3 # FIXME - there is no reason to use all of these different
4 # return codes, espcially when most of them are mapped to something
5 # else anyway. The combination of test number and return code
6 # figure out what failed.
12 # bug number for skipped test: LU-2828
13 ALWAYS_EXCEPT="$CONF_SANITY_EXCEPT 59 64"
14 # UPDATE THE COMMENT ABOVE WITH BUG NUMBERS WHEN CHANGING ALWAYS_EXCEPT!
18 if [ -r /etc/SuSE-release ]
20 local vers=`grep VERSION /etc/SuSE-release | awk '{print $3}'`
21 local patchlev=`grep PATCHLEVEL /etc/SuSE-release \
23 if [ $vers -eq 11 ] && [ $patchlev -eq 2 ]
31 if is_sles11; then # LU-2181
32 ALWAYS_EXCEPT="$ALWAYS_EXCEPT 23a 34b"
35 if [ "$FAILURE_MODE" = "HARD" ]; then
36 CONFIG_EXCEPTIONS="24a " && \
37 echo "Except the tests: $CONFIG_EXCEPTIONS for FAILURE_MODE=$FAILURE_MODE, bug 23573" && \
38 ALWAYS_EXCEPT="$ALWAYS_EXCEPT $CONFIG_EXCEPTIONS"
41 # bug number for skipped test:
42 # a tool to create lustre filesystem images
43 ALWAYS_EXCEPT="32newtarball $ALWAYS_EXCEPT"
46 PATH=$PWD/$SRCDIR:$SRCDIR:$SRCDIR/../utils:$PATH
48 PTLDEBUG=${PTLDEBUG:--1}
50 LUSTRE=${LUSTRE:-`dirname $0`/..}
51 RLUSTRE=${RLUSTRE:-$LUSTRE}
52 LUSTRE_TESTS_API_DIR=${LUSTRE_TESTS_API_DIR:-${LUSTRE}/tests/clientapi}
53 export MULTIOP=${MULTIOP:-multiop}
55 . $LUSTRE/tests/test-framework.sh
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 $SINGLEMDS) = "zfs" ] &&
78 # bug number for skipped test: LU-2778 LU-2059
79 ALWAYS_EXCEPT="$ALWAYS_EXCEPT 57b 50h"
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" $ORIG || return 5
1063 run_test 28 "permanent parameter setting"
1066 [ "$OSTCOUNT" -lt "2" ] && skip_env "$OSTCOUNT < 2, skipping" && return
1067 setup > /dev/null 2>&1
1071 local PARAM="$FSNAME-OST0001.osc.active"
1072 local PROC_ACT="osc.$FSNAME-OST0001-osc-[^M]*.active"
1073 local PROC_UUID="osc.$FSNAME-OST0001-osc-[^M]*.ost_server_uuid"
1075 ACTV=$(lctl get_param -n $PROC_ACT)
1077 set_conf_param_and_check client \
1078 "lctl get_param -n $PROC_ACT" "$PARAM" $DEAC || return 2
1079 # also check ost_server_uuid status
1080 RESULT=$(lctl get_param -n $PROC_UUID | grep DEACTIV)
1081 if [ -z "$RESULT" ]; then
1082 echo "Live client not deactivated: $(lctl get_param -n $PROC_UUID)"
1085 echo "Live client success: got $RESULT"
1089 for num in $(seq $MDSCOUNT); do
1090 local mdtosc=$(get_mdtosc_proc_path mds${num} $FSNAME-OST0001)
1091 local MPROC="osc.$mdtosc.active"
1096 RESULT=$(do_facet mds${num} " lctl get_param -n $MPROC")
1097 [ ${PIPESTATUS[0]} = 0 ] || error "Can't read $MPROC"
1098 if [ $RESULT -eq $DEAC ]; then
1099 echo -n "MDT deactivated also after"
1100 echo "$WAIT sec (got $RESULT)"
1104 if [ $WAIT -eq $MAX ]; then
1105 echo -n "MDT not deactivated: wanted $DEAC"
1109 echo "Waiting $(($MAX - $WAIT))secs for MDT deactivated"
1112 # test new client starts deactivated
1113 umount_client $MOUNT || return 200
1115 RESULT=$(lctl get_param -n $PROC_UUID | grep DEACTIV | grep NEW)
1116 if [ -z "$RESULT" ]; then
1117 echo "New client not deactivated from start: $(lctl get_param -n $PROC_UUID)"
1120 echo "New client success: got $RESULT"
1123 # make sure it reactivates
1124 set_conf_param_and_check client \
1125 "lctl get_param -n $PROC_ACT" "$PARAM" $ACTV || return 6
1127 umount_client $MOUNT
1130 #writeconf to remove all ost2 traces for subsequent tests
1131 writeconf_or_reformat
1133 run_test 29 "permanently remove an OST"
1138 echo Big config llog
1139 TEST="lctl get_param -n llite.$FSNAME-*.max_read_ahead_whole_mb"
1141 LIST=(1 2 3 4 5 4 3 2 1 2 3 4 5 4 3 2 1 2 3 4 5)
1142 for i in ${LIST[@]}; do
1143 set_conf_param_and_check client "$TEST" \
1144 "$FSNAME.llite.max_read_ahead_whole_mb" $i || return 3
1146 # make sure client restart still works
1147 umount_client $MOUNT
1148 mount_client $MOUNT || return 4
1149 [ "$($TEST)" -ne "$i" ] && error "Param didn't stick across restart $($TEST) != $i"
1152 echo Erase parameter setting
1153 do_facet mgs "$LCTL conf_param -d $FSNAME.llite.max_read_ahead_whole_mb" || return 6
1154 umount_client $MOUNT
1155 mount_client $MOUNT || return 6
1157 echo "deleted (default) value=$FINAL, orig=$ORIG"
1158 # assumes this parameter started at the default value
1159 [ "$FINAL" -eq "$ORIG" ] || fail "Deleted value=$FINAL, orig=$ORIG"
1163 run_test 30a "Big config llog and conf_param deletion"
1168 # Make a fake nid. Use the OST nid, and add 20 to the least significant
1169 # numerical part of it. Hopefully that's not already a failover address for
1171 OSTNID=$(do_facet ost1 "$LCTL get_param nis" | tail -1 | awk '{print $1}')
1172 ORIGVAL=$(echo $OSTNID | egrep -oi "[0-9]*@")
1173 NEWVAL=$((($(echo $ORIGVAL | egrep -oi "[0-9]*") + 20) % 256))
1174 NEW=$(echo $OSTNID | sed "s/$ORIGVAL/$NEWVAL@/")
1175 echo "Using fake nid $NEW"
1177 TEST="$LCTL get_param -n osc.$FSNAME-OST0000-osc-[^M]*.import | grep failover_nids | sed -n 's/.*\($NEW\).*/\1/p'"
1178 set_conf_param_and_check client "$TEST" \
1179 "$FSNAME-OST0000.failover.node" $NEW ||
1180 error "didn't add failover nid $NEW"
1181 NIDS=$($LCTL get_param -n osc.$FSNAME-OST0000-osc-[^M]*.import | grep failover_nids)
1183 NIDCOUNT=$(($(echo "$NIDS" | wc -w) - 1))
1184 echo "should have 2 failover nids: $NIDCOUNT"
1185 [ $NIDCOUNT -eq 2 ] || error "Failover nid not added"
1186 do_facet mgs "$LCTL conf_param -d $FSNAME-OST0000.failover.node" || error "conf_param delete failed"
1187 umount_client $MOUNT
1188 mount_client $MOUNT || return 3
1190 NIDS=$($LCTL get_param -n osc.$FSNAME-OST0000-osc-[^M]*.import | grep failover_nids)
1192 NIDCOUNT=$(($(echo "$NIDS" | wc -w) - 1))
1193 echo "only 1 final nid should remain: $NIDCOUNT"
1194 [ $NIDCOUNT -eq 1 ] || error "Failover nids not removed"
1198 run_test 30b "Remove failover nids"
1200 test_31() { # bug 10734
1201 # ipaddr must not exist
1202 mount -t lustre 4.3.2.1@tcp:/lustre $MOUNT || true
1205 run_test 31 "Connect to non-existent node (shouldn't crash)"
1209 T32_BLIMIT=20480 # Kbytes
1213 # This is not really a test but a tool to create new disk
1214 # image tarballs for the upgrade tests.
1216 # Disk image tarballs should be created on single-node
1217 # clusters by running this test with default configurations
1218 # plus a few mandatory environment settings that are verified
1219 # at the beginning of the test.
1221 test_32newtarball() {
1225 local tmp=$TMP/t32_image_create
1227 if [ $FSNAME != t32fs -o $MDSCOUNT -ne 1 -o \
1228 \( -z "$MDSDEV" -a -z "$MDSDEV1" \) -o $OSTCOUNT -ne 1 -o \
1229 -z "$OSTDEV1" ]; then
1230 error "Needs FSNAME=t32fs MDSCOUNT=1 MDSDEV1=<nonexistent_file>" \
1231 "(or MDSDEV, in the case of b1_8) OSTCOUNT=1" \
1232 "OSTDEV1=<nonexistent_file>"
1236 echo "Found stale $tmp"
1241 tar cf - -C $src . | tar xf - -C $tmp/src
1242 dd if=/dev/zero of=$tmp/src/t32_qf_old bs=1M \
1243 count=$(($T32_BLIMIT / 1024 / 2))
1244 chown $T32_QID.$T32_QID $tmp/src/t32_qf_old
1250 [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.3.50) ] &&
1251 $LFS quotacheck -ug /mnt/$FSNAME
1252 $LFS setquota -u $T32_QID -b 0 -B $T32_BLIMIT -i 0 -I $T32_ILIMIT \
1255 tar cf - -C $tmp/src . | tar xf - -C /mnt/$FSNAME
1262 ls -Rni --time-style=+%s >$tmp/img/list
1263 find . ! -name .lustre -type f -exec sha1sum {} \; |
1264 sort -k 2 >$tmp/img/sha1sums
1266 $LCTL get_param -n version | head -n 1 |
1267 sed -e 's/^lustre: *//' >$tmp/img/commit
1269 [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.3.50) ] &&
1270 $LFS quotaon -ug /mnt/$FSNAME
1271 $LFS quota -u $T32_QID -v /mnt/$FSNAME
1272 $LFS quota -v -u $T32_QID /mnt/$FSNAME |
1273 awk 'BEGIN { num='1' } { if ($1 == "'/mnt/$FSNAME'") \
1274 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1275 | tr -d "*" > $tmp/img/bspace
1276 $LFS quota -v -u $T32_QID /mnt/$FSNAME |
1277 awk 'BEGIN { num='5' } { if ($1 == "'/mnt/$FSNAME'") \
1278 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1279 | tr -d "*" > $tmp/img/ispace
1284 find -type f -exec sha1sum {} \; | sort -k 2 >$tmp/sha1sums.src
1287 if ! diff -u $tmp/sha1sums.src $tmp/img/sha1sums; then
1288 echo "Data verification failed"
1291 uname -r >$tmp/img/kernel
1292 uname -m >$tmp/img/arch
1294 mv ${MDSDEV1:-$MDSDEV} $tmp/img
1295 mv $OSTDEV1 $tmp/img
1297 version=$(sed -e 's/\(^[0-9]\+\.[0-9]\+\)\(.*$\)/\1/' $tmp/img/commit |
1298 sed -e 's/\./_/g') # E.g., "1.8.7" -> "1_8"
1301 tar cjvf $dst/disk$version-$(facet_fstype $SINGLEMDS).tar.bz2 -S *
1306 #run_test 32newtarball "Create a new test_32 disk image tarball for this version"
1309 # The list of applicable tarballs is returned via the caller's
1310 # variable "tarballs".
1313 local node=$(facet_active_host $SINGLEMDS)
1314 local r="do_node $node"
1316 if [ "$CLIENTONLY" ]; then
1317 skip "Client-only testing"
1321 if ! $r which $TUNEFS; then
1322 skip_env "tunefs.lustre required on $node"
1326 local IMGTYPE=$(facet_fstype $SINGLEMDS)
1328 tarballs=$($r find $RLUSTRE/tests -maxdepth 1 -name \'disk*-$IMGTYPE.tar.bz2\')
1330 if [ -z "$tarballs" ]; then
1331 skip "No applicable tarballs found"
1336 t32_test_cleanup() {
1340 if $shall_cleanup_lustre; then
1341 umount $tmp/mnt/lustre || rc=$?
1343 if $shall_cleanup_mdt; then
1344 $r umount -d $tmp/mnt/mdt || rc=$?
1346 if $shall_cleanup_mdt1; then
1347 $r umount -d $tmp/mnt/mdt1 || rc=$?
1349 if $shall_cleanup_ost; then
1350 $r umount -d $tmp/mnt/ost || rc=$?
1358 t32_bits_per_long() {
1360 # Yes, this is not meant to be perfect.
