3 # FIXME - there is no reason to use all of these different return codes,
4 # espcially when most of them are mapped to something else anyway.
5 # The tests should use error() to describe the failure more clearly,
6 # and reduce the need to look into the tests to see what failed.
12 # bug number for skipped test: LU-2828
13 ALWAYS_EXCEPT="$CONF_SANITY_EXCEPT 59 64"
14 # UPDATE THE COMMENT ABOVE WITH BUG NUMBERS WHEN CHANGING ALWAYS_EXCEPT!
18 if [ -r /etc/SuSE-release ]
20 local vers=`grep VERSION /etc/SuSE-release | awk '{print $3}'`
21 local patchlev=`grep PATCHLEVEL /etc/SuSE-release \
23 if [ $vers -eq 11 ] && [ $patchlev -eq 2 ]
31 if is_sles11; then # LU-2181
32 ALWAYS_EXCEPT="$ALWAYS_EXCEPT 23a 34b"
35 if [ "$FAILURE_MODE" = "HARD" ]; then
36 CONFIG_EXCEPTIONS="24a " && \
37 echo "Except the tests: $CONFIG_EXCEPTIONS for FAILURE_MODE=$FAILURE_MODE, bug 23573" && \
38 ALWAYS_EXCEPT="$ALWAYS_EXCEPT $CONFIG_EXCEPTIONS"
41 # bug number for skipped test:
42 # a tool to create lustre filesystem images
43 ALWAYS_EXCEPT="32newtarball $ALWAYS_EXCEPT"
46 PATH=$PWD/$SRCDIR:$SRCDIR:$SRCDIR/../utils:$PATH
48 PTLDEBUG=${PTLDEBUG:--1}
50 LUSTRE=${LUSTRE:-`dirname $0`/..}
51 RLUSTRE=${RLUSTRE:-$LUSTRE}
52 export MULTIOP=${MULTIOP:-multiop}
54 . $LUSTRE/tests/test-framework.sh
56 . ${CONFIG:=$LUSTRE/tests/cfg/$NAME.sh}
58 # use small MDS + OST size to speed formatting time
59 # do not use too small MDSSIZE/OSTSIZE, which affect the default jouranl size
60 # STORED_MDSSIZE is used in test_18
61 STORED_MDSSIZE=$MDSSIZE
62 STORED_OSTSIZE=$OSTSIZE
66 if ! combined_mgs_mds; then
67 # bug number for skipped test: 23954
68 ALWAYS_EXCEPT="$ALWAYS_EXCEPT 24b"
71 # pass "-E lazy_itable_init" to mke2fs to speed up the formatting time
72 if [[ "$LDISKFS_MKFS_OPTS" != *lazy_itable_init* ]]; then
73 LDISKFS_MKFS_OPTS=$(csa_add "$LDISKFS_MKFS_OPTS" -E lazy_itable_init)
76 [ $(facet_fstype $SINGLEMDS) = "zfs" ] &&
77 # bug number for skipped test: LU-2778 LU-4444
78 ALWAYS_EXCEPT="$ALWAYS_EXCEPT 57b 69"
83 require_dsh_mds || exit 0
84 require_dsh_ost || exit 0
86 [ "$SLOW" = "no" ] && EXCEPT_SLOW="30a 31 45 69"
92 # The MGS must be started before the OSTs for a new fs, so start
93 # and stop to generate the startup logs.
96 wait_osc_import_state mds ost FULL
101 reformat_and_config() {
103 if ! combined_mgs_mds ; then
109 writeconf_or_reformat() {
110 # There are at most 2 OSTs for write_conf test
111 # who knows if/where $TUNEFS is installed?
112 # Better reformat if it fails...
113 writeconf_all $MDSCOUNT 2 ||
114 { echo "tunefs failed, reformatting instead" &&
115 reformat_and_config && return 0; }
125 start mgs $(mgsdevname) $MGS_MOUNT_OPTS
131 local dev=$(mdsdevname $num)
134 echo "start mds service on `facet_active_host $facet`"
135 start $facet ${dev} $MDS_MOUNT_OPTS $@ || return 94
141 local dev=$(mdsdevname $num)
144 echo "stop mds service on `facet_active_host $facet`"
145 # These tests all use non-failover stop
146 stop $facet -f || return 97
152 for num in $(seq $MDSCOUNT); do
153 start_mdt $num $@ || return 94
158 if ! combined_mgs_mds ; then
166 for num in $(seq $MDSCOUNT); do
167 stop_mdt $num || return 97
172 echo "stop mgs service on `facet_active_host mgs`"
173 # These tests all use non-failover stop
174 stop mgs -f || return 97
178 echo "start ost1 service on `facet_active_host ost1`"
179 start ost1 `ostdevname 1` $OST_MOUNT_OPTS $@ || return 95
183 echo "stop ost1 service on `facet_active_host ost1`"
184 # These tests all use non-failover stop
185 stop ost1 -f || return 98
189 echo "start ost2 service on `facet_active_host ost2`"
190 start ost2 `ostdevname 2` $OST_MOUNT_OPTS $@ || return 92
194 echo "stop ost2 service on `facet_active_host ost2`"
195 # These tests all use non-failover stop
196 stop ost2 -f || return 93
201 echo "mount $FSNAME on ${MOUNTPATH}....."
202 zconf_mount `hostname` $MOUNTPATH || return 96
206 local mountopt="remount,$1"
208 echo "remount '$1' lustre on ${MOUNTPATH}....."
209 zconf_mount `hostname` $MOUNTPATH "$mountopt" || return 96
214 echo "umount lustre on ${MOUNTPATH}....."
215 zconf_umount `hostname` $MOUNTPATH || return 97
218 manual_umount_client(){
221 echo "manual umount lustre on ${MOUNT}...."
222 do_facet client "umount -d ${FORCE} $MOUNT"
228 start_mds || error "MDT start failed"
229 start_ost || error "OST start failed"
230 mount_client $MOUNT || error "client start failed"
231 client_up || error "client_up failed"
240 unload_modules_conf () {
241 if combined_mgs_mds || ! local_mode; then
242 unload_modules || return 1
247 stop_ost || return 202
248 stop_mds || return 201
249 unload_modules_conf || return 203
253 umount_client $MOUNT || return 200
254 cleanup_nocli || return $?
258 do_facet client "cp /etc/passwd $DIR/a" || return 71
259 do_facet client "rm $DIR/a" || return 72
260 # make sure lustre is actually mounted (touch will block,
261 # but grep won't, so do it after)
262 do_facet client "grep $MOUNT' ' /proc/mounts > /dev/null" || return 73
263 echo "setup single mount lustre success"
267 do_facet client "touch $DIR/a" || return 71
268 do_facet client "rm $DIR/a" || return 72
269 do_facet client "touch $DIR2/a" || return 73
270 do_facet client "rm $DIR2/a" || return 74
271 echo "setup double mount lustre success"
276 if [ "$ONLY" == "setup" ]; then
281 if [ "$ONLY" == "cleanup" ]; then
288 #create single point mountpoint
294 check_mount || return 41
297 run_test 0 "single mount setup"
300 start_mds || error "MDT start failed"
302 echo "start ost second time..."
303 start_ost && error "2nd OST start should fail"
304 mount_client $MOUNT || error "client start failed"
305 check_mount || return 42
308 run_test 1 "start up ost twice (should return errors)"
312 echo "start mds second time.."
313 start_mdt 1 && error "2nd MDT start should fail"
316 check_mount || return 43
319 run_test 2 "start up mds twice (should return err)"
323 #mount.lustre returns an error if already in mtab
324 mount_client $MOUNT && error "2nd client mount should fail"
325 check_mount || return 44
328 run_test 3 "mount client twice (should return err)"
332 touch $DIR/$tfile || return 85
336 # ok for ost to fail shutdown
337 if [ 202 -ne $eno ]; then
342 run_test 4 "force cleanup ost, then cleanup"
344 test_5a() { # was test_5
346 touch $DIR/$tfile || return 1
347 fuser -m -v $MOUNT && echo "$MOUNT is in use by user space process."
349 stop_mds -f || return 2
351 # cleanup may return an error from the failed
352 # disconnects; for now I'll consider this successful
353 # if all the modules have unloaded.
357 echo "killing umount"
358 kill -TERM $UMOUNT_PID
359 echo "waiting for umount to finish"
361 if grep " $MOUNT " /proc/mounts; then
362 echo "test 5: /proc/mounts after failed umount"
366 echo "killing umount"
367 kill -TERM $UMOUNT_PID
368 echo "waiting for umount to finish"
370 grep " $MOUNT " /proc/mounts && echo "test 5: /proc/mounts after second umount" && return 11
374 # stop_mds is a no-op here, and should not fail
375 cleanup_nocli || return $?
376 # df may have lingering entry
378 # mtab may have lingering entry
382 while [ "$WAIT" -ne "$MAX_WAIT" ]; do
384 grep -q $MOUNT" " /etc/mtab || break
385 echo "Waiting /etc/mtab updated ... "
386 WAIT=$(( WAIT + sleep))
388 [ "$WAIT" -eq "$MAX_WAIT" ] && error "/etc/mtab is not updated in $WAIT secs"
389 echo "/etc/mtab updated in $WAIT secs"
391 run_test 5a "force cleanup mds, then cleanup"
399 grep " $MOUNT " /etc/mtab && \
400 error false "unexpected entry in mtab before mount" && return 10
404 if ! combined_mgs_mds ; then
405 trap cleanup_5b EXIT ERR
410 [ -d $MOUNT ] || mkdir -p $MOUNT
411 mount_client $MOUNT && rc=1
412 grep " $MOUNT " /etc/mtab && \
413 error "$MOUNT entry in mtab after failed mount" && rc=11
415 # stop_mds is a no-op here, and should not fail
416 cleanup_nocli || rc=$?
417 if ! combined_mgs_mds ; then
422 run_test 5b "Try to start a client with no MGS (should return errs)"
425 grep " $MOUNT " /etc/mtab && \
426 error false "unexpected entry in mtab before mount" && return 10
431 [ -d $MOUNT ] || mkdir -p $MOUNT
432 local oldfs="${FSNAME}"
433 FSNAME="wrong.${FSNAME}"
434 mount_client $MOUNT || :
436 grep " $MOUNT " /etc/mtab && \
437 error "$MOUNT entry in mtab after failed mount" && rc=11
439 cleanup_nocli || rc=$?
442 run_test 5c "cleanup after failed mount (bug 2712) (should return errs)"
445 grep " $MOUNT " /etc/mtab && \
446 error false "unexpected entry in mtab before mount" && return 10
452 mount_client $MOUNT || rc=1
454 grep " $MOUNT " /etc/mtab && \
455 error "$MOUNT entry in mtab after unmount" && rc=11
458 run_test 5d "mount with ost down"
461 grep " $MOUNT " /etc/mtab && \
462 error false "unexpected entry in mtab before mount" && return 10
468 #define OBD_FAIL_PTLRPC_DELAY_SEND 0x506
469 do_facet client "lctl set_param fail_loc=0x80000506"
470 mount_client $MOUNT || echo "mount failed (not fatal)"
472 grep " $MOUNT " /etc/mtab && \
473 error "$MOUNT entry in mtab after unmount" && rc=11
476 run_test 5e "delayed connect, don't crash (bug 10268)"
479 if combined_mgs_mds ; then
480 skip "combined mgs and mds"
484 grep " $MOUNT " /etc/mtab && \
485 error false "unexpected entry in mtab before mount" && return 10
489 [ -d $MOUNT ] || mkdir -p $MOUNT
490 mount_client $MOUNT &
492 echo client_mount pid is $pid
496 if ! ps -f -p $pid >/dev/null; then
499 grep " $MOUNT " /etc/mtab && echo "test 5f: mtab after mount"
500 error "mount returns $rc, expected to hang"
509 # mount should succeed after start mds
512 [ $rc -eq 0 ] || error "mount returned $rc"
513 grep " $MOUNT " /etc/mtab && echo "test 5f: mtab after mount"
517 run_test 5f "mds down, cleanup after failed mount (bug 2712)"
522 mount_client ${MOUNT} || return 87
523 touch $DIR/a || return 86
526 run_test 6 "manual umount, then mount again"
531 cleanup_nocli || return $?
533 run_test 7 "manual umount, then cleanup"
538 check_mount2 || return 45
539 umount_client $MOUNT2
542 run_test 8 "double mount setup"
547 do_facet ost1 lctl set_param debug=\'inode trace\' || return 1
548 do_facet ost1 lctl set_param subsystem_debug=\'mds ost\' || return 1
550 CHECK_PTLDEBUG="`do_facet ost1 lctl get_param -n debug`"
551 if [ "$CHECK_PTLDEBUG" ] && { \
552 [ "$CHECK_PTLDEBUG" = "trace inode warning error emerg console" ] ||
553 [ "$CHECK_PTLDEBUG" = "trace inode" ]; }; then
554 echo "lnet.debug success"
556 echo "lnet.debug: want 'trace inode', have '$CHECK_PTLDEBUG'"
559 CHECK_SUBSYS="`do_facet ost1 lctl get_param -n subsystem_debug`"
560 if [ "$CHECK_SUBSYS" ] && [ "$CHECK_SUBSYS" = "mds ost" ]; then
561 echo "lnet.subsystem_debug success"
563 echo "lnet.subsystem_debug: want 'mds ost', have '$CHECK_SUBSYS'"
566 stop_ost || return $?
568 run_test 9 "test ptldebug and subsystem for mkfs"
576 do_facet $facet "test -b $dev" || rc=1
577 if [[ "$size" ]]; then
578 local in=$(do_facet $facet "dd if=$dev of=/dev/null bs=1k count=1 skip=$size 2>&1" |\
579 awk '($3 == "in") { print $1 }')
580 [[ $in = "1+0" ]] || rc=1
586 # Test 16 was to "verify that lustre will correct the mode of OBJECTS".
587 # But with new MDS stack we don't care about the mode of local objects
588 # anymore, so this test is removed. See bug 22944 for more details.
592 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
593 skip "Only applicable to ldiskfs-based MDTs"
598 check_mount || return 41
601 echo "Remove mds config log"
602 if ! combined_mgs_mds ; then
606 do_facet mgs "$DEBUGFS -w -R 'unlink CONFIGS/$FSNAME-MDT0000' \
607 $(mgsdevname) || return \$?" || return $?
609 if ! combined_mgs_mds ; then
614 start_mds && return 42
617 run_test 17 "Verify failed mds_postsetup won't fail assertion (2936) (should return errs)"
620 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
621 skip "Only applicable to ldiskfs-based MDTs"
625 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
630 # check if current MDSSIZE is large enough
631 [ $MDSSIZE -ge $MIN ] && OK=1 && myMDSSIZE=$MDSSIZE && \
632 log "use MDSSIZE=$MDSSIZE"
634 # check if the global config has a large enough MDSSIZE
635 [ -z "$OK" -a ! -z "$STORED_MDSSIZE" ] && [ $STORED_MDSSIZE -ge $MIN ] && \
636 OK=1 && myMDSSIZE=$STORED_MDSSIZE && \
637 log "use STORED_MDSSIZE=$STORED_MDSSIZE"
639 # check if the block device is large enough
640 is_blkdev $SINGLEMDS $MDSDEV $MIN
641 local large_enough=$?
642 if [ -n "$OK" ]; then
643 [ $large_enough -ne 0 ] && OK=""
645 [ $large_enough -eq 0 ] && OK=1 && myMDSSIZE=$MIN &&
646 log "use device $MDSDEV with MIN=$MIN"
649 # check if a loopback device has enough space for fs metadata (5%)
651 if [ -z "$OK" ]; then
652 local SPACE=$(do_facet $SINGLEMDS "[ -f $MDSDEV -o ! -e $MDSDEV ] && df -P \\\$(dirname $MDSDEV)" |
653 awk '($1 != "Filesystem") {print $4}')
654 ! [ -z "$SPACE" ] && [ $SPACE -gt $((MIN / 20)) ] && \
655 OK=1 && myMDSSIZE=$MIN && \
656 log "use file $MDSDEV with MIN=$MIN"
659 [ -z "$OK" ] && skip_env "$MDSDEV too small for ${MIN}kB MDS" && return
662 echo "mount mds with large journal..."
664 local OLD_MDSSIZE=$MDSSIZE
668 echo "mount lustre system..."
670 check_mount || return 41
672 echo "check journal size..."
673 local FOUNDSIZE=$(do_facet $SINGLEMDS "$DEBUGFS -c -R 'stat <8>' $MDSDEV" | awk '/Size: / { print $NF; exit;}')
674 if [ $FOUNDSIZE -gt $((32 * 1024 * 1024)) ]; then
675 log "Success: mkfs creates large journals. Size: $((FOUNDSIZE >> 20))M"
677 error "expected journal size > 32M, found $((FOUNDSIZE >> 20))M"
685 run_test 18 "check mkfs creates large journals"
688 start_mds || return 1
689 stop_mds -f || return 2
691 run_test 19a "start/stop MDS without OSTs"
694 start_ost || return 1
695 stop_ost -f || return 2
697 run_test 19b "start/stop OSTs without MDS"
700 # first format the ost/mdt
704 check_mount || return 43
706 remount_client ro $MOUNT || return 44
707 touch $DIR/$tfile && echo "$DIR/$tfile created incorrectly" && return 45
708 [ -e $DIR/$tfile ] && echo "$DIR/$tfile exists incorrectly" && return 46
709 remount_client rw $MOUNT || return 47
711 [ ! -f $DIR/$tfile ] && echo "$DIR/$tfile missing" && return 48
712 MCNT=`grep -c $MOUNT /etc/mtab`
713 [ "$MCNT" -ne 1 ] && echo "$MOUNT in /etc/mtab $MCNT times" && return 49
718 run_test 20 "remount ro,rw mounts work and doesn't break /etc/mtab"
723 wait_osc_import_state mds ost FULL
727 run_test 21a "start mds before ost, stop ost first"
732 wait_osc_import_state mds ost FULL
736 run_test 21b "start ost before mds, stop mds first"
742 wait_osc_import_state mds ost2 FULL
746 #writeconf to remove all ost2 traces for subsequent tests
747 writeconf_or_reformat
749 run_test 21c "start mds between two osts, stop mds last"
752 if combined_mgs_mds ; then
753 skip "need separate mgs device" && return 0
763 wait_osc_import_state mds ost2 FULL
769 #writeconf to remove all ost2 traces for subsequent tests
770 writeconf_or_reformat
773 run_test 21d "start mgs then ost and then mds"
778 echo Client mount with ost in logs, but none running
780 # wait until mds connected to ost and open client connection
781 wait_osc_import_state mds ost FULL
784 # check_mount will block trying to contact ost
785 mcreate $DIR/$tfile || return 40
786 rm -f $DIR/$tfile || return 42
790 echo Client mount with a running ost
793 # if gss enabled, wait full time to let connection from
794 # mds to ost be established, due to the mismatch between
795 # initial connect timeout and gss context negotiation timeout.
796 # This perhaps could be remove after AT landed.
797 echo "sleep $((TIMEOUT + TIMEOUT + TIMEOUT))s"
798 sleep $((TIMEOUT + TIMEOUT + TIMEOUT))
801 wait_osc_import_state mds ost FULL
802 wait_osc_import_state client ost FULL
803 check_mount || return 41
808 run_test 22 "start a client before osts (should return errs)"
810 test_23a() { # was test_23
814 # force down client so that recovering mds waits for reconnect
815 local running=$(grep -c $MOUNT /proc/mounts) || true
816 if [ $running -ne 0 ]; then
817 echo "Stopping client $MOUNT (opts: -f)"
821 # enter recovery on mds
823 # try to start a new client
824 mount_client $MOUNT &
826 MOUNT_PID=$(ps -ef | grep "t lustre" | grep -v grep | awk '{print $2}')
827 MOUNT_LUSTRE_PID=`ps -ef | grep mount.lustre | grep -v grep | awk '{print $2}'`
828 echo mount pid is ${MOUNT_PID}, mount.lustre pid is ${MOUNT_LUSTRE_PID}
830 ps --ppid $MOUNT_LUSTRE_PID
831 echo "waiting for mount to finish"
833 # "ctrl-c" sends SIGINT but it usually (in script) does not work on child process
834 # SIGTERM works but it does not spread to offspring processses
835 kill -s TERM $MOUNT_PID
836 kill -s TERM $MOUNT_LUSTRE_PID
837 # we can not wait $MOUNT_PID because it is not a child of this shell
843 while [ "$WAIT" -lt "$MAX_WAIT" ]; do
845 PID1=$(ps -ef | awk '{print $2}' | grep -w $MOUNT_PID)
846 PID2=$(ps -ef | awk '{print $2}' | grep -w $MOUNT_LUSTRE_PID)
849 [ -z "$PID1" -a -z "$PID2" ] && break
850 echo "waiting for mount to finish ... "
851 WAIT=$(( WAIT + sleep))
853 if [ "$WAIT" -eq "$MAX_WAIT" ]; then
854 error "MOUNT_PID $MOUNT_PID and "\
855 "MOUNT_LUSTRE_PID $MOUNT_LUSTRE_PID still not killed in $WAIT secs"
858 stop_mds || error "stopping MDSes failed"
859 stop_ost || error "stopping OSSes failed"
861 run_test 23a "interrupt client during recovery mount delay"
866 test_23b() { # was test_23
869 # Simulate -EINTR during mount OBD_FAIL_LDLM_CLOSE_THREAD
870 lctl set_param fail_loc=0x80000313
874 run_test 23b "Simulate -EINTR during mount"
876 fs2mds_HOST=$mds_HOST
877 fs2ost_HOST=$ost_HOST
879 MDSDEV1_2=$fs2mds_DEV
880 OSTDEV1_2=$fs2ost_DEV
881 OSTDEV2_2=$fs3ost_DEV
885 echo "umount $MOUNT2 ..."
