1339 sed -e 's/^lustre: *//' >$tmp/img/commit
1341 [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.3.50) ] &&
1342 $LFS quotaon -ug /mnt/$FSNAME
1343 $LFS quota -u $T32_QID -v /mnt/$FSNAME
1344 $LFS quota -v -u $T32_QID /mnt/$FSNAME |
1345 awk 'BEGIN { num='1' } { if ($1 == "'/mnt/$FSNAME'") \
1346 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1347 | tr -d "*" > $tmp/img/bspace
1348 $LFS quota -v -u $T32_QID /mnt/$FSNAME |
1349 awk 'BEGIN { num='5' } { if ($1 == "'/mnt/$FSNAME'") \
1350 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1351 | tr -d "*" > $tmp/img/ispace
1356 find -type f -exec sha1sum {} \; | sort -k 2 >$tmp/sha1sums.src
1359 if ! diff -u $tmp/sha1sums.src $tmp/img/sha1sums; then
1360 echo "Data verification failed"
1363 uname -r >$tmp/img/kernel
1364 uname -m >$tmp/img/arch
1366 mv ${MDSDEV1:-$MDSDEV} $tmp/img
1367 mv $OSTDEV1 $tmp/img
1369 version=$(sed -e 's/\(^[0-9]\+\.[0-9]\+\)\(.*$\)/\1/' $tmp/img/commit |
1370 sed -e 's/\./_/g') # E.g., "1.8.7" -> "1_8"
1373 tar cjvf $dst/disk$version-$(facet_fstype $SINGLEMDS).tar.bz2 -S *
1378 #run_test 32newtarball "Create a new test_32 disk image tarball for this version"
1381 # The list of applicable tarballs is returned via the caller's
1382 # variable "tarballs".
1385 local node=$(facet_active_host $SINGLEMDS)
1386 local r="do_node $node"
1388 if [ "$CLIENTONLY" ]; then
1389 skip "Client-only testing"
1393 if ! $r which $TUNEFS; then
1394 skip_env "tunefs.lustre required on $node"
1398 local IMGTYPE=$(facet_fstype $SINGLEMDS)
1400 tarballs=$($r find $RLUSTRE/tests -maxdepth 1 -name \'disk*-$IMGTYPE.tar.bz2\')
1402 if [ -z "$tarballs" ]; then
1403 skip "No applicable tarballs found"
1408 t32_test_cleanup() {
1412 if $shall_cleanup_lustre; then
1413 umount $tmp/mnt/lustre || rc=$?
1415 if $shall_cleanup_mdt; then
1416 $r umount -d $tmp/mnt/mdt || rc=$?
1418 if $shall_cleanup_mdt1; then
1419 $r umount -d $tmp/mnt/mdt1 || rc=$?
1421 if $shall_cleanup_ost; then
1422 $r umount -d $tmp/mnt/ost || rc=$?
1430 t32_bits_per_long() {
1432 # Yes, this is not meant to be perfect.
1442 t32_reload_modules() {
1444 local all_removed=false
1447 while ((i < 20)); do
1448 echo "Unloading modules on $node: Attempt $i"
1449 do_rpc_nodes $node $LUSTRE_RMMOD $(facet_fstype $SINGLEMDS) &&
1451 do_rpc_nodes $node check_mem_leak || return 1
1452 if $all_removed; then
1453 do_rpc_nodes $node load_modules
1459 echo "Unloading modules on $node: Given up"
1463 t32_wait_til_devices_gone() {
1469 echo wait for devices to go
1470 while ((i < 20)); do
1471 devices=$(do_rpc_nodes $node $LCTL device_list | wc -l)
1472 loops=$(do_rpc_nodes $node losetup -a | grep -c t32)
1473 ((devices == 0 && loops == 0)) && return 0
1477 echo "waiting for dev on $node: dev $devices loop $loops given up"
1478 do_rpc_nodes $node "losetup -a"
1479 do_rpc_nodes $node "$LCTL devices_list"
1483 t32_verify_quota() {
1487 local fstype=$(facet_fstype $SINGLEMDS)
1491 $LFS quota -u $T32_QID -v $mnt
1493 qval=$($LFS quota -v -u $T32_QID $mnt |
1494 awk 'BEGIN { num='1' } { if ($1 == "'$mnt'") \
1495 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1497 [ $qval -eq $img_bspace ] || {
1498 echo "bspace, act:$qval, exp:$img_bspace"
1502 qval=$($LFS quota -v -u $T32_QID $mnt |
1503 awk 'BEGIN { num='5' } { if ($1 == "'$mnt'") \
1504 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1506 [ $qval -eq $img_ispace ] || {
1507 echo "ispace, act:$qval, exp:$img_ispace"
1511 qval=$($LFS quota -v -u $T32_QID $mnt |
1512 awk 'BEGIN { num='3' } { if ($1 == "'$mnt'") \
1513 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1515 [ $qval -eq $T32_BLIMIT ] || {
1516 echo "blimit, act:$qval, exp:$T32_BLIMIT"
1520 qval=$($LFS quota -v -u $T32_QID $mnt |
1521 awk 'BEGIN { num='7' } { if ($1 == "'$mnt'") \
1522 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1524 [ $qval -eq $T32_ILIMIT ] || {
1525 echo "ilimit, act:$qval, exp:$T32_ILIMIT"
1529 do_node $node $LCTL conf_param $fsname.quota.mdt=ug
1530 cmd="$LCTL get_param -n osd-$fstype.$fsname-MDT0000"
1531 cmd=$cmd.quota_slave.enabled
1532 wait_update $node "$cmd" "ug" || {
1533 echo "Enable mdt quota failed"
1537 do_node $node $LCTL conf_param $fsname.quota.ost=ug
1538 cmd="$LCTL get_param -n osd-$fstype.$fsname-OST0000"
1539 cmd=$cmd.quota_slave.enabled
1540 wait_update $node "$cmd" "ug" || {
1541 echo "Enable ost quota failed"
1546 runas -u $T32_QID -g $T32_QID dd if=/dev/zero of=$mnt/t32_qf_new \
1547 bs=1M count=$(($T32_BLIMIT / 1024)) oflag=sync && {
1548 echo "Write succeed, but expect -EDQUOT"
1551 rm -f $mnt/t32_qf_new
1553 runas -u $T32_QID -g $T32_QID createmany -m $mnt/t32_qf_ \
1555 echo "Create succeed, but expect -EDQUOT"
1558 unlinkmany $mnt/t32_qf_ $T32_ILIMIT
1566 local dne_upgrade=${dne_upgrade:-"no"}
1567 local ff_convert=${ff_convert:-"no"}
1568 local shall_cleanup_mdt=false
1569 local shall_cleanup_mdt1=false
1570 local shall_cleanup_ost=false
1571 local shall_cleanup_lustre=false
1572 local node=$(facet_active_host $SINGLEMDS)
1573 local r="do_node $node"
1574 local node2=$(facet_active_host mds2)
1582 local nid=$($r $LCTL list_nids | head -1)
1588 local fstype=$(facet_fstype $SINGLEMDS)
1590 trap 'trap - RETURN; t32_test_cleanup' RETURN
1592 mkdir -p $tmp/mnt/lustre
1593 $r mkdir -p $tmp/mnt/{mdt,ost}
1594 $r tar xjvf $tarball -S -C $tmp || {
1595 error_noexit "Unpacking the disk image tarball"
1598 img_commit=$($r cat $tmp/commit)
1599 img_kernel=$($r cat $tmp/kernel)
1600 img_arch=$($r cat $tmp/arch)
1601 img_bspace=$($r cat $tmp/bspace)
1602 img_ispace=$($r cat $tmp/ispace)
1603 echo "Upgrading from $(basename $tarball), created with:"
1604 echo " Commit: $img_commit"
1605 echo " Kernel: $img_kernel"
1606 echo " Arch: $img_arch"
1608 local version=$(version_code $img_commit)
1609 [[ $version -gt $(version_code 2.4.0) ]] && ff_convert="no"
1611 $r $LCTL set_param debug="$PTLDEBUG"
1613 $r $TUNEFS --dryrun $tmp/mdt || {
1615 error_noexit "tunefs.lustre before mounting the MDT"
1618 if [ "$writeconf" ]; then
1619 mopts=loop,writeconf
1620 if [ $fstype == "ldiskfs" ]; then
1621 $r $TUNEFS --quota $tmp/mdt || {
1623 error_noexit "Enable mdt quota feature"
1628 if [ -n "$($LCTL list_nids | grep -v '\(tcp\|lo\)[[:digit:]]*$')" ]; then
1629 [[ $(lustre_version_code mgs) -ge $(version_code 2.3.59) ]] ||
1630 { skip "LU-2200: Cannot run over Inifiniband w/o lctl replace_nids "
1631 "(Need MGS version at least 2.3.59)"; return 0; }
1633 local osthost=$(facet_active_host ost1)
1634 local ostnid=$(do_node $osthost $LCTL list_nids | head -1)
1636 $r mount -t lustre -o loop,nosvc $tmp/mdt $tmp/mnt/mdt
1637 $r lctl replace_nids $fsname-OST0000 $ostnid
1638 $r lctl replace_nids $fsname-MDT0000 $nid
1639 $r umount -d $tmp/mnt/mdt
1642 mopts=loop,exclude=$fsname-OST0000
1645 t32_wait_til_devices_gone $node
1647 $r mount -t lustre -o $mopts $tmp/mdt $tmp/mnt/mdt || {
1649 error_noexit "Mounting the MDT"
1652 shall_cleanup_mdt=true
1654 if [ "$dne_upgrade" != "no" ]; then
1655 local fs2mdsdev=$(mdsdevname 1_2)
1656 local fs2mdsvdev=$(mdsvdevname 1_2)
1658 echo "mkfs new MDT on ${fs2mdsdev}...."
1659 if [ $(facet_fstype mds1) == ldiskfs ]; then
1660 mkfsoptions="--mkfsoptions=\\\"-J size=8\\\""
1663 add fs2mds $(mkfs_opts mds2 $fs2mdsdev $fsname) --reformat \
1664 $mkfsoptions $fs2mdsdev $fs2mdsvdev > /dev/null || {
1665 error_noexit "Mkfs new MDT failed"
1669 $r $TUNEFS --dryrun $fs2mdsdev || {
1670 error_noexit "tunefs.lustre before mounting the MDT"
1674 echo "mount new MDT....$fs2mdsdev"
1675 $r mkdir -p $tmp/mnt/mdt1
1676 $r mount -t lustre -o $mopts $fs2mdsdev $tmp/mnt/mdt1 || {
1677 error_noexit "mount mdt1 failed"
1680 shall_cleanup_mdt1=true
1683 uuid=$($r $LCTL get_param -n mdt.$fsname-MDT0000.uuid) || {
1684 error_noexit "Getting MDT UUID"
1687 if [ "$uuid" != $fsname-MDT0000_UUID ]; then
1688 error_noexit "Unexpected MDT UUID: \"$uuid\""
1692 $r $TUNEFS --dryrun $tmp/ost || {
1693 error_noexit "tunefs.lustre before mounting the OST"
1696 if [ "$writeconf" ]; then
1697 mopts=loop,mgsnode=$nid,$writeconf
1698 if [ $fstype == "ldiskfs" ]; then
1699 $r $TUNEFS --quota $tmp/ost || {
1701 error_noexit "Enable ost quota feature"
1706 mopts=loop,mgsnode=$nid
1708 $r mount -t lustre -o $mopts $tmp/ost $tmp/mnt/ost || {
1709 error_noexit "Mounting the OST"
1712 shall_cleanup_ost=true
1714 uuid=$($r $LCTL get_param -n obdfilter.$fsname-OST0000.uuid) || {
1715 error_noexit "Getting OST UUID"
1718 if [ "$uuid" != $fsname-OST0000_UUID ]; then
1719 error_noexit "Unexpected OST UUID: \"$uuid\""
1723 $r $LCTL conf_param $fsname-OST0000.osc.max_dirty_mb=15 || {
1724 error_noexit "Setting \"max_dirty_mb\""
1727 $r $LCTL conf_param $fsname-OST0000.failover.node=$nid || {
1728 error_noexit "Setting OST \"failover.node\""
1731 $r $LCTL conf_param $fsname-MDT0000.mdc.max_rpcs_in_flight=9 || {
1732 error_noexit "Setting \"max_rpcs_in_flight\""
1735 $r $LCTL conf_param $fsname-MDT0000.failover.node=$nid || {
1736 error_noexit "Setting MDT \"failover.node\""
1739 $r $LCTL pool_new $fsname.interop || {
1740 error_noexit "Setting \"interop\""
1743 $r $LCTL conf_param $fsname-MDT0000.lov.stripesize=4M || {
1744 error_noexit "Setting \"lov.stripesize\""
1748 if [ "$ff_convert" != "no" -a $(facet_fstype ost1) == "ldiskfs" ]; then
1749 $r $LCTL lfsck_start -M $fsname-OST0000 || {
1750 error_noexit "Start OI scrub on OST0"
1754 # The oi_scrub should be on ost1, but for test_32(),
1755 # all on the SINGLEMDS.