1370 t32_reload_modules() {
1372 local all_removed=false
1375 while ((i < 20)); do
1376 echo "Unloading modules on $node: Attempt $i"
1377 do_rpc_nodes $node $LUSTRE_RMMOD $(facet_fstype $SINGLEMDS) &&
1379 do_rpc_nodes $node check_mem_leak || return 1
1380 if $all_removed; then
1381 do_rpc_nodes $node load_modules
1387 echo "Unloading modules on $node: Given up"
1391 t32_wait_til_devices_gone() {
1396 echo wait for devices to go
1397 while ((i < 20)); do
1398 devices=$(do_rpc_nodes $node $LCTL device_list | wc -l)
1399 ((devices == 0)) && return 0
1403 echo "waiting for devices on $node: Given up"
1407 t32_verify_quota() {
1411 local fstype=$(facet_fstype $SINGLEMDS)
1415 $LFS quota -u $T32_QID -v $mnt
1417 qval=$($LFS quota -v -u $T32_QID $mnt |
1418 awk 'BEGIN { num='1' } { if ($1 == "'$mnt'") \
1419 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1421 [ $qval -eq $img_bspace ] || {
1422 echo "bspace, act:$qval, exp:$img_bspace"
1426 qval=$($LFS quota -v -u $T32_QID $mnt |
1427 awk 'BEGIN { num='5' } { if ($1 == "'$mnt'") \
1428 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1430 [ $qval -eq $img_ispace ] || {
1431 echo "ispace, act:$qval, exp:$img_ispace"
1435 qval=$($LFS quota -v -u $T32_QID $mnt |
1436 awk 'BEGIN { num='3' } { if ($1 == "'$mnt'") \
1437 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1439 [ $qval -eq $T32_BLIMIT ] || {
1440 echo "blimit, act:$qval, exp:$T32_BLIMIT"
1444 qval=$($LFS quota -v -u $T32_QID $mnt |
1445 awk 'BEGIN { num='7' } { if ($1 == "'$mnt'") \
1446 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1448 [ $qval -eq $T32_ILIMIT ] || {
1449 echo "ilimit, act:$qval, exp:$T32_ILIMIT"
1453 do_node $node $LCTL conf_param $fsname.quota.mdt=ug
1454 cmd="$LCTL get_param -n osd-$fstype.$fsname-MDT0000"
1455 cmd=$cmd.quota_slave.enabled
1456 wait_update $node "$cmd" "ug" || {
1457 echo "Enable mdt quota failed"
1461 do_node $node $LCTL conf_param $fsname.quota.ost=ug
1462 cmd="$LCTL get_param -n osd-$fstype.$fsname-OST0000"
1463 cmd=$cmd.quota_slave.enabled
1464 wait_update $node "$cmd" "ug" || {
1465 echo "Enable ost quota failed"
1470 runas -u $T32_QID -g $T32_QID dd if=/dev/zero of=$mnt/t32_qf_new \
1471 bs=1M count=$(($T32_BLIMIT / 1024)) oflag=sync && {
1472 echo "Write succeed, but expect -EDQUOT"
1475 rm -f $mnt/t32_qf_new
1477 runas -u $T32_QID -g $T32_QID createmany -m $mnt/t32_qf_ \
1479 echo "Create succeed, but expect -EDQUOT"
1482 unlinkmany $mnt/t32_qf_ $T32_ILIMIT
1490 local dne_upgrade=${dne_upgrade:-"no"}
1491 local ff_convert=${ff_convert:-"no"}
1492 local shall_cleanup_mdt=false
1493 local shall_cleanup_mdt1=false
1494 local shall_cleanup_ost=false
1495 local shall_cleanup_lustre=false
1496 local node=$(facet_active_host $SINGLEMDS)
1497 local r="do_node $node"
1498 local node2=$(facet_active_host mds2)
1506 local nid=$($r $LCTL list_nids | head -1)
1512 local fstype=$(facet_fstype $SINGLEMDS)
1514 trap 'trap - RETURN; t32_test_cleanup' RETURN
1516 mkdir -p $tmp/mnt/lustre
1517 $r mkdir -p $tmp/mnt/{mdt,ost}
1518 $r tar xjvf $tarball -S -C $tmp || {
1519 error_noexit "Unpacking the disk image tarball"
1522 img_commit=$($r cat $tmp/commit)
1523 img_kernel=$($r cat $tmp/kernel)
1524 img_arch=$($r cat $tmp/arch)
1525 img_bspace=$($r cat $tmp/bspace)
1526 img_ispace=$($r cat $tmp/ispace)
1527 echo "Upgrading from $(basename $tarball), created with:"
1528 echo " Commit: $img_commit"
1529 echo " Kernel: $img_kernel"
1530 echo " Arch: $img_arch"
1532 local version=$(version_code $img_commit)
1533 [[ $version -gt $(version_code 2.4.0) ]] && ff_convert="no"
1535 $r $LCTL set_param debug="$PTLDEBUG"
1537 $r $TUNEFS --dryrun $tmp/mdt || {
1538 error_noexit "tunefs.lustre before mounting the MDT"
1541 if [ "$writeconf" ]; then
1542 mopts=loop,writeconf
1543 if [ $fstype == "ldiskfs" ]; then
1544 $r $TUNEFS --quota $tmp/mdt || {
1545 error_noexit "Enable mdt quota feature"
1550 if [ -n "$($LCTL list_nids | grep -v '\(tcp\|lo\)[[:digit:]]*$')" ]; then
1551 [[ $(lustre_version_code mgs) -ge $(version_code 2.3.59) ]] ||
1552 { skip "LU-2200: Cannot run over Inifiniband w/o lctl replace_nids "
1553 "(Need MGS version at least 2.3.59)"; return 0; }
1555 local osthost=$(facet_active_host ost1)
1556 local ostnid=$(do_node $osthost $LCTL list_nids | head -1)
1558 $r mount -t lustre -o loop,nosvc $tmp/mdt $tmp/mnt/mdt
1559 $r lctl replace_nids $fsname-OST0000 $ostnid
1560 $r lctl replace_nids $fsname-MDT0000 $nid
1561 $r umount -d $tmp/mnt/mdt
1564 mopts=loop,exclude=$fsname-OST0000
1567 t32_wait_til_devices_gone $node
1569 $r mount -t lustre -o $mopts $tmp/mdt $tmp/mnt/mdt || {
1570 error_noexit "Mounting the MDT"
1573 shall_cleanup_mdt=true
1575 if [ "$dne_upgrade" != "no" ]; then
1576 local fs2mdsdev=$(mdsdevname 1_2)
1577 local fs2mdsvdev=$(mdsvdevname 1_2)
1579 echo "mkfs new MDT on ${fs2mdsdev}...."
1580 if [ $(facet_fstype mds1) == ldiskfs ]; then
1581 mkfsoptions="--mkfsoptions=\\\"-J size=8\\\""
1584 add fs2mds $(mkfs_opts mds2 $fs2mdsdev $fsname) --reformat \
1585 $mkfsoptions $fs2mdsdev $fs2mdsvdev > /dev/null || {
1586 error_noexit "Mkfs new MDT failed"
1590 $r $TUNEFS --dryrun $fs2mdsdev || {
1591 error_noexit "tunefs.lustre before mounting the MDT"
1595 echo "mount new MDT....$fs2mdsdev"
1596 $r mkdir -p $tmp/mnt/mdt1
1597 $r mount -t lustre -o $mopts $fs2mdsdev $tmp/mnt/mdt1 || {
1598 error_noexit "mount mdt1 failed"
1601 shall_cleanup_mdt1=true
1604 uuid=$($r $LCTL get_param -n mdt.$fsname-MDT0000.uuid) || {
1605 error_noexit "Getting MDT UUID"
1608 if [ "$uuid" != $fsname-MDT0000_UUID ]; then
1609 error_noexit "Unexpected MDT UUID: \"$uuid\""
1613 $r $TUNEFS --dryrun $tmp/ost || {
1614 error_noexit "tunefs.lustre before mounting the OST"
1617 if [ "$writeconf" ]; then
1618 mopts=loop,mgsnode=$nid,$writeconf
1619 if [ $fstype == "ldiskfs" ]; then
1620 $r $TUNEFS --quota $tmp/ost || {
1621 error_noexit "Enable ost quota feature"
1626 mopts=loop,mgsnode=$nid
1628 $r mount -t lustre -o $mopts $tmp/ost $tmp/mnt/ost || {
1629 error_noexit "Mounting the OST"
1632 shall_cleanup_ost=true
1634 uuid=$($r $LCTL get_param -n obdfilter.$fsname-OST0000.uuid) || {
1635 error_noexit "Getting OST UUID"
1638 if [ "$uuid" != $fsname-OST0000_UUID ]; then
1639 error_noexit "Unexpected OST UUID: \"$uuid\""
1643 $r $LCTL conf_param $fsname-OST0000.osc.max_dirty_mb=15 || {
1644 error_noexit "Setting \"max_dirty_mb\""
1647 $r $LCTL conf_param $fsname-OST0000.failover.node=$nid || {
1648 error_noexit "Setting OST \"failover.node\""
1651 $r $LCTL conf_param $fsname-MDT0000.mdc.max_rpcs_in_flight=9 || {
1652 error_noexit "Setting \"max_rpcs_in_flight\""
1655 $r $LCTL conf_param $fsname-MDT0000.failover.node=$nid || {
1656 error_noexit "Setting MDT \"failover.node\""
1659 $r $LCTL pool_new $fsname.interop || {
1660 error_noexit "Setting \"interop\""
1663 $r $LCTL conf_param $fsname-MDT0000.lov.stripesize=4M || {
1664 error_noexit "Setting \"lov.stripesize\""
1668 if [ "$ff_convert" != "no" -a $(facet_fstype ost1) == "ldiskfs" ]; then
1669 $r $LCTL lfsck_start -M $fsname-OST0000 || {
1670 error_noexit "Start OI scrub on OST0"
1674 # The oi_scrub should be on ost1, but for test_32(),
1675 # all on the SINGLEMDS.
1676 wait_update_facet $SINGLEMDS "$LCTL get_param -n \
1677 osd-ldiskfs.$fsname-OST0000.oi_scrub |
1678 awk '/^status/ { print \\\$2 }'" "completed" 30 || {
1679 error_noexit "Failed to get the expected 'completed'"
1683 local UPDATED=$($r $LCTL get_param -n \
1684 osd-ldiskfs.$fsname-OST0000.oi_scrub |
1685 awk '/^updated/ { print $2 }')
1686 [ $UPDATED -ge 1 ] || {
1687 error_noexit "Only $UPDATED objects have been converted"
1692 if [ "$dne_upgrade" != "no" ]; then
1693 $r $LCTL conf_param \
1694 $fsname-MDT0001.mdc.max_rpcs_in_flight=9 || {
1695 error_noexit "Setting MDT1 \"max_rpcs_in_flight\""
1698 $r $LCTL conf_param $fsname-MDT0001.failover.node=$nid || {
1699 error_noexit "Setting MDT1 \"failover.node\""
1702 $r $LCTL conf_param $fsname-MDT0001.lov.stripesize=4M || {
1703 error_noexit "Setting MDT1 \"lov.stripesize\""
1709 if [ "$writeconf" ]; then
1710 mount -t lustre $nid:/$fsname $tmp/mnt/lustre || {
1711 error_noexit "Mounting the client"
1714 shall_cleanup_lustre=true
1715 $LCTL set_param debug="$PTLDEBUG"
1717 t32_verify_quota $node $fsname $tmp/mnt/lustre || {
1718 error_noexit "verify quota failed"
1722 if [ "$dne_upgrade" != "no" ]; then
1723 $LFS mkdir -i 1 $tmp/mnt/lustre/remote_dir || {
1724 error_noexit "set remote dir failed"
1728 pushd $tmp/mnt/lustre
1729 tar -cf - . --exclude=./remote_dir |
1730 tar -xvf - -C remote_dir 1>/dev/null || {
1731 error_noexit "cp to remote dir failed"
1737 dd if=/dev/zero of=$tmp/mnt/lustre/tmp_file bs=10k count=10 || {
1738 error_noexit "dd failed"
1741 rm -rf $tmp/mnt/lustre/tmp_file || {
1742 error_noexit "rm failed"
1746 if $r test -f $tmp/sha1sums; then
1747 # LU-2393 - do both sorts on same node to ensure locale
1749 $r cat $tmp/sha1sums | sort -k 2 >$tmp/sha1sums.orig
1750 if [ "$dne_upgrade" != "no" ]; then
1751 pushd $tmp/mnt/lustre/remote_dir
1753 pushd $tmp/mnt/lustre
1756 find ! -name .lustre -type f -exec sha1sum {} \; |
1757 sort -k 2 >$tmp/sha1sums || {
1758 error_noexit "sha1sum"
1762 if ! diff -ub $tmp/sha1sums.orig $tmp/sha1sums; then
1763 error_noexit "sha1sum verification failed"
1767 echo "sha1sum verification skipped"
1770 if [ "$dne_upgrade" != "no" ]; then
1771 rm -rf $tmp/mnt/lustre/remote_dir || {
1772 error_noexit "remove remote dir failed"
1777 if $r test -f $tmp/list; then
1779 # There is not a Test Framework API to copy files to or
1780 # from a remote node.
1782 # LU-2393 - do both sorts on same node to ensure locale
1784 $r cat $tmp/list | sort -k 6 >$tmp/list.orig
1785 pushd $tmp/mnt/lustre
1786 ls -Rni --time-style=+%s | sort -k 6 >$tmp/list || {
1792 # 32-bit and 64-bit clients use different algorithms to
1793 # convert FIDs into inode numbers. Hence, remove the inode
1794 # numbers from the lists, if the original list was created
1795 # on an architecture with different number of bits per
1798 if [ $(t32_bits_per_long $(uname -m)) != \
1799 $(t32_bits_per_long $img_arch) ]; then
1800 echo "Different number of bits per \"long\" from the disk image"
1801 for list in list.orig list; do
1802 sed -i -e 's/^[0-9]\+[ \t]\+//' $tmp/$list
1805 if ! diff -ub $tmp/list.orig $tmp/list; then
1806 error_noexit "list verification failed"
1810 echo "list verification skipped"
1814 # When adding new data verification tests, please check for
1815 # the presence of the required reference files first, like
1816 # the "sha1sums" and "list" tests above, to avoid the need to
1817 # regenerate every image for each test addition.
1820 nrpcs_orig=$($LCTL get_param \
1821 -n mdc.*MDT0000*.max_rpcs_in_flight) || {
1822 error_noexit "Getting \"max_rpcs_in_flight\""
1825 nrpcs=$((nrpcs_orig + 5))
1826 $r $LCTL conf_param $fsname-MDT0000.mdc.max_rpcs_in_flight=$nrpcs || {
1827 error_noexit "Changing \"max_rpcs_in_flight\""
1830 wait_update $HOSTNAME "$LCTL get_param \
1831 -n mdc.*MDT0000*.max_rpcs_in_flight" $nrpcs || {
1832 error_noexit "Verifying \"max_rpcs_in_flight\""
1836 umount $tmp/mnt/lustre || {
1837 error_noexit "Unmounting the client"
1840 shall_cleanup_lustre=false
1842 if [ "$dne_upgrade" != "no" ]; then
1843 $r umount -d $tmp/mnt/mdt1 || {
1844 error_noexit "Unmounting the MDT2"
1847 shall_cleanup_mdt1=false
1850 $r umount -d $tmp/mnt/mdt || {
1851 error_noexit "Unmounting the MDT"
1854 shall_cleanup_mdt=false
1856 $r umount -d $tmp/mnt/ost || {
1857 error_noexit "Unmounting the OST"
1860 shall_cleanup_ost=false
1862 t32_reload_modules $node || {
1863 error_noexit "Reloading modules"
1867 # mount a second time to make sure we didnt leave upgrade flag on
1868 $r $TUNEFS --dryrun $tmp/mdt || {
1869 error_noexit "tunefs.lustre before remounting the MDT"
1872 $r mount -t lustre -o loop,exclude=$fsname-OST0000 $tmp/mdt \
1874 error_noexit "Remounting the MDT"
1877 shall_cleanup_mdt=true
1887 for tarball in $tarballs; do
1888 t32_test $tarball || let "rc += $?"
1892 run_test 32a "Upgrade (not live)"
1900 for tarball in $tarballs; do
1901 t32_test $tarball writeconf || let "rc += $?"
1905 run_test 32b "Upgrade with writeconf"
1912 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
1914 for tarball in $tarballs; do
1915 dne_upgrade=yes t32_test $tarball writeconf || rc=$?
1919 run_test 32c "dne upgrade test"
1927 for tarball in $tarballs; do
1928 ff_convert=yes t32_test $tarball || rc=$?
1932 run_test 32d "convert ff test"
1934 test_33a() { # bug 12333, was test_33
1936 local FSNAME2=test-123
1937 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
1940 [ -n "$ost1_HOST" ] && fs2ost_HOST=$ost1_HOST
1942 if [ -z "$fs2ost_DEV" -o -z "$fs2mds_DEV" ]; then
1943 local dev=${SINGLEMDS}_dev
1944 local MDSDEV=${!dev}
1945 is_blkdev $SINGLEMDS $MDSDEV && \
1946 skip_env "mixed loopback and real device not working" && return
1949 local fs2mdsdev=$(mdsdevname 1_2)
1950 local fs2ostdev=$(ostdevname 1_2)
1951 local fs2mdsvdev=$(mdsvdevname 1_2)
1952 local fs2ostvdev=$(ostvdevname 1_2)
1954 if [ $(facet_fstype mds1) == ldiskfs ]; then
1955 mkfsoptions="--mkfsoptions=\\\"-J size=8\\\"" # See bug 17931.