886 umount $MOUNT2 || true
887 echo "stopping fs2mds ..."
888 stop fs2mds -f || true
889 echo "stopping fs2ost ..."
890 stop fs2ost -f || true
894 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
896 if [ -z "$fs2ost_DEV" -o -z "$fs2mds_DEV" ]; then
897 is_blkdev $SINGLEMDS $MDSDEV && \
898 skip_env "mixed loopback and real device not working" && return
901 [ -n "$ost1_HOST" ] && fs2ost_HOST=$ost1_HOST
903 local fs2mdsdev=$(mdsdevname 1_2)
904 local fs2ostdev=$(ostdevname 1_2)
905 local fs2mdsvdev=$(mdsvdevname 1_2)
906 local fs2ostvdev=$(ostvdevname 1_2)
908 # test 8-char fsname as well
909 local FSNAME2=test1234
911 add fs2mds $(mkfs_opts mds1 ${fs2mdsdev} ) --nomgs --mgsnode=$MGSNID \
912 --fsname=${FSNAME2} --reformat $fs2mdsdev $fs2mdsvdev || exit 10
914 add fs2ost $(mkfs_opts ost1 ${fs2ostdev}) --fsname=${FSNAME2} \
915 --reformat $fs2ostdev $fs2ostvdev || exit 10
918 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS && trap cleanup_fs2 EXIT INT
919 start fs2ost $fs2ostdev $OST_MOUNT_OPTS
921 $MOUNT_CMD $MGSNID:/${FSNAME2} $MOUNT2 || return 1
923 check_mount || return 2
924 # files written on 1 should not show up on 2
925 cp /etc/passwd $DIR/$tfile
927 [ -e $MOUNT2/$tfile ] && error "File bleed" && return 7
930 cp /etc/passwd $MOUNT2/b || return 3
931 rm $MOUNT2/b || return 4
932 # 2 is actually mounted
933 grep $MOUNT2' ' /proc/mounts > /dev/null || return 5
935 facet_failover fs2mds
936 facet_failover fs2ost
939 # the MDS must remain up until last MDT
941 MDS=$(do_facet $SINGLEMDS "lctl get_param -n devices" | awk '($3 ~ "mdt" && $4 ~ "MDT") { print $4 }' | head -1)
942 [ -z "$MDS" ] && error "No MDT" && return 8
944 cleanup_nocli || return 6
946 run_test 24a "Multiple MDTs on a single node"
949 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
951 if [ -z "$fs2mds_DEV" ]; then
952 local dev=${SINGLEMDS}_dev
954 is_blkdev $SINGLEMDS $MDSDEV && \
955 skip_env "mixed loopback and real device not working" && return
958 local fs2mdsdev=$(mdsdevname 1_2)
959 local fs2mdsvdev=$(mdsvdevname 1_2)
961 add fs2mds $(mkfs_opts mds1 ${fs2mdsdev} ) --mgs --fsname=${FSNAME}2 \
962 --reformat $fs2mdsdev $fs2mdsvdev || exit 10
964 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS && return 2
968 run_test 24b "Multiple MGSs on a single node (should return err)"
972 check_mount || return 2
973 local MODULES=$($LCTL modules | awk '{ print $2 }')
974 rmmod $MODULES 2>/dev/null || true
977 run_test 25 "Verify modules are referenced"
981 # we need modules before mount for sysctl, so make sure...
982 do_facet $SINGLEMDS "lsmod | grep -q lustre || modprobe lustre"
983 #define OBD_FAIL_MDS_FS_SETUP 0x135
984 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000135"
985 start_mds && echo MDS started && return 1
986 lctl get_param -n devices
987 DEVS=$(lctl get_param -n devices | egrep -v MG | wc -l)
988 [ $DEVS -gt 0 ] && return 2
989 # start mds to drop writeconf setting
990 start_mds || return 3
992 unload_modules_conf || return $?
994 run_test 26 "MDT startup failure cleans LOV (should return errs)"
997 start_ost || return 1
998 start_mds || return 2
999 echo "Requeue thread should have started: "
1000 ps -e | grep ll_cfg_requeue
1001 set_conf_param_and_check ost1 \
1002 "lctl get_param -n obdfilter.$FSNAME-OST0000.client_cache_seconds" \
1003 "$FSNAME-OST0000.ost.client_cache_seconds" || return 3
1006 run_test 27a "Reacquire MGS lock if OST started first"
1011 local device=$(do_facet $SINGLEMDS "lctl get_param -n devices" |
1012 awk '($3 ~ "mdt" && $4 ~ "MDT0000") { print $4 }')
1014 facet_failover $SINGLEMDS
1015 set_conf_param_and_check $SINGLEMDS \
1016 "lctl get_param -n mdt.$device.identity_acquire_expire" \
1017 "$device.mdt.identity_acquire_expire" || return 3
1018 set_conf_param_and_check client \
1019 "lctl get_param -n mdc.$device-mdc-*.max_rpcs_in_flight"\
1020 "$device.mdc.max_rpcs_in_flight" || return 4
1024 run_test 27b "Reacquire MGS lock after failover"
1028 TEST="lctl get_param -n llite.$FSNAME-*.max_read_ahead_whole_mb"
1029 PARAM="$FSNAME.llite.max_read_ahead_whole_mb"
1031 FINAL=$(($ORIG + 1))
1032 set_conf_param_and_check client "$TEST" "$PARAM" $FINAL || return 3
1033 FINAL=$(($FINAL + 1))
1034 set_conf_param_and_check client "$TEST" "$PARAM" $FINAL || return 4
1035 umount_client $MOUNT || return 200
1038 if [ $RESULT -ne $FINAL ]; then
1039 echo "New config not seen: wanted $FINAL got $RESULT"
1042 echo "New config success: got $RESULT"
1044 set_conf_param_and_check client "$TEST" "$PARAM" $ORIG || return 5
1047 run_test 28 "permanent parameter setting"
1049 test_28a() { # LU-4221
1050 [[ $(lustre_version_code ost1) -ge $(version_code 2.5.52) ]] ||
1051 { skip "Need OST version at least 2.5.52" && return 0; }
1052 [ "$(facet_fstype ost1)" = "zfs" ] &&
1053 skip "LU-4221: no such proc params for ZFS OSTs" && return
1060 local device="$FSNAME-OST0000"
1064 # In this test we will set three kinds of proc parameters with
1066 # 1. the ones moved from the OFD to the OSD, and only their
1067 # symlinks kept in obdfilter
1068 # 2. non-symlink ones in the OFD
1069 # 3. non-symlink ones in the OSD
1072 # prepare a symlink parameter in the OFD
1073 name="writethrough_cache_enable"
1074 param="$device.ost.$name"
1075 cmd="$LCTL get_param -n obdfilter.$device.$name"
1077 # conf_param the symlink parameter in the OFD
1078 old=$(do_facet ost1 $cmd)
1079 new=$(((old + 1) % 2))
1080 set_conf_param_and_check ost1 "$cmd" "$param" $new ||
1081 error "lctl conf_param $device.ost.$param=$new failed"
1083 # conf_param the target parameter in the OSD
1084 param="$device.osd.$name"
1085 cmd="$LCTL get_param -n osd-*.$device.$name"
1086 set_conf_param_and_check ost1 "$cmd" "$param" $old ||
1087 error "lctl conf_param $device.osd.$param=$old failed"
1090 # prepare a non-symlink parameter in the OFD
1091 name="client_cache_seconds"
1092 param="$device.ost.$name"
1093 cmd="$LCTL get_param -n obdfilter.$device.$name"
1095 # conf_param the parameter in the OFD
1096 old=$(do_facet ost1 $cmd)
1098 set_conf_param_and_check ost1 "$cmd" "$param" $new ||
1099 error "lctl conf_param $device.ost.$param=$new failed"
1100 set_conf_param_and_check ost1 "$cmd" "$param" $old ||
1101 error "lctl conf_param $device.ost.$param=$old failed"
1104 # prepare a non-symlink parameter in the OSD
1105 name="lma_self_repair"
1106 param="$device.osd.$name"
1107 cmd="$LCTL get_param -n osd-*.$device.$name"
1109 # conf_param the parameter in the OSD
1110 old=$(do_facet ost1 $cmd)
1111 new=$(((old + 1) % 2))
1112 set_conf_param_and_check ost1 "$cmd" "$param" $new ||
1113 error "lctl conf_param $device.osd.$param=$new failed"
1114 set_conf_param_and_check ost1 "$cmd" "$param" $old ||
1115 error "lctl conf_param $device.osd.$param=$old failed"
1119 run_test 28a "set symlink parameters permanently with conf_param"
1122 [ "$OSTCOUNT" -lt "2" ] && skip_env "$OSTCOUNT < 2, skipping" && return
1123 setup > /dev/null 2>&1
1127 local PARAM="$FSNAME-OST0001.osc.active"
1128 local PROC_ACT="osc.$FSNAME-OST0001-osc-[^M]*.active"
1129 local PROC_UUID="osc.$FSNAME-OST0001-osc-[^M]*.ost_server_uuid"
1131 ACTV=$(lctl get_param -n $PROC_ACT)
1133 set_conf_param_and_check client \
1134 "lctl get_param -n $PROC_ACT" "$PARAM" $DEAC || return 2
1135 # also check ost_server_uuid status
1136 RESULT=$(lctl get_param -n $PROC_UUID | grep DEACTIV)
1137 if [ -z "$RESULT" ]; then
1138 echo "Live client not deactivated: $(lctl get_param -n $PROC_UUID)"
1141 echo "Live client success: got $RESULT"
1145 for num in $(seq $MDSCOUNT); do
1146 local mdtosc=$(get_mdtosc_proc_path mds${num} $FSNAME-OST0001)
1147 local MPROC="osc.$mdtosc.active"
1152 RESULT=$(do_facet mds${num} " lctl get_param -n $MPROC")
1153 [ ${PIPESTATUS[0]} = 0 ] || error "Can't read $MPROC"
1154 if [ $RESULT -eq $DEAC ]; then
1155 echo -n "MDT deactivated also after"
1156 echo "$WAIT sec (got $RESULT)"
1160 if [ $WAIT -eq $MAX ]; then
1161 echo -n "MDT not deactivated: wanted $DEAC"
1165 echo "Waiting $(($MAX - $WAIT))secs for MDT deactivated"
1168 # test new client starts deactivated
1169 umount_client $MOUNT || return 200
1171 RESULT=$(lctl get_param -n $PROC_UUID | grep DEACTIV | grep NEW)
1172 if [ -z "$RESULT" ]; then
1173 echo "New client not deactivated from start: $(lctl get_param -n $PROC_UUID)"
1176 echo "New client success: got $RESULT"
1179 # make sure it reactivates
1180 set_conf_param_and_check client \
1181 "lctl get_param -n $PROC_ACT" "$PARAM" $ACTV || return 6
1183 umount_client $MOUNT
1186 #writeconf to remove all ost2 traces for subsequent tests
1187 writeconf_or_reformat
1189 run_test 29 "permanently remove an OST"
1194 echo Big config llog
1195 TEST="lctl get_param -n llite.$FSNAME-*.max_read_ahead_whole_mb"
1197 LIST=(1 2 3 4 5 4 3 2 1 2 3 4 5 4 3 2 1 2 3 4 5)
1198 for i in ${LIST[@]}; do
1199 set_conf_param_and_check client "$TEST" \
1200 "$FSNAME.llite.max_read_ahead_whole_mb" $i || return 3
1202 # make sure client restart still works
1203 umount_client $MOUNT
1204 mount_client $MOUNT || return 4
1205 [ "$($TEST)" -ne "$i" ] && error "Param didn't stick across restart $($TEST) != $i"
1208 echo Erase parameter setting
1209 do_facet mgs "$LCTL conf_param -d $FSNAME.llite.max_read_ahead_whole_mb" || return 6
1210 umount_client $MOUNT
1211 mount_client $MOUNT || return 6
1213 echo "deleted (default) value=$FINAL, orig=$ORIG"
1214 # assumes this parameter started at the default value
1215 [ "$FINAL" -eq "$ORIG" ] || fail "Deleted value=$FINAL, orig=$ORIG"
1219 run_test 30a "Big config llog and conf_param deletion"
1224 # Make a fake nid. Use the OST nid, and add 20 to the least significant
1225 # numerical part of it. Hopefully that's not already a failover address for
1227 OSTNID=$(do_facet ost1 "$LCTL get_param nis" | tail -1 | awk '{print $1}')
1228 ORIGVAL=$(echo $OSTNID | egrep -oi "[0-9]*@")
1229 NEWVAL=$((($(echo $ORIGVAL | egrep -oi "[0-9]*") + 20) % 256))
1230 NEW=$(echo $OSTNID | sed "s/$ORIGVAL/$NEWVAL@/")
1231 echo "Using fake nid $NEW"
1233 TEST="$LCTL get_param -n osc.$FSNAME-OST0000-osc-[^M]*.import |
1234 grep failover_nids | sed -n 's/.*\($NEW\).*/\1/p'"
1235 set_conf_param_and_check client "$TEST" \
1236 "$FSNAME-OST0000.failover.node" $NEW ||
1237 error "didn't add failover nid $NEW"
1238 NIDS=$($LCTL get_param -n osc.$FSNAME-OST0000-osc-[^M]*.import |
1241 # The NIDS value is the failover nid strings and "[" and "]". So
1242 # we need to subtract the space taken by the delimiters. This has
1243 # changed from earlier version of Lustre but this test is run only
1244 # locally so this change will not break interop. See LU-3386
1245 NIDCOUNT=$(($(echo "$NIDS" | wc -w) - 3))
1246 echo "should have 2 failover nids: $NIDCOUNT"
1247 [ $NIDCOUNT -eq 2 ] || error "Failover nid not added"
1248 do_facet mgs "$LCTL conf_param -d $FSNAME-OST0000.failover.node" ||
1249 error "conf_param delete failed"
1250 umount_client $MOUNT
1251 mount_client $MOUNT || return 3
1253 NIDS=$($LCTL get_param -n osc.$FSNAME-OST0000-osc-[^M]*.import |
1256 NIDCOUNT=$(($(echo "$NIDS" | wc -w) - 3))
1257 echo "only 1 final nid should remain: $NIDCOUNT"
1258 [ $NIDCOUNT -eq 1 ] || error "Failover nids not removed"
1262 run_test 30b "Remove failover nids"
1264 test_31() { # bug 10734
1265 # ipaddr must not exist
1266 $MOUNT_CMD 4.3.2.1@tcp:/lustre $MOUNT || true
1269 run_test 31 "Connect to non-existent node (shouldn't crash)"
1273 T32_BLIMIT=20480 # Kbytes
1277 # This is not really a test but a tool to create new disk
1278 # image tarballs for the upgrade tests.
1280 # Disk image tarballs should be created on single-node
1281 # clusters by running this test with default configurations
1282 # plus a few mandatory environment settings that are verified
1283 # at the beginning of the test.
1285 test_32newtarball() {
1289 local tmp=$TMP/t32_image_create
1291 if [ $FSNAME != t32fs -o $MDSCOUNT -ne 1 -o \
1292 \( -z "$MDSDEV" -a -z "$MDSDEV1" \) -o $OSTCOUNT -ne 1 -o \
1293 -z "$OSTDEV1" ]; then
1294 error "Needs FSNAME=t32fs MDSCOUNT=1 MDSDEV1=<nonexistent_file>" \
1295 "(or MDSDEV, in the case of b1_8) OSTCOUNT=1" \
1296 "OSTDEV1=<nonexistent_file>"
1300 echo "Found stale $tmp"
1305 tar cf - -C $src . | tar xf - -C $tmp/src
1306 dd if=/dev/zero of=$tmp/src/t32_qf_old bs=1M \
1307 count=$(($T32_BLIMIT / 1024 / 2))
1308 chown $T32_QID.$T32_QID $tmp/src/t32_qf_old
1310 # format ost with comma-separated NIDs to verify LU-4460
1311 local failnid="$(h2$NETTYPE 1.2.3.4),$(h2$NETTYPE 4.3.2.1)"
1312 MGSNID="$MGSNID,$MGSNID" OSTOPT="--failnode=$failnid" formatall
1316 [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.3.50) ] &&
1317 $LFS quotacheck -ug /mnt/$FSNAME
1318 $LFS setquota -u $T32_QID -b 0 -B $T32_BLIMIT -i 0 -I $T32_ILIMIT \
1321 tar cf - -C $tmp/src . | tar xf - -C /mnt/$FSNAME
1328 ls -Rni --time-style=+%s >$tmp/img/list
1329 find . ! -name .lustre -type f -exec sha1sum {} \; |
1330 sort -k 2 >$tmp/img/sha1sums
1332 $LCTL get_param -n version | head -n 1 |
1333 sed -e 's/^lustre: *//' >$tmp/img/commit
1335 [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.3.50) ] &&
1336 $LFS quotaon -ug /mnt/$FSNAME
1337 $LFS quota -u $T32_QID -v /mnt/$FSNAME
1338 $LFS quota -v -u $T32_QID /mnt/$FSNAME |
1339 awk 'BEGIN { num='1' } { if ($1 == "'/mnt/$FSNAME'") \
1340 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1341 | tr -d "*" > $tmp/img/bspace
1342 $LFS quota -v -u $T32_QID /mnt/$FSNAME |
1343 awk 'BEGIN { num='5' } { if ($1 == "'/mnt/$FSNAME'") \
1344 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1345 | tr -d "*" > $tmp/img/ispace
1350 find -type f -exec sha1sum {} \; | sort -k 2 >$tmp/sha1sums.src
1353 if ! diff -u $tmp/sha1sums.src $tmp/img/sha1sums; then
1354 echo "Data verification failed"
1357 uname -r >$tmp/img/kernel
1358 uname -m >$tmp/img/arch
1360 mv ${MDSDEV1:-$MDSDEV} $tmp/img
1361 mv $OSTDEV1 $tmp/img
1363 version=$(sed -e 's/\(^[0-9]\+\.[0-9]\+\)\(.*$\)/\1/' $tmp/img/commit |
1364 sed -e 's/\./_/g') # E.g., "1.8.7" -> "1_8"
1367 tar cjvf $dst/disk$version-$(facet_fstype $SINGLEMDS).tar.bz2 -S *
1372 #run_test 32newtarball "Create a new test_32 disk image tarball for this version"
1375 # The list of applicable tarballs is returned via the caller's
1376 # variable "tarballs".
1379 local node=$(facet_active_host $SINGLEMDS)
1380 local r="do_node $node"
1382 if [ "$CLIENTONLY" ]; then
1383 skip "Client-only testing"
1387 if ! $r which $TUNEFS; then
1388 skip_env "tunefs.lustre required on $node"
1392 local IMGTYPE=$(facet_fstype $SINGLEMDS)
1394 tarballs=$($r find $RLUSTRE/tests -maxdepth 1 -name \'disk*-$IMGTYPE.tar.bz2\')
1396 if [ -z "$tarballs" ]; then
1397 skip "No applicable tarballs found"
1402 t32_test_cleanup() {
1404 local fstype=$(facet_fstype $SINGLEMDS)
1407 if $shall_cleanup_lustre; then
1408 umount $tmp/mnt/lustre || rc=$?
1410 if $shall_cleanup_mdt; then
1411 $r umount -d $tmp/mnt/mdt || rc=$?
1413 if $shall_cleanup_mdt1; then
1414 $r umount -d $tmp/mnt/mdt1 || rc=$?
1416 if $shall_cleanup_ost; then
1417 $r umount -d $tmp/mnt/ost || rc=$?
1422 if [ $fstype == "zfs" ]; then
1423 $r $ZPOOL destroy t32fs-mdt1 || rc=$?
1424 $r $ZPOOL destroy t32fs-ost1 || rc=$?
1429 t32_bits_per_long() {
1431 # Yes, this is not meant to be perfect.