1756 wait_update_facet $SINGLEMDS "$LCTL get_param -n \
1757 osd-ldiskfs.$fsname-OST0000.oi_scrub |
1758 awk '/^status/ { print \\\$2 }'" "completed" 30 || {
1759 error_noexit "Failed to get the expected 'completed'"
1763 local UPDATED=$($r $LCTL get_param -n \
1764 osd-ldiskfs.$fsname-OST0000.oi_scrub |
1765 awk '/^updated/ { print $2 }')
1766 [ $UPDATED -ge 1 ] || {
1767 error_noexit "Only $UPDATED objects have been converted"
1772 if [ "$dne_upgrade" != "no" ]; then
1773 $r $LCTL conf_param \
1774 $fsname-MDT0001.mdc.max_rpcs_in_flight=9 || {
1775 error_noexit "Setting MDT1 \"max_rpcs_in_flight\""
1778 $r $LCTL conf_param $fsname-MDT0001.failover.node=$nid || {
1779 error_noexit "Setting MDT1 \"failover.node\""
1782 $r $LCTL conf_param $fsname-MDT0001.lov.stripesize=4M || {
1783 error_noexit "Setting MDT1 \"lov.stripesize\""
1789 if [ "$writeconf" ]; then
1790 mount -t lustre $nid:/$fsname $tmp/mnt/lustre || {
1791 error_noexit "Mounting the client"
1794 shall_cleanup_lustre=true
1795 $LCTL set_param debug="$PTLDEBUG"
1797 t32_verify_quota $node $fsname $tmp/mnt/lustre || {
1798 error_noexit "verify quota failed"
1802 if [ "$dne_upgrade" != "no" ]; then
1803 $LFS mkdir -i 1 $tmp/mnt/lustre/remote_dir || {
1804 error_noexit "set remote dir failed"
1808 pushd $tmp/mnt/lustre
1809 tar -cf - . --exclude=./remote_dir |
1810 tar -xvf - -C remote_dir 1>/dev/null || {
1811 error_noexit "cp to remote dir failed"
1817 dd if=/dev/zero of=$tmp/mnt/lustre/tmp_file bs=10k count=10 || {
1818 error_noexit "dd failed"
1821 rm -rf $tmp/mnt/lustre/tmp_file || {
1822 error_noexit "rm failed"
1826 if $r test -f $tmp/sha1sums; then
1827 # LU-2393 - do both sorts on same node to ensure locale
1829 $r cat $tmp/sha1sums | sort -k 2 >$tmp/sha1sums.orig
1830 if [ "$dne_upgrade" != "no" ]; then
1831 pushd $tmp/mnt/lustre/remote_dir
1833 pushd $tmp/mnt/lustre
1836 find ! -name .lustre -type f -exec sha1sum {} \; |
1837 sort -k 2 >$tmp/sha1sums || {
1838 error_noexit "sha1sum"
1842 if ! diff -ub $tmp/sha1sums.orig $tmp/sha1sums; then
1843 error_noexit "sha1sum verification failed"
1847 echo "sha1sum verification skipped"
1850 if [ "$dne_upgrade" != "no" ]; then
1851 rm -rf $tmp/mnt/lustre/remote_dir || {
1852 error_noexit "remove remote dir failed"
1857 if $r test -f $tmp/list; then
1859 # There is not a Test Framework API to copy files to or
1860 # from a remote node.
1862 # LU-2393 - do both sorts on same node to ensure locale
1864 $r cat $tmp/list | sort -k 6 >$tmp/list.orig
1865 pushd $tmp/mnt/lustre
1866 ls -Rni --time-style=+%s | sort -k 6 >$tmp/list || {
1872 # 32-bit and 64-bit clients use different algorithms to
1873 # convert FIDs into inode numbers. Hence, remove the inode
1874 # numbers from the lists, if the original list was created
1875 # on an architecture with different number of bits per
1878 if [ $(t32_bits_per_long $(uname -m)) != \
1879 $(t32_bits_per_long $img_arch) ]; then
1880 echo "Different number of bits per \"long\" from the disk image"
1881 for list in list.orig list; do
1882 sed -i -e 's/^[0-9]\+[ \t]\+//' $tmp/$list
1885 if ! diff -ub $tmp/list.orig $tmp/list; then
1886 error_noexit "list verification failed"
1890 echo "list verification skipped"
1894 # When adding new data verification tests, please check for
1895 # the presence of the required reference files first, like
1896 # the "sha1sums" and "list" tests above, to avoid the need to
1897 # regenerate every image for each test addition.
1900 nrpcs_orig=$($LCTL get_param \
1901 -n mdc.*MDT0000*.max_rpcs_in_flight) || {
1902 error_noexit "Getting \"max_rpcs_in_flight\""
1905 nrpcs=$((nrpcs_orig + 5))
1906 $r $LCTL conf_param $fsname-MDT0000.mdc.max_rpcs_in_flight=$nrpcs || {
1907 error_noexit "Changing \"max_rpcs_in_flight\""
1910 wait_update $HOSTNAME "$LCTL get_param \
1911 -n mdc.*MDT0000*.max_rpcs_in_flight" $nrpcs || {
1912 error_noexit "Verifying \"max_rpcs_in_flight\""
1916 umount $tmp/mnt/lustre || {
1917 error_noexit "Unmounting the client"
1920 shall_cleanup_lustre=false
1922 if [ "$dne_upgrade" != "no" ]; then
1923 $r umount -d $tmp/mnt/mdt1 || {
1924 error_noexit "Unmounting the MDT2"
1927 shall_cleanup_mdt1=false
1930 $r umount -d $tmp/mnt/mdt || {
1931 error_noexit "Unmounting the MDT"
1934 shall_cleanup_mdt=false
1936 $r umount -d $tmp/mnt/ost || {
1937 error_noexit "Unmounting the OST"
1940 shall_cleanup_ost=false
1942 t32_reload_modules $node || {
1943 error_noexit "Reloading modules"
1947 # mount a second time to make sure we didnt leave upgrade flag on
1948 $r $TUNEFS --dryrun $tmp/mdt || {
1950 error_noexit "tunefs.lustre before remounting the MDT"
1953 $r mount -t lustre -o loop,exclude=$fsname-OST0000 $tmp/mdt \
1955 error_noexit "Remounting the MDT"
1958 shall_cleanup_mdt=true
1968 for tarball in $tarballs; do
1969 t32_test $tarball || let "rc += $?"
1973 run_test 32a "Upgrade (not live)"
1981 for tarball in $tarballs; do
1982 t32_test $tarball writeconf || let "rc += $?"
1986 run_test 32b "Upgrade with writeconf"
1993 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
1995 for tarball in $tarballs; do
1996 dne_upgrade=yes t32_test $tarball writeconf || rc=$?
2000 run_test 32c "dne upgrade test"
2008 for tarball in $tarballs; do
2009 ff_convert=yes t32_test $tarball || rc=$?
2013 run_test 32d "convert ff test"
2015 test_33a() { # bug 12333, was test_33
2017 local FSNAME2=test-123
2018 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
2021 [ -n "$ost1_HOST" ] && fs2ost_HOST=$ost1_HOST
2023 if [ -z "$fs2ost_DEV" -o -z "$fs2mds_DEV" ]; then
2024 local dev=${SINGLEMDS}_dev
2025 local MDSDEV=${!dev}
2026 is_blkdev $SINGLEMDS $MDSDEV && \
2027 skip_env "mixed loopback and real device not working" && return
2030 local fs2mdsdev=$(mdsdevname 1_2)
2031 local fs2ostdev=$(ostdevname 1_2)
2032 local fs2mdsvdev=$(mdsvdevname 1_2)
2033 local fs2ostvdev=$(ostvdevname 1_2)
2035 if [ $(facet_fstype mds1) == ldiskfs ]; then
2036 mkfsoptions="--mkfsoptions=\\\"-J size=8\\\"" # See bug 17931.
2039 add fs2mds $(mkfs_opts mds1 ${fs2mdsdev}) --mgs --fsname=${FSNAME2} \
2040 --reformat $mkfsoptions $fs2mdsdev $fs2mdsvdev || exit 10
2041 add fs2ost $(mkfs_opts ost1 ${fs2ostdev}) --mgsnode=$MGSNID \
2042 --fsname=${FSNAME2} --index=8191 --reformat $fs2ostdev \
2043 $fs2ostvdev || exit 10
2045 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS && trap cleanup_fs2 EXIT INT
2046 start fs2ost $fs2ostdev $OST_MOUNT_OPTS
2047 do_facet $SINGLEMDS "$LCTL conf_param $FSNAME2.sys.timeout=200" || rc=1
2049 mount -t lustre $MGSNID:/${FSNAME2} $MOUNT2 || rc=2
2052 cp /etc/hosts $MOUNT2/ || rc=3
2053 $LFS getstripe $MOUNT2/hosts
2058 cleanup_nocli || rc=6
2061 run_test 33a "Mount ost with a large index number"
2063 test_33b() { # was test_34
2066 do_facet client dd if=/dev/zero of=$MOUNT/24 bs=1024k count=1
2067 # Drop lock cancelation reply during umount
2068 #define OBD_FAIL_LDLM_CANCEL_NET 0x304
2069 do_facet client lctl set_param fail_loc=0x80000304
2070 #lctl set_param debug=-1
2071 umount_client $MOUNT
2074 run_test 33b "Drop cancel during umount"
2078 do_facet client "sh runmultiop_bg_pause $DIR/file O_c"
2079 manual_umount_client
2081 do_facet client killall -USR1 multiop
2082 if [ $rc -eq 0 ]; then
2083 error "umount not fail!"
2088 run_test 34a "umount with opened file should be fail"
2093 touch $DIR/$tfile || return 1
2094 stop_mds --force || return 2
2096 manual_umount_client --force
2098 if [ $rc -ne 0 ]; then
2099 error "mtab after failed umount - rc $rc"
2105 run_test 34b "force umount with failed mds should be normal"
2109 touch $DIR/$tfile || return 1
2110 stop_ost --force || return 2
2112 manual_umount_client --force
2114 if [ $rc -ne 0 ]; then
2115 error "mtab after failed umount - rc $rc"
2121 run_test 34c "force umount with failed ost should be normal"
2123 test_35a() { # bug 12459
2126 DBG_SAVE="`lctl get_param -n debug`"
2127 lctl set_param debug="ha"
2129 log "Set up a fake failnode for the MDS"
2131 local device=$(do_facet $SINGLEMDS "lctl get_param -n devices" |
2132 awk '($3 ~ "mdt" && $4 ~ "MDT") { print $4 }' | head -1)
2133 do_facet mgs "$LCTL conf_param \
2134 ${device}.failover.node=$(h2$NETTYPE $FAKENID)" || return 4
2136 log "Wait for RECONNECT_INTERVAL seconds (10s)"
2139 MSG="conf-sanity.sh test_35a `date +%F%kh%Mm%Ss`"
2142 log "Stopping the MDT: $device"
2143 stop_mdt 1 || return 5
2145 df $MOUNT > /dev/null 2>&1 &
2147 log "Restarting the MDT: $device"
2148 start_mdt 1 || return 6
2149 log "Wait for df ($DFPID) ... "
2152 lctl set_param debug="$DBG_SAVE"
2154 # retrieve from the log the first server that the client tried to
2155 # contact after the connection loss
2156 $LCTL dk $TMP/lustre-log-$TESTNAME.log
2157 NEXTCONN=`awk "/${MSG}/ {start = 1;}
2158 /import_select_connection.*$device-mdc.* using connection/ {
2160 if (\\\$NF ~ /$FAKENID/)
2166 }" $TMP/lustre-log-$TESTNAME.log`
2167 [ "$NEXTCONN" != "0" ] && log "The client didn't try to reconnect to the last active server (tried ${NEXTCONN} instead)" && return 7
2169 # remove nid settings
2170 writeconf_or_reformat
2172 run_test 35a "Reconnect to the last active server first"
2174 test_35b() { # bug 18674
2175 remote_mds || { skip "local MDS" && return 0; }
2179 $LCTL set_param debug="ha"
2181 MSG="conf-sanity.sh test_35b `date +%F%kh%Mm%Ss`"
2184 log "Set up a fake failnode for the MDS"
2186 local device=$(do_facet $SINGLEMDS "$LCTL get_param -n devices" |
2187 awk '($3 ~ "mdt" && $4 ~ "MDT") { print $4 }' | head -1)
2188 do_facet mgs "$LCTL conf_param \
2189 ${device}.failover.node=$(h2$NETTYPE $FAKENID)" || return 1
2191 local at_max_saved=0
2192 # adaptive timeouts may prevent seeing the issue
2193 if at_is_enabled; then
2194 at_max_saved=$(at_max_get mds)
2195 at_max_set 0 mds client
2198 mkdir -p $MOUNT/$tdir
2200 log "Injecting EBUSY on MDS"
2201 # Setting OBD_FAIL_MDS_RESEND=0x136
2202 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0x80000136" || return 2
2204 $LCTL set_param mdc.${FSNAME}*.stats=clear
2206 log "Creating a test file and stat it"
2207 touch $MOUNT/$tdir/$tfile
2208 stat $MOUNT/$tdir/$tfile
2210 log "Stop injecting EBUSY on MDS"
2211 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0" || return 3
2212 rm -f $MOUNT/$tdir/$tfile
2215 # restore adaptive timeout
2216 [ $at_max_saved -ne 0 ] && at_max_set $at_max_saved mds client
2218 $LCTL dk $TMP/lustre-log-$TESTNAME.log
2220 CONNCNT=`$LCTL get_param mdc.${FSNAME}*.stats | awk '/mds_connect/{print $2}'`
2222 # retrieve from the log if the client has ever tried to
2223 # contact the fake server after the loss of connection
2224 FAILCONN=`awk "BEGIN {ret = 0;}
2225 /import_select_connection.*${FSNAME}-MDT0000-mdc.* using connection/ {
2227 if (\\\$NF ~ /$FAKENID/) {
2232 END {print ret}" $TMP/lustre-log-$TESTNAME.log`
2234 [ "$FAILCONN" == "0" ] && \
2235 log "ERROR: The client reconnection has not been triggered" && \
2237 [ "$FAILCONN" == "2" ] && \
2238 log "ERROR: The client tried to reconnect to the failover server while the primary was busy" && \
2242 # When OBD_FAIL_MDS_RESEND is hit, we sleep for 2 * obd_timeout
2243 # Reconnects are supposed to be rate limited to one every 5s
2244 [ $CONNCNT -gt $((2 * $TIMEOUT / 5 + 1)) ] && \
2245 log "ERROR: Too many reconnects $CONNCNT" && \
2249 # remove nid settings
2250 writeconf_or_reformat
2252 run_test 35b "Continue reconnection retries, if the active server is busy"
2255 [ $OSTCOUNT -lt 2 ] && skip_env "skipping test for single OST" && return
2257 [ "$ost_HOST" = "`hostname`" -o "$ost1_HOST" = "`hostname`" ] || \
2258 { skip "remote OST" && return 0; }
2261 local FSNAME2=test1234
2262 local fs3ost_HOST=$ost_HOST
2263 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
2265 [ -n "$ost1_HOST" ] && fs2ost_HOST=$ost1_HOST && fs3ost_HOST=$ost1_HOST
2267 if [ -z "$fs2ost_DEV" -o -z "$fs2mds_DEV" -o -z "$fs3ost_DEV" ]; then
2268 is_blkdev $SINGLEMDS $MDSDEV && \
2269 skip_env "mixed loopback and real device not working" && return
2272 local fs2mdsdev=$(mdsdevname 1_2)
2273 local fs2ostdev=$(ostdevname 1_2)
2274 local fs3ostdev=$(ostdevname 2_2)
2275 local fs2mdsvdev=$(mdsvdevname 1_2)
2276 local fs2ostvdev=$(ostvdevname 1_2)
2277 local fs3ostvdev=$(ostvdevname 2_2)
2279 add fs2mds $(mkfs_opts mds1 ${fs2mdsdev}) --mgs --fsname=${FSNAME2} \
2280 --reformat $fs2mdsdev $fs2mdsvdev || exit 10
2281 # XXX after we support non 4K disk blocksize in ldiskfs, specify a
2282 # different one than the default value here.
