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 $SINGLEMDS) = "zfs" ] &&
77 # bug number for skipped test: LU-2778 LU-2059 LU-4444
78 ALWAYS_EXCEPT="$ALWAYS_EXCEPT 57b 50h 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="-o 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 " $MOUNT " /etc/mtab && \
450 error false "unexpected entry in mtab before mount" && return 10
452 [ "$(facet_fstype ost1)" = "zfs" ] &&
453 skip "LU-2059: no local config for ZFS OSTs" && return
459 mount_client $MOUNT || rc=1
461 grep " $MOUNT " /etc/mtab && \
462 error "$MOUNT entry in mtab after unmount" && rc=11
465 run_test 5d "mount with ost down"
468 grep " $MOUNT " /etc/mtab && \
469 error false "unexpected entry in mtab before mount" && return 10
475 #define OBD_FAIL_PTLRPC_DELAY_SEND 0x506
476 do_facet client "lctl set_param fail_loc=0x80000506"
477 mount_client $MOUNT || echo "mount failed (not fatal)"
479 grep " $MOUNT " /etc/mtab && \
480 error "$MOUNT entry in mtab after unmount" && rc=11
483 run_test 5e "delayed connect, don't crash (bug 10268)"
486 if combined_mgs_mds ; then
487 skip "combined mgs and mds"
491 grep " $MOUNT " /etc/mtab && \
492 error false "unexpected entry in mtab before mount" && return 10
496 [ -d $MOUNT ] || mkdir -p $MOUNT
497 mount_client $MOUNT &
499 echo client_mount pid is $pid
503 if ! ps -f -p $pid >/dev/null; then
506 grep " $MOUNT " /etc/mtab && echo "test 5f: mtab after mount"
507 error "mount returns $rc, expected to hang"
516 # mount should succeed after start mds
519 [ $rc -eq 0 ] || error "mount returned $rc"
520 grep " $MOUNT " /etc/mtab && echo "test 5f: mtab after mount"
524 run_test 5f "mds down, cleanup after failed mount (bug 2712)"
529 mount_client ${MOUNT} || return 87
530 touch $DIR/a || return 86
533 run_test 6 "manual umount, then mount again"
538 cleanup_nocli || return $?
540 run_test 7 "manual umount, then cleanup"
545 check_mount2 || return 45
546 umount_client $MOUNT2
549 run_test 8 "double mount setup"
554 do_facet ost1 lctl set_param debug=\'inode trace\' || return 1
555 do_facet ost1 lctl set_param subsystem_debug=\'mds ost\' || return 1
557 CHECK_PTLDEBUG="`do_facet ost1 lctl get_param -n debug`"
558 if [ "$CHECK_PTLDEBUG" ] && { \
559 [ "$CHECK_PTLDEBUG" = "trace inode warning error emerg console" ] ||
560 [ "$CHECK_PTLDEBUG" = "trace inode" ]; }; then
561 echo "lnet.debug success"
563 echo "lnet.debug: want 'trace inode', have '$CHECK_PTLDEBUG'"
566 CHECK_SUBSYS="`do_facet ost1 lctl get_param -n subsystem_debug`"
567 if [ "$CHECK_SUBSYS" ] && [ "$CHECK_SUBSYS" = "mds ost" ]; then
568 echo "lnet.subsystem_debug success"
570 echo "lnet.subsystem_debug: want 'mds ost', have '$CHECK_SUBSYS'"
573 stop_ost || return $?
575 run_test 9 "test ptldebug and subsystem for mkfs"
583 do_facet $facet "test -b $dev" || rc=1
584 if [[ "$size" ]]; then
585 local in=$(do_facet $facet "dd if=$dev of=/dev/null bs=1k count=1 skip=$size 2>&1" |\
586 awk '($3 == "in") { print $1 }')
587 [[ $in = "1+0" ]] || rc=1
593 # Test 16 was to "verify that lustre will correct the mode of OBJECTS".
594 # But with new MDS stack we don't care about the mode of local objects
595 # anymore, so this test is removed. See bug 22944 for more details.
599 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
600 skip "Only applicable to ldiskfs-based MDTs"
605 check_mount || return 41
608 echo "Remove mds config log"
609 if ! combined_mgs_mds ; then
613 do_facet mgs "$DEBUGFS -w -R 'unlink CONFIGS/$FSNAME-MDT0000' \
614 $(mgsdevname) || return \$?" || return $?
616 if ! combined_mgs_mds ; then
621 start_mds && return 42
624 run_test 17 "Verify failed mds_postsetup won't fail assertion (2936) (should return errs)"
627 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
628 skip "Only applicable to ldiskfs-based MDTs"
632 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
637 # check if current MDSSIZE is large enough
638 [ $MDSSIZE -ge $MIN ] && OK=1 && myMDSSIZE=$MDSSIZE && \
639 log "use MDSSIZE=$MDSSIZE"
641 # check if the global config has a large enough MDSSIZE
642 [ -z "$OK" -a ! -z "$STORED_MDSSIZE" ] && [ $STORED_MDSSIZE -ge $MIN ] && \
643 OK=1 && myMDSSIZE=$STORED_MDSSIZE && \
644 log "use STORED_MDSSIZE=$STORED_MDSSIZE"
646 # check if the block device is large enough
647 is_blkdev $SINGLEMDS $MDSDEV $MIN
648 local large_enough=$?
649 if [ -n "$OK" ]; then
650 [ $large_enough -ne 0 ] && OK=""
652 [ $large_enough -eq 0 ] && OK=1 && myMDSSIZE=$MIN &&
653 log "use device $MDSDEV with MIN=$MIN"
656 # check if a loopback device has enough space for fs metadata (5%)
658 if [ -z "$OK" ]; then
659 local SPACE=$(do_facet $SINGLEMDS "[ -f $MDSDEV -o ! -e $MDSDEV ] && df -P \\\$(dirname $MDSDEV)" |
660 awk '($1 != "Filesystem") {print $4}')
661 ! [ -z "$SPACE" ] && [ $SPACE -gt $((MIN / 20)) ] && \
662 OK=1 && myMDSSIZE=$MIN && \
663 log "use file $MDSDEV with MIN=$MIN"
666 [ -z "$OK" ] && skip_env "$MDSDEV too small for ${MIN}kB MDS" && return
669 echo "mount mds with large journal..."
671 local OLD_MDSSIZE=$MDSSIZE
675 echo "mount lustre system..."
677 check_mount || return 41
679 echo "check journal size..."
680 local FOUNDSIZE=$(do_facet $SINGLEMDS "$DEBUGFS -c -R 'stat <8>' $MDSDEV" | awk '/Size: / { print $NF; exit;}')
681 if [ $FOUNDSIZE -gt $((32 * 1024 * 1024)) ]; then
682 log "Success: mkfs creates large journals. Size: $((FOUNDSIZE >> 20))M"
684 error "expected journal size > 32M, found $((FOUNDSIZE >> 20))M"
692 run_test 18 "check mkfs creates large journals"
695 start_mds || return 1
696 stop_mds -f || return 2
698 run_test 19a "start/stop MDS without OSTs"
701 [ "$(facet_fstype ost1)" = "zfs" ] &&
702 skip "LU-2059: no local config for ZFS OSTs" && return
704 start_ost || return 1
705 stop_ost -f || return 2
707 run_test 19b "start/stop OSTs without MDS"
710 # first format the ost/mdt
714 check_mount || return 43
716 remount_client ro $MOUNT || return 44
717 touch $DIR/$tfile && echo "$DIR/$tfile created incorrectly" && return 45
718 [ -e $DIR/$tfile ] && echo "$DIR/$tfile exists incorrectly" && return 46
719 remount_client rw $MOUNT || return 47
721 [ ! -f $DIR/$tfile ] && echo "$DIR/$tfile missing" && return 48
722 MCNT=`grep -c $MOUNT /etc/mtab`
723 [ "$MCNT" -ne 1 ] && echo "$MOUNT in /etc/mtab $MCNT times" && return 49
728 run_test 20 "remount ro,rw mounts work and doesn't break /etc/mtab"
733 wait_osc_import_state mds ost FULL
737 run_test 21a "start mds before ost, stop ost first"
740 [ "$(facet_fstype ost1)" = "zfs" ] &&
741 skip "LU-2059: no local config for ZFS OSTs" && return
745 wait_osc_import_state mds ost FULL
749 run_test 21b "start ost before mds, stop mds first"
755 wait_osc_import_state mds ost2 FULL
759 #writeconf to remove all ost2 traces for subsequent tests
760 writeconf_or_reformat
762 run_test 21c "start mds between two osts, stop mds last"
765 if combined_mgs_mds ; then
766 skip "need separate mgs device" && return 0
776 wait_osc_import_state mds ost2 FULL
782 #writeconf to remove all ost2 traces for subsequent tests
783 writeconf_or_reformat
786 run_test 21d "start mgs then ost and then mds"
791 echo Client mount with ost in logs, but none running
793 # wait until mds connected to ost and open client connection
794 wait_osc_import_state mds ost FULL
797 # check_mount will block trying to contact ost
798 mcreate $DIR/$tfile || return 40
799 rm -f $DIR/$tfile || return 42
803 echo Client mount with a running ost
806 # if gss enabled, wait full time to let connection from
807 # mds to ost be established, due to the mismatch between
808 # initial connect timeout and gss context negotiation timeout.
809 # This perhaps could be remove after AT landed.
810 echo "sleep $((TIMEOUT + TIMEOUT + TIMEOUT))s"
811 sleep $((TIMEOUT + TIMEOUT + TIMEOUT))
814 wait_osc_import_state mds ost FULL
815 wait_osc_import_state client ost FULL
816 check_mount || return 41
821 run_test 22 "start a client before osts (should return errs)"
823 test_23a() { # was test_23
827 # force down client so that recovering mds waits for reconnect
828 local running=$(grep -c $MOUNT /proc/mounts) || true
829 if [ $running -ne 0 ]; then
830 echo "Stopping client $MOUNT (opts: -f)"
834 # enter recovery on mds
836 # try to start a new client
837 mount_client $MOUNT &
839 MOUNT_PID=$(ps -ef | grep "t lustre" | grep -v grep | awk '{print $2}')
840 MOUNT_LUSTRE_PID=`ps -ef | grep mount.lustre | grep -v grep | awk '{print $2}'`
841 echo mount pid is ${MOUNT_PID}, mount.lustre pid is ${MOUNT_LUSTRE_PID}
843 ps --ppid $MOUNT_LUSTRE_PID
844 echo "waiting for mount to finish"
846 # "ctrl-c" sends SIGINT but it usually (in script) does not work on child process
847 # SIGTERM works but it does not spread to offspring processses
848 kill -s TERM $MOUNT_PID
849 kill -s TERM $MOUNT_LUSTRE_PID
850 # we can not wait $MOUNT_PID because it is not a child of this shell
856 while [ "$WAIT" -lt "$MAX_WAIT" ]; do
858 PID1=$(ps -ef | awk '{print $2}' | grep -w $MOUNT_PID)
859 PID2=$(ps -ef | awk '{print $2}' | grep -w $MOUNT_LUSTRE_PID)
862 [ -z "$PID1" -a -z "$PID2" ] && break
863 echo "waiting for mount to finish ... "
864 WAIT=$(( WAIT + sleep))
866 if [ "$WAIT" -eq "$MAX_WAIT" ]; then
867 error "MOUNT_PID $MOUNT_PID and "\
868 "MOUNT_LUSTRE_PID $MOUNT_LUSTRE_PID still not killed in $WAIT secs"
871 stop_mds || error "stopping MDSes failed"
872 stop_ost || error "stopping OSSes failed"
874 run_test 23a "interrupt client during recovery mount delay"
879 test_23b() { # was test_23
882 # Simulate -EINTR during mount OBD_FAIL_LDLM_CLOSE_THREAD
883 lctl set_param fail_loc=0x80000313
887 run_test 23b "Simulate -EINTR during mount"
889 fs2mds_HOST=$mds_HOST
890 fs2ost_HOST=$ost_HOST
892 MDSDEV1_2=$fs2mds_DEV
893 OSTDEV1_2=$fs2ost_DEV
894 OSTDEV2_2=$fs3ost_DEV
898 echo "umount $MOUNT2 ..."
899 umount $MOUNT2 || true
900 echo "stopping fs2mds ..."
901 stop fs2mds -f || true
902 echo "stopping fs2ost ..."
903 stop fs2ost -f || true
907 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
909 if [ -z "$fs2ost_DEV" -o -z "$fs2mds_DEV" ]; then
910 is_blkdev $SINGLEMDS $MDSDEV && \
911 skip_env "mixed loopback and real device not working" && return
914 [ -n "$ost1_HOST" ] && fs2ost_HOST=$ost1_HOST
916 local fs2mdsdev=$(mdsdevname 1_2)
917 local fs2ostdev=$(ostdevname 1_2)
918 local fs2mdsvdev=$(mdsvdevname 1_2)
919 local fs2ostvdev=$(ostvdevname 1_2)
921 # test 8-char fsname as well
922 local FSNAME2=test1234
924 add fs2mds $(mkfs_opts mds1 ${fs2mdsdev} ) --nomgs --mgsnode=$MGSNID \
925 --fsname=${FSNAME2} --reformat $fs2mdsdev $fs2mdsvdev || exit 10
927 add fs2ost $(mkfs_opts ost1 ${fs2ostdev}) --fsname=${FSNAME2} \
928 --reformat $fs2ostdev $fs2ostvdev || exit 10
931 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS && trap cleanup_fs2 EXIT INT
932 start fs2ost $fs2ostdev $OST_MOUNT_OPTS
934 $MOUNT_CMD $MGSNID:/${FSNAME2} $MOUNT2 || return 1
936 check_mount || return 2
937 # files written on 1 should not show up on 2
938 cp /etc/passwd $DIR/$tfile
940 [ -e $MOUNT2/$tfile ] && error "File bleed" && return 7
943 cp /etc/passwd $MOUNT2/b || return 3
944 rm $MOUNT2/b || return 4
945 # 2 is actually mounted
946 grep $MOUNT2' ' /proc/mounts > /dev/null || return 5
948 facet_failover fs2mds
949 facet_failover fs2ost
952 # the MDS must remain up until last MDT
954 MDS=$(do_facet $SINGLEMDS "lctl get_param -n devices" | awk '($3 ~ "mdt" && $4 ~ "MDT") { print $4 }' | head -1)
955 [ -z "$MDS" ] && error "No MDT" && return 8
957 cleanup_nocli || return 6
959 run_test 24a "Multiple MDTs on a single node"
962 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
964 if [ -z "$fs2mds_DEV" ]; then
965 local dev=${SINGLEMDS}_dev
967 is_blkdev $SINGLEMDS $MDSDEV && \
968 skip_env "mixed loopback and real device not working" && return
971 local fs2mdsdev=$(mdsdevname 1_2)
972 local fs2mdsvdev=$(mdsvdevname 1_2)
974 add fs2mds $(mkfs_opts mds1 ${fs2mdsdev} ) --mgs --fsname=${FSNAME}2 \
975 --reformat $fs2mdsdev $fs2mdsvdev || exit 10
977 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS && return 2
980 run_test 24b "Multiple MGSs on a single node (should return err)"
984 check_mount || return 2
985 local MODULES=$($LCTL modules | awk '{ print $2 }')
986 rmmod $MODULES 2>/dev/null || true
989 run_test 25 "Verify modules are referenced"
993 # we need modules before mount for sysctl, so make sure...
994 do_facet $SINGLEMDS "lsmod | grep -q lustre || modprobe lustre"
995 #define OBD_FAIL_MDS_FS_SETUP 0x135
996 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000135"
997 start_mds && echo MDS started && return 1
998 lctl get_param -n devices
999 DEVS=$(lctl get_param -n devices | egrep -v MG | wc -l)
1000 [ $DEVS -gt 0 ] && return 2
1001 # start mds to drop writeconf setting
1002 start_mds || return 3
1003 stop_mds || return 4
1004 unload_modules_conf || return $?