1958 add fs2mds $(mkfs_opts mds1 ${fs2mdsdev}) --mgs --fsname=${FSNAME2} \
1959 --reformat $mkfsoptions $fs2mdsdev $fs2mdsvdev || exit 10
1960 add fs2ost $(mkfs_opts ost1 ${fs2ostdev}) --mgsnode=$MGSNID \
1961 --fsname=${FSNAME2} --index=8191 --reformat $fs2ostdev \
1962 $fs2ostvdev || exit 10
1964 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS && trap cleanup_fs2 EXIT INT
1965 start fs2ost $fs2ostdev $OST_MOUNT_OPTS
1966 do_facet $SINGLEMDS "$LCTL conf_param $FSNAME2.sys.timeout=200" || rc=1
1968 mount -t lustre $MGSNID:/${FSNAME2} $MOUNT2 || rc=2
1971 cp /etc/hosts $MOUNT2/ || rc=3
1972 $LFS getstripe $MOUNT2/hosts
1977 cleanup_nocli || rc=6
1980 run_test 33a "Mount ost with a large index number"
1982 test_33b() { # was test_34
1985 do_facet client dd if=/dev/zero of=$MOUNT/24 bs=1024k count=1
1986 # Drop lock cancelation reply during umount
1987 #define OBD_FAIL_LDLM_CANCEL_NET 0x304
1988 do_facet client lctl set_param fail_loc=0x80000304
1989 #lctl set_param debug=-1
1990 umount_client $MOUNT
1993 run_test 33b "Drop cancel during umount"
1997 do_facet client "sh runmultiop_bg_pause $DIR/file O_c"
1998 manual_umount_client
2000 do_facet client killall -USR1 multiop
2001 if [ $rc -eq 0 ]; then
2002 error "umount not fail!"
2007 run_test 34a "umount with opened file should be fail"
2012 touch $DIR/$tfile || return 1
2013 stop_mds --force || return 2
2015 manual_umount_client --force
2017 if [ $rc -ne 0 ]; then
2018 error "mtab after failed umount - rc $rc"
2024 run_test 34b "force umount with failed mds should be normal"
2028 touch $DIR/$tfile || return 1
2029 stop_ost --force || return 2
2031 manual_umount_client --force
2033 if [ $rc -ne 0 ]; then
2034 error "mtab after failed umount - rc $rc"
2040 run_test 34c "force umount with failed ost should be normal"
2042 test_35a() { # bug 12459
2045 DBG_SAVE="`lctl get_param -n debug`"
2046 lctl set_param debug="ha"
2048 log "Set up a fake failnode for the MDS"
2050 local device=$(do_facet $SINGLEMDS "lctl get_param -n devices" |
2051 awk '($3 ~ "mdt" && $4 ~ "MDT") { print $4 }' | head -1)
2052 do_facet mgs "$LCTL conf_param \
2053 ${device}.failover.node=$(h2$NETTYPE $FAKENID)" || return 4
2055 log "Wait for RECONNECT_INTERVAL seconds (10s)"
2058 MSG="conf-sanity.sh test_35a `date +%F%kh%Mm%Ss`"
2061 log "Stopping the MDT: $device"
2062 stop_mdt 1 || return 5
2064 df $MOUNT > /dev/null 2>&1 &
2066 log "Restarting the MDT: $device"
2067 start_mdt 1 || return 6
2068 log "Wait for df ($DFPID) ... "
2071 lctl set_param debug="$DBG_SAVE"
2073 # retrieve from the log the first server that the client tried to
2074 # contact after the connection loss
2075 $LCTL dk $TMP/lustre-log-$TESTNAME.log
2076 NEXTCONN=`awk "/${MSG}/ {start = 1;}
2077 /import_select_connection.*$device-mdc.* using connection/ {
2079 if (\\\$NF ~ /$FAKENID/)
2085 }" $TMP/lustre-log-$TESTNAME.log`
2086 [ "$NEXTCONN" != "0" ] && log "The client didn't try to reconnect to the last active server (tried ${NEXTCONN} instead)" && return 7
2088 # remove nid settings
2089 writeconf_or_reformat
2091 run_test 35a "Reconnect to the last active server first"
2093 test_35b() { # bug 18674
2094 remote_mds || { skip "local MDS" && return 0; }
2098 $LCTL set_param debug="ha"
2100 MSG="conf-sanity.sh test_35b `date +%F%kh%Mm%Ss`"
2103 log "Set up a fake failnode for the MDS"
2105 local device=$(do_facet $SINGLEMDS "$LCTL get_param -n devices" |
2106 awk '($3 ~ "mdt" && $4 ~ "MDT") { print $4 }' | head -1)
2107 do_facet mgs "$LCTL conf_param \
2108 ${device}.failover.node=$(h2$NETTYPE $FAKENID)" || return 1
2110 local at_max_saved=0
2111 # adaptive timeouts may prevent seeing the issue
2112 if at_is_enabled; then
2113 at_max_saved=$(at_max_get mds)
2114 at_max_set 0 mds client
2117 mkdir -p $MOUNT/$tdir
2119 log "Injecting EBUSY on MDS"
2120 # Setting OBD_FAIL_MDS_RESEND=0x136
2121 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0x80000136" || return 2
2123 $LCTL set_param mdc.${FSNAME}*.stats=clear
2125 log "Creating a test file and stat it"
2126 touch $MOUNT/$tdir/$tfile
2127 stat $MOUNT/$tdir/$tfile
2129 log "Stop injecting EBUSY on MDS"
2130 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0" || return 3
2131 rm -f $MOUNT/$tdir/$tfile
2134 # restore adaptive timeout
2135 [ $at_max_saved -ne 0 ] && at_max_set $at_max_saved mds client
2137 $LCTL dk $TMP/lustre-log-$TESTNAME.log
2139 CONNCNT=`$LCTL get_param mdc.${FSNAME}*.stats | awk '/mds_connect/{print $2}'`
2141 # retrieve from the log if the client has ever tried to
2142 # contact the fake server after the loss of connection
2143 FAILCONN=`awk "BEGIN {ret = 0;}
2144 /import_select_connection.*${FSNAME}-MDT0000-mdc.* using connection/ {
2146 if (\\\$NF ~ /$FAKENID/) {
2151 END {print ret}" $TMP/lustre-log-$TESTNAME.log`
2153 [ "$FAILCONN" == "0" ] && \
2154 log "ERROR: The client reconnection has not been triggered" && \
2156 [ "$FAILCONN" == "2" ] && \
2157 log "ERROR: The client tried to reconnect to the failover server while the primary was busy" && \
2161 # When OBD_FAIL_MDS_RESEND is hit, we sleep for 2 * obd_timeout
2162 # Reconnects are supposed to be rate limited to one every 5s
2163 [ $CONNCNT -gt $((2 * $TIMEOUT / 5 + 1)) ] && \
2164 log "ERROR: Too many reconnects $CONNCNT" && \
2168 # remove nid settings
2169 writeconf_or_reformat
2171 run_test 35b "Continue reconnection retries, if the active server is busy"
2174 [ $OSTCOUNT -lt 2 ] && skip_env "skipping test for single OST" && return
2176 [ "$ost_HOST" = "`hostname`" -o "$ost1_HOST" = "`hostname`" ] || \
2177 { skip "remote OST" && return 0; }
2180 local FSNAME2=test1234
2181 local fs3ost_HOST=$ost_HOST
2182 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
2184 [ -n "$ost1_HOST" ] && fs2ost_HOST=$ost1_HOST && fs3ost_HOST=$ost1_HOST
2186 if [ -z "$fs2ost_DEV" -o -z "$fs2mds_DEV" -o -z "$fs3ost_DEV" ]; then
2187 is_blkdev $SINGLEMDS $MDSDEV && \
2188 skip_env "mixed loopback and real device not working" && return
2191 local fs2mdsdev=$(mdsdevname 1_2)
2192 local fs2ostdev=$(ostdevname 1_2)
2193 local fs3ostdev=$(ostdevname 2_2)
2194 local fs2mdsvdev=$(mdsvdevname 1_2)
2195 local fs2ostvdev=$(ostvdevname 1_2)
2196 local fs3ostvdev=$(ostvdevname 2_2)
2198 add fs2mds $(mkfs_opts mds1 ${fs2mdsdev}) --mgs --fsname=${FSNAME2} \
2199 --reformat $fs2mdsdev $fs2mdsvdev || exit 10
2200 # XXX after we support non 4K disk blocksize in ldiskfs, specify a
2201 # different one than the default value here.
2202 add fs2ost $(mkfs_opts ost1 ${fs2ostdev}) --mgsnode=$MGSNID \
2203 --fsname=${FSNAME2} --reformat $fs2ostdev $fs2ostvdev || exit 10
2204 add fs3ost $(mkfs_opts ost1 ${fs3ostdev}) --mgsnode=$MGSNID \
2205 --fsname=${FSNAME2} --reformat $fs3ostdev $fs3ostvdev || exit 10
2207 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS
2208 start fs2ost $fs2ostdev $OST_MOUNT_OPTS
2209 start fs3ost $fs3ostdev $OST_MOUNT_OPTS
2211 mount -t lustre $MGSNID:/${FSNAME2} $MOUNT2 || return 1
2213 sleep 5 # until 11778 fixed
2215 dd if=/dev/zero of=$MOUNT2/$tfile bs=1M count=7 || return 2
2217 BKTOTAL=`lctl get_param -n obdfilter.*.kbytestotal | awk 'BEGIN{total=0}; {total+=$1}; END{print total}'`
2218 BKFREE=`lctl get_param -n obdfilter.*.kbytesfree | awk 'BEGIN{free=0}; {free+=$1}; END{print free}'`
2219 BKAVAIL=`lctl get_param -n obdfilter.*.kbytesavail | awk 'BEGIN{avail=0}; {avail+=$1}; END{print avail}'`
2220 STRING=`df -P $MOUNT2 | tail -n 1 | awk '{print $2","$3","$4}'`
2221 DFTOTAL=`echo $STRING | cut -d, -f1`
2222 DFUSED=`echo $STRING | cut -d, -f2`
2223 DFAVAIL=`echo $STRING | cut -d, -f3`
2224 DFFREE=$(($DFTOTAL - $DFUSED))
2226 ALLOWANCE=$((64 * $OSTCOUNT))
2228 if [ $DFTOTAL -lt $(($BKTOTAL - $ALLOWANCE)) ] ||
2229 [ $DFTOTAL -gt $(($BKTOTAL + $ALLOWANCE)) ] ; then
2230 echo "**** FAIL: df total($DFTOTAL) mismatch OST total($BKTOTAL)"
2233 if [ $DFFREE -lt $(($BKFREE - $ALLOWANCE)) ] ||
2234 [ $DFFREE -gt $(($BKFREE + $ALLOWANCE)) ] ; then
2235 echo "**** FAIL: df free($DFFREE) mismatch OST free($BKFREE)"
2238 if [ $DFAVAIL -lt $(($BKAVAIL - $ALLOWANCE)) ] ||
2239 [ $DFAVAIL -gt $(($BKAVAIL + $ALLOWANCE)) ] ; then
2240 echo "**** FAIL: df avail($DFAVAIL) mismatch OST avail($BKAVAIL)"
2245 stop fs3ost -f || return 200
2246 stop fs2ost -f || return 201
2247 stop fs2mds -f || return 202
2248 unload_modules_conf || return 203
2251 run_test 36 "df report consistency on OSTs with different block size"
2254 local mntpt=$(facet_mntpt $SINGLEMDS)
2255 local mdsdev=$(mdsdevname ${SINGLEMDS//mds/})
2256 local mdsdev_sym="$TMP/sym_mdt.img"
2257 local opts=$MDS_MOUNT_OPTS
2260 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
2261 skip "Currently only applicable to ldiskfs-based MDTs"
2265 echo "MDS : $mdsdev"
2266 echo "SYMLINK : $mdsdev_sym"
2267 do_facet $SINGLEMDS rm -f $mdsdev_sym
2269 do_facet $SINGLEMDS ln -s $mdsdev $mdsdev_sym
2271 echo "mount symlink device - $mdsdev_sym"
2273 if ! do_facet $SINGLEMDS test -b $mdsdev; then
2274 opts=$(csa_add "$opts" -o loop)
2276 mount_op=$(do_facet $SINGLEMDS mount -v -t lustre $opts \
2277 $mdsdev_sym $mntpt 2>&1)
2280 echo mount_op=$mount_op
2282 do_facet $SINGLEMDS "umount -d $mntpt && rm -f $mdsdev_sym"
2284 if $(echo $mount_op | grep -q "unable to set tunable"); then
2285 error "set tunables failed for symlink device"
2288 [ $rc -eq 0 ] || error "mount symlink $mdsdev_sym failed! rc=$rc"
2292 run_test 37 "verify set tunables works for symlink device"
2294 test_38() { # bug 14222
2295 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
2296 skip "Only applicable to ldiskfs-based MDTs"
2304 FILES=`find $SRC -type f -mtime +1 | head -n $COUNT`
2305 log "copying $(echo $FILES | wc -w) files to $DIR/$tdir"
2307 tar cf - $FILES | tar xf - -C $DIR/$tdir || \
2308 error "copying $SRC to $DIR/$tdir"
2310 umount_client $MOUNT
2312 log "rename lov_objid file on MDS"
2313 rm -f $TMP/lov_objid.orig
2315 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
2316 do_facet $SINGLEMDS "$DEBUGFS -c -R \\\"dump lov_objid $TMP/lov_objid.orig\\\" $MDSDEV"
2317 do_facet $SINGLEMDS "$DEBUGFS -w -R \\\"rm lov_objid\\\" $MDSDEV"
2319 do_facet $SINGLEMDS "od -Ax -td8 $TMP/lov_objid.orig"
2320 # check create in mds_lov_connect
2324 [ $V ] && log "verifying $DIR/$tdir/$f"
2325 diff -q $f $DIR/$tdir/$f || ERROR=y
2327 do_facet $SINGLEMDS "$DEBUGFS -c -R \\\"dump lov_objid $TMP/lov_objid.new\\\" $MDSDEV"
2328 do_facet $SINGLEMDS "od -Ax -td8 $TMP/lov_objid.new"
2329 [ "$ERROR" = "y" ] && error "old and new files are different after connect" || true
2331 # check it's updates in sync
2332 umount_client $MOUNT
2335 do_facet $SINGLEMDS dd if=/dev/zero of=$TMP/lov_objid.clear bs=4096 count=1
2336 do_facet $SINGLEMDS "$DEBUGFS -w -R \\\"rm lov_objid\\\" $MDSDEV"
2337 do_facet $SINGLEMDS "$DEBUGFS -w -R \\\"write $TMP/lov_objid.clear lov_objid\\\" $MDSDEV "
2342 [ $V ] && log "verifying $DIR/$tdir/$f"
2343 diff -q $f $DIR/$tdir/$f || ERROR=y
2345 do_facet $SINGLEMDS "$DEBUGFS -c -R \\\"dump lov_objid $TMP/lov_objid.new1\\\" $MDSDEV"
2346 do_facet $SINGLEMDS "od -Ax -td8 $TMP/lov_objid.new1"
2347 umount_client $MOUNT
2349 [ "$ERROR" = "y" ] && error "old and new files are different after sync" || true
2351 log "files compared the same"
2354 run_test 38 "MDS recreates missing lov_objid file from OST data"
2360 perl $SRCDIR/leak_finder.pl $TMP/debug 2>&1 | egrep '*** Leak:' &&
2361 error "memory leak detected" || true
2363 run_test 39 "leak_finder recognizes both LUSTRE and LNET malloc messages"
2365 test_40() { # bug 15759
2367 #define OBD_FAIL_TGT_TOOMANY_THREADS 0x706
2368 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0x80000706"
2372 run_test 40 "race during service thread startup"
2374 test_41a() { #bug 14134
2375 if [ $(facet_fstype $SINGLEMDS) == ldiskfs ] &&
2376 ! do_facet $SINGLEMDS test -b $(mdsdevname 1); then
2377 skip "Loop devices does not work with nosvc option"
2382 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
2384 start $SINGLEMDS $MDSDEV $MDS_MOUNT_OPTS -o nosvc -n
2385 start ost1 `ostdevname 1` $OST_MOUNT_OPTS
2386 start $SINGLEMDS $MDSDEV $MDS_MOUNT_OPTS -o nomgs,force
2388 mount_client $MOUNT || return 1
2391 echo "blah blah" > $MOUNT/$tfile
2394 umount_client $MOUNT
2395 stop ost1 -f || return 201
2396 stop_mds -f || return 202
2397 stop_mds -f || return 203
2398 unload_modules_conf || return 204
2401 run_test 41a "mount mds with --nosvc and --nomgs"
2404 if [ $(facet_fstype $SINGLEMDS) == ldiskfs ] &&