2283 add fs2ost $(mkfs_opts ost1 ${fs2ostdev}) --mgsnode=$MGSNID \
2284 --fsname=${FSNAME2} --reformat $fs2ostdev $fs2ostvdev || exit 10
2285 add fs3ost $(mkfs_opts ost1 ${fs3ostdev}) --mgsnode=$MGSNID \
2286 --fsname=${FSNAME2} --reformat $fs3ostdev $fs3ostvdev || exit 10
2288 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS
2289 start fs2ost $fs2ostdev $OST_MOUNT_OPTS
2290 start fs3ost $fs3ostdev $OST_MOUNT_OPTS
2292 mount -t lustre $MGSNID:/${FSNAME2} $MOUNT2 || return 1
2294 sleep 5 # until 11778 fixed
2296 dd if=/dev/zero of=$MOUNT2/$tfile bs=1M count=7 || return 2
2298 BKTOTAL=`lctl get_param -n obdfilter.*.kbytestotal | awk 'BEGIN{total=0}; {total+=$1}; END{print total}'`
2299 BKFREE=`lctl get_param -n obdfilter.*.kbytesfree | awk 'BEGIN{free=0}; {free+=$1}; END{print free}'`
2300 BKAVAIL=`lctl get_param -n obdfilter.*.kbytesavail | awk 'BEGIN{avail=0}; {avail+=$1}; END{print avail}'`
2301 STRING=`df -P $MOUNT2 | tail -n 1 | awk '{print $2","$3","$4}'`
2302 DFTOTAL=`echo $STRING | cut -d, -f1`
2303 DFUSED=`echo $STRING | cut -d, -f2`
2304 DFAVAIL=`echo $STRING | cut -d, -f3`
2305 DFFREE=$(($DFTOTAL - $DFUSED))
2307 ALLOWANCE=$((64 * $OSTCOUNT))
2309 if [ $DFTOTAL -lt $(($BKTOTAL - $ALLOWANCE)) ] ||
2310 [ $DFTOTAL -gt $(($BKTOTAL + $ALLOWANCE)) ] ; then
2311 echo "**** FAIL: df total($DFTOTAL) mismatch OST total($BKTOTAL)"
2314 if [ $DFFREE -lt $(($BKFREE - $ALLOWANCE)) ] ||
2315 [ $DFFREE -gt $(($BKFREE + $ALLOWANCE)) ] ; then
2316 echo "**** FAIL: df free($DFFREE) mismatch OST free($BKFREE)"
2319 if [ $DFAVAIL -lt $(($BKAVAIL - $ALLOWANCE)) ] ||
2320 [ $DFAVAIL -gt $(($BKAVAIL + $ALLOWANCE)) ] ; then
2321 echo "**** FAIL: df avail($DFAVAIL) mismatch OST avail($BKAVAIL)"
2326 stop fs3ost -f || return 200
2327 stop fs2ost -f || return 201
2328 stop fs2mds -f || return 202
2329 unload_modules_conf || return 203
2332 run_test 36 "df report consistency on OSTs with different block size"
2335 local mntpt=$(facet_mntpt $SINGLEMDS)
2336 local mdsdev=$(mdsdevname ${SINGLEMDS//mds/})
2337 local mdsdev_sym="$TMP/sym_mdt.img"
2338 local opts=$MDS_MOUNT_OPTS
2341 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
2342 skip "Currently only applicable to ldiskfs-based MDTs"
2346 echo "MDS : $mdsdev"
2347 echo "SYMLINK : $mdsdev_sym"
2348 do_facet $SINGLEMDS rm -f $mdsdev_sym
2350 do_facet $SINGLEMDS ln -s $mdsdev $mdsdev_sym
2352 echo "mount symlink device - $mdsdev_sym"
2354 if ! do_facet $SINGLEMDS test -b $mdsdev; then
2355 opts=$(csa_add "$opts" -o loop)
2357 mount_op=$(do_facet $SINGLEMDS mount -v -t lustre $opts \
2358 $mdsdev_sym $mntpt 2>&1)
2361 echo mount_op=$mount_op
2363 do_facet $SINGLEMDS "umount -d $mntpt && rm -f $mdsdev_sym"
2365 if $(echo $mount_op | grep -q "unable to set tunable"); then
2366 error "set tunables failed for symlink device"
2369 [ $rc -eq 0 ] || error "mount symlink $mdsdev_sym failed! rc=$rc"
2373 run_test 37 "verify set tunables works for symlink device"
2375 test_38() { # bug 14222
2376 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
2377 skip "Only applicable to ldiskfs-based MDTs"
2385 FILES=`find $SRC -type f -mtime +1 | head -n $COUNT`
2386 log "copying $(echo $FILES | wc -w) files to $DIR/$tdir"
2388 tar cf - $FILES | tar xf - -C $DIR/$tdir || \
2389 error "copying $SRC to $DIR/$tdir"
2391 umount_client $MOUNT
2393 log "rename lov_objid file on MDS"
2394 rm -f $TMP/lov_objid.orig
2396 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
2397 do_facet $SINGLEMDS "$DEBUGFS -c -R \\\"dump lov_objid $TMP/lov_objid.orig\\\" $MDSDEV"
2398 do_facet $SINGLEMDS "$DEBUGFS -w -R \\\"rm lov_objid\\\" $MDSDEV"
2400 do_facet $SINGLEMDS "od -Ax -td8 $TMP/lov_objid.orig"
2401 # check create in mds_lov_connect
2405 [ $V ] && log "verifying $DIR/$tdir/$f"
2406 diff -q $f $DIR/$tdir/$f || ERROR=y
2408 do_facet $SINGLEMDS "$DEBUGFS -c -R \\\"dump lov_objid $TMP/lov_objid.new\\\" $MDSDEV"
2409 do_facet $SINGLEMDS "od -Ax -td8 $TMP/lov_objid.new"
2410 [ "$ERROR" = "y" ] && error "old and new files are different after connect" || true
2412 # check it's updates in sync
2413 umount_client $MOUNT
2416 do_facet $SINGLEMDS dd if=/dev/zero of=$TMP/lov_objid.clear bs=4096 count=1
2417 do_facet $SINGLEMDS "$DEBUGFS -w -R \\\"rm lov_objid\\\" $MDSDEV"
2418 do_facet $SINGLEMDS "$DEBUGFS -w -R \\\"write $TMP/lov_objid.clear lov_objid\\\" $MDSDEV "
2423 [ $V ] && log "verifying $DIR/$tdir/$f"
2424 diff -q $f $DIR/$tdir/$f || ERROR=y
2426 do_facet $SINGLEMDS "$DEBUGFS -c -R \\\"dump lov_objid $TMP/lov_objid.new1\\\" $MDSDEV"
2427 do_facet $SINGLEMDS "od -Ax -td8 $TMP/lov_objid.new1"
2428 umount_client $MOUNT
2430 [ "$ERROR" = "y" ] && error "old and new files are different after sync" || true
2432 log "files compared the same"
2435 run_test 38 "MDS recreates missing lov_objid file from OST data"
2441 perl $SRCDIR/leak_finder.pl $TMP/debug 2>&1 | egrep '*** Leak:' &&
2442 error "memory leak detected" || true
2444 run_test 39 "leak_finder recognizes both LUSTRE and LNET malloc messages"
2446 test_40() { # bug 15759
2448 #define OBD_FAIL_TGT_TOOMANY_THREADS 0x706
2449 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0x80000706"
2453 run_test 40 "race during service thread startup"
2455 test_41a() { #bug 14134
2456 if [ $(facet_fstype $SINGLEMDS) == ldiskfs ] &&
2457 ! do_facet $SINGLEMDS test -b $(mdsdevname 1); then
2458 skip "Loop devices does not work with nosvc option"
2463 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
2465 start $SINGLEMDS $MDSDEV $MDS_MOUNT_OPTS -o nosvc -n
2466 start ost1 `ostdevname 1` $OST_MOUNT_OPTS
2467 start $SINGLEMDS $MDSDEV $MDS_MOUNT_OPTS -o nomgs,force
2469 mount_client $MOUNT || return 1
2472 echo "blah blah" > $MOUNT/$tfile
2475 umount_client $MOUNT
2476 stop ost1 -f || return 201
2477 stop_mds -f || return 202
2478 stop_mds -f || return 203
2479 unload_modules_conf || return 204
2482 run_test 41a "mount mds with --nosvc and --nomgs"
2485 if [ $(facet_fstype $SINGLEMDS) == ldiskfs ] &&
2486 ! do_facet $SINGLEMDS test -b $(mdsdevname 1); then
2487 skip "Loop devices does not work with nosvc option"
2491 ! combined_mgs_mds && skip "needs combined mgs device" && return 0
2495 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
2497 start $SINGLEMDS $MDSDEV $MDS_MOUNT_OPTS -o nosvc -n
2499 start $SINGLEMDS $MDSDEV $MDS_MOUNT_OPTS -o nomgs,force
2501 mount_client $MOUNT || return 1
2504 echo "blah blah" > $MOUNT/$tfile
2505 cat $MOUNT/$tfile || return 200
2507 umount_client $MOUNT
2508 stop_ost || return 201
2509 stop_mds -f || return 202
2510 stop_mds -f || return 203
2513 run_test 41b "mount mds with --nosvc and --nomgs on first mount"
2515 test_42() { #bug 14693
2517 check_mount || error "client was not mounted"
2519 do_facet mgs $LCTL conf_param $FSNAME.llite.some_wrong_param=10
2520 umount_client $MOUNT ||
2521 error "unmounting client failed with invalid llite param"
2522 mount_client $MOUNT ||
2523 error "mounting client failed with invalid llite param"
2525 do_facet mgs $LCTL conf_param $FSNAME.sys.some_wrong_param=20
2526 cleanup || error "stopping $FSNAME failed with invalid sys param"
2529 check_mount || "client was not mounted with invalid sys param"
2530 cleanup || error "stopping $FSNAME failed with invalid sys param"
2533 run_test 42 "allow client/server mount/unmount with invalid config param"
2536 [ $UID -ne 0 -o $RUNAS_ID -eq 0 ] && skip_env "run as root"
2538 chmod ugo+x $DIR || error "chmod 0 failed"
2539 set_conf_param_and_check mds \
2540 "lctl get_param -n mdt.$FSNAME-MDT0000.root_squash" \
2541 "$FSNAME.mdt.root_squash" \
2543 set_conf_param_and_check mds \
2544 "lctl get_param -n mdt.$FSNAME-MDT0000.nosquash_nids" \
2545 "$FSNAME.mdt.nosquash_nids" \
2549 # create set of test files
2551 echo "111" > $DIR/$tfile-userfile || error "write 1 failed"
2552 chmod go-rw $DIR/$tfile-userfile || error "chmod 1 failed"
2553 chown $RUNAS_ID.$RUNAS_ID $DIR/$tfile-userfile || error "chown failed"
2555 echo "222" > $DIR/$tfile-rootfile || error "write 2 failed"
2556 chmod go-rw $DIR/$tfile-rootfile || error "chmod 2 faield"
2558 mkdir $DIR/$tdir-rootdir -p || error "mkdir failed"
2559 chmod go-rwx $DIR/$tdir-rootdir || error "chmod 3 failed"
2560 touch $DIR/$tdir-rootdir/tfile-1 || error "touch failed"
2563 # check root_squash:
2564 # set root squash UID:GID to RUNAS_ID
2565 # root should be able to access only files owned by RUNAS_ID
2567 set_conf_param_and_check mds \
2568 "lctl get_param -n mdt.$FSNAME-MDT0000.root_squash" \
2569 "$FSNAME.mdt.root_squash" \
2570 "$RUNAS_ID:$RUNAS_ID"
2572 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-userfile)
2573 dd if=$DIR/$tfile-userfile 1>/dev/null 2>/dev/null || \
2574 error "$ST: root read permission is denied"
2575 echo "$ST: root read permission is granted - ok"
2578 dd conv=notrunc if=$DIR/$tfile-userfile 1>/dev/null 2>/dev/null || \
2579 error "$ST: root write permission is denied"
2580 echo "$ST: root write permission is granted - ok"
2582 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-rootfile)
2583 dd if=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null && \
2584 error "$ST: root read permission is granted"
2585 echo "$ST: root read permission is denied - ok"
2588 dd conv=notrunc of=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null && \
2589 error "$ST: root write permission is granted"
2590 echo "$ST: root write permission is denied - ok"
2592 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tdir-rootdir)
2593 rm $DIR/$tdir-rootdir/tfile-1 1>/dev/null 2>/dev/null && \
2594 error "$ST: root unlink permission is granted"
2595 echo "$ST: root unlink permission is denied - ok"
2597 touch $DIR/tdir-rootdir/tfile-2 1>/dev/null 2>/dev/null && \
2598 error "$ST: root create permission is granted"
2599 echo "$ST: root create permission is denied - ok"
2602 # check nosquash_nids:
2603 # put client's NID into nosquash_nids list,
2604 # root should be able to access root file after that
2606 local NIDLIST=$(lctl list_nids all | tr '\n' ' ')
2607 NIDLIST="2@elan $NIDLIST 192.168.0.[2,10]@tcp"
2608 NIDLIST=$(echo $NIDLIST | tr -s ' ' ' ')
2609 set_conf_param_and_check mds \
2610 "lctl get_param -n mdt.$FSNAME-MDT0000.nosquash_nids" \
2611 "$FSNAME-MDTall.mdt.nosquash_nids" \
2614 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-rootfile)
2615 dd if=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null || \
2616 error "$ST: root read permission is denied"
2617 echo "$ST: root read permission is granted - ok"
2620 dd conv=notrunc of=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null || \
2621 error "$ST: root write permission is denied"
2622 echo "$ST: root write permission is granted - ok"
2624 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tdir-rootdir)
2625 rm $DIR/$tdir-rootdir/tfile-1 || \
2626 error "$ST: root unlink permission is denied"
2627 echo "$ST: root unlink permission is granted - ok"
2628 touch $DIR/$tdir-rootdir/tfile-2 || \
2629 error "$ST: root create permission is denied"
2630 echo "$ST: root create permission is granted - ok"
2634 run_test 43 "check root_squash and nosquash_nids"
2636 umount_client $MOUNT
2641 check_mount || return 2
2642 UUID=$($LCTL get_param llite.${FSNAME}*.uuid | cut -d= -f2)
2644 UUIDS=$(do_facet $SINGLEMDS "$LCTL get_param mdt.${FSNAME}*.exports.*.uuid")
2645 for VAL in $UUIDS; do
2646 NID=$(echo $VAL | cut -d= -f1)
2647 CLUUID=$(echo $VAL | cut -d= -f2)
2648 [ "$UUID" = "$CLUUID" ] && STATS_FOUND=yes && break
2650 [ "$STATS_FOUND" = "no" ] && error "stats not found for client"
2654 run_test 44 "mounted client proc entry exists"
2658 check_mount || return 2
2663 #define OBD_FAIL_PTLRPC_LONG_UNLINK 0x50f
2664 do_facet client "lctl set_param fail_loc=0x50f"
2667 manual_umount_client --force || return 3
2668 do_facet client "lctl set_param fail_loc=0x0"
2670 mount_client $MOUNT || return 4
2674 run_test 45 "long unlink handling in ptlrpcd"
2681 umount_client $MOUNT2 || rc=$?