1441 t32_reload_modules() {
1443 local all_removed=false
1446 while ((i < 20)); do
1447 echo "Unloading modules on $node: Attempt $i"
1448 do_rpc_nodes $node $LUSTRE_RMMOD $(facet_fstype $SINGLEMDS) &&
1450 do_rpc_nodes $node check_mem_leak || return 1
1451 if $all_removed; then
1452 do_rpc_nodes $node load_modules
1458 echo "Unloading modules on $node: Given up"
1462 t32_wait_til_devices_gone() {
1468 echo wait for devices to go
1469 while ((i < 20)); do
1470 devices=$(do_rpc_nodes $node $LCTL device_list | wc -l)
1471 loops=$(do_rpc_nodes $node losetup -a | grep -c t32)
1472 ((devices == 0 && loops == 0)) && return 0
1476 echo "waiting for dev on $node: dev $devices loop $loops given up"
1477 do_rpc_nodes $node "losetup -a"
1478 do_rpc_nodes $node "$LCTL devices_list"
1482 t32_verify_quota() {
1486 local fstype=$(facet_fstype $SINGLEMDS)
1490 $LFS quota -u $T32_QID -v $mnt
1492 qval=$($LFS quota -v -u $T32_QID $mnt |
1493 awk 'BEGIN { num='1' } { if ($1 == "'$mnt'") \
1494 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1496 [ $qval -eq $img_bspace ] || {
1497 echo "bspace, act:$qval, exp:$img_bspace"
1501 qval=$($LFS quota -v -u $T32_QID $mnt |
1502 awk 'BEGIN { num='5' } { if ($1 == "'$mnt'") \
1503 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1505 [ $qval -eq $img_ispace ] || {
1506 echo "ispace, act:$qval, exp:$img_ispace"
1510 qval=$($LFS quota -v -u $T32_QID $mnt |
1511 awk 'BEGIN { num='3' } { if ($1 == "'$mnt'") \
1512 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1514 [ $qval -eq $T32_BLIMIT ] || {
1515 echo "blimit, act:$qval, exp:$T32_BLIMIT"
1519 qval=$($LFS quota -v -u $T32_QID $mnt |
1520 awk 'BEGIN { num='7' } { if ($1 == "'$mnt'") \
1521 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1523 [ $qval -eq $T32_ILIMIT ] || {
1524 echo "ilimit, act:$qval, exp:$T32_ILIMIT"
1528 do_node $node $LCTL conf_param $fsname.quota.mdt=ug
1529 cmd="$LCTL get_param -n osd-$fstype.$fsname-MDT0000"
1530 cmd=$cmd.quota_slave.enabled
1531 wait_update $node "$cmd" "ug" || {
1532 echo "Enable mdt quota failed"
1536 do_node $node $LCTL conf_param $fsname.quota.ost=ug
1537 cmd="$LCTL get_param -n osd-$fstype.$fsname-OST0000"
1538 cmd=$cmd.quota_slave.enabled
1539 wait_update $node "$cmd" "ug" || {
1540 echo "Enable ost quota failed"
1545 runas -u $T32_QID -g $T32_QID dd if=/dev/zero of=$mnt/t32_qf_new \
1546 bs=1M count=$(($T32_BLIMIT / 1024)) oflag=sync && {
1547 echo "Write succeed, but expect -EDQUOT"
1550 rm -f $mnt/t32_qf_new
1552 runas -u $T32_QID -g $T32_QID createmany -m $mnt/t32_qf_ \
1554 echo "Create succeed, but expect -EDQUOT"
1557 unlinkmany $mnt/t32_qf_ $T32_ILIMIT
1565 local dne_upgrade=${dne_upgrade:-"no"}
1566 local ff_convert=${ff_convert:-"no"}
1567 local shall_cleanup_mdt=false
1568 local shall_cleanup_mdt1=false
1569 local shall_cleanup_ost=false
1570 local shall_cleanup_lustre=false
1571 local node=$(facet_active_host $SINGLEMDS)
1572 local r="do_node $node"
1573 local node2=$(facet_active_host mds2)
1581 local nid=$($r $LCTL list_nids | head -1)
1587 local fstype=$(facet_fstype $SINGLEMDS)
1588 local mdt_dev=$tmp/mdt
1589 local ost_dev=$tmp/ost
1591 trap 'trap - RETURN; t32_test_cleanup' RETURN
1593 mkdir -p $tmp/mnt/lustre
1594 $r mkdir -p $tmp/mnt/{mdt,ost}
1595 $r tar xjvf $tarball -S -C $tmp || {
1596 error_noexit "Unpacking the disk image tarball"
1599 img_commit=$($r cat $tmp/commit)
1600 img_kernel=$($r cat $tmp/kernel)
1601 img_arch=$($r cat $tmp/arch)
1602 img_bspace=$($r cat $tmp/bspace)
1603 img_ispace=$($r cat $tmp/ispace)
1604 echo "Upgrading from $(basename $tarball), created with:"
1605 echo " Commit: $img_commit"
1606 echo " Kernel: $img_kernel"
1607 echo " Arch: $img_arch"
1608 echo "OST version: $(get_lustre_version ost1)"
1610 # The convertion can be made only when both of the following
1611 # conditions are satisfied:
1612 # - ost device img version < 2.3.64
1613 # - ost server version >= 2.5
1614 [ $(version_code $img_commit) -ge $(version_code 2.3.64) -o \
1615 $(lustre_version_code ost1) -lt $(version_code 2.5.0) ] &&
1618 if [ $fstype == "zfs" ]; then
1620 $r $ZPOOL import -f -d $tmp t32fs-mdt1
1621 $r $ZPOOL import -f -d $tmp t32fs-ost1
1622 mdt_dev=t32fs-mdt1/mdt1
1623 ost_dev=t32fs-ost1/ost1
1624 wait_update_facet $SINGLEMDS "$ZPOOL list |
1625 awk '/^t32fs-mdt1/ { print \\\$1 }'" "t32fs-mdt1" || {
1626 error_noexit "import zfs pool failed"
1631 $r $LCTL set_param debug="$PTLDEBUG"
1633 $r $TUNEFS --dryrun $mdt_dev || {
1635 error_noexit "tunefs.lustre before mounting the MDT"
1638 if [ "$writeconf" ]; then
1640 if [ $fstype == "ldiskfs" ]; then
1642 $r $TUNEFS --quota $mdt_dev || {
1644 error_noexit "Enable mdt quota feature"
1649 if [ -n "$($LCTL list_nids | grep -v '\(tcp\|lo\)[[:digit:]]*$')" ]; then
1650 [[ $(lustre_version_code mgs) -ge $(version_code 2.3.59) ]] ||
1651 { skip "LU-2200: Cannot run over Inifiniband w/o lctl replace_nids "
1652 "(Need MGS version at least 2.3.59)"; return 0; }
1654 local osthost=$(facet_active_host ost1)
1655 local ostnid=$(do_node $osthost $LCTL list_nids | head -1)
1658 if [ $fstype == "ldiskfs" ]; then
1661 $r $MOUNT_CMD -o $mopts $mdt_dev $tmp/mnt/mdt
1662 $r lctl replace_nids $fsname-OST0000 $ostnid
1663 $r lctl replace_nids $fsname-MDT0000 $nid
1664 $r umount -d $tmp/mnt/mdt
1667 mopts=exclude=$fsname-OST0000
1668 if [ $fstype == "ldiskfs" ]; then
1673 t32_wait_til_devices_gone $node
1675 $r $MOUNT_CMD -o $mopts $mdt_dev $tmp/mnt/mdt || {
1677 error_noexit "Mounting the MDT"
1680 shall_cleanup_mdt=true
1682 if [ "$dne_upgrade" != "no" ]; then
1683 local fs2mdsdev=$(mdsdevname 1_2)
1684 local fs2mdsvdev=$(mdsvdevname 1_2)
1686 echo "mkfs new MDT on ${fs2mdsdev}...."
1687 if [ $(facet_fstype mds1) == ldiskfs ]; then
1688 mkfsoptions="--mkfsoptions=\\\"-J size=8\\\""
1691 add fs2mds $(mkfs_opts mds2 $fs2mdsdev $fsname) --reformat \
1692 $mkfsoptions $fs2mdsdev $fs2mdsvdev > /dev/null || {
1693 error_noexit "Mkfs new MDT failed"
1697 $r $TUNEFS --dryrun $fs2mdsdev || {
1698 error_noexit "tunefs.lustre before mounting the MDT"
1702 echo "mount new MDT....$fs2mdsdev"
1703 $r mkdir -p $tmp/mnt/mdt1
1704 $r $MOUNT_CMD -o $mopts $fs2mdsdev $tmp/mnt/mdt1 || {
1705 error_noexit "mount mdt1 failed"
1709 $r $LCTL set_param -n mdt.${fsname}*.enable_remote_dir=1 ||
1710 error_noexit "enable remote dir create failed"
1712 shall_cleanup_mdt1=true
1715 uuid=$($r $LCTL get_param -n mdt.$fsname-MDT0000.uuid) || {
1716 error_noexit "Getting MDT UUID"
1719 if [ "$uuid" != $fsname-MDT0000_UUID ]; then
1720 error_noexit "Unexpected MDT UUID: \"$uuid\""
1724 $r $TUNEFS --dryrun $ost_dev || {
1725 error_noexit "tunefs.lustre before mounting the OST"
1728 if [ "$writeconf" ]; then
1729 mopts=mgsnode=$nid,$writeconf
1730 if [ $fstype == "ldiskfs" ]; then
1732 $r $TUNEFS --quota $ost_dev || {
1734 error_noexit "Enable ost quota feature"
1740 if [ $fstype == "ldiskfs" ]; then
1744 $r $MOUNT_CMD -o $mopts $ost_dev $tmp/mnt/ost || {
1745 error_noexit "Mounting the OST"
1748 shall_cleanup_ost=true
1750 uuid=$($r $LCTL get_param -n obdfilter.$fsname-OST0000.uuid) || {
1751 error_noexit "Getting OST UUID"
1754 if [ "$uuid" != $fsname-OST0000_UUID ]; then
1755 error_noexit "Unexpected OST UUID: \"$uuid\""
1759 $r $LCTL conf_param $fsname-OST0000.osc.max_dirty_mb=15 || {
1760 error_noexit "Setting \"max_dirty_mb\""
1763 $r $LCTL conf_param $fsname-OST0000.failover.node=$nid || {
1764 error_noexit "Setting OST \"failover.node\""
1767 $r $LCTL conf_param $fsname-MDT0000.mdc.max_rpcs_in_flight=9 || {
1768 error_noexit "Setting \"max_rpcs_in_flight\""
1771 $r $LCTL conf_param $fsname-MDT0000.failover.node=$nid || {
1772 error_noexit "Setting MDT \"failover.node\""
1775 $r $LCTL pool_new $fsname.interop || {
1776 error_noexit "Setting \"interop\""
1779 $r $LCTL conf_param $fsname-MDT0000.lov.stripesize=4M || {
1780 error_noexit "Setting \"lov.stripesize\""
1784 if [ "$ff_convert" != "no" -a $(facet_fstype ost1) == "ldiskfs" ]; then
1785 $r $LCTL lfsck_start -M $fsname-OST0000 || {
1786 error_noexit "Start OI scrub on OST0"
1790 # The oi_scrub should be on ost1, but for test_32(),
1791 # all on the SINGLEMDS.
1792 wait_update_facet $SINGLEMDS "$LCTL get_param -n \
1793 osd-ldiskfs.$fsname-OST0000.oi_scrub |
1794 awk '/^status/ { print \\\$2 }'" "completed" 30 || {
1795 error_noexit "Failed to get the expected 'completed'"
1799 local UPDATED=$($r $LCTL get_param -n \
1800 osd-ldiskfs.$fsname-OST0000.oi_scrub |
1801 awk '/^updated/ { print $2 }')
1802 [ $UPDATED -ge 1 ] || {
1803 error_noexit "Only $UPDATED objects have been converted"
1808 if [ "$dne_upgrade" != "no" ]; then
1809 $r $LCTL conf_param \
1810 $fsname-MDT0001.mdc.max_rpcs_in_flight=9 || {
1811 error_noexit "Setting MDT1 \"max_rpcs_in_flight\""
1814 $r $LCTL conf_param $fsname-MDT0001.failover.node=$nid || {
1815 error_noexit "Setting MDT1 \"failover.node\""
1818 $r $LCTL conf_param $fsname-MDT0001.lov.stripesize=4M || {
1819 error_noexit "Setting MDT1 \"lov.stripesize\""
1825 if [ "$writeconf" ]; then
1826 $MOUNT_CMD $nid:/$fsname $tmp/mnt/lustre || {
1827 error_noexit "Mounting the client"
1830 shall_cleanup_lustre=true
1831 $LCTL set_param debug="$PTLDEBUG"
1833 t32_verify_quota $node $fsname $tmp/mnt/lustre || {
1834 error_noexit "verify quota failed"
1838 if [ "$dne_upgrade" != "no" ]; then
1839 $LFS mkdir -i 1 $tmp/mnt/lustre/remote_dir || {
1840 error_noexit "set remote dir failed"
1844 pushd $tmp/mnt/lustre
1845 tar -cf - . --exclude=./remote_dir |
1846 tar -xvf - -C remote_dir 1>/dev/null || {
1847 error_noexit "cp to remote dir failed"
1853 dd if=/dev/zero of=$tmp/mnt/lustre/tmp_file bs=10k count=10 || {
1854 error_noexit "dd failed"
1857 rm -rf $tmp/mnt/lustre/tmp_file || {
1858 error_noexit "rm failed"
1862 if $r test -f $tmp/sha1sums; then
1863 # LU-2393 - do both sorts on same node to ensure locale
1865 $r cat $tmp/sha1sums | sort -k 2 >$tmp/sha1sums.orig
1866 if [ "$dne_upgrade" != "no" ]; then
1867 pushd $tmp/mnt/lustre/remote_dir
1869 pushd $tmp/mnt/lustre
1872 find ! -name .lustre -type f -exec sha1sum {} \; |
1873 sort -k 2 >$tmp/sha1sums || {
1874 error_noexit "sha1sum"
1878 if ! diff -ub $tmp/sha1sums.orig $tmp/sha1sums; then
1879 error_noexit "sha1sum verification failed"
1883 echo "sha1sum verification skipped"
1886 if [ "$dne_upgrade" != "no" ]; then
1887 rm -rf $tmp/mnt/lustre/remote_dir || {
1888 error_noexit "remove remote dir failed"
1893 if $r test -f $tmp/list; then
1895 # There is not a Test Framework API to copy files to or
1896 # from a remote node.
1898 # LU-2393 - do both sorts on same node to ensure locale
1900 $r cat $tmp/list | sort -k 6 >$tmp/list.orig
1901 pushd $tmp/mnt/lustre
1902 ls -Rni --time-style=+%s | sort -k 6 >$tmp/list || {
1908 # 32-bit and 64-bit clients use different algorithms to
1909 # convert FIDs into inode numbers. Hence, remove the inode
1910 # numbers from the lists, if the original list was created
1911 # on an architecture with different number of bits per
1914 if [ $(t32_bits_per_long $(uname -m)) != \
1915 $(t32_bits_per_long $img_arch) ]; then
1916 echo "Different number of bits per \"long\" from the disk image"
1917 for list in list.orig list; do
1918 sed -i -e 's/^[0-9]\+[ \t]\+//' $tmp/$list
1921 if ! diff -ub $tmp/list.orig $tmp/list; then
1922 error_noexit "list verification failed"
1926 echo "list verification skipped"
1930 # When adding new data verification tests, please check for
1931 # the presence of the required reference files first, like
1932 # the "sha1sums" and "list" tests above, to avoid the need to
1933 # regenerate every image for each test addition.
1936 nrpcs_orig=$($LCTL get_param \
1937 -n mdc.*MDT0000*.max_rpcs_in_flight) || {
1938 error_noexit "Getting \"max_rpcs_in_flight\""
1941 nrpcs=$((nrpcs_orig + 5))
1942 $r $LCTL conf_param $fsname-MDT0000.mdc.max_rpcs_in_flight=$nrpcs || {
1943 error_noexit "Changing \"max_rpcs_in_flight\""
1946 wait_update $HOSTNAME "$LCTL get_param \
1947 -n mdc.*MDT0000*.max_rpcs_in_flight" $nrpcs || {
1948 error_noexit "Verifying \"max_rpcs_in_flight\""
1952 umount $tmp/mnt/lustre || {
1953 error_noexit "Unmounting the client"
1956 shall_cleanup_lustre=false
1958 if [ "$dne_upgrade" != "no" ]; then
1959 $r umount -d $tmp/mnt/mdt1 || {
1960 error_noexit "Unmounting the MDT2"
1963 shall_cleanup_mdt1=false
1966 $r umount -d $tmp/mnt/mdt || {
1967 error_noexit "Unmounting the MDT"
1970 shall_cleanup_mdt=false
1972 $r umount -d $tmp/mnt/ost || {
1973 error_noexit "Unmounting the OST"
1976 shall_cleanup_ost=false
1978 t32_reload_modules $node || {
1979 error_noexit "Reloading modules"
1983 # mount a second time to make sure we didnt leave upgrade flag on
1984 $r $TUNEFS --dryrun $mdt_dev || {
1986 error_noexit "tunefs.lustre before remounting the MDT"
1990 mopts=exclude=$fsname-OST0000
1991 if [ $fstype == "ldiskfs" ]; then
1994 $r $MOUNT_CMD -o $mopts $mdt_dev $tmp/mnt/mdt || {
1995 error_noexit "Remounting the MDT"
1998 shall_cleanup_mdt=true
2008 for tarball in $tarballs; do
2009 t32_test $tarball || let "rc += $?"
2013 run_test 32a "Upgrade (not live)"
2021 for tarball in $tarballs; do
2022 t32_test $tarball writeconf || let "rc += $?"
2026 run_test 32b "Upgrade with writeconf"
2033 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
2035 for tarball in $tarballs; do
2036 dne_upgrade=yes t32_test $tarball writeconf || rc=$?
2040 run_test 32c "dne upgrade test"
2048 for tarball in $tarballs; do
2049 ff_convert=yes t32_test $tarball || rc=$?
2053 run_test 32d "convert ff test"
2055 test_33a() { # bug 12333, was test_33
2057 local FSNAME2=test-123
2058 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
2061 [ -n "$ost1_HOST" ] && fs2ost_HOST=$ost1_HOST
2063 if [ -z "$fs2ost_DEV" -o -z "$fs2mds_DEV" ]; then
2064 local dev=${SINGLEMDS}_dev
2065 local MDSDEV=${!dev}
2066 is_blkdev $SINGLEMDS $MDSDEV && \
2067 skip_env "mixed loopback and real device not working" && return
2070 local fs2mdsdev=$(mdsdevname 1_2)
2071 local fs2ostdev=$(ostdevname 1_2)
2072 local fs2mdsvdev=$(mdsvdevname 1_2)
2073 local fs2ostvdev=$(ostvdevname 1_2)
2075 if [ $(facet_fstype mds1) == ldiskfs ]; then
2076 mkfsoptions="--mkfsoptions=\\\"-J size=8\\\"" # See bug 17931.
2079 add fs2mds $(mkfs_opts mds1 ${fs2mdsdev}) --mgs --fsname=${FSNAME2} \
2080 --reformat $mkfsoptions $fs2mdsdev $fs2mdsvdev || exit 10
2081 add fs2ost $(mkfs_opts ost1 ${fs2ostdev}) --mgsnode=$MGSNID \
2082 --fsname=${FSNAME2} --index=8191 --reformat $fs2ostdev \
2083 $fs2ostvdev || exit 10
2085 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS && trap cleanup_fs2 EXIT INT
2086 start fs2ost $fs2ostdev $OST_MOUNT_OPTS
2087 do_facet $SINGLEMDS "$LCTL conf_param $FSNAME2.sys.timeout=200" || rc=1
2089 $MOUNT_CMD $MGSNID:/${FSNAME2} $MOUNT2 || rc=2
2092 cp /etc/hosts $MOUNT2/ || rc=3
2093 $LFS getstripe $MOUNT2/hosts
2098 cleanup_nocli || rc=6
2101 run_test 33a "Mount ost with a large index number"
2103 test_33b() { # was test_34
2106 do_facet client dd if=/dev/zero of=$MOUNT/24 bs=1024k count=1
2107 # Drop lock cancelation reply during umount
2108 #define OBD_FAIL_LDLM_CANCEL_NET 0x304
2109 do_facet client lctl set_param fail_loc=0x80000304
2110 #lctl set_param debug=-1
2111 umount_client $MOUNT
2114 run_test 33b "Drop cancel during umount"
2118 do_facet client "sh runmultiop_bg_pause $DIR/file O_c"
2119 manual_umount_client
2121 do_facet client killall -USR1 multiop
2122 if [ $rc -eq 0 ]; then
2123 error "umount not fail!"