2405 ! do_facet $SINGLEMDS test -b $(mdsdevname 1); then
2406 skip "Loop devices does not work with nosvc option"
2410 ! combined_mgs_mds && skip "needs combined mgs device" && return 0
2414 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
2416 start $SINGLEMDS $MDSDEV $MDS_MOUNT_OPTS -o nosvc -n
2418 start $SINGLEMDS $MDSDEV $MDS_MOUNT_OPTS -o nomgs,force
2420 mount_client $MOUNT || return 1
2423 echo "blah blah" > $MOUNT/$tfile
2424 cat $MOUNT/$tfile || return 200
2426 umount_client $MOUNT
2427 stop_ost || return 201
2428 stop_mds -f || return 202
2429 stop_mds -f || return 203
2432 run_test 41b "mount mds with --nosvc and --nomgs on first mount"
2434 test_42() { #bug 14693
2436 check_mount || error "client was not mounted"
2438 do_facet mgs $LCTL conf_param $FSNAME.llite.some_wrong_param=10
2439 umount_client $MOUNT ||
2440 error "unmounting client failed with invalid llite param"
2441 mount_client $MOUNT ||
2442 error "mounting client failed with invalid llite param"
2444 do_facet mgs $LCTL conf_param $FSNAME.sys.some_wrong_param=20
2445 cleanup || error "stopping $FSNAME failed with invalid sys param"
2448 check_mount || "client was not mounted with invalid sys param"
2449 cleanup || error "stopping $FSNAME failed with invalid sys param"
2452 run_test 42 "allow client/server mount/unmount with invalid config param"
2455 [ $UID -ne 0 -o $RUNAS_ID -eq 0 ] && skip_env "run as root"
2457 chmod ugo+x $DIR || error "chmod 0 failed"
2458 set_conf_param_and_check mds \
2459 "lctl get_param -n mdt.$FSNAME-MDT0000.root_squash" \
2460 "$FSNAME.mdt.root_squash" \
2462 set_conf_param_and_check mds \
2463 "lctl get_param -n mdt.$FSNAME-MDT0000.nosquash_nids" \
2464 "$FSNAME.mdt.nosquash_nids" \
2468 # create set of test files
2470 echo "111" > $DIR/$tfile-userfile || error "write 1 failed"
2471 chmod go-rw $DIR/$tfile-userfile || error "chmod 1 failed"
2472 chown $RUNAS_ID.$RUNAS_ID $DIR/$tfile-userfile || error "chown failed"
2474 echo "222" > $DIR/$tfile-rootfile || error "write 2 failed"
2475 chmod go-rw $DIR/$tfile-rootfile || error "chmod 2 faield"
2477 mkdir $DIR/$tdir-rootdir -p || error "mkdir failed"
2478 chmod go-rwx $DIR/$tdir-rootdir || error "chmod 3 failed"
2479 touch $DIR/$tdir-rootdir/tfile-1 || error "touch failed"
2482 # check root_squash:
2483 # set root squash UID:GID to RUNAS_ID
2484 # root should be able to access only files owned by RUNAS_ID
2486 set_conf_param_and_check mds \
2487 "lctl get_param -n mdt.$FSNAME-MDT0000.root_squash" \
2488 "$FSNAME.mdt.root_squash" \
2489 "$RUNAS_ID:$RUNAS_ID"
2491 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-userfile)
2492 dd if=$DIR/$tfile-userfile 1>/dev/null 2>/dev/null || \
2493 error "$ST: root read permission is denied"
2494 echo "$ST: root read permission is granted - ok"
2497 dd conv=notrunc if=$DIR/$tfile-userfile 1>/dev/null 2>/dev/null || \
2498 error "$ST: root write permission is denied"
2499 echo "$ST: root write permission is granted - ok"
2501 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-rootfile)
2502 dd if=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null && \
2503 error "$ST: root read permission is granted"
2504 echo "$ST: root read permission is denied - ok"
2507 dd conv=notrunc of=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null && \
2508 error "$ST: root write permission is granted"
2509 echo "$ST: root write permission is denied - ok"
2511 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tdir-rootdir)
2512 rm $DIR/$tdir-rootdir/tfile-1 1>/dev/null 2>/dev/null && \
2513 error "$ST: root unlink permission is granted"
2514 echo "$ST: root unlink permission is denied - ok"
2516 touch $DIR/tdir-rootdir/tfile-2 1>/dev/null 2>/dev/null && \
2517 error "$ST: root create permission is granted"
2518 echo "$ST: root create permission is denied - ok"
2521 # check nosquash_nids:
2522 # put client's NID into nosquash_nids list,
2523 # root should be able to access root file after that
2525 local NIDLIST=$(lctl list_nids all | tr '\n' ' ')
2526 NIDLIST="2@elan $NIDLIST 192.168.0.[2,10]@tcp"
2527 NIDLIST=$(echo $NIDLIST | tr -s ' ' ' ')
2528 set_conf_param_and_check mds \
2529 "lctl get_param -n mdt.$FSNAME-MDT0000.nosquash_nids" \
2530 "$FSNAME-MDTall.mdt.nosquash_nids" \
2533 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-rootfile)
2534 dd if=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null || \
2535 error "$ST: root read permission is denied"
2536 echo "$ST: root read permission is granted - ok"
2539 dd conv=notrunc of=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null || \
2540 error "$ST: root write permission is denied"
2541 echo "$ST: root write permission is granted - ok"
2543 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tdir-rootdir)
2544 rm $DIR/$tdir-rootdir/tfile-1 || \
2545 error "$ST: root unlink permission is denied"
2546 echo "$ST: root unlink permission is granted - ok"
2547 touch $DIR/$tdir-rootdir/tfile-2 || \
2548 error "$ST: root create permission is denied"
2549 echo "$ST: root create permission is granted - ok"
2553 run_test 43 "check root_squash and nosquash_nids"
2555 umount_client $MOUNT
2560 check_mount || return 2
2561 UUID=$($LCTL get_param llite.${FSNAME}*.uuid | cut -d= -f2)
2563 UUIDS=$(do_facet $SINGLEMDS "$LCTL get_param mdt.${FSNAME}*.exports.*.uuid")
2564 for VAL in $UUIDS; do
2565 NID=$(echo $VAL | cut -d= -f1)
2566 CLUUID=$(echo $VAL | cut -d= -f2)
2567 [ "$UUID" = "$CLUUID" ] && STATS_FOUND=yes && break
2569 [ "$STATS_FOUND" = "no" ] && error "stats not found for client"
2573 run_test 44 "mounted client proc entry exists"
2577 check_mount || return 2
2582 #define OBD_FAIL_PTLRPC_LONG_UNLINK 0x50f
2583 do_facet client "lctl set_param fail_loc=0x50f"
2586 manual_umount_client --force || return 3
2587 do_facet client "lctl set_param fail_loc=0x0"
2589 mount_client $MOUNT || return 4
2593 run_test 45 "long unlink handling in ptlrpcd"
2600 umount_client $MOUNT2 || rc=$?
2601 umount_client $MOUNT || rc=$?
2602 while [ $count -gt 0 ]; do
2603 stop ost${count} -f || rc=$?
2607 cleanup_nocli || rc=$?
2608 #writeconf to remove all ost2 traces for subsequent tests
2609 writeconf_or_reformat
2614 echo "Testing with $OSTCOUNT OSTs"
2616 start_mds || return 1
2617 #first client should see only one ost
2618 start_ost || return 2
2619 wait_osc_import_state mds ost FULL
2621 mount_client $MOUNT || return 3
2622 trap "cleanup_46a $OSTCOUNT" EXIT ERR
2625 for (( i=2; i<=$OSTCOUNT; i++ )); do
2626 start ost$i `ostdevname $i` $OST_MOUNT_OPTS || return $((i+2))
2629 # wait until osts in sync
2630 for (( i=2; i<=$OSTCOUNT; i++ )); do
2631 wait_osc_import_state mds ost$i FULL
2632 wait_osc_import_state client ost$i FULL
2635 #second client see all ost's
2637 mount_client $MOUNT2 || return 8
2638 $LFS setstripe -c -1 $MOUNT2 || return 9
2639 $LFS getstripe $MOUNT2 || return 10
2641 echo "ok" > $MOUNT2/widestripe
2642 $LFS getstripe $MOUNT2/widestripe || return 11
2643 # fill acl buffer for avoid expand lsm to them
2644 awk -F : '{if (FNR < 25) { print "u:"$1":rwx" }}' /etc/passwd | while read acl; do
2645 setfacl -m $acl $MOUNT2/widestripe
2649 stat $MOUNT/widestripe || return 12
2651 cleanup_46a $OSTCOUNT || { echo "cleanup_46a failed!" && return 13; }
2654 run_test 46a "handle ost additional - wide striped file"
2659 check_mount || return 2
2660 $LCTL set_param ldlm.namespaces.$FSNAME-*-*-*.lru_size=100
2664 for ns in $($LCTL get_param ldlm.namespaces.$FSNAME-*-*-*.lru_size); do
2665 if echo $ns | grep "MDT[[:digit:]]*"; then
2668 lrs=$(echo $ns | sed 's/.*lru_size=//')
2669 lru_size[count]=$lrs
2674 facet_failover $SINGLEMDS
2675 client_up || return 3
2678 for ns in $($LCTL get_param ldlm.namespaces.$FSNAME-*-*-*.lru_size); do
2679 if echo $ns | grep "MDT[[:digit:]]*"; then
2682 lrs=$(echo $ns | sed 's/.*lru_size=//')
2683 if ! test "$lrs" -eq "${lru_size[count]}"; then
2684 n=$(echo $ns | sed -e 's/ldlm.namespaces.//' -e 's/.lru_size=.*//')
2685 error "$n has lost lru_size: $lrs vs. ${lru_size[count]}"
2693 run_test 47 "server restart does not make client loss lru_resize settings"
2698 # reformat after this test is needed - if test will failed
2699 # we will have unkillable file at FS
2703 test_48() { # bug 17636
2706 check_mount || return 2
2708 $LFS setstripe -c -1 $MOUNT || return 9
2709 $LFS getstripe $MOUNT || return 10
2711 echo "ok" > $MOUNT/widestripe
2712 $LFS getstripe $MOUNT/widestripe || return 11
2714 trap cleanup_48 EXIT ERR
2716 # fill acl buffer for avoid expand lsm to them
2717 getent passwd | awk -F : '{ print "u:"$1":rwx" }' | while read acl; do
2718 setfacl -m $acl $MOUNT/widestripe
2721 stat $MOUNT/widestripe || return 12
2726 run_test 48 "too many acls on file"
2728 # check PARAM_SYS_LDLM_TIMEOUT option of MKFS.LUSTRE
2729 test_49() { # bug 17710
2730 local timeout_orig=$TIMEOUT
2731 local ldlm_timeout_orig=$LDLM_TIMEOUT
2732 local LOCAL_TIMEOUT=20
2734 LDLM_TIMEOUT=$LOCAL_TIMEOUT
2735 TIMEOUT=$LOCAL_TIMEOUT
2739 check_mount || return 1
2741 echo "check ldlm_timout..."
2742 LDLM_MDS="`do_facet $SINGLEMDS lctl get_param -n ldlm_timeout`"
2743 LDLM_OST1="`do_facet ost1 lctl get_param -n ldlm_timeout`"
2744 LDLM_CLIENT="`do_facet client lctl get_param -n ldlm_timeout`"
2746 if [ $LDLM_MDS -ne $LDLM_OST1 ] || [ $LDLM_MDS -ne $LDLM_CLIENT ]; then
2747 error "Different LDLM_TIMEOUT:$LDLM_MDS $LDLM_OST1 $LDLM_CLIENT"
2750 if [ $LDLM_MDS -ne $((LOCAL_TIMEOUT / 3)) ]; then
2751 error "LDLM_TIMEOUT($LDLM_MDS) is not correct"
2754 umount_client $MOUNT
2755 stop_ost || return 2
2756 stop_mds || return 3
2758 LDLM_TIMEOUT=$((LOCAL_TIMEOUT - 1))
2762 check_mount || return 7
2764 LDLM_MDS="`do_facet $SINGLEMDS lctl get_param -n ldlm_timeout`"
2765 LDLM_OST1="`do_facet ost1 lctl get_param -n ldlm_timeout`"
2766 LDLM_CLIENT="`do_facet client lctl get_param -n ldlm_timeout`"
2768 if [ $LDLM_MDS -ne $LDLM_OST1 ] || [ $LDLM_MDS -ne $LDLM_CLIENT ]; then
2769 error "Different LDLM_TIMEOUT:$LDLM_MDS $LDLM_OST1 $LDLM_CLIENT"
2772 if [ $LDLM_MDS -ne $((LOCAL_TIMEOUT - 1)) ]; then
2773 error "LDLM_TIMEOUT($LDLM_MDS) is not correct"
2776 cleanup || return $?
2778 LDLM_TIMEOUT=$ldlm_timeout_orig
2779 TIMEOUT=$timeout_orig
2781 run_test 49 "check PARAM_SYS_LDLM_TIMEOUT option of MKFS.LUSTRE"
2784 # Test both statfs and lfs df and fail if either one fails
2785 multiop_bg_pause $1 f_
2788 killall -USR1 multiop
2789 [ $RC1 -ne 0 ] && log "lazystatfs multiop failed"
2790 wait $PID || { RC1=$?; log "multiop return error "; }
2797 if [ $RC2 -eq 0 ]; then
2799 log "lazystatfs df failed"
2803 [[ $RC1 -ne 0 || $RC2 -eq 0 ]] && RC=1
2809 lctl set_param llite.$FSNAME-*.lazystatfs=1
2812 lazystatfs $MOUNT || error "lazystatfs failed but no down servers"
2814 cleanup || return $?
2816 run_test 50a "lazystatfs all servers available =========================="
2820 lctl set_param llite.$FSNAME-*.lazystatfs=1
2823 # Wait for client to detect down OST
2824 stop_ost || error "Unable to stop OST1"
2825 wait_osc_import_state mds ost DISCONN
2827 lazystatfs $MOUNT || error "lazystatfs should don't have returned EIO"
2829 umount_client $MOUNT || error "Unable to unmount client"
2830 stop_mds || error "Unable to stop MDS"
2832 run_test 50b "lazystatfs all servers down =========================="
2835 start_mds || error "Unable to start MDS"
2836 start_ost || error "Unable to start OST1"
2837 start_ost2 || error "Unable to start OST2"
2838 mount_client $MOUNT || error "Unable to mount client"
2839 lctl set_param llite.$FSNAME-*.lazystatfs=1
2842 # Wait for client to detect down OST
2843 stop_ost || error "Unable to stop OST1"
2844 wait_osc_import_state mds ost DISCONN
2845 lazystatfs $MOUNT || error "lazystatfs failed with one down server"
2847 umount_client $MOUNT || error "Unable to unmount client"
2848 stop_ost2 || error "Unable to stop OST2"
2849 stop_mds || error "Unable to stop MDS"
2850 #writeconf to remove all ost2 traces for subsequent tests
2851 writeconf_or_reformat
2853 run_test 50c "lazystatfs one server down =========================="
2856 start_mds || error "Unable to start MDS"
2857 start_ost || error "Unable to start OST1"
2858 start_ost2 || error "Unable to start OST2"
2859 mount_client $MOUNT || error "Unable to mount client"
2860 lctl set_param llite.$FSNAME-*.lazystatfs=1
2863 # Issue the statfs during the window where the client still
2864 # belives the OST to be available but it is in fact down.