2682 umount_client $MOUNT || rc=$?
2683 while [ $count -gt 0 ]; do
2684 stop ost${count} -f || rc=$?
2688 cleanup_nocli || rc=$?
2689 #writeconf to remove all ost2 traces for subsequent tests
2690 writeconf_or_reformat
2695 echo "Testing with $OSTCOUNT OSTs"
2697 start_mds || return 1
2698 #first client should see only one ost
2699 start_ost || return 2
2700 wait_osc_import_state mds ost FULL
2702 mount_client $MOUNT || return 3
2703 trap "cleanup_46a $OSTCOUNT" EXIT ERR
2706 for (( i=2; i<=$OSTCOUNT; i++ )); do
2707 start ost$i `ostdevname $i` $OST_MOUNT_OPTS || return $((i+2))
2710 # wait until osts in sync
2711 for (( i=2; i<=$OSTCOUNT; i++ )); do
2712 wait_osc_import_state mds ost$i FULL
2713 wait_osc_import_state client ost$i FULL
2716 #second client see all ost's
2718 mount_client $MOUNT2 || return 8
2719 $LFS setstripe -c -1 $MOUNT2 || return 9
2720 $LFS getstripe $MOUNT2 || return 10
2722 echo "ok" > $MOUNT2/widestripe
2723 $LFS getstripe $MOUNT2/widestripe || return 11
2724 # fill acl buffer for avoid expand lsm to them
2725 awk -F : '{if (FNR < 25) { print "u:"$1":rwx" }}' /etc/passwd | while read acl; do
2726 setfacl -m $acl $MOUNT2/widestripe
2730 stat $MOUNT/widestripe || return 12
2732 cleanup_46a $OSTCOUNT || { echo "cleanup_46a failed!" && return 13; }
2735 run_test 46a "handle ost additional - wide striped file"
2740 check_mount || return 2
2741 $LCTL set_param ldlm.namespaces.$FSNAME-*-*-*.lru_size=100
2745 for ns in $($LCTL get_param ldlm.namespaces.$FSNAME-*-*-*.lru_size); do
2746 if echo $ns | grep "MDT[[:digit:]]*"; then
2749 lrs=$(echo $ns | sed 's/.*lru_size=//')
2750 lru_size[count]=$lrs
2755 facet_failover $SINGLEMDS
2756 client_up || return 3
2759 for ns in $($LCTL get_param ldlm.namespaces.$FSNAME-*-*-*.lru_size); do
2760 if echo $ns | grep "MDT[[:digit:]]*"; then
2763 lrs=$(echo $ns | sed 's/.*lru_size=//')
2764 if ! test "$lrs" -eq "${lru_size[count]}"; then
2765 n=$(echo $ns | sed -e 's/ldlm.namespaces.//' -e 's/.lru_size=.*//')
2766 error "$n has lost lru_size: $lrs vs. ${lru_size[count]}"
2774 run_test 47 "server restart does not make client loss lru_resize settings"
2779 # reformat after this test is needed - if test will failed
2780 # we will have unkillable file at FS
2784 test_48() { # bug 17636
2787 check_mount || return 2
2789 $LFS setstripe -c -1 $MOUNT || return 9
2790 $LFS getstripe $MOUNT || return 10
2792 echo "ok" > $MOUNT/widestripe
2793 $LFS getstripe $MOUNT/widestripe || return 11
2795 trap cleanup_48 EXIT ERR
2797 # fill acl buffer for avoid expand lsm to them
2798 getent passwd | awk -F : '{ print "u:"$1":rwx" }' | while read acl; do
2799 setfacl -m $acl $MOUNT/widestripe
2802 stat $MOUNT/widestripe || return 12
2807 run_test 48 "too many acls on file"
2809 # check PARAM_SYS_LDLM_TIMEOUT option of MKFS.LUSTRE
2810 test_49a() { # bug 17710
2811 local timeout_orig=$TIMEOUT
2812 local ldlm_timeout_orig=$LDLM_TIMEOUT
2813 local LOCAL_TIMEOUT=20
2815 LDLM_TIMEOUT=$LOCAL_TIMEOUT
2816 TIMEOUT=$LOCAL_TIMEOUT
2820 check_mount || error "client mount failed"
2822 echo "check ldlm_timout..."
2823 local LDLM_MDS="$(do_facet $SINGLEMDS lctl get_param -n ldlm_timeout)"
2824 local LDLM_OST1="$(do_facet ost1 lctl get_param -n ldlm_timeout)"
2825 local LDLM_CLIENT="$(do_facet client lctl get_param -n ldlm_timeout)"
2827 if [ $LDLM_MDS -ne $LDLM_OST1 -o $LDLM_MDS -ne $LDLM_CLIENT ]; then
2828 error "Different LDLM_TIMEOUT:$LDLM_MDS $LDLM_OST1 $LDLM_CLIENT"
2831 if [ $LDLM_MDS -ne $((LOCAL_TIMEOUT / 3)) ]; then
2832 error "LDLM_TIMEOUT($LDLM_MDS) is not $((LOCAL_TIMEOUT / 3))"
2835 umount_client $MOUNT
2836 stop_ost || error "problem stopping OSS"
2837 stop_mds || error "problem stopping MDS"
2839 LDLM_TIMEOUT=$ldlm_timeout_orig
2840 TIMEOUT=$timeout_orig
2842 run_test 49a "check PARAM_SYS_LDLM_TIMEOUT option of mkfs.lustre"
2844 test_49b() { # bug 17710
2845 local timeout_orig=$TIMEOUT
2846 local ldlm_timeout_orig=$LDLM_TIMEOUT
2847 local LOCAL_TIMEOUT=20
2849 LDLM_TIMEOUT=$((LOCAL_TIMEOUT - 1))
2850 TIMEOUT=$LOCAL_TIMEOUT
2854 check_mount || error "client mount failed"
2856 local LDLM_MDS="$(do_facet $SINGLEMDS lctl get_param -n ldlm_timeout)"
2857 local LDLM_OST1="$(do_facet ost1 lctl get_param -n ldlm_timeout)"
2858 local LDLM_CLIENT="$(do_facet client lctl get_param -n ldlm_timeout)"
2860 if [ $LDLM_MDS -ne $LDLM_OST1 -o $LDLM_MDS -ne $LDLM_CLIENT ]; then
2861 error "Different LDLM_TIMEOUT:$LDLM_MDS $LDLM_OST1 $LDLM_CLIENT"
2864 if [ $LDLM_MDS -ne $((LOCAL_TIMEOUT - 1)) ]; then
2865 error "LDLM_TIMEOUT($LDLM_MDS) is not $((LOCAL_TIMEOUT - 1))"
2868 cleanup || error "cleanup failed"
2870 LDLM_TIMEOUT=$ldlm_timeout_orig
2871 TIMEOUT=$timeout_orig
2873 run_test 49b "check PARAM_SYS_LDLM_TIMEOUT option of mkfs.lustre"
2876 # Test both statfs and lfs df and fail if either one fails
2877 multiop_bg_pause $1 f_
2880 killall -USR1 multiop
2881 [ $RC1 -ne 0 ] && log "lazystatfs multiop failed"
2882 wait $PID || { RC1=$?; log "multiop return error "; }
2889 if [ $RC2 -eq 0 ]; then
2891 log "lazystatfs df failed"
2895 [[ $RC1 -ne 0 || $RC2 -eq 0 ]] && RC=1
2901 lctl set_param llite.$FSNAME-*.lazystatfs=1
2904 lazystatfs $MOUNT || error "lazystatfs failed but no down servers"
2906 cleanup || return $?
2908 run_test 50a "lazystatfs all servers available =========================="
2912 lctl set_param llite.$FSNAME-*.lazystatfs=1
2915 # Wait for client to detect down OST
2916 stop_ost || error "Unable to stop OST1"
2917 wait_osc_import_state mds ost DISCONN
2919 lazystatfs $MOUNT || error "lazystatfs should don't have returned EIO"
2921 umount_client $MOUNT || error "Unable to unmount client"
2922 stop_mds || error "Unable to stop MDS"
2924 run_test 50b "lazystatfs all servers down =========================="
2927 start_mds || error "Unable to start MDS"
2928 start_ost || error "Unable to start OST1"
2929 start_ost2 || error "Unable to start OST2"
2930 mount_client $MOUNT || error "Unable to mount client"
2931 lctl set_param llite.$FSNAME-*.lazystatfs=1
2934 # Wait for client to detect down OST
2935 stop_ost || error "Unable to stop OST1"
2936 wait_osc_import_state mds ost DISCONN
2937 lazystatfs $MOUNT || error "lazystatfs failed with one down server"
2939 umount_client $MOUNT || error "Unable to unmount client"
2940 stop_ost2 || error "Unable to stop OST2"
2941 stop_mds || error "Unable to stop MDS"
2942 #writeconf to remove all ost2 traces for subsequent tests
2943 writeconf_or_reformat
2945 run_test 50c "lazystatfs one server down =========================="
2948 start_mds || error "Unable to start MDS"
2949 start_ost || error "Unable to start OST1"
2950 start_ost2 || error "Unable to start OST2"
2951 mount_client $MOUNT || error "Unable to mount client"
2952 lctl set_param llite.$FSNAME-*.lazystatfs=1
2955 # Issue the statfs during the window where the client still
2956 # belives the OST to be available but it is in fact down.
2957 # No failure just a statfs which hangs for a timeout interval.