1006 run_test 26 "MDT startup failure cleans LOV (should return errs)"
1009 [ "$(facet_fstype ost1)" = "zfs" ] &&
1010 skip "LU-2059: no local config for ZFS OSTs" && return
1012 start_ost || return 1
1013 start_mds || return 2
1014 echo "Requeue thread should have started: "
1015 ps -e | grep ll_cfg_requeue
1016 set_conf_param_and_check ost1 \
1017 "lctl get_param -n obdfilter.$FSNAME-OST0000.client_cache_seconds" \
1018 "$FSNAME-OST0000.ost.client_cache_seconds" || return 3
1021 run_test 27a "Reacquire MGS lock if OST started first"
1026 local device=$(do_facet $SINGLEMDS "lctl get_param -n devices" |
1027 awk '($3 ~ "mdt" && $4 ~ "MDT0000") { print $4 }')
1029 facet_failover $SINGLEMDS
1030 set_conf_param_and_check $SINGLEMDS \
1031 "lctl get_param -n mdt.$device.identity_acquire_expire" \
1032 "$device.mdt.identity_acquire_expire" || return 3
1033 set_conf_param_and_check client \
1034 "lctl get_param -n mdc.$device-mdc-*.max_rpcs_in_flight"\
1035 "$device.mdc.max_rpcs_in_flight" || return 4
1039 run_test 27b "Reacquire MGS lock after failover"
1043 TEST="lctl get_param -n llite.$FSNAME-*.max_read_ahead_whole_mb"
1044 PARAM="$FSNAME.llite.max_read_ahead_whole_mb"
1046 FINAL=$(($ORIG + 1))
1047 set_conf_param_and_check client "$TEST" "$PARAM" $FINAL || return 3
1048 FINAL=$(($FINAL + 1))
1049 set_conf_param_and_check client "$TEST" "$PARAM" $FINAL || return 4
1050 umount_client $MOUNT || return 200
1053 if [ $RESULT -ne $FINAL ]; then
1054 echo "New config not seen: wanted $FINAL got $RESULT"
1057 echo "New config success: got $RESULT"
1059 set_conf_param_and_check client "$TEST" "$PARAM" $ORIG || return 5
1062 run_test 28 "permanent parameter setting"
1064 test_28a() { # LU-4221
1065 [[ $(lustre_version_code ost1) -ge $(version_code 2.5.52) ]] ||
1066 { skip "Need OST version at least 2.5.52" && return 0; }
1067 [ "$(facet_fstype ost1)" = "zfs" ] &&
1068 skip "LU-4221: no such proc params for ZFS OSTs" && return
1075 local device="$FSNAME-OST0000"
1079 # In this test we will set three kinds of proc parameters with
1081 # 1. the ones moved from the OFD to the OSD, and only their
1082 # symlinks kept in obdfilter
1083 # 2. non-symlink ones in the OFD
1084 # 3. non-symlink ones in the OSD
1087 # prepare a symlink parameter in the OFD
1088 name="writethrough_cache_enable"
1089 param="$device.ost.$name"
1090 cmd="$LCTL get_param -n obdfilter.$device.$name"
1092 # conf_param the symlink parameter in the OFD
1093 old=$(do_facet ost1 $cmd)
1094 new=$(((old + 1) % 2))
1095 set_conf_param_and_check ost1 "$cmd" "$param" $new ||
1096 error "lctl conf_param $device.ost.$param=$new failed"
1098 # conf_param the target parameter in the OSD
1099 param="$device.osd.$name"
1100 cmd="$LCTL get_param -n osd-*.$device.$name"
1101 set_conf_param_and_check ost1 "$cmd" "$param" $old ||
1102 error "lctl conf_param $device.osd.$param=$old failed"
1105 # prepare a non-symlink parameter in the OFD
1106 name="client_cache_seconds"
1107 param="$device.ost.$name"
1108 cmd="$LCTL get_param -n obdfilter.$device.$name"
1110 # conf_param the parameter in the OFD
1111 old=$(do_facet ost1 $cmd)
1113 set_conf_param_and_check ost1 "$cmd" "$param" $new ||
1114 error "lctl conf_param $device.ost.$param=$new failed"
1115 set_conf_param_and_check ost1 "$cmd" "$param" $old ||
1116 error "lctl conf_param $device.ost.$param=$old failed"
1119 # prepare a non-symlink parameter in the OSD
1120 name="lma_self_repair"
1121 param="$device.osd.$name"
1122 cmd="$LCTL get_param -n osd-*.$device.$name"
1124 # conf_param the parameter in the OSD
1125 old=$(do_facet ost1 $cmd)
1126 new=$(((old + 1) % 2))
1127 set_conf_param_and_check ost1 "$cmd" "$param" $new ||
1128 error "lctl conf_param $device.osd.$param=$new failed"
1129 set_conf_param_and_check ost1 "$cmd" "$param" $old ||
1130 error "lctl conf_param $device.osd.$param=$old failed"
1134 run_test 28a "set symlink parameters permanently with conf_param"
1137 [ "$OSTCOUNT" -lt "2" ] && skip_env "$OSTCOUNT < 2, skipping" && return
1138 setup > /dev/null 2>&1
1142 local PARAM="$FSNAME-OST0001.osc.active"
1143 local PROC_ACT="osc.$FSNAME-OST0001-osc-[^M]*.active"
1144 local PROC_UUID="osc.$FSNAME-OST0001-osc-[^M]*.ost_server_uuid"
1146 ACTV=$(lctl get_param -n $PROC_ACT)
1148 set_conf_param_and_check client \
1149 "lctl get_param -n $PROC_ACT" "$PARAM" $DEAC || return 2
1150 # also check ost_server_uuid status
1151 RESULT=$(lctl get_param -n $PROC_UUID | grep DEACTIV)
1152 if [ -z "$RESULT" ]; then
1153 echo "Live client not deactivated: $(lctl get_param -n $PROC_UUID)"
1156 echo "Live client success: got $RESULT"
1160 for num in $(seq $MDSCOUNT); do
1161 local mdtosc=$(get_mdtosc_proc_path mds${num} $FSNAME-OST0001)
1162 local MPROC="osc.$mdtosc.active"
1167 RESULT=$(do_facet mds${num} " lctl get_param -n $MPROC")
1168 [ ${PIPESTATUS[0]} = 0 ] || error "Can't read $MPROC"
1169 if [ $RESULT -eq $DEAC ]; then
1170 echo -n "MDT deactivated also after"
1171 echo "$WAIT sec (got $RESULT)"
1175 if [ $WAIT -eq $MAX ]; then
1176 echo -n "MDT not deactivated: wanted $DEAC"
1180 echo "Waiting $(($MAX - $WAIT))secs for MDT deactivated"
1183 # test new client starts deactivated
1184 umount_client $MOUNT || return 200
1186 RESULT=$(lctl get_param -n $PROC_UUID | grep DEACTIV | grep NEW)
1187 if [ -z "$RESULT" ]; then
1188 echo "New client not deactivated from start: $(lctl get_param -n $PROC_UUID)"
1191 echo "New client success: got $RESULT"
1194 # make sure it reactivates
1195 set_conf_param_and_check client \
1196 "lctl get_param -n $PROC_ACT" "$PARAM" $ACTV || return 6
1198 umount_client $MOUNT
1201 #writeconf to remove all ost2 traces for subsequent tests
1202 writeconf_or_reformat
1204 run_test 29 "permanently remove an OST"
1209 echo Big config llog
1210 TEST="lctl get_param -n llite.$FSNAME-*.max_read_ahead_whole_mb"
1212 LIST=(1 2 3 4 5 4 3 2 1 2 3 4 5 4 3 2 1 2 3 4 5)
1213 for i in ${LIST[@]}; do
1214 set_conf_param_and_check client "$TEST" \
1215 "$FSNAME.llite.max_read_ahead_whole_mb" $i || return 3
1217 # make sure client restart still works
1218 umount_client $MOUNT
1219 mount_client $MOUNT || return 4
1220 [ "$($TEST)" -ne "$i" ] && error "Param didn't stick across restart $($TEST) != $i"
1223 echo Erase parameter setting
1224 do_facet mgs "$LCTL conf_param -d $FSNAME.llite.max_read_ahead_whole_mb" || return 6
1225 umount_client $MOUNT
1226 mount_client $MOUNT || return 6
1228 echo "deleted (default) value=$FINAL, orig=$ORIG"
1229 # assumes this parameter started at the default value
1230 [ "$FINAL" -eq "$ORIG" ] || fail "Deleted value=$FINAL, orig=$ORIG"
1234 run_test 30a "Big config llog and conf_param deletion"
1239 # Make a fake nid. Use the OST nid, and add 20 to the least significant
1240 # numerical part of it. Hopefully that's not already a failover address for
1242 OSTNID=$(do_facet ost1 "$LCTL get_param nis" | tail -1 | awk '{print $1}')
1243 ORIGVAL=$(echo $OSTNID | egrep -oi "[0-9]*@")
1244 NEWVAL=$((($(echo $ORIGVAL | egrep -oi "[0-9]*") + 20) % 256))
1245 NEW=$(echo $OSTNID | sed "s/$ORIGVAL/$NEWVAL@/")
1246 echo "Using fake nid $NEW"
1248 TEST="$LCTL get_param -n osc.$FSNAME-OST0000-osc-[^M]*.import | grep failover_nids | sed -n 's/.*\($NEW\).*/\1/p'"
1249 set_conf_param_and_check client "$TEST" \
1250 "$FSNAME-OST0000.failover.node" $NEW ||
1251 error "didn't add failover nid $NEW"
1252 NIDS=$($LCTL get_param -n osc.$FSNAME-OST0000-osc-[^M]*.import | grep failover_nids)
1254 NIDCOUNT=$(($(echo "$NIDS" | wc -w) - 1))
1255 echo "should have 2 failover nids: $NIDCOUNT"
1256 [ $NIDCOUNT -eq 2 ] || error "Failover nid not added"
1257 do_facet mgs "$LCTL conf_param -d $FSNAME-OST0000.failover.node" || error "conf_param delete failed"
1258 umount_client $MOUNT
1259 mount_client $MOUNT || return 3
1261 NIDS=$($LCTL get_param -n osc.$FSNAME-OST0000-osc-[^M]*.import | grep failover_nids)
1263 NIDCOUNT=$(($(echo "$NIDS" | wc -w) - 1))
1264 echo "only 1 final nid should remain: $NIDCOUNT"
1265 [ $NIDCOUNT -eq 1 ] || error "Failover nids not removed"
1269 run_test 30b "Remove failover nids"
1271 test_31() { # bug 10734
1272 # ipaddr must not exist
1273 $MOUNT_CMD 4.3.2.1@tcp:/lustre $MOUNT || true
1276 run_test 31 "Connect to non-existent node (shouldn't crash)"
1280 T32_BLIMIT=20480 # Kbytes
1284 # This is not really a test but a tool to create new disk
1285 # image tarballs for the upgrade tests.
1287 # Disk image tarballs should be created on single-node
1288 # clusters by running this test with default configurations
1289 # plus a few mandatory environment settings that are verified
1290 # at the beginning of the test.
1292 test_32newtarball() {
1296 local tmp=$TMP/t32_image_create
1298 if [ $FSNAME != t32fs -o $MDSCOUNT -ne 1 -o \
1299 \( -z "$MDSDEV" -a -z "$MDSDEV1" \) -o $OSTCOUNT -ne 1 -o \
1300 -z "$OSTDEV1" ]; then
1301 error "Needs FSNAME=t32fs MDSCOUNT=1 MDSDEV1=<nonexistent_file>" \
1302 "(or MDSDEV, in the case of b1_8) OSTCOUNT=1" \
1303 "OSTDEV1=<nonexistent_file>"
1307 echo "Found stale $tmp"
1312 tar cf - -C $src . | tar xf - -C $tmp/src
1313 dd if=/dev/zero of=$tmp/src/t32_qf_old bs=1M \
1314 count=$(($T32_BLIMIT / 1024 / 2))
1315 chown $T32_QID.$T32_QID $tmp/src/t32_qf_old
1317 # format ost with comma-separated NIDs to verify LU-4460
1318 local failnid="$(h2$NETTYPE 1.2.3.4),$(h2$NETTYPE 4.3.2.1)"
1319 MGSNID="$MGSNID,$MGSNID" OSTOPT="--failnode=$failnid" formatall
1323 [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.3.50) ] &&
1324 $LFS quotacheck -ug /mnt/$FSNAME
1325 $LFS setquota -u $T32_QID -b 0 -B $T32_BLIMIT -i 0 -I $T32_ILIMIT \
1328 tar cf - -C $tmp/src . | tar xf - -C /mnt/$FSNAME
1335 ls -Rni --time-style=+%s >$tmp/img/list
1336 find . ! -name .lustre -type f -exec sha1sum {} \; |
1337 sort -k 2 >$tmp/img/sha1sums
1339 $LCTL get_param -n version | head -n 1 |
1340 sed -e 's/^lustre: *//' >$tmp/img/commit
1342 [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.3.50) ] &&
1343 $LFS quotaon -ug /mnt/$FSNAME
1344 $LFS quota -u $T32_QID -v /mnt/$FSNAME
1345 $LFS quota -v -u $T32_QID /mnt/$FSNAME |
1346 awk 'BEGIN { num='1' } { if ($1 == "'/mnt/$FSNAME'") \
1347 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1348 | tr -d "*" > $tmp/img/bspace
1349 $LFS quota -v -u $T32_QID /mnt/$FSNAME |
1350 awk 'BEGIN { num='5' } { if ($1 == "'/mnt/$FSNAME'") \
1351 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1352 | tr -d "*" > $tmp/img/ispace
1357 find -type f -exec sha1sum {} \; | sort -k 2 >$tmp/sha1sums.src
1360 if ! diff -u $tmp/sha1sums.src $tmp/img/sha1sums; then
1361 echo "Data verification failed"
1364 uname -r >$tmp/img/kernel
1365 uname -m >$tmp/img/arch
1367 mv ${MDSDEV1:-$MDSDEV} $tmp/img
1368 mv $OSTDEV1 $tmp/img
1370 version=$(sed -e 's/\(^[0-9]\+\.[0-9]\+\)\(.*$\)/\1/' $tmp/img/commit |
1371 sed -e 's/\./_/g') # E.g., "1.8.7" -> "1_8"
1374 tar cjvf $dst/disk$version-$(facet_fstype $SINGLEMDS).tar.bz2 -S *
1379 #run_test 32newtarball "Create a new test_32 disk image tarball for this version"
1382 # The list of applicable tarballs is returned via the caller's
1383 # variable "tarballs".
1386 local node=$(facet_active_host $SINGLEMDS)
1387 local r="do_node $node"
1389 if [ "$CLIENTONLY" ]; then
1390 skip "Client-only testing"
1394 if ! $r which $TUNEFS; then
1395 skip_env "tunefs.lustre required on $node"
1399 local IMGTYPE=$(facet_fstype $SINGLEMDS)
1401 tarballs=$($r find $RLUSTRE/tests -maxdepth 1 -name \'disk*-$IMGTYPE.tar.bz2\')
1403 if [ -z "$tarballs" ]; then
1404 skip "No applicable tarballs found"
1409 t32_test_cleanup() {
1411 local fstype=$(facet_fstype $SINGLEMDS)
1414 if $shall_cleanup_lustre; then
1415 umount $tmp/mnt/lustre || rc=$?
1417 if $shall_cleanup_mdt; then
1418 $r umount -d $tmp/mnt/mdt || rc=$?
1420 if $shall_cleanup_mdt1; then
1421 $r umount -d $tmp/mnt/mdt1 || rc=$?
1423 if $shall_cleanup_ost; then
1424 $r umount -d $tmp/mnt/ost || rc=$?
1429 if [ $fstype == "zfs" ]; then
1430 $r $ZPOOL destroy t32fs-mdt1 || rc=$?
1431 $r $ZPOOL destroy t32fs-ost1 || rc=$?
1436 t32_bits_per_long() {
1438 # Yes, this is not meant to be perfect.
1448 t32_reload_modules() {
1450 local all_removed=false
1453 while ((i < 20)); do
1454 echo "Unloading modules on $node: Attempt $i"
1455 do_rpc_nodes $node $LUSTRE_RMMOD $(facet_fstype $SINGLEMDS) &&
1457 do_rpc_nodes $node check_mem_leak || return 1
1458 if $all_removed; then
1459 do_rpc_nodes $node load_modules
1465 echo "Unloading modules on $node: Given up"
1469 t32_wait_til_devices_gone() {
1475 echo wait for devices to go
1476 while ((i < 20)); do
1477 devices=$(do_rpc_nodes $node $LCTL device_list | wc -l)
1478 loops=$(do_rpc_nodes $node losetup -a | grep -c t32)
1479 ((devices == 0 && loops == 0)) && return 0
1483 echo "waiting for dev on $node: dev $devices loop $loops given up"
1484 do_rpc_nodes $node "losetup -a"
1485 do_rpc_nodes $node "$LCTL devices_list"
1489 t32_verify_quota() {
1493 local fstype=$(facet_fstype $SINGLEMDS)
1497 $LFS quota -u $T32_QID -v $mnt
1499 qval=$($LFS quota -v -u $T32_QID $mnt |
1500 awk 'BEGIN { num='1' } { if ($1 == "'$mnt'") \
1501 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1503 [ $qval -eq $img_bspace ] || {
1504 echo "bspace, act:$qval, exp:$img_bspace"
1508 qval=$($LFS quota -v -u $T32_QID $mnt |
1509 awk 'BEGIN { num='5' } { if ($1 == "'$mnt'") \
1510 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1512 [ $qval -eq $img_ispace ] || {
1513 echo "ispace, act:$qval, exp:$img_ispace"
1517 qval=$($LFS quota -v -u $T32_QID $mnt |
1518 awk 'BEGIN { num='3' } { if ($1 == "'$mnt'") \
1519 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1521 [ $qval -eq $T32_BLIMIT ] || {
1522 echo "blimit, act:$qval, exp:$T32_BLIMIT"
1526 qval=$($LFS quota -v -u $T32_QID $mnt |
1527 awk 'BEGIN { num='7' } { if ($1 == "'$mnt'") \
1528 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1530 [ $qval -eq $T32_ILIMIT ] || {
1531 echo "ilimit, act:$qval, exp:$T32_ILIMIT"
1535 do_node $node $LCTL conf_param $fsname.quota.mdt=ug
1536 cmd="$LCTL get_param -n osd-$fstype.$fsname-MDT0000"
1537 cmd=$cmd.quota_slave.enabled
1538 wait_update $node "$cmd" "ug" || {
1539 echo "Enable mdt quota failed"
1543 do_node $node $LCTL conf_param $fsname.quota.ost=ug
1544 cmd="$LCTL get_param -n osd-$fstype.$fsname-OST0000"
1545 cmd=$cmd.quota_slave.enabled
1546 wait_update $node "$cmd" "ug" || {
1547 echo "Enable ost quota failed"
1552 runas -u $T32_QID -g $T32_QID dd if=/dev/zero of=$mnt/t32_qf_new \
1553 bs=1M count=$(($T32_BLIMIT / 1024)) oflag=sync && {
1554 echo "Write succeed, but expect -EDQUOT"
1557 rm -f $mnt/t32_qf_new
1559 runas -u $T32_QID -g $T32_QID createmany -m $mnt/t32_qf_ \
1561 echo "Create succeed, but expect -EDQUOT"
1564 unlinkmany $mnt/t32_qf_ $T32_ILIMIT
1572 local dne_upgrade=${dne_upgrade:-"no"}
1573 local ff_convert=${ff_convert:-"no"}
1574 local shall_cleanup_mdt=false
1575 local shall_cleanup_mdt1=false
1576 local shall_cleanup_ost=false
1577 local shall_cleanup_lustre=false
1578 local node=$(facet_active_host $SINGLEMDS)
1579 local r="do_node $node"
1580 local node2=$(facet_active_host mds2)
1588 local nid=$($r $LCTL list_nids | head -1)
1594 local fstype=$(facet_fstype $SINGLEMDS)
1595 local mdt_dev=$tmp/mdt
1596 local ost_dev=$tmp/ost
1598 trap 'trap - RETURN; t32_test_cleanup' RETURN
1600 mkdir -p $tmp/mnt/lustre
1601 $r mkdir -p $tmp/mnt/{mdt,ost}
1602 $r tar xjvf $tarball -S -C $tmp || {
1603 error_noexit "Unpacking the disk image tarball"
1606 img_commit=$($r cat $tmp/commit)
1607 img_kernel=$($r cat $tmp/kernel)
1608 img_arch=$($r cat $tmp/arch)
1609 img_bspace=$($r cat $tmp/bspace)
1610 img_ispace=$($r cat $tmp/ispace)
1611 echo "Upgrading from $(basename $tarball), created with:"
1612 echo " Commit: $img_commit"
1613 echo " Kernel: $img_kernel"
1614 echo " Arch: $img_arch"
1616 local version=$(version_code $img_commit)
1617 [[ $version -ge $(version_code 2.5.0) ]] && ff_convert="no"
1619 if [ $fstype == "zfs" ]; then
1621 $r $ZPOOL import -f -d $tmp t32fs-mdt1
1622 $r $ZPOOL import -f -d $tmp t32fs-ost1
1623 mdt_dev=t32fs-mdt1/mdt1
1624 ost_dev=t32fs-ost1/ost1
1625 wait_update_facet $SINGLEMDS "$ZPOOL list |
1626 awk '/^t32fs-mdt1/ { print \\\$1 }'" "t32fs-mdt1" || {
1627 error_noexit "import zfs pool failed"
1632 $r $LCTL set_param debug="$PTLDEBUG"
1634 $r $TUNEFS --dryrun $mdt_dev || {
1636 error_noexit "tunefs.lustre before mounting the MDT"
1639 if [ "$writeconf" ]; then
1641 if [ $fstype == "ldiskfs" ]; then
1643 $r $TUNEFS --quota $mdt_dev || {
1645 error_noexit "Enable mdt quota feature"
1650 if [ -n "$($LCTL list_nids | grep -v '\(tcp\|lo\)[[:digit:]]*$')" ]; then
1651 [[ $(lustre_version_code mgs) -ge $(version_code 2.3.59) ]] ||
1652 { skip "LU-2200: Cannot run over Inifiniband w/o lctl replace_nids "
1653 "(Need MGS version at least 2.3.59)"; return 0; }
1655 local osthost=$(facet_active_host ost1)
1656 local ostnid=$(do_node $osthost $LCTL list_nids | head -1)
1659 if [ $fstype == "ldiskfs" ]; then
1662 $r $MOUNT_CMD -o $mopts $mdt_dev $tmp/mnt/mdt
1663 $r lctl replace_nids $fsname-OST0000 $ostnid
1664 $r lctl replace_nids $fsname-MDT0000 $nid
1665 $r umount -d $tmp/mnt/mdt
1668 mopts=exclude=$fsname-OST0000
1669 if [ $fstype == "ldiskfs" ]; then
1674 t32_wait_til_devices_gone $node
1676 $r $MOUNT_CMD -o $mopts $mdt_dev $tmp/mnt/mdt || {
1678 error_noexit "Mounting the MDT"
1681 shall_cleanup_mdt=true
1683 if [ "$dne_upgrade" != "no" ]; then
1684 local fs2mdsdev=$(mdsdevname 1_2)
1685 local fs2mdsvdev=$(mdsvdevname 1_2)
1687 echo "mkfs new MDT on ${fs2mdsdev}...."