2128 run_test 34a "umount with opened file should be fail"
2133 touch $DIR/$tfile || return 1
2134 stop_mds --force || return 2
2136 manual_umount_client --force
2138 if [ $rc -ne 0 ]; then
2139 error "mtab after failed umount - rc $rc"
2145 run_test 34b "force umount with failed mds should be normal"
2149 touch $DIR/$tfile || return 1
2150 stop_ost --force || return 2
2152 manual_umount_client --force
2154 if [ $rc -ne 0 ]; then
2155 error "mtab after failed umount - rc $rc"
2161 run_test 34c "force umount with failed ost should be normal"
2163 test_35a() { # bug 12459
2166 DBG_SAVE="`lctl get_param -n debug`"
2167 lctl set_param debug="ha"
2169 log "Set up a fake failnode for the MDS"
2171 local device=$(do_facet $SINGLEMDS "lctl get_param -n devices" |
2172 awk '($3 ~ "mdt" && $4 ~ "MDT") { print $4 }' | head -1)
2173 do_facet mgs "$LCTL conf_param \
2174 ${device}.failover.node=$(h2$NETTYPE $FAKENID)" || return 4
2176 log "Wait for RECONNECT_INTERVAL seconds (10s)"
2179 MSG="conf-sanity.sh test_35a `date +%F%kh%Mm%Ss`"
2182 log "Stopping the MDT: $device"
2183 stop_mdt 1 || return 5
2185 df $MOUNT > /dev/null 2>&1 &
2187 log "Restarting the MDT: $device"
2188 start_mdt 1 || return 6
2189 log "Wait for df ($DFPID) ... "
2192 lctl set_param debug="$DBG_SAVE"
2194 # retrieve from the log the first server that the client tried to
2195 # contact after the connection loss
2196 $LCTL dk $TMP/lustre-log-$TESTNAME.log
2197 NEXTCONN=`awk "/${MSG}/ {start = 1;}
2198 /import_select_connection.*$device-mdc.* using connection/ {
2200 if (\\\$NF ~ /$FAKENID/)
2206 }" $TMP/lustre-log-$TESTNAME.log`
2207 [ "$NEXTCONN" != "0" ] && log "The client didn't try to reconnect to the last active server (tried ${NEXTCONN} instead)" && return 7
2209 # remove nid settings
2210 writeconf_or_reformat
2212 run_test 35a "Reconnect to the last active server first"
2214 test_35b() { # bug 18674
2215 remote_mds || { skip "local MDS" && return 0; }
2219 $LCTL set_param debug="ha"
2221 MSG="conf-sanity.sh test_35b `date +%F%kh%Mm%Ss`"
2224 log "Set up a fake failnode for the MDS"
2226 local device=$(do_facet $SINGLEMDS "$LCTL get_param -n devices" |
2227 awk '($3 ~ "mdt" && $4 ~ "MDT") { print $4 }' | head -1)
2228 do_facet mgs "$LCTL conf_param \
2229 ${device}.failover.node=$(h2$NETTYPE $FAKENID)" || return 1
2231 local at_max_saved=0
2232 # adaptive timeouts may prevent seeing the issue
2233 if at_is_enabled; then
2234 at_max_saved=$(at_max_get mds)
2235 at_max_set 0 mds client
2238 mkdir -p $MOUNT/$tdir
2240 log "Injecting EBUSY on MDS"
2241 # Setting OBD_FAIL_MDS_RESEND=0x136
2242 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0x80000136" || return 2
2244 $LCTL set_param mdc.${FSNAME}*.stats=clear
2246 log "Creating a test file and stat it"
2247 touch $MOUNT/$tdir/$tfile
2248 stat $MOUNT/$tdir/$tfile
2250 log "Stop injecting EBUSY on MDS"
2251 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0" || return 3
2252 rm -f $MOUNT/$tdir/$tfile
2255 # restore adaptive timeout
2256 [ $at_max_saved -ne 0 ] && at_max_set $at_max_saved mds client
2258 $LCTL dk $TMP/lustre-log-$TESTNAME.log
2260 CONNCNT=`$LCTL get_param mdc.${FSNAME}*.stats | awk '/mds_connect/{print $2}'`
2262 # retrieve from the log if the client has ever tried to
2263 # contact the fake server after the loss of connection
2264 FAILCONN=`awk "BEGIN {ret = 0;}
2265 /import_select_connection.*${FSNAME}-MDT0000-mdc.* using connection/ {
2267 if (\\\$NF ~ /$FAKENID/) {
2272 END {print ret}" $TMP/lustre-log-$TESTNAME.log`
2274 [ "$FAILCONN" == "0" ] && \
2275 log "ERROR: The client reconnection has not been triggered" && \
2277 [ "$FAILCONN" == "2" ] && \
2278 log "ERROR: The client tried to reconnect to the failover server while the primary was busy" && \
2282 # When OBD_FAIL_MDS_RESEND is hit, we sleep for 2 * obd_timeout
2283 # Reconnects are supposed to be rate limited to one every 5s
2284 [ $CONNCNT -gt $((2 * $TIMEOUT / 5 + 1)) ] && \
2285 log "ERROR: Too many reconnects $CONNCNT" && \
2289 # remove nid settings
2290 writeconf_or_reformat
2292 run_test 35b "Continue reconnection retries, if the active server is busy"
2295 [ $OSTCOUNT -lt 2 ] && skip_env "skipping test for single OST" && return
2297 [ "$ost_HOST" = "`hostname`" -o "$ost1_HOST" = "`hostname`" ] || \
2298 { skip "remote OST" && return 0; }
2301 local FSNAME2=test1234
2302 local fs3ost_HOST=$ost_HOST
2303 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
2305 [ -n "$ost1_HOST" ] && fs2ost_HOST=$ost1_HOST && fs3ost_HOST=$ost1_HOST
2307 if [ -z "$fs2ost_DEV" -o -z "$fs2mds_DEV" -o -z "$fs3ost_DEV" ]; then
2308 is_blkdev $SINGLEMDS $MDSDEV && \
2309 skip_env "mixed loopback and real device not working" && return
2312 local fs2mdsdev=$(mdsdevname 1_2)
2313 local fs2ostdev=$(ostdevname 1_2)
2314 local fs3ostdev=$(ostdevname 2_2)
2315 local fs2mdsvdev=$(mdsvdevname 1_2)
2316 local fs2ostvdev=$(ostvdevname 1_2)
2317 local fs3ostvdev=$(ostvdevname 2_2)
2319 add fs2mds $(mkfs_opts mds1 ${fs2mdsdev}) --mgs --fsname=${FSNAME2} \
2320 --reformat $fs2mdsdev $fs2mdsvdev || exit 10
2321 # XXX after we support non 4K disk blocksize in ldiskfs, specify a
2322 # different one than the default value here.
2323 add fs2ost $(mkfs_opts ost1 ${fs2ostdev}) --mgsnode=$MGSNID \
2324 --fsname=${FSNAME2} --reformat $fs2ostdev $fs2ostvdev || exit 10
2325 add fs3ost $(mkfs_opts ost2 ${fs3ostdev}) --mgsnode=$MGSNID \
2326 --fsname=${FSNAME2} --reformat $fs3ostdev $fs3ostvdev || exit 10
2328 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS
2329 start fs2ost $fs2ostdev $OST_MOUNT_OPTS
2330 start fs3ost $fs3ostdev $OST_MOUNT_OPTS
2332 $MOUNT_CMD $MGSNID:/${FSNAME2} $MOUNT2 || return 1
2334 sleep 5 # until 11778 fixed
2336 dd if=/dev/zero of=$MOUNT2/$tfile bs=1M count=7 || return 2
2338 BKTOTAL=`lctl get_param -n obdfilter.*.kbytestotal | awk 'BEGIN{total=0}; {total+=$1}; END{print total}'`
2339 BKFREE=`lctl get_param -n obdfilter.*.kbytesfree | awk 'BEGIN{free=0}; {free+=$1}; END{print free}'`
2340 BKAVAIL=`lctl get_param -n obdfilter.*.kbytesavail | awk 'BEGIN{avail=0}; {avail+=$1}; END{print avail}'`
2341 STRING=`df -P $MOUNT2 | tail -n 1 | awk '{print $2","$3","$4}'`
2342 DFTOTAL=`echo $STRING | cut -d, -f1`
2343 DFUSED=`echo $STRING | cut -d, -f2`
2344 DFAVAIL=`echo $STRING | cut -d, -f3`
2345 DFFREE=$(($DFTOTAL - $DFUSED))
2347 ALLOWANCE=$((64 * $OSTCOUNT))
2349 if [ $DFTOTAL -lt $(($BKTOTAL - $ALLOWANCE)) ] ||
2350 [ $DFTOTAL -gt $(($BKTOTAL + $ALLOWANCE)) ] ; then
2351 echo "**** FAIL: df total($DFTOTAL) mismatch OST total($BKTOTAL)"
2354 if [ $DFFREE -lt $(($BKFREE - $ALLOWANCE)) ] ||
2355 [ $DFFREE -gt $(($BKFREE + $ALLOWANCE)) ] ; then
2356 echo "**** FAIL: df free($DFFREE) mismatch OST free($BKFREE)"
2359 if [ $DFAVAIL -lt $(($BKAVAIL - $ALLOWANCE)) ] ||
2360 [ $DFAVAIL -gt $(($BKAVAIL + $ALLOWANCE)) ] ; then
2361 echo "**** FAIL: df avail($DFAVAIL) mismatch OST avail($BKAVAIL)"
2366 stop fs3ost -f || return 200
2367 stop fs2ost -f || return 201
2368 stop fs2mds -f || return 202
2369 unload_modules_conf || return 203
2372 run_test 36 "df report consistency on OSTs with different block size"
2375 local mntpt=$(facet_mntpt $SINGLEMDS)
2376 local mdsdev=$(mdsdevname ${SINGLEMDS//mds/})
2377 local mdsdev_sym="$TMP/sym_mdt.img"
2378 local opts=$MDS_MOUNT_OPTS
2381 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
2382 skip "Currently only applicable to ldiskfs-based MDTs"
2386 echo "MDS : $mdsdev"
2387 echo "SYMLINK : $mdsdev_sym"
2388 do_facet $SINGLEMDS rm -f $mdsdev_sym
2390 do_facet $SINGLEMDS ln -s $mdsdev $mdsdev_sym
2392 echo "mount symlink device - $mdsdev_sym"
2394 if ! do_facet $SINGLEMDS test -b $mdsdev; then
2395 opts=$(csa_add "$opts" -o loop)
2397 mount_op=$(do_facet $SINGLEMDS mount -v -t lustre $opts \
2398 $mdsdev_sym $mntpt 2>&1)
2401 echo mount_op=$mount_op
2403 do_facet $SINGLEMDS "umount -d $mntpt && rm -f $mdsdev_sym"
2405 if $(echo $mount_op | grep -q "unable to set tunable"); then
2406 error "set tunables failed for symlink device"
2409 [ $rc -eq 0 ] || error "mount symlink $mdsdev_sym failed! rc=$rc"
2413 run_test 37 "verify set tunables works for symlink device"
2415 test_38() { # bug 14222
2416 local fstype=$(facet_fstype $SINGLEMDS)
2417 local mntpt=$(facet_mntpt $SINGLEMDS)
2422 local SRC="/etc /bin"
2423 local FILES=$(find $SRC -type f -mtime +1 | head -n $COUNT)
2424 log "copying $(echo $FILES | wc -w) files to $DIR/$tdir"
2426 tar cf - $FILES | tar xf - -C $DIR/$tdir ||
2427 error "copying $SRC to $DIR/$tdir"
2429 umount_client $MOUNT
2430 do_facet $SINGLEMDS "$LCTL get_param osp.*.prealloc_next_id"
2432 log "delete lov_objid file on MDS"
2434 mount_fstype $SINGLEMDS || error "mount MDS failed (1)"
2436 do_facet $SINGLEMDS "od -Ax -td8 $mntpt/lov_objid; rm $mntpt/lov_objid"
2438 unmount_fstype $SINGLEMDS || error "umount failed (1)"
2440 # check create in mds_lov_connect
2444 [ $V ] && log "verifying $DIR/$tdir/$f"
2445 diff -q $f $DIR/$tdir/$f || ERROR=y
2447 do_facet $SINGLEMDS "$LCTL get_param osp.*.prealloc_next_id"
2448 if [ "$ERROR" = "y" ]; then
2449 # check it's updates in sync
2450 umount_client $MOUNT
2452 mount_fstype $SIGNLEMDS
2453 do_facet $SINGLEMDS "od -Ax -td8 $mntpt/lov_objid"
2454 unmount_fstype $SINGLEMDS
2455 error "old and new files are different after connect" || true
2457 touch $DIR/$tdir/f2 || error "f2 file create failed"
2459 # check it's updates in sync
2460 umount_client $MOUNT
2463 mount_fstype $SINGLEMDS || error "mount MDS failed (3)"
2465 do_facet $SINGLEMDS "od -Ax -td8 $mntpt/lov_objid"
2466 do_facet $SINGLEMDS dd if=/dev/zero of=$mntpt/lov_objid.clear count=8
2468 unmount_fstype $SINGLEMDS || error "umount failed (3)"
2473 [ $V ] && log "verifying $DIR/$tdir/$f"
2474 diff -q $f $DIR/$tdir/$f || ERROR=y
2476 touch $DIR/$tdir/f3 || error "f3 file create failed"
2477 do_facet $SINGLEMDS "$LCTL get_param osp.*.prealloc_next_id"
2478 umount_client $MOUNT
2480 mount_fstype $SINGLEMDS || error "mount MDS failed (4)"
2481 do_facet $SINGLEMDS "od -Ax -td8 $mntpt/lov_objid"
2482 unmount_fstype $SINGLEMDS || error "umount failed (4)"
2484 [ "$ERROR" = "y" ] &&
2485 error "old and new files are different after sync" || true
2487 log "files compared the same"
2490 run_test 38 "MDS recreates missing lov_objid file from OST data"
2496 perl $SRCDIR/leak_finder.pl $TMP/debug 2>&1 | egrep '*** Leak:' &&
2497 error "memory leak detected" || true
2499 run_test 39 "leak_finder recognizes both LUSTRE and LNET malloc messages"
2501 test_40() { # bug 15759
2503 #define OBD_FAIL_TGT_TOOMANY_THREADS 0x706
2504 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0x80000706"
2508 run_test 40 "race during service thread startup"
2510 test_41a() { #bug 14134
2511 if [ $(facet_fstype $SINGLEMDS) == ldiskfs ] &&
2512 ! do_facet $SINGLEMDS test -b $(mdsdevname 1); then
2513 skip "Loop devices does not work with nosvc option"
2518 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
2520 start $SINGLEMDS $MDSDEV $MDS_MOUNT_OPTS -o nosvc -n
2521 start ost1 `ostdevname 1` $OST_MOUNT_OPTS
2522 start $SINGLEMDS $MDSDEV $MDS_MOUNT_OPTS -o nomgs,force
2524 mount_client $MOUNT || return 1
2527 echo "blah blah" > $MOUNT/$tfile
2530 umount_client $MOUNT
2531 stop ost1 -f || return 201
2532 stop_mds -f || return 202
2533 stop_mds -f || return 203
2534 unload_modules_conf || return 204
2537 run_test 41a "mount mds with --nosvc and --nomgs"
2540 if [ $(facet_fstype $SINGLEMDS) == ldiskfs ] &&
2541 ! do_facet $SINGLEMDS test -b $(mdsdevname 1); then
2542 skip "Loop devices does not work with nosvc option"
2546 ! combined_mgs_mds && skip "needs combined mgs device" && return 0
2550 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
2552 start $SINGLEMDS $MDSDEV $MDS_MOUNT_OPTS -o nosvc -n
2554 start $SINGLEMDS $MDSDEV $MDS_MOUNT_OPTS -o nomgs,force
2556 mount_client $MOUNT || return 1
2559 echo "blah blah" > $MOUNT/$tfile
2560 cat $MOUNT/$tfile || return 200
2562 umount_client $MOUNT
2563 stop_ost || return 201
2564 stop_mds -f || return 202
2565 stop_mds -f || return 203
2568 run_test 41b "mount mds with --nosvc and --nomgs on first mount"
2571 local server_version=$(lustre_version_code $SINGLEMDS)
2573 [[ $server_version -ge $(version_code 2.6.52) ]] ||
2574 [[ $server_version -ge $(version_code 2.5.26) &&
2575 $server_version -lt $(version_code 2.5.50) ]] ||
2576 [[ $server_version -ge $(version_code 2.5.4) &&
2577 $server_version -lt $(version_code 2.5.11) ]] ||
2578 { skip "Need MDS version 2.5.4+ or 2.5.26+ or 2.6.52+"; return; }
2581 # MDT concurent start
2582 #define OBD_FAIL_TGT_DELAY_CONNECT 0x703
2583 do_facet $SINGLEMDS "lctl set_param fail_loc=0x703"
2584 start mds1 $(mdsdevname 1) $MDS_MOUNT_OPTS &
2587 do_facet $SINGLEMDS "lctl set_param fail_loc=0x0"
2588 start mds1 $(mdsdevname 1) $MDS_MOUNT_OPTS &
2594 if [ $rc == 0 ] && [ $rc2 == 114 ]; then
2595 echo "1st MDT start succeed"
2596 echo "2nd MDT start failed with EALREADY"
2597 elif [ $rc2 == 0 ] && [ $rc == 114 ]; then
2598 echo "1st MDT start failed with EALREADY"
2599 echo "2nd MDT start succeed"
2602 error "unexpected concurent MDT mounts result, rc=$rc rc2=$rc2"
2605 # OST concurent start
2606 #define OBD_FAIL_TGT_DELAY_CONNECT 0x703
2607 do_facet ost1 "lctl set_param fail_loc=0x703"
2608 start ost1 $(ostdevname 1) $OST_MOUNT_OPTS &
2611 do_facet ost1 "lctl set_param fail_loc=0x0"
2612 start ost1 $(ostdevname 1) $OST_MOUNT_OPTS &
2618 if [ $rc == 0 ] && [ $rc2 == 114 ]; then
2619 echo "1st OST start succeed"
2620 echo "2nd OST start failed with EALREADY"
2621 elif [ $rc2 == 0 ] && [ $rc == 114 ]; then
2622 echo "1st OST start failed with EALREADY"
2623 echo "2nd OST start succeed"
2627 error "unexpected concurent OST mounts result, rc=$rc rc2=$rc2"
2633 # verify everything ok
2638 error "MDT(s) start failed"
2646 error "OST(s) start failed"
2654 error "client start failed"
2661 error "client mount failed"
2665 run_test 41c "concurent mounts of MDT/OST should all fail but one"
2667 test_42() { #bug 14693
2669 check_mount || error "client was not mounted"
2671 do_facet mgs $LCTL conf_param $FSNAME.llite.some_wrong_param=10
2672 umount_client $MOUNT ||
2673 error "unmounting client failed with invalid llite param"
2674 mount_client $MOUNT ||
2675 error "mounting client failed with invalid llite param"
2677 do_facet mgs $LCTL conf_param $FSNAME.sys.some_wrong_param=20
2678 cleanup || error "stopping $FSNAME failed with invalid sys param"
2681 check_mount || "client was not mounted with invalid sys param"
2682 cleanup || error "stopping $FSNAME failed with invalid sys param"
2685 run_test 42 "allow client/server mount/unmount with invalid config param"
2688 [[ $(lustre_version_code mgs) -ge $(version_code 2.5.58) ]] ||
2689 { skip "Need MDS version at least 2.5.58" && return 0; }
2690 [ $UID -ne 0 -o $RUNAS_ID -eq 0 ] && skip_env "run as root"
2693 USER1=$(cat /etc/passwd | grep :$ID1:$ID1: | cut -d: -f1)
2694 [ -z "$USER1" ] && skip_env "missing user with uid=$ID1 gid=$ID1" &&
2698 chmod ugo+x $DIR || error "chmod 0 failed"
2699 set_conf_param_and_check mds \
2700 "lctl get_param -n mdt.$FSNAME-MDT0000.root_squash" \
2701 "$FSNAME.mdt.root_squash" \
2703 wait_update $HOSTNAME \
2704 "lctl get_param -n llite.${FSNAME}*.root_squash" \
2706 error "check llite root_squash failed!"
2707 set_conf_param_and_check mds \
2708 "lctl get_param -n mdt.$FSNAME-MDT0000.nosquash_nids" \
2709 "$FSNAME.mdt.nosquash_nids" \
2711 wait_update $HOSTNAME \
2712 "lctl get_param -n llite.${FSNAME}*.nosquash_nids" \
2714 error "check llite nosquash_nids failed!"
2717 # create set of test files
2719 echo "111" > $DIR/$tfile-userfile || error "write 1 failed"
2720 chmod go-rw $DIR/$tfile-userfile || error "chmod 1 failed"
2721 chown $RUNAS_ID.$RUNAS_ID $DIR/$tfile-userfile || error "chown failed"
2723 echo "222" > $DIR/$tfile-rootfile || error "write 2 failed"
2724 chmod go-rw $DIR/$tfile-rootfile || error "chmod 2 faield"
2726 mkdir $DIR/$tdir-rootdir -p || error "mkdir failed"
2727 chmod go-rwx $DIR/$tdir-rootdir || error "chmod 3 failed"
2728 touch $DIR/$tdir-rootdir/tfile-1 || error "touch failed"
2730 echo "777" > $DIR/$tfile-user1file || error "write 7 failed"
2731 chmod go-rw $DIR/$tfile-user1file || error "chmod 7 failed"
2732 chown $ID1.$ID1 $DIR/$tfile-user1file || error "chown failed"
2735 # check root_squash:
2736 # set root squash UID:GID to RUNAS_ID
2737 # root should be able to access only files owned by RUNAS_ID
2739 set_conf_param_and_check mds \
2740 "lctl get_param -n mdt.$FSNAME-MDT0000.root_squash" \
2741 "$FSNAME.mdt.root_squash" \
2742 "$RUNAS_ID:$RUNAS_ID"
2743 wait_update $HOSTNAME \
2744 "lctl get_param -n llite.${FSNAME}*.root_squash" \
2745 "$RUNAS_ID:$RUNAS_ID" ||
2746 error "check llite root_squash failed!"
2748 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-userfile)
2749 dd if=$DIR/$tfile-userfile 1>/dev/null 2>/dev/null || \
2750 error "$ST: root read permission is denied"
2751 echo "$ST: root read permission is granted - ok"
2754 dd conv=notrunc of=$DIR/$tfile-userfile 1>/dev/null 2>/dev/null || \
2755 error "$ST: root write permission is denied"
2756 echo "$ST: root write permission is granted - ok"
2758 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-rootfile)
2759 dd if=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null && \
2760 error "$ST: root read permission is granted"
2761 echo "$ST: root read permission is denied - ok"
2764 dd conv=notrunc of=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null && \
2765 error "$ST: root write permission is granted"
2766 echo "$ST: root write permission is denied - ok"
2768 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tdir-rootdir)
2769 rm $DIR/$tdir-rootdir/tfile-1 1>/dev/null 2>/dev/null && \
2770 error "$ST: root unlink permission is granted"
2771 echo "$ST: root unlink permission is denied - ok"
2773 touch $DIR/tdir-rootdir/tfile-2 1>/dev/null 2>/dev/null && \
2774 error "$ST: root create permission is granted"
2775 echo "$ST: root create permission is denied - ok"
2779 # check root_squash is enforced independently
2780 # of client cache content
2782 # access file by USER1, keep access open
2783 # root should be denied access to user file
2785 runas -u $ID1 tail -f $DIR/$tfile-user1file 1>/dev/null 2>&1 &
2789 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-user1file)
2790 dd if=$DIR/$tfile-user1file 1>/dev/null 2>&1 &&
2791 { kill $pid; error "$ST: root read permission is granted"; }
2792 echo "$ST: root read permission is denied - ok"
2795 dd conv=notrunc of=$DIR/$tfile-user1file 1>/dev/null 2>&1 &&
2796 { kill $pid; error "$ST: root write permission is granted"; }
2797 echo "$ST: root write permission is denied - ok"
2803 # check nosquash_nids:
2804 # put client's NID into nosquash_nids list,
2805 # root should be able to access root file after that
2807 local NIDLIST=$(lctl list_nids all | tr '\n' ' ')
2808 NIDLIST="2@elan $NIDLIST 192.168.0.[2,10]@tcp"
2809 NIDLIST=$(echo $NIDLIST | tr -s ' ' ' ')
2810 set_conf_param_and_check mds \
2811 "lctl get_param -n mdt.$FSNAME-MDT0000.nosquash_nids" \
2812 "$FSNAME-MDTall.mdt.nosquash_nids" \
2814 wait_update $HOSTNAME \
2815 "lctl get_param -n llite.${FSNAME}*.nosquash_nids" \
2817 error "check llite nosquash_nids failed!"
