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"
235 if ! combined_mgs_mds ; then
244 unload_modules_conf () {
245 if combined_mgs_mds || ! local_mode; then
246 unload_modules || return 1
251 stop_ost || return 202
252 stop_mds || return 201
253 unload_modules_conf || return 203
257 umount_client $MOUNT || return 200
258 cleanup_nocli || return $?
262 do_facet client "cp /etc/passwd $DIR/a" || return 71
263 do_facet client "rm $DIR/a" || return 72
264 # make sure lustre is actually mounted (touch will block,
265 # but grep won't, so do it after)
266 do_facet client "grep $MOUNT' ' /proc/mounts > /dev/null" || return 73
267 echo "setup single mount lustre success"
271 do_facet client "touch $DIR/a" || return 71
272 do_facet client "rm $DIR/a" || return 72
273 do_facet client "touch $DIR2/a" || return 73
274 do_facet client "rm $DIR2/a" || return 74
275 echo "setup double mount lustre success"
280 if [ "$ONLY" == "setup" ]; then
285 if [ "$ONLY" == "cleanup" ]; then
292 #create single point mountpoint
298 check_mount || return 41
301 run_test 0 "single mount setup"
304 start_mds || error "MDT start failed"
306 echo "start ost second time..."
307 start_ost && error "2nd OST start should fail"
308 mount_client $MOUNT || error "client start failed"
309 check_mount || return 42
312 run_test 1 "start up ost twice (should return errors)"
316 echo "start mds second time.."
317 start_mdt 1 && error "2nd MDT start should fail"
320 check_mount || return 43
323 run_test 2 "start up mds twice (should return err)"
327 #mount.lustre returns an error if already in mtab
328 mount_client $MOUNT && error "2nd client mount should fail"
329 check_mount || return 44
332 run_test 3 "mount client twice (should return err)"
336 touch $DIR/$tfile || return 85
340 # ok for ost to fail shutdown
341 if [ 202 -ne $eno ]; then
346 run_test 4 "force cleanup ost, then cleanup"
348 test_5a() { # was test_5
350 touch $DIR/$tfile || return 1
351 fuser -m -v $MOUNT && echo "$MOUNT is in use by user space process."
353 stop_mds -f || return 2
355 # cleanup may return an error from the failed
356 # disconnects; for now I'll consider this successful
357 # if all the modules have unloaded.
361 echo "killing umount"
362 kill -TERM $UMOUNT_PID
363 echo "waiting for umount to finish"
365 if grep " $MOUNT " /proc/mounts; then
366 echo "test 5: /proc/mounts after failed umount"
370 echo "killing umount"
371 kill -TERM $UMOUNT_PID
372 echo "waiting for umount to finish"
374 grep " $MOUNT " /proc/mounts && echo "test 5: /proc/mounts after second umount" && return 11
378 # stop_mds is a no-op here, and should not fail
379 cleanup_nocli || return $?
380 # df may have lingering entry
382 # mtab may have lingering entry
386 while [ "$WAIT" -ne "$MAX_WAIT" ]; do
388 grep -q $MOUNT" " /etc/mtab || break
389 echo "Waiting /etc/mtab updated ... "
390 WAIT=$(( WAIT + sleep))
392 [ "$WAIT" -eq "$MAX_WAIT" ] && error "/etc/mtab is not updated in $WAIT secs"
393 echo "/etc/mtab updated in $WAIT secs"
395 run_test 5a "force cleanup mds, then cleanup"
403 grep " $MOUNT " /etc/mtab && \
404 error false "unexpected entry in mtab before mount" && return 10
408 if ! combined_mgs_mds ; then
409 trap cleanup_5b EXIT ERR
414 [ -d $MOUNT ] || mkdir -p $MOUNT
415 mount_client $MOUNT && rc=1
416 grep " $MOUNT " /etc/mtab && \
417 error "$MOUNT entry in mtab after failed mount" && rc=11
419 # stop_mds is a no-op here, and should not fail
420 cleanup_nocli || rc=$?
421 if ! combined_mgs_mds ; then
426 run_test 5b "Try to start a client with no MGS (should return errs)"
429 grep " $MOUNT " /etc/mtab && \
430 error false "unexpected entry in mtab before mount" && return 10
435 [ -d $MOUNT ] || mkdir -p $MOUNT
436 local oldfs="${FSNAME}"
437 FSNAME="wrong.${FSNAME}"
438 mount_client $MOUNT || :
440 grep " $MOUNT " /etc/mtab && \
441 error "$MOUNT entry in mtab after failed mount" && rc=11
443 cleanup_nocli || rc=$?
446 run_test 5c "cleanup after failed mount (bug 2712) (should return errs)"
449 grep " $MOUNT " /etc/mtab && \
450 error false "unexpected entry in mtab before mount" && return 10
456 mount_client $MOUNT || rc=1
458 grep " $MOUNT " /etc/mtab && \
459 error "$MOUNT entry in mtab after unmount" && rc=11
462 run_test 5d "mount with ost down"
465 grep " $MOUNT " /etc/mtab && \
466 error false "unexpected entry in mtab before mount" && return 10
472 #define OBD_FAIL_PTLRPC_DELAY_SEND 0x506
473 do_facet client "lctl set_param fail_loc=0x80000506"
474 mount_client $MOUNT || echo "mount failed (not fatal)"
476 grep " $MOUNT " /etc/mtab && \
477 error "$MOUNT entry in mtab after unmount" && rc=11
480 run_test 5e "delayed connect, don't crash (bug 10268)"
483 if combined_mgs_mds ; then
484 skip "combined mgs and mds"
488 grep " $MOUNT " /etc/mtab && \
489 error false "unexpected entry in mtab before mount" && return 10
493 [ -d $MOUNT ] || mkdir -p $MOUNT
494 mount_client $MOUNT &
496 echo client_mount pid is $pid
500 if ! ps -f -p $pid >/dev/null; then
503 grep " $MOUNT " /etc/mtab && echo "test 5f: mtab after mount"
504 error "mount returns $rc, expected to hang"
513 # mount should succeed after start mds
516 [ $rc -eq 0 ] || error "mount returned $rc"
517 grep " $MOUNT " /etc/mtab && echo "test 5f: mtab after mount"
521 run_test 5f "mds down, cleanup after failed mount (bug 2712)"
526 mount_client ${MOUNT} || return 87
527 touch $DIR/a || return 86
530 run_test 6 "manual umount, then mount again"
535 cleanup_nocli || return $?
537 run_test 7 "manual umount, then cleanup"
542 check_mount2 || return 45
543 umount_client $MOUNT2
546 run_test 8 "double mount setup"
551 do_facet ost1 lctl set_param debug=\'inode trace\' || return 1
552 do_facet ost1 lctl set_param subsystem_debug=\'mds ost\' || return 1
554 CHECK_PTLDEBUG="`do_facet ost1 lctl get_param -n debug`"
555 if [ "$CHECK_PTLDEBUG" ] && { \
556 [ "$CHECK_PTLDEBUG" = "trace inode warning error emerg console" ] ||
557 [ "$CHECK_PTLDEBUG" = "trace inode" ]; }; then
558 echo "lnet.debug success"
560 echo "lnet.debug: want 'trace inode', have '$CHECK_PTLDEBUG'"
563 CHECK_SUBSYS="`do_facet ost1 lctl get_param -n subsystem_debug`"
564 if [ "$CHECK_SUBSYS" ] && [ "$CHECK_SUBSYS" = "mds ost" ]; then
565 echo "lnet.subsystem_debug success"
567 echo "lnet.subsystem_debug: want 'mds ost', have '$CHECK_SUBSYS'"
570 stop_ost || return $?
572 run_test 9 "test ptldebug and subsystem for mkfs"
580 do_facet $facet "test -b $dev" || rc=1
581 if [[ "$size" ]]; then
582 local in=$(do_facet $facet "dd if=$dev of=/dev/null bs=1k count=1 skip=$size 2>&1" |\
583 awk '($3 == "in") { print $1 }')
584 [[ $in = "1+0" ]] || rc=1
590 # Test 16 was to "verify that lustre will correct the mode of OBJECTS".
591 # But with new MDS stack we don't care about the mode of local objects
592 # anymore, so this test is removed. See bug 22944 for more details.
596 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
597 skip "Only applicable to ldiskfs-based MDTs"
602 check_mount || return 41
605 echo "Remove mds config log"
606 if ! combined_mgs_mds ; then
610 do_facet mgs "$DEBUGFS -w -R 'unlink CONFIGS/$FSNAME-MDT0000' \
611 $(mgsdevname) || return \$?" || return $?
613 if ! combined_mgs_mds ; then
618 start_mds && return 42
621 run_test 17 "Verify failed mds_postsetup won't fail assertion (2936) (should return errs)"
624 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
625 skip "Only applicable to ldiskfs-based MDTs"
629 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
634 # check if current MDSSIZE is large enough
635 [ $MDSSIZE -ge $MIN ] && OK=1 && myMDSSIZE=$MDSSIZE && \
636 log "use MDSSIZE=$MDSSIZE"
638 # check if the global config has a large enough MDSSIZE
639 [ -z "$OK" -a ! -z "$STORED_MDSSIZE" ] && [ $STORED_MDSSIZE -ge $MIN ] && \
640 OK=1 && myMDSSIZE=$STORED_MDSSIZE && \
641 log "use STORED_MDSSIZE=$STORED_MDSSIZE"
643 # check if the block device is large enough
644 is_blkdev $SINGLEMDS $MDSDEV $MIN
645 local large_enough=$?
646 if [ -n "$OK" ]; then
647 [ $large_enough -ne 0 ] && OK=""
649 [ $large_enough -eq 0 ] && OK=1 && myMDSSIZE=$MIN &&
650 log "use device $MDSDEV with MIN=$MIN"
653 # check if a loopback device has enough space for fs metadata (5%)
655 if [ -z "$OK" ]; then
656 local SPACE=$(do_facet $SINGLEMDS "[ -f $MDSDEV -o ! -e $MDSDEV ] && df -P \\\$(dirname $MDSDEV)" |
657 awk '($1 != "Filesystem") {print $4}')
658 ! [ -z "$SPACE" ] && [ $SPACE -gt $((MIN / 20)) ] && \
659 OK=1 && myMDSSIZE=$MIN && \
660 log "use file $MDSDEV with MIN=$MIN"
663 [ -z "$OK" ] && skip_env "$MDSDEV too small for ${MIN}kB MDS" && return
666 echo "mount mds with large journal..."
668 local OLD_MDSSIZE=$MDSSIZE
672 echo "mount lustre system..."
674 check_mount || return 41
676 echo "check journal size..."
677 local FOUNDSIZE=$(do_facet $SINGLEMDS "$DEBUGFS -c -R 'stat <8>' $MDSDEV" | awk '/Size: / { print $NF; exit;}')
678 if [ $FOUNDSIZE -gt $((32 * 1024 * 1024)) ]; then
679 log "Success: mkfs creates large journals. Size: $((FOUNDSIZE >> 20))M"
681 error "expected journal size > 32M, found $((FOUNDSIZE >> 20))M"
689 run_test 18 "check mkfs creates large journals"
692 start_mds || return 1
693 stop_mds -f || return 2
695 run_test 19a "start/stop MDS without OSTs"
698 start_ost || return 1
699 stop_ost -f || return 2
701 run_test 19b "start/stop OSTs without MDS"
704 # first format the ost/mdt
708 check_mount || return 43
710 remount_client ro $MOUNT || return 44
711 touch $DIR/$tfile && echo "$DIR/$tfile created incorrectly" && return 45
712 [ -e $DIR/$tfile ] && echo "$DIR/$tfile exists incorrectly" && return 46
713 remount_client rw $MOUNT || return 47
715 [ ! -f $DIR/$tfile ] && echo "$DIR/$tfile missing" && return 48
716 MCNT=`grep -c $MOUNT /etc/mtab`
717 [ "$MCNT" -ne 1 ] && echo "$MOUNT in /etc/mtab $MCNT times" && return 49
722 run_test 20 "remount ro,rw mounts work and doesn't break /etc/mtab"
727 wait_osc_import_state mds ost FULL
731 run_test 21a "start mds before ost, stop ost first"
736 wait_osc_import_state mds ost FULL
740 run_test 21b "start ost before mds, stop mds first"
746 wait_osc_import_state mds ost2 FULL
750 #writeconf to remove all ost2 traces for subsequent tests
751 writeconf_or_reformat
753 run_test 21c "start mds between two osts, stop mds last"
756 if combined_mgs_mds ; then
757 skip "need separate mgs device" && return 0
767 wait_osc_import_state mds ost2 FULL
773 #writeconf to remove all ost2 traces for subsequent tests
774 writeconf_or_reformat
777 run_test 21d "start mgs then ost and then mds"
782 echo Client mount with ost in logs, but none running
784 # wait until mds connected to ost and open client connection
785 wait_osc_import_state mds ost FULL
788 # check_mount will block trying to contact ost
789 mcreate $DIR/$tfile || return 40
790 rm -f $DIR/$tfile || return 42
794 echo Client mount with a running ost
797 # if gss enabled, wait full time to let connection from
798 # mds to ost be established, due to the mismatch between
799 # initial connect timeout and gss context negotiation timeout.
800 # This perhaps could be remove after AT landed.
801 echo "sleep $((TIMEOUT + TIMEOUT + TIMEOUT))s"
802 sleep $((TIMEOUT + TIMEOUT + TIMEOUT))
805 wait_osc_import_state mds ost FULL
806 wait_osc_import_state client ost FULL
807 check_mount || return 41
812 run_test 22 "start a client before osts (should return errs)"
814 test_23a() { # was test_23
818 # force down client so that recovering mds waits for reconnect
819 local running=$(grep -c $MOUNT /proc/mounts) || true
820 if [ $running -ne 0 ]; then
821 echo "Stopping client $MOUNT (opts: -f)"
825 # enter recovery on mds
827 # try to start a new client
828 mount_client $MOUNT &
830 MOUNT_PID=$(ps -ef | grep "t lustre" | grep -v grep | awk '{print $2}')
831 MOUNT_LUSTRE_PID=`ps -ef | grep mount.lustre | grep -v grep | awk '{print $2}'`
832 echo mount pid is ${MOUNT_PID}, mount.lustre pid is ${MOUNT_LUSTRE_PID}
834 ps --ppid $MOUNT_LUSTRE_PID
835 echo "waiting for mount to finish"
837 # "ctrl-c" sends SIGINT but it usually (in script) does not work on child process
838 # SIGTERM works but it does not spread to offspring processses
839 kill -s TERM $MOUNT_PID
840 kill -s TERM $MOUNT_LUSTRE_PID
841 # we can not wait $MOUNT_PID because it is not a child of this shell
847 while [ "$WAIT" -lt "$MAX_WAIT" ]; do
849 PID1=$(ps -ef | awk '{print $2}' | grep -w $MOUNT_PID)
850 PID2=$(ps -ef | awk '{print $2}' | grep -w $MOUNT_LUSTRE_PID)
853 [ -z "$PID1" -a -z "$PID2" ] && break
854 echo "waiting for mount to finish ... "
855 WAIT=$(( WAIT + sleep))
857 if [ "$WAIT" -eq "$MAX_WAIT" ]; then
858 error "MOUNT_PID $MOUNT_PID and "\
859 "MOUNT_LUSTRE_PID $MOUNT_LUSTRE_PID still not killed in $WAIT secs"
862 stop_mds || error "stopping MDSes failed"
863 stop_ost || error "stopping OSSes failed"
865 run_test 23a "interrupt client during recovery mount delay"
870 test_23b() { # was test_23
873 # Simulate -EINTR during mount OBD_FAIL_LDLM_CLOSE_THREAD
874 lctl set_param fail_loc=0x80000313
878 run_test 23b "Simulate -EINTR during mount"
880 fs2mds_HOST=$mds_HOST
881 fs2ost_HOST=$ost_HOST
883 MDSDEV1_2=$fs2mds_DEV
884 OSTDEV1_2=$fs2ost_DEV
885 OSTDEV2_2=$fs3ost_DEV
889 echo "umount $MOUNT2 ..."
890 umount $MOUNT2 || true
891 echo "stopping fs2mds ..."
892 stop fs2mds -f || true
893 echo "stopping fs2ost ..."
894 stop fs2ost -f || true
898 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
900 if [ -z "$fs2ost_DEV" -o -z "$fs2mds_DEV" ]; then
901 is_blkdev $SINGLEMDS $MDSDEV && \
902 skip_env "mixed loopback and real device not working" && return
905 [ -n "$ost1_HOST" ] && fs2ost_HOST=$ost1_HOST
907 local fs2mdsdev=$(mdsdevname 1_2)
908 local fs2ostdev=$(ostdevname 1_2)
909 local fs2mdsvdev=$(mdsvdevname 1_2)
910 local fs2ostvdev=$(ostvdevname 1_2)
912 # test 8-char fsname as well
913 local FSNAME2=test1234
915 add fs2mds $(mkfs_opts mds1 ${fs2mdsdev} ) --nomgs --mgsnode=$MGSNID \
916 --fsname=${FSNAME2} --reformat $fs2mdsdev $fs2mdsvdev || exit 10
918 add fs2ost $(mkfs_opts ost1 ${fs2ostdev}) --fsname=${FSNAME2} \
919 --reformat $fs2ostdev $fs2ostvdev || exit 10
922 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS && trap cleanup_fs2 EXIT INT
923 start fs2ost $fs2ostdev $OST_MOUNT_OPTS
925 $MOUNT_CMD $MGSNID:/${FSNAME2} $MOUNT2 || return 1
927 check_mount || return 2
928 # files written on 1 should not show up on 2
929 cp /etc/passwd $DIR/$tfile
931 [ -e $MOUNT2/$tfile ] && error "File bleed" && return 7
934 cp /etc/passwd $MOUNT2/b || return 3
935 rm $MOUNT2/b || return 4
936 # 2 is actually mounted
937 grep $MOUNT2' ' /proc/mounts > /dev/null || return 5
939 facet_failover fs2mds
940 facet_failover fs2ost
943 # the MDS must remain up until last MDT
945 MDS=$(do_facet $SINGLEMDS "lctl get_param -n devices" | awk '($3 ~ "mdt" && $4 ~ "MDT") { print $4 }' | head -1)
946 [ -z "$MDS" ] && error "No MDT" && return 8
948 cleanup_nocli || return 6
950 run_test 24a "Multiple MDTs on a single node"
953 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
955 if [ -z "$fs2mds_DEV" ]; then
956 local dev=${SINGLEMDS}_dev
958 is_blkdev $SINGLEMDS $MDSDEV && \
959 skip_env "mixed loopback and real device not working" && return
962 local fs2mdsdev=$(mdsdevname 1_2)
963 local fs2mdsvdev=$(mdsvdevname 1_2)
965 add fs2mds $(mkfs_opts mds1 ${fs2mdsdev} ) --mgs --fsname=${FSNAME}2 \
966 --reformat $fs2mdsdev $fs2mdsvdev || exit 10
968 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS && return 2
972 run_test 24b "Multiple MGSs on a single node (should return err)"
976 check_mount || return 2
977 local MODULES=$($LCTL modules | awk '{ print $2 }')
978 rmmod $MODULES 2>/dev/null || true
981 run_test 25 "Verify modules are referenced"
985 # we need modules before mount for sysctl, so make sure...
986 do_facet $SINGLEMDS "lsmod | grep -q lustre || modprobe lustre"
987 #define OBD_FAIL_MDS_FS_SETUP 0x135
988 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000135"
989 start_mds && echo MDS started && return 1
990 lctl get_param -n devices
991 DEVS=$(lctl get_param -n devices | egrep -v MG | wc -l)
992 [ $DEVS -gt 0 ] && return 2
993 # start mds to drop writeconf setting
994 start_mds || return 3
996 unload_modules_conf || return $?
