3 # FIXME - there is no reason to use all of these different
4 # return codes, espcially when most of them are mapped to something
5 # else anyway. The combination of test number and return code
6 # figure out what failed.
12 # bug number for skipped test: LU-2828
13 ALWAYS_EXCEPT="$CONF_SANITY_EXCEPT 59 64"
14 # UPDATE THE COMMENT ABOVE WITH BUG NUMBERS WHEN CHANGING ALWAYS_EXCEPT!
18 if [ -r /etc/SuSE-release ]
20 local vers=`grep VERSION /etc/SuSE-release | awk '{print $3}'`
21 local patchlev=`grep PATCHLEVEL /etc/SuSE-release \
23 if [ $vers -eq 11 ] && [ $patchlev -eq 2 ]
31 if is_sles11; then # LU-2181
32 ALWAYS_EXCEPT="$ALWAYS_EXCEPT 23a 34b"
35 if [ "$FAILURE_MODE" = "HARD" ]; then
36 CONFIG_EXCEPTIONS="24a " && \
37 echo "Except the tests: $CONFIG_EXCEPTIONS for FAILURE_MODE=$FAILURE_MODE, bug 23573" && \
38 ALWAYS_EXCEPT="$ALWAYS_EXCEPT $CONFIG_EXCEPTIONS"
41 # bug number for skipped test:
42 # a tool to create lustre filesystem images
43 ALWAYS_EXCEPT="32newtarball $ALWAYS_EXCEPT"
46 PATH=$PWD/$SRCDIR:$SRCDIR:$SRCDIR/../utils:$PATH
48 PTLDEBUG=${PTLDEBUG:--1}
50 LUSTRE=${LUSTRE:-`dirname $0`/..}
51 RLUSTRE=${RLUSTRE:-$LUSTRE}
52 LUSTRE_TESTS_API_DIR=${LUSTRE_TESTS_API_DIR:-${LUSTRE}/tests/clientapi}
53 export MULTIOP=${MULTIOP:-multiop}
55 . $LUSTRE/tests/test-framework.sh
58 # use small MDS + OST size to speed formatting time
59 # do not use too small MDSSIZE/OSTSIZE, which affect the default jouranl size
62 . ${CONFIG:=$LUSTRE/tests/cfg/$NAME.sh}
64 if ! combined_mgs_mds; then
65 # bug number for skipped test: 23954
66 ALWAYS_EXCEPT="$ALWAYS_EXCEPT 24b"
69 # STORED_MDSSIZE is used in test_18
70 if [ -n "$MDSSIZE" ]; then
71 STORED_MDSSIZE=$MDSSIZE
74 # pass "-E lazy_itable_init" to mke2fs to speed up the formatting time
75 if [[ "$LDISKFS_MKFS_OPTS" != *lazy_itable_init* ]]; then
76 LDISKFS_MKFS_OPTS=$(csa_add "$LDISKFS_MKFS_OPTS" -E lazy_itable_init)
79 [ $(facet_fstype $SINGLEMDS) = "zfs" ] &&
80 # bug number for skipped test: LU-2778 LU-2059 LU-4444
81 ALWAYS_EXCEPT="$ALWAYS_EXCEPT 57b 50h 69"
86 require_dsh_mds || exit 0
87 require_dsh_ost || exit 0
89 [ "$SLOW" = "no" ] && EXCEPT_SLOW="30a 31 45 69"
95 # The MGS must be started before the OSTs for a new fs, so start
96 # and stop to generate the startup logs.
99 wait_osc_import_state mds ost FULL
104 reformat_and_config() {
106 if ! combined_mgs_mds ; then
112 writeconf_or_reformat() {
113 # There are at most 2 OSTs for write_conf test
114 # who knows if/where $TUNEFS is installed?
115 # Better reformat if it fails...