2958 stop_ost || error "Unable to stop OST1"
2959 lazystatfs $MOUNT || error "lazystatfs failed with one down server"
2961 umount_client $MOUNT || error "Unable to unmount client"
2962 stop_ost2 || error "Unable to stop OST2"
2963 stop_mds || error "Unable to stop MDS"
2964 #writeconf to remove all ost2 traces for subsequent tests
2965 writeconf_or_reformat
2967 run_test 50d "lazystatfs client/server conn race =========================="
2974 start_mds || return 1
2975 #first client should see only one ost
2976 start_ost || return 2
2977 wait_osc_import_state mds ost FULL
2979 # Wait for client to detect down OST
2980 stop_ost || error "Unable to stop OST1"
2981 wait_osc_import_state mds ost DISCONN
2983 mount_client $MOUNT || error "Unable to mount client"
2984 lctl set_param llite.$FSNAME-*.lazystatfs=0
2986 multiop_bg_pause $MOUNT _f
2990 if [ $RC1 -ne 0 ]; then
2991 log "multiop failed $RC1"
2994 sleep $(( $TIMEOUT+1 ))
2996 [ $? -ne 0 ] && error "process isn't sleep"
2997 start_ost || error "Unable to start OST1"
2998 wait $pid || error "statfs failed"
3001 umount_client $MOUNT || error "Unable to unmount client"
3002 stop_ost || error "Unable to stop OST1"
3003 stop_mds || error "Unable to stop MDS"
3005 run_test 50e "normal statfs all servers down =========================="
3010 CONN_PROC="osc.$FSNAME-OST0001-osc-[M]*.ost_server_uuid"
3012 start_mds || error "Unable to start mds"
3013 #first client should see only one ost
3014 start_ost || error "Unable to start OST1"
3015 wait_osc_import_state mds ost FULL
3017 start_ost2 || error "Unable to start OST2"
3018 wait_osc_import_state mds ost2 FULL
3020 # Wait for client to detect down OST
3021 stop_ost2 || error "Unable to stop OST2"
3023 wait_osc_import_state mds ost2 DISCONN
3024 mount_client $MOUNT || error "Unable to mount client"
3025 lctl set_param llite.$FSNAME-*.lazystatfs=0
3027 multiop_bg_pause $MOUNT _f
3031 if [ $RC1 -ne 0 ]; then
3032 log "lazystatfs multiop failed $RC1"
3035 sleep $(( $TIMEOUT+1 ))
3037 [ $? -ne 0 ] && error "process isn't sleep"
3038 start_ost2 || error "Unable to start OST2"
3039 wait $pid || error "statfs failed"
3040 stop_ost2 || error "Unable to stop OST2"
3043 umount_client $MOUNT || error "Unable to unmount client"
3044 stop_ost || error "Unable to stop OST1"
3045 stop_mds || error "Unable to stop MDS"
3046 #writeconf to remove all ost2 traces for subsequent tests
3047 writeconf_or_reformat
3049 run_test 50f "normal statfs one server in down =========================="
3052 [ "$OSTCOUNT" -lt "2" ] && skip_env "$OSTCOUNT < 2, skipping" && return
3054 start_ost2 || error "Unable to start OST2"
3055 wait_osc_import_state mds ost2 FULL
3056 wait_osc_import_state client ost2 FULL
3058 local PARAM="${FSNAME}-OST0001.osc.active"
3060 $LFS setstripe -c -1 $DIR/$tfile || error "Unable to lfs setstripe"
3061 do_facet mgs $LCTL conf_param $PARAM=0 || error "Unable to deactivate OST"
3063 umount_client $MOUNT || error "Unable to unmount client"
3064 mount_client $MOUNT || error "Unable to mount client"
3065 # This df should not cause a panic
3068 do_facet mgs $LCTL conf_param $PARAM=1 || error "Unable to activate OST"
3070 umount_client $MOUNT || error "Unable to unmount client"
3071 stop_ost2 || error "Unable to stop OST2"
3072 stop_ost || error "Unable to stop OST1"
3073 stop_mds || error "Unable to stop MDS"
3074 #writeconf to remove all ost2 traces for subsequent tests
3075 writeconf_or_reformat
3077 run_test 50g "deactivated OST should not cause panic====================="
3081 # prepare MDT/OST, make OSC inactive for OST1
3082 [ "$OSTCOUNT" -lt "2" ] && skip_env "$OSTCOUNT < 2, skipping" && return
3083 do_facet ost1 "$TUNEFS --param osc.active=0 `ostdevname 1`" ||
3084 error "tunefs OST1 failed"
3085 start_mds || error "Unable to start MDT"
3086 start_ost || error "Unable to start OST1"
3087 start_ost2 || error "Unable to start OST2"
3088 mount_client $MOUNT || error "client start failed"
3092 # activatate OSC for OST1
3093 local TEST="$LCTL get_param -n osc.${FSNAME}-OST0000-osc-[!M]*.active"
3094 set_conf_param_and_check client \
3095 "$TEST" "${FSNAME}-OST0000.osc.active" 1 ||
3096 error "Unable to activate OST1"
3098 mkdir -p $DIR/$tdir/2
3099 $LFS setstripe -c -1 -i 0 $DIR/$tdir/2
3100 sleep 1 && echo "create a file after OST1 is activated"
3102 createmany -o $DIR/$tdir/2/$tfile-%d 1
3104 # check OSC import is working
3105 stat $DIR/$tdir/2/* >/dev/null 2>&1 ||
3106 error "some OSC imports are still not connected"
3109 umount_client $MOUNT || error "Unable to umount client"
3110 stop_ost2 || error "Unable to stop OST2"
3113 run_test 50h "LU-642: activate deactivated OST ==="
3116 local LOCAL_TIMEOUT=20
3120 check_mount || return 1
3123 $LFS setstripe -c -1 $MOUNT/d1
3124 #define OBD_FAIL_MDS_REINT_DELAY 0x142
3125 do_facet $SINGLEMDS "lctl set_param fail_loc=0x142"
3126 touch $MOUNT/d1/f1 &
3129 start_ost2 || return 2
3131 stop_ost2 || return 3
3133 #writeconf to remove all ost2 traces for subsequent tests
3134 writeconf_or_reformat
3136 run_test 51 "Verify that mdt_reint handles RMF_MDT_MD correctly when an OST is added"
3145 do_node $node mkdir -p $dest
3146 [ $? -eq 0 ] || { error "Unable to create directory"; return 1; }
3148 do_node $node 'tar cf - '$@' | tar xf - -C '$dest';
3149 [ \"\${PIPESTATUS[*]}\" = \"0 0\" ] || exit 1'
3150 [ $? -eq 0 ] || { error "Unable to tar files"; return 2; }
3152 do_node $node 'getfattr -d -m "[a-z]*\\." '$@' > '$xattrs
3153 [ $? -eq 0 ] || { error "Unable to read xattrs"; return 3; }
3163 local backup2=${TMP}/backup2
3165 do_node $node mkdir -p $backup2
3166 [ $? -eq 0 ] || { error "Unable to create directory"; return 1; }
3168 do_node $node 'tar cf - '$@' | tar xf - -C '$backup2';
3169 [ \"\${PIPESTATUS[*]}\" = \"0 0\" ] || exit 1'
3170 [ $? -eq 0 ] || { error "Unable to tar files to diff"; return 2; }
3172 do_node $node "diff -rq $backup $backup2"
3173 [ $? -eq 0 ] || { error "contents differ"; return 3; }
3175 local xattrs2=${TMP}/xattrs2
3176 do_node $node 'getfattr -d -m "[a-z]*\\." '$@' > '$xattrs2
3177 [ $? -eq 0 ] || { error "Unable to read xattrs to diff"; return 4; }
3179 do_node $node "diff $xattrs $xattrs2"
3180 [ $? -eq 0 ] || { error "xattrs differ"; return 5; }
3182 do_node $node "rm -rf $backup2 $xattrs2"
3183 [ $? -eq 0 ] || { error "Unable to delete temporary files"; return 6; }
3187 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
3188 skip "Only applicable to ldiskfs-based MDTs"
3193 [ $? -eq 0 ] || { error "Unable to start MDS"; return 1; }
3195 [ $? -eq 0 ] || { error "Unable to start OST1"; return 2; }
3197 [ $? -eq 0 ] || { error "Unable to mount client"; return 3; }
3200 local ost1mnt=$(facet_mntpt ost1)
3201 local ost1node=$(facet_active_host ost1)
3202 local ost1tmp=$TMP/conf52
3206 [ $? -eq 0 ] || { error "Unable to create tdir"; return 4; }
3207 touch $TMP/modified_first
3208 [ $? -eq 0 ] || { error "Unable to create temporary file"; return 5; }
3209 local mtime=$(stat -c %Y $TMP/modified_first)
3210 do_node $ost1node "mkdir -p $ost1tmp && touch -m -d @$mtime $ost1tmp/modified_first"
3212 [ $? -eq 0 ] || { error "Unable to create temporary file"; return 6; }
3215 $LFS setstripe -c -1 -S 1M $DIR/$tdir
3216 [ $? -eq 0 ] || { error "lfs setstripe failed"; return 7; }
3218 for (( i=0; i < nrfiles; i++ )); do
3219 multiop $DIR/$tdir/$tfile-$i Ow1048576w1048576w524288c
3220 [ $? -eq 0 ] || { error "multiop failed"; return 8; }
3226 echo backup files to $TMP/files
3227 local files=$(find $DIR/$tdir -type f -newer $TMP/modified_first)
3228 copy_files_xattrs `hostname` $TMP/files $TMP/file_xattrs $files
3229 [ $? -eq 0 ] || { error "Unable to copy files"; return 9; }
3231 umount_client $MOUNT
3232 [ $? -eq 0 ] || { error "Unable to umount client"; return 10; }
3234 [ $? -eq 0 ] || { error "Unable to stop ost1"; return 11; }
3236 echo mount ost1 as ldiskfs
3237 do_node $ost1node mkdir -p $ost1mnt
3238 [ $? -eq 0 ] || { error "Unable to create $ost1mnt"; return 23; }
3239 if ! do_node $ost1node test -b $ost1_dev; then
3242 do_node $ost1node mount -t $(facet_fstype ost1) $loop $ost1_dev \
3244 [ $? -eq 0 ] || { error "Unable to mount ost1 as ldiskfs"; return 12; }
3247 echo backup objects to $ost1tmp/objects
3248 local objects=$(do_node $ost1node 'find '$ost1mnt'/O/[0-9]* -type f'\
3249 '-size +0 -newer '$ost1tmp'/modified_first -regex ".*\/[0-9]+"')
3250 copy_files_xattrs $ost1node $ost1tmp/objects $ost1tmp/object_xattrs \
3252 [ $? -eq 0 ] || { error "Unable to copy objects"; return 13; }
3254 # move objects to lost+found
3255 do_node $ost1node 'mv '$objects' '${ost1mnt}'/lost+found'
3256 [ $? -eq 0 ] || { error "Unable to move objects"; return 14; }
3259 do_node $ost1node "ll_recover_lost_found_objs -d $ost1mnt/lost+found"
3260 [ $? -eq 0 ] || { error "ll_recover_lost_found_objs failed"; return 15; }
3262 # compare restored objects against saved ones
3263 diff_files_xattrs $ost1node $ost1tmp/objects $ost1tmp/object_xattrs $objects
3264 [ $? -eq 0 ] || { error "Unable to diff objects"; return 16; }
3266 do_node $ost1node "umount $ost1mnt"
3267 [ $? -eq 0 ] || { error "Unable to umount ost1 as ldiskfs"; return 17; }
3270 [ $? -eq 0 ] || { error "Unable to start ost1"; return 18; }
3272 [ $? -eq 0 ] || { error "Unable to mount client"; return 19; }
3275 diff_files_xattrs `hostname` $TMP/files $TMP/file_xattrs $files
3276 [ $? -eq 0 ] || { error "Unable to diff files"; return 20; }
3278 rm -rf $TMP/files $TMP/file_xattrs
3279 [ $? -eq 0 ] || { error "Unable to delete temporary files"; return 21; }
3280 do_node $ost1node "rm -rf $ost1tmp"
3281 [ $? -eq 0 ] || { error "Unable to delete temporary files"; return 22; }
3284 run_test 52 "check recovering objects from lost+found"
3286 # Checks threads_min/max/started for some service
3288 # Arguments: service name (OST or MDT), facet (e.g., ost1, $SINGLEMDS), and a
3289 # parameter pattern prefix like 'ost.*.ost'.
3302 local msg="Insane $modname thread counts"
3303 local ncpts=$(check_cpt_number $facet)
3307 check_mount || return 41
3309 # We need to expand $parampat, but it may match multiple parameters, so
3310 # we'll pick the first one
3311 if ! paramp=$(do_facet $facet "lctl get_param -N ${parampat}.threads_min"|head -1); then
3312 error "Couldn't expand ${parampat}.threads_min parameter name"
3316 # Remove the .threads_min part
3317 paramp=${paramp%.threads_min}
3319 # Check for sanity in defaults
3320 tmin=$(do_facet $facet "lctl get_param -n ${paramp}.threads_min" || echo 0)
3321 tmax=$(do_facet $facet "lctl get_param -n ${paramp}.threads_max" || echo 0)
3322 tstarted=$(do_facet $facet "lctl get_param -n ${paramp}.threads_started" || echo 0)
3323 lassert 23 "$msg (PDSH problems?)" '(($tstarted && $tmin && $tmax))' || return $?
3324 lassert 24 "$msg" '(($tstarted >= $tmin && $tstarted <= $tmax ))' || return $?
3325 nthrs=$(expr $tmax - $tmin)
3326 if [ $nthrs -lt $ncpts ]; then
3332 [ $tmin -eq $tmax -a $tmin -eq $tstarted ] &&
3333 skip_env "module parameter forced $facet thread count" &&
3334 tmin=3 && tmax=$((3 * tmax))
3336 # Check that we can change min/max
3337 do_facet $facet "lctl set_param ${paramp}.threads_min=$((tmin + nthrs))"
3338 do_facet $facet "lctl set_param ${paramp}.threads_max=$((tmax - nthrs))"
3339 tmin2=$(do_facet $facet "lctl get_param -n ${paramp}.threads_min" || echo 0)
3340 tmax2=$(do_facet $facet "lctl get_param -n ${paramp}.threads_max" || echo 0)
3341 lassert 25 "$msg" '(($tmin2 == ($tmin + $nthrs) && $tmax2 == ($tmax - $nthrs)))' || return $?
3343 # Check that we can set min/max to the same value
3344 tmin=$(do_facet $facet "lctl get_param -n ${paramp}.threads_min" || echo 0)
3345 do_facet $facet "lctl set_param ${paramp}.threads_max=$tmin"
3346 tmin2=$(do_facet $facet "lctl get_param -n ${paramp}.threads_min" || echo 0)
3347 tmax2=$(do_facet $facet "lctl get_param -n ${paramp}.threads_max" || echo 0)
3348 lassert 26 "$msg" '(($tmin2 == $tmin && $tmax2 == $tmin))' || return $?
3350 # Check that we can't set max < min
3351 do_facet $facet "lctl set_param ${paramp}.threads_max=$((tmin - 1))"
3352 tmin2=$(do_facet $facet "lctl get_param -n ${paramp}.threads_min" || echo 0)
3353 tmax2=$(do_facet $facet "lctl get_param -n ${paramp}.threads_max" || echo 0)
3354 lassert 27 "$msg" '(($tmin2 <= $tmax2))' || return $?
3356 # We need to ensure that we get the module options desired; to do this
3357 # we set LOAD_MODULES_REMOTE=true and we call setmodopts below.
3358 LOAD_MODULES_REMOTE=true
3361 local newvalue="${opts}=$(expr $basethr \* $ncpts)"
3362 setmodopts -a $modname "$newvalue" oldvalue
3366 check_mount || return 41
3368 # Restore previous setting of MODOPTS_*
3369 setmodopts $modname "$oldvalue"
3371 # Check that $opts took
3372 tmin=$(do_facet $facet "lctl get_param -n ${paramp}.threads_min")
3373 tmax=$(do_facet $facet "lctl get_param -n ${paramp}.threads_max")
3374 tstarted=$(do_facet $facet "lctl get_param -n ${paramp}.threads_started")
3375 lassert 28 "$msg" '(($tstarted >= $tmin && $tstarted <= $tmax ))' || return $?
3384 thread_sanity OST ost1 'ost.*.ost' 'oss_num_threads' '16'
3387 run_test 53a "check OSS thread count params"
3391 local mds=$(do_facet $SINGLEMDS "lctl get_param -N mds.*.*.threads_max \
3393 if [ -z "$mds" ]; then
3394 #running this on an old MDT
3395 thread_sanity MDT $SINGLEMDS 'mdt.*.*.' 'mdt_num_threads' 16
3397 thread_sanity MDT $SINGLEMDS 'mds.*.*.' 'mds_num_threads' 16
3401 run_test 53b "check MDS thread count params"
3404 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
3405 skip "Only applicable to ldiskfs-based MDTs"
3409 do_rpc_nodes $(facet_host ost1) run_llverdev $(ostdevname 1) -p
3410 [ $? -eq 0 ] || error "llverdev failed!"
3413 run_test 54a "test llverdev and partial verify of device"
3416 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
3417 skip "Only applicable to ldiskfs-based MDTs"
3422 run_llverfs $MOUNT -p
3423 [ $? -eq 0 ] || error "llverfs failed!"