2819 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-rootfile)
2820 dd if=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null || \
2821 error "$ST: root read permission is denied"
2822 echo "$ST: root read permission is granted - ok"
2825 dd conv=notrunc of=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null || \
2826 error "$ST: root write permission is denied"
2827 echo "$ST: root write permission is granted - ok"
2829 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tdir-rootdir)
2830 rm $DIR/$tdir-rootdir/tfile-1 || \
2831 error "$ST: root unlink permission is denied"
2832 echo "$ST: root unlink permission is granted - ok"
2833 touch $DIR/$tdir-rootdir/tfile-2 || \
2834 error "$ST: root create permission is denied"
2835 echo "$ST: root create permission is granted - ok"
2839 run_test 43 "check root_squash and nosquash_nids"
2841 umount_client $MOUNT
2846 check_mount || return 2
2847 UUID=$($LCTL get_param llite.${FSNAME}*.uuid | cut -d= -f2)
2849 UUIDS=$(do_facet $SINGLEMDS "$LCTL get_param mdt.${FSNAME}*.exports.*.uuid")
2850 for VAL in $UUIDS; do
2851 NID=$(echo $VAL | cut -d= -f1)
2852 CLUUID=$(echo $VAL | cut -d= -f2)
2853 [ "$UUID" = "$CLUUID" ] && STATS_FOUND=yes && break
2855 [ "$STATS_FOUND" = "no" ] && error "stats not found for client"
2859 run_test 44 "mounted client proc entry exists"
2863 check_mount || return 2
2868 #define OBD_FAIL_PTLRPC_LONG_UNLINK 0x50f
2869 do_facet client "lctl set_param fail_loc=0x50f"
2872 manual_umount_client --force || return 3
2873 do_facet client "lctl set_param fail_loc=0x0"
2875 mount_client $MOUNT || return 4
2879 run_test 45 "long unlink handling in ptlrpcd"
2886 umount_client $MOUNT2 || rc=$?
2887 umount_client $MOUNT || rc=$?
2888 while [ $count -gt 0 ]; do
2889 stop ost${count} -f || rc=$?
2893 cleanup_nocli || rc=$?
2894 #writeconf to remove all ost2 traces for subsequent tests
2895 writeconf_or_reformat
2900 echo "Testing with $OSTCOUNT OSTs"
2902 start_mds || return 1
2903 #first client should see only one ost
2904 start_ost || return 2
2905 wait_osc_import_state mds ost FULL
2907 mount_client $MOUNT || return 3
2908 trap "cleanup_46a $OSTCOUNT" EXIT ERR
2911 for (( i=2; i<=$OSTCOUNT; i++ )); do
2912 start ost$i `ostdevname $i` $OST_MOUNT_OPTS || return $((i+2))
2915 # wait until osts in sync
2916 for (( i=2; i<=$OSTCOUNT; i++ )); do
2917 wait_osc_import_state mds ost$i FULL
2918 wait_osc_import_state client ost$i FULL
2921 #second client see all ost's
2923 mount_client $MOUNT2 || return 8
2924 $LFS setstripe -c -1 $MOUNT2 || return 9
2925 $LFS getstripe $MOUNT2 || return 10
2927 echo "ok" > $MOUNT2/widestripe
2928 $LFS getstripe $MOUNT2/widestripe || return 11
2929 # fill acl buffer for avoid expand lsm to them
2930 awk -F : '{if (FNR < 25) { print "u:"$1":rwx" }}' /etc/passwd | while read acl; do
2931 setfacl -m $acl $MOUNT2/widestripe
2935 stat $MOUNT/widestripe || return 12
2937 cleanup_46a $OSTCOUNT || { echo "cleanup_46a failed!" && return 13; }
2940 run_test 46a "handle ost additional - wide striped file"
2945 check_mount || return 2
2946 $LCTL set_param ldlm.namespaces.$FSNAME-*-*-*.lru_size=100
2950 for ns in $($LCTL get_param ldlm.namespaces.$FSNAME-*-*-*.lru_size); do
2951 if echo $ns | grep "MDT[[:digit:]]*"; then
2954 lrs=$(echo $ns | sed 's/.*lru_size=//')
2955 lru_size[count]=$lrs
2960 facet_failover $SINGLEMDS
2961 client_up || return 3
2964 for ns in $($LCTL get_param ldlm.namespaces.$FSNAME-*-*-*.lru_size); do
2965 if echo $ns | grep "MDT[[:digit:]]*"; then
2968 lrs=$(echo $ns | sed 's/.*lru_size=//')
2969 if ! test "$lrs" -eq "${lru_size[count]}"; then
2970 n=$(echo $ns | sed -e 's/ldlm.namespaces.//' -e 's/.lru_size=.*//')
2971 error "$n has lost lru_size: $lrs vs. ${lru_size[count]}"
2979 run_test 47 "server restart does not make client loss lru_resize settings"
2984 # reformat after this test is needed - if test will failed
2985 # we will have unkillable file at FS
2989 test_48() { # bug 17636
2992 check_mount || return 2
2994 $LFS setstripe -c -1 $MOUNT || return 9
2995 $LFS getstripe $MOUNT || return 10
2997 echo "ok" > $MOUNT/widestripe
2998 $LFS getstripe $MOUNT/widestripe || return 11
3000 trap cleanup_48 EXIT ERR
3002 # fill acl buffer for avoid expand lsm to them
3003 getent passwd | awk -F : '{ print "u:"$1":rwx" }' | while read acl; do
3004 setfacl -m $acl $MOUNT/widestripe
3007 stat $MOUNT/widestripe || return 12
3012 run_test 48 "too many acls on file"
3014 # check PARAM_SYS_LDLM_TIMEOUT option of MKFS.LUSTRE
3015 test_49a() { # bug 17710
3016 local timeout_orig=$TIMEOUT
3017 local ldlm_timeout_orig=$LDLM_TIMEOUT
3018 local LOCAL_TIMEOUT=20
3020 LDLM_TIMEOUT=$LOCAL_TIMEOUT
3021 TIMEOUT=$LOCAL_TIMEOUT
3025 check_mount || error "client mount failed"
3027 echo "check ldlm_timout..."
3028 local LDLM_MDS="$(do_facet $SINGLEMDS lctl get_param -n ldlm_timeout)"
3029 local LDLM_OST1="$(do_facet ost1 lctl get_param -n ldlm_timeout)"
3030 local LDLM_CLIENT="$(do_facet client lctl get_param -n ldlm_timeout)"
3032 if [ $LDLM_MDS -ne $LDLM_OST1 -o $LDLM_MDS -ne $LDLM_CLIENT ]; then
3033 error "Different LDLM_TIMEOUT:$LDLM_MDS $LDLM_OST1 $LDLM_CLIENT"
3036 if [ $LDLM_MDS -ne $((LOCAL_TIMEOUT / 3)) ]; then
3037 error "LDLM_TIMEOUT($LDLM_MDS) is not $((LOCAL_TIMEOUT / 3))"
3040 umount_client $MOUNT
3041 stop_ost || error "problem stopping OSS"
3042 stop_mds || error "problem stopping MDS"
3044 LDLM_TIMEOUT=$ldlm_timeout_orig
3045 TIMEOUT=$timeout_orig
3047 run_test 49a "check PARAM_SYS_LDLM_TIMEOUT option of mkfs.lustre"
3049 test_49b() { # bug 17710
3050 local timeout_orig=$TIMEOUT
3051 local ldlm_timeout_orig=$LDLM_TIMEOUT
3052 local LOCAL_TIMEOUT=20
3054 LDLM_TIMEOUT=$((LOCAL_TIMEOUT - 1))
3055 TIMEOUT=$LOCAL_TIMEOUT
3059 check_mount || error "client mount failed"
3061 local LDLM_MDS="$(do_facet $SINGLEMDS lctl get_param -n ldlm_timeout)"
3062 local LDLM_OST1="$(do_facet ost1 lctl get_param -n ldlm_timeout)"
3063 local LDLM_CLIENT="$(do_facet client lctl get_param -n ldlm_timeout)"
3065 if [ $LDLM_MDS -ne $LDLM_OST1 -o $LDLM_MDS -ne $LDLM_CLIENT ]; then
3066 error "Different LDLM_TIMEOUT:$LDLM_MDS $LDLM_OST1 $LDLM_CLIENT"
3069 if [ $LDLM_MDS -ne $((LOCAL_TIMEOUT - 1)) ]; then
3070 error "LDLM_TIMEOUT($LDLM_MDS) is not $((LOCAL_TIMEOUT - 1))"
3073 cleanup || error "cleanup failed"
3075 LDLM_TIMEOUT=$ldlm_timeout_orig
3076 TIMEOUT=$timeout_orig
3078 run_test 49b "check PARAM_SYS_LDLM_TIMEOUT option of mkfs.lustre"
3081 # Test both statfs and lfs df and fail if either one fails
3082 multiop_bg_pause $1 f_
3085 killall -USR1 multiop
3086 [ $RC1 -ne 0 ] && log "lazystatfs multiop failed"
3087 wait $PID || { RC1=$?; log "multiop return error "; }
3094 if [ $RC2 -eq 0 ]; then
3096 log "lazystatfs df failed"
3100 [[ $RC1 -ne 0 || $RC2 -eq 0 ]] && RC=1
3106 lctl set_param llite.$FSNAME-*.lazystatfs=1
3109 lazystatfs $MOUNT || error "lazystatfs failed but no down servers"
3111 cleanup || return $?
3113 run_test 50a "lazystatfs all servers available =========================="
3117 lctl set_param llite.$FSNAME-*.lazystatfs=1
3120 # Wait for client to detect down OST
3121 stop_ost || error "Unable to stop OST1"
3122 wait_osc_import_state mds ost DISCONN
3124 lazystatfs $MOUNT || error "lazystatfs should don't have returned EIO"
3126 umount_client $MOUNT || error "Unable to unmount client"
3127 stop_mds || error "Unable to stop MDS"
3129 run_test 50b "lazystatfs all servers down =========================="
3132 start_mds || error "Unable to start MDS"
3133 start_ost || error "Unable to start OST1"
3134 start_ost2 || error "Unable to start OST2"
3135 mount_client $MOUNT || error "Unable to mount client"
3136 lctl set_param llite.$FSNAME-*.lazystatfs=1
3139 # Wait for client to detect down OST
3140 stop_ost || error "Unable to stop OST1"
3141 wait_osc_import_state mds ost DISCONN
3142 lazystatfs $MOUNT || error "lazystatfs failed with one down server"
3144 umount_client $MOUNT || error "Unable to unmount client"
3145 stop_ost2 || error "Unable to stop OST2"
3146 stop_mds || error "Unable to stop MDS"
3147 #writeconf to remove all ost2 traces for subsequent tests
3148 writeconf_or_reformat
3150 run_test 50c "lazystatfs one server down =========================="
3153 start_mds || error "Unable to start MDS"
3154 start_ost || error "Unable to start OST1"
3155 start_ost2 || error "Unable to start OST2"
3156 mount_client $MOUNT || error "Unable to mount client"
3157 lctl set_param llite.$FSNAME-*.lazystatfs=1
3160 # Issue the statfs during the window where the client still
3161 # belives the OST to be available but it is in fact down.
3162 # No failure just a statfs which hangs for a timeout interval.
3163 stop_ost || error "Unable to stop OST1"
3164 lazystatfs $MOUNT || error "lazystatfs failed with one down server"
3166 umount_client $MOUNT || error "Unable to unmount client"
3167 stop_ost2 || error "Unable to stop OST2"
3168 stop_mds || error "Unable to stop MDS"
3169 #writeconf to remove all ost2 traces for subsequent tests
3170 writeconf_or_reformat
3172 run_test 50d "lazystatfs client/server conn race =========================="
3179 start_mds || return 1
3180 #first client should see only one ost
3181 start_ost || return 2
3182 wait_osc_import_state mds ost FULL
3184 # Wait for client to detect down OST
3185 stop_ost || error "Unable to stop OST1"
3186 wait_osc_import_state mds ost DISCONN
3188 mount_client $MOUNT || error "Unable to mount client"
3189 lctl set_param llite.$FSNAME-*.lazystatfs=0
3191 multiop_bg_pause $MOUNT _f
3195 if [ $RC1 -ne 0 ]; then
3196 log "multiop failed $RC1"
3199 sleep $(( $TIMEOUT+1 ))
3201 [ $? -ne 0 ] && error "process isn't sleep"
3202 start_ost || error "Unable to start OST1"
3203 wait $pid || error "statfs failed"
3206 umount_client $MOUNT || error "Unable to unmount client"
3207 stop_ost || error "Unable to stop OST1"
3208 stop_mds || error "Unable to stop MDS"
3210 run_test 50e "normal statfs all servers down =========================="
3215 CONN_PROC="osc.$FSNAME-OST0001-osc-[M]*.ost_server_uuid"
3217 start_mds || error "Unable to start mds"
3218 #first client should see only one ost
3219 start_ost || error "Unable to start OST1"
3220 wait_osc_import_state mds ost FULL
3222 start_ost2 || error "Unable to start OST2"
3223 wait_osc_import_state mds ost2 FULL
3225 # Wait for client to detect down OST
3226 stop_ost2 || error "Unable to stop OST2"
3228 wait_osc_import_state mds ost2 DISCONN
3229 mount_client $MOUNT || error "Unable to mount client"
3230 lctl set_param llite.$FSNAME-*.lazystatfs=0
3232 multiop_bg_pause $MOUNT _f
3236 if [ $RC1 -ne 0 ]; then
3237 log "lazystatfs multiop failed $RC1"
3240 sleep $(( $TIMEOUT+1 ))
3242 [ $? -ne 0 ] && error "process isn't sleep"
3243 start_ost2 || error "Unable to start OST2"
3244 wait $pid || error "statfs failed"
3245 stop_ost2 || error "Unable to stop OST2"
3248 umount_client $MOUNT || error "Unable to unmount client"
3249 stop_ost || error "Unable to stop OST1"
3250 stop_mds || error "Unable to stop MDS"
3251 #writeconf to remove all ost2 traces for subsequent tests
3252 writeconf_or_reformat
3254 run_test 50f "normal statfs one server in down =========================="
3257 [ "$OSTCOUNT" -lt "2" ] && skip_env "$OSTCOUNT < 2, skipping" && return
3259 start_ost2 || error "Unable to start OST2"
3260 wait_osc_import_state mds ost2 FULL
3261 wait_osc_import_state client ost2 FULL
3263 local PARAM="${FSNAME}-OST0001.osc.active"
3265 $LFS setstripe -c -1 $DIR/$tfile || error "Unable to lfs setstripe"
3266 do_facet mgs $LCTL conf_param $PARAM=0 || error "Unable to deactivate OST"
3268 umount_client $MOUNT || error "Unable to unmount client"
3269 mount_client $MOUNT || error "Unable to mount client"
3270 # This df should not cause a panic
3273 do_facet mgs $LCTL conf_param $PARAM=1 || error "Unable to activate OST"
3275 umount_client $MOUNT || error "Unable to unmount client"
3276 stop_ost2 || error "Unable to stop OST2"
3277 stop_ost || error "Unable to stop OST1"
3278 stop_mds || error "Unable to stop MDS"
3279 #writeconf to remove all ost2 traces for subsequent tests
3280 writeconf_or_reformat
3282 run_test 50g "deactivated OST should not cause panic====================="
3286 # prepare MDT/OST, make OSC inactive for OST1
3287 [ "$OSTCOUNT" -lt "2" ] && skip_env "$OSTCOUNT < 2, skipping" && return
3289 [ $(facet_fstype ost1) == zfs ] && import_zpool ost1
3290 do_facet ost1 "$TUNEFS --param osc.active=0 `ostdevname 1`" ||
3291 error "tunefs OST1 failed"
3292 start_mds || error "Unable to start MDT"
3293 start_ost || error "Unable to start OST1"
3294 start_ost2 || error "Unable to start OST2"
3295 mount_client $MOUNT || error "client start failed"
3299 # activatate OSC for OST1
3300 local TEST="$LCTL get_param -n osc.${FSNAME}-OST0000-osc-[!M]*.active"
3301 set_conf_param_and_check client \
3302 "$TEST" "${FSNAME}-OST0000.osc.active" 1 ||
3303 error "Unable to activate OST1"
3305 mkdir -p $DIR/$tdir/2
3306 $LFS setstripe -c -1 -i 0 $DIR/$tdir/2
3307 sleep 1 && echo "create a file after OST1 is activated"
3309 createmany -o $DIR/$tdir/2/$tfile-%d 1
3311 # check OSC import is working
3312 stat $DIR/$tdir/2/* >/dev/null 2>&1 ||
3313 error "some OSC imports are still not connected"
3316 umount_client $MOUNT || error "Unable to umount client"
3317 stop_ost2 || error "Unable to stop OST2"
3320 run_test 50h "LU-642: activate deactivated OST ==="
3323 local LOCAL_TIMEOUT=20
3327 check_mount || return 1
3330 $LFS setstripe -c -1 $MOUNT/d1
3331 #define OBD_FAIL_MDS_REINT_DELAY 0x142
3332 do_facet $SINGLEMDS "lctl set_param fail_loc=0x142"
3333 touch $MOUNT/d1/f1 &
3336 start_ost2 || return 2
3338 stop_ost2 || return 3
3340 #writeconf to remove all ost2 traces for subsequent tests
3341 writeconf_or_reformat
3343 run_test 51 "Verify that mdt_reint handles RMF_MDT_MD correctly when an OST is added"
3352 do_node $node mkdir -p $dest
3353 [ $? -eq 0 ] || { error "Unable to create directory"; return 1; }
3355 do_node $node 'tar cf - '$@' | tar xf - -C '$dest';
3356 [ \"\${PIPESTATUS[*]}\" = \"0 0\" ] || exit 1'
3357 [ $? -eq 0 ] || { error "Unable to tar files"; return 2; }
3359 do_node $node 'getfattr -d -m "[a-z]*\\." '$@' > '$xattrs
3360 [ $? -eq 0 ] || { error "Unable to read xattrs"; return 3; }
3370 local backup2=${TMP}/backup2
3372 do_node $node mkdir -p $backup2
3373 [ $? -eq 0 ] || { error "Unable to create directory"; return 1; }
3375 do_node $node 'tar cf - '$@' | tar xf - -C '$backup2';
3376 [ \"\${PIPESTATUS[*]}\" = \"0 0\" ] || exit 1'
3377 [ $? -eq 0 ] || { error "Unable to tar files to diff"; return 2; }
3379 do_node $node "diff -rq $backup $backup2"
3380 [ $? -eq 0 ] || { error "contents differ"; return 3; }
3382 local xattrs2=${TMP}/xattrs2
3383 do_node $node 'getfattr -d -m "[a-z]*\\." '$@' > '$xattrs2
3384 [ $? -eq 0 ] || { error "Unable to read xattrs to diff"; return 4; }
3386 do_node $node "diff $xattrs $xattrs2"
3387 [ $? -eq 0 ] || { error "xattrs differ"; return 5; }
3389 do_node $node "rm -rf $backup2 $xattrs2"
3390 [ $? -eq 0 ] || { error "Unable to delete temporary files"; return 6; }
3394 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
3395 skip "Only applicable to ldiskfs-based MDTs"
3400 [ $? -eq 0 ] || { error "Unable to start MDS"; return 1; }
3402 [ $? -eq 0 ] || { error "Unable to start OST1"; return 2; }
3404 [ $? -eq 0 ] || { error "Unable to mount client"; return 3; }
3407 local ost1mnt=$(facet_mntpt ost1)
3408 local ost1node=$(facet_active_host ost1)
3409 local ost1tmp=$TMP/conf52
3413 [ $? -eq 0 ] || { error "Unable to create tdir"; return 4; }
3414 touch $TMP/modified_first
3415 [ $? -eq 0 ] || { error "Unable to create temporary file"; return 5; }
3416 local mtime=$(stat -c %Y $TMP/modified_first)
3417 do_node $ost1node "mkdir -p $ost1tmp && touch -m -d @$mtime $ost1tmp/modified_first"
3419 [ $? -eq 0 ] || { error "Unable to create temporary file"; return 6; }
3422 $LFS setstripe -c -1 -S 1M $DIR/$tdir
3423 [ $? -eq 0 ] || { error "lfs setstripe failed"; return 7; }
3425 for (( i=0; i < nrfiles; i++ )); do
3426 multiop $DIR/$tdir/$tfile-$i Ow1048576w1048576w524288c
3427 [ $? -eq 0 ] || { error "multiop failed"; return 8; }
3433 echo backup files to $TMP/files
3434 local files=$(find $DIR/$tdir -type f -newer $TMP/modified_first)
3435 copy_files_xattrs `hostname` $TMP/files $TMP/file_xattrs $files
3436 [ $? -eq 0 ] || { error "Unable to copy files"; return 9; }
3438 umount_client $MOUNT
3439 [ $? -eq 0 ] || { error "Unable to umount client"; return 10; }
3441 [ $? -eq 0 ] || { error "Unable to stop ost1"; return 11; }
3443 echo mount ost1 as ldiskfs
3444 do_node $ost1node mkdir -p $ost1mnt
3445 [ $? -eq 0 ] || { error "Unable to create $ost1mnt"; return 23; }
3446 if ! do_node $ost1node test -b $ost1_dev; then
3449 do_node $ost1node mount -t $(facet_fstype ost1) $loop $ost1_dev \
3451 [ $? -eq 0 ] || { error "Unable to mount ost1 as ldiskfs"; return 12; }
3454 echo backup objects to $ost1tmp/objects
3455 local objects=$(do_node $ost1node 'find '$ost1mnt'/O/[0-9]* -type f'\
3456 '-size +0 -newer '$ost1tmp'/modified_first -regex ".*\/[0-9]+"')
3457 copy_files_xattrs $ost1node $ost1tmp/objects $ost1tmp/object_xattrs \
3459 [ $? -eq 0 ] || { error "Unable to copy objects"; return 13; }
3461 # move objects to lost+found
3462 do_node $ost1node 'mv '$objects' '${ost1mnt}'/lost+found'
3463 [ $? -eq 0 ] || { error "Unable to move objects"; return 14; }
3465 # recover objects dry-run
3466 echo "ll_recover_lost_found_objs dry_run"
3467 do_node $ost1node "ll_recover_lost_found_objs -n -d $ost1mnt/O"
3468 [ $? -eq 0 ] || { error "ll_recover_lost_found_objs failed"; return 15; }
3471 echo "ll_recover_lost_found_objs fix run"
3472 do_node $ost1node "ll_recover_lost_found_objs -d $ost1mnt/lost+found"
3473 [ $? -eq 0 ] || { error "ll_recover_lost_found_objs failed"; return 15; }
3475 # compare restored objects against saved ones
3476 diff_files_xattrs $ost1node $ost1tmp/objects $ost1tmp/object_xattrs $objects
3477 [ $? -eq 0 ] || { error "Unable to diff objects"; return 16; }
3479 do_node $ost1node "umount $ost1mnt"
3480 [ $? -eq 0 ] || { error "Unable to umount ost1 as ldiskfs"; return 17; }
3483 [ $? -eq 0 ] || { error "Unable to start ost1"; return 18; }
3485 [ $? -eq 0 ] || { error "Unable to mount client"; return 19; }
3488 diff_files_xattrs `hostname` $TMP/files $TMP/file_xattrs $files
3489 [ $? -eq 0 ] || { error "Unable to diff files"; return 20; }
3491 rm -rf $TMP/files $TMP/file_xattrs
3492 [ $? -eq 0 ] || { error "Unable to delete temporary files"; return 21; }
3493 do_node $ost1node "rm -rf $ost1tmp"
3494 [ $? -eq 0 ] || { error "Unable to delete temporary files"; return 22; }
3497 run_test 52 "check recovering objects from lost+found"
3499 # Checks threads_min/max/started for some service
3501 # Arguments: service name (OST or MDT), facet (e.g., ost1, $SINGLEMDS), and a
3502 # parameter pattern prefix like 'ost.*.ost'.