998 run_test 26 "MDT startup failure cleans LOV (should return errs)"
1001 start_ost || return 1
1002 start_mds || return 2
1003 echo "Requeue thread should have started: "
1004 ps -e | grep ll_cfg_requeue
1005 set_conf_param_and_check ost1 \
1006 "lctl get_param -n obdfilter.$FSNAME-OST0000.client_cache_seconds" \
1007 "$FSNAME-OST0000.ost.client_cache_seconds" || return 3
1010 run_test 27a "Reacquire MGS lock if OST started first"
1015 local device=$(do_facet $SINGLEMDS "lctl get_param -n devices" |
1016 awk '($3 ~ "mdt" && $4 ~ "MDT0000") { print $4 }')
1018 facet_failover $SINGLEMDS
1019 set_conf_param_and_check $SINGLEMDS \
1020 "lctl get_param -n mdt.$device.identity_acquire_expire" \
1021 "$device.mdt.identity_acquire_expire" || return 3
1022 set_conf_param_and_check client \
1023 "lctl get_param -n mdc.$device-mdc-*.max_rpcs_in_flight"\
1024 "$device.mdc.max_rpcs_in_flight" || return 4
1028 run_test 27b "Reacquire MGS lock after failover"
1032 TEST="lctl get_param -n llite.$FSNAME-*.max_read_ahead_whole_mb"
1033 PARAM="$FSNAME.llite.max_read_ahead_whole_mb"
1035 FINAL=$(($ORIG + 1))
1036 set_conf_param_and_check client "$TEST" "$PARAM" $FINAL || return 3
1037 FINAL=$(($FINAL + 1))
1038 set_conf_param_and_check client "$TEST" "$PARAM" $FINAL || return 4
1039 umount_client $MOUNT || return 200
1042 if [ $RESULT -ne $FINAL ]; then
1043 echo "New config not seen: wanted $FINAL got $RESULT"
1046 echo "New config success: got $RESULT"
1048 set_conf_param_and_check client "$TEST" "$PARAM" $ORIG || return 5
1051 run_test 28 "permanent parameter setting"
1053 test_28a() { # LU-4221
1054 [[ $(lustre_version_code ost1) -ge $(version_code 2.5.52) ]] ||
1055 { skip "Need OST version at least 2.5.52" && return 0; }
1056 [ "$(facet_fstype ost1)" = "zfs" ] &&
1057 skip "LU-4221: no such proc params for ZFS OSTs" && return
1064 local device="$FSNAME-OST0000"
1068 # In this test we will set three kinds of proc parameters with
1070 # 1. the ones moved from the OFD to the OSD, and only their
1071 # symlinks kept in obdfilter
1072 # 2. non-symlink ones in the OFD
1073 # 3. non-symlink ones in the OSD
1076 # prepare a symlink parameter in the OFD
1077 name="writethrough_cache_enable"
1078 param="$device.ost.$name"
1079 cmd="$LCTL get_param -n obdfilter.$device.$name"
1081 # conf_param the symlink parameter in the OFD
1082 old=$(do_facet ost1 $cmd)
1083 new=$(((old + 1) % 2))
1084 set_conf_param_and_check ost1 "$cmd" "$param" $new ||
1085 error "lctl conf_param $device.ost.$param=$new failed"
1087 # conf_param the target parameter in the OSD
1088 param="$device.osd.$name"
1089 cmd="$LCTL get_param -n osd-*.$device.$name"
1090 set_conf_param_and_check ost1 "$cmd" "$param" $old ||
1091 error "lctl conf_param $device.osd.$param=$old failed"
1094 # prepare a non-symlink parameter in the OFD
1095 name="client_cache_seconds"
1096 param="$device.ost.$name"
1097 cmd="$LCTL get_param -n obdfilter.$device.$name"
1099 # conf_param the parameter in the OFD
1100 old=$(do_facet ost1 $cmd)
1102 set_conf_param_and_check ost1 "$cmd" "$param" $new ||
1103 error "lctl conf_param $device.ost.$param=$new failed"
1104 set_conf_param_and_check ost1 "$cmd" "$param" $old ||
1105 error "lctl conf_param $device.ost.$param=$old failed"
1108 # prepare a non-symlink parameter in the OSD
1109 name="lma_self_repair"
1110 param="$device.osd.$name"
1111 cmd="$LCTL get_param -n osd-*.$device.$name"
1113 # conf_param the parameter in the OSD
1114 old=$(do_facet ost1 $cmd)
1115 new=$(((old + 1) % 2))
1116 set_conf_param_and_check ost1 "$cmd" "$param" $new ||
1117 error "lctl conf_param $device.osd.$param=$new failed"
1118 set_conf_param_and_check ost1 "$cmd" "$param" $old ||
1119 error "lctl conf_param $device.osd.$param=$old failed"
1123 run_test 28a "set symlink parameters permanently with conf_param"
1126 [ "$OSTCOUNT" -lt "2" ] && skip_env "$OSTCOUNT < 2, skipping" && return
1127 setup > /dev/null 2>&1
1131 local PARAM="$FSNAME-OST0001.osc.active"
1132 local PROC_ACT="osc.$FSNAME-OST0001-osc-[^M]*.active"
1133 local PROC_UUID="osc.$FSNAME-OST0001-osc-[^M]*.ost_server_uuid"
1135 ACTV=$(lctl get_param -n $PROC_ACT)
1137 set_conf_param_and_check client \
1138 "lctl get_param -n $PROC_ACT" "$PARAM" $DEAC || return 2
1139 # also check ost_server_uuid status
1140 RESULT=$(lctl get_param -n $PROC_UUID | grep DEACTIV)
1141 if [ -z "$RESULT" ]; then
1142 echo "Live client not deactivated: $(lctl get_param -n $PROC_UUID)"
1145 echo "Live client success: got $RESULT"
1149 for num in $(seq $MDSCOUNT); do
1150 local mdtosc=$(get_mdtosc_proc_path mds${num} $FSNAME-OST0001)
1151 local MPROC="osc.$mdtosc.active"
1156 RESULT=$(do_facet mds${num} " lctl get_param -n $MPROC")
1157 [ ${PIPESTATUS[0]} = 0 ] || error "Can't read $MPROC"
1158 if [ $RESULT -eq $DEAC ]; then
1159 echo -n "MDT deactivated also after"
1160 echo "$WAIT sec (got $RESULT)"
1164 if [ $WAIT -eq $MAX ]; then
1165 echo -n "MDT not deactivated: wanted $DEAC"
1169 echo "Waiting $(($MAX - $WAIT))secs for MDT deactivated"
1172 # test new client starts deactivated
1173 umount_client $MOUNT || return 200
1175 RESULT=$(lctl get_param -n $PROC_UUID | grep DEACTIV | grep NEW)
1176 if [ -z "$RESULT" ]; then
1177 echo "New client not deactivated from start: $(lctl get_param -n $PROC_UUID)"
1180 echo "New client success: got $RESULT"
1183 # make sure it reactivates
1184 set_conf_param_and_check client \
1185 "lctl get_param -n $PROC_ACT" "$PARAM" $ACTV || return 6
1187 umount_client $MOUNT
1190 #writeconf to remove all ost2 traces for subsequent tests
1191 writeconf_or_reformat
1193 run_test 29 "permanently remove an OST"
1198 echo Big config llog
1199 TEST="lctl get_param -n llite.$FSNAME-*.max_read_ahead_whole_mb"
1201 LIST=(1 2 3 4 5 4 3 2 1 2 3 4 5 4 3 2 1 2 3 4 5)
1202 for i in ${LIST[@]}; do
1203 set_conf_param_and_check client "$TEST" \
1204 "$FSNAME.llite.max_read_ahead_whole_mb" $i || return 3
1206 # make sure client restart still works
1207 umount_client $MOUNT
1208 mount_client $MOUNT || return 4
1209 [ "$($TEST)" -ne "$i" ] && error "Param didn't stick across restart $($TEST) != $i"
1212 echo Erase parameter setting
1213 do_facet mgs "$LCTL conf_param -d $FSNAME.llite.max_read_ahead_whole_mb" || return 6
1214 umount_client $MOUNT
1215 mount_client $MOUNT || return 6
1217 echo "deleted (default) value=$FINAL, orig=$ORIG"
1218 # assumes this parameter started at the default value
1219 [ "$FINAL" -eq "$ORIG" ] || fail "Deleted value=$FINAL, orig=$ORIG"
1223 run_test 30a "Big config llog and conf_param deletion"
1228 # Make a fake nid. Use the OST nid, and add 20 to the least significant
1229 # numerical part of it. Hopefully that's not already a failover address for
1231 OSTNID=$(do_facet ost1 "$LCTL get_param nis" | tail -1 | awk '{print $1}')
1232 ORIGVAL=$(echo $OSTNID | egrep -oi "[0-9]*@")
1233 NEWVAL=$((($(echo $ORIGVAL | egrep -oi "[0-9]*") + 20) % 256))
1234 NEW=$(echo $OSTNID | sed "s/$ORIGVAL/$NEWVAL@/")
1235 echo "Using fake nid $NEW"
1237 TEST="$LCTL get_param -n osc.$FSNAME-OST0000-osc-[^M]*.import |
1238 grep failover_nids | sed -n 's/.*\($NEW\).*/\1/p'"
1239 set_conf_param_and_check client "$TEST" \
1240 "$FSNAME-OST0000.failover.node" $NEW ||
1241 error "didn't add failover nid $NEW"
1242 NIDS=$($LCTL get_param -n osc.$FSNAME-OST0000-osc-[^M]*.import |
1245 # The NIDS value is the failover nid strings and "[" and "]". So
1246 # we need to subtract the space taken by the delimiters. This has
1247 # changed from earlier version of Lustre but this test is run only
1248 # locally so this change will not break interop. See LU-3386
1249 NIDCOUNT=$(($(echo "$NIDS" | wc -w) - 3))
1250 echo "should have 2 failover nids: $NIDCOUNT"
1251 [ $NIDCOUNT -eq 2 ] || error "Failover nid not added"
1252 do_facet mgs "$LCTL conf_param -d $FSNAME-OST0000.failover.node" ||
1253 error "conf_param delete failed"
1254 umount_client $MOUNT
1255 mount_client $MOUNT || return 3
1257 NIDS=$($LCTL get_param -n osc.$FSNAME-OST0000-osc-[^M]*.import |
1260 NIDCOUNT=$(($(echo "$NIDS" | wc -w) - 3))
1261 echo "only 1 final nid should remain: $NIDCOUNT"
1262 [ $NIDCOUNT -eq 1 ] || error "Failover nids not removed"
1266 run_test 30b "Remove failover nids"
1268 test_31() { # bug 10734
1269 # ipaddr must not exist
1270 $MOUNT_CMD 4.3.2.1@tcp:/lustre $MOUNT || true
1273 run_test 31 "Connect to non-existent node (shouldn't crash)"
1277 T32_BLIMIT=20480 # Kbytes
1281 # This is not really a test but a tool to create new disk
1282 # image tarballs for the upgrade tests.
1284 # Disk image tarballs should be created on single-node
1285 # clusters by running this test with default configurations
1286 # plus a few mandatory environment settings that are verified
1287 # at the beginning of the test.
1289 test_32newtarball() {
1293 local tmp=$TMP/t32_image_create
1295 if [ $FSNAME != t32fs -o $MDSCOUNT -ne 1 -o \
1296 \( -z "$MDSDEV" -a -z "$MDSDEV1" \) -o $OSTCOUNT -ne 1 -o \
1297 -z "$OSTDEV1" ]; then
1298 error "Needs FSNAME=t32fs MDSCOUNT=1 MDSDEV1=<nonexistent_file>" \
1299 "(or MDSDEV, in the case of b1_8) OSTCOUNT=1" \
1300 "OSTDEV1=<nonexistent_file>"
1304 echo "Found stale $tmp"
1309 tar cf - -C $src . | tar xf - -C $tmp/src
1310 dd if=/dev/zero of=$tmp/src/t32_qf_old bs=1M \
1311 count=$(($T32_BLIMIT / 1024 / 2))
1312 chown $T32_QID.$T32_QID $tmp/src/t32_qf_old
1314 # format ost with comma-separated NIDs to verify LU-4460
1315 local failnid="$(h2$NETTYPE 1.2.3.4),$(h2$NETTYPE 4.3.2.1)"
1316 MGSNID="$MGSNID,$MGSNID" OSTOPT="--failnode=$failnid" formatall
1320 [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.3.50) ] &&
1321 $LFS quotacheck -ug /mnt/$FSNAME
1322 $LFS setquota -u $T32_QID -b 0 -B $T32_BLIMIT -i 0 -I $T32_ILIMIT \
1325 tar cf - -C $tmp/src . | tar xf - -C /mnt/$FSNAME
1332 ls -Rni --time-style=+%s >$tmp/img/list
1333 find . ! -name .lustre -type f -exec sha1sum {} \; |
1334 sort -k 2 >$tmp/img/sha1sums
1336 $LCTL get_param -n version | head -n 1 |
1337 sed -e 's/^lustre: *//' >$tmp/img/commit
1339 [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.3.50) ] &&
1340 $LFS quotaon -ug /mnt/$FSNAME
1341 $LFS quota -u $T32_QID -v /mnt/$FSNAME
1342 $LFS quota -v -u $T32_QID /mnt/$FSNAME |
1343 awk 'BEGIN { num='1' } { if ($1 == "'/mnt/$FSNAME'") \
1344 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1345 | tr -d "*" > $tmp/img/bspace
1346 $LFS quota -v -u $T32_QID /mnt/$FSNAME |
1347 awk 'BEGIN { num='5' } { if ($1 == "'/mnt/$FSNAME'") \
1348 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1349 | tr -d "*" > $tmp/img/ispace
1354 find -type f -exec sha1sum {} \; | sort -k 2 >$tmp/sha1sums.src
1357 if ! diff -u $tmp/sha1sums.src $tmp/img/sha1sums; then
1358 echo "Data verification failed"
1361 uname -r >$tmp/img/kernel
1362 uname -m >$tmp/img/arch
1364 mv ${MDSDEV1:-$MDSDEV} $tmp/img
1365 mv $OSTDEV1 $tmp/img
1367 version=$(sed -e 's/\(^[0-9]\+\.[0-9]\+\)\(.*$\)/\1/' $tmp/img/commit |
1368 sed -e 's/\./_/g') # E.g., "1.8.7" -> "1_8"
1371 tar cjvf $dst/disk$version-$(facet_fstype $SINGLEMDS).tar.bz2 -S *
1376 #run_test 32newtarball "Create a new test_32 disk image tarball for this version"
1379 # The list of applicable tarballs is returned via the caller's
1380 # variable "tarballs".
1383 local node=$(facet_active_host $SINGLEMDS)
1384 local r="do_node $node"
1386 if [ "$CLIENTONLY" ]; then
1387 skip "Client-only testing"
1391 if ! $r which $TUNEFS; then
1392 skip_env "tunefs.lustre required on $node"
1396 local IMGTYPE=$(facet_fstype $SINGLEMDS)
1398 tarballs=$($r find $RLUSTRE/tests -maxdepth 1 -name \'disk*-$IMGTYPE.tar.bz2\')
1400 if [ -z "$tarballs" ]; then
1401 skip "No applicable tarballs found"
1406 t32_test_cleanup() {
1408 local fstype=$(facet_fstype $SINGLEMDS)
1411 if $shall_cleanup_lustre; then
1412 umount $tmp/mnt/lustre || rc=$?
1414 if $shall_cleanup_mdt; then
1415 $r umount -d $tmp/mnt/mdt || rc=$?
1417 if $shall_cleanup_mdt1; then
1418 $r umount -d $tmp/mnt/mdt1 || rc=$?
1420 if $shall_cleanup_ost; then
1421 $r umount -d $tmp/mnt/ost || rc=$?
1426 if [ $fstype == "zfs" ]; then
1427 $r $ZPOOL destroy t32fs-mdt1 || rc=$?
1428 $r $ZPOOL destroy t32fs-ost1 || rc=$?
1433 t32_bits_per_long() {
1435 # Yes, this is not meant to be perfect.