2865 # No failure just a statfs which hangs for a timeout interval.
2866 stop_ost || error "Unable to stop OST1"
2867 lazystatfs $MOUNT || error "lazystatfs failed with one down server"
2869 umount_client $MOUNT || error "Unable to unmount client"
2870 stop_ost2 || error "Unable to stop OST2"
2871 stop_mds || error "Unable to stop MDS"
2872 #writeconf to remove all ost2 traces for subsequent tests
2873 writeconf_or_reformat
2875 run_test 50d "lazystatfs client/server conn race =========================="
2882 start_mds || return 1
2883 #first client should see only one ost
2884 start_ost || return 2
2885 wait_osc_import_state mds ost FULL
2887 # Wait for client to detect down OST
2888 stop_ost || error "Unable to stop OST1"
2889 wait_osc_import_state mds ost DISCONN
2891 mount_client $MOUNT || error "Unable to mount client"
2892 lctl set_param llite.$FSNAME-*.lazystatfs=0
2894 multiop_bg_pause $MOUNT _f
2898 if [ $RC1 -ne 0 ]; then
2899 log "multiop failed $RC1"
2902 sleep $(( $TIMEOUT+1 ))
2904 [ $? -ne 0 ] && error "process isn't sleep"
2905 start_ost || error "Unable to start OST1"
2906 wait $pid || error "statfs failed"
2909 umount_client $MOUNT || error "Unable to unmount client"
2910 stop_ost || error "Unable to stop OST1"
2911 stop_mds || error "Unable to stop MDS"
2913 run_test 50e "normal statfs all servers down =========================="
2918 CONN_PROC="osc.$FSNAME-OST0001-osc-[M]*.ost_server_uuid"
2920 start_mds || error "Unable to start mds"
2921 #first client should see only one ost
2922 start_ost || error "Unable to start OST1"
2923 wait_osc_import_state mds ost FULL
2925 start_ost2 || error "Unable to start OST2"
2926 wait_osc_import_state mds ost2 FULL
2928 # Wait for client to detect down OST
2929 stop_ost2 || error "Unable to stop OST2"
2931 wait_osc_import_state mds ost2 DISCONN
2932 mount_client $MOUNT || error "Unable to mount client"
2933 lctl set_param llite.$FSNAME-*.lazystatfs=0
2935 multiop_bg_pause $MOUNT _f
2939 if [ $RC1 -ne 0 ]; then
2940 log "lazystatfs multiop failed $RC1"
2943 sleep $(( $TIMEOUT+1 ))
2945 [ $? -ne 0 ] && error "process isn't sleep"
2946 start_ost2 || error "Unable to start OST2"
2947 wait $pid || error "statfs failed"
2948 stop_ost2 || error "Unable to stop OST2"
2951 umount_client $MOUNT || error "Unable to unmount client"
2952 stop_ost || error "Unable to stop OST1"
2953 stop_mds || error "Unable to stop MDS"
2954 #writeconf to remove all ost2 traces for subsequent tests
2955 writeconf_or_reformat
2957 run_test 50f "normal statfs one server in down =========================="
2960 [ "$OSTCOUNT" -lt "2" ] && skip_env "$OSTCOUNT < 2, skipping" && return
2962 start_ost2 || error "Unable to start OST2"
2963 wait_osc_import_state mds ost2 FULL
2964 wait_osc_import_state client ost2 FULL
2966 local PARAM="${FSNAME}-OST0001.osc.active"
2968 $LFS setstripe -c -1 $DIR/$tfile || error "Unable to lfs setstripe"
2969 do_facet mgs $LCTL conf_param $PARAM=0 || error "Unable to deactivate OST"
2971 umount_client $MOUNT || error "Unable to unmount client"
2972 mount_client $MOUNT || error "Unable to mount client"
2973 # This df should not cause a panic
2976 do_facet mgs $LCTL conf_param $PARAM=1 || error "Unable to activate OST"
2978 umount_client $MOUNT || error "Unable to unmount client"
2979 stop_ost2 || error "Unable to stop OST2"
2980 stop_ost || error "Unable to stop OST1"
2981 stop_mds || error "Unable to stop MDS"
2982 #writeconf to remove all ost2 traces for subsequent tests
2983 writeconf_or_reformat
2985 run_test 50g "deactivated OST should not cause panic====================="
2989 # prepare MDT/OST, make OSC inactive for OST1
2990 [ "$OSTCOUNT" -lt "2" ] && skip_env "$OSTCOUNT < 2, skipping" && return
2991 do_facet ost1 "$TUNEFS --param osc.active=0 `ostdevname 1`" ||
2992 error "tunefs OST1 failed"
2993 start_mds || error "Unable to start MDT"
2994 start_ost || error "Unable to start OST1"
2995 start_ost2 || error "Unable to start OST2"
2996 mount_client $MOUNT || error "client start failed"
3000 # activatate OSC for OST1
3001 local TEST="$LCTL get_param -n osc.${FSNAME}-OST0000-osc-[!M]*.active"
3002 set_conf_param_and_check client \
3003 "$TEST" "${FSNAME}-OST0000.osc.active" 1 ||
3004 error "Unable to activate OST1"
3006 mkdir -p $DIR/$tdir/2
3007 $LFS setstripe -c -1 -i 0 $DIR/$tdir/2
3008 sleep 1 && echo "create a file after OST1 is activated"
3010 createmany -o $DIR/$tdir/2/$tfile-%d 1
3012 # check OSC import is working
3013 stat $DIR/$tdir/2/* >/dev/null 2>&1 ||
3014 error "some OSC imports are still not connected"
3017 umount_client $MOUNT || error "Unable to umount client"
3018 stop_ost2 || error "Unable to stop OST2"
3021 run_test 50h "LU-642: activate deactivated OST ==="
3024 local LOCAL_TIMEOUT=20
3028 check_mount || return 1
3031 $LFS setstripe -c -1 $MOUNT/d1
3032 #define OBD_FAIL_MDS_REINT_DELAY 0x142
3033 do_facet $SINGLEMDS "lctl set_param fail_loc=0x142"
3034 touch $MOUNT/d1/f1 &
3037 start_ost2 || return 2
3039 stop_ost2 || return 3
3041 #writeconf to remove all ost2 traces for subsequent tests
3042 writeconf_or_reformat
3044 run_test 51 "Verify that mdt_reint handles RMF_MDT_MD correctly when an OST is added"
3053 do_node $node mkdir -p $dest
3054 [ $? -eq 0 ] || { error "Unable to create directory"; return 1; }
3056 do_node $node 'tar cf - '$@' | tar xf - -C '$dest';
3057 [ \"\${PIPESTATUS[*]}\" = \"0 0\" ] || exit 1'
3058 [ $? -eq 0 ] || { error "Unable to tar files"; return 2; }
3060 do_node $node 'getfattr -d -m "[a-z]*\\." '$@' > '$xattrs
3061 [ $? -eq 0 ] || { error "Unable to read xattrs"; return 3; }
3071 local backup2=${TMP}/backup2
3073 do_node $node mkdir -p $backup2
3074 [ $? -eq 0 ] || { error "Unable to create directory"; return 1; }
3076 do_node $node 'tar cf - '$@' | tar xf - -C '$backup2';
3077 [ \"\${PIPESTATUS[*]}\" = \"0 0\" ] || exit 1'
3078 [ $? -eq 0 ] || { error "Unable to tar files to diff"; return 2; }
3080 do_node $node "diff -rq $backup $backup2"
3081 [ $? -eq 0 ] || { error "contents differ"; return 3; }
3083 local xattrs2=${TMP}/xattrs2
3084 do_node $node 'getfattr -d -m "[a-z]*\\." '$@' > '$xattrs2
3085 [ $? -eq 0 ] || { error "Unable to read xattrs to diff"; return 4; }
3087 do_node $node "diff $xattrs $xattrs2"
3088 [ $? -eq 0 ] || { error "xattrs differ"; return 5; }
3090 do_node $node "rm -rf $backup2 $xattrs2"
3091 [ $? -eq 0 ] || { error "Unable to delete temporary files"; return 6; }
3095 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
3096 skip "Only applicable to ldiskfs-based MDTs"
3101 [ $? -eq 0 ] || { error "Unable to start MDS"; return 1; }
3103 [ $? -eq 0 ] || { error "Unable to start OST1"; return 2; }
3105 [ $? -eq 0 ] || { error "Unable to mount client"; return 3; }
3108 local ost1mnt=$(facet_mntpt ost1)
3109 local ost1node=$(facet_active_host ost1)
3110 local ost1tmp=$TMP/conf52
3114 [ $? -eq 0 ] || { error "Unable to create tdir"; return 4; }
3115 touch $TMP/modified_first
3116 [ $? -eq 0 ] || { error "Unable to create temporary file"; return 5; }
3117 local mtime=$(stat -c %Y $TMP/modified_first)
3118 do_node $ost1node "mkdir -p $ost1tmp && touch -m -d @$mtime $ost1tmp/modified_first"
3120 [ $? -eq 0 ] || { error "Unable to create temporary file"; return 6; }
3123 $LFS setstripe -c -1 -S 1M $DIR/$tdir
3124 [ $? -eq 0 ] || { error "lfs setstripe failed"; return 7; }
3126 for (( i=0; i < nrfiles; i++ )); do
3127 multiop $DIR/$tdir/$tfile-$i Ow1048576w1048576w524288c
3128 [ $? -eq 0 ] || { error "multiop failed"; return 8; }
3134 echo backup files to $TMP/files
3135 local files=$(find $DIR/$tdir -type f -newer $TMP/modified_first)
3136 copy_files_xattrs `hostname` $TMP/files $TMP/file_xattrs $files
3137 [ $? -eq 0 ] || { error "Unable to copy files"; return 9; }
3139 umount_client $MOUNT
3140 [ $? -eq 0 ] || { error "Unable to umount client"; return 10; }
3142 [ $? -eq 0 ] || { error "Unable to stop ost1"; return 11; }
3144 echo mount ost1 as ldiskfs
3145 do_node $ost1node mkdir -p $ost1mnt
3146 [ $? -eq 0 ] || { error "Unable to create $ost1mnt"; return 23; }
3147 if ! do_node $ost1node test -b $ost1_dev; then
3150 do_node $ost1node mount -t $(facet_fstype ost1) $loop $ost1_dev \
3152 [ $? -eq 0 ] || { error "Unable to mount ost1 as ldiskfs"; return 12; }
3155 echo backup objects to $ost1tmp/objects
3156 local objects=$(do_node $ost1node 'find '$ost1mnt'/O/[0-9]* -type f'\
3157 '-size +0 -newer '$ost1tmp'/modified_first -regex ".*\/[0-9]+"')
3158 copy_files_xattrs $ost1node $ost1tmp/objects $ost1tmp/object_xattrs \
3160 [ $? -eq 0 ] || { error "Unable to copy objects"; return 13; }
3162 # move objects to lost+found
3163 do_node $ost1node 'mv '$objects' '${ost1mnt}'/lost+found'
3164 [ $? -eq 0 ] || { error "Unable to move objects"; return 14; }
3167 do_node $ost1node "ll_recover_lost_found_objs -d $ost1mnt/lost+found"
3168 [ $? -eq 0 ] || { error "ll_recover_lost_found_objs failed"; return 15; }
3170 # compare restored objects against saved ones
3171 diff_files_xattrs $ost1node $ost1tmp/objects $ost1tmp/object_xattrs $objects
3172 [ $? -eq 0 ] || { error "Unable to diff objects"; return 16; }
3174 do_node $ost1node "umount $ost1mnt"
3175 [ $? -eq 0 ] || { error "Unable to umount ost1 as ldiskfs"; return 17; }
3178 [ $? -eq 0 ] || { error "Unable to start ost1"; return 18; }
3180 [ $? -eq 0 ] || { error "Unable to mount client"; return 19; }
3183 diff_files_xattrs `hostname` $TMP/files $TMP/file_xattrs $files
3184 [ $? -eq 0 ] || { error "Unable to diff files"; return 20; }
3186 rm -rf $TMP/files $TMP/file_xattrs
3187 [ $? -eq 0 ] || { error "Unable to delete temporary files"; return 21; }
3188 do_node $ost1node "rm -rf $ost1tmp"
3189 [ $? -eq 0 ] || { error "Unable to delete temporary files"; return 22; }
3192 run_test 52 "check recovering objects from lost+found"
3194 # Checks threads_min/max/started for some service
3196 # Arguments: service name (OST or MDT), facet (e.g., ost1, $SINGLEMDS), and a
3197 # parameter pattern prefix like 'ost.*.ost'.
3210 local msg="Insane $modname thread counts"
3211 local ncpts=$(check_cpt_number $facet)
3215 check_mount || return 41
3217 # We need to expand $parampat, but it may match multiple parameters, so
3218 # we'll pick the first one
3219 if ! paramp=$(do_facet $facet "lctl get_param -N ${parampat}.threads_min"|head -1); then
3220 error "Couldn't expand ${parampat}.threads_min parameter name"
3224 # Remove the .threads_min part
3225 paramp=${paramp%.threads_min}
3227 # Check for sanity in defaults
3228 tmin=$(do_facet $facet "lctl get_param -n ${paramp}.threads_min" || echo 0)
3229 tmax=$(do_facet $facet "lctl get_param -n ${paramp}.threads_max" || echo 0)
3230 tstarted=$(do_facet $facet "lctl get_param -n ${paramp}.threads_started" || echo 0)
3231 lassert 23 "$msg (PDSH problems?)" '(($tstarted && $tmin && $tmax))' || return $?
3232 lassert 24 "$msg" '(($tstarted >= $tmin && $tstarted <= $tmax ))' || return $?
3233 nthrs=$(expr $tmax - $tmin)
3234 if [ $nthrs -lt $ncpts ]; then
3240 [ $tmin -eq $tmax -a $tmin -eq $tstarted ] &&
3241 skip_env "module parameter forced $facet thread count" &&
3242 tmin=3 && tmax=$((3 * tmax))
3244 # Check that we can change min/max
3245 do_facet $facet "lctl set_param ${paramp}.threads_min=$((tmin + nthrs))"
3246 do_facet $facet "lctl set_param ${paramp}.threads_max=$((tmax - nthrs))"
3247 tmin2=$(do_facet $facet "lctl get_param -n ${paramp}.threads_min" || echo 0)
3248 tmax2=$(do_facet $facet "lctl get_param -n ${paramp}.threads_max" || echo 0)
3249 lassert 25 "$msg" '(($tmin2 == ($tmin + $nthrs) && $tmax2 == ($tmax - $nthrs)))' || return $?
3251 # Check that we can set min/max to the same value
3252 tmin=$(do_facet $facet "lctl get_param -n ${paramp}.threads_min" || echo 0)
3253 do_facet $facet "lctl set_param ${paramp}.threads_max=$tmin"
3254 tmin2=$(do_facet $facet "lctl get_param -n ${paramp}.threads_min" || echo 0)
3255 tmax2=$(do_facet $facet "lctl get_param -n ${paramp}.threads_max" || echo 0)
3256 lassert 26 "$msg" '(($tmin2 == $tmin && $tmax2 == $tmin))' || return $?