1688 if [ $(facet_fstype mds1) == ldiskfs ]; then
1689 mkfsoptions="--mkfsoptions=\\\"-J size=8\\\""
1692 add fs2mds $(mkfs_opts mds2 $fs2mdsdev $fsname) --reformat \
1693 $mkfsoptions $fs2mdsdev $fs2mdsvdev > /dev/null || {
1694 error_noexit "Mkfs new MDT failed"
1698 $r $TUNEFS --dryrun $fs2mdsdev || {
1699 error_noexit "tunefs.lustre before mounting the MDT"
1703 echo "mount new MDT....$fs2mdsdev"
1704 $r mkdir -p $tmp/mnt/mdt1
1705 $r $MOUNT_CMD -o $mopts $fs2mdsdev $tmp/mnt/mdt1 || {
1706 error_noexit "mount mdt1 failed"
1709 shall_cleanup_mdt1=true
1712 uuid=$($r $LCTL get_param -n mdt.$fsname-MDT0000.uuid) || {
1713 error_noexit "Getting MDT UUID"
1716 if [ "$uuid" != $fsname-MDT0000_UUID ]; then
1717 error_noexit "Unexpected MDT UUID: \"$uuid\""
1721 $r $TUNEFS --dryrun $ost_dev || {
1722 error_noexit "tunefs.lustre before mounting the OST"
1725 if [ "$writeconf" ]; then
1726 mopts=mgsnode=$nid,$writeconf
1727 if [ $fstype == "ldiskfs" ]; then
1729 $r $TUNEFS --quota $ost_dev || {
1731 error_noexit "Enable ost quota feature"
1737 if [ $fstype == "ldiskfs" ]; then
1741 $r $MOUNT_CMD -o $mopts $ost_dev $tmp/mnt/ost || {
1742 error_noexit "Mounting the OST"
1745 shall_cleanup_ost=true
1747 uuid=$($r $LCTL get_param -n obdfilter.$fsname-OST0000.uuid) || {
1748 error_noexit "Getting OST UUID"
1751 if [ "$uuid" != $fsname-OST0000_UUID ]; then
1752 error_noexit "Unexpected OST UUID: \"$uuid\""
1756 $r $LCTL conf_param $fsname-OST0000.osc.max_dirty_mb=15 || {
1757 error_noexit "Setting \"max_dirty_mb\""
1760 $r $LCTL conf_param $fsname-OST0000.failover.node=$nid || {
1761 error_noexit "Setting OST \"failover.node\""
1764 $r $LCTL conf_param $fsname-MDT0000.mdc.max_rpcs_in_flight=9 || {
1765 error_noexit "Setting \"max_rpcs_in_flight\""
1768 $r $LCTL conf_param $fsname-MDT0000.failover.node=$nid || {
1769 error_noexit "Setting MDT \"failover.node\""
1772 $r $LCTL pool_new $fsname.interop || {
1773 error_noexit "Setting \"interop\""
1776 $r $LCTL conf_param $fsname-MDT0000.lov.stripesize=4M || {
1777 error_noexit "Setting \"lov.stripesize\""
1781 if [ "$ff_convert" != "no" -a $(facet_fstype ost1) == "ldiskfs" ]; then
1782 $r $LCTL lfsck_start -M $fsname-OST0000 || {
1783 error_noexit "Start OI scrub on OST0"
1787 # The oi_scrub should be on ost1, but for test_32(),
1788 # all on the SINGLEMDS.
1789 wait_update_facet $SINGLEMDS "$LCTL get_param -n \
1790 osd-ldiskfs.$fsname-OST0000.oi_scrub |
1791 awk '/^status/ { print \\\$2 }'" "completed" 30 || {
1792 error_noexit "Failed to get the expected 'completed'"
1796 local UPDATED=$($r $LCTL get_param -n \
1797 osd-ldiskfs.$fsname-OST0000.oi_scrub |
1798 awk '/^updated/ { print $2 }')
1799 [ $UPDATED -ge 1 ] || {
1800 error_noexit "Only $UPDATED objects have been converted"
1805 if [ "$dne_upgrade" != "no" ]; then
1806 $r $LCTL conf_param \
1807 $fsname-MDT0001.mdc.max_rpcs_in_flight=9 || {
1808 error_noexit "Setting MDT1 \"max_rpcs_in_flight\""
1811 $r $LCTL conf_param $fsname-MDT0001.failover.node=$nid || {
1812 error_noexit "Setting MDT1 \"failover.node\""
1815 $r $LCTL conf_param $fsname-MDT0001.lov.stripesize=4M || {
1816 error_noexit "Setting MDT1 \"lov.stripesize\""
1822 if [ "$writeconf" ]; then
1823 $MOUNT_CMD $nid:/$fsname $tmp/mnt/lustre || {
1824 error_noexit "Mounting the client"
1827 shall_cleanup_lustre=true
1828 $LCTL set_param debug="$PTLDEBUG"
1830 t32_verify_quota $node $fsname $tmp/mnt/lustre || {
1831 error_noexit "verify quota failed"
1835 if [ "$dne_upgrade" != "no" ]; then
1836 $LFS mkdir -i 1 $tmp/mnt/lustre/remote_dir || {
1837 error_noexit "set remote dir failed"
1841 pushd $tmp/mnt/lustre
1842 tar -cf - . --exclude=./remote_dir |
1843 tar -xvf - -C remote_dir 1>/dev/null || {
1844 error_noexit "cp to remote dir failed"
1850 dd if=/dev/zero of=$tmp/mnt/lustre/tmp_file bs=10k count=10 || {
1851 error_noexit "dd failed"
1854 rm -rf $tmp/mnt/lustre/tmp_file || {
1855 error_noexit "rm failed"
1859 if $r test -f $tmp/sha1sums; then
1860 # LU-2393 - do both sorts on same node to ensure locale
1862 $r cat $tmp/sha1sums | sort -k 2 >$tmp/sha1sums.orig
1863 if [ "$dne_upgrade" != "no" ]; then
1864 pushd $tmp/mnt/lustre/remote_dir
1866 pushd $tmp/mnt/lustre
1869 find ! -name .lustre -type f -exec sha1sum {} \; |
1870 sort -k 2 >$tmp/sha1sums || {
1871 error_noexit "sha1sum"
1875 if ! diff -ub $tmp/sha1sums.orig $tmp/sha1sums; then
1876 error_noexit "sha1sum verification failed"
1880 echo "sha1sum verification skipped"
1883 if [ "$dne_upgrade" != "no" ]; then
1884 rm -rf $tmp/mnt/lustre/remote_dir || {
1885 error_noexit "remove remote dir failed"
1890 if $r test -f $tmp/list; then
1892 # There is not a Test Framework API to copy files to or
1893 # from a remote node.
1895 # LU-2393 - do both sorts on same node to ensure locale
1897 $r cat $tmp/list | sort -k 6 >$tmp/list.orig
1898 pushd $tmp/mnt/lustre
1899 ls -Rni --time-style=+%s | sort -k 6 >$tmp/list || {
1905 # 32-bit and 64-bit clients use different algorithms to
1906 # convert FIDs into inode numbers. Hence, remove the inode
1907 # numbers from the lists, if the original list was created
1908 # on an architecture with different number of bits per
1911 if [ $(t32_bits_per_long $(uname -m)) != \
1912 $(t32_bits_per_long $img_arch) ]; then
1913 echo "Different number of bits per \"long\" from the disk image"
1914 for list in list.orig list; do
1915 sed -i -e 's/^[0-9]\+[ \t]\+//' $tmp/$list
1918 if ! diff -ub $tmp/list.orig $tmp/list; then
1919 error_noexit "list verification failed"
1923 echo "list verification skipped"
1927 # When adding new data verification tests, please check for
1928 # the presence of the required reference files first, like
1929 # the "sha1sums" and "list" tests above, to avoid the need to
1930 # regenerate every image for each test addition.
1933 nrpcs_orig=$($LCTL get_param \
1934 -n mdc.*MDT0000*.max_rpcs_in_flight) || {
1935 error_noexit "Getting \"max_rpcs_in_flight\""
1938 nrpcs=$((nrpcs_orig + 5))
1939 $r $LCTL conf_param $fsname-MDT0000.mdc.max_rpcs_in_flight=$nrpcs || {
1940 error_noexit "Changing \"max_rpcs_in_flight\""
1943 wait_update $HOSTNAME "$LCTL get_param \
1944 -n mdc.*MDT0000*.max_rpcs_in_flight" $nrpcs || {
1945 error_noexit "Verifying \"max_rpcs_in_flight\""
1949 umount $tmp/mnt/lustre || {
1950 error_noexit "Unmounting the client"
1953 shall_cleanup_lustre=false
1955 if [ "$dne_upgrade" != "no" ]; then
1956 $r umount -d $tmp/mnt/mdt1 || {
1957 error_noexit "Unmounting the MDT2"
1960 shall_cleanup_mdt1=false
1963 $r umount -d $tmp/mnt/mdt || {
1964 error_noexit "Unmounting the MDT"
1967 shall_cleanup_mdt=false
1969 $r umount -d $tmp/mnt/ost || {
1970 error_noexit "Unmounting the OST"
1973 shall_cleanup_ost=false
1975 t32_reload_modules $node || {
1976 error_noexit "Reloading modules"
1980 # mount a second time to make sure we didnt leave upgrade flag on
1981 $r $TUNEFS --dryrun $mdt_dev || {
1983 error_noexit "tunefs.lustre before remounting the MDT"
1987 mopts=exclude=$fsname-OST0000
1988 if [ $fstype == "ldiskfs" ]; then
1991 $r $MOUNT_CMD -o $mopts $mdt_dev $tmp/mnt/mdt || {
1992 error_noexit "Remounting the MDT"
1995 shall_cleanup_mdt=true
2005 for tarball in $tarballs; do
2006 t32_test $tarball || let "rc += $?"
2010 run_test 32a "Upgrade (not live)"
2018 for tarball in $tarballs; do
2019 t32_test $tarball writeconf || let "rc += $?"
2023 run_test 32b "Upgrade with writeconf"
2030 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
2032 for tarball in $tarballs; do
2033 dne_upgrade=yes t32_test $tarball writeconf || rc=$?
2037 run_test 32c "dne upgrade test"
2045 for tarball in $tarballs; do
2046 ff_convert=yes t32_test $tarball || rc=$?
2050 run_test 32d "convert ff test"
2052 test_33a() { # bug 12333, was test_33
2054 local FSNAME2=test-123
2055 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
2058 [ -n "$ost1_HOST" ] && fs2ost_HOST=$ost1_HOST
2060 if [ -z "$fs2ost_DEV" -o -z "$fs2mds_DEV" ]; then
2061 local dev=${SINGLEMDS}_dev
2062 local MDSDEV=${!dev}
2063 is_blkdev $SINGLEMDS $MDSDEV && \
2064 skip_env "mixed loopback and real device not working" && return
2067 local fs2mdsdev=$(mdsdevname 1_2)
2068 local fs2ostdev=$(ostdevname 1_2)
2069 local fs2mdsvdev=$(mdsvdevname 1_2)
2070 local fs2ostvdev=$(ostvdevname 1_2)
2072 if [ $(facet_fstype mds1) == ldiskfs ]; then
2073 mkfsoptions="--mkfsoptions=\\\"-J size=8\\\"" # See bug 17931.
2076 add fs2mds $(mkfs_opts mds1 ${fs2mdsdev}) --mgs --fsname=${FSNAME2} \
2077 --reformat $mkfsoptions $fs2mdsdev $fs2mdsvdev || exit 10
2078 add fs2ost $(mkfs_opts ost1 ${fs2ostdev}) --mgsnode=$MGSNID \
2079 --fsname=${FSNAME2} --index=8191 --reformat $fs2ostdev \
2080 $fs2ostvdev || exit 10
2082 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS && trap cleanup_fs2 EXIT INT
2083 start fs2ost $fs2ostdev $OST_MOUNT_OPTS
2084 do_facet $SINGLEMDS "$LCTL conf_param $FSNAME2.sys.timeout=200" || rc=1
2086 $MOUNT_CMD $MGSNID:/${FSNAME2} $MOUNT2 || rc=2
2089 cp /etc/hosts $MOUNT2/ || rc=3
2090 $LFS getstripe $MOUNT2/hosts
2095 cleanup_nocli || rc=6
2098 run_test 33a "Mount ost with a large index number"
2100 test_33b() { # was test_34
2103 do_facet client dd if=/dev/zero of=$MOUNT/24 bs=1024k count=1
2104 # Drop lock cancelation reply during umount
2105 #define OBD_FAIL_LDLM_CANCEL_NET 0x304
2106 do_facet client lctl set_param fail_loc=0x80000304
2107 #lctl set_param debug=-1
2108 umount_client $MOUNT
2111 run_test 33b "Drop cancel during umount"
2115 do_facet client "sh runmultiop_bg_pause $DIR/file O_c"
2116 manual_umount_client
2118 do_facet client killall -USR1 multiop
2119 if [ $rc -eq 0 ]; then
2120 error "umount not fail!"