1445 t32_reload_modules() {
1447 local all_removed=false
1450 while ((i < 20)); do
1451 echo "Unloading modules on $node: Attempt $i"
1452 do_rpc_nodes $node $LUSTRE_RMMOD $(facet_fstype $SINGLEMDS) &&
1454 do_rpc_nodes $node check_mem_leak || return 1
1455 if $all_removed; then
1456 do_rpc_nodes $node load_modules
1462 echo "Unloading modules on $node: Given up"
1466 t32_wait_til_devices_gone() {
1472 echo wait for devices to go
1473 while ((i < 20)); do
1474 devices=$(do_rpc_nodes $node $LCTL device_list | wc -l)
1475 loops=$(do_rpc_nodes $node losetup -a | grep -c t32)
1476 ((devices == 0 && loops == 0)) && return 0
1480 echo "waiting for dev on $node: dev $devices loop $loops given up"
1481 do_rpc_nodes $node "losetup -a"
1482 do_rpc_nodes $node "$LCTL devices_list"
1486 t32_verify_quota() {
1490 local fstype=$(facet_fstype $SINGLEMDS)
1494 $LFS quota -u $T32_QID -v $mnt
1496 qval=$($LFS quota -v -u $T32_QID $mnt |
1497 awk 'BEGIN { num='1' } { if ($1 == "'$mnt'") \
1498 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1500 [ $qval -eq $img_bspace ] || {
1501 echo "bspace, act:$qval, exp:$img_bspace"
1505 qval=$($LFS quota -v -u $T32_QID $mnt |
1506 awk 'BEGIN { num='5' } { if ($1 == "'$mnt'") \
1507 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1509 [ $qval -eq $img_ispace ] || {
1510 echo "ispace, act:$qval, exp:$img_ispace"
1514 qval=$($LFS quota -v -u $T32_QID $mnt |
1515 awk 'BEGIN { num='3' } { if ($1 == "'$mnt'") \
1516 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1518 [ $qval -eq $T32_BLIMIT ] || {
1519 echo "blimit, act:$qval, exp:$T32_BLIMIT"
1523 qval=$($LFS quota -v -u $T32_QID $mnt |
1524 awk 'BEGIN { num='7' } { if ($1 == "'$mnt'") \
1525 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1527 [ $qval -eq $T32_ILIMIT ] || {
1528 echo "ilimit, act:$qval, exp:$T32_ILIMIT"
1532 do_node $node $LCTL conf_param $fsname.quota.mdt=ug
1533 cmd="$LCTL get_param -n osd-$fstype.$fsname-MDT0000"
1534 cmd=$cmd.quota_slave.enabled
1535 wait_update $node "$cmd" "ug" || {
1536 echo "Enable mdt quota failed"
1540 do_node $node $LCTL conf_param $fsname.quota.ost=ug
1541 cmd="$LCTL get_param -n osd-$fstype.$fsname-OST0000"
1542 cmd=$cmd.quota_slave.enabled
1543 wait_update $node "$cmd" "ug" || {
1544 echo "Enable ost quota failed"
1549 runas -u $T32_QID -g $T32_QID dd if=/dev/zero of=$mnt/t32_qf_new \
1550 bs=1M count=$(($T32_BLIMIT / 1024)) oflag=sync && {
1551 echo "Write succeed, but expect -EDQUOT"
1554 rm -f $mnt/t32_qf_new
1556 runas -u $T32_QID -g $T32_QID createmany -m $mnt/t32_qf_ \
1558 echo "Create succeed, but expect -EDQUOT"
1561 unlinkmany $mnt/t32_qf_ $T32_ILIMIT
1569 local dne_upgrade=${dne_upgrade:-"no"}
1570 local ff_convert=${ff_convert:-"no"}
1571 local shall_cleanup_mdt=false
1572 local shall_cleanup_mdt1=false
1573 local shall_cleanup_ost=false
1574 local shall_cleanup_lustre=false
1575 local node=$(facet_active_host $SINGLEMDS)
1576 local r="do_node $node"
1577 local node2=$(facet_active_host mds2)
1585 local nid=$($r $LCTL list_nids | head -1)
1591 local fstype=$(facet_fstype $SINGLEMDS)
1592 local mdt_dev=$tmp/mdt
1593 local ost_dev=$tmp/ost
1595 trap 'trap - RETURN; t32_test_cleanup' RETURN
1597 mkdir -p $tmp/mnt/lustre
1598 $r mkdir -p $tmp/mnt/{mdt,ost}
1599 $r tar xjvf $tarball -S -C $tmp || {
1600 error_noexit "Unpacking the disk image tarball"
1603 img_commit=$($r cat $tmp/commit)
1604 img_kernel=$($r cat $tmp/kernel)
1605 img_arch=$($r cat $tmp/arch)
1606 img_bspace=$($r cat $tmp/bspace)
1607 img_ispace=$($r cat $tmp/ispace)
1608 echo "Upgrading from $(basename $tarball), created with:"
1609 echo " Commit: $img_commit"
1610 echo " Kernel: $img_kernel"
1611 echo " Arch: $img_arch"
1613 local version=$(version_code $img_commit)
1614 [[ $version -ge $(version_code 2.5.0) ]] && ff_convert="no"
1616 if [ $fstype == "zfs" ]; then
1618 $r $ZPOOL import -f -d $tmp t32fs-mdt1
1619 $r $ZPOOL import -f -d $tmp t32fs-ost1
1620 mdt_dev=t32fs-mdt1/mdt1
1621 ost_dev=t32fs-ost1/ost1
1622 wait_update_facet $SINGLEMDS "$ZPOOL list |
1623 awk '/^t32fs-mdt1/ { print \\\$1 }'" "t32fs-mdt1" || {
1624 error_noexit "import zfs pool failed"
1629 $r $LCTL set_param debug="$PTLDEBUG"
1631 $r $TUNEFS --dryrun $mdt_dev || {
1633 error_noexit "tunefs.lustre before mounting the MDT"
1636 if [ "$writeconf" ]; then
1638 if [ $fstype == "ldiskfs" ]; then
1640 $r $TUNEFS --quota $mdt_dev || {
1642 error_noexit "Enable mdt quota feature"
1647 if [ -n "$($LCTL list_nids | grep -v '\(tcp\|lo\)[[:digit:]]*$')" ]; then
1648 [[ $(lustre_version_code mgs) -ge $(version_code 2.3.59) ]] ||
1649 { skip "LU-2200: Cannot run over Inifiniband w/o lctl replace_nids "
1650 "(Need MGS version at least 2.3.59)"; return 0; }
1652 local osthost=$(facet_active_host ost1)
1653 local ostnid=$(do_node $osthost $LCTL list_nids | head -1)
1656 if [ $fstype == "ldiskfs" ]; then
1659 $r $MOUNT_CMD -o $mopts $mdt_dev $tmp/mnt/mdt
1660 $r lctl replace_nids $fsname-OST0000 $ostnid
1661 $r lctl replace_nids $fsname-MDT0000 $nid
1662 $r umount -d $tmp/mnt/mdt
1665 mopts=exclude=$fsname-OST0000
1666 if [ $fstype == "ldiskfs" ]; then
1671 t32_wait_til_devices_gone $node
1673 $r $MOUNT_CMD -o $mopts $mdt_dev $tmp/mnt/mdt || {
1675 error_noexit "Mounting the MDT"
1678 shall_cleanup_mdt=true
1680 if [ "$dne_upgrade" != "no" ]; then
1681 local fs2mdsdev=$(mdsdevname 1_2)
1682 local fs2mdsvdev=$(mdsvdevname 1_2)
1684 echo "mkfs new MDT on ${fs2mdsdev}...."
1685 if [ $(facet_fstype mds1) == ldiskfs ]; then
1686 mkfsoptions="--mkfsoptions=\\\"-J size=8\\\""
1689 add fs2mds $(mkfs_opts mds2 $fs2mdsdev $fsname) --reformat \
1690 $mkfsoptions $fs2mdsdev $fs2mdsvdev > /dev/null || {
1691 error_noexit "Mkfs new MDT failed"
1695 $r $TUNEFS --dryrun $fs2mdsdev || {
1696 error_noexit "tunefs.lustre before mounting the MDT"
1700 echo "mount new MDT....$fs2mdsdev"
1701 $r mkdir -p $tmp/mnt/mdt1
1702 $r $MOUNT_CMD -o $mopts $fs2mdsdev $tmp/mnt/mdt1 || {
1703 error_noexit "mount mdt1 failed"
1707 $r $LCTL set_param -n mdt.${fsname}*.enable_remote_dir=1 ||
1708 error_noexit "enable remote dir create failed"
1710 shall_cleanup_mdt1=true
1713 uuid=$($r $LCTL get_param -n mdt.$fsname-MDT0000.uuid) || {
1714 error_noexit "Getting MDT UUID"
1717 if [ "$uuid" != $fsname-MDT0000_UUID ]; then
1718 error_noexit "Unexpected MDT UUID: \"$uuid\""
1722 $r $TUNEFS --dryrun $ost_dev || {
1723 error_noexit "tunefs.lustre before mounting the OST"
1726 if [ "$writeconf" ]; then
1727 mopts=mgsnode=$nid,$writeconf
1728 if [ $fstype == "ldiskfs" ]; then
1730 $r $TUNEFS --quota $ost_dev || {
1732 error_noexit "Enable ost quota feature"
1738 if [ $fstype == "ldiskfs" ]; then
1742 $r $MOUNT_CMD -o $mopts $ost_dev $tmp/mnt/ost || {
1743 error_noexit "Mounting the OST"
1746 shall_cleanup_ost=true
1748 uuid=$($r $LCTL get_param -n obdfilter.$fsname-OST0000.uuid) || {
1749 error_noexit "Getting OST UUID"
1752 if [ "$uuid" != $fsname-OST0000_UUID ]; then
1753 error_noexit "Unexpected OST UUID: \"$uuid\""
1757 $r $LCTL conf_param $fsname-OST0000.osc.max_dirty_mb=15 || {
1758 error_noexit "Setting \"max_dirty_mb\""
1761 $r $LCTL conf_param $fsname-OST0000.failover.node=$nid || {
1762 error_noexit "Setting OST \"failover.node\""
1765 $r $LCTL conf_param $fsname-MDT0000.mdc.max_rpcs_in_flight=9 || {
1766 error_noexit "Setting \"max_rpcs_in_flight\""
1769 $r $LCTL conf_param $fsname-MDT0000.failover.node=$nid || {
1770 error_noexit "Setting MDT \"failover.node\""
1773 $r $LCTL pool_new $fsname.interop || {
1774 error_noexit "Setting \"interop\""
1777 $r $LCTL conf_param $fsname-MDT0000.lov.stripesize=4M || {
1778 error_noexit "Setting \"lov.stripesize\""
1782 if [ "$ff_convert" != "no" -a $(facet_fstype ost1) == "ldiskfs" ]; then
1783 $r $LCTL lfsck_start -M $fsname-OST0000 || {
1784 error_noexit "Start OI scrub on OST0"
1788 # The oi_scrub should be on ost1, but for test_32(),
1789 # all on the SINGLEMDS.
1790 wait_update_facet $SINGLEMDS "$LCTL get_param -n \
1791 osd-ldiskfs.$fsname-OST0000.oi_scrub |
1792 awk '/^status/ { print \\\$2 }'" "completed" 30 || {
1793 error_noexit "Failed to get the expected 'completed'"
1797 local UPDATED=$($r $LCTL get_param -n \
1798 osd-ldiskfs.$fsname-OST0000.oi_scrub |
1799 awk '/^updated/ { print $2 }')
1800 [ $UPDATED -ge 1 ] || {
1801 error_noexit "Only $UPDATED objects have been converted"
1806 if [ "$dne_upgrade" != "no" ]; then
1807 $r $LCTL conf_param \
1808 $fsname-MDT0001.mdc.max_rpcs_in_flight=9 || {
1809 error_noexit "Setting MDT1 \"max_rpcs_in_flight\""
1812 $r $LCTL conf_param $fsname-MDT0001.failover.node=$nid || {
1813 error_noexit "Setting MDT1 \"failover.node\""
1816 $r $LCTL conf_param $fsname-MDT0001.lov.stripesize=4M || {
1817 error_noexit "Setting MDT1 \"lov.stripesize\""
1823 if [ "$writeconf" ]; then
1824 $MOUNT_CMD $nid:/$fsname $tmp/mnt/lustre || {
1825 error_noexit "Mounting the client"
1828 shall_cleanup_lustre=true
1829 $LCTL set_param debug="$PTLDEBUG"
1831 t32_verify_quota $node $fsname $tmp/mnt/lustre || {
1832 error_noexit "verify quota failed"
1836 if [ "$dne_upgrade" != "no" ]; then
1837 $LFS mkdir -i 1 $tmp/mnt/lustre/remote_dir || {
1838 error_noexit "set remote dir failed"
1842 pushd $tmp/mnt/lustre
1843 tar -cf - . --exclude=./remote_dir |
1844 tar -xvf - -C remote_dir 1>/dev/null || {
1845 error_noexit "cp to remote dir failed"
1851 dd if=/dev/zero of=$tmp/mnt/lustre/tmp_file bs=10k count=10 || {
1852 error_noexit "dd failed"
1855 rm -rf $tmp/mnt/lustre/tmp_file || {
1856 error_noexit "rm failed"
1860 if $r test -f $tmp/sha1sums; then
1861 # LU-2393 - do both sorts on same node to ensure locale
1863 $r cat $tmp/sha1sums | sort -k 2 >$tmp/sha1sums.orig
1864 if [ "$dne_upgrade" != "no" ]; then
1865 pushd $tmp/mnt/lustre/remote_dir
1867 pushd $tmp/mnt/lustre
1870 find ! -name .lustre -type f -exec sha1sum {} \; |
1871 sort -k 2 >$tmp/sha1sums || {
1872 error_noexit "sha1sum"
1876 if ! diff -ub $tmp/sha1sums.orig $tmp/sha1sums; then
1877 error_noexit "sha1sum verification failed"
1881 echo "sha1sum verification skipped"
1884 if [ "$dne_upgrade" != "no" ]; then
1885 rm -rf $tmp/mnt/lustre/remote_dir || {
1886 error_noexit "remove remote dir failed"
1891 if $r test -f $tmp/list; then
1893 # There is not a Test Framework API to copy files to or
1894 # from a remote node.
1896 # LU-2393 - do both sorts on same node to ensure locale
1898 $r cat $tmp/list | sort -k 6 >$tmp/list.orig
1899 pushd $tmp/mnt/lustre
1900 ls -Rni --time-style=+%s | sort -k 6 >$tmp/list || {
1906 # 32-bit and 64-bit clients use different algorithms to
1907 # convert FIDs into inode numbers. Hence, remove the inode
1908 # numbers from the lists, if the original list was created
1909 # on an architecture with different number of bits per
1912 if [ $(t32_bits_per_long $(uname -m)) != \
1913 $(t32_bits_per_long $img_arch) ]; then
1914 echo "Different number of bits per \"long\" from the disk image"
1915 for list in list.orig list; do
1916 sed -i -e 's/^[0-9]\+[ \t]\+//' $tmp/$list
1919 if ! diff -ub $tmp/list.orig $tmp/list; then
1920 error_noexit "list verification failed"
1924 echo "list verification skipped"
1928 # When adding new data verification tests, please check for
1929 # the presence of the required reference files first, like
1930 # the "sha1sums" and "list" tests above, to avoid the need to
1931 # regenerate every image for each test addition.
1934 nrpcs_orig=$($LCTL get_param \
1935 -n mdc.*MDT0000*.max_rpcs_in_flight) || {
1936 error_noexit "Getting \"max_rpcs_in_flight\""
1939 nrpcs=$((nrpcs_orig + 5))
1940 $r $LCTL conf_param $fsname-MDT0000.mdc.max_rpcs_in_flight=$nrpcs || {
1941 error_noexit "Changing \"max_rpcs_in_flight\""
1944 wait_update $HOSTNAME "$LCTL get_param \
1945 -n mdc.*MDT0000*.max_rpcs_in_flight" $nrpcs || {
1946 error_noexit "Verifying \"max_rpcs_in_flight\""
1950 umount $tmp/mnt/lustre || {
1951 error_noexit "Unmounting the client"
1954 shall_cleanup_lustre=false
1956 if [ "$dne_upgrade" != "no" ]; then
1957 $r umount -d $tmp/mnt/mdt1 || {
1958 error_noexit "Unmounting the MDT2"
1961 shall_cleanup_mdt1=false
1964 $r umount -d $tmp/mnt/mdt || {
1965 error_noexit "Unmounting the MDT"
1968 shall_cleanup_mdt=false
1970 $r umount -d $tmp/mnt/ost || {
1971 error_noexit "Unmounting the OST"
1974 shall_cleanup_ost=false
1976 t32_reload_modules $node || {
1977 error_noexit "Reloading modules"
1981 # mount a second time to make sure we didnt leave upgrade flag on
1982 $r $TUNEFS --dryrun $mdt_dev || {
1984 error_noexit "tunefs.lustre before remounting the MDT"
1988 mopts=exclude=$fsname-OST0000
1989 if [ $fstype == "ldiskfs" ]; then
1992 $r $MOUNT_CMD -o $mopts $mdt_dev $tmp/mnt/mdt || {
1993 error_noexit "Remounting the MDT"
1996 shall_cleanup_mdt=true
2006 for tarball in $tarballs; do
2007 t32_test $tarball || let "rc += $?"
2011 run_test 32a "Upgrade (not live)"
2019 for tarball in $tarballs; do
2020 t32_test $tarball writeconf || let "rc += $?"
2024 run_test 32b "Upgrade with writeconf"
2031 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
2033 for tarball in $tarballs; do
2034 dne_upgrade=yes t32_test $tarball writeconf || rc=$?
2038 run_test 32c "dne upgrade test"
2046 for tarball in $tarballs; do
2047 ff_convert=yes t32_test $tarball || rc=$?
2051 run_test 32d "convert ff test"
2053 test_33a() { # bug 12333, was test_33
2055 local FSNAME2=test-123
2056 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
2059 [ -n "$ost1_HOST" ] && fs2ost_HOST=$ost1_HOST
2061 if [ -z "$fs2ost_DEV" -o -z "$fs2mds_DEV" ]; then
2062 local dev=${SINGLEMDS}_dev
2063 local MDSDEV=${!dev}
2064 is_blkdev $SINGLEMDS $MDSDEV && \
2065 skip_env "mixed loopback and real device not working" && return
2068 local fs2mdsdev=$(mdsdevname 1_2)
2069 local fs2ostdev=$(ostdevname 1_2)
2070 local fs2mdsvdev=$(mdsvdevname 1_2)
2071 local fs2ostvdev=$(ostvdevname 1_2)
2073 if [ $(facet_fstype mds1) == ldiskfs ]; then
2074 mkfsoptions="--mkfsoptions=\\\"-J size=8\\\"" # See bug 17931.
2077 add fs2mds $(mkfs_opts mds1 ${fs2mdsdev}) --mgs --fsname=${FSNAME2} \
2078 --reformat $mkfsoptions $fs2mdsdev $fs2mdsvdev || exit 10
2079 add fs2ost $(mkfs_opts ost1 ${fs2ostdev}) --mgsnode=$MGSNID \
2080 --fsname=${FSNAME2} --index=8191 --reformat $fs2ostdev \
2081 $fs2ostvdev || exit 10
2083 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS && trap cleanup_fs2 EXIT INT
2084 start fs2ost $fs2ostdev $OST_MOUNT_OPTS
2085 do_facet $SINGLEMDS "$LCTL conf_param $FSNAME2.sys.timeout=200" || rc=1
2087 $MOUNT_CMD $MGSNID:/${FSNAME2} $MOUNT2 || rc=2
2090 cp /etc/hosts $MOUNT2/ || rc=3
2091 $LFS getstripe $MOUNT2/hosts
2096 cleanup_nocli || rc=6
2099 run_test 33a "Mount ost with a large index number"
2101 test_33b() { # was test_34
2104 do_facet client dd if=/dev/zero of=$MOUNT/24 bs=1024k count=1
2105 # Drop lock cancelation reply during umount
2106 #define OBD_FAIL_LDLM_CANCEL_NET 0x304
2107 do_facet client lctl set_param fail_loc=0x80000304
2108 #lctl set_param debug=-1
2109 umount_client $MOUNT
2112 run_test 33b "Drop cancel during umount"
2116 do_facet client "sh runmultiop_bg_pause $DIR/file O_c"
2117 manual_umount_client
2119 do_facet client killall -USR1 multiop
2120 if [ $rc -eq 0 ]; then
2121 error "umount not fail!"