3426 run_test 54b "test llverfs and partial verify of filesystem"
3430 local max_ost_index=$1
3431 echo -n $(((max_ost_index + 1) * 8))
3435 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
3436 skip "Only applicable to ldiskfs-based MDTs"
3440 local mdsdev=$(mdsdevname 1)
3441 local mdsvdev=$(mdsvdevname 1)
3445 add mds1 $(mkfs_opts mds1 ${mdsdev}) --reformat $mdsdev \
3447 add ost1 $(mkfs_opts ost1 $(ostdevname 1)) --index=$i \
3448 --reformat $(ostdevname 1) $(ostvdevname 1)
3454 echo checking size of lov_objid for ost index $i
3455 LOV_OBJID_SIZE=$(do_facet mds1 "$DEBUGFS -R 'stat lov_objid' $mdsdev 2>/dev/null" | grep ^User | awk '{print $6}')
3456 if [ "$LOV_OBJID_SIZE" != $(lov_objid_size $i) ]; then
3457 error "lov_objid size has to be $(lov_objid_size $i), not $LOV_OBJID_SIZE"
3459 echo ok, lov_objid size is correct: $LOV_OBJID_SIZE
3466 run_test 55 "check lov_objid size"
3469 local mds_journal_size_orig=$MDSJOURNALSIZE
3473 for num in $(seq 1 $MDSCOUNT); do
3474 add mds${num} $(mkfs_opts mds${num} $(mdsdevname $num)) \
3475 --reformat $(mdsdevname $num) $(mdsvdevname $num)
3477 add ost1 $(mkfs_opts ost1 $(ostdevname 1)) --index=1000 --reformat \
3478 $(ostdevname 1) $(ostvdevname 1)
3479 add ost2 $(mkfs_opts ost2 $(ostdevname 2)) --index=10000 --reformat \
3480 $(ostdevname 2) $(ostvdevname 2)
3484 start_ost2 || error "Unable to start second ost"
3485 mount_client $MOUNT || error "Unable to mount client"
3489 MDSJOURNALSIZE=$mds_journal_size_orig
3492 run_test 56 "check big indexes"
3494 test_57a() { # bug 22656
3495 local NID=$(do_facet ost1 "$LCTL get_param nis" | tail -1 | awk '{print $1}')
3496 writeconf_or_reformat
3497 [ $(facet_fstype ost1) == zfs ] && import_zpool ost1
3498 do_facet ost1 "$TUNEFS --failnode=$NID `ostdevname 1`" || error "tunefs failed"
3500 start_ost && error "OST registration from failnode should fail"
3503 run_test 57a "initial registration from failnode should fail (should return errs)"
3506 local NID=$(do_facet ost1 "$LCTL get_param nis" | tail -1 | awk '{print $1}')
3507 writeconf_or_reformat
3508 [ $(facet_fstype ost1) == zfs ] && import_zpool ost1
3509 do_facet ost1 "$TUNEFS --servicenode=$NID `ostdevname 1`" || error "tunefs failed"
3511 start_ost || error "OST registration from servicenode should not fail"
3514 run_test 57b "initial registration from servicenode should not fail"
3517 do_facet mgs $LCTL get_param mgs.MGS.live.$FSNAME | grep OST | wc -l
3520 test_58() { # bug 22658
3521 if [ $(facet_fstype mds) != ldiskfs ]; then
3522 skip "Only applicable to ldiskfs-based MDTs"
3527 createmany -o $DIR/$tdir/$tfile-%d 100
3528 # make sure that OSTs do not cancel llog cookies before we unmount the MDS
3529 #define OBD_FAIL_OBD_LOG_CANCEL_NET 0x601
3530 do_facet $SINGLEMDS "lctl set_param fail_loc=0x601"
3531 unlinkmany $DIR/$tdir/$tfile-%d 100
3534 local MNTDIR=$(facet_mntpt $SINGLEMDS)
3535 local devname=$(mdsdevname ${SINGLEMDS//mds/})
3537 if ! do_facet $SINGLEMDS "test -b $devname"; then
3541 # remove all files from the OBJECTS dir
3542 do_facet $SINGLEMDS "mount -t ldiskfs $opts $devname $MNTDIR"
3543 do_facet $SINGLEMDS "find $MNTDIR/O/1/d* -type f -delete"
3544 do_facet $SINGLEMDS "umount -d $MNTDIR"
3545 # restart MDS with missing llog files
3547 do_facet mds "lctl set_param fail_loc=0"
3550 run_test 58 "missing llog files must not prevent MDT from mounting"
3553 start_mgsmds >> /dev/null
3554 local C1=$(count_osts)
3555 if [ $C1 -eq 0 ]; then
3556 start_ost >> /dev/null
3560 echo "original ost count: $C1 (expect > 0)"
3561 [ $C1 -gt 0 ] || error "No OSTs in $FSNAME log"
3562 start_mgsmds -o writeconf >> /dev/null || error "MDT start failed"
3563 local C2=$(count_osts)
3564 echo "after mdt writeconf count: $C2 (expect 0)"
3565 [ $C2 -gt 0 ] && error "MDT writeconf should erase OST logs"
3566 echo "OST start without writeconf should fail:"
3567 start_ost >> /dev/null && error "OST start without writeconf didn't fail"
3568 echo "OST start with writeconf should succeed:"
3569 start_ost -o writeconf >> /dev/null || error "OST1 start failed"
3570 local C3=$(count_osts)
3571 echo "after ost writeconf count: $C3 (expect 1)"
3572 [ $C3 -eq 1 ] || error "new OST writeconf should add:"
3573 start_ost2 -o writeconf >> /dev/null || error "OST2 start failed"
3574 local C4=$(count_osts)
3575 echo "after ost2 writeconf count: $C4 (expect 2)"
3576 [ $C4 -eq 2 ] || error "OST2 writeconf should add log"
3577 stop_ost2 >> /dev/null
3578 cleanup_nocli >> /dev/null
3579 #writeconf to remove all ost2 traces for subsequent tests
3580 writeconf_or_reformat
3582 run_test 59 "writeconf mount option"
3584 test_60() { # LU-471
3587 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
3588 skip "Only applicable to ldiskfs-based MDTs"
3592 for num in $(seq $MDSCOUNT); do
3593 add mds${num} $(mkfs_opts mds${num} $(mdsdevname $num)) \
3594 --mkfsoptions='\" -E stride=64 -O ^uninit_bg\"' \
3595 --reformat $(mdsdevname $num) $(mdsvdevname $num) ||
3599 dump=$(do_facet $SINGLEMDS dumpe2fs $(mdsdevname 1))
3601 [ $rc -eq 0 ] || error "dumpe2fs $(mdsdevname 1) failed"
3603 # MDT default has dirdata feature
3604 echo $dump | grep dirdata > /dev/null || error "dirdata is not set"
3605 # we disable uninit_bg feature
3606 echo $dump | grep uninit_bg > /dev/null && error "uninit_bg is set"
3607 # we set stride extended options
3608 echo $dump | grep stride > /dev/null || error "stride is not set"
3611 run_test 60 "check mkfs.lustre --mkfsoptions -E -O options setting"
3614 local reformat=false
3616 [ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.1.53) ] ||
3617 { skip "Need MDS version at least 2.1.53"; return 0; }
3619 if [ $(facet_fstype $SINGLEMDS) == ldiskfs ] &&
3620 ! large_xattr_enabled; then
3622 LDISKFS_MKFS_OPTS+=" -O large_xattr"
3624 for num in $(seq $MDSCOUNT); do
3625 add mds${num} $(mkfs_opts mds$num $(mdsdevname $num)) \
3626 --reformat $(mdsdevname $num) $(mdsvdevname $num) ||
3627 error "add mds $num failed"
3631 setup_noconfig || error "setting up the filesystem failed"
3632 client_up || error "starting client failed"
3634 local file=$DIR/$tfile
3637 local large_value="$(generate_string $(max_xattr_size))"
3638 local small_value="bar"
3640 local name="trusted.big"
3641 log "save large xattr $name on $file"
3642 setfattr -n $name -v $large_value $file ||
3643 error "saving $name on $file failed"
3645 local new_value=$(get_xattr_value $name $file)
3646 [[ "$new_value" != "$large_value" ]] &&
3647 error "$name different after saving"
3649 log "shrink value of $name on $file"
3650 setfattr -n $name -v $small_value $file ||
3651 error "shrinking value of $name on $file failed"
3653 new_value=$(get_xattr_value $name $file)
3654 [[ "$new_value" != "$small_value" ]] &&
3655 error "$name different after shrinking"
3657 log "grow value of $name on $file"
3658 setfattr -n $name -v $large_value $file ||
3659 error "growing value of $name on $file failed"
3661 new_value=$(get_xattr_value $name $file)
3662 [[ "$new_value" != "$large_value" ]] &&
3663 error "$name different after growing"
3665 log "check value of $name on $file after remounting MDS"
3667 new_value=$(get_xattr_value $name $file)
3668 [[ "$new_value" != "$large_value" ]] &&
3669 error "$name different after remounting MDS"
3671 log "remove large xattr $name from $file"
3672 setfattr -x $name $file || error "removing $name from $file failed"
3677 LDISKFS_MKFS_OPTS=${LDISKFS_MKFS_OPTS% -O large_xattr}
3681 run_test 61 "large xattr"
3684 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
3685 skip "Only applicable to ldiskfs-based MDTs"
3690 local mdsdev=$(mdsdevname 1)
3691 local ostdev=$(ostdevname 1)
3693 [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.2.51) ]] ||
3694 { skip "Need MDS version at least 2.2.51"; return 0; }
3696 echo "disable journal for mds"
3697 do_facet mds tune2fs -O ^has_journal $mdsdev || error "tune2fs failed"
3698 start_mds && error "MDT start should fail"
3699 echo "disable journal for ost"
3700 do_facet ost1 tune2fs -O ^has_journal $ostdev || error "tune2fs failed"
3701 start_ost && error "OST start should fail"
3702 cleanup || return $?
3705 run_test 62 "start with disabled journal"
3708 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
3709 skip "Only applicable to ldiskfs-based MDTs"
3713 local inode_slab=$(do_facet $SINGLEMDS \
3714 "awk '/ldiskfs_inode_cache/ { print \\\$5 }' /proc/slabinfo")
3715 if [ -z "$inode_slab" ]; then
3716 skip "ldiskfs module has not been loaded"
3720 echo "$inode_slab ldisk inodes per page"
3721 [ "$inode_slab" -ge "3" ] ||
3722 error "ldisk inode size is too big, $inode_slab objs per page"
3725 run_test 63 "Verify each page can at least hold 3 ldisk inodes"
3730 start_ost2 || error "Unable to start second ost"
3731 mount_client $MOUNT || error "Unable to mount client"
3732 stop_ost2 || error "Unable to stop second ost"
3734 $LFS df --lazy || error "lfs df failed"
3735 cleanup || return $?
3736 #writeconf to remove all ost2 traces for subsequent tests
3737 writeconf_or_reformat
3739 run_test 64 "check lfs df --lazy "
3741 test_65() { # LU-2237
3742 # Currently, the test is only valid for ldiskfs backend
3743 [ "$(facet_fstype $SINGLEMDS)" != "ldiskfs" ] &&
3744 skip "non-ldiskfs backend" && return
3746 local devname=$(mdsdevname ${SINGLEMDS//mds/})
3747 local brpt=$(facet_mntpt brpt)
3750 if ! do_facet $SINGLEMDS "test -b $devname"; then
3755 local obj=$(do_facet $SINGLEMDS \
3756 "$DEBUGFS -c -R \\\"stat last_rcvd\\\" $devname" |
3758 if [ -z "$obj" ]; then
3759 # The MDT may be just re-formatted, mount the MDT for the
3760 # first time to guarantee the "last_rcvd" file is there.
3761 start_mds || error "fail to mount the MDS for the first time"
3765 # remove the "last_rcvd" file
3766 do_facet $SINGLEMDS "mkdir -p $brpt"
3767 do_facet $SINGLEMDS \
3768 "mount -t $(facet_fstype $SINGLEMDS) $opts $devname $brpt"
3769 do_facet $SINGLEMDS "rm -f ${brpt}/last_rcvd"
3770 do_facet $SINGLEMDS "umount -d $brpt"
3772 # restart MDS, the "last_rcvd" file should be recreated.
3773 start_mds || error "fail to restart the MDS"
3775 obj=$(do_facet $SINGLEMDS \
3776 "$DEBUGFS -c -R \\\"stat last_rcvd\\\" $devname" | grep Inode)
3777 [ -n "$obj" ] || error "fail to re-create the last_rcvd"
3779 run_test 65 "re-create the lost last_rcvd file when server mount"
3782 [[ $(lustre_version_code mgs) -ge $(version_code 2.3.59) ]] ||
3783 { skip "Need MGS version at least 2.3.59"; return 0; }
3786 local OST1_NID=$(do_facet ost1 $LCTL list_nids | head -1)
3787 local MDS_NID=$(do_facet $SINGLEMDS $LCTL list_nids | head -1)
3789 echo "replace_nids should fail if MDS, OSTs and clients are UP"
3790 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 $OST1_NID &&
3791 error "replace_nids fail"
3793 umount_client $MOUNT || error "unmounting client failed"
3794 echo "replace_nids should fail if MDS and OSTs are UP"
3795 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 $OST1_NID &&
3796 error "replace_nids fail"
3799 echo "replace_nids should fail if MDS is UP"
3800 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 $OST1_NID &&
3801 error "replace_nids fail"
3803 stop_mds || error "stopping mds failed"
3805 if combined_mgs_mds; then
3806 start_mdt 1 "-o nosvc" ||
3807 error "starting mds with nosvc option failed"
3810 echo "command should accept two parameters"
3811 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 &&
3812 error "command should accept two params"
3814 echo "correct device name should be passed"
3815 do_facet mgs $LCTL replace_nids $FSNAME-WRONG0000 $OST1_NID &&
3816 error "wrong devname"
3818 echo "wrong nids list should not destroy the system"
3819 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 "wrong nids list" &&
3822 echo "replace OST nid"
3823 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 $OST1_NID ||
3824 error "replace nids failed"
3826 echo "command should accept two parameters"
3827 do_facet mgs $LCTL replace_nids $FSNAME-MDT0000 &&
3828 error "command should accept two params"
3830 echo "wrong nids list should not destroy the system"
3831 do_facet mgs $LCTL replace_nids $FSNAME-MDT0000 "wrong nids list" &&
3834 echo "replace MDS nid"
3835 do_facet mgs $LCTL replace_nids $FSNAME-MDT0000 $MDS_NID ||
3836 error "replace nids failed"
3838 if ! combined_mgs_mds ; then
3845 check_mount || error "error after nid replace"
3846 cleanup || error "cleanup failed"
3849 run_test 66 "replace nids"
3851 test_67() { #LU-2950
3852 local legacy="$TMP/legacy_lnet_config"
3853 local new="$TMP/new_routes_test"
3854 local out="$TMP/config_out_file"
3855 local verify="$TMP/conv_verify"
3856 local verify_conf="$TMP/conf_verify"
3858 # Create the legacy file that will be run through the
3859 # lustre_routes_conversion script
3860 cat <<- LEGACY_LNET_CONFIG > $legacy
3861 tcp1 23 192.168.213.1@tcp:1; tcp5 34 193.30.4.3@tcp:4;
3862 tcp2 54 10.1.3.2@tcp;
3863 tcp3 10.3.4.3@tcp:3;
3867 # Create the verification file to verify the output of
3868 # lustre_routes_conversion script against.