3515 local msg="Insane $modname thread counts"
3516 local ncpts=$(check_cpt_number $facet)
3520 check_mount || return 41
3522 # We need to expand $parampat, but it may match multiple parameters, so
3523 # we'll pick the first one
3524 if ! paramp=$(do_facet $facet "lctl get_param -N ${parampat}.threads_min"|head -1); then
3525 error "Couldn't expand ${parampat}.threads_min parameter name"
3529 # Remove the .threads_min part
3530 paramp=${paramp%.threads_min}
3532 # Check for sanity in defaults
3533 tmin=$(do_facet $facet "lctl get_param -n ${paramp}.threads_min" || echo 0)
3534 tmax=$(do_facet $facet "lctl get_param -n ${paramp}.threads_max" || echo 0)
3535 tstarted=$(do_facet $facet "lctl get_param -n ${paramp}.threads_started" || echo 0)
3536 lassert 23 "$msg (PDSH problems?)" '(($tstarted && $tmin && $tmax))' || return $?
3537 lassert 24 "$msg" '(($tstarted >= $tmin && $tstarted <= $tmax ))' || return $?
3538 nthrs=$(expr $tmax - $tmin)
3539 if [ $nthrs -lt $ncpts ]; then
3545 [ $tmin -eq $tmax -a $tmin -eq $tstarted ] &&
3546 skip_env "module parameter forced $facet thread count" &&
3547 tmin=3 && tmax=$((3 * tmax))
3549 # Check that we can change min/max
3550 do_facet $facet "lctl set_param ${paramp}.threads_min=$((tmin + nthrs))"
3551 do_facet $facet "lctl set_param ${paramp}.threads_max=$((tmax - nthrs))"
3552 tmin2=$(do_facet $facet "lctl get_param -n ${paramp}.threads_min" || echo 0)
3553 tmax2=$(do_facet $facet "lctl get_param -n ${paramp}.threads_max" || echo 0)
3554 lassert 25 "$msg" '(($tmin2 == ($tmin + $nthrs) && $tmax2 == ($tmax - $nthrs)))' || return $?
3556 # Check that we can set min/max to the same value
3557 tmin=$(do_facet $facet "lctl get_param -n ${paramp}.threads_min" || echo 0)
3558 do_facet $facet "lctl set_param ${paramp}.threads_max=$tmin"
3559 tmin2=$(do_facet $facet "lctl get_param -n ${paramp}.threads_min" || echo 0)
3560 tmax2=$(do_facet $facet "lctl get_param -n ${paramp}.threads_max" || echo 0)
3561 lassert 26 "$msg" '(($tmin2 == $tmin && $tmax2 == $tmin))' || return $?
3563 # Check that we can't set max < min
3564 do_facet $facet "lctl set_param ${paramp}.threads_max=$((tmin - 1))"
3565 tmin2=$(do_facet $facet "lctl get_param -n ${paramp}.threads_min" || echo 0)
3566 tmax2=$(do_facet $facet "lctl get_param -n ${paramp}.threads_max" || echo 0)
3567 lassert 27 "$msg" '(($tmin2 <= $tmax2))' || return $?
3569 # We need to ensure that we get the module options desired; to do this
3570 # we set LOAD_MODULES_REMOTE=true and we call setmodopts below.
3571 LOAD_MODULES_REMOTE=true
3574 local newvalue="${opts}=$(expr $basethr \* $ncpts)"
3575 setmodopts -a $modname "$newvalue" oldvalue
3579 check_mount || return 41
3581 # Restore previous setting of MODOPTS_*
3582 setmodopts $modname "$oldvalue"
3584 # Check that $opts took
3585 tmin=$(do_facet $facet "lctl get_param -n ${paramp}.threads_min")
3586 tmax=$(do_facet $facet "lctl get_param -n ${paramp}.threads_max")
3587 tstarted=$(do_facet $facet "lctl get_param -n ${paramp}.threads_started")
3588 lassert 28 "$msg" '(($tstarted >= $tmin && $tstarted <= $tmax ))' || return $?
3597 thread_sanity OST ost1 'ost.*.ost' 'oss_num_threads' '16'
3600 run_test 53a "check OSS thread count params"
3604 local mds=$(do_facet $SINGLEMDS "lctl get_param -N mds.*.*.threads_max \
3606 if [ -z "$mds" ]; then
3607 #running this on an old MDT
3608 thread_sanity MDT $SINGLEMDS 'mdt.*.*.' 'mdt_num_threads' 16
3610 thread_sanity MDT $SINGLEMDS 'mds.*.*.' 'mds_num_threads' 16
3614 run_test 53b "check MDS thread count params"
3617 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
3618 skip "Only applicable to ldiskfs-based MDTs"
3622 do_rpc_nodes $(facet_host ost1) run_llverdev $(ostdevname 1) -p
3623 [ $? -eq 0 ] || error "llverdev failed!"
3626 run_test 54a "test llverdev and partial verify of device"
3629 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
3630 skip "Only applicable to ldiskfs-based MDTs"
3635 run_llverfs $MOUNT -p
3636 [ $? -eq 0 ] || error "llverfs failed!"
3639 run_test 54b "test llverfs and partial verify of filesystem"
3643 local max_ost_index=$1
3644 echo -n $(((max_ost_index + 1) * 8))
3648 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
3649 skip "Only applicable to ldiskfs-based MDTs"
3653 local mdsdev=$(mdsdevname 1)
3654 local mdsvdev=$(mdsvdevname 1)
3658 add mds1 $(mkfs_opts mds1 ${mdsdev}) --reformat $mdsdev \
3660 add ost1 $(mkfs_opts ost1 $(ostdevname 1)) --index=$i \
3661 --reformat $(ostdevname 1) $(ostvdevname 1)
3667 echo checking size of lov_objid for ost index $i
3668 LOV_OBJID_SIZE=$(do_facet mds1 "$DEBUGFS -R 'stat lov_objid' $mdsdev 2>/dev/null" | grep ^User | awk '{print $6}')
3669 if [ "$LOV_OBJID_SIZE" != $(lov_objid_size $i) ]; then
3670 error "lov_objid size has to be $(lov_objid_size $i), not $LOV_OBJID_SIZE"
3672 echo ok, lov_objid size is correct: $LOV_OBJID_SIZE
3679 run_test 55 "check lov_objid size"
3682 local mds_journal_size_orig=$MDSJOURNALSIZE
3687 for num in $(seq 1 $MDSCOUNT); do
3688 add mds${num} $(mkfs_opts mds${num} $(mdsdevname $num)) \
3689 --reformat $(mdsdevname $num) $(mdsvdevname $num)
3691 add ost1 $(mkfs_opts ost1 $(ostdevname 1)) --index=10000 --reformat \
3692 $(ostdevname 1) $(ostvdevname 1)
3693 add ost2 $(mkfs_opts ost2 $(ostdevname 2)) --index=1000 --reformat \
3694 $(ostdevname 2) $(ostvdevname 2)
3697 start_ost || error "Unable to start first ost (idx 10000)"
3698 start_ost2 || error "Unable to start second ost (idx 1000)"
3699 mount_client $MOUNT || error "Unable to mount client"
3702 wait_osc_import_state mds ost1 FULL
3703 wait_osc_import_state mds ost2 FULL
3704 $LFS setstripe --stripe-count=-1 $DIR/$tfile || error "Unable to create"
3705 n=$($LFS getstripe --stripe-count $DIR/$tfile)
3706 [ "$n" -eq 2 ] || error "Stripe count not two: $n"
3709 MDSJOURNALSIZE=$mds_journal_size_orig
3712 run_test 56 "check big OST indexes and out-of-index-order start"
3714 test_57a() { # bug 22656
3715 do_rpc_nodes $(facet_active_host ost1) load_modules_local
3716 local NID=$(do_facet ost1 "$LCTL get_param nis" | tail -1 | awk '{print $1}')
3717 writeconf_or_reformat
3718 [ $(facet_fstype ost1) == zfs ] && import_zpool ost1
3719 do_facet ost1 "$TUNEFS --failnode=$NID `ostdevname 1`" || error "tunefs failed"
3721 start_ost && error "OST registration from failnode should fail"
3724 run_test 57a "initial registration from failnode should fail (should return errs)"
3727 do_rpc_nodes $(facet_active_host ost1) load_modules_local
3728 local NID=$(do_facet ost1 "$LCTL get_param nis" | tail -1 | awk '{print $1}')
3729 writeconf_or_reformat
3730 [ $(facet_fstype ost1) == zfs ] && import_zpool ost1
3731 do_facet ost1 "$TUNEFS --servicenode=$NID `ostdevname 1`" || error "tunefs failed"
3733 start_ost || error "OST registration from servicenode should not fail"
3736 run_test 57b "initial registration from servicenode should not fail"
3739 do_facet mgs $LCTL get_param mgs.MGS.live.$FSNAME | grep OST | wc -l
3742 test_58() { # bug 22658
3745 createmany -o $DIR/$tdir/$tfile-%d 100
3746 # make sure that OSTs do not cancel llog cookies before we unmount the MDS
3747 #define OBD_FAIL_OBD_LOG_CANCEL_NET 0x601
3748 do_facet $SINGLEMDS "lctl set_param fail_loc=0x601"
3749 unlinkmany $DIR/$tdir/$tfile-%d 100
3752 local MNTDIR=$(facet_mntpt $SINGLEMDS)
3753 local devname=$(mdsdevname ${SINGLEMDS//mds/})
3755 # remove all files from the OBJECTS dir
3756 mount_fstype $SINGLEMDS
3758 do_facet $SINGLEMDS "find $MNTDIR/O/1/d* -type f -delete"
3760 unmount_fstype $SINGLEMDS
3761 # restart MDS with missing llog files
3763 do_facet mds "lctl set_param fail_loc=0"
3766 run_test 58 "missing llog files must not prevent MDT from mounting"
3769 start_mgsmds >> /dev/null
3770 local C1=$(count_osts)
3771 if [ $C1 -eq 0 ]; then
3772 start_ost >> /dev/null
3776 echo "original ost count: $C1 (expect > 0)"
3777 [ $C1 -gt 0 ] || error "No OSTs in $FSNAME log"
3778 start_mgsmds -o writeconf >> /dev/null || error "MDT start failed"
3779 local C2=$(count_osts)
3780 echo "after mdt writeconf count: $C2 (expect 0)"
3781 [ $C2 -gt 0 ] && error "MDT writeconf should erase OST logs"
3782 echo "OST start without writeconf should fail:"
3783 start_ost >> /dev/null && error "OST start without writeconf didn't fail"
3784 echo "OST start with writeconf should succeed:"
3785 start_ost -o writeconf >> /dev/null || error "OST1 start failed"
3786 local C3=$(count_osts)
3787 echo "after ost writeconf count: $C3 (expect 1)"
3788 [ $C3 -eq 1 ] || error "new OST writeconf should add:"
3789 start_ost2 -o writeconf >> /dev/null || error "OST2 start failed"
3790 local C4=$(count_osts)
3791 echo "after ost2 writeconf count: $C4 (expect 2)"
3792 [ $C4 -eq 2 ] || error "OST2 writeconf should add log"
3793 stop_ost2 >> /dev/null
3794 cleanup_nocli >> /dev/null
3795 #writeconf to remove all ost2 traces for subsequent tests
3796 writeconf_or_reformat
3798 run_test 59 "writeconf mount option"
3800 test_60() { # LU-471
3803 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
3804 skip "Only applicable to ldiskfs-based MDTs"
3808 for num in $(seq $MDSCOUNT); do
3809 add mds${num} $(mkfs_opts mds${num} $(mdsdevname $num)) \
3810 --mkfsoptions='\" -E stride=64 -O ^uninit_bg\"' \
3811 --reformat $(mdsdevname $num) $(mdsvdevname $num) ||
3815 dump=$(do_facet $SINGLEMDS dumpe2fs $(mdsdevname 1))
3817 [ $rc -eq 0 ] || error "dumpe2fs $(mdsdevname 1) failed"
3819 # MDT default has dirdata feature
3820 echo $dump | grep dirdata > /dev/null || error "dirdata is not set"
3821 # we disable uninit_bg feature
3822 echo $dump | grep uninit_bg > /dev/null && error "uninit_bg is set"
3823 # we set stride extended options
3824 echo $dump | grep stride > /dev/null || error "stride is not set"
3827 run_test 60 "check mkfs.lustre --mkfsoptions -E -O options setting"
3830 local reformat=false
3832 [ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.1.53) ] ||
3833 { skip "Need MDS version at least 2.1.53"; return 0; }
3835 if [ $(facet_fstype $SINGLEMDS) == ldiskfs ] &&
3836 ! large_xattr_enabled; then
3838 LDISKFS_MKFS_OPTS+=" -O large_xattr"
3840 for num in $(seq $MDSCOUNT); do
3841 add mds${num} $(mkfs_opts mds$num $(mdsdevname $num)) \
3842 --reformat $(mdsdevname $num) $(mdsvdevname $num) ||
3843 error "add mds $num failed"
3847 setup_noconfig || error "setting up the filesystem failed"
3848 client_up || error "starting client failed"
3850 local file=$DIR/$tfile
3853 local large_value="$(generate_string $(max_xattr_size))"
3854 local small_value="bar"
3856 local name="trusted.big"
3857 log "save large xattr $name on $file"
3858 setfattr -n $name -v $large_value $file ||
3859 error "saving $name on $file failed"
3861 local new_value=$(get_xattr_value $name $file)
3862 [[ "$new_value" != "$large_value" ]] &&
3863 error "$name different after saving"
3865 log "shrink value of $name on $file"
3866 setfattr -n $name -v $small_value $file ||
3867 error "shrinking value of $name on $file failed"
3869 new_value=$(get_xattr_value $name $file)
3870 [[ "$new_value" != "$small_value" ]] &&
3871 error "$name different after shrinking"
3873 log "grow value of $name on $file"
3874 setfattr -n $name -v $large_value $file ||
3875 error "growing value of $name on $file failed"
3877 new_value=$(get_xattr_value $name $file)
3878 [[ "$new_value" != "$large_value" ]] &&
3879 error "$name different after growing"
3881 log "check value of $name on $file after remounting MDS"
3883 new_value=$(get_xattr_value $name $file)
3884 [[ "$new_value" != "$large_value" ]] &&
3885 error "$name different after remounting MDS"
3887 log "remove large xattr $name from $file"
3888 setfattr -x $name $file || error "removing $name from $file failed"
3893 LDISKFS_MKFS_OPTS=${LDISKFS_MKFS_OPTS% -O large_xattr}
3897 run_test 61 "large xattr"
3900 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
3901 skip "Only applicable to ldiskfs-based MDTs"
3906 local mdsdev=$(mdsdevname 1)
3907 local ostdev=$(ostdevname 1)
3909 [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.2.51) ]] ||
3910 { skip "Need MDS version at least 2.2.51"; return 0; }
3912 echo "disable journal for mds"
3913 do_facet mds tune2fs -O ^has_journal $mdsdev || error "tune2fs failed"
3914 start_mds && error "MDT start should fail"
3915 echo "disable journal for ost"
3916 do_facet ost1 tune2fs -O ^has_journal $ostdev || error "tune2fs failed"
3917 start_ost && error "OST start should fail"
3918 cleanup || return $?
3921 run_test 62 "start with disabled journal"
3924 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
3925 skip "Only applicable to ldiskfs-based MDTs"
3929 local inode_slab=$(do_facet $SINGLEMDS \
3930 "awk '/ldiskfs_inode_cache/ { print \\\$5 }' /proc/slabinfo")
3931 if [ -z "$inode_slab" ]; then
3932 skip "ldiskfs module has not been loaded"
3936 echo "$inode_slab ldisk inodes per page"
3937 [ "$inode_slab" -ge "3" ] ||
3938 error "ldisk inode size is too big, $inode_slab objs per page"
3941 run_test 63 "Verify each page can at least hold 3 ldisk inodes"
3946 start_ost2 || error "Unable to start second ost"
3947 mount_client $MOUNT || error "Unable to mount client"
3948 stop_ost2 || error "Unable to stop second ost"
3950 $LFS df --lazy || error "lfs df failed"
3951 cleanup || return $?
3952 #writeconf to remove all ost2 traces for subsequent tests
3953 writeconf_or_reformat
3955 run_test 64 "check lfs df --lazy "
3957 test_65() { # LU-2237
3958 # Currently, the test is only valid for ldiskfs backend
3959 [ "$(facet_fstype $SINGLEMDS)" != "ldiskfs" ] &&
3960 skip "non-ldiskfs backend" && return
3962 local devname=$(mdsdevname ${SINGLEMDS//mds/})
3963 local brpt=$(facet_mntpt brpt)
3966 if ! do_facet $SINGLEMDS "test -b $devname"; then
3971 local obj=$(do_facet $SINGLEMDS \
3972 "$DEBUGFS -c -R \\\"stat last_rcvd\\\" $devname" |
3974 if [ -z "$obj" ]; then
3975 # The MDT may be just re-formatted, mount the MDT for the
3976 # first time to guarantee the "last_rcvd" file is there.
3977 start_mds || error "fail to mount the MDS for the first time"
3981 # remove the "last_rcvd" file
3982 do_facet $SINGLEMDS "mkdir -p $brpt"
3983 do_facet $SINGLEMDS \
3984 "mount -t $(facet_fstype $SINGLEMDS) $opts $devname $brpt"
3985 do_facet $SINGLEMDS "rm -f ${brpt}/last_rcvd"
3986 do_facet $SINGLEMDS "umount -d $brpt"
3988 # restart MDS, the "last_rcvd" file should be recreated.
3989 start_mds || error "fail to restart the MDS"
3991 obj=$(do_facet $SINGLEMDS \
3992 "$DEBUGFS -c -R \\\"stat last_rcvd\\\" $devname" | grep Inode)
3993 [ -n "$obj" ] || error "fail to re-create the last_rcvd"
3995 run_test 65 "re-create the lost last_rcvd file when server mount"
3998 [[ $(lustre_version_code mgs) -ge $(version_code 2.3.59) ]] ||
3999 { skip "Need MGS version at least 2.3.59"; return 0; }
4002 local OST1_NID=$(do_facet ost1 $LCTL list_nids | head -1)
4003 local MDS_NID=$(do_facet $SINGLEMDS $LCTL list_nids | head -1)
4005 echo "replace_nids should fail if MDS, OSTs and clients are UP"
4006 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 $OST1_NID &&
4007 error "replace_nids fail"
4009 umount_client $MOUNT || error "unmounting client failed"
4010 echo "replace_nids should fail if MDS and OSTs are UP"
4011 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 $OST1_NID &&
4012 error "replace_nids fail"
4015 echo "replace_nids should fail if MDS is UP"
4016 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 $OST1_NID &&
4017 error "replace_nids fail"
4019 stop_mds || error "stopping mds failed"
4021 if combined_mgs_mds; then
4022 start_mdt 1 "-o nosvc" ||
4023 error "starting mds with nosvc option failed"
4026 echo "command should accept two parameters"
4027 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 &&
4028 error "command should accept two params"
4030 echo "correct device name should be passed"
4031 do_facet mgs $LCTL replace_nids $FSNAME-WRONG0000 $OST1_NID &&
4032 error "wrong devname"
4034 echo "wrong nids list should not destroy the system"
4035 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 "wrong nids list" &&
4038 echo "replace OST nid"
4039 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 $OST1_NID ||
4040 error "replace nids failed"
4042 echo "command should accept two parameters"
4043 do_facet mgs $LCTL replace_nids $FSNAME-MDT0000 &&
4044 error "command should accept two params"
4046 echo "wrong nids list should not destroy the system"
4047 do_facet mgs $LCTL replace_nids $FSNAME-MDT0000 "wrong nids list" &&
4050 echo "replace MDS nid"
4051 do_facet mgs $LCTL replace_nids $FSNAME-MDT0000 $MDS_NID ||
4052 error "replace nids failed"
4054 if ! combined_mgs_mds ; then
4061 check_mount || error "error after nid replace"
4062 cleanup || error "cleanup failed"
4065 run_test 66 "replace nids"
4067 test_67() { #LU-2950
4068 local legacy="$TMP/legacy_lnet_config"
4069 local new="$TMP/new_routes_test"
4070 local out="$TMP/config_out_file"
4071 local verify="$TMP/conv_verify"
4072 local verify_conf="$TMP/conf_verify"
4074 # Create the legacy file that will be run through the
4075 # lustre_routes_conversion script
4076 cat <<- LEGACY_LNET_CONFIG > $legacy
4077 tcp1 23 192.168.213.1@tcp:1; tcp5 34 193.30.4.3@tcp:4;
4078 tcp2 54 10.1.3.2@tcp;
4079 tcp3 10.3.4.3@tcp:3;
4083 # Create the verification file to verify the output of
4084 # lustre_routes_conversion script against.