2126 run_test 34a "umount with opened file should be fail"
2131 touch $DIR/$tfile || return 1
2132 stop_mds --force || return 2
2134 manual_umount_client --force
2136 if [ $rc -ne 0 ]; then
2137 error "mtab after failed umount - rc $rc"
2143 run_test 34b "force umount with failed mds should be normal"
2147 touch $DIR/$tfile || return 1
2148 stop_ost --force || return 2
2150 manual_umount_client --force
2152 if [ $rc -ne 0 ]; then
2153 error "mtab after failed umount - rc $rc"
2159 run_test 34c "force umount with failed ost should be normal"
2161 test_35a() { # bug 12459
2164 DBG_SAVE="`lctl get_param -n debug`"
2165 lctl set_param debug="ha"
2167 log "Set up a fake failnode for the MDS"
2169 local device=$(do_facet $SINGLEMDS "lctl get_param -n devices" |
2170 awk '($3 ~ "mdt" && $4 ~ "MDT") { print $4 }' | head -1)
2171 do_facet mgs "$LCTL conf_param \
2172 ${device}.failover.node=$(h2$NETTYPE $FAKENID)" || return 4
2174 log "Wait for RECONNECT_INTERVAL seconds (10s)"
2177 MSG="conf-sanity.sh test_35a `date +%F%kh%Mm%Ss`"
2180 log "Stopping the MDT: $device"
2181 stop_mdt 1 || return 5
2183 df $MOUNT > /dev/null 2>&1 &
2185 log "Restarting the MDT: $device"
2186 start_mdt 1 || return 6
2187 log "Wait for df ($DFPID) ... "
2190 lctl set_param debug="$DBG_SAVE"
2192 # retrieve from the log the first server that the client tried to
2193 # contact after the connection loss
2194 $LCTL dk $TMP/lustre-log-$TESTNAME.log
2195 NEXTCONN=`awk "/${MSG}/ {start = 1;}
2196 /import_select_connection.*$device-mdc.* using connection/ {
2198 if (\\\$NF ~ /$FAKENID/)
2204 }" $TMP/lustre-log-$TESTNAME.log`
2205 [ "$NEXTCONN" != "0" ] && log "The client didn't try to reconnect to the last active server (tried ${NEXTCONN} instead)" && return 7
2207 # remove nid settings
2208 writeconf_or_reformat
2210 run_test 35a "Reconnect to the last active server first"
2212 test_35b() { # bug 18674
2213 remote_mds || { skip "local MDS" && return 0; }
2217 $LCTL set_param debug="ha"
2219 MSG="conf-sanity.sh test_35b `date +%F%kh%Mm%Ss`"
2222 log "Set up a fake failnode for the MDS"
2224 local device=$(do_facet $SINGLEMDS "$LCTL get_param -n devices" |
2225 awk '($3 ~ "mdt" && $4 ~ "MDT") { print $4 }' | head -1)
2226 do_facet mgs "$LCTL conf_param \
2227 ${device}.failover.node=$(h2$NETTYPE $FAKENID)" || return 1
2229 local at_max_saved=0
2230 # adaptive timeouts may prevent seeing the issue
2231 if at_is_enabled; then
2232 at_max_saved=$(at_max_get mds)
2233 at_max_set 0 mds client
2236 mkdir -p $MOUNT/$tdir
2238 log "Injecting EBUSY on MDS"
2239 # Setting OBD_FAIL_MDS_RESEND=0x136
2240 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0x80000136" || return 2
2242 $LCTL set_param mdc.${FSNAME}*.stats=clear
2244 log "Creating a test file and stat it"
2245 touch $MOUNT/$tdir/$tfile
2246 stat $MOUNT/$tdir/$tfile
2248 log "Stop injecting EBUSY on MDS"
2249 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0" || return 3
2250 rm -f $MOUNT/$tdir/$tfile
2253 # restore adaptive timeout
2254 [ $at_max_saved -ne 0 ] && at_max_set $at_max_saved mds client
2256 $LCTL dk $TMP/lustre-log-$TESTNAME.log
2258 CONNCNT=`$LCTL get_param mdc.${FSNAME}*.stats | awk '/mds_connect/{print $2}'`
2260 # retrieve from the log if the client has ever tried to
2261 # contact the fake server after the loss of connection
2262 FAILCONN=`awk "BEGIN {ret = 0;}
2263 /import_select_connection.*${FSNAME}-MDT0000-mdc.* using connection/ {
2265 if (\\\$NF ~ /$FAKENID/) {
2270 END {print ret}" $TMP/lustre-log-$TESTNAME.log`
2272 [ "$FAILCONN" == "0" ] && \
2273 log "ERROR: The client reconnection has not been triggered" && \
2275 [ "$FAILCONN" == "2" ] && \
2276 log "ERROR: The client tried to reconnect to the failover server while the primary was busy" && \
2280 # When OBD_FAIL_MDS_RESEND is hit, we sleep for 2 * obd_timeout
2281 # Reconnects are supposed to be rate limited to one every 5s
2282 [ $CONNCNT -gt $((2 * $TIMEOUT / 5 + 1)) ] && \
2283 log "ERROR: Too many reconnects $CONNCNT" && \
2287 # remove nid settings
2288 writeconf_or_reformat
2290 run_test 35b "Continue reconnection retries, if the active server is busy"
2293 [ $OSTCOUNT -lt 2 ] && skip_env "skipping test for single OST" && return
2295 [ "$ost_HOST" = "`hostname`" -o "$ost1_HOST" = "`hostname`" ] || \
2296 { skip "remote OST" && return 0; }
2299 local FSNAME2=test1234
2300 local fs3ost_HOST=$ost_HOST
2301 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
2303 [ -n "$ost1_HOST" ] && fs2ost_HOST=$ost1_HOST && fs3ost_HOST=$ost1_HOST
2305 if [ -z "$fs2ost_DEV" -o -z "$fs2mds_DEV" -o -z "$fs3ost_DEV" ]; then
2306 is_blkdev $SINGLEMDS $MDSDEV && \
2307 skip_env "mixed loopback and real device not working" && return
2310 local fs2mdsdev=$(mdsdevname 1_2)
2311 local fs2ostdev=$(ostdevname 1_2)
2312 local fs3ostdev=$(ostdevname 2_2)
2313 local fs2mdsvdev=$(mdsvdevname 1_2)
2314 local fs2ostvdev=$(ostvdevname 1_2)
2315 local fs3ostvdev=$(ostvdevname 2_2)
2317 add fs2mds $(mkfs_opts mds1 ${fs2mdsdev}) --mgs --fsname=${FSNAME2} \
2318 --reformat $fs2mdsdev $fs2mdsvdev || exit 10
2319 # XXX after we support non 4K disk blocksize in ldiskfs, specify a
2320 # different one than the default value here.
2321 add fs2ost $(mkfs_opts ost1 ${fs2ostdev}) --mgsnode=$MGSNID \
2322 --fsname=${FSNAME2} --reformat $fs2ostdev $fs2ostvdev || exit 10
2323 add fs3ost $(mkfs_opts ost2 ${fs3ostdev}) --mgsnode=$MGSNID \
2324 --fsname=${FSNAME2} --reformat $fs3ostdev $fs3ostvdev || exit 10
2326 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS
2327 start fs2ost $fs2ostdev $OST_MOUNT_OPTS
2328 start fs3ost $fs3ostdev $OST_MOUNT_OPTS
2330 $MOUNT_CMD $MGSNID:/${FSNAME2} $MOUNT2 || return 1
2332 sleep 5 # until 11778 fixed
2334 dd if=/dev/zero of=$MOUNT2/$tfile bs=1M count=7 || return 2
2336 BKTOTAL=`lctl get_param -n obdfilter.*.kbytestotal | awk 'BEGIN{total=0}; {total+=$1}; END{print total}'`
2337 BKFREE=`lctl get_param -n obdfilter.*.kbytesfree | awk 'BEGIN{free=0}; {free+=$1}; END{print free}'`
2338 BKAVAIL=`lctl get_param -n obdfilter.*.kbytesavail | awk 'BEGIN{avail=0}; {avail+=$1}; END{print avail}'`
2339 STRING=`df -P $MOUNT2 | tail -n 1 | awk '{print $2","$3","$4}'`
2340 DFTOTAL=`echo $STRING | cut -d, -f1`
2341 DFUSED=`echo $STRING | cut -d, -f2`
2342 DFAVAIL=`echo $STRING | cut -d, -f3`
2343 DFFREE=$(($DFTOTAL - $DFUSED))
2345 ALLOWANCE=$((64 * $OSTCOUNT))
2347 if [ $DFTOTAL -lt $(($BKTOTAL - $ALLOWANCE)) ] ||
2348 [ $DFTOTAL -gt $(($BKTOTAL + $ALLOWANCE)) ] ; then
2349 echo "**** FAIL: df total($DFTOTAL) mismatch OST total($BKTOTAL)"
2352 if [ $DFFREE -lt $(($BKFREE - $ALLOWANCE)) ] ||
2353 [ $DFFREE -gt $(($BKFREE + $ALLOWANCE)) ] ; then
2354 echo "**** FAIL: df free($DFFREE) mismatch OST free($BKFREE)"
2357 if [ $DFAVAIL -lt $(($BKAVAIL - $ALLOWANCE)) ] ||
2358 [ $DFAVAIL -gt $(($BKAVAIL + $ALLOWANCE)) ] ; then
2359 echo "**** FAIL: df avail($DFAVAIL) mismatch OST avail($BKAVAIL)"
2364 stop fs3ost -f || return 200
2365 stop fs2ost -f || return 201
2366 stop fs2mds -f || return 202
2367 unload_modules_conf || return 203
2370 run_test 36 "df report consistency on OSTs with different block size"
2373 local mntpt=$(facet_mntpt $SINGLEMDS)
2374 local mdsdev=$(mdsdevname ${SINGLEMDS//mds/})
2375 local mdsdev_sym="$TMP/sym_mdt.img"
2376 local opts=$MDS_MOUNT_OPTS
2379 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
2380 skip "Currently only applicable to ldiskfs-based MDTs"
2384 echo "MDS : $mdsdev"
2385 echo "SYMLINK : $mdsdev_sym"
2386 do_facet $SINGLEMDS rm -f $mdsdev_sym
2388 do_facet $SINGLEMDS ln -s $mdsdev $mdsdev_sym
2390 echo "mount symlink device - $mdsdev_sym"
2392 if ! do_facet $SINGLEMDS test -b $mdsdev; then
2393 opts=$(csa_add "$opts" -o loop)
2395 mount_op=$(do_facet $SINGLEMDS mount -v -t lustre $opts \
2396 $mdsdev_sym $mntpt 2>&1)
2399 echo mount_op=$mount_op
2401 do_facet $SINGLEMDS "umount -d $mntpt && rm -f $mdsdev_sym"
2403 if $(echo $mount_op | grep -q "unable to set tunable"); then
2404 error "set tunables failed for symlink device"
2407 [ $rc -eq 0 ] || error "mount symlink $mdsdev_sym failed! rc=$rc"
2411 run_test 37 "verify set tunables works for symlink device"
2413 test_38() { # bug 14222
2414 local fstype=$(facet_fstype $SINGLEMDS)
2415 local mntpt=$(facet_mntpt $SINGLEMDS)
2420 local SRC="/etc /bin"
2421 local FILES=$(find $SRC -type f -mtime +1 | head -n $COUNT)
2422 log "copying $(echo $FILES | wc -w) files to $DIR/$tdir"
2424 tar cf - $FILES | tar xf - -C $DIR/$tdir ||
2425 error "copying $SRC to $DIR/$tdir"
2427 umount_client $MOUNT
2428 do_facet $SINGLEMDS "$LCTL get_param osp.*.prealloc_next_id"
2430 log "delete lov_objid file on MDS"
2432 mount_fstype $SINGLEMDS || error "mount MDS failed (1)"
2434 do_facet $SINGLEMDS "od -Ax -td8 $mntpt/lov_objid; rm $mntpt/lov_objid"
2436 unmount_fstype $SINGLEMDS || error "umount failed (1)"
2438 # check create in mds_lov_connect
2442 [ $V ] && log "verifying $DIR/$tdir/$f"
2443 diff -q $f $DIR/$tdir/$f || ERROR=y
2445 do_facet $SINGLEMDS "$LCTL get_param osp.*.prealloc_next_id"
2446 if [ "$ERROR" = "y" ]; then
2447 # check it's updates in sync
2448 umount_client $MOUNT
2450 mount_fstype $SIGNLEMDS
2451 do_facet $SINGLEMDS "od -Ax -td8 $mntpt/lov_objid"
2452 unmount_fstype $SINGLEMDS
2453 error "old and new files are different after connect" || true
2455 touch $DIR/$tdir/f2 || error "f2 file create failed"
2457 # check it's updates in sync
2458 umount_client $MOUNT
2461 mount_fstype $SINGLEMDS || error "mount MDS failed (3)"
2463 do_facet $SINGLEMDS "od -Ax -td8 $mntpt/lov_objid"
2464 do_facet $SINGLEMDS dd if=/dev/zero of=$mntpt/lov_objid.clear count=8
2466 unmount_fstype $SINGLEMDS || error "umount failed (3)"
2471 [ $V ] && log "verifying $DIR/$tdir/$f"
2472 diff -q $f $DIR/$tdir/$f || ERROR=y
2474 touch $DIR/$tdir/f3 || error "f3 file create failed"
2475 do_facet $SINGLEMDS "$LCTL get_param osp.*.prealloc_next_id"
2476 umount_client $MOUNT
2478 mount_fstype $SINGLEMDS || error "mount MDS failed (4)"
2479 do_facet $SINGLEMDS "od -Ax -td8 $mntpt/lov_objid"
2480 unmount_fstype $SINGLEMDS || error "umount failed (4)"
2482 [ "$ERROR" = "y" ] &&
2483 error "old and new files are different after sync" || true
2485 log "files compared the same"
2488 run_test 38 "MDS recreates missing lov_objid file from OST data"
2494 perl $SRCDIR/leak_finder.pl $TMP/debug 2>&1 | egrep '*** Leak:' &&
2495 error "memory leak detected" || true
2497 run_test 39 "leak_finder recognizes both LUSTRE and LNET malloc messages"
2499 test_40() { # bug 15759
2501 #define OBD_FAIL_TGT_TOOMANY_THREADS 0x706
2502 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0x80000706"
2506 run_test 40 "race during service thread startup"
2508 test_41a() { #bug 14134
2509 if [ $(facet_fstype $SINGLEMDS) == ldiskfs ] &&
2510 ! do_facet $SINGLEMDS test -b $(mdsdevname 1); then
2511 skip "Loop devices does not work with nosvc option"
2516 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
2518 start $SINGLEMDS $MDSDEV $MDS_MOUNT_OPTS -o nosvc -n
2519 start ost1 `ostdevname 1` $OST_MOUNT_OPTS
2520 start $SINGLEMDS $MDSDEV $MDS_MOUNT_OPTS -o nomgs,force
2522 mount_client $MOUNT || return 1
2525 echo "blah blah" > $MOUNT/$tfile
2528 umount_client $MOUNT
2529 stop ost1 -f || return 201
2530 stop_mds -f || return 202
2531 stop_mds -f || return 203
2532 unload_modules_conf || return 204
2535 run_test 41a "mount mds with --nosvc and --nomgs"
2538 if [ $(facet_fstype $SINGLEMDS) == ldiskfs ] &&
2539 ! do_facet $SINGLEMDS test -b $(mdsdevname 1); then
2540 skip "Loop devices does not work with nosvc option"
2544 ! combined_mgs_mds && skip "needs combined mgs device" && return 0
2548 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
2550 start $SINGLEMDS $MDSDEV $MDS_MOUNT_OPTS -o nosvc -n
2552 start $SINGLEMDS $MDSDEV $MDS_MOUNT_OPTS -o nomgs,force
2554 mount_client $MOUNT || return 1
2557 echo "blah blah" > $MOUNT/$tfile
2558 cat $MOUNT/$tfile || return 200
2560 umount_client $MOUNT
2561 stop_ost || return 201
2562 stop_mds -f || return 202
2563 stop_mds -f || return 203
2566 run_test 41b "mount mds with --nosvc and --nomgs on first mount"
2570 # MDT concurent start
2571 #define OBD_FAIL_TGT_DELAY_CONNECT 0x703
2572 do_facet $SINGLEMDS "lctl set_param fail_loc=0x703"
2573 start mds1 $(mdsdevname 1) $MDS_MOUNT_OPTS &
2576 do_facet $SINGLEMDS "lctl set_param fail_loc=0x0"
2577 start mds1 $(mdsdevname 1) $MDS_MOUNT_OPTS &
2583 if [ $rc == 0 ] && [ $rc2 == 114 ]; then
2584 echo "1st MDT start succeed"
2585 echo "2nd MDT start failed with EALREADY"
2586 elif [ $rc2 == 0 ] && [ $rc == 114 ]; then
2587 echo "1st MDT start failed with EALREADY"
2588 echo "2nd MDT start succeed"
2591 error "unexpected concurent MDT mounts result, rc=$rc rc2=$rc2"
2594 # OST concurent start
2595 #define OBD_FAIL_TGT_DELAY_CONNECT 0x703
2596 do_facet ost1 "lctl set_param fail_loc=0x703"
2597 start ost1 $(ostdevname 1) $OST_MOUNT_OPTS &
2600 do_facet ost1 "lctl set_param fail_loc=0x0"
2601 start ost1 $(ostdevname 1) $OST_MOUNT_OPTS &
2607 if [ $rc == 0 ] && [ $rc2 == 114 ]; then
2608 echo "1st OST start succeed"
2609 echo "2nd OST start failed with EALREADY"
2610 elif [ $rc2 == 0 ] && [ $rc == 114 ]; then
2611 echo "1st OST start failed with EALREADY"
2612 echo "2nd OST start succeed"
2616 error "unexpected concurent OST mounts result, rc=$rc rc2=$rc2"
2622 # verify everything ok
2627 error "MDT(s) start failed"
2635 error "OST(s) start failed"
2643 error "client start failed"
2650 error "client mount failed"
2654 run_test 41c "concurent mounts of MDT/OST should all fail but one"
2656 test_42() { #bug 14693
2658 check_mount || error "client was not mounted"
2660 do_facet mgs $LCTL conf_param $FSNAME.llite.some_wrong_param=10
2661 umount_client $MOUNT ||
2662 error "unmounting client failed with invalid llite param"
2663 mount_client $MOUNT ||
2664 error "mounting client failed with invalid llite param"
2666 do_facet mgs $LCTL conf_param $FSNAME.sys.some_wrong_param=20
2667 cleanup || error "stopping $FSNAME failed with invalid sys param"
2670 check_mount || "client was not mounted with invalid sys param"
2671 cleanup || error "stopping $FSNAME failed with invalid sys param"
2674 run_test 42 "allow client/server mount/unmount with invalid config param"
2677 [[ $(lustre_version_code mgs) -ge $(version_code 2.5.58) ]] ||
2678 { skip "Need MDS version at least 2.5.58" && return 0; }
2679 [ $UID -ne 0 -o $RUNAS_ID -eq 0 ] && skip_env "run as root"
2682 USER1=$(cat /etc/passwd | grep :$ID1:$ID1: | cut -d: -f1)
2683 [ -z "$USER1" ] && skip_env "missing user with uid=$ID1 gid=$ID1" &&
2687 chmod ugo+x $DIR || error "chmod 0 failed"
2688 set_conf_param_and_check mds \
2689 "lctl get_param -n mdt.$FSNAME-MDT0000.root_squash" \
2690 "$FSNAME.mdt.root_squash" \
2692 wait_update $HOSTNAME \
2693 "lctl get_param -n llite.${FSNAME}*.root_squash" \
2695 error "check llite root_squash failed!"
2696 set_conf_param_and_check mds \
2697 "lctl get_param -n mdt.$FSNAME-MDT0000.nosquash_nids" \
2698 "$FSNAME.mdt.nosquash_nids" \
2700 wait_update $HOSTNAME \
2701 "lctl get_param -n llite.${FSNAME}*.nosquash_nids" \
2703 error "check llite nosquash_nids failed!"
2706 # create set of test files
2708 echo "111" > $DIR/$tfile-userfile || error "write 1 failed"
2709 chmod go-rw $DIR/$tfile-userfile || error "chmod 1 failed"
2710 chown $RUNAS_ID.$RUNAS_ID $DIR/$tfile-userfile || error "chown failed"
2712 echo "222" > $DIR/$tfile-rootfile || error "write 2 failed"
2713 chmod go-rw $DIR/$tfile-rootfile || error "chmod 2 faield"
2715 mkdir $DIR/$tdir-rootdir -p || error "mkdir failed"
2716 chmod go-rwx $DIR/$tdir-rootdir || error "chmod 3 failed"
2717 touch $DIR/$tdir-rootdir/tfile-1 || error "touch failed"
2719 echo "777" > $DIR/$tfile-user1file || error "write 7 failed"
2720 chmod go-rw $DIR/$tfile-user1file || error "chmod 7 failed"
2721 chown $ID1.$ID1 $DIR/$tfile-user1file || error "chown failed"
2724 # check root_squash:
2725 # set root squash UID:GID to RUNAS_ID
2726 # root should be able to access only files owned by RUNAS_ID
2728 set_conf_param_and_check mds \
2729 "lctl get_param -n mdt.$FSNAME-MDT0000.root_squash" \
2730 "$FSNAME.mdt.root_squash" \
2731 "$RUNAS_ID:$RUNAS_ID"
2732 wait_update $HOSTNAME \
2733 "lctl get_param -n llite.${FSNAME}*.root_squash" \
2734 "$RUNAS_ID:$RUNAS_ID" ||
2735 error "check llite root_squash failed!"