3869 cat <<- VERIFY_LNET_CONFIG > $verify
3870 tcp1: { gateway: 192.168.213.1@tcp, hop: 23, priority: 1 }
3871 tcp5: { gateway: 193.30.4.3@tcp, hop: 34, priority: 4 }
3872 tcp2: { gateway: 10.1.3.2@tcp, hop: 54 }
3873 tcp3: { gateway: 10.3.4.3@tcp, priority: 3 }
3874 tcp4: { gateway: 10.3.3.4@tcp }
3877 # Create the verification file to verify the output of
3878 # lustre_routes_config script against
3879 cat <<- VERIFY_LNET_CONFIG > $verify_conf
3880 lctl --net tcp1 add_route 192.168.213.1@tcp 23 1
3881 lctl --net tcp5 add_route 193.30.4.3@tcp 34 4
3882 lctl --net tcp2 add_route 10.1.3.2@tcp 54 4
3883 lctl --net tcp3 add_route 10.3.4.3@tcp 1 3
3884 lctl --net tcp4 add_route 10.3.3.4@tcp 1 3
3887 lustre_routes_conversion $legacy $new > /dev/null
3888 if [ -f $new ]; then
3889 # verify the conversion output
3890 cmp -s $new $verify > /dev/null
3891 if [ $? -eq 1 ]; then
3892 error "routes conversion failed"
3895 lustre_routes_config --dry-run --verbose $new > $out
3896 # check that the script succeeded
3897 cmp -s $out $verify_conf > /dev/null
3898 if [ $? -eq 1 ]; then
3899 error "routes config failed"
3902 error "routes conversion test failed"
3904 # remove generated files
3905 rm -f $new $legacy $verify $verify_conf $out
3907 run_test 67 "test routes conversion and configuration"
3915 [ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.4.53) ] ||
3916 { skip "Need MDS version at least 2.4.53"; return 0; }
3918 umount_client $MOUNT || error "umount client failed"
3920 start_mdt 1 || error "MDT start failed"
3923 # START-END - the sequences we'll be reserving
3924 START=$(do_facet $SINGLEMDS \
3925 lctl get_param -n seq.ctl*.space | awk -F'[[ ]' '{print $2}')
3926 END=$((START + (1 << 30)))
3927 do_facet $SINGLEMDS \
3928 lctl set_param seq.ctl*.fldb="[$START-$END\):0:mdt"
3930 # reset the sequences MDT0000 has already assigned
3931 do_facet $SINGLEMDS \
3932 lctl set_param seq.srv*MDT0000.space=clear
3934 # remount to let the client allocate new sequence
3935 mount_client $MOUNT || error "mount client failed"
3938 do_facet $SINGLEMDS \
3939 lctl get_param seq.srv*MDT0000.space
3940 $LFS path2fid $DIR/$tfile
3942 local old_ifs="$IFS"
3944 fid=($($LFS path2fid $DIR/$tfile))
3948 if [[ $seq < $END ]]; then
3949 error "used reserved sequence $seq?"
3951 cleanup || return $?
3953 run_test 68 "be able to reserve specific sequences in FLDB"
3956 local server_version=$(lustre_version_code $SINGLEMDS)
3958 [[ $server_version -lt $(version_code 2.4.2) ]] &&
3959 skip "Need MDS version at least 2.4.2" && return
3961 [[ $server_version -ge $(version_code 2.4.50) ]] &&
3962 [[ $server_version -lt $(version_code 2.5.0) ]] &&
3963 skip "Need MDS version at least 2.5.0" && return
3967 # use OST0000 since it probably has the most creations
3968 local OSTNAME=$(ostname_from_index 0)
3969 local mdtosc_proc1=$(get_mdtosc_proc_path mds1 $OSTNAME)
3970 local last_id=$(do_facet mds1 lctl get_param -n \
3971 osc.$mdtosc_proc1.prealloc_last_id)
3973 # Want to have OST LAST_ID over 1.5 * OST_MAX_PRECREATE to
3974 # verify that the LAST_ID recovery is working properly. If
3975 # not, then the OST will refuse to allow the MDS connect
3976 # because the LAST_ID value is too different from the MDS
3977 #define OST_MAX_PRECREATE=20000
3978 local num_create=$((20000 * 5))
3981 $LFS setstripe -i 0 $DIR/$tdir
3982 createmany -o $DIR/$tdir/$tfile- $num_create ||
3983 error "createmany: failed to create $num_create files: $?"
3984 # delete all of the files with objects on OST0 so the
3985 # filesystem is not inconsistent later on
3986 $LFS find $MOUNT --ost 0 | xargs rm
3988 stop_ost || error "OST0 stop failure"
3989 add ost1 $(mkfs_opts ost1 $(ostdevname 1)) --reformat --replace \
3990 $(ostdevname 1) $(ostvdevname 1) ||
3991 error "reformat and replace $ostdev failed"
3992 start_ost || error "OST0 restart failure"
3993 wait_osc_import_state mds ost FULL
3995 touch $DIR/$tdir/$tfile-last || error "create file after reformat"
3996 local idx=$($LFS getstripe -i $DIR/$tdir/$tfile-last)
3997 [ $idx -ne 0 ] && error "$DIR/$tdir/$tfile-last on $idx not 0" || true
4001 run_test 69 "replace an OST with the same index"
4004 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
4009 start_mdt 1 || error "MDT0 start fail"
4011 start_ost || error "OST0 start fail"
4013 start_mdt 2 || error "MDT1 start fail"
4015 mount_client $MOUNT || error "mount client fails"
4017 mkdir -p $DIR/$tdir || error "create dir fail"
4019 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
4020 error "create remote dir fail"
4022 rm -rf $DIR/$tdir || error "delete dir fail"
4023 cleanup || return $?
4025 run_test 70a "start MDT0, then OST, then MDT1"
4028 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
4031 start_ost || error "OST0 start fail"
4033 start_mdt 1 || error "MDT0 start fail"
4034 start_mdt 2 || error "MDT1 start fail"
4036 mount_client $MOUNT || error "mount client fails"
4038 mkdir -p $DIR/$tdir || error "create dir fail"
4040 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
4041 error "create remote dir fail"
4043 rm -rf $DIR/$tdir || error "delete dir fail"
4045 cleanup || return $?
4047 run_test 70b "start OST, MDT1, MDT0"
4050 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
4053 start_mdt 1 || error "MDT0 start fail"
4054 start_mdt 2 || error "MDT1 start fail"
4055 start_ost || error "OST0 start fail"
4057 mount_client $MOUNT || error "mount client fails"
4058 stop_mdt 1 || error "MDT1 start fail"
4060 local mdc_for_mdt1=$($LCTL dl | grep MDT0000-mdc | awk '{print $4}')
4061 echo "deactivate $mdc_for_mdt1"
4062 $LCTL --device $mdc_for_mdt1 deactivate || return 1
4064 mkdir -p $DIR/$tdir && error "mkdir succeed"
4066 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir &&
4067 error "create remote dir succeed"
4069 cleanup || return $?
4071 run_test 70c "stop MDT0, mkdir fail, create remote dir fail"
4074 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
4077 start_mdt 1 || error "MDT0 start fail"
4078 start_mdt 2 || error "MDT1 start fail"
4079 start_ost || error "OST0 start fail"
4081 mount_client $MOUNT || error "mount client fails"
4083 stop_mdt 2 || error "MDT1 start fail"
4085 local mdc_for_mdt2=$($LCTL dl | grep MDT0001-mdc |
4087 echo "deactivate $mdc_for_mdt2"
4088 $LCTL --device $mdc_for_mdt2 deactivate ||
4089 error "set $mdc_for_mdt2 deactivate failed"
4091 mkdir -p $DIR/$tdir || error "mkdir fail"
4092 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir &&
4093 error "create remote dir succeed"
4095 rm -rf $DIR/$tdir || error "delete dir fail"
4097 cleanup || return $?
4099 run_test 70d "stop MDT1, mkdir succeed, create remote dir fail"
4102 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
4103 if combined_mgs_mds; then
4104 skip "needs separate MGS/MDT" && return
4108 start_mdt 1 || error "MDT0 start fail"
4109 start_ost || error "OST0 start fail"
4110 start_mdt 2 || error "MDT1 start fail"
4111 start_ost2 || error "OST1 start fail"
4113 mount_client $MOUNT || error "mount client fails"
4115 mkdir -p $DIR/$tdir || error "mkdir fail"
4116 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
4117 error "create remote dir succeed"
4119 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
4120 rm -rf $DIR/$tdir || error "delete dir fail"
4122 umount_client $MOUNT
4123 stop_mdt 1 || error "MDT0 stop fail"
4124 stop_mdt 2 || error "MDT1 stop fail"
4125 stop_ost || error "OST0 stop fail"
4126 stop_ost2 || error "OST1 stop fail"
4128 run_test 71a "start MDT0 OST0, MDT1, OST1"
4131 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
4132 if combined_mgs_mds; then
4133 skip "needs separate MGS/MDT" && return
4137 start_mdt 2 || error "MDT1 start fail"
4138 start_ost || error "OST0 start fail"
4139 start_mdt 1 || error "MDT0 start fail"
4140 start_ost2 || error "OST1 start fail"
4142 mount_client $MOUNT || error "mount client fails"
4144 mkdir -p $DIR/$tdir || error "mkdir fail"
4145 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
4146 error "create remote dir succeed"
4148 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
4149 rm -rf $DIR/$tdir || error "delete dir fail"
4151 umount_client $MOUNT
4152 stop_mdt 1 || error "MDT0 stop fail"
4153 stop_mdt 2 || error "MDT1 stop fail"
4154 stop_ost || error "OST0 stop fail"
4155 stop_ost2 || error "OST1 stop fail"
4157 run_test 71b "start MDT1, OST0, MDT0, OST1"
4160 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
4161 if combined_mgs_mds; then
4162 skip "needs separate MGS/MDT" && return
4166 start_ost || error "OST0 start fail"
4167 start_ost2 || error "OST1 start fail"
4168 start_mdt 2 || error "MDT1 start fail"
4169 start_mdt 1 || error "MDT0 start fail"
4171 mount_client $MOUNT || error "mount client fails"
4173 mkdir -p $DIR/$tdir || error "mkdir fail"
4174 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
4175 error "create remote dir succeed"
4177 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
4178 rm -rf $DIR/$tdir || error "delete dir fail"
4180 umount_client $MOUNT
4181 stop_mdt 1 || error "MDT0 stop fail"
4182 stop_mdt 2 || error "MDT1 stop fail"
4183 stop_ost || error "OST0 stop fail"
4184 stop_ost2 || error "OST1 stop fail"
4187 run_test 71c "start OST0, OST1, MDT1, MDT0"
4190 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
4191 if combined_mgs_mds; then
4192 skip "needs separate MGS/MDT" && return
4196 start_ost || error "OST0 start fail"
4197 start_mdt 2 || error "MDT0 start fail"
4198 start_mdt 1 || error "MDT0 start fail"
4199 start_ost2 || error "OST1 start fail"
4201 mount_client $MOUNT || error "mount client fails"
4203 mkdir -p $DIR/$tdir || error "mkdir fail"
4204 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
4205 error "create remote dir succeed"
4207 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
4208 rm -rf $DIR/$tdir || error "delete dir fail"
4210 umount_client $MOUNT
4211 stop_mdt 1 || error "MDT0 stop fail"
4212 stop_mdt 2 || error "MDT1 stop fail"
4213 stop_ost || error "OST0 stop fail"
4214 stop_ost2 || error "OST1 stop fail"
4217 run_test 71d "start OST0, MDT1, MDT0, OST1"
4220 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
4221 if combined_mgs_mds; then
4222 skip "needs separate MGS/MDT" && return
4226 start_ost || error "OST0 start fail"
4227 start_mdt 2 || error "MDT1 start fail"
4228 start_ost2 || error "OST1 start fail"
4229 start_mdt 1 || error "MDT0 start fail"
4231 mount_client $MOUNT || error "mount client fails"
4233 mkdir -p $DIR/$tdir || error "mkdir fail"
4234 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
4235 error "create remote dir succeed"
4237 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
4238 rm -rf $DIR/$tdir || error "delete dir fail"
4240 umount_client $MOUNT
4241 stop_mdt 1 || error "MDT0 stop fail"
4242 stop_mdt 2 || error "MDT1 stop fail"
4243 stop_ost || error "OST0 stop fail"
4244 stop_ost2 || error "OST1 stop fail"
4247 run_test 71e "start OST0, MDT1, OST1, MDT0"
4249 test_72() { #LU-2634
4250 local mdsdev=$(mdsdevname 1)
4251 local ostdev=$(ostdevname 1)
4252 local cmd="$E2FSCK -fnvd $mdsdev"
4255 [ "$(facet_fstype $SINGLEMDS)" != "ldiskfs" ] &&
4256 skip "ldiskfs only test" && return
4258 #tune MDT with "-O extents"
4260 for num in $(seq $MDSCOUNT); do
4261 add mds${num} $(mkfs_opts mds$num $(mdsdevname $num)) \
4262 --reformat $(mdsdevname $num) $(mdsvdevname $num) ||
4263 error "add mds $num failed"
4264 $TUNE2FS -O extents $(mdsdevname $num)
4267 add ost1 $(mkfs_opts ost1 $ostdev) --reformat $ostdev ||
4268 error "add $ostdev failed"
4269 start_mgsmds || error "start mds failed"
4270 start_ost || error "start ost failed"
4271 mount_client $MOUNT || error "mount client failed"
4273 #create some short symlinks
4275 createmany -o $DIR/$tdir/$tfile-%d $fn
4276 echo "create $fn short symlinks"
4277 for i in $(seq -w 1 $fn); do
4278 ln -s $DIR/$tdir/$tfile-$i $MOUNT/$tfile-$i
4283 umount_client $MOUNT || error "umount client failed"
4284 stop_mds || error "stop mds failed"
4285 stop_ost || error "stop ost failed"
4288 run_e2fsck $(facet_active_host $SINGLEMDS) $mdsdev "-n"
4290 run_test 72 "test fast symlink with extents flag enabled"
4292 test_73() { #LU-3006
4294 [ $(facet_fstype ost1) == zfs ] && import_zpool ost1
4295 do_facet ost1 "$TUNEFS --failnode=1.2.3.4@$NETTYPE $(ostdevname 1)" ||
4296 error "1st tunefs failed"
4297 start_mgsmds || error "start mds failed"
4298 start_ost || error "start ost failed"
4299 mount_client $MOUNT || error "mount client failed"
4300 lctl get_param -n osc.*OST0000-osc-[^M]*.import | grep failover_nids |
4301 grep 1.2.3.4@$NETTYPE || error "failover nids haven't changed"
4302 umount_client $MOUNT || error "umount client failed"
4306 run_test 73 "failnode to update from mountdata properly"
4308 test_74() { # LU-1606
4309 for TESTPROG in $LUSTRE_TESTS_API_DIR/*.c; do
4310 gcc -Wall -Werror $LUSTRE_TESTS_API_DIR/simple_test.c \
4312 -L$LUSTRE/utils -llustreapi ||
4313 error "client api broken"
4315 cleanup || return $?