4085 cat <<- VERIFY_LNET_CONFIG > $verify
4086 tcp1: { gateway: 192.168.213.1@tcp, hop: 23, priority: 1 }
4087 tcp5: { gateway: 193.30.4.3@tcp, hop: 34, priority: 4 }
4088 tcp2: { gateway: 10.1.3.2@tcp, hop: 54 }
4089 tcp3: { gateway: 10.3.4.3@tcp, priority: 3 }
4090 tcp4: { gateway: 10.3.3.4@tcp }
4093 # Create the verification file to verify the output of
4094 # lustre_routes_config script against
4095 cat <<- VERIFY_LNET_CONFIG > $verify_conf
4096 lctl --net tcp1 add_route 192.168.213.1@tcp 23 1
4097 lctl --net tcp5 add_route 193.30.4.3@tcp 34 4
4098 lctl --net tcp2 add_route 10.1.3.2@tcp 54 4
4099 lctl --net tcp3 add_route 10.3.4.3@tcp 1 3
4100 lctl --net tcp4 add_route 10.3.3.4@tcp 1 3
4103 lustre_routes_conversion $legacy $new > /dev/null
4104 if [ -f $new ]; then
4105 # verify the conversion output
4106 cmp -s $new $verify > /dev/null
4107 if [ $? -eq 1 ]; then
4108 error "routes conversion failed"
4111 lustre_routes_config --dry-run --verbose $new > $out
4112 # check that the script succeeded
4113 cmp -s $out $verify_conf > /dev/null
4114 if [ $? -eq 1 ]; then
4115 error "routes config failed"
4118 error "routes conversion test failed"
4120 # remove generated files
4121 rm -f $new $legacy $verify $verify_conf $out
4123 run_test 67 "test routes conversion and configuration"
4131 [ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.4.53) ] ||
4132 { skip "Need MDS version at least 2.4.53"; return 0; }
4134 umount_client $MOUNT || error "umount client failed"
4136 start_mdt 1 || error "MDT start failed"
4139 # START-END - the sequences we'll be reserving
4140 START=$(do_facet $SINGLEMDS \
4141 lctl get_param -n seq.ctl*.space | awk -F'[[ ]' '{print $2}')
4142 END=$((START + (1 << 30)))
4143 do_facet $SINGLEMDS \
4144 lctl set_param seq.ctl*.fldb="[$START-$END\):0:mdt"
4146 # reset the sequences MDT0000 has already assigned
4147 do_facet $SINGLEMDS \
4148 lctl set_param seq.srv*MDT0000.space=clear
4150 # remount to let the client allocate new sequence
4151 mount_client $MOUNT || error "mount client failed"
4154 do_facet $SINGLEMDS \
4155 lctl get_param seq.srv*MDT0000.space
4156 $LFS path2fid $DIR/$tfile
4158 local old_ifs="$IFS"
4160 fid=($($LFS path2fid $DIR/$tfile))
4164 if [[ $seq < $END ]]; then
4165 error "used reserved sequence $seq?"
4167 cleanup || return $?
4169 run_test 68 "be able to reserve specific sequences in FLDB"
4172 local server_version=$(lustre_version_code $SINGLEMDS)
4174 [[ $server_version -lt $(version_code 2.4.2) ]] &&
4175 skip "Need MDS version at least 2.4.2" && return
4177 [[ $server_version -ge $(version_code 2.4.50) ]] &&
4178 [[ $server_version -lt $(version_code 2.5.0) ]] &&
4179 skip "Need MDS version at least 2.5.0" && return
4183 # use OST0000 since it probably has the most creations
4184 local OSTNAME=$(ostname_from_index 0)
4185 local mdtosc_proc1=$(get_mdtosc_proc_path mds1 $OSTNAME)
4186 local last_id=$(do_facet mds1 lctl get_param -n \
4187 osc.$mdtosc_proc1.prealloc_last_id)
4189 # Want to have OST LAST_ID over 1.5 * OST_MAX_PRECREATE to
4190 # verify that the LAST_ID recovery is working properly. If
4191 # not, then the OST will refuse to allow the MDS connect
4192 # because the LAST_ID value is too different from the MDS
4193 #define OST_MAX_PRECREATE=20000
4194 local num_create=$((20000 * 5))
4197 $LFS setstripe -i 0 $DIR/$tdir
4198 createmany -o $DIR/$tdir/$tfile- $num_create ||
4199 error "createmany: failed to create $num_create files: $?"
4200 # delete all of the files with objects on OST0 so the
4201 # filesystem is not inconsistent later on
4202 $LFS find $MOUNT --ost 0 | xargs rm
4204 stop_ost || error "OST0 stop failure"
4205 add ost1 $(mkfs_opts ost1 $(ostdevname 1)) --reformat --replace \
4206 $(ostdevname 1) $(ostvdevname 1) ||
4207 error "reformat and replace $ostdev failed"
4208 start_ost || error "OST0 restart failure"
4209 wait_osc_import_state mds ost FULL
4211 touch $DIR/$tdir/$tfile-last || error "create file after reformat"
4212 local idx=$($LFS getstripe -i $DIR/$tdir/$tfile-last)
4213 [ $idx -ne 0 ] && error "$DIR/$tdir/$tfile-last on $idx not 0" || true
4217 run_test 69 "replace an OST with the same index"
4220 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
4225 start_mdt 1 || error "MDT0 start fail"
4227 start_ost || error "OST0 start fail"
4229 start_mdt 2 || error "MDT1 start fail"
4231 mount_client $MOUNT || error "mount client fails"
4233 mkdir -p $DIR/$tdir || error "create dir fail"
4235 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
4236 error "create remote dir fail"
4238 rm -rf $DIR/$tdir || error "delete dir fail"
4239 cleanup || return $?
4241 run_test 70a "start MDT0, then OST, then MDT1"
4244 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
4247 start_ost || error "OST0 start fail"
4249 start_mdt 1 || error "MDT0 start fail"
4250 start_mdt 2 || error "MDT1 start fail"
4252 mount_client $MOUNT || error "mount client fails"
4254 mkdir -p $DIR/$tdir || error "create dir fail"
4256 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
4257 error "create remote dir fail"
4259 rm -rf $DIR/$tdir || error "delete dir fail"
4261 cleanup || return $?
4263 run_test 70b "start OST, MDT1, MDT0"
4266 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
4269 start_mdt 1 || error "MDT0 start fail"
4270 start_mdt 2 || error "MDT1 start fail"
4271 start_ost || error "OST0 start fail"
4273 mount_client $MOUNT || error "mount client fails"
4274 stop_mdt 1 || error "MDT1 start fail"
4276 local mdc_for_mdt1=$($LCTL dl | grep MDT0000-mdc | awk '{print $4}')
4277 echo "deactivate $mdc_for_mdt1"
4278 $LCTL --device $mdc_for_mdt1 deactivate || return 1
4280 mkdir -p $DIR/$tdir && error "mkdir succeed"
4282 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir &&
4283 error "create remote dir succeed"
4285 cleanup || return $?
4287 run_test 70c "stop MDT0, mkdir fail, create remote dir fail"
4290 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
4293 start_mdt 1 || error "MDT0 start fail"
4294 start_mdt 2 || error "MDT1 start fail"
4295 start_ost || error "OST0 start fail"
4297 mount_client $MOUNT || error "mount client fails"
4299 stop_mdt 2 || error "MDT1 start fail"
4301 local mdc_for_mdt2=$($LCTL dl | grep MDT0001-mdc |
4303 echo "deactivate $mdc_for_mdt2"
4304 $LCTL --device $mdc_for_mdt2 deactivate ||
4305 error "set $mdc_for_mdt2 deactivate failed"
4307 mkdir -p $DIR/$tdir || error "mkdir fail"
4308 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir &&
4309 error "create remote dir succeed"
4311 rm -rf $DIR/$tdir || error "delete dir fail"
4313 cleanup || return $?
4315 run_test 70d "stop MDT1, mkdir succeed, create remote dir fail"
4318 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
4319 if combined_mgs_mds; then
4320 skip "needs separate MGS/MDT" && return
4324 start_mdt 1 || error "MDT0 start fail"
4325 start_ost || error "OST0 start fail"
4326 start_mdt 2 || error "MDT1 start fail"
4327 start_ost2 || error "OST1 start fail"
4329 mount_client $MOUNT || error "mount client fails"
4331 mkdir -p $DIR/$tdir || error "mkdir fail"
4332 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
4333 error "create remote dir succeed"
4335 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
4336 rm -rf $DIR/$tdir || error "delete dir fail"
4338 umount_client $MOUNT
4339 stop_mdt 1 || error "MDT0 stop fail"
4340 stop_mdt 2 || error "MDT1 stop fail"
4341 stop_ost || error "OST0 stop fail"
4342 stop_ost2 || error "OST1 stop fail"
4344 run_test 71a "start MDT0 OST0, MDT1, OST1"
4347 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
4348 if combined_mgs_mds; then
4349 skip "needs separate MGS/MDT" && return
4353 start_mdt 2 || error "MDT1 start fail"
4354 start_ost || error "OST0 start fail"
4355 start_mdt 1 || error "MDT0 start fail"
4356 start_ost2 || error "OST1 start fail"
4358 mount_client $MOUNT || error "mount client fails"
4360 mkdir -p $DIR/$tdir || error "mkdir fail"
4361 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
4362 error "create remote dir succeed"
4364 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
4365 rm -rf $DIR/$tdir || error "delete dir fail"
4367 umount_client $MOUNT
4368 stop_mdt 1 || error "MDT0 stop fail"
4369 stop_mdt 2 || error "MDT1 stop fail"
4370 stop_ost || error "OST0 stop fail"
4371 stop_ost2 || error "OST1 stop fail"
4373 run_test 71b "start MDT1, OST0, MDT0, OST1"
4376 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
4377 if combined_mgs_mds; then
4378 skip "needs separate MGS/MDT" && return
4382 start_ost || error "OST0 start fail"
4383 start_ost2 || error "OST1 start fail"
4384 start_mdt 2 || error "MDT1 start fail"
4385 start_mdt 1 || error "MDT0 start fail"
4387 mount_client $MOUNT || error "mount client fails"
4389 mkdir -p $DIR/$tdir || error "mkdir fail"
4390 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
4391 error "create remote dir succeed"
4393 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
4394 rm -rf $DIR/$tdir || error "delete dir fail"
4396 umount_client $MOUNT
4397 stop_mdt 1 || error "MDT0 stop fail"
4398 stop_mdt 2 || error "MDT1 stop fail"
4399 stop_ost || error "OST0 stop fail"
4400 stop_ost2 || error "OST1 stop fail"
4403 run_test 71c "start OST0, OST1, MDT1, MDT0"
4406 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
4407 if combined_mgs_mds; then
4408 skip "needs separate MGS/MDT" && return
4412 start_ost || error "OST0 start fail"
4413 start_mdt 2 || error "MDT0 start fail"
4414 start_mdt 1 || error "MDT0 start fail"
4415 start_ost2 || error "OST1 start fail"
4417 mount_client $MOUNT || error "mount client fails"
4419 mkdir -p $DIR/$tdir || error "mkdir fail"
4420 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
4421 error "create remote dir succeed"
4423 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
4424 rm -rf $DIR/$tdir || error "delete dir fail"
4426 umount_client $MOUNT
4427 stop_mdt 1 || error "MDT0 stop fail"
4428 stop_mdt 2 || error "MDT1 stop fail"
4429 stop_ost || error "OST0 stop fail"
4430 stop_ost2 || error "OST1 stop fail"
4433 run_test 71d "start OST0, MDT1, MDT0, OST1"
4436 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
4437 if combined_mgs_mds; then
4438 skip "needs separate MGS/MDT" && return
4442 start_ost || error "OST0 start fail"
4443 start_mdt 2 || error "MDT1 start fail"
4444 start_ost2 || error "OST1 start fail"
4445 start_mdt 1 || error "MDT0 start fail"
4447 mount_client $MOUNT || error "mount client fails"
4449 mkdir -p $DIR/$tdir || error "mkdir fail"
4450 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
4451 error "create remote dir succeed"
4453 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
4454 rm -rf $DIR/$tdir || error "delete dir fail"
4456 umount_client $MOUNT
4457 stop_mdt 1 || error "MDT0 stop fail"
4458 stop_mdt 2 || error "MDT1 stop fail"
4459 stop_ost || error "OST0 stop fail"
4460 stop_ost2 || error "OST1 stop fail"
4463 run_test 71e "start OST0, MDT1, OST1, MDT0"
4465 test_72() { #LU-2634
4466 local mdsdev=$(mdsdevname 1)
4467 local ostdev=$(ostdevname 1)
4468 local cmd="$E2FSCK -fnvd $mdsdev"
4471 [ "$(facet_fstype $SINGLEMDS)" != "ldiskfs" ] &&
4472 skip "ldiskfs only test" && return
4474 #tune MDT with "-O extents"
4476 for num in $(seq $MDSCOUNT); do
4477 add mds${num} $(mkfs_opts mds$num $(mdsdevname $num)) \
4478 --reformat $(mdsdevname $num) $(mdsvdevname $num) ||
4479 error "add mds $num failed"
4480 $TUNE2FS -O extents $(mdsdevname $num)
4483 add ost1 $(mkfs_opts ost1 $ostdev) --reformat $ostdev ||
4484 error "add $ostdev failed"
4485 start_mgsmds || error "start mds failed"
4486 start_ost || error "start ost failed"
4487 mount_client $MOUNT || error "mount client failed"
4489 #create some short symlinks
4491 createmany -o $DIR/$tdir/$tfile-%d $fn
4492 echo "create $fn short symlinks"
4493 for i in $(seq -w 1 $fn); do
4494 ln -s $DIR/$tdir/$tfile-$i $MOUNT/$tfile-$i
4499 umount_client $MOUNT || error "umount client failed"
4500 stop_mds || error "stop mds failed"
4501 stop_ost || error "stop ost failed"
4504 run_e2fsck $(facet_active_host $SINGLEMDS) $mdsdev "-n"
4506 run_test 72 "test fast symlink with extents flag enabled"
4508 test_73() { #LU-3006
4510 [ $(facet_fstype ost1) == zfs ] && import_zpool ost1
4511 do_facet ost1 "$TUNEFS --failnode=1.2.3.4@$NETTYPE $(ostdevname 1)" ||
4512 error "1st tunefs failed"
4513 start_mgsmds || error "start mds failed"
4514 start_ost || error "start ost failed"
4515 mount_client $MOUNT || error "mount client failed"
4516 lctl get_param -n osc.*OST0000-osc-[^M]*.import | grep failover_nids |
4517 grep 1.2.3.4@$NETTYPE || error "failover nids haven't changed"
4518 umount_client $MOUNT || error "umount client failed"
4522 run_test 73 "failnode to update from mountdata properly"
4524 test_75() { # LU-2374
4525 [[ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.4.1) ]] &&
4526 skip "Need MDS version at least 2.4.1" && return
4529 local opts_mds="$(mkfs_opts mds1 $(mdsdevname 1)) \
4530 --reformat $(mdsdevname 1) $(mdsvdevname 1)"
4531 local opts_ost="$(mkfs_opts ost1 $(ostdevname 1)) \
4532 --reformat $(ostdevname 1) $(ostvdevname 1)"
4534 #check with default parameters
4535 add mds1 $opts_mds || error "add mds1 failed for default params"
4536 add ost1 $opts_ost || error "add ost1 failed for default params"
4538 opts_mds=$(echo $opts_mds | sed -e "s/--mdt//")
4539 opts_mds=$(echo $opts_mds |
4540 sed -e "s/--index=$index/--index=$index --mdt/")
4541 opts_ost=$(echo $opts_ost | sed -e "s/--ost//")
4542 opts_ost=$(echo $opts_ost |
4543 sed -e "s/--index=$index/--index=$index --ost/")
4545 add mds1 $opts_mds || error "add mds1 failed for new params"
4546 add ost1 $opts_ost || error "add ost1 failed for new params"
4549 run_test 75 "The order of --index should be irrelevant"
4552 [[ $(lustre_version_code mgs) -ge $(version_code 2.4.52) ]] ||
4553 { skip "Need MDS version at least 2.4.52" && return 0; }
4555 local MDMB_PARAM="osc.*.max_dirty_mb"
4556 echo "Change MGS params"
4557 local MAX_DIRTY_MB=$($LCTL get_param -n $MDMB_PARAM |
4559 echo "max_dirty_mb: $MAX_DIRTY_MB"
4560 local NEW_MAX_DIRTY_MB=$((MAX_DIRTY_MB + MAX_DIRTY_MB))
4561 echo "new_max_dirty_mb: $NEW_MAX_DIRTY_MB"
4562 do_facet mgs $LCTL set_param -P $MDMB_PARAM=$NEW_MAX_DIRTY_MB
4563 wait_update $HOSTNAME "lctl get_param -n $MDMB_PARAM |
4564 head -1" $NEW_MAX_DIRTY_MB
4565 MAX_DIRTY_MB=$($LCTL get_param -n $MDMB_PARAM | head -1)
4566 echo "$MAX_DIRTY_MB"
4567 [ $MAX_DIRTY_MB = $NEW_MAX_DIRTY_MB ] ||
4568 error "error while apply max_dirty_mb"
4570 echo "Check the value is stored after remount"
4573 wait_update $HOSTNAME "lctl get_param -n $MDMB_PARAM |
4574 head -1" $NEW_MAX_DIRTY_MB
4575 MAX_DIRTY_MB=$($LCTL get_param -n $MDMB_PARAM | head -1)
4576 [ $MAX_DIRTY_MB = $NEW_MAX_DIRTY_MB ] ||
4577 error "max_dirty_mb is not saved after remount"
4579 echo "Change OST params"
4580 CLIENT_PARAM="obdfilter.*.client_cache_count"
4581 local CLIENT_CACHE_COUNT
4582 CLIENT_CACHE_COUNT=$(do_facet ost1 $LCTL get_param -n $CLIENT_PARAM |
4584 echo "client_cache_count: $CLIENT_CACHE_COUNT"
4585 NEW_CLIENT_CACHE_COUNT=$((CLIENT_CACHE_COUNT+CLIENT_CACHE_COUNT))
4586 echo "new_client_cache_count: $NEW_CLIENT_CACHE_COUNT"
4587 do_facet mgs $LCTL set_param -P $CLIENT_PARAM=$NEW_CLIENT_CACHE_COUNT
4588 wait_update $(facet_host ost1) "lctl get_param -n $CLIENT_PARAM |
4589 head -1" $NEW_CLIENT_CACHE_COUNT
4590 CLIENT_CACHE_COUNT=$(do_facet ost1 $LCTL get_param -n $CLIENT_PARAM |
4592 echo "$CLIENT_CACHE_COUNT"
4593 [ $CLIENT_CACHE_COUNT = $NEW_CLIENT_CACHE_COUNT ] ||
4594 error "error while apply client_cache_count"
4596 echo "Check the value is stored after remount"
4599 wait_update $(facet_host ost1) "lctl get_param -n $CLIENT_PARAM |
4600 head -1" $NEW_CLIENT_CACHE_COUNT
4601 CLIENT_CACHE_COUNT=$(do_facet ost1 $LCTL get_param -n $CLIENT_PARAM |
4603 echo "$CLIENT_CACHE_COUNT"
4604 [ $CLIENT_CACHE_COUNT = $NEW_CLIENT_CACHE_COUNT ] ||
4605 error "client_cache_count is not saved after remount"
4608 run_test 76a "set permanent params set_param -P"
4610 test_76b() { # LU-4783
4611 [[ $(lustre_version_code mgs) -ge $(version_code 2.5.57) ]] ||
4612 { skip "Need MGS version at least 2.5.57" && return 0; }
4615 do_facet mgs $LCTL get_param mgs.MGS.live.params ||
4616 error "start params log failed"
4619 run_test 76b "verify params log setup correctly"
4621 test_77() { # LU-3445
4622 local server_version=$(lustre_version_code $SINGLEMDS)
4624 [[ $server_version -ge $(version_code 2.2.60) ]] &&
4625 [[ $server_version -le $(version_code 2.4.0) ]] &&
4626 skip "Need MDS version < 2.2.60 or > 2.4.0" && return
4628 if [[ -z "$fs2ost_DEV" || -z "$fs2mds_DEV" ]]; then
4629 is_blkdev $SINGLEMDS $(mdsdevname ${SINGLEMDS//mds/}) &&
4630 skip_env "mixed loopback and real device not working" && return
4633 local fs2mdsdev=$(mdsdevname 1_2)
4634 local fs2ostdev=$(ostdevname 1_2)
4635 local fs2mdsvdev=$(mdsvdevname 1_2)
4636 local fs2ostvdev=$(ostvdevname 1_2)
4637 local fsname=test1234
4639 local failnid="$(h2$NETTYPE 1.2.3.4),$(h2$NETTYPE 4.3.2.1)"