2737 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-userfile)
2738 dd if=$DIR/$tfile-userfile 1>/dev/null 2>/dev/null || \
2739 error "$ST: root read permission is denied"
2740 echo "$ST: root read permission is granted - ok"
2743 dd conv=notrunc of=$DIR/$tfile-userfile 1>/dev/null 2>/dev/null || \
2744 error "$ST: root write permission is denied"
2745 echo "$ST: root write permission is granted - ok"
2747 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-rootfile)
2748 dd if=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null && \
2749 error "$ST: root read permission is granted"
2750 echo "$ST: root read permission is denied - ok"
2753 dd conv=notrunc of=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null && \
2754 error "$ST: root write permission is granted"
2755 echo "$ST: root write permission is denied - ok"
2757 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tdir-rootdir)
2758 rm $DIR/$tdir-rootdir/tfile-1 1>/dev/null 2>/dev/null && \
2759 error "$ST: root unlink permission is granted"
2760 echo "$ST: root unlink permission is denied - ok"
2762 touch $DIR/tdir-rootdir/tfile-2 1>/dev/null 2>/dev/null && \
2763 error "$ST: root create permission is granted"
2764 echo "$ST: root create permission is denied - ok"
2768 # check root_squash is enforced independently
2769 # of client cache content
2771 # access file by USER1, keep access open
2772 # root should be denied access to user file
2774 runas -u $ID1 tail -f $DIR/$tfile-user1file 1>/dev/null 2>&1 &
2778 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-user1file)
2779 dd if=$DIR/$tfile-user1file 1>/dev/null 2>&1 &&
2780 { kill $pid; error "$ST: root read permission is granted"; }
2781 echo "$ST: root read permission is denied - ok"
2784 dd conv=notrunc of=$DIR/$tfile-user1file 1>/dev/null 2>&1 &&
2785 { kill $pid; error "$ST: root write permission is granted"; }
2786 echo "$ST: root write permission is denied - ok"
2792 # check nosquash_nids:
2793 # put client's NID into nosquash_nids list,
2794 # root should be able to access root file after that
2796 local NIDLIST=$(lctl list_nids all | tr '\n' ' ')
2797 NIDLIST="2@elan $NIDLIST 192.168.0.[2,10]@tcp"
2798 NIDLIST=$(echo $NIDLIST | tr -s ' ' ' ')
2799 set_conf_param_and_check mds \
2800 "lctl get_param -n mdt.$FSNAME-MDT0000.nosquash_nids" \
2801 "$FSNAME-MDTall.mdt.nosquash_nids" \
2803 wait_update $HOSTNAME \
2804 "lctl get_param -n llite.${FSNAME}*.nosquash_nids" \
2806 error "check llite nosquash_nids failed!"
2808 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-rootfile)
2809 dd if=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null || \
2810 error "$ST: root read permission is denied"
2811 echo "$ST: root read permission is granted - ok"
2814 dd conv=notrunc of=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null || \
2815 error "$ST: root write permission is denied"
2816 echo "$ST: root write permission is granted - ok"
2818 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tdir-rootdir)
2819 rm $DIR/$tdir-rootdir/tfile-1 || \
2820 error "$ST: root unlink permission is denied"
2821 echo "$ST: root unlink permission is granted - ok"
2822 touch $DIR/$tdir-rootdir/tfile-2 || \
2823 error "$ST: root create permission is denied"
2824 echo "$ST: root create permission is granted - ok"
2828 run_test 43 "check root_squash and nosquash_nids"
2830 umount_client $MOUNT
2835 check_mount || return 2
2836 UUID=$($LCTL get_param llite.${FSNAME}*.uuid | cut -d= -f2)
2838 UUIDS=$(do_facet $SINGLEMDS "$LCTL get_param mdt.${FSNAME}*.exports.*.uuid")
2839 for VAL in $UUIDS; do
2840 NID=$(echo $VAL | cut -d= -f1)
2841 CLUUID=$(echo $VAL | cut -d= -f2)
2842 [ "$UUID" = "$CLUUID" ] && STATS_FOUND=yes && break
2844 [ "$STATS_FOUND" = "no" ] && error "stats not found for client"
2848 run_test 44 "mounted client proc entry exists"
2852 check_mount || return 2
2857 #define OBD_FAIL_PTLRPC_LONG_UNLINK 0x50f
2858 do_facet client "lctl set_param fail_loc=0x50f"
2861 manual_umount_client --force || return 3
2862 do_facet client "lctl set_param fail_loc=0x0"
2864 mount_client $MOUNT || return 4
2868 run_test 45 "long unlink handling in ptlrpcd"
2875 umount_client $MOUNT2 || rc=$?
2876 umount_client $MOUNT || rc=$?
2877 while [ $count -gt 0 ]; do
2878 stop ost${count} -f || rc=$?
2882 cleanup_nocli || rc=$?
2883 #writeconf to remove all ost2 traces for subsequent tests
2884 writeconf_or_reformat
2889 echo "Testing with $OSTCOUNT OSTs"
2891 start_mds || return 1
2892 #first client should see only one ost
2893 start_ost || return 2
2894 wait_osc_import_state mds ost FULL
2896 mount_client $MOUNT || return 3
2897 trap "cleanup_46a $OSTCOUNT" EXIT ERR
2900 for (( i=2; i<=$OSTCOUNT; i++ )); do
2901 start ost$i `ostdevname $i` $OST_MOUNT_OPTS || return $((i+2))
2904 # wait until osts in sync
2905 for (( i=2; i<=$OSTCOUNT; i++ )); do
2906 wait_osc_import_state mds ost$i FULL
2907 wait_osc_import_state client ost$i FULL
2910 #second client see all ost's
2912 mount_client $MOUNT2 || return 8
2913 $LFS setstripe -c -1 $MOUNT2 || return 9
2914 $LFS getstripe $MOUNT2 || return 10
2916 echo "ok" > $MOUNT2/widestripe
2917 $LFS getstripe $MOUNT2/widestripe || return 11
2918 # fill acl buffer for avoid expand lsm to them
2919 awk -F : '{if (FNR < 25) { print "u:"$1":rwx" }}' /etc/passwd | while read acl; do
2920 setfacl -m $acl $MOUNT2/widestripe
2924 stat $MOUNT/widestripe || return 12
2926 cleanup_46a $OSTCOUNT || { echo "cleanup_46a failed!" && return 13; }
2929 run_test 46a "handle ost additional - wide striped file"
2934 check_mount || return 2
2935 $LCTL set_param ldlm.namespaces.$FSNAME-*-*-*.lru_size=100
2939 for ns in $($LCTL get_param ldlm.namespaces.$FSNAME-*-*-*.lru_size); do
2940 if echo $ns | grep "MDT[[:digit:]]*"; then
2943 lrs=$(echo $ns | sed 's/.*lru_size=//')
2944 lru_size[count]=$lrs
2949 facet_failover $SINGLEMDS
2950 client_up || return 3
2953 for ns in $($LCTL get_param ldlm.namespaces.$FSNAME-*-*-*.lru_size); do
2954 if echo $ns | grep "MDT[[:digit:]]*"; then
2957 lrs=$(echo $ns | sed 's/.*lru_size=//')
2958 if ! test "$lrs" -eq "${lru_size[count]}"; then
2959 n=$(echo $ns | sed -e 's/ldlm.namespaces.//' -e 's/.lru_size=.*//')
2960 error "$n has lost lru_size: $lrs vs. ${lru_size[count]}"
2968 run_test 47 "server restart does not make client loss lru_resize settings"
2973 # reformat after this test is needed - if test will failed
2974 # we will have unkillable file at FS
2978 test_48() { # bug 17636
2981 check_mount || return 2
2983 $LFS setstripe -c -1 $MOUNT || return 9
2984 $LFS getstripe $MOUNT || return 10
2986 echo "ok" > $MOUNT/widestripe
2987 $LFS getstripe $MOUNT/widestripe || return 11
2989 trap cleanup_48 EXIT ERR
2991 # fill acl buffer for avoid expand lsm to them
2992 getent passwd | awk -F : '{ print "u:"$1":rwx" }' | while read acl; do
2993 setfacl -m $acl $MOUNT/widestripe
2996 stat $MOUNT/widestripe || return 12
3001 run_test 48 "too many acls on file"
3003 # check PARAM_SYS_LDLM_TIMEOUT option of MKFS.LUSTRE
3004 test_49a() { # bug 17710
3005 local timeout_orig=$TIMEOUT
3006 local ldlm_timeout_orig=$LDLM_TIMEOUT
3007 local LOCAL_TIMEOUT=20
3009 LDLM_TIMEOUT=$LOCAL_TIMEOUT
3010 TIMEOUT=$LOCAL_TIMEOUT
3014 check_mount || error "client mount failed"
3016 echo "check ldlm_timout..."
3017 local LDLM_MDS="$(do_facet $SINGLEMDS lctl get_param -n ldlm_timeout)"
3018 local LDLM_OST1="$(do_facet ost1 lctl get_param -n ldlm_timeout)"
3019 local LDLM_CLIENT="$(do_facet client lctl get_param -n ldlm_timeout)"
3021 if [ $LDLM_MDS -ne $LDLM_OST1 -o $LDLM_MDS -ne $LDLM_CLIENT ]; then
3022 error "Different LDLM_TIMEOUT:$LDLM_MDS $LDLM_OST1 $LDLM_CLIENT"
3025 if [ $LDLM_MDS -ne $((LOCAL_TIMEOUT / 3)) ]; then
3026 error "LDLM_TIMEOUT($LDLM_MDS) is not $((LOCAL_TIMEOUT / 3))"
3029 umount_client $MOUNT
3030 stop_ost || error "problem stopping OSS"
3031 stop_mds || error "problem stopping MDS"
3033 LDLM_TIMEOUT=$ldlm_timeout_orig
3034 TIMEOUT=$timeout_orig
3036 run_test 49a "check PARAM_SYS_LDLM_TIMEOUT option of mkfs.lustre"
3038 test_49b() { # bug 17710
3039 local timeout_orig=$TIMEOUT
3040 local ldlm_timeout_orig=$LDLM_TIMEOUT
3041 local LOCAL_TIMEOUT=20
3043 LDLM_TIMEOUT=$((LOCAL_TIMEOUT - 1))
3044 TIMEOUT=$LOCAL_TIMEOUT
3048 check_mount || error "client mount failed"
3050 local LDLM_MDS="$(do_facet $SINGLEMDS lctl get_param -n ldlm_timeout)"
3051 local LDLM_OST1="$(do_facet ost1 lctl get_param -n ldlm_timeout)"
3052 local LDLM_CLIENT="$(do_facet client lctl get_param -n ldlm_timeout)"
3054 if [ $LDLM_MDS -ne $LDLM_OST1 -o $LDLM_MDS -ne $LDLM_CLIENT ]; then
3055 error "Different LDLM_TIMEOUT:$LDLM_MDS $LDLM_OST1 $LDLM_CLIENT"
3058 if [ $LDLM_MDS -ne $((LOCAL_TIMEOUT - 1)) ]; then
3059 error "LDLM_TIMEOUT($LDLM_MDS) is not $((LOCAL_TIMEOUT - 1))"
3062 cleanup || error "cleanup failed"
3064 LDLM_TIMEOUT=$ldlm_timeout_orig
3065 TIMEOUT=$timeout_orig
3067 run_test 49b "check PARAM_SYS_LDLM_TIMEOUT option of mkfs.lustre"
3070 # Test both statfs and lfs df and fail if either one fails
3071 multiop_bg_pause $1 f_
3074 killall -USR1 multiop
3075 [ $RC1 -ne 0 ] && log "lazystatfs multiop failed"
3076 wait $PID || { RC1=$?; log "multiop return error "; }
3083 if [ $RC2 -eq 0 ]; then
3085 log "lazystatfs df failed"
3089 [[ $RC1 -ne 0 || $RC2 -eq 0 ]] && RC=1
3095 lctl set_param llite.$FSNAME-*.lazystatfs=1
3098 lazystatfs $MOUNT || error "lazystatfs failed but no down servers"
3100 cleanup || return $?
3102 run_test 50a "lazystatfs all servers available =========================="
3106 lctl set_param llite.$FSNAME-*.lazystatfs=1
3109 # Wait for client to detect down OST
3110 stop_ost || error "Unable to stop OST1"
3111 wait_osc_import_state mds ost DISCONN
3113 lazystatfs $MOUNT || error "lazystatfs should don't have returned EIO"
3115 umount_client $MOUNT || error "Unable to unmount client"
3116 stop_mds || error "Unable to stop MDS"
3118 run_test 50b "lazystatfs all servers down =========================="
3121 start_mds || error "Unable to start MDS"
3122 start_ost || error "Unable to start OST1"
3123 start_ost2 || error "Unable to start OST2"
3124 mount_client $MOUNT || error "Unable to mount client"
3125 lctl set_param llite.$FSNAME-*.lazystatfs=1
3128 # Wait for client to detect down OST
3129 stop_ost || error "Unable to stop OST1"
3130 wait_osc_import_state mds ost DISCONN
3131 lazystatfs $MOUNT || error "lazystatfs failed with one down server"
3133 umount_client $MOUNT || error "Unable to unmount client"
3134 stop_ost2 || error "Unable to stop OST2"
3135 stop_mds || error "Unable to stop MDS"
3136 #writeconf to remove all ost2 traces for subsequent tests
3137 writeconf_or_reformat
3139 run_test 50c "lazystatfs one server down =========================="
3142 start_mds || error "Unable to start MDS"
3143 start_ost || error "Unable to start OST1"
3144 start_ost2 || error "Unable to start OST2"
3145 mount_client $MOUNT || error "Unable to mount client"
3146 lctl set_param llite.$FSNAME-*.lazystatfs=1
3149 # Issue the statfs during the window where the client still
3150 # belives the OST to be available but it is in fact down.
3151 # No failure just a statfs which hangs for a timeout interval.