2125 run_test 34a "umount with opened file should be fail"
2130 touch $DIR/$tfile || return 1
2131 stop_mds --force || return 2
2133 manual_umount_client --force
2135 if [ $rc -ne 0 ]; then
2136 error "mtab after failed umount - rc $rc"
2142 run_test 34b "force umount with failed mds should be normal"
2146 touch $DIR/$tfile || return 1
2147 stop_ost --force || return 2
2149 manual_umount_client --force
2151 if [ $rc -ne 0 ]; then
2152 error "mtab after failed umount - rc $rc"
2158 run_test 34c "force umount with failed ost should be normal"
2160 test_35a() { # bug 12459
2163 DBG_SAVE="`lctl get_param -n debug`"
2164 lctl set_param debug="ha"
2166 log "Set up a fake failnode for the MDS"
2168 local device=$(do_facet $SINGLEMDS "lctl get_param -n devices" |
2169 awk '($3 ~ "mdt" && $4 ~ "MDT") { print $4 }' | head -1)
2170 do_facet mgs "$LCTL conf_param \
2171 ${device}.failover.node=$(h2$NETTYPE $FAKENID)" || return 4
2173 log "Wait for RECONNECT_INTERVAL seconds (10s)"
2176 MSG="conf-sanity.sh test_35a `date +%F%kh%Mm%Ss`"
2179 log "Stopping the MDT: $device"
2180 stop_mdt 1 || return 5
2182 df $MOUNT > /dev/null 2>&1 &
2184 log "Restarting the MDT: $device"
2185 start_mdt 1 || return 6
2186 log "Wait for df ($DFPID) ... "
2189 lctl set_param debug="$DBG_SAVE"
2191 # retrieve from the log the first server that the client tried to
2192 # contact after the connection loss
2193 $LCTL dk $TMP/lustre-log-$TESTNAME.log
2194 NEXTCONN=`awk "/${MSG}/ {start = 1;}
2195 /import_select_connection.*$device-mdc.* using connection/ {
2197 if (\\\$NF ~ /$FAKENID/)
2203 }" $TMP/lustre-log-$TESTNAME.log`
2204 [ "$NEXTCONN" != "0" ] && log "The client didn't try to reconnect to the last active server (tried ${NEXTCONN} instead)" && return 7
2206 # remove nid settings
2207 writeconf_or_reformat
2209 run_test 35a "Reconnect to the last active server first"
2211 test_35b() { # bug 18674
2212 remote_mds || { skip "local MDS" && return 0; }
2216 $LCTL set_param debug="ha"
2218 MSG="conf-sanity.sh test_35b `date +%F%kh%Mm%Ss`"
2221 log "Set up a fake failnode for the MDS"
2223 local device=$(do_facet $SINGLEMDS "$LCTL get_param -n devices" |
2224 awk '($3 ~ "mdt" && $4 ~ "MDT") { print $4 }' | head -1)
2225 do_facet mgs "$LCTL conf_param \
2226 ${device}.failover.node=$(h2$NETTYPE $FAKENID)" || return 1
2228 local at_max_saved=0
2229 # adaptive timeouts may prevent seeing the issue
2230 if at_is_enabled; then
2231 at_max_saved=$(at_max_get mds)
2232 at_max_set 0 mds client
2235 mkdir -p $MOUNT/$tdir
2237 log "Injecting EBUSY on MDS"
2238 # Setting OBD_FAIL_MDS_RESEND=0x136
2239 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0x80000136" || return 2
2241 $LCTL set_param mdc.${FSNAME}*.stats=clear
2243 log "Creating a test file and stat it"
2244 touch $MOUNT/$tdir/$tfile
2245 stat $MOUNT/$tdir/$tfile
2247 log "Stop injecting EBUSY on MDS"
2248 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0" || return 3
2249 rm -f $MOUNT/$tdir/$tfile
2252 # restore adaptive timeout
2253 [ $at_max_saved -ne 0 ] && at_max_set $at_max_saved mds client
2255 $LCTL dk $TMP/lustre-log-$TESTNAME.log
2257 CONNCNT=`$LCTL get_param mdc.${FSNAME}*.stats | awk '/mds_connect/{print $2}'`
2259 # retrieve from the log if the client has ever tried to
2260 # contact the fake server after the loss of connection
2261 FAILCONN=`awk "BEGIN {ret = 0;}
2262 /import_select_connection.*${FSNAME}-MDT0000-mdc.* using connection/ {
2264 if (\\\$NF ~ /$FAKENID/) {
2269 END {print ret}" $TMP/lustre-log-$TESTNAME.log`
2271 [ "$FAILCONN" == "0" ] && \
2272 log "ERROR: The client reconnection has not been triggered" && \
2274 [ "$FAILCONN" == "2" ] && \
2275 log "ERROR: The client tried to reconnect to the failover server while the primary was busy" && \
2279 # When OBD_FAIL_MDS_RESEND is hit, we sleep for 2 * obd_timeout
2280 # Reconnects are supposed to be rate limited to one every 5s
2281 [ $CONNCNT -gt $((2 * $TIMEOUT / 5 + 1)) ] && \
2282 log "ERROR: Too many reconnects $CONNCNT" && \
2286 # remove nid settings
2287 writeconf_or_reformat
2289 run_test 35b "Continue reconnection retries, if the active server is busy"
2292 [ $OSTCOUNT -lt 2 ] && skip_env "skipping test for single OST" && return
2294 [ "$ost_HOST" = "`hostname`" -o "$ost1_HOST" = "`hostname`" ] || \
2295 { skip "remote OST" && return 0; }
2298 local FSNAME2=test1234
2299 local fs3ost_HOST=$ost_HOST
2300 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
2302 [ -n "$ost1_HOST" ] && fs2ost_HOST=$ost1_HOST && fs3ost_HOST=$ost1_HOST
2304 if [ -z "$fs2ost_DEV" -o -z "$fs2mds_DEV" -o -z "$fs3ost_DEV" ]; then
2305 is_blkdev $SINGLEMDS $MDSDEV && \
2306 skip_env "mixed loopback and real device not working" && return
2309 local fs2mdsdev=$(mdsdevname 1_2)
2310 local fs2ostdev=$(ostdevname 1_2)
2311 local fs3ostdev=$(ostdevname 2_2)
2312 local fs2mdsvdev=$(mdsvdevname 1_2)
2313 local fs2ostvdev=$(ostvdevname 1_2)
2314 local fs3ostvdev=$(ostvdevname 2_2)
2316 add fs2mds $(mkfs_opts mds1 ${fs2mdsdev}) --mgs --fsname=${FSNAME2} \
2317 --reformat $fs2mdsdev $fs2mdsvdev || exit 10
2318 # XXX after we support non 4K disk blocksize in ldiskfs, specify a
2319 # different one than the default value here.
2320 add fs2ost $(mkfs_opts ost1 ${fs2ostdev}) --mgsnode=$MGSNID \
2321 --fsname=${FSNAME2} --reformat $fs2ostdev $fs2ostvdev || exit 10
2322 add fs3ost $(mkfs_opts ost2 ${fs3ostdev}) --mgsnode=$MGSNID \
2323 --fsname=${FSNAME2} --reformat $fs3ostdev $fs3ostvdev || exit 10
2325 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS
2326 start fs2ost $fs2ostdev $OST_MOUNT_OPTS
2327 start fs3ost $fs3ostdev $OST_MOUNT_OPTS
2329 $MOUNT_CMD $MGSNID:/${FSNAME2} $MOUNT2 || return 1
2331 sleep 5 # until 11778 fixed
2333 dd if=/dev/zero of=$MOUNT2/$tfile bs=1M count=7 || return 2
2335 BKTOTAL=`lctl get_param -n obdfilter.*.kbytestotal | awk 'BEGIN{total=0}; {total+=$1}; END{print total}'`
2336 BKFREE=`lctl get_param -n obdfilter.*.kbytesfree | awk 'BEGIN{free=0}; {free+=$1}; END{print free}'`
2337 BKAVAIL=`lctl get_param -n obdfilter.*.kbytesavail | awk 'BEGIN{avail=0}; {avail+=$1}; END{print avail}'`
2338 STRING=`df -P $MOUNT2 | tail -n 1 | awk '{print $2","$3","$4}'`
2339 DFTOTAL=`echo $STRING | cut -d, -f1`
2340 DFUSED=`echo $STRING | cut -d, -f2`
2341 DFAVAIL=`echo $STRING | cut -d, -f3`
2342 DFFREE=$(($DFTOTAL - $DFUSED))
2344 ALLOWANCE=$((64 * $OSTCOUNT))
2346 if [ $DFTOTAL -lt $(($BKTOTAL - $ALLOWANCE)) ] ||
2347 [ $DFTOTAL -gt $(($BKTOTAL + $ALLOWANCE)) ] ; then
2348 echo "**** FAIL: df total($DFTOTAL) mismatch OST total($BKTOTAL)"
2351 if [ $DFFREE -lt $(($BKFREE - $ALLOWANCE)) ] ||
2352 [ $DFFREE -gt $(($BKFREE + $ALLOWANCE)) ] ; then
2353 echo "**** FAIL: df free($DFFREE) mismatch OST free($BKFREE)"
2356 if [ $DFAVAIL -lt $(($BKAVAIL - $ALLOWANCE)) ] ||
2357 [ $DFAVAIL -gt $(($BKAVAIL + $ALLOWANCE)) ] ; then
2358 echo "**** FAIL: df avail($DFAVAIL) mismatch OST avail($BKAVAIL)"
2363 stop fs3ost -f || return 200
2364 stop fs2ost -f || return 201
2365 stop fs2mds -f || return 202
2366 unload_modules_conf || return 203
2369 run_test 36 "df report consistency on OSTs with different block size"
2372 local mntpt=$(facet_mntpt $SINGLEMDS)
2373 local mdsdev=$(mdsdevname ${SINGLEMDS//mds/})
2374 local mdsdev_sym="$TMP/sym_mdt.img"
2375 local opts=$MDS_MOUNT_OPTS
2378 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
2379 skip "Currently only applicable to ldiskfs-based MDTs"
2383 echo "MDS : $mdsdev"
2384 echo "SYMLINK : $mdsdev_sym"
2385 do_facet $SINGLEMDS rm -f $mdsdev_sym
2387 do_facet $SINGLEMDS ln -s $mdsdev $mdsdev_sym
2389 echo "mount symlink device - $mdsdev_sym"
2391 if ! do_facet $SINGLEMDS test -b $mdsdev; then
2392 opts=$(csa_add "$opts" -o loop)
2394 mount_op=$(do_facet $SINGLEMDS mount -v -t lustre $opts \
2395 $mdsdev_sym $mntpt 2>&1)
2398 echo mount_op=$mount_op
2400 do_facet $SINGLEMDS "umount -d $mntpt && rm -f $mdsdev_sym"
2402 if $(echo $mount_op | grep -q "unable to set tunable"); then
2403 error "set tunables failed for symlink device"
2406 [ $rc -eq 0 ] || error "mount symlink $mdsdev_sym failed! rc=$rc"
2410 run_test 37 "verify set tunables works for symlink device"
2412 test_38() { # bug 14222
2413 local fstype=$(facet_fstype $SINGLEMDS)
2414 local mntpt=$(facet_mntpt $SINGLEMDS)
2419 local SRC="/etc /bin"
2420 local FILES=$(find $SRC -type f -mtime +1 | head -n $COUNT)
2421 log "copying $(echo $FILES | wc -w) files to $DIR/$tdir"
2423 tar cf - $FILES | tar xf - -C $DIR/$tdir ||
2424 error "copying $SRC to $DIR/$tdir"
2426 umount_client $MOUNT
2427 do_facet $SINGLEMDS "$LCTL get_param osp.*.prealloc_next_id"
2429 log "delete lov_objid file on MDS"
2431 mount_fstype $SINGLEMDS || error "mount MDS failed (1)"
2433 do_facet $SINGLEMDS "od -Ax -td8 $mntpt/lov_objid; rm $mntpt/lov_objid"
2435 unmount_fstype $SINGLEMDS || error "umount failed (1)"
2437 # check create in mds_lov_connect
2441 [ $V ] && log "verifying $DIR/$tdir/$f"
2442 diff -q $f $DIR/$tdir/$f || ERROR=y
2444 do_facet $SINGLEMDS "$LCTL get_param osp.*.prealloc_next_id"
2445 if [ "$ERROR" = "y" ]; then
2446 # check it's updates in sync
2447 umount_client $MOUNT
2449 mount_fstype $SIGNLEMDS
2450 do_facet $SINGLEMDS "od -Ax -td8 $mntpt/lov_objid"
2451 unmount_fstype $SINGLEMDS
2452 error "old and new files are different after connect" || true
2454 touch $DIR/$tdir/f2 || error "f2 file create failed"
2456 # check it's updates in sync
2457 umount_client $MOUNT
2460 mount_fstype $SINGLEMDS || error "mount MDS failed (3)"
2462 do_facet $SINGLEMDS "od -Ax -td8 $mntpt/lov_objid"
2463 do_facet $SINGLEMDS dd if=/dev/zero of=$mntpt/lov_objid.clear count=8
2465 unmount_fstype $SINGLEMDS || error "umount failed (3)"
2470 [ $V ] && log "verifying $DIR/$tdir/$f"
2471 diff -q $f $DIR/$tdir/$f || ERROR=y
2473 touch $DIR/$tdir/f3 || error "f3 file create failed"
2474 do_facet $SINGLEMDS "$LCTL get_param osp.*.prealloc_next_id"
2475 umount_client $MOUNT
2477 mount_fstype $SINGLEMDS || error "mount MDS failed (4)"
2478 do_facet $SINGLEMDS "od -Ax -td8 $mntpt/lov_objid"
2479 unmount_fstype $SINGLEMDS || error "umount failed (4)"
2481 [ "$ERROR" = "y" ] &&
2482 error "old and new files are different after sync" || true
2484 log "files compared the same"
2487 run_test 38 "MDS recreates missing lov_objid file from OST data"
2493 perl $SRCDIR/leak_finder.pl $TMP/debug 2>&1 | egrep '*** Leak:' &&
2494 error "memory leak detected" || true
2496 run_test 39 "leak_finder recognizes both LUSTRE and LNET malloc messages"
2498 test_40() { # bug 15759
2500 #define OBD_FAIL_TGT_TOOMANY_THREADS 0x706
2501 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0x80000706"
2505 run_test 40 "race during service thread startup"
2507 test_41a() { #bug 14134
2508 if [ $(facet_fstype $SINGLEMDS) == ldiskfs ] &&
2509 ! do_facet $SINGLEMDS test -b $(mdsdevname 1); then
2510 skip "Loop devices does not work with nosvc option"
2515 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
2517 start $SINGLEMDS $MDSDEV $MDS_MOUNT_OPTS -o nosvc -n
2518 start ost1 `ostdevname 1` $OST_MOUNT_OPTS
2519 start $SINGLEMDS $MDSDEV $MDS_MOUNT_OPTS -o nomgs,force
2521 mount_client $MOUNT || return 1
2524 echo "blah blah" > $MOUNT/$tfile
2527 umount_client $MOUNT
2528 stop ost1 -f || return 201
2529 stop_mds -f || return 202
2530 stop_mds -f || return 203
2531 unload_modules_conf || return 204
2534 run_test 41a "mount mds with --nosvc and --nomgs"
2537 if [ $(facet_fstype $SINGLEMDS) == ldiskfs ] &&
2538 ! do_facet $SINGLEMDS test -b $(mdsdevname 1); then
2539 skip "Loop devices does not work with nosvc option"
2543 ! combined_mgs_mds && skip "needs combined mgs device" && return 0
2547 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
2549 start $SINGLEMDS $MDSDEV $MDS_MOUNT_OPTS -o nosvc -n
2551 start $SINGLEMDS $MDSDEV $MDS_MOUNT_OPTS -o nomgs,force
2553 mount_client $MOUNT || return 1
2556 echo "blah blah" > $MOUNT/$tfile
2557 cat $MOUNT/$tfile || return 200
2559 umount_client $MOUNT
2560 stop_ost || return 201
2561 stop_mds -f || return 202
2562 stop_mds -f || return 203
2565 run_test 41b "mount mds with --nosvc and --nomgs on first mount"
2567 test_42() { #bug 14693
2569 check_mount || error "client was not mounted"
2571 do_facet mgs $LCTL conf_param $FSNAME.llite.some_wrong_param=10
2572 umount_client $MOUNT ||
2573 error "unmounting client failed with invalid llite param"
2574 mount_client $MOUNT ||
2575 error "mounting client failed with invalid llite param"
2577 do_facet mgs $LCTL conf_param $FSNAME.sys.some_wrong_param=20
2578 cleanup || error "stopping $FSNAME failed with invalid sys param"
2581 check_mount || "client was not mounted with invalid sys param"
2582 cleanup || error "stopping $FSNAME failed with invalid sys param"
2585 run_test 42 "allow client/server mount/unmount with invalid config param"
2588 [ $UID -ne 0 -o $RUNAS_ID -eq 0 ] && skip_env "run as root"
2591 USER1=$(cat /etc/passwd | grep :$ID1:$ID1: | cut -d: -f1)
2592 [ -z "$USER1" ] && skip_env "missing user with uid=$ID1 gid=$ID1" &&
2596 chmod ugo+x $DIR || error "chmod 0 failed"
2597 set_conf_param_and_check mds \
2598 "lctl get_param -n mdt.$FSNAME-MDT0000.root_squash" \
2599 "$FSNAME.mdt.root_squash" \
2601 wait_update $HOSTNAME \
2602 "lctl get_param -n llite.${FSNAME}*.root_squash" \
2604 error "check llite root_squash failed!"
2605 set_conf_param_and_check mds \
2606 "lctl get_param -n mdt.$FSNAME-MDT0000.nosquash_nids" \
2607 "$FSNAME.mdt.nosquash_nids" \
2609 wait_update $HOSTNAME \
2610 "lctl get_param -n llite.${FSNAME}*.nosquash_nids" \
2612 error "check llite nosquash_nids failed!"
2615 # create set of test files
2617 echo "111" > $DIR/$tfile-userfile || error "write 1 failed"
2618 chmod go-rw $DIR/$tfile-userfile || error "chmod 1 failed"
2619 chown $RUNAS_ID.$RUNAS_ID $DIR/$tfile-userfile || error "chown failed"
2621 echo "222" > $DIR/$tfile-rootfile || error "write 2 failed"
2622 chmod go-rw $DIR/$tfile-rootfile || error "chmod 2 faield"
2624 mkdir $DIR/$tdir-rootdir -p || error "mkdir failed"
2625 chmod go-rwx $DIR/$tdir-rootdir || error "chmod 3 failed"
2626 touch $DIR/$tdir-rootdir/tfile-1 || error "touch failed"
2628 echo "777" > $DIR/$tfile-user1file || error "write 7 failed"
2629 chmod go-rw $DIR/$tfile-user1file || error "chmod 7 failed"
2630 chown $ID1.$ID1 $DIR/$tfile-user1file || error "chown failed"
2633 # check root_squash:
2634 # set root squash UID:GID to RUNAS_ID
2635 # root should be able to access only files owned by RUNAS_ID
2637 set_conf_param_and_check mds \
2638 "lctl get_param -n mdt.$FSNAME-MDT0000.root_squash" \
2639 "$FSNAME.mdt.root_squash" \
2640 "$RUNAS_ID:$RUNAS_ID"
2641 wait_update $HOSTNAME \
2642 "lctl get_param -n llite.${FSNAME}*.root_squash" \
2643 "$RUNAS_ID:$RUNAS_ID" ||
2644 error "check llite root_squash failed!"
2646 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-userfile)
2647 dd if=$DIR/$tfile-userfile 1>/dev/null 2>/dev/null || \
2648 error "$ST: root read permission is denied"
2649 echo "$ST: root read permission is granted - ok"
2652 dd conv=notrunc of=$DIR/$tfile-userfile 1>/dev/null 2>/dev/null || \
2653 error "$ST: root write permission is denied"
2654 echo "$ST: root write permission is granted - ok"
2656 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-rootfile)
2657 dd if=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null && \
2658 error "$ST: root read permission is granted"
2659 echo "$ST: root read permission is denied - ok"
2662 dd conv=notrunc of=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null && \
2663 error "$ST: root write permission is granted"
2664 echo "$ST: root write permission is denied - ok"
2666 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tdir-rootdir)
2667 rm $DIR/$tdir-rootdir/tfile-1 1>/dev/null 2>/dev/null && \
2668 error "$ST: root unlink permission is granted"
2669 echo "$ST: root unlink permission is denied - ok"
2671 touch $DIR/tdir-rootdir/tfile-2 1>/dev/null 2>/dev/null && \
2672 error "$ST: root create permission is granted"
2673 echo "$ST: root create permission is denied - ok"
2677 # check root_squash is enforced independently
2678 # of client cache content
2680 # access file by USER1, keep access open
2681 # root should be denied access to user file
2683 runas -u $ID1 tail -f $DIR/$tfile-user1file 1>/dev/null 2>&1 &
2687 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-user1file)
2688 dd if=$DIR/$tfile-user1file 1>/dev/null 2>&1 &&
2689 { kill $pid; error "$ST: root read permission is granted"; }
2690 echo "$ST: root read permission is denied - ok"
2693 dd conv=notrunc of=$DIR/$tfile-user1file 1>/dev/null 2>&1 &&
2694 { kill $pid; error "$ST: root write permission is granted"; }
2695 echo "$ST: root write permission is denied - ok"
2701 # check nosquash_nids:
2702 # put client's NID into nosquash_nids list,
2703 # root should be able to access root file after that
2705 local NIDLIST=$(lctl list_nids all | tr '\n' ' ')
2706 NIDLIST="2@elan $NIDLIST 192.168.0.[2,10]@tcp"
2707 NIDLIST=$(echo $NIDLIST | tr -s ' ' ' ')
2708 set_conf_param_and_check mds \
2709 "lctl get_param -n mdt.$FSNAME-MDT0000.nosquash_nids" \
2710 "$FSNAME-MDTall.mdt.nosquash_nids" \
2712 wait_update $HOSTNAME \
2713 "lctl get_param -n llite.${FSNAME}*.nosquash_nids" \
2715 error "check llite nosquash_nids failed!"