4641 add fs2mds $(mkfs_opts mds1 $fs2mdsdev) --mgs --fsname=$fsname \
4642 --reformat $fs2mdsdev $fs2mdsvdev || error "add fs2mds failed"
4643 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS && trap cleanup_fs2 EXIT INT ||
4644 error "start fs2mds failed"
4646 mgsnid=$(do_facet fs2mds $LCTL list_nids | xargs | tr ' ' ,)
4647 [[ $mgsnid = *,* ]] || mgsnid+=",$mgsnid"
4649 add fs2ost $(mkfs_opts ost1 $fs2ostdev) --mgsnode=$mgsnid \
4650 --failnode=$failnid --fsname=$fsname \
4651 --reformat $fs2ostdev $fs2ostvdev ||
4652 error "add fs2ost failed"
4653 start fs2ost $fs2ostdev $OST_MOUNT_OPTS || error "start fs2ost failed"
4656 $MOUNT_CMD $mgsnid:/$fsname $MOUNT2 || error "mount $MOUNT2 failed"
4657 DIR=$MOUNT2 MOUNT=$MOUNT2 check_mount || error "check $MOUNT2 failed"
4660 run_test 77 "comma-separated MGS NIDs and failover node NIDs"
4663 [[ $(facet_fstype $SINGLEMDS) != ldiskfs ||
4664 $(facet_fstype ost1) != ldiskfs ]] &&
4665 skip "only applicable to ldiskfs-based MDTs and OSTs" && return
4667 # reformat the Lustre filesystem with a smaller size
4668 local saved_MDSSIZE=$MDSSIZE
4669 local saved_OSTSIZE=$OSTSIZE
4670 MDSSIZE=$((MDSSIZE - 20000))
4671 OSTSIZE=$((OSTSIZE - 20000))
4672 reformat || error "(1) reformat Lustre filesystem failed"
4673 MDSSIZE=$saved_MDSSIZE
4674 OSTSIZE=$saved_OSTSIZE
4676 # mount the Lustre filesystem
4677 setup_noconfig || error "(2) setup Lustre filesystem failed"
4680 log "create test files"
4684 mkdir -p $MOUNT/$tdir || error "(3) mkdir $MOUNT/$tdir failed"
4685 for i in $(seq $num_files); do
4686 file=$MOUNT/$tdir/$tfile-$i
4687 dd if=/dev/urandom of=$file count=1 bs=1M ||
4688 error "(4) create $file failed"
4691 # unmount the Lustre filesystem
4692 cleanup || error "(5) cleanup Lustre filesystem failed"
4694 # run e2fsck on the MDT and OST devices
4695 local mds_host=$(facet_active_host $SINGLEMDS)
4696 local ost_host=$(facet_active_host ost1)
4697 local mds_dev=$(mdsdevname ${SINGLEMDS//mds/})
4698 local ost_dev=$(ostdevname 1)
4700 run_e2fsck $mds_host $mds_dev "-y"
4701 run_e2fsck $ost_host $ost_dev "-y"
4703 # get the original block count of the MDT and OST filesystems
4704 local mds_orig_blks=$(get_block_count $SINGLEMDS $mds_dev)
4705 local ost_orig_blks=$(get_block_count ost1 $ost_dev)
4707 # expand the MDT and OST filesystems to the device size
4708 run_resize2fs $SINGLEMDS $mds_dev "" || error "expand $SINGLEMDS failed"
4709 run_resize2fs ost1 $ost_dev "" || error "expand ost1 failed"
4711 # run e2fsck on the MDT and OST devices again
4712 run_e2fsck $mds_host $mds_dev "-y"
4713 run_e2fsck $ost_host $ost_dev "-y"
4715 # mount the Lustre filesystem
4719 log "check files after expanding the MDT and OST filesystems"
4720 for i in $(seq $num_files); do
4721 file=$MOUNT/$tdir/$tfile-$i
4722 $CHECKSTAT -t file -s 1048576 $file ||
4723 error "(6) checkstat $file failed"
4727 log "create more files after expanding the MDT and OST filesystems"
4728 for i in $(seq $((num_files + 1)) $((num_files + 10))); do
4729 file=$MOUNT/$tdir/$tfile-$i
4730 dd if=/dev/urandom of=$file count=1 bs=1M ||
4731 error "(7) create $file failed"
4734 # unmount the Lustre filesystem
4735 cleanup || error "(8) cleanup Lustre filesystem failed"
4737 # run e2fsck on the MDT and OST devices
4738 run_e2fsck $mds_host $mds_dev "-y"
4739 run_e2fsck $ost_host $ost_dev "-y"
4741 # get the maximum block count of the MDT and OST filesystems
4742 local mds_max_blks=$(get_block_count $SINGLEMDS $mds_dev)
4743 local ost_max_blks=$(get_block_count ost1 $ost_dev)
4745 # get the minimum block count of the MDT and OST filesystems
4746 local mds_min_blks=$(run_resize2fs $SINGLEMDS $mds_dev "" "-P" 2>&1 |
4747 grep minimum | sed -e 's/^.*filesystem: //g')
4748 local ost_min_blks=$(run_resize2fs ost1 $ost_dev "" "-P" 2>&1 |
4749 grep minimum | sed -e 's/^.*filesystem: //g')
4751 # shrink the MDT and OST filesystems to a smaller size
4755 if [[ $mds_max_blks -gt $mds_min_blks &&
4756 $mds_max_blks -gt $mds_orig_blks ]]; then
4757 [[ $mds_orig_blks -gt $mds_min_blks ]] &&
4758 base_blks=$mds_orig_blks || base_blks=$mds_min_blks
4759 new_blks=$(( (mds_max_blks - base_blks) / 2 + base_blks ))
4760 run_resize2fs $SINGLEMDS $mds_dev $new_blks ||
4761 error "shrink $SINGLEMDS to $new_blks failed"
4765 if [[ $ost_max_blks -gt $ost_min_blks &&
4766 $ost_max_blks -gt $ost_orig_blks ]]; then
4767 [[ $ost_orig_blks -gt $ost_min_blks ]] &&
4768 base_blks=$ost_orig_blks || base_blks=$ost_min_blks
4769 new_blks=$(( (ost_max_blks - base_blks) / 2 + base_blks ))
4770 run_resize2fs ost1 $ost_dev $new_blks ||
4771 error "shrink ost1 to $new_blks failed"
4775 # check whether the MDT or OST filesystem was shrunk or not
4777 combined_mgs_mds || stop_mgs || error "(9) stop mgs failed"
4778 reformat || error "(10) reformat Lustre filesystem failed"
4782 # run e2fsck on the MDT and OST devices again
4783 run_e2fsck $mds_host $mds_dev "-y"
4784 run_e2fsck $ost_host $ost_dev "-y"
4786 # mount the Lustre filesystem again
4790 log "check files after shrinking the MDT and OST filesystems"
4791 for i in $(seq $((num_files + 10))); do
4792 file=$MOUNT/$tdir/$tfile-$i
4793 $CHECKSTAT -t file -s 1048576 $file ||
4794 error "(11) checkstat $file failed"
4797 # unmount and reformat the Lustre filesystem
4798 cleanup || error "(12) cleanup Lustre filesystem failed"
4799 combined_mgs_mds || stop_mgs || error "(13) stop mgs failed"
4800 reformat || error "(14) reformat Lustre filesystem failed"
4802 run_test 78 "run resize2fs on MDT and OST filesystems"
4804 test_79() { # LU-4227
4805 [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.5.59) ]] ||
4806 { skip "Need MDS version at least 2.5.59"; return 0; }
4808 local mdsdev1=$(mdsdevname 1)
4809 local mdsvdev1=$(mdsvdevname 1)
4810 local mdsdev2=$(mdsdevname 2)
4811 local mdsvdev2=$(mdsvdevname 2)
4812 local ostdev1=$(ostdevname 1)
4813 local ostvdev1=$(ostvdevname 1)
4814 local opts_mds1="$(mkfs_opts mds1 $mdsdev1) --reformat"
4815 local opts_mds2="$(mkfs_opts mds2 $mdsdev2) --reformat"
4816 local opts_ost1="$(mkfs_opts ost1 $ostdev1) --reformat"
4819 # remove --mgs/--mgsnode from mkfs.lustre options
4820 opts_mds1=$(echo $opts_mds1 | sed -e "s/--mgs//")
4822 mgsnode_opt=$(echo $opts_mds2 |
4823 awk '{ for ( i = 1; i < NF; i++ )
4824 if ( $i ~ "--mgsnode" ) { print $i; break } }')
4825 [ -n $mgsnode_opt ] &&
4826 opts_mds2=$(echo $opts_mds2 | sed -e "s/$mgsnode_opt//")
4828 mgsnode_opt=$(echo $opts_ost1 |
4829 awk '{ for ( i = 1; i < NF; i++ )
4830 if ( $i ~ "--mgsnode" ) { print $i; break } }')
4831 [ -n $mgsnode_opt ] &&
4832 opts_ost1=$(echo $opts_ost1 | sed -e "s/$mgsnode_opt//")
4834 # -MGS, format a mdt without --mgs option
4835 add mds1 $opts_mds1 $mdsdev1 $mdsvdev1 &&
4836 error "Must specify --mgs when formatting mdt combined with mgs"
4838 # +MGS, format a mdt/ost without --mgsnode option
4839 add mds1 $(mkfs_opts mds1 $mdsdev1) --reformat $mdsdev1 $mdsvdev1 \
4840 > /dev/null || error "start mds1 failed"
4841 add mds2 $opts_mds2 $mdsdev2 $mdsvdev2 &&
4842 error "Must specify --mgsnode when formatting a mdt"
4843 add ost1 $opts_ost1 $ostdev1 $ostvdev2 &&
4844 error "Must specify --mgsnode when formatting an ost"
4848 run_test 79 "format MDT/OST without mgs option (should return errors)"
4853 uuid=$(do_facet ost1 lctl get_param -n mgc.*.uuid)
4854 #define OBD_FAIL_MGS_PAUSE_TARGET_CON 0x906
4855 do_facet ost1 "lctl set_param fail_val=10 fail_loc=0x906"
4856 do_facet mgs "lctl set_param fail_val=10 fail_loc=0x906"
4857 do_facet mgs "lctl set_param -n mgs/MGS/evict_client $uuid"
4861 do_facet ost1 "lctl set_param fail_loc=0"
4864 run_test 80 "mgc import reconnect race"
4866 # Save the original values of $OSTCOUNT and $OSTINDEX$i.
4868 local new_ostcount=$1
4869 saved_ostcount=$OSTCOUNT
4870 OSTCOUNT=$new_ostcount
4874 for i in $(seq $OSTCOUNT); do
4876 eval saved_ostindex$i=${!index}
4881 # Restore the original values of $OSTCOUNT and $OSTINDEX$i.
4882 restore_ostindex() {
4887 for i in $(seq $OSTCOUNT); do
4888 index=saved_ostindex$i
4889 eval OSTINDEX$i=${!index}
4891 OSTCOUNT=$saved_ostcount
4896 # The main purpose of this test is to ensure the OST_INDEX_LIST functions as
4897 # expected. This test uses OST_INDEX_LIST to format OSTs with a randomly
4898 # assigned index and ensures we can mount such a formatted file system
4899 test_81() { # LU-4665
4900 [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.6.54) ]] ||
4901 { skip "Need MDS version at least 2.6.54" && return; }
4902 [[ $OSTCOUNT -ge 3 ]] || { skip_env "Need at least 3 OSTs" && return; }
4906 # Each time RANDOM is referenced, a random integer between 0 and 32767
4909 local saved_ostindex1=$OSTINDEX1
4910 for i in 65535 $((RANDOM + 65536)); do
4911 echo -e "\nFormat ost1 with --index=$i, should fail"
4913 if add ost1 $(mkfs_opts ost1 $(ostdevname 1)) --reformat \
4914 $(ostdevname 1) $(ostvdevname 1); then
4915 OSTINDEX1=$saved_ostindex1
4916 error "format ost1 with --index=$i should fail"
4919 OSTINDEX1=$saved_ostindex1
4923 # Format OSTs with random sparse indices.
4924 trap "restore_ostindex" EXIT
4925 echo -e "\nFormat $OSTCOUNT OSTs with sparse indices"
4926 OST_INDEX_LIST=[0,$((RANDOM * 2 % 65533 + 1)),65534] formatall
4928 # Setup and check Lustre filesystem.
4929 start_mgsmds || error "start_mgsmds failed"
4930 for i in $(seq $OSTCOUNT); do
4931 start ost$i $(ostdevname $i) $OST_MOUNT_OPTS ||
4932 error "start ost$i failed"
4935 mount_client $MOUNT || error "mount client $MOUNT failed"
4936 check_mount || error "check client $MOUNT failed"
4939 local max_easize=$($LCTL get_param -n llite.*.max_easize)
4940 [[ $max_easize -eq 128 ]] ||
4941 error "max_easize is $max_easize, should be 128 bytes"
4945 run_test 81 "sparse OST indexing"
4947 # Wait OSTs to be active on both client and MDT side.
4949 local cmd="$LCTL get_param -n lov.$FSNAME-clilov-*.target_obd |
4950 awk 'BEGIN {c = 0} /ACTIVE/{c += 1} END {printf \\\"%d\\\", c}'"
4951 wait_update $HOSTNAME "eval $cmd" $OSTCOUNT ||
4952 error "wait_update OSTs up on client failed"
4954 cmd="$LCTL get_param -n lod.$FSNAME-MDT*-*.target_obd | sort -u |
4955 awk 'BEGIN {c = 0} /ACTIVE/{c += 1} END {printf \\\"%d\\\", c}'"
4956 wait_update_facet $SINGLEMDS "eval $cmd" $OSTCOUNT ||
4957 error "wait_update OSTs up on MDT failed"
4960 # Here we exercise the stripe placement functionality on a file system that
4961 # has formatted the OST with a random index. With the file system the following
4962 # functionality is tested:
4964 # 1. Creating a new file with a specific stripe layout.
4966 # 2. Modifiy a existing empty file with a specific stripe layout.
4968 # 3. Ensure we fail to set the stripe layout of a file that already has one.
4970 # 4. If ost-index is defined we need to ensure it is the first entry in the
4971 # ost index list returned by lfs getstripe.
4973 # 5. Lastly ensure this functionality fails with directories.
4974 test_82a() { # LU-4665
4975 [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.6.54) ]] ||
4976 { skip "Need MDS version at least 2.6.54" && return; }
4977 [[ $OSTCOUNT -ge 3 ]] || { skip_env "Need at least 3 OSTs" && return; }
4983 # Format OSTs with random sparse indices.
4987 for i in $(seq $OSTCOUNT); do
4988 index=$((RANDOM * 2))
4989 ost_indices+=" $index"
4991 ost_indices=$(comma_list $ost_indices)
4993 trap "restore_ostindex" EXIT
4994 echo -e "\nFormat $OSTCOUNT OSTs with sparse indices $ost_indices"
4995 OST_INDEX_LIST=[$ost_indices] formatall
4997 # Setup Lustre filesystem.
4998 start_mgsmds || error "start_mgsmds failed"
4999 for i in $(seq $OSTCOUNT); do
5000 start ost$i $(ostdevname $i) $OST_MOUNT_OPTS ||
5001 error "start ost$i failed"
5004 mount_client $MOUNT || error "mount client $MOUNT failed"
5007 $LFS df $MOUNT || error "$LFS df $MOUNT failed"
5008 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
5010 # 1. If the file does not exist, new file will be created
5011 # with specified OSTs.
5012 local file=$DIR/$tdir/$tfile-1
5013 local cmd="$SETSTRIPE -o $ost_indices $file"
5015 eval $cmd || error "$cmd failed"
5016 check_stripe_count $file $OSTCOUNT
5017 check_obdidx $file $ost_indices
5018 dd if=/dev/urandom of=$file count=1 bs=1M > /dev/null 2>&1 ||
5019 error "write $file failed"
5021 # 2. If the file already exists and is an empty file, the file
5022 # will be attached with specified layout.
5023 file=$DIR/$tdir/$tfile-2
5024 mcreate $file || error "mcreate $file failed"
5025 cmd="$SETSTRIPE -o $ost_indices $file"
5027 eval $cmd || error "$cmd failed"
5028 dd if=/dev/urandom of=$file count=1 bs=1M > /dev/null 2>&1 ||
5029 error "write $file failed"
5030 check_stripe_count $file $OSTCOUNT
5031 check_obdidx $file $ost_indices
5033 # 3. If the file already has a valid layout attached, the command
5034 # should fail with EBUSY.
5036 eval $cmd && error "stripe is already set on $file, $cmd should fail"
5038 # 4. If [--stripe-index|-i <start_ost_idx>] is used, the index must
5039 # be in the OST indices list.
5040 local start_ost_idx=${ost_indices##*,}
5041 file=$DIR/$tdir/$tfile-3
5042 cmd="$SETSTRIPE -o $ost_indices -i $start_ost_idx $file"
5044 eval $cmd || error "$cmd failed"
5045 check_stripe_count $file $OSTCOUNT
5046 check_obdidx $file $ost_indices
5047 check_start_ost_idx $file $start_ost_idx
5049 file=$DIR/$tdir/$tfile-4
5051 cmd+=" -o $(exclude_items_from_list $ost_indices $start_ost_idx)"
5052 cmd+=" -i $start_ost_idx $file"
5054 eval $cmd && error "index $start_ost_idx should be in $ost_indices"
5056 # 5. Specifying OST indices for directory should fail with ENOSUPP.
5057 local dir=$DIR/$tdir/$tdir
5058 mkdir $dir || error "mkdir $dir failed"
5059 cmd="$SETSTRIPE -o $ost_indices $dir"
5061 eval $cmd && error "$cmd should fail, specifying OST indices" \
5062 "for directory is not supported"
5066 run_test 82a "specify OSTs for file (succeed) or directory (fail)"
5071 # Remove OSTs from a pool and destroy the pool.
5072 destroy_pool $ost_pool || true
5077 # Test 82b is run to ensure that if the user supplies a pool with a specific
5078 # stripe layout that it behaves proprerly. It should fail in the case that
5079 # the supplied OST index list points to OSTs not contained in the user
5081 test_82b() { # LU-4665
5082 [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.6.54) ]] ||
5083 { skip "Need MDS version at least 2.6.54" && return; }
5084 [[ $OSTCOUNT -ge 4 ]] || { skip_env "Need at least 4 OSTs" && return; }
5090 # Format OSTs with random sparse indices.
5094 for i in $(seq $OSTCOUNT); do
5095 index=$((RANDOM * 2))
5096 ost_indices+=" $index"
5098 ost_indices=$(comma_list $ost_indices)
5100 trap "restore_ostindex" EXIT
5101 echo -e "\nFormat $OSTCOUNT OSTs with sparse indices $ost_indices"
5102 OST_INDEX_LIST=[$ost_indices] formatall
5104 # Setup Lustre filesystem.
5105 start_mgsmds || error "start_mgsmds failed"
5106 for i in $(seq $OSTCOUNT); do
5107 start ost$i $(ostdevname $i) $OST_MOUNT_OPTS ||
5108 error "start ost$i failed"
5111 mount_client $MOUNT || error "mount client $MOUNT failed"
5113 $LFS df $MOUNT || error "$LFS df $MOUNT failed"
5114 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
5116 # Create a new pool and add OSTs into it.
5117 local ost_pool=$FSNAME.$TESTNAME
5118 create_pool $ost_pool || error "create OST pool $ost_pool failed"
5121 trap "cleanup_82b" EXIT
5123 local ost_idx_in_list=${ost_indices##*,}
5124 local ost_idx_in_pool=$(exclude_items_from_list $ost_indices \
5127 local ost_targets="$FSNAME-OST["
5128 for i in ${ost_idx_in_pool//,/ }; do
5129 ost_targets=$ost_targets$(printf "%04x," $i)
5131 ost_targets="${ost_targets%,}]"
5133 local ost_targets_uuid=$(for i in ${ost_idx_in_pool//,/ }; \
5134 do printf "$FSNAME-OST%04x_UUID\n" $i; done |
5135 sort -u | tr '\n' ' ')
5137 local cmd="$LCTL pool_add $ost_pool $ost_targets"
5138 do_facet mgs $cmd || error "$cmd failed"
5139 wait_update $HOSTNAME "$LCTL get_param -n lov.$FSNAME-*.pools.$TESTNAME|
5140 sort -u | tr '\n' ' ' " "$ost_targets_uuid" ||
5141 error "wait_update $ost_pool failed"
5142 pool_list $ost_pool || error "list OST pool $ost_pool failed"
5144 # If [--pool|-p <pool_name>] is set with [--ost-list|-o <ost_indices>],
5145 # then the OSTs must be the members of the pool.
5146 local file=$DIR/$tdir/$tfile
5147 cmd="$SETSTRIPE -p $ost_pool -o $ost_idx_in_list $file"
5149 eval $cmd && error "OST with index $ost_idx_in_list should be" \
5150 "in OST pool $ost_pool"
5152 # Only select OST $ost_idx_in_list from $ost_pool for file.
5153 ost_idx_in_list=${ost_idx_in_pool#*,}
5154 cmd="$SETSTRIPE -p $ost_pool -o $ost_idx_in_list $file"
5156 eval $cmd || error "$cmd failed"
5157 cmd="$GETSTRIPE $file"
5159 eval $cmd || error "$cmd failed"
5160 check_stripe_count $file 2
5161 check_obdidx $file $ost_idx_in_list
5162 dd if=/dev/urandom of=$file count=1 bs=1M > /dev/null 2>&1 ||
5163 error "write $file failed"
5167 run_test 82b "specify OSTs for file with --pool and --ost-list options"
5169 if ! combined_mgs_mds ; then
5175 # restore the values of MDSSIZE and OSTSIZE
5176 MDSSIZE=$STORED_MDSSIZE
5177 OSTSIZE=$STORED_OSTSIZE