3258 # Check that we can't set max < min
3259 do_facet $facet "lctl set_param ${paramp}.threads_max=$((tmin - 1))"
3260 tmin2=$(do_facet $facet "lctl get_param -n ${paramp}.threads_min" || echo 0)
3261 tmax2=$(do_facet $facet "lctl get_param -n ${paramp}.threads_max" || echo 0)
3262 lassert 27 "$msg" '(($tmin2 <= $tmax2))' || return $?
3264 # We need to ensure that we get the module options desired; to do this
3265 # we set LOAD_MODULES_REMOTE=true and we call setmodopts below.
3266 LOAD_MODULES_REMOTE=true
3269 local newvalue="${opts}=$(expr $basethr \* $ncpts)"
3270 setmodopts -a $modname "$newvalue" oldvalue
3274 check_mount || return 41
3276 # Restore previous setting of MODOPTS_*
3277 setmodopts $modname "$oldvalue"
3279 # Check that $opts took
3280 tmin=$(do_facet $facet "lctl get_param -n ${paramp}.threads_min")
3281 tmax=$(do_facet $facet "lctl get_param -n ${paramp}.threads_max")
3282 tstarted=$(do_facet $facet "lctl get_param -n ${paramp}.threads_started")
3283 lassert 28 "$msg" '(($tstarted >= $tmin && $tstarted <= $tmax ))' || return $?
3292 thread_sanity OST ost1 'ost.*.ost' 'oss_num_threads' '16'
3295 run_test 53a "check OSS thread count params"
3299 local mds=$(do_facet $SINGLEMDS "lctl get_param -N mds.*.*.threads_max \
3301 if [ -z "$mds" ]; then
3302 #running this on an old MDT
3303 thread_sanity MDT $SINGLEMDS 'mdt.*.*.' 'mdt_num_threads' 16
3305 thread_sanity MDT $SINGLEMDS 'mds.*.*.' 'mds_num_threads' 16
3309 run_test 53b "check MDS thread count params"
3312 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
3313 skip "Only applicable to ldiskfs-based MDTs"
3317 do_rpc_nodes $(facet_host ost1) run_llverdev $(ostdevname 1) -p
3318 [ $? -eq 0 ] || error "llverdev failed!"
3321 run_test 54a "test llverdev and partial verify of device"
3324 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
3325 skip "Only applicable to ldiskfs-based MDTs"
3330 run_llverfs $MOUNT -p
3331 [ $? -eq 0 ] || error "llverfs failed!"
3334 run_test 54b "test llverfs and partial verify of filesystem"
3338 local max_ost_index=$1
3339 echo -n $(((max_ost_index + 1) * 8))
3343 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
3344 skip "Only applicable to ldiskfs-based MDTs"
3348 local mdsdev=$(mdsdevname 1)
3349 local mdsvdev=$(mdsvdevname 1)
3353 add mds1 $(mkfs_opts mds1 ${mdsdev}) --reformat $mdsdev \
3355 add ost1 $(mkfs_opts ost1 $(ostdevname 1)) --index=$i \
3356 --reformat $(ostdevname 1) $(ostvdevname 1)
3362 echo checking size of lov_objid for ost index $i
3363 LOV_OBJID_SIZE=$(do_facet mds1 "$DEBUGFS -R 'stat lov_objid' $mdsdev 2>/dev/null" | grep ^User | awk '{print $6}')
3364 if [ "$LOV_OBJID_SIZE" != $(lov_objid_size $i) ]; then
3365 error "lov_objid size has to be $(lov_objid_size $i), not $LOV_OBJID_SIZE"
3367 echo ok, lov_objid size is correct: $LOV_OBJID_SIZE
3374 run_test 55 "check lov_objid size"
3377 local mds_journal_size_orig=$MDSJOURNALSIZE
3381 for num in $(seq 1 $MDSCOUNT); do
3382 add mds${num} $(mkfs_opts mds${num} $(mdsdevname $num)) \
3383 --reformat $(mdsdevname $num) $(mdsvdevname $num)
3385 add ost1 $(mkfs_opts ost1 $(ostdevname 1)) --index=1000 --reformat \
3386 $(ostdevname 1) $(ostvdevname 1)
3387 add ost2 $(mkfs_opts ost2 $(ostdevname 2)) --index=10000 --reformat \
3388 $(ostdevname 2) $(ostvdevname 2)
3392 start_ost2 || error "Unable to start second ost"
3393 mount_client $MOUNT || error "Unable to mount client"
3397 MDSJOURNALSIZE=$mds_journal_size_orig
3400 run_test 56 "check big indexes"
3402 test_57a() { # bug 22656
3403 local NID=$(do_facet ost1 "$LCTL get_param nis" | tail -1 | awk '{print $1}')
3404 writeconf_or_reformat
3405 [ $(facet_fstype ost1) == zfs ] && import_zpool ost1
3406 do_facet ost1 "$TUNEFS --failnode=$NID `ostdevname 1`" || error "tunefs failed"
3408 start_ost && error "OST registration from failnode should fail"
3411 run_test 57a "initial registration from failnode should fail (should return errs)"
3414 local NID=$(do_facet ost1 "$LCTL get_param nis" | tail -1 | awk '{print $1}')
3415 writeconf_or_reformat
3416 [ $(facet_fstype ost1) == zfs ] && import_zpool ost1
3417 do_facet ost1 "$TUNEFS --servicenode=$NID `ostdevname 1`" || error "tunefs failed"
3419 start_ost || error "OST registration from servicenode should not fail"
3422 run_test 57b "initial registration from servicenode should not fail"
3425 do_facet mgs $LCTL get_param mgs.MGS.live.$FSNAME | grep OST | wc -l
3428 test_58() { # bug 22658
3429 if [ $(facet_fstype mds) != ldiskfs ]; then
3430 skip "Only applicable to ldiskfs-based MDTs"
3435 createmany -o $DIR/$tdir/$tfile-%d 100
3436 # make sure that OSTs do not cancel llog cookies before we unmount the MDS
3437 #define OBD_FAIL_OBD_LOG_CANCEL_NET 0x601
3438 do_facet $SINGLEMDS "lctl set_param fail_loc=0x601"
3439 unlinkmany $DIR/$tdir/$tfile-%d 100
3442 local MNTDIR=$(facet_mntpt $SINGLEMDS)
3443 local devname=$(mdsdevname ${SINGLEMDS//mds/})
3445 if ! do_facet $SINGLEMDS "test -b $devname"; then
3449 # remove all files from the OBJECTS dir
3450 do_facet $SINGLEMDS "mount -t ldiskfs $opts $devname $MNTDIR"
3451 do_facet $SINGLEMDS "find $MNTDIR/O/1/d* -type f -delete"
3452 do_facet $SINGLEMDS "umount -d $MNTDIR"
3453 # restart MDS with missing llog files
3455 do_facet mds "lctl set_param fail_loc=0"
3458 run_test 58 "missing llog files must not prevent MDT from mounting"
3461 start_mgsmds >> /dev/null
3462 local C1=$(count_osts)
3463 if [ $C1 -eq 0 ]; then
3464 start_ost >> /dev/null
3468 echo "original ost count: $C1 (expect > 0)"
3469 [ $C1 -gt 0 ] || error "No OSTs in $FSNAME log"
3470 start_mgsmds -o writeconf >> /dev/null || error "MDT start failed"
3471 local C2=$(count_osts)
3472 echo "after mdt writeconf count: $C2 (expect 0)"
3473 [ $C2 -gt 0 ] && error "MDT writeconf should erase OST logs"
3474 echo "OST start without writeconf should fail:"
3475 start_ost >> /dev/null && error "OST start without writeconf didn't fail"
3476 echo "OST start with writeconf should succeed:"
3477 start_ost -o writeconf >> /dev/null || error "OST1 start failed"
3478 local C3=$(count_osts)
3479 echo "after ost writeconf count: $C3 (expect 1)"
3480 [ $C3 -eq 1 ] || error "new OST writeconf should add:"
3481 start_ost2 -o writeconf >> /dev/null || error "OST2 start failed"
3482 local C4=$(count_osts)
3483 echo "after ost2 writeconf count: $C4 (expect 2)"
3484 [ $C4 -eq 2 ] || error "OST2 writeconf should add log"
3485 stop_ost2 >> /dev/null
3486 cleanup_nocli >> /dev/null
3487 #writeconf to remove all ost2 traces for subsequent tests
3488 writeconf_or_reformat
3490 run_test 59 "writeconf mount option"
3492 test_60() { # LU-471
3495 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
3496 skip "Only applicable to ldiskfs-based MDTs"
3500 for num in $(seq $MDSCOUNT); do
3501 add mds${num} $(mkfs_opts mds${num} $(mdsdevname $num)) \
3502 --mkfsoptions='\" -E stride=64 -O ^uninit_bg\"' \
3503 --reformat $(mdsdevname $num) $(mdsvdevname $num) ||
3507 dump=$(do_facet $SINGLEMDS dumpe2fs $(mdsdevname 1))
3509 [ $rc -eq 0 ] || error "dumpe2fs $(mdsdevname 1) failed"
3511 # MDT default has dirdata feature
3512 echo $dump | grep dirdata > /dev/null || error "dirdata is not set"
3513 # we disable uninit_bg feature
3514 echo $dump | grep uninit_bg > /dev/null && error "uninit_bg is set"
3515 # we set stride extended options
3516 echo $dump | grep stride > /dev/null || error "stride is not set"
3519 run_test 60 "check mkfs.lustre --mkfsoptions -E -O options setting"
3522 local reformat=false
3524 [ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.1.53) ] ||
3525 { skip "Need MDS version at least 2.1.53"; return 0; }
3527 if [ $(facet_fstype $SINGLEMDS) == ldiskfs ] &&
3528 ! large_xattr_enabled; then
3530 LDISKFS_MKFS_OPTS+=" -O large_xattr"
3532 for num in $(seq $MDSCOUNT); do
3533 add mds${num} $(mkfs_opts mds$num $(mdsdevname $num)) \
3534 --reformat $(mdsdevname $num) $(mdsvdevname $num) ||
3535 error "add mds $num failed"
3539 setup_noconfig || error "setting up the filesystem failed"
3540 client_up || error "starting client failed"
3542 local file=$DIR/$tfile
3545 local large_value="$(generate_string $(max_xattr_size))"
3546 local small_value="bar"
3548 local name="trusted.big"
3549 log "save large xattr $name on $file"
3550 setfattr -n $name -v $large_value $file ||
3551 error "saving $name on $file failed"
3553 local new_value=$(get_xattr_value $name $file)
3554 [[ "$new_value" != "$large_value" ]] &&
3555 error "$name different after saving"
3557 log "shrink value of $name on $file"
3558 setfattr -n $name -v $small_value $file ||
3559 error "shrinking value of $name on $file failed"
3561 new_value=$(get_xattr_value $name $file)
3562 [[ "$new_value" != "$small_value" ]] &&
3563 error "$name different after shrinking"
3565 log "grow value of $name on $file"
3566 setfattr -n $name -v $large_value $file ||
3567 error "growing value of $name on $file failed"
3569 new_value=$(get_xattr_value $name $file)
3570 [[ "$new_value" != "$large_value" ]] &&
3571 error "$name different after growing"
3573 log "check value of $name on $file after remounting MDS"
3575 new_value=$(get_xattr_value $name $file)
3576 [[ "$new_value" != "$large_value" ]] &&
3577 error "$name different after remounting MDS"
3579 log "remove large xattr $name from $file"
3580 setfattr -x $name $file || error "removing $name from $file failed"
3585 LDISKFS_MKFS_OPTS=${LDISKFS_MKFS_OPTS% -O large_xattr}
3589 run_test 61 "large xattr"
3592 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
3593 skip "Only applicable to ldiskfs-based MDTs"
3598 local mdsdev=$(mdsdevname 1)
3599 local ostdev=$(ostdevname 1)
3601 [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.2.51) ]] ||
3602 { skip "Need MDS version at least 2.2.51"; return 0; }
3604 echo "disable journal for mds"
3605 do_facet mds tune2fs -O ^has_journal $mdsdev || error "tune2fs failed"
3606 start_mds && error "MDT start should fail"
3607 echo "disable journal for ost"
3608 do_facet ost1 tune2fs -O ^has_journal $ostdev || error "tune2fs failed"
3609 start_ost && error "OST start should fail"
3610 cleanup || return $?
3613 run_test 62 "start with disabled journal"
3616 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
3617 skip "Only applicable to ldiskfs-based MDTs"
3621 local inode_slab=$(do_facet $SINGLEMDS \
3622 "awk '/ldiskfs_inode_cache/ { print \\\$5 }' /proc/slabinfo")
3623 if [ -z "$inode_slab" ]; then
3624 skip "ldiskfs module has not been loaded"
3628 echo "$inode_slab ldisk inodes per page"
3629 [ "$inode_slab" -ge "3" ] ||
3630 error "ldisk inode size is too big, $inode_slab objs per page"
3633 run_test 63 "Verify each page can at least hold 3 ldisk inodes"
3638 start_ost2 || error "Unable to start second ost"
3639 mount_client $MOUNT || error "Unable to mount client"
3640 stop_ost2 || error "Unable to stop second ost"
3642 $LFS df --lazy || error "lfs df failed"
3643 cleanup || return $?
3644 #writeconf to remove all ost2 traces for subsequent tests
3645 writeconf_or_reformat
3647 run_test 64 "check lfs df --lazy "
3649 test_65() { # LU-2237
3650 # Currently, the test is only valid for ldiskfs backend
3651 [ "$(facet_fstype $SINGLEMDS)" != "ldiskfs" ] &&
3652 skip "non-ldiskfs backend" && return
3654 local devname=$(mdsdevname ${SINGLEMDS//mds/})
3655 local brpt=$(facet_mntpt brpt)
3658 if ! do_facet $SINGLEMDS "test -b $devname"; then
3663 local obj=$(do_facet $SINGLEMDS \
3664 "$DEBUGFS -c -R \\\"stat last_rcvd\\\" $devname" |
3666 if [ -z "$obj" ]; then
3667 # The MDT may be just re-formatted, mount the MDT for the
3668 # first time to guarantee the "last_rcvd" file is there.
3669 start_mds || error "fail to mount the MDS for the first time"
3673 # remove the "last_rcvd" file
3674 do_facet $SINGLEMDS "mkdir -p $brpt"
3675 do_facet $SINGLEMDS \
3676 "mount -t $(facet_fstype $SINGLEMDS) $opts $devname $brpt"
3677 do_facet $SINGLEMDS "rm -f ${brpt}/last_rcvd"
3678 do_facet $SINGLEMDS "umount -d $brpt"
3680 # restart MDS, the "last_rcvd" file should be recreated.