2717 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-rootfile)
2718 dd if=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null || \
2719 error "$ST: root read permission is denied"
2720 echo "$ST: root read permission is granted - ok"
2723 dd conv=notrunc of=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null || \
2724 error "$ST: root write permission is denied"
2725 echo "$ST: root write permission is granted - ok"
2727 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tdir-rootdir)
2728 rm $DIR/$tdir-rootdir/tfile-1 || \
2729 error "$ST: root unlink permission is denied"
2730 echo "$ST: root unlink permission is granted - ok"
2731 touch $DIR/$tdir-rootdir/tfile-2 || \
2732 error "$ST: root create permission is denied"
2733 echo "$ST: root create permission is granted - ok"
2737 run_test 43 "check root_squash and nosquash_nids"
2739 umount_client $MOUNT
2744 check_mount || return 2
2745 UUID=$($LCTL get_param llite.${FSNAME}*.uuid | cut -d= -f2)
2747 UUIDS=$(do_facet $SINGLEMDS "$LCTL get_param mdt.${FSNAME}*.exports.*.uuid")
2748 for VAL in $UUIDS; do
2749 NID=$(echo $VAL | cut -d= -f1)
2750 CLUUID=$(echo $VAL | cut -d= -f2)
2751 [ "$UUID" = "$CLUUID" ] && STATS_FOUND=yes && break
2753 [ "$STATS_FOUND" = "no" ] && error "stats not found for client"
2757 run_test 44 "mounted client proc entry exists"
2761 check_mount || return 2
2766 #define OBD_FAIL_PTLRPC_LONG_UNLINK 0x50f
2767 do_facet client "lctl set_param fail_loc=0x50f"
2770 manual_umount_client --force || return 3
2771 do_facet client "lctl set_param fail_loc=0x0"
2773 mount_client $MOUNT || return 4
2777 run_test 45 "long unlink handling in ptlrpcd"
2784 umount_client $MOUNT2 || rc=$?
2785 umount_client $MOUNT || rc=$?
2786 while [ $count -gt 0 ]; do
2787 stop ost${count} -f || rc=$?
2791 cleanup_nocli || rc=$?
2792 #writeconf to remove all ost2 traces for subsequent tests
2793 writeconf_or_reformat
2798 echo "Testing with $OSTCOUNT OSTs"
2800 start_mds || return 1
2801 #first client should see only one ost
2802 start_ost || return 2
2803 wait_osc_import_state mds ost FULL
2805 mount_client $MOUNT || return 3
2806 trap "cleanup_46a $OSTCOUNT" EXIT ERR
2809 for (( i=2; i<=$OSTCOUNT; i++ )); do
2810 start ost$i `ostdevname $i` $OST_MOUNT_OPTS || return $((i+2))
2813 # wait until osts in sync
2814 for (( i=2; i<=$OSTCOUNT; i++ )); do
2815 wait_osc_import_state mds ost$i FULL
2816 wait_osc_import_state client ost$i FULL
2819 #second client see all ost's
2821 mount_client $MOUNT2 || return 8
2822 $LFS setstripe -c -1 $MOUNT2 || return 9
2823 $LFS getstripe $MOUNT2 || return 10
2825 echo "ok" > $MOUNT2/widestripe
2826 $LFS getstripe $MOUNT2/widestripe || return 11
2827 # fill acl buffer for avoid expand lsm to them
2828 awk -F : '{if (FNR < 25) { print "u:"$1":rwx" }}' /etc/passwd | while read acl; do
2829 setfacl -m $acl $MOUNT2/widestripe
2833 stat $MOUNT/widestripe || return 12
2835 cleanup_46a $OSTCOUNT || { echo "cleanup_46a failed!" && return 13; }
2838 run_test 46a "handle ost additional - wide striped file"
2843 check_mount || return 2
2844 $LCTL set_param ldlm.namespaces.$FSNAME-*-*-*.lru_size=100
2848 for ns in $($LCTL get_param ldlm.namespaces.$FSNAME-*-*-*.lru_size); do
2849 if echo $ns | grep "MDT[[:digit:]]*"; then
2852 lrs=$(echo $ns | sed 's/.*lru_size=//')
2853 lru_size[count]=$lrs
2858 facet_failover $SINGLEMDS
2859 client_up || return 3
2862 for ns in $($LCTL get_param ldlm.namespaces.$FSNAME-*-*-*.lru_size); do
2863 if echo $ns | grep "MDT[[:digit:]]*"; then
2866 lrs=$(echo $ns | sed 's/.*lru_size=//')
2867 if ! test "$lrs" -eq "${lru_size[count]}"; then
2868 n=$(echo $ns | sed -e 's/ldlm.namespaces.//' -e 's/.lru_size=.*//')
2869 error "$n has lost lru_size: $lrs vs. ${lru_size[count]}"
2877 run_test 47 "server restart does not make client loss lru_resize settings"
2882 # reformat after this test is needed - if test will failed
2883 # we will have unkillable file at FS
2887 test_48() { # bug 17636
2890 check_mount || return 2
2892 $LFS setstripe -c -1 $MOUNT || return 9
2893 $LFS getstripe $MOUNT || return 10
2895 echo "ok" > $MOUNT/widestripe
2896 $LFS getstripe $MOUNT/widestripe || return 11
2898 trap cleanup_48 EXIT ERR
2900 # fill acl buffer for avoid expand lsm to them
2901 getent passwd | awk -F : '{ print "u:"$1":rwx" }' | while read acl; do
2902 setfacl -m $acl $MOUNT/widestripe
2905 stat $MOUNT/widestripe || return 12
2910 run_test 48 "too many acls on file"
2912 # check PARAM_SYS_LDLM_TIMEOUT option of MKFS.LUSTRE
2913 test_49a() { # bug 17710
2914 local timeout_orig=$TIMEOUT
2915 local ldlm_timeout_orig=$LDLM_TIMEOUT
2916 local LOCAL_TIMEOUT=20
2918 LDLM_TIMEOUT=$LOCAL_TIMEOUT
2919 TIMEOUT=$LOCAL_TIMEOUT
2923 check_mount || error "client mount failed"
2925 echo "check ldlm_timout..."
2926 local LDLM_MDS="$(do_facet $SINGLEMDS lctl get_param -n ldlm_timeout)"
2927 local LDLM_OST1="$(do_facet ost1 lctl get_param -n ldlm_timeout)"
2928 local LDLM_CLIENT="$(do_facet client lctl get_param -n ldlm_timeout)"
2930 if [ $LDLM_MDS -ne $LDLM_OST1 -o $LDLM_MDS -ne $LDLM_CLIENT ]; then
2931 error "Different LDLM_TIMEOUT:$LDLM_MDS $LDLM_OST1 $LDLM_CLIENT"
2934 if [ $LDLM_MDS -ne $((LOCAL_TIMEOUT / 3)) ]; then
2935 error "LDLM_TIMEOUT($LDLM_MDS) is not $((LOCAL_TIMEOUT / 3))"
2938 umount_client $MOUNT
2939 stop_ost || error "problem stopping OSS"
2940 stop_mds || error "problem stopping MDS"
2942 LDLM_TIMEOUT=$ldlm_timeout_orig
2943 TIMEOUT=$timeout_orig
2945 run_test 49a "check PARAM_SYS_LDLM_TIMEOUT option of mkfs.lustre"
2947 test_49b() { # bug 17710
2948 local timeout_orig=$TIMEOUT
2949 local ldlm_timeout_orig=$LDLM_TIMEOUT
2950 local LOCAL_TIMEOUT=20
2952 LDLM_TIMEOUT=$((LOCAL_TIMEOUT - 1))
2953 TIMEOUT=$LOCAL_TIMEOUT
2957 check_mount || error "client mount failed"
2959 local LDLM_MDS="$(do_facet $SINGLEMDS lctl get_param -n ldlm_timeout)"
2960 local LDLM_OST1="$(do_facet ost1 lctl get_param -n ldlm_timeout)"
2961 local LDLM_CLIENT="$(do_facet client lctl get_param -n ldlm_timeout)"
2963 if [ $LDLM_MDS -ne $LDLM_OST1 -o $LDLM_MDS -ne $LDLM_CLIENT ]; then
2964 error "Different LDLM_TIMEOUT:$LDLM_MDS $LDLM_OST1 $LDLM_CLIENT"
2967 if [ $LDLM_MDS -ne $((LOCAL_TIMEOUT - 1)) ]; then
2968 error "LDLM_TIMEOUT($LDLM_MDS) is not $((LOCAL_TIMEOUT - 1))"
2971 cleanup || error "cleanup failed"
2973 LDLM_TIMEOUT=$ldlm_timeout_orig
2974 TIMEOUT=$timeout_orig
2976 run_test 49b "check PARAM_SYS_LDLM_TIMEOUT option of mkfs.lustre"
2979 # Test both statfs and lfs df and fail if either one fails
2980 multiop_bg_pause $1 f_
2983 killall -USR1 multiop
2984 [ $RC1 -ne 0 ] && log "lazystatfs multiop failed"
2985 wait $PID || { RC1=$?; log "multiop return error "; }
2992 if [ $RC2 -eq 0 ]; then
2994 log "lazystatfs df failed"
2998 [[ $RC1 -ne 0 || $RC2 -eq 0 ]] && RC=1
3004 lctl set_param llite.$FSNAME-*.lazystatfs=1
3007 lazystatfs $MOUNT || error "lazystatfs failed but no down servers"
3009 cleanup || return $?
3011 run_test 50a "lazystatfs all servers available =========================="
3015 lctl set_param llite.$FSNAME-*.lazystatfs=1
3018 # Wait for client to detect down OST
3019 stop_ost || error "Unable to stop OST1"
3020 wait_osc_import_state mds ost DISCONN
3022 lazystatfs $MOUNT || error "lazystatfs should don't have returned EIO"
3024 umount_client $MOUNT || error "Unable to unmount client"
3025 stop_mds || error "Unable to stop MDS"
3027 run_test 50b "lazystatfs all servers down =========================="
3030 start_mds || error "Unable to start MDS"
3031 start_ost || error "Unable to start OST1"
3032 start_ost2 || error "Unable to start OST2"
3033 mount_client $MOUNT || error "Unable to mount client"
3034 lctl set_param llite.$FSNAME-*.lazystatfs=1
3037 # Wait for client to detect down OST
3038 stop_ost || error "Unable to stop OST1"
3039 wait_osc_import_state mds ost DISCONN
3040 lazystatfs $MOUNT || error "lazystatfs failed with one down server"
3042 umount_client $MOUNT || error "Unable to unmount client"
3043 stop_ost2 || error "Unable to stop OST2"
3044 stop_mds || error "Unable to stop MDS"
3045 #writeconf to remove all ost2 traces for subsequent tests
3046 writeconf_or_reformat
3048 run_test 50c "lazystatfs one server down =========================="
3051 start_mds || error "Unable to start MDS"
3052 start_ost || error "Unable to start OST1"
3053 start_ost2 || error "Unable to start OST2"
3054 mount_client $MOUNT || error "Unable to mount client"
3055 lctl set_param llite.$FSNAME-*.lazystatfs=1
3058 # Issue the statfs during the window where the client still
3059 # belives the OST to be available but it is in fact down.
3060 # No failure just a statfs which hangs for a timeout interval.
3061 stop_ost || error "Unable to stop OST1"
3062 lazystatfs $MOUNT || error "lazystatfs failed with one down server"
3064 umount_client $MOUNT || error "Unable to unmount client"
3065 stop_ost2 || error "Unable to stop OST2"
3066 stop_mds || error "Unable to stop MDS"
3067 #writeconf to remove all ost2 traces for subsequent tests
3068 writeconf_or_reformat
3070 run_test 50d "lazystatfs client/server conn race =========================="
3077 start_mds || return 1
3078 #first client should see only one ost
3079 start_ost || return 2
3080 wait_osc_import_state mds ost FULL
3082 # Wait for client to detect down OST
3083 stop_ost || error "Unable to stop OST1"
3084 wait_osc_import_state mds ost DISCONN
3086 mount_client $MOUNT || error "Unable to mount client"
3087 lctl set_param llite.$FSNAME-*.lazystatfs=0
3089 multiop_bg_pause $MOUNT _f
3093 if [ $RC1 -ne 0 ]; then
3094 log "multiop failed $RC1"
3097 sleep $(( $TIMEOUT+1 ))
3099 [ $? -ne 0 ] && error "process isn't sleep"
3100 start_ost || error "Unable to start OST1"
3101 wait $pid || error "statfs failed"
3104 umount_client $MOUNT || error "Unable to unmount client"
3105 stop_ost || error "Unable to stop OST1"
3106 stop_mds || error "Unable to stop MDS"
3108 run_test 50e "normal statfs all servers down =========================="
3113 CONN_PROC="osc.$FSNAME-OST0001-osc-[M]*.ost_server_uuid"
3115 start_mds || error "Unable to start mds"
3116 #first client should see only one ost
3117 start_ost || error "Unable to start OST1"
3118 wait_osc_import_state mds ost FULL
3120 start_ost2 || error "Unable to start OST2"
3121 wait_osc_import_state mds ost2 FULL
3123 # Wait for client to detect down OST
3124 stop_ost2 || error "Unable to stop OST2"
3126 wait_osc_import_state mds ost2 DISCONN
3127 mount_client $MOUNT || error "Unable to mount client"
3128 lctl set_param llite.$FSNAME-*.lazystatfs=0
3130 multiop_bg_pause $MOUNT _f
3134 if [ $RC1 -ne 0 ]; then
3135 log "lazystatfs multiop failed $RC1"
3138 sleep $(( $TIMEOUT+1 ))
3140 [ $? -ne 0 ] && error "process isn't sleep"
3141 start_ost2 || error "Unable to start OST2"
3142 wait $pid || error "statfs failed"
3143 stop_ost2 || error "Unable to stop OST2"
3146 umount_client $MOUNT || error "Unable to unmount client"
3147 stop_ost || error "Unable to stop OST1"
3148 stop_mds || error "Unable to stop MDS"
3149 #writeconf to remove all ost2 traces for subsequent tests
3150 writeconf_or_reformat
3152 run_test 50f "normal statfs one server in down =========================="
3155 [ "$OSTCOUNT" -lt "2" ] && skip_env "$OSTCOUNT < 2, skipping" && return
3157 start_ost2 || error "Unable to start OST2"
3158 wait_osc_import_state mds ost2 FULL
3159 wait_osc_import_state client ost2 FULL
3161 local PARAM="${FSNAME}-OST0001.osc.active"
3163 $LFS setstripe -c -1 $DIR/$tfile || error "Unable to lfs setstripe"
3164 do_facet mgs $LCTL conf_param $PARAM=0 || error "Unable to deactivate OST"
3166 umount_client $MOUNT || error "Unable to unmount client"
3167 mount_client $MOUNT || error "Unable to mount client"
3168 # This df should not cause a panic
3171 do_facet mgs $LCTL conf_param $PARAM=1 || error "Unable to activate OST"
3173 umount_client $MOUNT || error "Unable to unmount client"
3174 stop_ost2 || error "Unable to stop OST2"
3175 stop_ost || error "Unable to stop OST1"
3176 stop_mds || error "Unable to stop MDS"
3177 #writeconf to remove all ost2 traces for subsequent tests
3178 writeconf_or_reformat
3180 run_test 50g "deactivated OST should not cause panic====================="
3184 # prepare MDT/OST, make OSC inactive for OST1
3185 [ "$OSTCOUNT" -lt "2" ] && skip_env "$OSTCOUNT < 2, skipping" && return
3186 do_facet ost1 "$TUNEFS --param osc.active=0 `ostdevname 1`" ||
3187 error "tunefs OST1 failed"
3188 start_mds || error "Unable to start MDT"
3189 start_ost || error "Unable to start OST1"
3190 start_ost2 || error "Unable to start OST2"
3191 mount_client $MOUNT || error "client start failed"
3195 # activatate OSC for OST1
3196 local TEST="$LCTL get_param -n osc.${FSNAME}-OST0000-osc-[!M]*.active"
3197 set_conf_param_and_check client \
3198 "$TEST" "${FSNAME}-OST0000.osc.active" 1 ||
3199 error "Unable to activate OST1"
3201 mkdir -p $DIR/$tdir/2
3202 $LFS setstripe -c -1 -i 0 $DIR/$tdir/2
3203 sleep 1 && echo "create a file after OST1 is activated"
3205 createmany -o $DIR/$tdir/2/$tfile-%d 1
3207 # check OSC import is working
3208 stat $DIR/$tdir/2/* >/dev/null 2>&1 ||
3209 error "some OSC imports are still not connected"
3212 umount_client $MOUNT || error "Unable to umount client"
3213 stop_ost2 || error "Unable to stop OST2"
3216 run_test 50h "LU-642: activate deactivated OST ==="
3219 local LOCAL_TIMEOUT=20
3223 check_mount || return 1
3226 $LFS setstripe -c -1 $MOUNT/d1
3227 #define OBD_FAIL_MDS_REINT_DELAY 0x142
3228 do_facet $SINGLEMDS "lctl set_param fail_loc=0x142"
3229 touch $MOUNT/d1/f1 &
3232 start_ost2 || return 2
3234 stop_ost2 || return 3
3236 #writeconf to remove all ost2 traces for subsequent tests
3237 writeconf_or_reformat
3239 run_test 51 "Verify that mdt_reint handles RMF_MDT_MD correctly when an OST is added"
3248 do_node $node mkdir -p $dest
3249 [ $? -eq 0 ] || { error "Unable to create directory"; return 1; }
3251 do_node $node 'tar cf - '$@' | tar xf - -C '$dest';
3252 [ \"\${PIPESTATUS[*]}\" = \"0 0\" ] || exit 1'
3253 [ $? -eq 0 ] || { error "Unable to tar files"; return 2; }
3255 do_node $node 'getfattr -d -m "[a-z]*\\." '$@' > '$xattrs
3256 [ $? -eq 0 ] || { error "Unable to read xattrs"; return 3; }
3266 local backup2=${TMP}/backup2
3268 do_node $node mkdir -p $backup2
3269 [ $? -eq 0 ] || { error "Unable to create directory"; return 1; }
3271 do_node $node 'tar cf - '$@' | tar xf - -C '$backup2';
3272 [ \"\${PIPESTATUS[*]}\" = \"0 0\" ] || exit 1'
3273 [ $? -eq 0 ] || { error "Unable to tar files to diff"; return 2; }
3275 do_node $node "diff -rq $backup $backup2"
3276 [ $? -eq 0 ] || { error "contents differ"; return 3; }
3278 local xattrs2=${TMP}/xattrs2
3279 do_node $node 'getfattr -d -m "[a-z]*\\." '$@' > '$xattrs2
3280 [ $? -eq 0 ] || { error "Unable to read xattrs to diff"; return 4; }
3282 do_node $node "diff $xattrs $xattrs2"
3283 [ $? -eq 0 ] || { error "xattrs differ"; return 5; }
3285 do_node $node "rm -rf $backup2 $xattrs2"
3286 [ $? -eq 0 ] || { error "Unable to delete temporary files"; return 6; }
3290 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
3291 skip "Only applicable to ldiskfs-based MDTs"
3296 [ $? -eq 0 ] || { error "Unable to start MDS"; return 1; }
3298 [ $? -eq 0 ] || { error "Unable to start OST1"; return 2; }
3300 [ $? -eq 0 ] || { error "Unable to mount client"; return 3; }
3303 local ost1mnt=$(facet_mntpt ost1)
3304 local ost1node=$(facet_active_host ost1)