4317 run_test 74 "Lustre client api program can compile and link"
4319 test_75() { # LU-2374
4320 [[ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.4.1) ]] &&
4321 skip "Need MDS version at least 2.4.1" && return
4324 local opts_mds="$(mkfs_opts mds1 $(mdsdevname 1)) \
4325 --reformat $(mdsdevname 1) $(mdsvdevname 1)"
4326 local opts_ost="$(mkfs_opts ost1 $(ostdevname 1)) \
4327 --reformat $(ostdevname 1) $(ostvdevname 1)"
4329 #check with default parameters
4330 add mds1 $opts_mds || error "add mds1 failed for default params"
4331 add ost1 $opts_ost || error "add ost1 failed for default params"
4333 opts_mds=$(echo $opts_mds | sed -e "s/--mdt//")
4334 opts_mds=$(echo $opts_mds |
4335 sed -e "s/--index=$index/--index=$index --mdt/")
4336 opts_ost=$(echo $opts_ost | sed -e "s/--ost//")
4337 opts_ost=$(echo $opts_ost |
4338 sed -e "s/--index=$index/--index=$index --ost/")
4340 add mds1 $opts_mds || error "add mds1 failed for new params"
4341 add ost1 $opts_ost || error "add ost1 failed for new params"
4344 run_test 75 "The order of --index should be irrelevant"
4347 [[ $(lustre_version_code mgs) -ge $(version_code 2.4.52) ]] ||
4348 { skip "Need MDS version at least 2.4.52" && return 0; }
4350 local MDMB_PARAM="osc.*.max_dirty_mb"
4351 echo "Change MGS params"
4352 local MAX_DIRTY_MB=$($LCTL get_param -n $MDMB_PARAM |
4354 echo "max_dirty_mb: $MAX_DIRTY_MB"
4355 local NEW_MAX_DIRTY_MB=$((MAX_DIRTY_MB + MAX_DIRTY_MB))
4356 echo "new_max_dirty_mb: $NEW_MAX_DIRTY_MB"
4357 do_facet mgs $LCTL set_param -P $MDMB_PARAM=$NEW_MAX_DIRTY_MB
4358 wait_update $HOSTNAME "lctl get_param -n $MDMB_PARAM |
4359 head -1" $NEW_MAX_DIRTY_MB
4360 MAX_DIRTY_MB=$($LCTL get_param -n $MDMB_PARAM | head -1)
4361 echo "$MAX_DIRTY_MB"
4362 [ $MAX_DIRTY_MB = $NEW_MAX_DIRTY_MB ] ||
4363 error "error while apply max_dirty_mb"
4365 echo "Check the value is stored after remount"
4368 wait_update $HOSTNAME "lctl get_param -n $MDMB_PARAM |
4369 head -1" $NEW_MAX_DIRTY_MB
4370 MAX_DIRTY_MB=$($LCTL get_param -n $MDMB_PARAM | head -1)
4371 [ $MAX_DIRTY_MB = $NEW_MAX_DIRTY_MB ] ||
4372 error "max_dirty_mb is not saved after remount"
4374 echo "Change OST params"
4375 CLIENT_PARAM="obdfilter.*.client_cache_count"
4376 local CLIENT_CACHE_COUNT
4377 CLIENT_CACHE_COUNT=$(do_facet ost1 $LCTL get_param -n $CLIENT_PARAM |
4379 echo "client_cache_count: $CLIENT_CACHE_COUNT"
4380 NEW_CLIENT_CACHE_COUNT=$((CLIENT_CACHE_COUNT+CLIENT_CACHE_COUNT))
4381 echo "new_client_cache_count: $NEW_CLIENT_CACHE_COUNT"
4382 do_facet mgs $LCTL set_param -P $CLIENT_PARAM=$NEW_CLIENT_CACHE_COUNT
4383 wait_update $(facet_host ost1) "lctl get_param -n $CLIENT_PARAM |
4384 head -1" $NEW_CLIENT_CACHE_COUNT
4385 CLIENT_CACHE_COUNT=$(do_facet ost1 $LCTL get_param -n $CLIENT_PARAM |
4387 echo "$CLIENT_CACHE_COUNT"
4388 [ $CLIENT_CACHE_COUNT = $NEW_CLIENT_CACHE_COUNT ] ||
4389 error "error while apply client_cache_count"
4391 echo "Check the value is stored after remount"
4394 wait_update $(facet_host ost1) "lctl get_param -n $CLIENT_PARAM |
4395 head -1" $NEW_CLIENT_CACHE_COUNT
4396 CLIENT_CACHE_COUNT=$(do_facet ost1 $LCTL get_param -n $CLIENT_PARAM |
4398 echo "$CLIENT_CACHE_COUNT"
4399 [ $CLIENT_CACHE_COUNT = $NEW_CLIENT_CACHE_COUNT ] ||
4400 error "client_cache_count is not saved after remount"
4403 run_test 76 "set permanent params set_param -P"
4405 test_77() { # LU-3445
4406 local server_version=$(lustre_version_code $SINGLEMDS)
4408 [[ $server_version -ge $(version_code 2.2.60) ]] &&
4409 [[ $server_version -le $(version_code 2.4.0) ]] &&
4410 skip "Need MDS version < 2.2.60 or > 2.4.0" && return
4412 if [[ -z "$fs2ost_DEV" || -z "$fs2mds_DEV" ]]; then
4413 is_blkdev $SINGLEMDS $(mdsdevname ${SINGLEMDS//mds/}) &&
4414 skip_env "mixed loopback and real device not working" && return
4417 local fs2mdsdev=$(mdsdevname 1_2)
4418 local fs2ostdev=$(ostdevname 1_2)
4419 local fs2mdsvdev=$(mdsvdevname 1_2)
4420 local fs2ostvdev=$(ostvdevname 1_2)
4421 local fsname=test1234
4423 local failnid="$(h2$NETTYPE 1.2.3.4),$(h2$NETTYPE 4.3.2.1)"
4425 add fs2mds $(mkfs_opts mds1 $fs2mdsdev) --mgs --fsname=$fsname \
4426 --reformat $fs2mdsdev $fs2mdsvdev || error "add fs2mds failed"
4427 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS && trap cleanup_fs2 EXIT INT ||
4428 error "start fs2mds failed"
4430 mgsnid=$(do_facet fs2mds $LCTL list_nids | xargs | tr ' ' ,)
4431 [[ $mgsnid = *,* ]] || mgsnid+=",$mgsnid"
4433 add fs2ost $(mkfs_opts ost1 $fs2ostdev) --mgsnode=$mgsnid \
4434 --failnode=$failnid --fsname=$fsname \
4435 --reformat $fs2ostdev $fs2ostvdev ||
4436 error "add fs2ost failed"
4437 start fs2ost $fs2ostdev $OST_MOUNT_OPTS || error "start fs2ost failed"
4440 mount -t lustre $mgsnid:/$fsname $MOUNT2 || error "mount $MOUNT2 failed"
4441 DIR=$MOUNT2 MOUNT=$MOUNT2 check_mount || error "check $MOUNT2 failed"
4444 run_test 77 "comma-separated MGS NIDs and failover node NIDs"
4447 [[ $(facet_fstype $SINGLEMDS) != ldiskfs ||
4448 $(facet_fstype ost1) != ldiskfs ]] &&
4449 skip "only applicable to ldiskfs-based MDTs and OSTs" && return
4451 # reformat the Lustre filesystem with a smaller size
4452 local saved_MDSSIZE=$MDSSIZE
4453 local saved_OSTSIZE=$OSTSIZE
4454 MDSSIZE=$((MDSSIZE - 20000))
4455 OSTSIZE=$((OSTSIZE - 20000))
4456 reformat || error "(1) reformat Lustre filesystem failed"
4457 MDSSIZE=$saved_MDSSIZE
4458 OSTSIZE=$saved_OSTSIZE
4460 # mount the Lustre filesystem
4461 setup_noconfig || error "(2) setup Lustre filesystem failed"
4464 log "create test files"
4468 mkdir -p $MOUNT/$tdir || error "(3) mkdir $MOUNT/$tdir failed"
4469 for i in $(seq $num_files); do
4470 file=$MOUNT/$tdir/$tfile-$i
4471 dd if=/dev/urandom of=$file count=1 bs=1M ||
4472 error "(4) create $file failed"
4475 # unmount the Lustre filesystem
4476 cleanup || error "(5) cleanup Lustre filesystem failed"
4478 # run e2fsck on the MDT and OST devices
4479 local mds_host=$(facet_active_host $SINGLEMDS)
4480 local ost_host=$(facet_active_host ost1)
4481 local mds_dev=$(mdsdevname ${SINGLEMDS//mds/})
4482 local ost_dev=$(ostdevname 1)
4484 run_e2fsck $mds_host $mds_dev "-y"
4485 run_e2fsck $ost_host $ost_dev "-y"
4487 # get the original block count of the MDT and OST filesystems
4488 local mds_orig_blks=$(get_block_count $SINGLEMDS $mds_dev)
4489 local ost_orig_blks=$(get_block_count ost1 $ost_dev)
4491 # expand the MDT and OST filesystems to the device size
4492 run_resize2fs $SINGLEMDS $mds_dev "" || error "expand $SINGLEMDS failed"
4493 run_resize2fs ost1 $ost_dev "" || error "expand ost1 failed"
4495 # run e2fsck on the MDT and OST devices again
4496 run_e2fsck $mds_host $mds_dev "-y"
4497 run_e2fsck $ost_host $ost_dev "-y"
4499 # mount the Lustre filesystem
4503 log "check files after expanding the MDT and OST filesystems"
4504 for i in $(seq $num_files); do
4505 file=$MOUNT/$tdir/$tfile-$i
4506 $CHECKSTAT -t file -s 1048576 $file ||
4507 error "(6) checkstat $file failed"
4511 log "create more files after expanding the MDT and OST filesystems"
4512 for i in $(seq $((num_files + 1)) $((num_files + 10))); do
4513 file=$MOUNT/$tdir/$tfile-$i
4514 dd if=/dev/urandom of=$file count=1 bs=1M ||
4515 error "(7) create $file failed"
4518 # unmount the Lustre filesystem
4519 cleanup || error "(8) cleanup Lustre filesystem failed"
4521 # run e2fsck on the MDT and OST devices
4522 run_e2fsck $mds_host $mds_dev "-y"
4523 run_e2fsck $ost_host $ost_dev "-y"
4525 # get the maximum block count of the MDT and OST filesystems
4526 local mds_max_blks=$(get_block_count $SINGLEMDS $mds_dev)
4527 local ost_max_blks=$(get_block_count ost1 $ost_dev)
4529 # get the minimum block count of the MDT and OST filesystems
4530 local mds_min_blks=$(run_resize2fs $SINGLEMDS $mds_dev "" "-P" 2>&1 |
4531 grep minimum | sed -e 's/^.*filesystem: //g')
4532 local ost_min_blks=$(run_resize2fs ost1 $ost_dev "" "-P" 2>&1 |
4533 grep minimum | sed -e 's/^.*filesystem: //g')
4535 # shrink the MDT and OST filesystems to a smaller size
4539 if [[ $mds_max_blks -gt $mds_min_blks &&
4540 $mds_max_blks -gt $mds_orig_blks ]]; then
4541 [[ $mds_orig_blks -gt $mds_min_blks ]] &&
4542 base_blks=$mds_orig_blks || base_blks=$mds_min_blks
4543 new_blks=$(( (mds_max_blks - base_blks) / 2 + base_blks ))
4544 run_resize2fs $SINGLEMDS $mds_dev $new_blks ||
4545 error "shrink $SINGLEMDS to $new_blks failed"
4549 if [[ $ost_max_blks -gt $ost_min_blks &&
4550 $ost_max_blks -gt $ost_orig_blks ]]; then
4551 [[ $ost_orig_blks -gt $ost_min_blks ]] &&
4552 base_blks=$ost_orig_blks || base_blks=$ost_min_blks
4553 new_blks=$(( (ost_max_blks - base_blks) / 2 + base_blks ))
4554 run_resize2fs ost1 $ost_dev $new_blks ||
4555 error "shrink ost1 to $new_blks failed"
4559 # check whether the MDT or OST filesystem was shrunk or not
4561 combined_mgs_mds || stop_mgs || error "(9) stop mgs failed"
4562 reformat || error "(10) reformat Lustre filesystem failed"
4566 # run e2fsck on the MDT and OST devices again
4567 run_e2fsck $mds_host $mds_dev "-y"
4568 run_e2fsck $ost_host $ost_dev "-y"
4570 # mount the Lustre filesystem again
4574 log "check files after shrinking the MDT and OST filesystems"
4575 for i in $(seq $((num_files + 10))); do
4576 file=$MOUNT/$tdir/$tfile-$i
4577 $CHECKSTAT -t file -s 1048576 $file ||
4578 error "(11) checkstat $file failed"
4581 # unmount and reformat the Lustre filesystem
4582 cleanup || error "(12) cleanup Lustre filesystem failed"
4583 combined_mgs_mds || stop_mgs || error "(13) stop mgs failed"
4584 reformat || error "(14) reformat Lustre filesystem failed"
4586 run_test 78 "run resize2fs on MDT and OST filesystems"
4588 if ! combined_mgs_mds ; then
4594 # restore the values of MDSSIZE and OSTSIZE
4595 MDSSIZE=$STORED_MDSSIZE
4596 OSTSIZE=$STORED_OSTSIZE