3681 start_mds || error "fail to restart the MDS"
3683 obj=$(do_facet $SINGLEMDS \
3684 "$DEBUGFS -c -R \\\"stat last_rcvd\\\" $devname" | grep Inode)
3685 [ -n "$obj" ] || error "fail to re-create the last_rcvd"
3687 run_test 65 "re-create the lost last_rcvd file when server mount"
3690 [[ $(lustre_version_code mgs) -ge $(version_code 2.3.59) ]] ||
3691 { skip "Need MGS version at least 2.3.59"; return 0; }
3694 local OST1_NID=$(do_facet ost1 $LCTL list_nids | head -1)
3695 local MDS_NID=$(do_facet $SINGLEMDS $LCTL list_nids | head -1)
3697 echo "replace_nids should fail if MDS, OSTs and clients are UP"
3698 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 $OST1_NID &&
3699 error "replace_nids fail"
3701 umount_client $MOUNT || error "unmounting client failed"
3702 echo "replace_nids should fail if MDS and OSTs are UP"
3703 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 $OST1_NID &&
3704 error "replace_nids fail"
3707 echo "replace_nids should fail if MDS is UP"
3708 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 $OST1_NID &&
3709 error "replace_nids fail"
3711 stop_mds || error "stopping mds failed"
3713 if combined_mgs_mds; then
3714 start_mdt 1 "-o nosvc" ||
3715 error "starting mds with nosvc option failed"
3718 echo "command should accept two parameters"
3719 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 &&
3720 error "command should accept two params"
3722 echo "correct device name should be passed"
3723 do_facet mgs $LCTL replace_nids $FSNAME-WRONG0000 $OST1_NID &&
3724 error "wrong devname"
3726 echo "wrong nids list should not destroy the system"
3727 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 "wrong nids list" &&
3730 echo "replace OST nid"
3731 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 $OST1_NID ||
3732 error "replace nids failed"
3734 echo "command should accept two parameters"
3735 do_facet mgs $LCTL replace_nids $FSNAME-MDT0000 &&
3736 error "command should accept two params"
3738 echo "wrong nids list should not destroy the system"
3739 do_facet mgs $LCTL replace_nids $FSNAME-MDT0000 "wrong nids list" &&
3742 echo "replace MDS nid"
3743 do_facet mgs $LCTL replace_nids $FSNAME-MDT0000 $MDS_NID ||
3744 error "replace nids failed"
3746 if ! combined_mgs_mds ; then
3753 check_mount || error "error after nid replace"
3754 cleanup || error "cleanup failed"
3757 run_test 66 "replace nids"
3759 test_67() { #LU-2950
3760 local legacy="$TMP/legacy_lnet_config"
3761 local new="$TMP/new_routes_test"
3762 local out="$TMP/config_out_file"
3763 local verify="$TMP/conv_verify"
3764 local verify_conf="$TMP/conf_verify"
3766 # Create the legacy file that will be run through the
3767 # lustre_routes_conversion script
3768 cat <<- LEGACY_LNET_CONFIG > $legacy
3769 tcp1 23 192.168.213.1@tcp:1; tcp5 34 193.30.4.3@tcp:4;
3770 tcp2 54 10.1.3.2@tcp;
3771 tcp3 10.3.4.3@tcp:3;
3775 # Create the verification file to verify the output of
3776 # lustre_routes_conversion script against.
3777 cat <<- VERIFY_LNET_CONFIG > $verify
3778 tcp1: { gateway: 192.168.213.1@tcp, hop: 23, priority: 1 }
3779 tcp5: { gateway: 193.30.4.3@tcp, hop: 34, priority: 4 }
3780 tcp2: { gateway: 10.1.3.2@tcp, hop: 54 }
3781 tcp3: { gateway: 10.3.4.3@tcp, priority: 3 }
3782 tcp4: { gateway: 10.3.3.4@tcp }
3785 # Create the verification file to verify the output of
3786 # lustre_routes_config script against
3787 cat <<- VERIFY_LNET_CONFIG > $verify_conf
3788 lctl --net tcp1 add_route 192.168.213.1@tcp 23 1
3789 lctl --net tcp5 add_route 193.30.4.3@tcp 34 4
3790 lctl --net tcp2 add_route 10.1.3.2@tcp 54 4
3791 lctl --net tcp3 add_route 10.3.4.3@tcp 1 3
3792 lctl --net tcp4 add_route 10.3.3.4@tcp 1 3
3795 lustre_routes_conversion $legacy $new > /dev/null
3796 if [ -f $new ]; then
3797 # verify the conversion output
3798 cmp -s $new $verify > /dev/null
3799 if [ $? -eq 1 ]; then
3800 error "routes conversion failed"
3803 lustre_routes_config --dry-run --verbose $new > $out
3804 # check that the script succeeded
3805 cmp -s $out $verify_conf > /dev/null
3806 if [ $? -eq 1 ]; then
3807 error "routes config failed"
3810 error "routes conversion test failed"
3812 # remove generated files
3813 rm -f $new $legacy $verify $verify_conf $out
3815 run_test 67 "test routes conversion and configuration"
3823 [ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.4.53) ] ||
3824 { skip "Need MDS version at least 2.4.53"; return 0; }
3826 umount_client $MOUNT || error "umount client failed"
3828 start_mdt 1 || error "MDT start failed"
3831 # START-END - the sequences we'll be reserving
3832 START=$(do_facet $SINGLEMDS \
3833 lctl get_param -n seq.ctl*.space | awk -F'[[ ]' '{print $2}')
3834 END=$((START + (1 << 30)))
3835 do_facet $SINGLEMDS \
3836 lctl set_param seq.ctl*.fldb="[$START-$END\):0:mdt"
3838 # reset the sequences MDT0000 has already assigned
3839 do_facet $SINGLEMDS \
3840 lctl set_param seq.srv*MDT0000.space=clear
3842 # remount to let the client allocate new sequence
3843 mount_client $MOUNT || error "mount client failed"
3846 do_facet $SINGLEMDS \
3847 lctl get_param seq.srv*MDT0000.space
3848 $LFS path2fid $DIR/$tfile
3850 local old_ifs="$IFS"
3852 fid=($($LFS path2fid $DIR/$tfile))
3856 if [[ $seq < $END ]]; then
3857 error "used reserved sequence $seq?"
3859 cleanup || return $?
3861 run_test 68 "be able to reserve specific sequences in FLDB"
3864 local server_version=$(lustre_version_code $SINGLEMDS)
3866 [[ $server_version -lt $(version_code 2.4.2) ]] &&
3867 skip "Need MDS version at least 2.4.2" && return
3869 [[ $server_version -ge $(version_code 2.4.50) ]] &&
3870 [[ $server_version -lt $(version_code 2.5.0) ]] &&
3871 skip "Need MDS version at least 2.5.0" && return
3875 # use OST0000 since it probably has the most creations
3876 local OSTNAME=$(ostname_from_index 0)
3877 local mdtosc_proc1=$(get_mdtosc_proc_path mds1 $OSTNAME)
3878 local last_id=$(do_facet mds1 lctl get_param -n \
3879 osc.$mdtosc_proc1.prealloc_last_id)
3881 # Want to have OST LAST_ID over 1.5 * OST_MAX_PRECREATE to
3882 # verify that the LAST_ID recovery is working properly. If
3883 # not, then the OST will refuse to allow the MDS connect
3884 # because the LAST_ID value is too different from the MDS
3885 #define OST_MAX_PRECREATE=20000
3886 local num_create=$((20000 * 5))
3889 $LFS setstripe -i 0 $DIR/$tdir
3890 createmany -o $DIR/$tdir/$tfile- $num_create
3891 # delete all of the files with objects on OST0 so the
3892 # filesystem is not inconsistent later on
3893 $LFS find $MOUNT --ost 0 | xargs rm
3895 stop_ost || error "OST0 stop failure"
3896 add ost1 $(mkfs_opts ost1 $(ostdevname 1)) --reformat --replace \
3897 $(ostdevname 1) $(ostvdevname 1) ||
3898 error "reformat and replace $ostdev failed"
3899 start_ost || error "OST0 restart failure"
3900 wait_osc_import_state mds ost FULL
3902 touch $DIR/$tdir/$tfile-last || error "create file after reformat"
3903 local idx=$($LFS getstripe -i $DIR/$tdir/$tfile-last)
3904 [ $idx -ne 0 ] && error "$DIR/$tdir/$tfile-last on $idx not 0" || true
3908 run_test 69 "replace an OST with the same index"
3911 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
3916 start_mdt 1 || error "MDT0 start fail"
3918 start_ost || error "OST0 start fail"
3920 start_mdt 2 || error "MDT1 start fail"
3922 mount_client $MOUNT || error "mount client fails"
3924 mkdir -p $DIR/$tdir || error "create dir fail"
3926 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
3927 error "create remote dir fail"
3929 rm -rf $DIR/$tdir || error "delete dir fail"
3930 cleanup || return $?
3932 run_test 70a "start MDT0, then OST, then MDT1"
3935 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
3938 start_ost || error "OST0 start fail"
3940 start_mdt 1 || error "MDT0 start fail"
3941 start_mdt 2 || error "MDT1 start fail"
3943 mount_client $MOUNT || error "mount client fails"
3945 mkdir -p $DIR/$tdir || error "create dir fail"
3947 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
3948 error "create remote dir fail"
3950 rm -rf $DIR/$tdir || error "delete dir fail"
3952 cleanup || return $?
3954 run_test 70b "start OST, MDT1, MDT0"
3957 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
3960 start_mdt 1 || error "MDT0 start fail"
3961 start_mdt 2 || error "MDT1 start fail"
3962 start_ost || error "OST0 start fail"
3964 mount_client $MOUNT || error "mount client fails"
3965 stop_mdt 1 || error "MDT1 start fail"
3967 local mdc_for_mdt1=$($LCTL dl | grep MDT0000-mdc | awk '{print $4}')
3968 echo "deactivate $mdc_for_mdt1"
3969 $LCTL --device $mdc_for_mdt1 deactivate || return 1
3971 mkdir -p $DIR/$tdir && error "mkdir succeed"
3973 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir &&
3974 error "create remote dir succeed"
3976 cleanup || return $?
3978 run_test 70c "stop MDT0, mkdir fail, create remote dir fail"
3981 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
3984 start_mdt 1 || error "MDT0 start fail"
3985 start_mdt 2 || error "MDT1 start fail"
3986 start_ost || error "OST0 start fail"
3988 mount_client $MOUNT || error "mount client fails"
3990 stop_mdt 2 || error "MDT1 start fail"
3992 local mdc_for_mdt2=$($LCTL dl | grep MDT0001-mdc |
3994 echo "deactivate $mdc_for_mdt2"
3995 $LCTL --device $mdc_for_mdt2 deactivate ||
3996 error "set $mdc_for_mdt2 deactivate failed"
3998 mkdir -p $DIR/$tdir || error "mkdir fail"
3999 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir &&
4000 error "create remote dir succeed"
4002 rm -rf $DIR/$tdir || error "delete dir fail"
4004 cleanup || return $?
4006 run_test 70d "stop MDT1, mkdir succeed, create remote dir fail"
4009 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
4010 if combined_mgs_mds; then
4011 skip "needs separate MGS/MDT" && return
4015 start_mdt 1 || error "MDT0 start fail"
4016 start_ost || error "OST0 start fail"
4017 start_mdt 2 || error "MDT1 start fail"
4018 start_ost2 || error "OST1 start fail"
4020 mount_client $MOUNT || error "mount client fails"
4022 mkdir -p $DIR/$tdir || error "mkdir fail"
4023 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
4024 error "create remote dir succeed"
4026 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
4027 rm -rf $DIR/$tdir || error "delete dir fail"
4029 umount_client $MOUNT
4030 stop_mdt 1 || error "MDT0 stop fail"
4031 stop_mdt 2 || error "MDT1 stop fail"
4032 stop_ost || error "OST0 stop fail"
4033 stop_ost2 || error "OST1 stop fail"
4035 run_test 71a "start MDT0 OST0, MDT1, OST1"
4038 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
4039 if combined_mgs_mds; then
4040 skip "needs separate MGS/MDT" && return
4044 start_mdt 2 || error "MDT1 start fail"
4045 start_ost || error "OST0 start fail"
4046 start_mdt 1 || error "MDT0 start fail"
4047 start_ost2 || error "OST1 start fail"
4049 mount_client $MOUNT || error "mount client fails"
4051 mkdir -p $DIR/$tdir || error "mkdir fail"
4052 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
4053 error "create remote dir succeed"
4055 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
4056 rm -rf $DIR/$tdir || error "delete dir fail"
4058 umount_client $MOUNT
4059 stop_mdt 1 || error "MDT0 stop fail"
4060 stop_mdt 2 || error "MDT1 stop fail"
4061 stop_ost || error "OST0 stop fail"
4062 stop_ost2 || error "OST1 stop fail"
4064 run_test 71b "start MDT1, OST0, MDT0, OST1"
4067 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
4068 if combined_mgs_mds; then
4069 skip "needs separate MGS/MDT" && return
4073 start_ost || error "OST0 start fail"
4074 start_ost2 || error "OST1 start fail"
4075 start_mdt 2 || error "MDT1 start fail"
4076 start_mdt 1 || error "MDT0 start fail"
4078 mount_client $MOUNT || error "mount client fails"
4080 mkdir -p $DIR/$tdir || error "mkdir fail"
4081 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
4082 error "create remote dir succeed"
4084 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
4085 rm -rf $DIR/$tdir || error "delete dir fail"
4087 umount_client $MOUNT
4088 stop_mdt 1 || error "MDT0 stop fail"
4089 stop_mdt 2 || error "MDT1 stop fail"
4090 stop_ost || error "OST0 stop fail"
4091 stop_ost2 || error "OST1 stop fail"
4094 run_test 71c "start OST0, OST1, MDT1, MDT0"
4097 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
4098 if combined_mgs_mds; then
4099 skip "needs separate MGS/MDT" && return
4103 start_ost || error "OST0 start fail"
4104 start_mdt 2 || error "MDT0 start fail"
4105 start_mdt 1 || error "MDT0 start fail"
4106 start_ost2 || error "OST1 start fail"
4108 mount_client $MOUNT || error "mount client fails"
4110 mkdir -p $DIR/$tdir || error "mkdir fail"
4111 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
4112 error "create remote dir succeed"
4114 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
4115 rm -rf $DIR/$tdir || error "delete dir fail"
4117 umount_client $MOUNT
4118 stop_mdt 1 || error "MDT0 stop fail"
4119 stop_mdt 2 || error "MDT1 stop fail"
4120 stop_ost || error "OST0 stop fail"
4121 stop_ost2 || error "OST1 stop fail"
4124 run_test 71d "start OST0, MDT1, MDT0, OST1"
4127 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
4128 if combined_mgs_mds; then
4129 skip "needs separate MGS/MDT" && return
4133 start_ost || error "OST0 start fail"
4134 start_mdt 2 || error "MDT1 start fail"
4135 start_ost2 || error "OST1 start fail"
4136 start_mdt 1 || error "MDT0 start fail"
4138 mount_client $MOUNT || error "mount client fails"
4140 mkdir -p $DIR/$tdir || error "mkdir fail"
4141 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
4142 error "create remote dir succeed"
4144 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
4145 rm -rf $DIR/$tdir || error "delete dir fail"
4147 umount_client $MOUNT
4148 stop_mdt 1 || error "MDT0 stop fail"
4149 stop_mdt 2 || error "MDT1 stop fail"
4150 stop_ost || error "OST0 stop fail"
4151 stop_ost2 || error "OST1 stop fail"
4154 run_test 71e "start OST0, MDT1, OST1, MDT0"
4156 test_72() { #LU-2634
4157 local mdsdev=$(mdsdevname 1)
4158 local ostdev=$(ostdevname 1)
4159 local cmd="$E2FSCK -fnvd $mdsdev"
4162 [ "$(facet_fstype $SINGLEMDS)" != "ldiskfs" ] &&
4163 skip "ldiskfs only test" && return
4165 #tune MDT with "-O extents"
4167 for num in $(seq $MDSCOUNT); do
4168 add mds${num} $(mkfs_opts mds$num $(mdsdevname $num)) \
4169 --reformat $(mdsdevname $num) $(mdsvdevname $num) ||
4170 error "add mds $num failed"
4171 $TUNE2FS -O extents $(mdsdevname $num)
4174 add ost1 $(mkfs_opts ost1 $ostdev) --reformat $ostdev ||
4175 error "add $ostdev failed"
4176 start_mgsmds || error "start mds failed"
4177 start_ost || error "start ost failed"
4178 mount_client $MOUNT || error "mount client failed"
4180 #create some short symlinks
4182 createmany -o $DIR/$tdir/$tfile-%d $fn
4183 echo "create $fn short symlinks"
4184 for i in $(seq -w 1 $fn); do
4185 ln -s $DIR/$tdir/$tfile-$i $MOUNT/$tfile-$i
4190 umount_client $MOUNT || error "umount client failed"
4191 stop_mds || error "stop mds failed"
4192 stop_ost || error "stop ost failed"
4195 run_e2fsck $(facet_active_host $SINGLEMDS) $mdsdev "-n"
4197 run_test 72 "test fast symlink with extents flag enabled"
4199 test_73() { #LU-3006
4201 [ $(facet_fstype ost1) == zfs ] && import_zpool ost1
4202 do_facet ost1 "$TUNEFS --failnode=1.2.3.4@$NETTYPE $(ostdevname 1)" ||
4203 error "1st tunefs failed"
4204 start_mgsmds || error "start mds failed"
4205 start_ost || error "start ost failed"
4206 mount_client $MOUNT || error "mount client failed"
4207 lctl get_param -n osc.*OST0000-osc-[^M]*.import | grep failover_nids |
4208 grep 1.2.3.4@$NETTYPE || error "failover nids haven't changed"
4209 umount_client $MOUNT || error "umount client failed"
4213 run_test 73 "failnode to update from mountdata properly"
4215 test_74() { # LU-1606
4216 for TESTPROG in $LUSTRE_TESTS_API_DIR/*.c; do
4217 gcc -Wall -Werror $LUSTRE_TESTS_API_DIR/simple_test.c \
4219 -L$LUSTRE/utils -llustreapi ||
4220 error "client api broken"
4222 cleanup || return $?