3305 local ost1tmp=$TMP/conf52
3309 [ $? -eq 0 ] || { error "Unable to create tdir"; return 4; }
3310 touch $TMP/modified_first
3311 [ $? -eq 0 ] || { error "Unable to create temporary file"; return 5; }
3312 local mtime=$(stat -c %Y $TMP/modified_first)
3313 do_node $ost1node "mkdir -p $ost1tmp && touch -m -d @$mtime $ost1tmp/modified_first"
3315 [ $? -eq 0 ] || { error "Unable to create temporary file"; return 6; }
3318 $LFS setstripe -c -1 -S 1M $DIR/$tdir
3319 [ $? -eq 0 ] || { error "lfs setstripe failed"; return 7; }
3321 for (( i=0; i < nrfiles; i++ )); do
3322 multiop $DIR/$tdir/$tfile-$i Ow1048576w1048576w524288c
3323 [ $? -eq 0 ] || { error "multiop failed"; return 8; }
3329 echo backup files to $TMP/files
3330 local files=$(find $DIR/$tdir -type f -newer $TMP/modified_first)
3331 copy_files_xattrs `hostname` $TMP/files $TMP/file_xattrs $files
3332 [ $? -eq 0 ] || { error "Unable to copy files"; return 9; }
3334 umount_client $MOUNT
3335 [ $? -eq 0 ] || { error "Unable to umount client"; return 10; }
3337 [ $? -eq 0 ] || { error "Unable to stop ost1"; return 11; }
3339 echo mount ost1 as ldiskfs
3340 do_node $ost1node mkdir -p $ost1mnt
3341 [ $? -eq 0 ] || { error "Unable to create $ost1mnt"; return 23; }
3342 if ! do_node $ost1node test -b $ost1_dev; then
3345 do_node $ost1node mount -t $(facet_fstype ost1) $loop $ost1_dev \
3347 [ $? -eq 0 ] || { error "Unable to mount ost1 as ldiskfs"; return 12; }
3350 echo backup objects to $ost1tmp/objects
3351 local objects=$(do_node $ost1node 'find '$ost1mnt'/O/[0-9]* -type f'\
3352 '-size +0 -newer '$ost1tmp'/modified_first -regex ".*\/[0-9]+"')
3353 copy_files_xattrs $ost1node $ost1tmp/objects $ost1tmp/object_xattrs \
3355 [ $? -eq 0 ] || { error "Unable to copy objects"; return 13; }
3357 # move objects to lost+found
3358 do_node $ost1node 'mv '$objects' '${ost1mnt}'/lost+found'
3359 [ $? -eq 0 ] || { error "Unable to move objects"; return 14; }
3361 # recover objects dry-run
3362 echo "ll_recover_lost_found_objs dry_run"
3363 do_node $ost1node "ll_recover_lost_found_objs -n -d $ost1mnt/O"
3364 [ $? -eq 0 ] || { error "ll_recover_lost_found_objs failed"; return 15; }
3367 echo "ll_recover_lost_found_objs fix run"
3368 do_node $ost1node "ll_recover_lost_found_objs -d $ost1mnt/lost+found"
3369 [ $? -eq 0 ] || { error "ll_recover_lost_found_objs failed"; return 15; }
3371 # compare restored objects against saved ones
3372 diff_files_xattrs $ost1node $ost1tmp/objects $ost1tmp/object_xattrs $objects
3373 [ $? -eq 0 ] || { error "Unable to diff objects"; return 16; }
3375 do_node $ost1node "umount $ost1mnt"
3376 [ $? -eq 0 ] || { error "Unable to umount ost1 as ldiskfs"; return 17; }
3379 [ $? -eq 0 ] || { error "Unable to start ost1"; return 18; }
3381 [ $? -eq 0 ] || { error "Unable to mount client"; return 19; }
3384 diff_files_xattrs `hostname` $TMP/files $TMP/file_xattrs $files
3385 [ $? -eq 0 ] || { error "Unable to diff files"; return 20; }
3387 rm -rf $TMP/files $TMP/file_xattrs
3388 [ $? -eq 0 ] || { error "Unable to delete temporary files"; return 21; }
3389 do_node $ost1node "rm -rf $ost1tmp"
3390 [ $? -eq 0 ] || { error "Unable to delete temporary files"; return 22; }
3393 run_test 52 "check recovering objects from lost+found"
3395 # Checks threads_min/max/started for some service
3397 # Arguments: service name (OST or MDT), facet (e.g., ost1, $SINGLEMDS), and a
3398 # parameter pattern prefix like 'ost.*.ost'.
3411 local msg="Insane $modname thread counts"
3412 local ncpts=$(check_cpt_number $facet)
3416 check_mount || return 41
3418 # We need to expand $parampat, but it may match multiple parameters, so
3419 # we'll pick the first one
3420 if ! paramp=$(do_facet $facet "lctl get_param -N ${parampat}.threads_min"|head -1); then
3421 error "Couldn't expand ${parampat}.threads_min parameter name"
3425 # Remove the .threads_min part
3426 paramp=${paramp%.threads_min}
3428 # Check for sanity in defaults
3429 tmin=$(do_facet $facet "lctl get_param -n ${paramp}.threads_min" || echo 0)
3430 tmax=$(do_facet $facet "lctl get_param -n ${paramp}.threads_max" || echo 0)
3431 tstarted=$(do_facet $facet "lctl get_param -n ${paramp}.threads_started" || echo 0)
3432 lassert 23 "$msg (PDSH problems?)" '(($tstarted && $tmin && $tmax))' || return $?
3433 lassert 24 "$msg" '(($tstarted >= $tmin && $tstarted <= $tmax ))' || return $?
3434 nthrs=$(expr $tmax - $tmin)
3435 if [ $nthrs -lt $ncpts ]; then
3441 [ $tmin -eq $tmax -a $tmin -eq $tstarted ] &&
3442 skip_env "module parameter forced $facet thread count" &&
3443 tmin=3 && tmax=$((3 * tmax))
3445 # Check that we can change min/max
3446 do_facet $facet "lctl set_param ${paramp}.threads_min=$((tmin + nthrs))"
3447 do_facet $facet "lctl set_param ${paramp}.threads_max=$((tmax - nthrs))"
3448 tmin2=$(do_facet $facet "lctl get_param -n ${paramp}.threads_min" || echo 0)
3449 tmax2=$(do_facet $facet "lctl get_param -n ${paramp}.threads_max" || echo 0)
3450 lassert 25 "$msg" '(($tmin2 == ($tmin + $nthrs) && $tmax2 == ($tmax - $nthrs)))' || return $?
3452 # Check that we can set min/max to the same value
3453 tmin=$(do_facet $facet "lctl get_param -n ${paramp}.threads_min" || echo 0)
3454 do_facet $facet "lctl set_param ${paramp}.threads_max=$tmin"
3455 tmin2=$(do_facet $facet "lctl get_param -n ${paramp}.threads_min" || echo 0)
3456 tmax2=$(do_facet $facet "lctl get_param -n ${paramp}.threads_max" || echo 0)
3457 lassert 26 "$msg" '(($tmin2 == $tmin && $tmax2 == $tmin))' || return $?
3459 # Check that we can't set max < min
3460 do_facet $facet "lctl set_param ${paramp}.threads_max=$((tmin - 1))"
3461 tmin2=$(do_facet $facet "lctl get_param -n ${paramp}.threads_min" || echo 0)
3462 tmax2=$(do_facet $facet "lctl get_param -n ${paramp}.threads_max" || echo 0)
3463 lassert 27 "$msg" '(($tmin2 <= $tmax2))' || return $?
3465 # We need to ensure that we get the module options desired; to do this
3466 # we set LOAD_MODULES_REMOTE=true and we call setmodopts below.
3467 LOAD_MODULES_REMOTE=true
3470 local newvalue="${opts}=$(expr $basethr \* $ncpts)"
3471 setmodopts -a $modname "$newvalue" oldvalue
3475 check_mount || return 41
3477 # Restore previous setting of MODOPTS_*
3478 setmodopts $modname "$oldvalue"
3480 # Check that $opts took
3481 tmin=$(do_facet $facet "lctl get_param -n ${paramp}.threads_min")
3482 tmax=$(do_facet $facet "lctl get_param -n ${paramp}.threads_max")
3483 tstarted=$(do_facet $facet "lctl get_param -n ${paramp}.threads_started")
3484 lassert 28 "$msg" '(($tstarted >= $tmin && $tstarted <= $tmax ))' || return $?
3493 thread_sanity OST ost1 'ost.*.ost' 'oss_num_threads' '16'
3496 run_test 53a "check OSS thread count params"
3500 local mds=$(do_facet $SINGLEMDS "lctl get_param -N mds.*.*.threads_max \
3502 if [ -z "$mds" ]; then
3503 #running this on an old MDT
3504 thread_sanity MDT $SINGLEMDS 'mdt.*.*.' 'mdt_num_threads' 16
3506 thread_sanity MDT $SINGLEMDS 'mds.*.*.' 'mds_num_threads' 16
3510 run_test 53b "check MDS thread count params"
3513 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
3514 skip "Only applicable to ldiskfs-based MDTs"
3518 do_rpc_nodes $(facet_host ost1) run_llverdev $(ostdevname 1) -p
3519 [ $? -eq 0 ] || error "llverdev failed!"
3522 run_test 54a "test llverdev and partial verify of device"
3525 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
3526 skip "Only applicable to ldiskfs-based MDTs"
3531 run_llverfs $MOUNT -p
3532 [ $? -eq 0 ] || error "llverfs failed!"
3535 run_test 54b "test llverfs and partial verify of filesystem"
3539 local max_ost_index=$1
3540 echo -n $(((max_ost_index + 1) * 8))
3544 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
3545 skip "Only applicable to ldiskfs-based MDTs"
3549 local mdsdev=$(mdsdevname 1)
3550 local mdsvdev=$(mdsvdevname 1)
3554 add mds1 $(mkfs_opts mds1 ${mdsdev}) --reformat $mdsdev \
3556 add ost1 $(mkfs_opts ost1 $(ostdevname 1)) --index=$i \
3557 --reformat $(ostdevname 1) $(ostvdevname 1)
3563 echo checking size of lov_objid for ost index $i
3564 LOV_OBJID_SIZE=$(do_facet mds1 "$DEBUGFS -R 'stat lov_objid' $mdsdev 2>/dev/null" | grep ^User | awk '{print $6}')
3565 if [ "$LOV_OBJID_SIZE" != $(lov_objid_size $i) ]; then
3566 error "lov_objid size has to be $(lov_objid_size $i), not $LOV_OBJID_SIZE"
3568 echo ok, lov_objid size is correct: $LOV_OBJID_SIZE
3575 run_test 55 "check lov_objid size"
3578 local mds_journal_size_orig=$MDSJOURNALSIZE
3582 for num in $(seq 1 $MDSCOUNT); do
3583 add mds${num} $(mkfs_opts mds${num} $(mdsdevname $num)) \
3584 --reformat $(mdsdevname $num) $(mdsvdevname $num)
3586 add ost1 $(mkfs_opts ost1 $(ostdevname 1)) --index=1000 --reformat \
3587 $(ostdevname 1) $(ostvdevname 1)
3588 add ost2 $(mkfs_opts ost2 $(ostdevname 2)) --index=10000 --reformat \
3589 $(ostdevname 2) $(ostvdevname 2)
3593 start_ost2 || error "Unable to start second ost"
3594 mount_client $MOUNT || error "Unable to mount client"
3598 MDSJOURNALSIZE=$mds_journal_size_orig
3601 run_test 56 "check big indexes"
3603 test_57a() { # bug 22656
3604 local NID=$(do_facet ost1 "$LCTL get_param nis" | tail -1 | awk '{print $1}')
3605 writeconf_or_reformat
3606 [ $(facet_fstype ost1) == zfs ] && import_zpool ost1
3607 do_facet ost1 "$TUNEFS --failnode=$NID `ostdevname 1`" || error "tunefs failed"
3609 start_ost && error "OST registration from failnode should fail"
3612 run_test 57a "initial registration from failnode should fail (should return errs)"
3615 local NID=$(do_facet ost1 "$LCTL get_param nis" | tail -1 | awk '{print $1}')
3616 writeconf_or_reformat
3617 [ $(facet_fstype ost1) == zfs ] && import_zpool ost1
3618 do_facet ost1 "$TUNEFS --servicenode=$NID `ostdevname 1`" || error "tunefs failed"
3620 start_ost || error "OST registration from servicenode should not fail"
3623 run_test 57b "initial registration from servicenode should not fail"
3626 do_facet mgs $LCTL get_param mgs.MGS.live.$FSNAME | grep OST | wc -l
3629 test_58() { # bug 22658
3632 createmany -o $DIR/$tdir/$tfile-%d 100
3633 # make sure that OSTs do not cancel llog cookies before we unmount the MDS
3634 #define OBD_FAIL_OBD_LOG_CANCEL_NET 0x601
3635 do_facet $SINGLEMDS "lctl set_param fail_loc=0x601"
3636 unlinkmany $DIR/$tdir/$tfile-%d 100
3639 local MNTDIR=$(facet_mntpt $SINGLEMDS)
3640 local devname=$(mdsdevname ${SINGLEMDS//mds/})
3642 # remove all files from the OBJECTS dir
3643 mount_fstype $SINGLEMDS
3645 do_facet $SINGLEMDS "find $MNTDIR/O/1/d* -type f -delete"
3647 unmount_fstype $SINGLEMDS
3648 # restart MDS with missing llog files
3650 do_facet mds "lctl set_param fail_loc=0"
3653 run_test 58 "missing llog files must not prevent MDT from mounting"
3656 start_mgsmds >> /dev/null
3657 local C1=$(count_osts)
3658 if [ $C1 -eq 0 ]; then
3659 start_ost >> /dev/null
3663 echo "original ost count: $C1 (expect > 0)"
3664 [ $C1 -gt 0 ] || error "No OSTs in $FSNAME log"
3665 start_mgsmds -o writeconf >> /dev/null || error "MDT start failed"
3666 local C2=$(count_osts)
3667 echo "after mdt writeconf count: $C2 (expect 0)"
3668 [ $C2 -gt 0 ] && error "MDT writeconf should erase OST logs"
3669 echo "OST start without writeconf should fail:"
3670 start_ost >> /dev/null && error "OST start without writeconf didn't fail"
3671 echo "OST start with writeconf should succeed:"
3672 start_ost -o writeconf >> /dev/null || error "OST1 start failed"
3673 local C3=$(count_osts)
3674 echo "after ost writeconf count: $C3 (expect 1)"
3675 [ $C3 -eq 1 ] || error "new OST writeconf should add:"
3676 start_ost2 -o writeconf >> /dev/null || error "OST2 start failed"
3677 local C4=$(count_osts)
3678 echo "after ost2 writeconf count: $C4 (expect 2)"
3679 [ $C4 -eq 2 ] || error "OST2 writeconf should add log"
3680 stop_ost2 >> /dev/null
3681 cleanup_nocli >> /dev/null
3682 #writeconf to remove all ost2 traces for subsequent tests
3683 writeconf_or_reformat
3685 run_test 59 "writeconf mount option"
3687 test_60() { # LU-471
3690 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
3691 skip "Only applicable to ldiskfs-based MDTs"
3695 for num in $(seq $MDSCOUNT); do
3696 add mds${num} $(mkfs_opts mds${num} $(mdsdevname $num)) \
3697 --mkfsoptions='\" -E stride=64 -O ^uninit_bg\"' \
3698 --reformat $(mdsdevname $num) $(mdsvdevname $num) ||
3702 dump=$(do_facet $SINGLEMDS dumpe2fs $(mdsdevname 1))
3704 [ $rc -eq 0 ] || error "dumpe2fs $(mdsdevname 1) failed"
3706 # MDT default has dirdata feature
3707 echo $dump | grep dirdata > /dev/null || error "dirdata is not set"
3708 # we disable uninit_bg feature
3709 echo $dump | grep uninit_bg > /dev/null && error "uninit_bg is set"
3710 # we set stride extended options
3711 echo $dump | grep stride > /dev/null || error "stride is not set"
3714 run_test 60 "check mkfs.lustre --mkfsoptions -E -O options setting"
3717 local reformat=false
3719 [ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.1.53) ] ||
3720 { skip "Need MDS version at least 2.1.53"; return 0; }
3722 if [ $(facet_fstype $SINGLEMDS) == ldiskfs ] &&
3723 ! large_xattr_enabled; then
3725 LDISKFS_MKFS_OPTS+=" -O large_xattr"
3727 for num in $(seq $MDSCOUNT); do
3728 add mds${num} $(mkfs_opts mds$num $(mdsdevname $num)) \
3729 --reformat $(mdsdevname $num) $(mdsvdevname $num) ||
3730 error "add mds $num failed"
3734 setup_noconfig || error "setting up the filesystem failed"
3735 client_up || error "starting client failed"
3737 local file=$DIR/$tfile
3740 local large_value="$(generate_string $(max_xattr_size))"
3741 local small_value="bar"
3743 local name="trusted.big"
3744 log "save large xattr $name on $file"
3745 setfattr -n $name -v $large_value $file ||
3746 error "saving $name on $file failed"
3748 local new_value=$(get_xattr_value $name $file)
3749 [[ "$new_value" != "$large_value" ]] &&
3750 error "$name different after saving"
3752 log "shrink value of $name on $file"
3753 setfattr -n $name -v $small_value $file ||
3754 error "shrinking value of $name on $file failed"
3756 new_value=$(get_xattr_value $name $file)
3757 [[ "$new_value" != "$small_value" ]] &&
3758 error "$name different after shrinking"
3760 log "grow value of $name on $file"
3761 setfattr -n $name -v $large_value $file ||
3762 error "growing value of $name on $file failed"
3764 new_value=$(get_xattr_value $name $file)
3765 [[ "$new_value" != "$large_value" ]] &&
3766 error "$name different after growing"
3768 log "check value of $name on $file after remounting MDS"
3770 new_value=$(get_xattr_value $name $file)
3771 [[ "$new_value" != "$large_value" ]] &&
3772 error "$name different after remounting MDS"
3774 log "remove large xattr $name from $file"
3775 setfattr -x $name $file || error "removing $name from $file failed"
3780 LDISKFS_MKFS_OPTS=${LDISKFS_MKFS_OPTS% -O large_xattr}
3784 run_test 61 "large xattr"
3787 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
3788 skip "Only applicable to ldiskfs-based MDTs"
3793 local mdsdev=$(mdsdevname 1)
3794 local ostdev=$(ostdevname 1)
3796 [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.2.51) ]] ||
3797 { skip "Need MDS version at least 2.2.51"; return 0; }
3799 echo "disable journal for mds"
3800 do_facet mds tune2fs -O ^has_journal $mdsdev || error "tune2fs failed"
3801 start_mds && error "MDT start should fail"
3802 echo "disable journal for ost"
3803 do_facet ost1 tune2fs -O ^has_journal $ostdev || error "tune2fs failed"
3804 start_ost && error "OST start should fail"
3805 cleanup || return $?