3152 stop_ost || error "Unable to stop OST1"
3153 lazystatfs $MOUNT || error "lazystatfs failed with one down server"
3155 umount_client $MOUNT || error "Unable to unmount client"
3156 stop_ost2 || error "Unable to stop OST2"
3157 stop_mds || error "Unable to stop MDS"
3158 #writeconf to remove all ost2 traces for subsequent tests
3159 writeconf_or_reformat
3161 run_test 50d "lazystatfs client/server conn race =========================="
3168 start_mds || return 1
3169 #first client should see only one ost
3170 start_ost || return 2
3171 wait_osc_import_state mds ost FULL
3173 # Wait for client to detect down OST
3174 stop_ost || error "Unable to stop OST1"
3175 wait_osc_import_state mds ost DISCONN
3177 mount_client $MOUNT || error "Unable to mount client"
3178 lctl set_param llite.$FSNAME-*.lazystatfs=0
3180 multiop_bg_pause $MOUNT _f
3184 if [ $RC1 -ne 0 ]; then
3185 log "multiop failed $RC1"
3188 sleep $(( $TIMEOUT+1 ))
3190 [ $? -ne 0 ] && error "process isn't sleep"
3191 start_ost || error "Unable to start OST1"
3192 wait $pid || error "statfs failed"
3195 umount_client $MOUNT || error "Unable to unmount client"
3196 stop_ost || error "Unable to stop OST1"
3197 stop_mds || error "Unable to stop MDS"
3199 run_test 50e "normal statfs all servers down =========================="
3204 CONN_PROC="osc.$FSNAME-OST0001-osc-[M]*.ost_server_uuid"
3206 start_mds || error "Unable to start mds"
3207 #first client should see only one ost
3208 start_ost || error "Unable to start OST1"
3209 wait_osc_import_state mds ost FULL
3211 start_ost2 || error "Unable to start OST2"
3212 wait_osc_import_state mds ost2 FULL
3214 # Wait for client to detect down OST
3215 stop_ost2 || error "Unable to stop OST2"
3217 wait_osc_import_state mds ost2 DISCONN
3218 mount_client $MOUNT || error "Unable to mount client"
3219 lctl set_param llite.$FSNAME-*.lazystatfs=0
3221 multiop_bg_pause $MOUNT _f
3225 if [ $RC1 -ne 0 ]; then
3226 log "lazystatfs multiop failed $RC1"
3229 sleep $(( $TIMEOUT+1 ))
3231 [ $? -ne 0 ] && error "process isn't sleep"
3232 start_ost2 || error "Unable to start OST2"
3233 wait $pid || error "statfs failed"
3234 stop_ost2 || error "Unable to stop OST2"
3237 umount_client $MOUNT || error "Unable to unmount client"
3238 stop_ost || error "Unable to stop OST1"
3239 stop_mds || error "Unable to stop MDS"
3240 #writeconf to remove all ost2 traces for subsequent tests
3241 writeconf_or_reformat
3243 run_test 50f "normal statfs one server in down =========================="
3246 [ "$OSTCOUNT" -lt "2" ] && skip_env "$OSTCOUNT < 2, skipping" && return
3248 start_ost2 || error "Unable to start OST2"
3249 wait_osc_import_state mds ost2 FULL
3250 wait_osc_import_state client ost2 FULL
3252 local PARAM="${FSNAME}-OST0001.osc.active"
3254 $LFS setstripe -c -1 $DIR/$tfile || error "Unable to lfs setstripe"
3255 do_facet mgs $LCTL conf_param $PARAM=0 || error "Unable to deactivate OST"
3257 umount_client $MOUNT || error "Unable to unmount client"
3258 mount_client $MOUNT || error "Unable to mount client"
3259 # This df should not cause a panic
3262 do_facet mgs $LCTL conf_param $PARAM=1 || error "Unable to activate OST"
3264 umount_client $MOUNT || error "Unable to unmount client"
3265 stop_ost2 || error "Unable to stop OST2"
3266 stop_ost || error "Unable to stop OST1"
3267 stop_mds || error "Unable to stop MDS"
3268 #writeconf to remove all ost2 traces for subsequent tests
3269 writeconf_or_reformat
3271 run_test 50g "deactivated OST should not cause panic====================="
3275 # prepare MDT/OST, make OSC inactive for OST1
3276 [ "$OSTCOUNT" -lt "2" ] && skip_env "$OSTCOUNT < 2, skipping" && return
3278 [ $(facet_fstype ost1) == zfs ] && import_zpool ost1
3279 do_facet ost1 "$TUNEFS --param osc.active=0 `ostdevname 1`" ||
3280 error "tunefs OST1 failed"
3281 start_mds || error "Unable to start MDT"
3282 start_ost || error "Unable to start OST1"
3283 start_ost2 || error "Unable to start OST2"
3284 mount_client $MOUNT || error "client start failed"
3288 # activatate OSC for OST1
3289 local TEST="$LCTL get_param -n osc.${FSNAME}-OST0000-osc-[!M]*.active"
3290 set_conf_param_and_check client \
3291 "$TEST" "${FSNAME}-OST0000.osc.active" 1 ||
3292 error "Unable to activate OST1"
3294 mkdir -p $DIR/$tdir/2
3295 $LFS setstripe -c -1 -i 0 $DIR/$tdir/2
3296 sleep 1 && echo "create a file after OST1 is activated"
3298 createmany -o $DIR/$tdir/2/$tfile-%d 1
3300 # check OSC import is working
3301 stat $DIR/$tdir/2/* >/dev/null 2>&1 ||
3302 error "some OSC imports are still not connected"
3305 umount_client $MOUNT || error "Unable to umount client"
3306 stop_ost2 || error "Unable to stop OST2"
3309 run_test 50h "LU-642: activate deactivated OST ==="
3312 local LOCAL_TIMEOUT=20
3316 check_mount || return 1
3319 $LFS setstripe -c -1 $MOUNT/d1
3320 #define OBD_FAIL_MDS_REINT_DELAY 0x142
3321 do_facet $SINGLEMDS "lctl set_param fail_loc=0x142"
3322 touch $MOUNT/d1/f1 &
3325 start_ost2 || return 2
3327 stop_ost2 || return 3
3329 #writeconf to remove all ost2 traces for subsequent tests
3330 writeconf_or_reformat
3332 run_test 51 "Verify that mdt_reint handles RMF_MDT_MD correctly when an OST is added"
3341 do_node $node mkdir -p $dest
3342 [ $? -eq 0 ] || { error "Unable to create directory"; return 1; }
3344 do_node $node 'tar cf - '$@' | tar xf - -C '$dest';
3345 [ \"\${PIPESTATUS[*]}\" = \"0 0\" ] || exit 1'
3346 [ $? -eq 0 ] || { error "Unable to tar files"; return 2; }
3348 do_node $node 'getfattr -d -m "[a-z]*\\." '$@' > '$xattrs
3349 [ $? -eq 0 ] || { error "Unable to read xattrs"; return 3; }
3359 local backup2=${TMP}/backup2
3361 do_node $node mkdir -p $backup2
3362 [ $? -eq 0 ] || { error "Unable to create directory"; return 1; }
3364 do_node $node 'tar cf - '$@' | tar xf - -C '$backup2';
3365 [ \"\${PIPESTATUS[*]}\" = \"0 0\" ] || exit 1'
3366 [ $? -eq 0 ] || { error "Unable to tar files to diff"; return 2; }
3368 do_node $node "diff -rq $backup $backup2"
3369 [ $? -eq 0 ] || { error "contents differ"; return 3; }
3371 local xattrs2=${TMP}/xattrs2
3372 do_node $node 'getfattr -d -m "[a-z]*\\." '$@' > '$xattrs2
3373 [ $? -eq 0 ] || { error "Unable to read xattrs to diff"; return 4; }
3375 do_node $node "diff $xattrs $xattrs2"
3376 [ $? -eq 0 ] || { error "xattrs differ"; return 5; }
3378 do_node $node "rm -rf $backup2 $xattrs2"
3379 [ $? -eq 0 ] || { error "Unable to delete temporary files"; return 6; }
3383 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
3384 skip "Only applicable to ldiskfs-based MDTs"
3389 [ $? -eq 0 ] || { error "Unable to start MDS"; return 1; }
3391 [ $? -eq 0 ] || { error "Unable to start OST1"; return 2; }
3393 [ $? -eq 0 ] || { error "Unable to mount client"; return 3; }
3396 local ost1mnt=$(facet_mntpt ost1)
3397 local ost1node=$(facet_active_host ost1)
3398 local ost1tmp=$TMP/conf52
3402 [ $? -eq 0 ] || { error "Unable to create tdir"; return 4; }
3403 touch $TMP/modified_first
3404 [ $? -eq 0 ] || { error "Unable to create temporary file"; return 5; }
3405 local mtime=$(stat -c %Y $TMP/modified_first)
3406 do_node $ost1node "mkdir -p $ost1tmp && touch -m -d @$mtime $ost1tmp/modified_first"
3408 [ $? -eq 0 ] || { error "Unable to create temporary file"; return 6; }
3411 $LFS setstripe -c -1 -S 1M $DIR/$tdir
3412 [ $? -eq 0 ] || { error "lfs setstripe failed"; return 7; }
3414 for (( i=0; i < nrfiles; i++ )); do
3415 multiop $DIR/$tdir/$tfile-$i Ow1048576w1048576w524288c
3416 [ $? -eq 0 ] || { error "multiop failed"; return 8; }
3422 echo backup files to $TMP/files
3423 local files=$(find $DIR/$tdir -type f -newer $TMP/modified_first)
3424 copy_files_xattrs `hostname` $TMP/files $TMP/file_xattrs $files
3425 [ $? -eq 0 ] || { error "Unable to copy files"; return 9; }
3427 umount_client $MOUNT
3428 [ $? -eq 0 ] || { error "Unable to umount client"; return 10; }
3430 [ $? -eq 0 ] || { error "Unable to stop ost1"; return 11; }
3432 echo mount ost1 as ldiskfs
3433 do_node $ost1node mkdir -p $ost1mnt
3434 [ $? -eq 0 ] || { error "Unable to create $ost1mnt"; return 23; }
3435 if ! do_node $ost1node test -b $ost1_dev; then
3438 do_node $ost1node mount -t $(facet_fstype ost1) $loop $ost1_dev \
3440 [ $? -eq 0 ] || { error "Unable to mount ost1 as ldiskfs"; return 12; }
3443 echo backup objects to $ost1tmp/objects
3444 local objects=$(do_node $ost1node 'find '$ost1mnt'/O/[0-9]* -type f'\
3445 '-size +0 -newer '$ost1tmp'/modified_first -regex ".*\/[0-9]+"')
3446 copy_files_xattrs $ost1node $ost1tmp/objects $ost1tmp/object_xattrs \
3448 [ $? -eq 0 ] || { error "Unable to copy objects"; return 13; }
3450 # move objects to lost+found
3451 do_node $ost1node 'mv '$objects' '${ost1mnt}'/lost+found'
3452 [ $? -eq 0 ] || { error "Unable to move objects"; return 14; }
3454 # recover objects dry-run
3455 echo "ll_recover_lost_found_objs dry_run"
3456 do_node $ost1node "ll_recover_lost_found_objs -n -d $ost1mnt/O"
3457 [ $? -eq 0 ] || { error "ll_recover_lost_found_objs failed"; return 15; }
3460 echo "ll_recover_lost_found_objs fix run"
3461 do_node $ost1node "ll_recover_lost_found_objs -d $ost1mnt/lost+found"
3462 [ $? -eq 0 ] || { error "ll_recover_lost_found_objs failed"; return 15; }
3464 # compare restored objects against saved ones
3465 diff_files_xattrs $ost1node $ost1tmp/objects $ost1tmp/object_xattrs $objects
3466 [ $? -eq 0 ] || { error "Unable to diff objects"; return 16; }
3468 do_node $ost1node "umount $ost1mnt"
3469 [ $? -eq 0 ] || { error "Unable to umount ost1 as ldiskfs"; return 17; }
3472 [ $? -eq 0 ] || { error "Unable to start ost1"; return 18; }
3474 [ $? -eq 0 ] || { error "Unable to mount client"; return 19; }
3477 diff_files_xattrs `hostname` $TMP/files $TMP/file_xattrs $files
3478 [ $? -eq 0 ] || { error "Unable to diff files"; return 20; }
3480 rm -rf $TMP/files $TMP/file_xattrs
3481 [ $? -eq 0 ] || { error "Unable to delete temporary files"; return 21; }
3482 do_node $ost1node "rm -rf $ost1tmp"
3483 [ $? -eq 0 ] || { error "Unable to delete temporary files"; return 22; }
3486 run_test 52 "check recovering objects from lost+found"
3488 # Checks threads_min/max/started for some service
3490 # Arguments: service name (OST or MDT), facet (e.g., ost1, $SINGLEMDS), and a
3491 # parameter pattern prefix like 'ost.*.ost'.
3504 local msg="Insane $modname thread counts"
3505 local ncpts=$(check_cpt_number $facet)
3509 check_mount || return 41
3511 # We need to expand $parampat, but it may match multiple parameters, so
3512 # we'll pick the first one
3513 if ! paramp=$(do_facet $facet "lctl get_param -N ${parampat}.threads_min"|head -1); then
3514 error "Couldn't expand ${parampat}.threads_min parameter name"
3518 # Remove the .threads_min part
3519 paramp=${paramp%.threads_min}
3521 # Check for sanity in defaults
3522 tmin=$(do_facet $facet "lctl get_param -n ${paramp}.threads_min" || echo 0)
3523 tmax=$(do_facet $facet "lctl get_param -n ${paramp}.threads_max" || echo 0)
3524 tstarted=$(do_facet $facet "lctl get_param -n ${paramp}.threads_started" || echo 0)
3525 lassert 23 "$msg (PDSH problems?)" '(($tstarted && $tmin && $tmax))' || return $?
3526 lassert 24 "$msg" '(($tstarted >= $tmin && $tstarted <= $tmax ))' || return $?
3527 nthrs=$(expr $tmax - $tmin)
3528 if [ $nthrs -lt $ncpts ]; then
3534 [ $tmin -eq $tmax -a $tmin -eq $tstarted ] &&
3535 skip_env "module parameter forced $facet thread count" &&
3536 tmin=3 && tmax=$((3 * tmax))
3538 # Check that we can change min/max
3539 do_facet $facet "lctl set_param ${paramp}.threads_min=$((tmin + nthrs))"
3540 do_facet $facet "lctl set_param ${paramp}.threads_max=$((tmax - nthrs))"
3541 tmin2=$(do_facet $facet "lctl get_param -n ${paramp}.threads_min" || echo 0)
3542 tmax2=$(do_facet $facet "lctl get_param -n ${paramp}.threads_max" || echo 0)
3543 lassert 25 "$msg" '(($tmin2 == ($tmin + $nthrs) && $tmax2 == ($tmax - $nthrs)))' || return $?
3545 # Check that we can set min/max to the same value
3546 tmin=$(do_facet $facet "lctl get_param -n ${paramp}.threads_min" || echo 0)
3547 do_facet $facet "lctl set_param ${paramp}.threads_max=$tmin"
3548 tmin2=$(do_facet $facet "lctl get_param -n ${paramp}.threads_min" || echo 0)
3549 tmax2=$(do_facet $facet "lctl get_param -n ${paramp}.threads_max" || echo 0)
3550 lassert 26 "$msg" '(($tmin2 == $tmin && $tmax2 == $tmin))' || return $?
3552 # Check that we can't set max < min
3553 do_facet $facet "lctl set_param ${paramp}.threads_max=$((tmin - 1))"
3554 tmin2=$(do_facet $facet "lctl get_param -n ${paramp}.threads_min" || echo 0)
3555 tmax2=$(do_facet $facet "lctl get_param -n ${paramp}.threads_max" || echo 0)
3556 lassert 27 "$msg" '(($tmin2 <= $tmax2))' || return $?
3558 # We need to ensure that we get the module options desired; to do this
3559 # we set LOAD_MODULES_REMOTE=true and we call setmodopts below.
3560 LOAD_MODULES_REMOTE=true
3563 local newvalue="${opts}=$(expr $basethr \* $ncpts)"
3564 setmodopts -a $modname "$newvalue" oldvalue
3568 check_mount || return 41
3570 # Restore previous setting of MODOPTS_*
3571 setmodopts $modname "$oldvalue"
3573 # Check that $opts took
3574 tmin=$(do_facet $facet "lctl get_param -n ${paramp}.threads_min")
3575 tmax=$(do_facet $facet "lctl get_param -n ${paramp}.threads_max")
3576 tstarted=$(do_facet $facet "lctl get_param -n ${paramp}.threads_started")
3577 lassert 28 "$msg" '(($tstarted >= $tmin && $tstarted <= $tmax ))' || return $?
3586 thread_sanity OST ost1 'ost.*.ost' 'oss_num_threads' '16'
3589 run_test 53a "check OSS thread count params"
3593 local mds=$(do_facet $SINGLEMDS "lctl get_param -N mds.*.*.threads_max \
3595 if [ -z "$mds" ]; then
3596 #running this on an old MDT
3597 thread_sanity MDT $SINGLEMDS 'mdt.*.*.' 'mdt_num_threads' 16
3599 thread_sanity MDT $SINGLEMDS 'mds.*.*.' 'mds_num_threads' 16
3603 run_test 53b "check MDS thread count params"
3606 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
3607 skip "Only applicable to ldiskfs-based MDTs"
3611 do_rpc_nodes $(facet_host ost1) run_llverdev $(ostdevname 1) -p
3612 [ $? -eq 0 ] || error "llverdev failed!"
3615 run_test 54a "test llverdev and partial verify of device"
3618 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
3619 skip "Only applicable to ldiskfs-based MDTs"
3624 run_llverfs $MOUNT -p
3625 [ $? -eq 0 ] || error "llverfs failed!"
3628 run_test 54b "test llverfs and partial verify of filesystem"
3632 local max_ost_index=$1
3633 echo -n $(((max_ost_index + 1) * 8))
3637 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
3638 skip "Only applicable to ldiskfs-based MDTs"
3642 local mdsdev=$(mdsdevname 1)
3643 local mdsvdev=$(mdsvdevname 1)
3647 add mds1 $(mkfs_opts mds1 ${mdsdev}) --reformat $mdsdev \
3649 add ost1 $(mkfs_opts ost1 $(ostdevname 1)) --index=$i \
3650 --reformat $(ostdevname 1) $(ostvdevname 1)
3656 echo checking size of lov_objid for ost index $i
3657 LOV_OBJID_SIZE=$(do_facet mds1 "$DEBUGFS -R 'stat lov_objid' $mdsdev 2>/dev/null" | grep ^User | awk '{print $6}')
3658 if [ "$LOV_OBJID_SIZE" != $(lov_objid_size $i) ]; then
3659 error "lov_objid size has to be $(lov_objid_size $i), not $LOV_OBJID_SIZE"
3661 echo ok, lov_objid size is correct: $LOV_OBJID_SIZE
3668 run_test 55 "check lov_objid size"
3671 local mds_journal_size_orig=$MDSJOURNALSIZE
3675 for num in $(seq 1 $MDSCOUNT); do
3676 add mds${num} $(mkfs_opts mds${num} $(mdsdevname $num)) \
3677 --reformat $(mdsdevname $num) $(mdsvdevname $num)
3679 add ost1 $(mkfs_opts ost1 $(ostdevname 1)) --index=1000 --reformat \
3680 $(ostdevname 1) $(ostvdevname 1)
3681 add ost2 $(mkfs_opts ost2 $(ostdevname 2)) --index=10000 --reformat \
3682 $(ostdevname 2) $(ostvdevname 2)
3686 start_ost2 || error "Unable to start second ost"
3687 mount_client $MOUNT || error "Unable to mount client"
3691 MDSJOURNALSIZE=$mds_journal_size_orig
3694 run_test 56 "check big indexes"
3696 test_57a() { # bug 22656
3697 local NID=$(do_facet ost1 "$LCTL get_param nis" | tail -1 | awk '{print $1}')
3698 writeconf_or_reformat
3699 [ $(facet_fstype ost1) == zfs ] && import_zpool ost1
3700 do_facet ost1 "$TUNEFS --failnode=$NID `ostdevname 1`" || error "tunefs failed"
3702 start_ost && error "OST registration from failnode should fail"
3705 run_test 57a "initial registration from failnode should fail (should return errs)"
3708 local NID=$(do_facet ost1 "$LCTL get_param nis" | tail -1 | awk '{print $1}')
3709 writeconf_or_reformat
3710 [ $(facet_fstype ost1) == zfs ] && import_zpool ost1
3711 do_facet ost1 "$TUNEFS --servicenode=$NID `ostdevname 1`" || error "tunefs failed"
3713 start_ost || error "OST registration from servicenode should not fail"
3716 run_test 57b "initial registration from servicenode should not fail"
3719 do_facet mgs $LCTL get_param mgs.MGS.live.$FSNAME | grep OST | wc -l
3722 test_58() { # bug 22658
3725 createmany -o $DIR/$tdir/$tfile-%d 100
3726 # make sure that OSTs do not cancel llog cookies before we unmount the MDS
3727 #define OBD_FAIL_OBD_LOG_CANCEL_NET 0x601
3728 do_facet $SINGLEMDS "lctl set_param fail_loc=0x601"
3729 unlinkmany $DIR/$tdir/$tfile-%d 100
3732 local MNTDIR=$(facet_mntpt $SINGLEMDS)
3733 local devname=$(mdsdevname ${SINGLEMDS//mds/})
3735 # remove all files from the OBJECTS dir
3736 mount_fstype $SINGLEMDS
3738 do_facet $SINGLEMDS "find $MNTDIR/O/1/d* -type f -delete"
3740 unmount_fstype $SINGLEMDS
3741 # restart MDS with missing llog files
3743 do_facet mds "lctl set_param fail_loc=0"
3746 run_test 58 "missing llog files must not prevent MDT from mounting"
3749 start_mgsmds >> /dev/null
3750 local C1=$(count_osts)
3751 if [ $C1 -eq 0 ]; then
3752 start_ost >> /dev/null
3756 echo "original ost count: $C1 (expect > 0)"
3757 [ $C1 -gt 0 ] || error "No OSTs in $FSNAME log"
3758 start_mgsmds -o writeconf >> /dev/null || error "MDT start failed"
3759 local C2=$(count_osts)
3760 echo "after mdt writeconf count: $C2 (expect 0)"
3761 [ $C2 -gt 0 ] && error "MDT writeconf should erase OST logs"
3762 echo "OST start without writeconf should fail:"
3763 start_ost >> /dev/null && error "OST start without writeconf didn't fail"
3764 echo "OST start with writeconf should succeed:"
3765 start_ost -o writeconf >> /dev/null || error "OST1 start failed"
3766 local C3=$(count_osts)
3767 echo "after ost writeconf count: $C3 (expect 1)"
3768 [ $C3 -eq 1 ] || error "new OST writeconf should add:"
3769 start_ost2 -o writeconf >> /dev/null || error "OST2 start failed"
3770 local C4=$(count_osts)
3771 echo "after ost2 writeconf count: $C4 (expect 2)"
3772 [ $C4 -eq 2 ] || error "OST2 writeconf should add log"
3773 stop_ost2 >> /dev/null
3774 cleanup_nocli >> /dev/null
3775 #writeconf to remove all ost2 traces for subsequent tests
3776 writeconf_or_reformat
3778 run_test 59 "writeconf mount option"
3780 test_60() { # LU-471
3783 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
3784 skip "Only applicable to ldiskfs-based MDTs"
3788 for num in $(seq $MDSCOUNT); do
3789 add mds${num} $(mkfs_opts mds${num} $(mdsdevname $num)) \
3790 --mkfsoptions='\" -E stride=64 -O ^uninit_bg\"' \
3791 --reformat $(mdsdevname $num) $(mdsvdevname $num) ||
3795 dump=$(do_facet $SINGLEMDS dumpe2fs $(mdsdevname 1))
3797 [ $rc -eq 0 ] || error "dumpe2fs $(mdsdevname 1) failed"
3799 # MDT default has dirdata feature
3800 echo $dump | grep dirdata > /dev/null || error "dirdata is not set"
3801 # we disable uninit_bg feature
3802 echo $dump | grep uninit_bg > /dev/null && error "uninit_bg is set"
3803 # we set stride extended options
3804 echo $dump | grep stride > /dev/null || error "stride is not set"
3807 run_test 60 "check mkfs.lustre --mkfsoptions -E -O options setting"
3810 local reformat=false
3812 [ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.1.53) ] ||
3813 { skip "Need MDS version at least 2.1.53"; return 0; }
3815 if [ $(facet_fstype $SINGLEMDS) == ldiskfs ] &&
3816 ! large_xattr_enabled; then
3818 LDISKFS_MKFS_OPTS+=" -O large_xattr"
3820 for num in $(seq $MDSCOUNT); do
3821 add mds${num} $(mkfs_opts mds$num $(mdsdevname $num)) \
3822 --reformat $(mdsdevname $num) $(mdsvdevname $num) ||
3823 error "add mds $num failed"
3827 setup_noconfig || error "setting up the filesystem failed"
3828 client_up || error "starting client failed"
3830 local file=$DIR/$tfile
3833 local large_value="$(generate_string $(max_xattr_size))"
3834 local small_value="bar"
3836 local name="trusted.big"
3837 log "save large xattr $name on $file"
3838 setfattr -n $name -v $large_value $file ||
3839 error "saving $name on $file failed"
3841 local new_value=$(get_xattr_value $name $file)
3842 [[ "$new_value" != "$large_value" ]] &&
3843 error "$name different after saving"
3845 log "shrink value of $name on $file"
3846 setfattr -n $name -v $small_value $file ||
3847 error "shrinking value of $name on $file failed"
3849 new_value=$(get_xattr_value $name $file)
3850 [[ "$new_value" != "$small_value" ]] &&
3851 error "$name different after shrinking"
3853 log "grow value of $name on $file"
3854 setfattr -n $name -v $large_value $file ||
3855 error "growing value of $name on $file failed"
3857 new_value=$(get_xattr_value $name $file)
3858 [[ "$new_value" != "$large_value" ]] &&
3859 error "$name different after growing"
3861 log "check value of $name on $file after remounting MDS"
3863 new_value=$(get_xattr_value $name $file)
3864 [[ "$new_value" != "$large_value" ]] &&
3865 error "$name different after remounting MDS"
3867 log "remove large xattr $name from $file"
3868 setfattr -x $name $file || error "removing $name from $file failed"
3873 LDISKFS_MKFS_OPTS=${LDISKFS_MKFS_OPTS% -O large_xattr}
3877 run_test 61 "large xattr"
3880 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
3881 skip "Only applicable to ldiskfs-based MDTs"
3886 local mdsdev=$(mdsdevname 1)
3887 local ostdev=$(ostdevname 1)
3889 [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.2.51) ]] ||
3890 { skip "Need MDS version at least 2.2.51"; return 0; }
3892 echo "disable journal for mds"
3893 do_facet mds tune2fs -O ^has_journal $mdsdev || error "tune2fs failed"
3894 start_mds && error "MDT start should fail"
3895 echo "disable journal for ost"
3896 do_facet ost1 tune2fs -O ^has_journal $ostdev || error "tune2fs failed"
3897 start_ost && error "OST start should fail"
3898 cleanup || return $?