4224 run_test 74 "Lustre client api program can compile and link"
4226 test_75() { # LU-2374
4227 [[ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.4.1) ]] &&
4228 skip "Need MDS version at least 2.4.1" && return
4231 local opts_mds="$(mkfs_opts mds1 $(mdsdevname 1)) \
4232 --reformat $(mdsdevname 1) $(mdsvdevname 1)"
4233 local opts_ost="$(mkfs_opts ost1 $(ostdevname 1)) \
4234 --reformat $(ostdevname 1) $(ostvdevname 1)"
4236 #check with default parameters
4237 add mds1 $opts_mds || error "add mds1 failed for default params"
4238 add ost1 $opts_ost || error "add ost1 failed for default params"
4240 opts_mds=$(echo $opts_mds | sed -e "s/--mdt//")
4241 opts_mds=$(echo $opts_mds |
4242 sed -e "s/--index=$index/--index=$index --mdt/")
4243 opts_ost=$(echo $opts_ost | sed -e "s/--ost//")
4244 opts_ost=$(echo $opts_ost |
4245 sed -e "s/--index=$index/--index=$index --ost/")
4247 add mds1 $opts_mds || error "add mds1 failed for new params"
4248 add ost1 $opts_ost || error "add ost1 failed for new params"
4251 run_test 75 "The order of --index should be irrelevant"
4254 [[ $(lustre_version_code mgs) -ge $(version_code 2.4.52) ]] ||
4255 { skip "Need MDS version at least 2.4.52" && return 0; }
4257 local MDMB_PARAM="osc.*.max_dirty_mb"
4258 echo "Change MGS params"
4259 local MAX_DIRTY_MB=$($LCTL get_param -n $MDMB_PARAM |
4261 echo "max_dirty_mb: $MAX_DIRTY_MB"
4262 local NEW_MAX_DIRTY_MB=$((MAX_DIRTY_MB + MAX_DIRTY_MB))
4263 echo "new_max_dirty_mb: $NEW_MAX_DIRTY_MB"
4264 do_facet mgs $LCTL set_param -P $MDMB_PARAM=$NEW_MAX_DIRTY_MB
4265 wait_update $HOSTNAME "lctl get_param -n $MDMB_PARAM |
4266 head -1" $NEW_MAX_DIRTY_MB
4267 MAX_DIRTY_MB=$($LCTL get_param -n $MDMB_PARAM | head -1)
4268 echo "$MAX_DIRTY_MB"
4269 [ $MAX_DIRTY_MB = $NEW_MAX_DIRTY_MB ] ||
4270 error "error while apply max_dirty_mb"
4272 echo "Check the value is stored after remount"
4275 wait_update $HOSTNAME "lctl get_param -n $MDMB_PARAM |
4276 head -1" $NEW_MAX_DIRTY_MB
4277 MAX_DIRTY_MB=$($LCTL get_param -n $MDMB_PARAM | head -1)
4278 [ $MAX_DIRTY_MB = $NEW_MAX_DIRTY_MB ] ||
4279 error "max_dirty_mb is not saved after remount"
4281 echo "Change OST params"
4282 CLIENT_PARAM="obdfilter.*.client_cache_count"
4283 local CLIENT_CACHE_COUNT
4284 CLIENT_CACHE_COUNT=$(do_facet ost1 $LCTL get_param -n $CLIENT_PARAM |
4286 echo "client_cache_count: $CLIENT_CACHE_COUNT"
4287 NEW_CLIENT_CACHE_COUNT=$((CLIENT_CACHE_COUNT+CLIENT_CACHE_COUNT))
4288 echo "new_client_cache_count: $NEW_CLIENT_CACHE_COUNT"
4289 do_facet mgs $LCTL set_param -P $CLIENT_PARAM=$NEW_CLIENT_CACHE_COUNT
4290 wait_update $(facet_host ost1) "lctl get_param -n $CLIENT_PARAM |
4291 head -1" $NEW_CLIENT_CACHE_COUNT
4292 CLIENT_CACHE_COUNT=$(do_facet ost1 $LCTL get_param -n $CLIENT_PARAM |
4294 echo "$CLIENT_CACHE_COUNT"
4295 [ $CLIENT_CACHE_COUNT = $NEW_CLIENT_CACHE_COUNT ] ||
4296 error "error while apply client_cache_count"
4298 echo "Check the value is stored after remount"
4301 wait_update $(facet_host ost1) "lctl get_param -n $CLIENT_PARAM |
4302 head -1" $NEW_CLIENT_CACHE_COUNT
4303 CLIENT_CACHE_COUNT=$(do_facet ost1 $LCTL get_param -n $CLIENT_PARAM |
4305 echo "$CLIENT_CACHE_COUNT"
4306 [ $CLIENT_CACHE_COUNT = $NEW_CLIENT_CACHE_COUNT ] ||
4307 error "client_cache_count is not saved after remount"
4310 run_test 76 "set permanent params set_param -P"
4312 test_77() { # LU-3445
4313 local server_version=$(lustre_version_code $SINGLEMDS)
4315 [[ $server_version -ge $(version_code 2.2.60) ]] &&
4316 [[ $server_version -le $(version_code 2.4.0) ]] &&
4317 skip "Need MDS version < 2.2.60 or > 2.4.0" && return
4319 if [[ -z "$fs2ost_DEV" || -z "$fs2mds_DEV" ]]; then
4320 is_blkdev $SINGLEMDS $(mdsdevname ${SINGLEMDS//mds/}) &&
4321 skip_env "mixed loopback and real device not working" && return
4324 local fs2mdsdev=$(mdsdevname 1_2)
4325 local fs2ostdev=$(ostdevname 1_2)
4326 local fs2mdsvdev=$(mdsvdevname 1_2)
4327 local fs2ostvdev=$(ostvdevname 1_2)
4328 local fsname=test1234
4330 local failnid="$(h2$NETTYPE 1.2.3.4),$(h2$NETTYPE 4.3.2.1)"
4332 add fs2mds $(mkfs_opts mds1 $fs2mdsdev) --mgs --fsname=$fsname \
4333 --reformat $fs2mdsdev $fs2mdsvdev || error "add fs2mds failed"
4334 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS && trap cleanup_fs2 EXIT INT ||
4335 error "start fs2mds failed"
4337 mgsnid=$(do_facet fs2mds $LCTL list_nids | xargs | tr ' ' ,)
4338 [[ $mgsnid = *,* ]] || mgsnid+=",$mgsnid"
4340 add fs2ost $(mkfs_opts ost1 $fs2ostdev) --mgsnode=$mgsnid \
4341 --failnode=$failnid --fsname=$fsname \
4342 --reformat $fs2ostdev $fs2ostvdev ||
4343 error "add fs2ost failed"
4344 start fs2ost $fs2ostdev $OST_MOUNT_OPTS || error "start fs2ost failed"
4347 mount -t lustre $mgsnid:/$fsname $MOUNT2 || error "mount $MOUNT2 failed"
4348 DIR=$MOUNT2 MOUNT=$MOUNT2 check_mount || error "check $MOUNT2 failed"
4351 run_test 77 "comma-separated MGS NIDs and failover node NIDs"
4354 [[ $(facet_fstype $SINGLEMDS) != ldiskfs ||
4355 $(facet_fstype ost1) != ldiskfs ]] &&
4356 skip "only applicable to ldiskfs-based MDTs and OSTs" && return
4358 # reformat the Lustre filesystem with a smaller size
4359 local saved_MDSSIZE=$MDSSIZE
4360 local saved_OSTSIZE=$OSTSIZE
4361 MDSSIZE=$((MDSSIZE - 20000))
4362 OSTSIZE=$((OSTSIZE - 20000))
4363 reformat || error "(1) reformat Lustre filesystem failed"
4364 MDSSIZE=$saved_MDSSIZE
4365 OSTSIZE=$saved_OSTSIZE
4367 # mount the Lustre filesystem
4368 setup_noconfig || error "(2) setup Lustre filesystem failed"
4371 log "create test files"
4375 mkdir -p $MOUNT/$tdir || error "(3) mkdir $MOUNT/$tdir failed"
4376 for i in $(seq $num_files); do
4377 file=$MOUNT/$tdir/$tfile-$i
4378 dd if=/dev/urandom of=$file count=1 bs=1M ||
4379 error "(4) create $file failed"
4382 # unmount the Lustre filesystem
4383 cleanup || error "(5) cleanup Lustre filesystem failed"
4385 # run e2fsck on the MDT and OST devices
4386 local mds_host=$(facet_active_host $SINGLEMDS)
4387 local ost_host=$(facet_active_host ost1)
4388 local mds_dev=$(mdsdevname ${SINGLEMDS//mds/})
4389 local ost_dev=$(ostdevname 1)
4391 run_e2fsck $mds_host $mds_dev "-y"
4392 run_e2fsck $ost_host $ost_dev "-y"
4394 # get the original block count of the MDT and OST filesystems
4395 local mds_orig_blks=$(get_block_count $SINGLEMDS $mds_dev)
4396 local ost_orig_blks=$(get_block_count ost1 $ost_dev)
4398 # expand the MDT and OST filesystems to the device size
4399 run_resize2fs $SINGLEMDS $mds_dev "" || error "expand $SINGLEMDS failed"
4400 run_resize2fs ost1 $ost_dev "" || error "expand ost1 failed"
4402 # run e2fsck on the MDT and OST devices again
4403 run_e2fsck $mds_host $mds_dev "-y"
4404 run_e2fsck $ost_host $ost_dev "-y"
4406 # mount the Lustre filesystem
4410 log "check files after expanding the MDT and OST filesystems"
4411 for i in $(seq $num_files); do
4412 file=$MOUNT/$tdir/$tfile-$i
4413 $CHECKSTAT -t file -s 1048576 $file ||
4414 error "(6) checkstat $file failed"
4418 log "create more files after expanding the MDT and OST filesystems"
4419 for i in $(seq $((num_files + 1)) $((num_files + 10))); do
4420 file=$MOUNT/$tdir/$tfile-$i
4421 dd if=/dev/urandom of=$file count=1 bs=1M ||
4422 error "(7) create $file failed"
4425 # unmount the Lustre filesystem
4426 cleanup || error "(8) cleanup Lustre filesystem failed"
4428 # run e2fsck on the MDT and OST devices
4429 run_e2fsck $mds_host $mds_dev "-y"
4430 run_e2fsck $ost_host $ost_dev "-y"
4432 # get the maximum block count of the MDT and OST filesystems
4433 local mds_max_blks=$(get_block_count $SINGLEMDS $mds_dev)
4434 local ost_max_blks=$(get_block_count ost1 $ost_dev)
4436 # get the minimum block count of the MDT and OST filesystems
4437 local mds_min_blks=$(run_resize2fs $SINGLEMDS $mds_dev "" "-P" 2>&1 |
4438 grep minimum | sed -e 's/^.*filesystem: //g')
4439 local ost_min_blks=$(run_resize2fs ost1 $ost_dev "" "-P" 2>&1 |
4440 grep minimum | sed -e 's/^.*filesystem: //g')
4442 # shrink the MDT and OST filesystems to a smaller size
4446 if [[ $mds_max_blks -gt $mds_min_blks &&
4447 $mds_max_blks -gt $mds_orig_blks ]]; then
4448 [[ $mds_orig_blks -gt $mds_min_blks ]] &&
4449 base_blks=$mds_orig_blks || base_blks=$mds_min_blks
4450 new_blks=$(( (mds_max_blks - base_blks) / 2 + base_blks ))
4451 run_resize2fs $SINGLEMDS $mds_dev $new_blks ||
4452 error "shrink $SINGLEMDS to $new_blks failed"
4456 if [[ $ost_max_blks -gt $ost_min_blks &&
4457 $ost_max_blks -gt $ost_orig_blks ]]; then
4458 [[ $ost_orig_blks -gt $ost_min_blks ]] &&
4459 base_blks=$ost_orig_blks || base_blks=$ost_min_blks
4460 new_blks=$(( (ost_max_blks - base_blks) / 2 + base_blks ))
4461 run_resize2fs ost1 $ost_dev $new_blks ||
4462 error "shrink ost1 to $new_blks failed"
4466 # check whether the MDT or OST filesystem was shrunk or not
4468 combined_mgs_mds || stop_mgs || error "(9) stop mgs failed"
4469 reformat || error "(10) reformat Lustre filesystem failed"
4473 # run e2fsck on the MDT and OST devices again
4474 run_e2fsck $mds_host $mds_dev "-y"
4475 run_e2fsck $ost_host $ost_dev "-y"
4477 # mount the Lustre filesystem again
4481 log "check files after shrinking the MDT and OST filesystems"
4482 for i in $(seq $((num_files + 10))); do
4483 file=$MOUNT/$tdir/$tfile-$i
4484 $CHECKSTAT -t file -s 1048576 $file ||
4485 error "(11) checkstat $file failed"
4488 # unmount and reformat the Lustre filesystem
4489 cleanup || error "(12) cleanup Lustre filesystem failed"
4490 combined_mgs_mds || stop_mgs || error "(13) stop mgs failed"
4491 reformat || error "(14) reformat Lustre filesystem failed"
4493 run_test 78 "run resize2fs on MDT and OST filesystems"
4495 if ! combined_mgs_mds ; then
4501 # restore the values of MDSSIZE and OSTSIZE
4502 MDSSIZE=$STORED_MDSSIZE
4503 OSTSIZE=$STORED_OSTSIZE