3808 run_test 62 "start with disabled journal"
3811 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
3812 skip "Only applicable to ldiskfs-based MDTs"
3816 local inode_slab=$(do_facet $SINGLEMDS \
3817 "awk '/ldiskfs_inode_cache/ { print \\\$5 }' /proc/slabinfo")
3818 if [ -z "$inode_slab" ]; then
3819 skip "ldiskfs module has not been loaded"
3823 echo "$inode_slab ldisk inodes per page"
3824 [ "$inode_slab" -ge "3" ] ||
3825 error "ldisk inode size is too big, $inode_slab objs per page"
3828 run_test 63 "Verify each page can at least hold 3 ldisk inodes"
3833 start_ost2 || error "Unable to start second ost"
3834 mount_client $MOUNT || error "Unable to mount client"
3835 stop_ost2 || error "Unable to stop second ost"
3837 $LFS df --lazy || error "lfs df failed"
3838 cleanup || return $?
3839 #writeconf to remove all ost2 traces for subsequent tests
3840 writeconf_or_reformat
3842 run_test 64 "check lfs df --lazy "
3844 test_65() { # LU-2237
3845 # Currently, the test is only valid for ldiskfs backend
3846 [ "$(facet_fstype $SINGLEMDS)" != "ldiskfs" ] &&
3847 skip "non-ldiskfs backend" && return
3849 local devname=$(mdsdevname ${SINGLEMDS//mds/})
3850 local brpt=$(facet_mntpt brpt)
3853 if ! do_facet $SINGLEMDS "test -b $devname"; then
3858 local obj=$(do_facet $SINGLEMDS \
3859 "$DEBUGFS -c -R \\\"stat last_rcvd\\\" $devname" |
3861 if [ -z "$obj" ]; then
3862 # The MDT may be just re-formatted, mount the MDT for the
3863 # first time to guarantee the "last_rcvd" file is there.
3864 start_mds || error "fail to mount the MDS for the first time"
3868 # remove the "last_rcvd" file
3869 do_facet $SINGLEMDS "mkdir -p $brpt"
3870 do_facet $SINGLEMDS \
3871 "mount -t $(facet_fstype $SINGLEMDS) $opts $devname $brpt"
3872 do_facet $SINGLEMDS "rm -f ${brpt}/last_rcvd"
3873 do_facet $SINGLEMDS "umount -d $brpt"
3875 # restart MDS, the "last_rcvd" file should be recreated.
3876 start_mds || error "fail to restart the MDS"
3878 obj=$(do_facet $SINGLEMDS \
3879 "$DEBUGFS -c -R \\\"stat last_rcvd\\\" $devname" | grep Inode)
3880 [ -n "$obj" ] || error "fail to re-create the last_rcvd"
3882 run_test 65 "re-create the lost last_rcvd file when server mount"
3885 [[ $(lustre_version_code mgs) -ge $(version_code 2.3.59) ]] ||
3886 { skip "Need MGS version at least 2.3.59"; return 0; }
3889 local OST1_NID=$(do_facet ost1 $LCTL list_nids | head -1)
3890 local MDS_NID=$(do_facet $SINGLEMDS $LCTL list_nids | head -1)
3892 echo "replace_nids should fail if MDS, OSTs and clients are UP"
3893 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 $OST1_NID &&
3894 error "replace_nids fail"
3896 umount_client $MOUNT || error "unmounting client failed"
3897 echo "replace_nids should fail if MDS and OSTs are UP"
3898 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 $OST1_NID &&
3899 error "replace_nids fail"
3902 echo "replace_nids should fail if MDS is UP"
3903 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 $OST1_NID &&
3904 error "replace_nids fail"
3906 stop_mds || error "stopping mds failed"
3908 if combined_mgs_mds; then
3909 start_mdt 1 "-o nosvc" ||
3910 error "starting mds with nosvc option failed"
3913 echo "command should accept two parameters"
3914 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 &&
3915 error "command should accept two params"
3917 echo "correct device name should be passed"
3918 do_facet mgs $LCTL replace_nids $FSNAME-WRONG0000 $OST1_NID &&
3919 error "wrong devname"
3921 echo "wrong nids list should not destroy the system"
3922 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 "wrong nids list" &&
3925 echo "replace OST nid"
3926 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 $OST1_NID ||
3927 error "replace nids failed"
3929 echo "command should accept two parameters"
3930 do_facet mgs $LCTL replace_nids $FSNAME-MDT0000 &&
3931 error "command should accept two params"
3933 echo "wrong nids list should not destroy the system"
3934 do_facet mgs $LCTL replace_nids $FSNAME-MDT0000 "wrong nids list" &&
3937 echo "replace MDS nid"
3938 do_facet mgs $LCTL replace_nids $FSNAME-MDT0000 $MDS_NID ||
3939 error "replace nids failed"
3941 if ! combined_mgs_mds ; then
3948 check_mount || error "error after nid replace"
3949 cleanup || error "cleanup failed"
3952 run_test 66 "replace nids"
3954 test_67() { #LU-2950
3955 local legacy="$TMP/legacy_lnet_config"
3956 local new="$TMP/new_routes_test"
3957 local out="$TMP/config_out_file"
3958 local verify="$TMP/conv_verify"
3959 local verify_conf="$TMP/conf_verify"
3961 # Create the legacy file that will be run through the
3962 # lustre_routes_conversion script
3963 cat <<- LEGACY_LNET_CONFIG > $legacy
3964 tcp1 23 192.168.213.1@tcp:1; tcp5 34 193.30.4.3@tcp:4;
3965 tcp2 54 10.1.3.2@tcp;
3966 tcp3 10.3.4.3@tcp:3;
3970 # Create the verification file to verify the output of
3971 # lustre_routes_conversion script against.
3972 cat <<- VERIFY_LNET_CONFIG > $verify
3973 tcp1: { gateway: 192.168.213.1@tcp, hop: 23, priority: 1 }
3974 tcp5: { gateway: 193.30.4.3@tcp, hop: 34, priority: 4 }
3975 tcp2: { gateway: 10.1.3.2@tcp, hop: 54 }
3976 tcp3: { gateway: 10.3.4.3@tcp, priority: 3 }
3977 tcp4: { gateway: 10.3.3.4@tcp }
3980 # Create the verification file to verify the output of
3981 # lustre_routes_config script against
3982 cat <<- VERIFY_LNET_CONFIG > $verify_conf
3983 lctl --net tcp1 add_route 192.168.213.1@tcp 23 1
3984 lctl --net tcp5 add_route 193.30.4.3@tcp 34 4
3985 lctl --net tcp2 add_route 10.1.3.2@tcp 54 4
3986 lctl --net tcp3 add_route 10.3.4.3@tcp 1 3
3987 lctl --net tcp4 add_route 10.3.3.4@tcp 1 3
3990 lustre_routes_conversion $legacy $new > /dev/null
3991 if [ -f $new ]; then
3992 # verify the conversion output
3993 cmp -s $new $verify > /dev/null
3994 if [ $? -eq 1 ]; then
3995 error "routes conversion failed"
3998 lustre_routes_config --dry-run --verbose $new > $out
3999 # check that the script succeeded
4000 cmp -s $out $verify_conf > /dev/null
4001 if [ $? -eq 1 ]; then
4002 error "routes config failed"
4005 error "routes conversion test failed"
4007 # remove generated files
4008 rm -f $new $legacy $verify $verify_conf $out
4010 run_test 67 "test routes conversion and configuration"
4018 [ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.4.53) ] ||
4019 { skip "Need MDS version at least 2.4.53"; return 0; }
4021 umount_client $MOUNT || error "umount client failed"
4023 start_mdt 1 || error "MDT start failed"
4026 # START-END - the sequences we'll be reserving
4027 START=$(do_facet $SINGLEMDS \
4028 lctl get_param -n seq.ctl*.space | awk -F'[[ ]' '{print $2}')
4029 END=$((START + (1 << 30)))
4030 do_facet $SINGLEMDS \
4031 lctl set_param seq.ctl*.fldb="[$START-$END\):0:mdt"
4033 # reset the sequences MDT0000 has already assigned
4034 do_facet $SINGLEMDS \
4035 lctl set_param seq.srv*MDT0000.space=clear
4037 # remount to let the client allocate new sequence
4038 mount_client $MOUNT || error "mount client failed"
4041 do_facet $SINGLEMDS \
4042 lctl get_param seq.srv*MDT0000.space
4043 $LFS path2fid $DIR/$tfile
4045 local old_ifs="$IFS"
4047 fid=($($LFS path2fid $DIR/$tfile))
4051 if [[ $seq < $END ]]; then
4052 error "used reserved sequence $seq?"
4054 cleanup || return $?
4056 run_test 68 "be able to reserve specific sequences in FLDB"
4059 local server_version=$(lustre_version_code $SINGLEMDS)
4061 [[ $server_version -lt $(version_code 2.4.2) ]] &&
4062 skip "Need MDS version at least 2.4.2" && return
4064 [[ $server_version -ge $(version_code 2.4.50) ]] &&
4065 [[ $server_version -lt $(version_code 2.5.0) ]] &&
4066 skip "Need MDS version at least 2.5.0" && return
4070 # use OST0000 since it probably has the most creations
4071 local OSTNAME=$(ostname_from_index 0)
4072 local mdtosc_proc1=$(get_mdtosc_proc_path mds1 $OSTNAME)
4073 local last_id=$(do_facet mds1 lctl get_param -n \
4074 osc.$mdtosc_proc1.prealloc_last_id)
4076 # Want to have OST LAST_ID over 1.5 * OST_MAX_PRECREATE to
4077 # verify that the LAST_ID recovery is working properly. If
4078 # not, then the OST will refuse to allow the MDS connect
4079 # because the LAST_ID value is too different from the MDS
4080 #define OST_MAX_PRECREATE=20000
4081 local num_create=$((20000 * 5))
4084 $LFS setstripe -i 0 $DIR/$tdir
4085 createmany -o $DIR/$tdir/$tfile- $num_create ||
4086 error "createmany: failed to create $num_create files: $?"
4087 # delete all of the files with objects on OST0 so the
4088 # filesystem is not inconsistent later on
4089 $LFS find $MOUNT --ost 0 | xargs rm
4091 stop_ost || error "OST0 stop failure"
4092 add ost1 $(mkfs_opts ost1 $(ostdevname 1)) --reformat --replace \
4093 $(ostdevname 1) $(ostvdevname 1) ||
4094 error "reformat and replace $ostdev failed"
4095 start_ost || error "OST0 restart failure"
4096 wait_osc_import_state mds ost FULL
4098 touch $DIR/$tdir/$tfile-last || error "create file after reformat"
4099 local idx=$($LFS getstripe -i $DIR/$tdir/$tfile-last)
4100 [ $idx -ne 0 ] && error "$DIR/$tdir/$tfile-last on $idx not 0" || true
4104 run_test 69 "replace an OST with the same index"
4107 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
4112 start_mdt 1 || error "MDT0 start fail"
4114 start_ost || error "OST0 start fail"
4116 start_mdt 2 || error "MDT1 start fail"
4118 mount_client $MOUNT || error "mount client fails"
4120 mkdir -p $DIR/$tdir || error "create dir fail"
4122 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
4123 error "create remote dir fail"
4125 rm -rf $DIR/$tdir || error "delete dir fail"
4126 cleanup || return $?
4128 run_test 70a "start MDT0, then OST, then MDT1"
4131 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
4134 start_ost || error "OST0 start fail"
4136 start_mdt 1 || error "MDT0 start fail"
4137 start_mdt 2 || error "MDT1 start fail"
4139 mount_client $MOUNT || error "mount client fails"
4141 mkdir -p $DIR/$tdir || error "create dir fail"
4143 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
4144 error "create remote dir fail"
4146 rm -rf $DIR/$tdir || error "delete dir fail"
4148 cleanup || return $?
4150 run_test 70b "start OST, MDT1, MDT0"
4153 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
4156 start_mdt 1 || error "MDT0 start fail"
4157 start_mdt 2 || error "MDT1 start fail"
4158 start_ost || error "OST0 start fail"
4160 mount_client $MOUNT || error "mount client fails"
4161 stop_mdt 1 || error "MDT1 start fail"
4163 local mdc_for_mdt1=$($LCTL dl | grep MDT0000-mdc | awk '{print $4}')
4164 echo "deactivate $mdc_for_mdt1"
4165 $LCTL --device $mdc_for_mdt1 deactivate || return 1
4167 mkdir -p $DIR/$tdir && error "mkdir succeed"
4169 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir &&
4170 error "create remote dir succeed"
4172 cleanup || return $?
4174 run_test 70c "stop MDT0, mkdir fail, create remote dir fail"
4177 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
4180 start_mdt 1 || error "MDT0 start fail"
4181 start_mdt 2 || error "MDT1 start fail"
4182 start_ost || error "OST0 start fail"
4184 mount_client $MOUNT || error "mount client fails"
4186 stop_mdt 2 || error "MDT1 start fail"
4188 local mdc_for_mdt2=$($LCTL dl | grep MDT0001-mdc |
4190 echo "deactivate $mdc_for_mdt2"
4191 $LCTL --device $mdc_for_mdt2 deactivate ||
4192 error "set $mdc_for_mdt2 deactivate failed"
4194 mkdir -p $DIR/$tdir || error "mkdir fail"
4195 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir &&
4196 error "create remote dir succeed"
4198 rm -rf $DIR/$tdir || error "delete dir fail"
4200 cleanup || return $?