3901 run_test 62 "start with disabled journal"
3904 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
3905 skip "Only applicable to ldiskfs-based MDTs"
3909 local inode_slab=$(do_facet $SINGLEMDS \
3910 "awk '/ldiskfs_inode_cache/ { print \\\$5 }' /proc/slabinfo")
3911 if [ -z "$inode_slab" ]; then
3912 skip "ldiskfs module has not been loaded"
3916 echo "$inode_slab ldisk inodes per page"
3917 [ "$inode_slab" -ge "3" ] ||
3918 error "ldisk inode size is too big, $inode_slab objs per page"
3921 run_test 63 "Verify each page can at least hold 3 ldisk inodes"
3926 start_ost2 || error "Unable to start second ost"
3927 mount_client $MOUNT || error "Unable to mount client"
3928 stop_ost2 || error "Unable to stop second ost"
3930 $LFS df --lazy || error "lfs df failed"
3931 cleanup || return $?
3932 #writeconf to remove all ost2 traces for subsequent tests
3933 writeconf_or_reformat
3935 run_test 64 "check lfs df --lazy "
3937 test_65() { # LU-2237
3938 # Currently, the test is only valid for ldiskfs backend
3939 [ "$(facet_fstype $SINGLEMDS)" != "ldiskfs" ] &&
3940 skip "non-ldiskfs backend" && return
3942 local devname=$(mdsdevname ${SINGLEMDS//mds/})
3943 local brpt=$(facet_mntpt brpt)
3946 if ! do_facet $SINGLEMDS "test -b $devname"; then
3951 local obj=$(do_facet $SINGLEMDS \
3952 "$DEBUGFS -c -R \\\"stat last_rcvd\\\" $devname" |
3954 if [ -z "$obj" ]; then
3955 # The MDT may be just re-formatted, mount the MDT for the
3956 # first time to guarantee the "last_rcvd" file is there.
3957 start_mds || error "fail to mount the MDS for the first time"
3961 # remove the "last_rcvd" file
3962 do_facet $SINGLEMDS "mkdir -p $brpt"
3963 do_facet $SINGLEMDS \
3964 "mount -t $(facet_fstype $SINGLEMDS) $opts $devname $brpt"
3965 do_facet $SINGLEMDS "rm -f ${brpt}/last_rcvd"
3966 do_facet $SINGLEMDS "umount -d $brpt"
3968 # restart MDS, the "last_rcvd" file should be recreated.
3969 start_mds || error "fail to restart the MDS"
3971 obj=$(do_facet $SINGLEMDS \
3972 "$DEBUGFS -c -R \\\"stat last_rcvd\\\" $devname" | grep Inode)
3973 [ -n "$obj" ] || error "fail to re-create the last_rcvd"
3975 run_test 65 "re-create the lost last_rcvd file when server mount"
3978 [[ $(lustre_version_code mgs) -ge $(version_code 2.3.59) ]] ||
3979 { skip "Need MGS version at least 2.3.59"; return 0; }
3982 local OST1_NID=$(do_facet ost1 $LCTL list_nids | head -1)
3983 local MDS_NID=$(do_facet $SINGLEMDS $LCTL list_nids | head -1)
3985 echo "replace_nids should fail if MDS, OSTs and clients are UP"
3986 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 $OST1_NID &&
3987 error "replace_nids fail"
3989 umount_client $MOUNT || error "unmounting client failed"
3990 echo "replace_nids should fail if MDS and OSTs are UP"
3991 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 $OST1_NID &&
3992 error "replace_nids fail"
3995 echo "replace_nids should fail if MDS is UP"
3996 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 $OST1_NID &&
3997 error "replace_nids fail"
3999 stop_mds || error "stopping mds failed"
4001 if combined_mgs_mds; then
4002 start_mdt 1 "-o nosvc" ||
4003 error "starting mds with nosvc option failed"
4006 echo "command should accept two parameters"
4007 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 &&
4008 error "command should accept two params"
4010 echo "correct device name should be passed"
4011 do_facet mgs $LCTL replace_nids $FSNAME-WRONG0000 $OST1_NID &&
4012 error "wrong devname"
4014 echo "wrong nids list should not destroy the system"
4015 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 "wrong nids list" &&
4018 echo "replace OST nid"
4019 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 $OST1_NID ||
4020 error "replace nids failed"
4022 echo "command should accept two parameters"
4023 do_facet mgs $LCTL replace_nids $FSNAME-MDT0000 &&
4024 error "command should accept two params"
4026 echo "wrong nids list should not destroy the system"
4027 do_facet mgs $LCTL replace_nids $FSNAME-MDT0000 "wrong nids list" &&
4030 echo "replace MDS nid"
4031 do_facet mgs $LCTL replace_nids $FSNAME-MDT0000 $MDS_NID ||
4032 error "replace nids failed"
4034 if ! combined_mgs_mds ; then
4041 check_mount || error "error after nid replace"
4042 cleanup || error "cleanup failed"
4045 run_test 66 "replace nids"
4047 test_67() { #LU-2950
4048 local legacy="$TMP/legacy_lnet_config"
4049 local new="$TMP/new_routes_test"
4050 local out="$TMP/config_out_file"
4051 local verify="$TMP/conv_verify"
4052 local verify_conf="$TMP/conf_verify"
4054 # Create the legacy file that will be run through the
4055 # lustre_routes_conversion script
4056 cat <<- LEGACY_LNET_CONFIG > $legacy
4057 tcp1 23 192.168.213.1@tcp:1; tcp5 34 193.30.4.3@tcp:4;
4058 tcp2 54 10.1.3.2@tcp;
4059 tcp3 10.3.4.3@tcp:3;
4063 # Create the verification file to verify the output of
4064 # lustre_routes_conversion script against.
4065 cat <<- VERIFY_LNET_CONFIG > $verify
4066 tcp1: { gateway: 192.168.213.1@tcp, hop: 23, priority: 1 }
4067 tcp5: { gateway: 193.30.4.3@tcp, hop: 34, priority: 4 }
4068 tcp2: { gateway: 10.1.3.2@tcp, hop: 54 }
4069 tcp3: { gateway: 10.3.4.3@tcp, priority: 3 }
4070 tcp4: { gateway: 10.3.3.4@tcp }
4073 # Create the verification file to verify the output of
4074 # lustre_routes_config script against
4075 cat <<- VERIFY_LNET_CONFIG > $verify_conf
4076 lctl --net tcp1 add_route 192.168.213.1@tcp 23 1
4077 lctl --net tcp5 add_route 193.30.4.3@tcp 34 4
4078 lctl --net tcp2 add_route 10.1.3.2@tcp 54 4
4079 lctl --net tcp3 add_route 10.3.4.3@tcp 1 3
4080 lctl --net tcp4 add_route 10.3.3.4@tcp 1 3
4083 lustre_routes_conversion $legacy $new > /dev/null
4084 if [ -f $new ]; then
4085 # verify the conversion output
4086 cmp -s $new $verify > /dev/null
4087 if [ $? -eq 1 ]; then
4088 error "routes conversion failed"
4091 lustre_routes_config --dry-run --verbose $new > $out
4092 # check that the script succeeded
4093 cmp -s $out $verify_conf > /dev/null
4094 if [ $? -eq 1 ]; then
4095 error "routes config failed"
4098 error "routes conversion test failed"
4100 # remove generated files
4101 rm -f $new $legacy $verify $verify_conf $out
4103 run_test 67 "test routes conversion and configuration"
4111 [ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.4.53) ] ||
4112 { skip "Need MDS version at least 2.4.53"; return 0; }
4114 umount_client $MOUNT || error "umount client failed"
4116 start_mdt 1 || error "MDT start failed"
4119 # START-END - the sequences we'll be reserving
4120 START=$(do_facet $SINGLEMDS \
4121 lctl get_param -n seq.ctl*.space | awk -F'[[ ]' '{print $2}')
4122 END=$((START + (1 << 30)))
4123 do_facet $SINGLEMDS \
4124 lctl set_param seq.ctl*.fldb="[$START-$END\):0:mdt"
4126 # reset the sequences MDT0000 has already assigned
4127 do_facet $SINGLEMDS \
4128 lctl set_param seq.srv*MDT0000.space=clear
4130 # remount to let the client allocate new sequence
4131 mount_client $MOUNT || error "mount client failed"
4134 do_facet $SINGLEMDS \
4135 lctl get_param seq.srv*MDT0000.space
4136 $LFS path2fid $DIR/$tfile
4138 local old_ifs="$IFS"
4140 fid=($($LFS path2fid $DIR/$tfile))
4144 if [[ $seq < $END ]]; then
4145 error "used reserved sequence $seq?"
4147 cleanup || return $?
4149 run_test 68 "be able to reserve specific sequences in FLDB"
4152 local server_version=$(lustre_version_code $SINGLEMDS)
4154 [[ $server_version -lt $(version_code 2.4.2) ]] &&
4155 skip "Need MDS version at least 2.4.2" && return
4157 [[ $server_version -ge $(version_code 2.4.50) ]] &&
4158 [[ $server_version -lt $(version_code 2.5.0) ]] &&
4159 skip "Need MDS version at least 2.5.0" && return
4163 # use OST0000 since it probably has the most creations
4164 local OSTNAME=$(ostname_from_index 0)
4165 local mdtosc_proc1=$(get_mdtosc_proc_path mds1 $OSTNAME)
4166 local last_id=$(do_facet mds1 lctl get_param -n \
4167 osc.$mdtosc_proc1.prealloc_last_id)
4169 # Want to have OST LAST_ID over 1.5 * OST_MAX_PRECREATE to
4170 # verify that the LAST_ID recovery is working properly. If
4171 # not, then the OST will refuse to allow the MDS connect
4172 # because the LAST_ID value is too different from the MDS
4173 #define OST_MAX_PRECREATE=20000
4174 local num_create=$((20000 * 5))
4177 $LFS setstripe -i 0 $DIR/$tdir
4178 createmany -o $DIR/$tdir/$tfile- $num_create ||
4179 error "createmany: failed to create $num_create files: $?"
4180 # delete all of the files with objects on OST0 so the
4181 # filesystem is not inconsistent later on
4182 $LFS find $MOUNT --ost 0 | xargs rm
4184 stop_ost || error "OST0 stop failure"
4185 add ost1 $(mkfs_opts ost1 $(ostdevname 1)) --reformat --replace \
4186 $(ostdevname 1) $(ostvdevname 1) ||
4187 error "reformat and replace $ostdev failed"
4188 start_ost || error "OST0 restart failure"
4189 wait_osc_import_state mds ost FULL
4191 touch $DIR/$tdir/$tfile-last || error "create file after reformat"
4192 local idx=$($LFS getstripe -i $DIR/$tdir/$tfile-last)
4193 [ $idx -ne 0 ] && error "$DIR/$tdir/$tfile-last on $idx not 0" || true
4197 run_test 69 "replace an OST with the same index"
4200 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
4205 start_mdt 1 || error "MDT0 start fail"
4207 start_ost || error "OST0 start fail"
4209 start_mdt 2 || error "MDT1 start fail"
4211 mount_client $MOUNT || error "mount client fails"
4213 mkdir -p $DIR/$tdir || error "create dir fail"
4215 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
4216 error "create remote dir fail"
4218 rm -rf $DIR/$tdir || error "delete dir fail"
4219 cleanup || return $?
4221 run_test 70a "start MDT0, then OST, then MDT1"
4224 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
4227 start_ost || error "OST0 start fail"
4229 start_mdt 1 || error "MDT0 start fail"
4230 start_mdt 2 || error "MDT1 start fail"
4232 mount_client $MOUNT || error "mount client fails"
4234 mkdir -p $DIR/$tdir || error "create dir fail"
4236 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
4237 error "create remote dir fail"
4239 rm -rf $DIR/$tdir || error "delete dir fail"
4241 cleanup || return $?
4243 run_test 70b "start OST, MDT1, MDT0"
4246 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
4249 start_mdt 1 || error "MDT0 start fail"
4250 start_mdt 2 || error "MDT1 start fail"
4251 start_ost || error "OST0 start fail"
4253 mount_client $MOUNT || error "mount client fails"
4254 stop_mdt 1 || error "MDT1 start fail"
4256 local mdc_for_mdt1=$($LCTL dl | grep MDT0000-mdc | awk '{print $4}')
4257 echo "deactivate $mdc_for_mdt1"
4258 $LCTL --device $mdc_for_mdt1 deactivate || return 1
4260 mkdir -p $DIR/$tdir && error "mkdir succeed"
4262 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir &&
4263 error "create remote dir succeed"
4265 cleanup || return $?
4267 run_test 70c "stop MDT0, mkdir fail, create remote dir fail"
4270 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
4273 start_mdt 1 || error "MDT0 start fail"
4274 start_mdt 2 || error "MDT1 start fail"
4275 start_ost || error "OST0 start fail"
4277 mount_client $MOUNT || error "mount client fails"
4279 stop_mdt 2 || error "MDT1 start fail"
4281 local mdc_for_mdt2=$($LCTL dl | grep MDT0001-mdc |
4283 echo "deactivate $mdc_for_mdt2"
4284 $LCTL --device $mdc_for_mdt2 deactivate ||
4285 error "set $mdc_for_mdt2 deactivate failed"
4287 mkdir -p $DIR/$tdir || error "mkdir fail"
4288 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir &&
4289 error "create remote dir succeed"
4291 rm -rf $DIR/$tdir || error "delete dir fail"
4293 cleanup || return $?