4202 run_test 70d "stop MDT1, mkdir succeed, create remote dir fail"
4205 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
4206 if combined_mgs_mds; then
4207 skip "needs separate MGS/MDT" && return
4211 start_mdt 1 || error "MDT0 start fail"
4212 start_ost || error "OST0 start fail"
4213 start_mdt 2 || error "MDT1 start fail"
4214 start_ost2 || error "OST1 start fail"
4216 mount_client $MOUNT || error "mount client fails"
4218 mkdir -p $DIR/$tdir || error "mkdir fail"
4219 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
4220 error "create remote dir succeed"
4222 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
4223 rm -rf $DIR/$tdir || error "delete dir fail"
4225 umount_client $MOUNT
4226 stop_mdt 1 || error "MDT0 stop fail"
4227 stop_mdt 2 || error "MDT1 stop fail"
4228 stop_ost || error "OST0 stop fail"
4229 stop_ost2 || error "OST1 stop fail"
4231 run_test 71a "start MDT0 OST0, MDT1, OST1"
4234 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
4235 if combined_mgs_mds; then
4236 skip "needs separate MGS/MDT" && return
4240 start_mdt 2 || error "MDT1 start fail"
4241 start_ost || error "OST0 start fail"
4242 start_mdt 1 || error "MDT0 start fail"
4243 start_ost2 || error "OST1 start fail"
4245 mount_client $MOUNT || error "mount client fails"
4247 mkdir -p $DIR/$tdir || error "mkdir fail"
4248 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
4249 error "create remote dir succeed"
4251 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
4252 rm -rf $DIR/$tdir || error "delete dir fail"
4254 umount_client $MOUNT
4255 stop_mdt 1 || error "MDT0 stop fail"
4256 stop_mdt 2 || error "MDT1 stop fail"
4257 stop_ost || error "OST0 stop fail"
4258 stop_ost2 || error "OST1 stop fail"
4260 run_test 71b "start MDT1, OST0, MDT0, OST1"
4263 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
4264 if combined_mgs_mds; then
4265 skip "needs separate MGS/MDT" && return
4269 start_ost || error "OST0 start fail"
4270 start_ost2 || error "OST1 start fail"
4271 start_mdt 2 || error "MDT1 start fail"
4272 start_mdt 1 || error "MDT0 start fail"
4274 mount_client $MOUNT || error "mount client fails"
4276 mkdir -p $DIR/$tdir || error "mkdir fail"
4277 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
4278 error "create remote dir succeed"
4280 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
4281 rm -rf $DIR/$tdir || error "delete dir fail"
4283 umount_client $MOUNT
4284 stop_mdt 1 || error "MDT0 stop fail"
4285 stop_mdt 2 || error "MDT1 stop fail"
4286 stop_ost || error "OST0 stop fail"
4287 stop_ost2 || error "OST1 stop fail"
4290 run_test 71c "start OST0, OST1, MDT1, MDT0"
4293 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
4294 if combined_mgs_mds; then
4295 skip "needs separate MGS/MDT" && return
4299 start_ost || error "OST0 start fail"
4300 start_mdt 2 || error "MDT0 start fail"
4301 start_mdt 1 || error "MDT0 start fail"
4302 start_ost2 || error "OST1 start fail"
4304 mount_client $MOUNT || error "mount client fails"
4306 mkdir -p $DIR/$tdir || error "mkdir fail"
4307 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
4308 error "create remote dir succeed"
4310 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
4311 rm -rf $DIR/$tdir || error "delete dir fail"
4313 umount_client $MOUNT
4314 stop_mdt 1 || error "MDT0 stop fail"
4315 stop_mdt 2 || error "MDT1 stop fail"
4316 stop_ost || error "OST0 stop fail"
4317 stop_ost2 || error "OST1 stop fail"
4320 run_test 71d "start OST0, MDT1, MDT0, OST1"
4323 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
4324 if combined_mgs_mds; then
4325 skip "needs separate MGS/MDT" && return
4329 start_ost || error "OST0 start fail"
4330 start_mdt 2 || error "MDT1 start fail"
4331 start_ost2 || error "OST1 start fail"
4332 start_mdt 1 || error "MDT0 start fail"
4334 mount_client $MOUNT || error "mount client fails"
4336 mkdir -p $DIR/$tdir || error "mkdir fail"
4337 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
4338 error "create remote dir succeed"
4340 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
4341 rm -rf $DIR/$tdir || error "delete dir fail"
4343 umount_client $MOUNT
4344 stop_mdt 1 || error "MDT0 stop fail"
4345 stop_mdt 2 || error "MDT1 stop fail"
4346 stop_ost || error "OST0 stop fail"
4347 stop_ost2 || error "OST1 stop fail"
4350 run_test 71e "start OST0, MDT1, OST1, MDT0"
4352 test_72() { #LU-2634
4353 local mdsdev=$(mdsdevname 1)
4354 local ostdev=$(ostdevname 1)
4355 local cmd="$E2FSCK -fnvd $mdsdev"
4358 [ "$(facet_fstype $SINGLEMDS)" != "ldiskfs" ] &&
4359 skip "ldiskfs only test" && return
4361 #tune MDT with "-O extents"
4363 for num in $(seq $MDSCOUNT); do
4364 add mds${num} $(mkfs_opts mds$num $(mdsdevname $num)) \
4365 --reformat $(mdsdevname $num) $(mdsvdevname $num) ||
4366 error "add mds $num failed"
4367 $TUNE2FS -O extents $(mdsdevname $num)
4370 add ost1 $(mkfs_opts ost1 $ostdev) --reformat $ostdev ||
4371 error "add $ostdev failed"
4372 start_mgsmds || error "start mds failed"
4373 start_ost || error "start ost failed"
4374 mount_client $MOUNT || error "mount client failed"
4376 #create some short symlinks
4378 createmany -o $DIR/$tdir/$tfile-%d $fn
4379 echo "create $fn short symlinks"
4380 for i in $(seq -w 1 $fn); do
4381 ln -s $DIR/$tdir/$tfile-$i $MOUNT/$tfile-$i
4386 umount_client $MOUNT || error "umount client failed"
4387 stop_mds || error "stop mds failed"
4388 stop_ost || error "stop ost failed"
4391 run_e2fsck $(facet_active_host $SINGLEMDS) $mdsdev "-n"
4393 run_test 72 "test fast symlink with extents flag enabled"
4395 test_73() { #LU-3006
4397 [ $(facet_fstype ost1) == zfs ] && import_zpool ost1
4398 do_facet ost1 "$TUNEFS --failnode=1.2.3.4@$NETTYPE $(ostdevname 1)" ||
4399 error "1st tunefs failed"
4400 start_mgsmds || error "start mds failed"
4401 start_ost || error "start ost failed"
4402 mount_client $MOUNT || error "mount client failed"
4403 lctl get_param -n osc.*OST0000-osc-[^M]*.import | grep failover_nids |
4404 grep 1.2.3.4@$NETTYPE || error "failover nids haven't changed"
4405 umount_client $MOUNT || error "umount client failed"
4409 run_test 73 "failnode to update from mountdata properly"
4411 test_75() { # LU-2374
4412 [[ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.4.1) ]] &&
4413 skip "Need MDS version at least 2.4.1" && return
4416 local opts_mds="$(mkfs_opts mds1 $(mdsdevname 1)) \
4417 --reformat $(mdsdevname 1) $(mdsvdevname 1)"
4418 local opts_ost="$(mkfs_opts ost1 $(ostdevname 1)) \
4419 --reformat $(ostdevname 1) $(ostvdevname 1)"
4421 #check with default parameters
4422 add mds1 $opts_mds || error "add mds1 failed for default params"
4423 add ost1 $opts_ost || error "add ost1 failed for default params"
4425 opts_mds=$(echo $opts_mds | sed -e "s/--mdt//")
4426 opts_mds=$(echo $opts_mds |
4427 sed -e "s/--index=$index/--index=$index --mdt/")
4428 opts_ost=$(echo $opts_ost | sed -e "s/--ost//")
4429 opts_ost=$(echo $opts_ost |
4430 sed -e "s/--index=$index/--index=$index --ost/")
4432 add mds1 $opts_mds || error "add mds1 failed for new params"
4433 add ost1 $opts_ost || error "add ost1 failed for new params"
4436 run_test 75 "The order of --index should be irrelevant"
4439 [[ $(lustre_version_code mgs) -ge $(version_code 2.4.52) ]] ||
4440 { skip "Need MDS version at least 2.4.52" && return 0; }
4442 local MDMB_PARAM="osc.*.max_dirty_mb"
4443 echo "Change MGS params"
4444 local MAX_DIRTY_MB=$($LCTL get_param -n $MDMB_PARAM |
4446 echo "max_dirty_mb: $MAX_DIRTY_MB"
4447 local NEW_MAX_DIRTY_MB=$((MAX_DIRTY_MB + MAX_DIRTY_MB))
4448 echo "new_max_dirty_mb: $NEW_MAX_DIRTY_MB"
4449 do_facet mgs $LCTL set_param -P $MDMB_PARAM=$NEW_MAX_DIRTY_MB
4450 wait_update $HOSTNAME "lctl get_param -n $MDMB_PARAM |
4451 head -1" $NEW_MAX_DIRTY_MB
4452 MAX_DIRTY_MB=$($LCTL get_param -n $MDMB_PARAM | head -1)
4453 echo "$MAX_DIRTY_MB"
4454 [ $MAX_DIRTY_MB = $NEW_MAX_DIRTY_MB ] ||
4455 error "error while apply max_dirty_mb"
4457 echo "Check the value is stored after remount"
4460 wait_update $HOSTNAME "lctl get_param -n $MDMB_PARAM |
4461 head -1" $NEW_MAX_DIRTY_MB
4462 MAX_DIRTY_MB=$($LCTL get_param -n $MDMB_PARAM | head -1)
4463 [ $MAX_DIRTY_MB = $NEW_MAX_DIRTY_MB ] ||
4464 error "max_dirty_mb is not saved after remount"
4466 echo "Change OST params"
4467 CLIENT_PARAM="obdfilter.*.client_cache_count"
4468 local CLIENT_CACHE_COUNT
4469 CLIENT_CACHE_COUNT=$(do_facet ost1 $LCTL get_param -n $CLIENT_PARAM |
4471 echo "client_cache_count: $CLIENT_CACHE_COUNT"
4472 NEW_CLIENT_CACHE_COUNT=$((CLIENT_CACHE_COUNT+CLIENT_CACHE_COUNT))
4473 echo "new_client_cache_count: $NEW_CLIENT_CACHE_COUNT"
4474 do_facet mgs $LCTL set_param -P $CLIENT_PARAM=$NEW_CLIENT_CACHE_COUNT
4475 wait_update $(facet_host ost1) "lctl get_param -n $CLIENT_PARAM |
4476 head -1" $NEW_CLIENT_CACHE_COUNT
4477 CLIENT_CACHE_COUNT=$(do_facet ost1 $LCTL get_param -n $CLIENT_PARAM |
4479 echo "$CLIENT_CACHE_COUNT"
4480 [ $CLIENT_CACHE_COUNT = $NEW_CLIENT_CACHE_COUNT ] ||
4481 error "error while apply client_cache_count"
4483 echo "Check the value is stored after remount"
4486 wait_update $(facet_host ost1) "lctl get_param -n $CLIENT_PARAM |
4487 head -1" $NEW_CLIENT_CACHE_COUNT
4488 CLIENT_CACHE_COUNT=$(do_facet ost1 $LCTL get_param -n $CLIENT_PARAM |
4490 echo "$CLIENT_CACHE_COUNT"
4491 [ $CLIENT_CACHE_COUNT = $NEW_CLIENT_CACHE_COUNT ] ||
4492 error "client_cache_count is not saved after remount"
4495 run_test 76a "set permanent params set_param -P"
4497 test_76b() { # LU-4783
4498 [[ $(lustre_version_code mgs) -ge $(version_code 2.5.57) ]] ||
4499 { skip "Need MGS version at least 2.5.57" && return 0; }
4502 do_facet mgs $LCTL get_param mgs.MGS.live.params ||
4503 error "start params log failed"
4506 run_test 76b "verify params log setup correctly"
4508 test_77() { # LU-3445
4509 local server_version=$(lustre_version_code $SINGLEMDS)
4511 [[ $server_version -ge $(version_code 2.2.60) ]] &&
4512 [[ $server_version -le $(version_code 2.4.0) ]] &&
4513 skip "Need MDS version < 2.2.60 or > 2.4.0" && return
4515 if [[ -z "$fs2ost_DEV" || -z "$fs2mds_DEV" ]]; then
4516 is_blkdev $SINGLEMDS $(mdsdevname ${SINGLEMDS//mds/}) &&
4517 skip_env "mixed loopback and real device not working" && return
4520 local fs2mdsdev=$(mdsdevname 1_2)
4521 local fs2ostdev=$(ostdevname 1_2)
4522 local fs2mdsvdev=$(mdsvdevname 1_2)
4523 local fs2ostvdev=$(ostvdevname 1_2)
4524 local fsname=test1234
4526 local failnid="$(h2$NETTYPE 1.2.3.4),$(h2$NETTYPE 4.3.2.1)"
4528 add fs2mds $(mkfs_opts mds1 $fs2mdsdev) --mgs --fsname=$fsname \
4529 --reformat $fs2mdsdev $fs2mdsvdev || error "add fs2mds failed"
4530 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS && trap cleanup_fs2 EXIT INT ||
4531 error "start fs2mds failed"
4533 mgsnid=$(do_facet fs2mds $LCTL list_nids | xargs | tr ' ' ,)
4534 [[ $mgsnid = *,* ]] || mgsnid+=",$mgsnid"
4536 add fs2ost $(mkfs_opts ost1 $fs2ostdev) --mgsnode=$mgsnid \
4537 --failnode=$failnid --fsname=$fsname \
4538 --reformat $fs2ostdev $fs2ostvdev ||
4539 error "add fs2ost failed"
4540 start fs2ost $fs2ostdev $OST_MOUNT_OPTS || error "start fs2ost failed"
4543 $MOUNT_CMD $mgsnid:/$fsname $MOUNT2 || error "mount $MOUNT2 failed"
4544 DIR=$MOUNT2 MOUNT=$MOUNT2 check_mount || error "check $MOUNT2 failed"
4547 run_test 77 "comma-separated MGS NIDs and failover node NIDs"
4550 [[ $(facet_fstype $SINGLEMDS) != ldiskfs ||
4551 $(facet_fstype ost1) != ldiskfs ]] &&
4552 skip "only applicable to ldiskfs-based MDTs and OSTs" && return
4554 # reformat the Lustre filesystem with a smaller size
4555 local saved_MDSSIZE=$MDSSIZE
4556 local saved_OSTSIZE=$OSTSIZE
4557 MDSSIZE=$((MDSSIZE - 20000))
4558 OSTSIZE=$((OSTSIZE - 20000))
4559 reformat || error "(1) reformat Lustre filesystem failed"
4560 MDSSIZE=$saved_MDSSIZE
4561 OSTSIZE=$saved_OSTSIZE
4563 # mount the Lustre filesystem
4564 setup_noconfig || error "(2) setup Lustre filesystem failed"
4567 log "create test files"
4571 mkdir -p $MOUNT/$tdir || error "(3) mkdir $MOUNT/$tdir failed"
4572 for i in $(seq $num_files); do
4573 file=$MOUNT/$tdir/$tfile-$i
4574 dd if=/dev/urandom of=$file count=1 bs=1M ||
4575 error "(4) create $file failed"
4578 # unmount the Lustre filesystem
4579 cleanup || error "(5) cleanup Lustre filesystem failed"
4581 # run e2fsck on the MDT and OST devices
4582 local mds_host=$(facet_active_host $SINGLEMDS)
4583 local ost_host=$(facet_active_host ost1)
4584 local mds_dev=$(mdsdevname ${SINGLEMDS//mds/})
4585 local ost_dev=$(ostdevname 1)
4587 run_e2fsck $mds_host $mds_dev "-y"
4588 run_e2fsck $ost_host $ost_dev "-y"
4590 # get the original block count of the MDT and OST filesystems
4591 local mds_orig_blks=$(get_block_count $SINGLEMDS $mds_dev)
4592 local ost_orig_blks=$(get_block_count ost1 $ost_dev)
4594 # expand the MDT and OST filesystems to the device size
4595 run_resize2fs $SINGLEMDS $mds_dev "" || error "expand $SINGLEMDS failed"
4596 run_resize2fs ost1 $ost_dev "" || error "expand ost1 failed"
4598 # run e2fsck on the MDT and OST devices again
4599 run_e2fsck $mds_host $mds_dev "-y"
4600 run_e2fsck $ost_host $ost_dev "-y"
4602 # mount the Lustre filesystem
4606 log "check files after expanding the MDT and OST filesystems"
4607 for i in $(seq $num_files); do
4608 file=$MOUNT/$tdir/$tfile-$i
4609 $CHECKSTAT -t file -s 1048576 $file ||
4610 error "(6) checkstat $file failed"
4614 log "create more files after expanding the MDT and OST filesystems"
4615 for i in $(seq $((num_files + 1)) $((num_files + 10))); do
4616 file=$MOUNT/$tdir/$tfile-$i
4617 dd if=/dev/urandom of=$file count=1 bs=1M ||
4618 error "(7) create $file failed"
4621 # unmount the Lustre filesystem
4622 cleanup || error "(8) cleanup Lustre filesystem failed"
4624 # run e2fsck on the MDT and OST devices
4625 run_e2fsck $mds_host $mds_dev "-y"
4626 run_e2fsck $ost_host $ost_dev "-y"
4628 # get the maximum block count of the MDT and OST filesystems
4629 local mds_max_blks=$(get_block_count $SINGLEMDS $mds_dev)
4630 local ost_max_blks=$(get_block_count ost1 $ost_dev)
4632 # get the minimum block count of the MDT and OST filesystems
4633 local mds_min_blks=$(run_resize2fs $SINGLEMDS $mds_dev "" "-P" 2>&1 |
4634 grep minimum | sed -e 's/^.*filesystem: //g')
4635 local ost_min_blks=$(run_resize2fs ost1 $ost_dev "" "-P" 2>&1 |
4636 grep minimum | sed -e 's/^.*filesystem: //g')
4638 # shrink the MDT and OST filesystems to a smaller size
4642 if [[ $mds_max_blks -gt $mds_min_blks &&
4643 $mds_max_blks -gt $mds_orig_blks ]]; then
4644 [[ $mds_orig_blks -gt $mds_min_blks ]] &&
4645 base_blks=$mds_orig_blks || base_blks=$mds_min_blks
4646 new_blks=$(( (mds_max_blks - base_blks) / 2 + base_blks ))
4647 run_resize2fs $SINGLEMDS $mds_dev $new_blks ||
4648 error "shrink $SINGLEMDS to $new_blks failed"
4652 if [[ $ost_max_blks -gt $ost_min_blks &&
4653 $ost_max_blks -gt $ost_orig_blks ]]; then
4654 [[ $ost_orig_blks -gt $ost_min_blks ]] &&
4655 base_blks=$ost_orig_blks || base_blks=$ost_min_blks
4656 new_blks=$(( (ost_max_blks - base_blks) / 2 + base_blks ))
4657 run_resize2fs ost1 $ost_dev $new_blks ||
4658 error "shrink ost1 to $new_blks failed"
4662 # check whether the MDT or OST filesystem was shrunk or not
4664 combined_mgs_mds || stop_mgs || error "(9) stop mgs failed"
4665 reformat || error "(10) reformat Lustre filesystem failed"
4669 # run e2fsck on the MDT and OST devices again
4670 run_e2fsck $mds_host $mds_dev "-y"
4671 run_e2fsck $ost_host $ost_dev "-y"
4673 # mount the Lustre filesystem again
4677 log "check files after shrinking the MDT and OST filesystems"
4678 for i in $(seq $((num_files + 10))); do
4679 file=$MOUNT/$tdir/$tfile-$i
4680 $CHECKSTAT -t file -s 1048576 $file ||
4681 error "(11) checkstat $file failed"
4684 # unmount and reformat the Lustre filesystem
4685 cleanup || error "(12) cleanup Lustre filesystem failed"
4686 combined_mgs_mds || stop_mgs || error "(13) stop mgs failed"
4687 reformat || error "(14) reformat Lustre filesystem failed"
4689 run_test 78 "run resize2fs on MDT and OST filesystems"
4691 if ! combined_mgs_mds ; then
4697 # restore the values of MDSSIZE and OSTSIZE
4698 MDSSIZE=$STORED_MDSSIZE
4699 OSTSIZE=$STORED_OSTSIZE