4295 run_test 70d "stop MDT1, mkdir succeed, create remote dir fail"
4298 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
4299 if combined_mgs_mds; then
4300 skip "needs separate MGS/MDT" && return
4304 start_mdt 1 || error "MDT0 start fail"
4305 start_ost || error "OST0 start fail"
4306 start_mdt 2 || error "MDT1 start fail"
4307 start_ost2 || error "OST1 start fail"
4309 mount_client $MOUNT || error "mount client fails"
4311 mkdir -p $DIR/$tdir || error "mkdir fail"
4312 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
4313 error "create remote dir succeed"
4315 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
4316 rm -rf $DIR/$tdir || error "delete dir fail"
4318 umount_client $MOUNT
4319 stop_mdt 1 || error "MDT0 stop fail"
4320 stop_mdt 2 || error "MDT1 stop fail"
4321 stop_ost || error "OST0 stop fail"
4322 stop_ost2 || error "OST1 stop fail"
4324 run_test 71a "start MDT0 OST0, MDT1, OST1"
4327 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
4328 if combined_mgs_mds; then
4329 skip "needs separate MGS/MDT" && return
4333 start_mdt 2 || error "MDT1 start fail"
4334 start_ost || error "OST0 start fail"
4335 start_mdt 1 || error "MDT0 start fail"
4336 start_ost2 || error "OST1 start fail"
4338 mount_client $MOUNT || error "mount client fails"
4340 mkdir -p $DIR/$tdir || error "mkdir fail"
4341 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
4342 error "create remote dir succeed"
4344 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
4345 rm -rf $DIR/$tdir || error "delete dir fail"
4347 umount_client $MOUNT
4348 stop_mdt 1 || error "MDT0 stop fail"
4349 stop_mdt 2 || error "MDT1 stop fail"
4350 stop_ost || error "OST0 stop fail"
4351 stop_ost2 || error "OST1 stop fail"
4353 run_test 71b "start MDT1, OST0, MDT0, OST1"
4356 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
4357 if combined_mgs_mds; then
4358 skip "needs separate MGS/MDT" && return
4362 start_ost || error "OST0 start fail"
4363 start_ost2 || error "OST1 start fail"
4364 start_mdt 2 || error "MDT1 start fail"
4365 start_mdt 1 || error "MDT0 start fail"
4367 mount_client $MOUNT || error "mount client fails"
4369 mkdir -p $DIR/$tdir || error "mkdir fail"
4370 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
4371 error "create remote dir succeed"
4373 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
4374 rm -rf $DIR/$tdir || error "delete dir fail"
4376 umount_client $MOUNT
4377 stop_mdt 1 || error "MDT0 stop fail"
4378 stop_mdt 2 || error "MDT1 stop fail"
4379 stop_ost || error "OST0 stop fail"
4380 stop_ost2 || error "OST1 stop fail"
4383 run_test 71c "start OST0, OST1, MDT1, MDT0"
4386 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
4387 if combined_mgs_mds; then
4388 skip "needs separate MGS/MDT" && return
4392 start_ost || error "OST0 start fail"
4393 start_mdt 2 || error "MDT0 start fail"
4394 start_mdt 1 || error "MDT0 start fail"
4395 start_ost2 || error "OST1 start fail"
4397 mount_client $MOUNT || error "mount client fails"
4399 mkdir -p $DIR/$tdir || error "mkdir fail"
4400 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
4401 error "create remote dir succeed"
4403 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
4404 rm -rf $DIR/$tdir || error "delete dir fail"
4406 umount_client $MOUNT
4407 stop_mdt 1 || error "MDT0 stop fail"
4408 stop_mdt 2 || error "MDT1 stop fail"
4409 stop_ost || error "OST0 stop fail"
4410 stop_ost2 || error "OST1 stop fail"
4413 run_test 71d "start OST0, MDT1, MDT0, OST1"
4416 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
4417 if combined_mgs_mds; then
4418 skip "needs separate MGS/MDT" && return
4422 start_ost || error "OST0 start fail"
4423 start_mdt 2 || error "MDT1 start fail"
4424 start_ost2 || error "OST1 start fail"
4425 start_mdt 1 || error "MDT0 start fail"
4427 mount_client $MOUNT || error "mount client fails"
4429 mkdir -p $DIR/$tdir || error "mkdir fail"
4430 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
4431 error "create remote dir succeed"
4433 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
4434 rm -rf $DIR/$tdir || error "delete dir fail"
4436 umount_client $MOUNT
4437 stop_mdt 1 || error "MDT0 stop fail"
4438 stop_mdt 2 || error "MDT1 stop fail"
4439 stop_ost || error "OST0 stop fail"
4440 stop_ost2 || error "OST1 stop fail"
4443 run_test 71e "start OST0, MDT1, OST1, MDT0"
4445 test_72() { #LU-2634
4446 local mdsdev=$(mdsdevname 1)
4447 local ostdev=$(ostdevname 1)
4448 local cmd="$E2FSCK -fnvd $mdsdev"
4451 [ "$(facet_fstype $SINGLEMDS)" != "ldiskfs" ] &&
4452 skip "ldiskfs only test" && return
4454 #tune MDT with "-O extents"
4456 for num in $(seq $MDSCOUNT); do
4457 add mds${num} $(mkfs_opts mds$num $(mdsdevname $num)) \
4458 --reformat $(mdsdevname $num) $(mdsvdevname $num) ||
4459 error "add mds $num failed"
4460 $TUNE2FS -O extents $(mdsdevname $num)
4463 add ost1 $(mkfs_opts ost1 $ostdev) --reformat $ostdev ||
4464 error "add $ostdev failed"
4465 start_mgsmds || error "start mds failed"
4466 start_ost || error "start ost failed"
4467 mount_client $MOUNT || error "mount client failed"
4469 #create some short symlinks
4471 createmany -o $DIR/$tdir/$tfile-%d $fn
4472 echo "create $fn short symlinks"
4473 for i in $(seq -w 1 $fn); do
4474 ln -s $DIR/$tdir/$tfile-$i $MOUNT/$tfile-$i
4479 umount_client $MOUNT || error "umount client failed"
4480 stop_mds || error "stop mds failed"
4481 stop_ost || error "stop ost failed"
4484 run_e2fsck $(facet_active_host $SINGLEMDS) $mdsdev "-n"
4486 run_test 72 "test fast symlink with extents flag enabled"
4488 test_73() { #LU-3006
4490 [ $(facet_fstype ost1) == zfs ] && import_zpool ost1
4491 do_facet ost1 "$TUNEFS --failnode=1.2.3.4@$NETTYPE $(ostdevname 1)" ||
4492 error "1st tunefs failed"
4493 start_mgsmds || error "start mds failed"
4494 start_ost || error "start ost failed"
4495 mount_client $MOUNT || error "mount client failed"
4496 lctl get_param -n osc.*OST0000-osc-[^M]*.import | grep failover_nids |
4497 grep 1.2.3.4@$NETTYPE || error "failover nids haven't changed"
4498 umount_client $MOUNT || error "umount client failed"
4502 run_test 73 "failnode to update from mountdata properly"
4504 test_75() { # LU-2374
4505 [[ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.4.1) ]] &&
4506 skip "Need MDS version at least 2.4.1" && return
4509 local opts_mds="$(mkfs_opts mds1 $(mdsdevname 1)) \
4510 --reformat $(mdsdevname 1) $(mdsvdevname 1)"
4511 local opts_ost="$(mkfs_opts ost1 $(ostdevname 1)) \
4512 --reformat $(ostdevname 1) $(ostvdevname 1)"
4514 #check with default parameters
4515 add mds1 $opts_mds || error "add mds1 failed for default params"
4516 add ost1 $opts_ost || error "add ost1 failed for default params"
4518 opts_mds=$(echo $opts_mds | sed -e "s/--mdt//")
4519 opts_mds=$(echo $opts_mds |
4520 sed -e "s/--index=$index/--index=$index --mdt/")
4521 opts_ost=$(echo $opts_ost | sed -e "s/--ost//")
4522 opts_ost=$(echo $opts_ost |
4523 sed -e "s/--index=$index/--index=$index --ost/")
4525 add mds1 $opts_mds || error "add mds1 failed for new params"
4526 add ost1 $opts_ost || error "add ost1 failed for new params"
4529 run_test 75 "The order of --index should be irrelevant"
4532 [[ $(lustre_version_code mgs) -ge $(version_code 2.4.52) ]] ||
4533 { skip "Need MDS version at least 2.4.52" && return 0; }
4535 local MDMB_PARAM="osc.*.max_dirty_mb"
4536 echo "Change MGS params"
4537 local MAX_DIRTY_MB=$($LCTL get_param -n $MDMB_PARAM |
4539 echo "max_dirty_mb: $MAX_DIRTY_MB"
4540 local NEW_MAX_DIRTY_MB=$((MAX_DIRTY_MB + MAX_DIRTY_MB))
4541 echo "new_max_dirty_mb: $NEW_MAX_DIRTY_MB"
4542 do_facet mgs $LCTL set_param -P $MDMB_PARAM=$NEW_MAX_DIRTY_MB
4543 wait_update $HOSTNAME "lctl get_param -n $MDMB_PARAM |
4544 head -1" $NEW_MAX_DIRTY_MB
4545 MAX_DIRTY_MB=$($LCTL get_param -n $MDMB_PARAM | head -1)
4546 echo "$MAX_DIRTY_MB"
4547 [ $MAX_DIRTY_MB = $NEW_MAX_DIRTY_MB ] ||
4548 error "error while apply max_dirty_mb"
4550 echo "Check the value is stored after remount"
4553 wait_update $HOSTNAME "lctl get_param -n $MDMB_PARAM |
4554 head -1" $NEW_MAX_DIRTY_MB
4555 MAX_DIRTY_MB=$($LCTL get_param -n $MDMB_PARAM | head -1)
4556 [ $MAX_DIRTY_MB = $NEW_MAX_DIRTY_MB ] ||
4557 error "max_dirty_mb is not saved after remount"
4559 echo "Change OST params"
4560 CLIENT_PARAM="obdfilter.*.client_cache_count"
4561 local CLIENT_CACHE_COUNT
4562 CLIENT_CACHE_COUNT=$(do_facet ost1 $LCTL get_param -n $CLIENT_PARAM |
4564 echo "client_cache_count: $CLIENT_CACHE_COUNT"
4565 NEW_CLIENT_CACHE_COUNT=$((CLIENT_CACHE_COUNT+CLIENT_CACHE_COUNT))
4566 echo "new_client_cache_count: $NEW_CLIENT_CACHE_COUNT"
4567 do_facet mgs $LCTL set_param -P $CLIENT_PARAM=$NEW_CLIENT_CACHE_COUNT
4568 wait_update $(facet_host ost1) "lctl get_param -n $CLIENT_PARAM |
4569 head -1" $NEW_CLIENT_CACHE_COUNT
4570 CLIENT_CACHE_COUNT=$(do_facet ost1 $LCTL get_param -n $CLIENT_PARAM |
4572 echo "$CLIENT_CACHE_COUNT"
4573 [ $CLIENT_CACHE_COUNT = $NEW_CLIENT_CACHE_COUNT ] ||
4574 error "error while apply client_cache_count"
4576 echo "Check the value is stored after remount"
4579 wait_update $(facet_host ost1) "lctl get_param -n $CLIENT_PARAM |
4580 head -1" $NEW_CLIENT_CACHE_COUNT
4581 CLIENT_CACHE_COUNT=$(do_facet ost1 $LCTL get_param -n $CLIENT_PARAM |
4583 echo "$CLIENT_CACHE_COUNT"
4584 [ $CLIENT_CACHE_COUNT = $NEW_CLIENT_CACHE_COUNT ] ||
4585 error "client_cache_count is not saved after remount"
4588 run_test 76a "set permanent params set_param -P"
4590 test_76b() { # LU-4783
4591 [[ $(lustre_version_code mgs) -ge $(version_code 2.5.57) ]] ||
4592 { skip "Need MGS version at least 2.5.57" && return 0; }
4595 do_facet mgs $LCTL get_param mgs.MGS.live.params ||
4596 error "start params log failed"
4599 run_test 76b "verify params log setup correctly"
4601 test_77() { # LU-3445
4602 local server_version=$(lustre_version_code $SINGLEMDS)
4604 [[ $server_version -ge $(version_code 2.2.60) ]] &&
4605 [[ $server_version -le $(version_code 2.4.0) ]] &&
4606 skip "Need MDS version < 2.2.60 or > 2.4.0" && return
4608 if [[ -z "$fs2ost_DEV" || -z "$fs2mds_DEV" ]]; then
4609 is_blkdev $SINGLEMDS $(mdsdevname ${SINGLEMDS//mds/}) &&
4610 skip_env "mixed loopback and real device not working" && return
4613 local fs2mdsdev=$(mdsdevname 1_2)
4614 local fs2ostdev=$(ostdevname 1_2)
4615 local fs2mdsvdev=$(mdsvdevname 1_2)
4616 local fs2ostvdev=$(ostvdevname 1_2)
4617 local fsname=test1234
4619 local failnid="$(h2$NETTYPE 1.2.3.4),$(h2$NETTYPE 4.3.2.1)"
4621 add fs2mds $(mkfs_opts mds1 $fs2mdsdev) --mgs --fsname=$fsname \
4622 --reformat $fs2mdsdev $fs2mdsvdev || error "add fs2mds failed"
4623 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS && trap cleanup_fs2 EXIT INT ||
4624 error "start fs2mds failed"
4626 mgsnid=$(do_facet fs2mds $LCTL list_nids | xargs | tr ' ' ,)
4627 [[ $mgsnid = *,* ]] || mgsnid+=",$mgsnid"
4629 add fs2ost $(mkfs_opts ost1 $fs2ostdev) --mgsnode=$mgsnid \
4630 --failnode=$failnid --fsname=$fsname \
4631 --reformat $fs2ostdev $fs2ostvdev ||
4632 error "add fs2ost failed"
4633 start fs2ost $fs2ostdev $OST_MOUNT_OPTS || error "start fs2ost failed"
4636 $MOUNT_CMD $mgsnid:/$fsname $MOUNT2 || error "mount $MOUNT2 failed"
4637 DIR=$MOUNT2 MOUNT=$MOUNT2 check_mount || error "check $MOUNT2 failed"
4640 run_test 77 "comma-separated MGS NIDs and failover node NIDs"
4643 [[ $(facet_fstype $SINGLEMDS) != ldiskfs ||
4644 $(facet_fstype ost1) != ldiskfs ]] &&
4645 skip "only applicable to ldiskfs-based MDTs and OSTs" && return
4647 # reformat the Lustre filesystem with a smaller size
4648 local saved_MDSSIZE=$MDSSIZE
4649 local saved_OSTSIZE=$OSTSIZE
4650 MDSSIZE=$((MDSSIZE - 20000))
4651 OSTSIZE=$((OSTSIZE - 20000))
4652 reformat || error "(1) reformat Lustre filesystem failed"
4653 MDSSIZE=$saved_MDSSIZE
4654 OSTSIZE=$saved_OSTSIZE
4656 # mount the Lustre filesystem
4657 setup_noconfig || error "(2) setup Lustre filesystem failed"
4660 log "create test files"
4664 mkdir -p $MOUNT/$tdir || error "(3) mkdir $MOUNT/$tdir failed"
4665 for i in $(seq $num_files); do
4666 file=$MOUNT/$tdir/$tfile-$i
4667 dd if=/dev/urandom of=$file count=1 bs=1M ||
4668 error "(4) create $file failed"
4671 # unmount the Lustre filesystem
4672 cleanup || error "(5) cleanup Lustre filesystem failed"
4674 # run e2fsck on the MDT and OST devices
4675 local mds_host=$(facet_active_host $SINGLEMDS)
4676 local ost_host=$(facet_active_host ost1)
4677 local mds_dev=$(mdsdevname ${SINGLEMDS//mds/})
4678 local ost_dev=$(ostdevname 1)
4680 run_e2fsck $mds_host $mds_dev "-y"
4681 run_e2fsck $ost_host $ost_dev "-y"
4683 # get the original block count of the MDT and OST filesystems
4684 local mds_orig_blks=$(get_block_count $SINGLEMDS $mds_dev)
4685 local ost_orig_blks=$(get_block_count ost1 $ost_dev)
4687 # expand the MDT and OST filesystems to the device size
4688 run_resize2fs $SINGLEMDS $mds_dev "" || error "expand $SINGLEMDS failed"
4689 run_resize2fs ost1 $ost_dev "" || error "expand ost1 failed"
4691 # run e2fsck on the MDT and OST devices again
4692 run_e2fsck $mds_host $mds_dev "-y"
4693 run_e2fsck $ost_host $ost_dev "-y"
4695 # mount the Lustre filesystem
4699 log "check files after expanding the MDT and OST filesystems"
4700 for i in $(seq $num_files); do
4701 file=$MOUNT/$tdir/$tfile-$i
4702 $CHECKSTAT -t file -s 1048576 $file ||
4703 error "(6) checkstat $file failed"
4707 log "create more files after expanding the MDT and OST filesystems"
4708 for i in $(seq $((num_files + 1)) $((num_files + 10))); do
4709 file=$MOUNT/$tdir/$tfile-$i
4710 dd if=/dev/urandom of=$file count=1 bs=1M ||
4711 error "(7) create $file failed"
4714 # unmount the Lustre filesystem
4715 cleanup || error "(8) cleanup Lustre filesystem failed"
4717 # run e2fsck on the MDT and OST devices
4718 run_e2fsck $mds_host $mds_dev "-y"
4719 run_e2fsck $ost_host $ost_dev "-y"
4721 # get the maximum block count of the MDT and OST filesystems
4722 local mds_max_blks=$(get_block_count $SINGLEMDS $mds_dev)
4723 local ost_max_blks=$(get_block_count ost1 $ost_dev)
4725 # get the minimum block count of the MDT and OST filesystems
4726 local mds_min_blks=$(run_resize2fs $SINGLEMDS $mds_dev "" "-P" 2>&1 |
4727 grep minimum | sed -e 's/^.*filesystem: //g')
4728 local ost_min_blks=$(run_resize2fs ost1 $ost_dev "" "-P" 2>&1 |
4729 grep minimum | sed -e 's/^.*filesystem: //g')
4731 # shrink the MDT and OST filesystems to a smaller size
4735 if [[ $mds_max_blks -gt $mds_min_blks &&
4736 $mds_max_blks -gt $mds_orig_blks ]]; then
4737 [[ $mds_orig_blks -gt $mds_min_blks ]] &&
4738 base_blks=$mds_orig_blks || base_blks=$mds_min_blks
4739 new_blks=$(( (mds_max_blks - base_blks) / 2 + base_blks ))
4740 run_resize2fs $SINGLEMDS $mds_dev $new_blks ||
4741 error "shrink $SINGLEMDS to $new_blks failed"
4745 if [[ $ost_max_blks -gt $ost_min_blks &&
4746 $ost_max_blks -gt $ost_orig_blks ]]; then
4747 [[ $ost_orig_blks -gt $ost_min_blks ]] &&
4748 base_blks=$ost_orig_blks || base_blks=$ost_min_blks
4749 new_blks=$(( (ost_max_blks - base_blks) / 2 + base_blks ))
4750 run_resize2fs ost1 $ost_dev $new_blks ||
4751 error "shrink ost1 to $new_blks failed"
4755 # check whether the MDT or OST filesystem was shrunk or not
4757 combined_mgs_mds || stop_mgs || error "(9) stop mgs failed"
4758 reformat || error "(10) reformat Lustre filesystem failed"
4762 # run e2fsck on the MDT and OST devices again
4763 run_e2fsck $mds_host $mds_dev "-y"
4764 run_e2fsck $ost_host $ost_dev "-y"
4766 # mount the Lustre filesystem again
4770 log "check files after shrinking the MDT and OST filesystems"
4771 for i in $(seq $((num_files + 10))); do
4772 file=$MOUNT/$tdir/$tfile-$i
4773 $CHECKSTAT -t file -s 1048576 $file ||
4774 error "(11) checkstat $file failed"
4777 # unmount and reformat the Lustre filesystem
4778 cleanup || error "(12) cleanup Lustre filesystem failed"
4779 combined_mgs_mds || stop_mgs || error "(13) stop mgs failed"
4780 reformat || error "(14) reformat Lustre filesystem failed"
4782 run_test 78 "run resize2fs on MDT and OST filesystems"
4784 test_79() { # LU-4227
4785 [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.5.59) ]] ||
4786 { skip "Need MDS version at least 2.5.59"; return 0; }
4788 local mdsdev1=$(mdsdevname 1)
4789 local mdsvdev1=$(mdsvdevname 1)
4790 local mdsdev2=$(mdsdevname 2)
4791 local mdsvdev2=$(mdsvdevname 2)
4792 local ostdev1=$(ostdevname 1)
4793 local ostvdev1=$(ostvdevname 1)
4794 local opts_mds1="$(mkfs_opts mds1 $mdsdev1) --reformat"
4795 local opts_mds2="$(mkfs_opts mds2 $mdsdev2) --reformat"
4796 local opts_ost1="$(mkfs_opts ost1 $ostdev1) --reformat"
4799 # remove --mgs/--mgsnode from mkfs.lustre options
4800 opts_mds1=$(echo $opts_mds1 | sed -e "s/--mgs//")
4802 mgsnode_opt=$(echo $opts_mds2 |
4803 awk '{ for ( i = 1; i < NF; i++ )
4804 if ( $i ~ "--mgsnode" ) { print $i; break } }')
4805 [ -n $mgsnode_opt ] &&
4806 opts_mds2=$(echo $opts_mds2 | sed -e "s/$mgsnode_opt//")
4808 mgsnode_opt=$(echo $opts_ost1 |
4809 awk '{ for ( i = 1; i < NF; i++ )
4810 if ( $i ~ "--mgsnode" ) { print $i; break } }')
4811 [ -n $mgsnode_opt ] &&
4812 opts_ost1=$(echo $opts_ost1 | sed -e "s/$mgsnode_opt//")
4814 # -MGS, format a mdt without --mgs option
4815 add mds1 $opts_mds1 $mdsdev1 $mdsvdev1 &&
4816 error "Must specify --mgs when formatting mdt combined with mgs"
4818 # +MGS, format a mdt/ost without --mgsnode option
4819 add mds1 $(mkfs_opts mds1 $mdsdev1) --reformat $mdsdev1 $mdsvdev1 \
4820 > /dev/null || error "start mds1 failed"
4821 add mds2 $opts_mds2 $mdsdev2 $mdsvdev2 &&
4822 error "Must specify --mgsnode when formatting a mdt"
4823 add ost1 $opts_ost1 $ostdev1 $ostvdev2 &&
4824 error "Must specify --mgsnode when formatting an ost"
4828 run_test 79 "format MDT/OST without mgs option (should return errors)"
4833 uuid=$(do_facet ost1 lctl get_param -n mgc.*.uuid)
4834 #define OBD_FAIL_MGS_PAUSE_TARGET_CON 0x906
4835 do_facet ost1 "lctl set_param fail_val=10 fail_loc=0x906"
4836 do_facet mgs "lctl set_param fail_val=10 fail_loc=0x906"
4837 do_facet mgs "lctl set_param -n mgs/MGS/evict_client $uuid"
4841 do_facet ost1 "lctl set_param fail_loc=0"
4844 run_test 80 "mgc import reconnect race"
4846 if ! combined_mgs_mds ; then
4852 # restore the values of MDSSIZE and OSTSIZE
4853 MDSSIZE=$STORED_MDSSIZE
4854 OSTSIZE=$STORED_OSTSIZE