116 writeconf_all $MDSCOUNT 2 ||
117 { echo "tunefs failed, reformatting instead" &&
118 reformat_and_config && return 0; }
128 start mgs $MGSDEV $MGS_MOUNT_OPTS
134 local dev=$(mdsdevname $num)
137 echo "start mds service on `facet_active_host $facet`"
138 start $facet ${dev} $MDS_MOUNT_OPTS $@ || return 94
144 local dev=$(mdsdevname $num)
147 echo "stop mds service on `facet_active_host $facet`"
148 # These tests all use non-failover stop
149 stop $facet -f || return 97
155 for num in $(seq $MDSCOUNT); do
156 start_mdt $num $@ || return 94
161 if ! combined_mgs_mds ; then
169 for num in $(seq $MDSCOUNT); do
170 stop_mdt $num || return 97
175 echo "stop mgs service on `facet_active_host mgs`"
176 # These tests all use non-failover stop
177 stop mgs -f || return 97
181 echo "start ost1 service on `facet_active_host ost1`"
182 start ost1 `ostdevname 1` $OST_MOUNT_OPTS $@ || return 95
186 echo "stop ost1 service on `facet_active_host ost1`"
187 # These tests all use non-failover stop
188 stop ost1 -f || return 98
192 echo "start ost2 service on `facet_active_host ost2`"
193 start ost2 `ostdevname 2` $OST_MOUNT_OPTS $@ || return 92
197 echo "stop ost2 service on `facet_active_host ost2`"
198 # These tests all use non-failover stop
199 stop ost2 -f || return 93
204 echo "mount $FSNAME on ${MOUNTPATH}....."
205 zconf_mount `hostname` $MOUNTPATH || return 96
209 local mountopt="-o remount,$1"
211 echo "remount '$1' lustre on ${MOUNTPATH}....."
212 zconf_mount `hostname` $MOUNTPATH "$mountopt" || return 96
217 echo "umount lustre on ${MOUNTPATH}....."
218 zconf_umount `hostname` $MOUNTPATH || return 97
221 manual_umount_client(){
224 echo "manual umount lustre on ${MOUNT}...."
225 do_facet client "umount -d ${FORCE} $MOUNT"
231 start_mds || error "MDT start failed"
232 start_ost || error "OST start failed"
233 mount_client $MOUNT || error "client start failed"
234 client_up || error "client_up failed"
238 if ! combined_mgs_mds ; then
247 unload_modules_conf () {
248 if combined_mgs_mds || ! local_mode; then
249 unload_modules || return 1
254 stop_ost || return 202
255 stop_mds || return 201
256 unload_modules_conf || return 203
260 umount_client $MOUNT || return 200
261 cleanup_nocli || return $?
265 do_facet client "cp /etc/passwd $DIR/a" || return 71
266 do_facet client "rm $DIR/a" || return 72
267 # make sure lustre is actually mounted (touch will block,
268 # but grep won't, so do it after)
269 do_facet client "grep $MOUNT' ' /proc/mounts > /dev/null" || return 73
270 echo "setup single mount lustre success"
274 do_facet client "touch $DIR/a" || return 71
275 do_facet client "rm $DIR/a" || return 72
276 do_facet client "touch $DIR2/a" || return 73
277 do_facet client "rm $DIR2/a" || return 74
278 echo "setup double mount lustre success"
283 if [ "$ONLY" == "setup" ]; then
288 if [ "$ONLY" == "cleanup" ]; then
295 #create single point mountpoint
301 check_mount || return 41
304 run_test 0 "single mount setup"
307 start_mds || error "MDT start failed"
309 echo "start ost second time..."
310 start_ost && error "2nd OST start should fail"
311 mount_client $MOUNT || error "client start failed"
312 check_mount || return 42
315 run_test 1 "start up ost twice (should return errors)"
319 echo "start mds second time.."
320 start_mdt 1 && error "2nd MDT start should fail"
323 check_mount || return 43
326 run_test 2 "start up mds twice (should return err)"
330 #mount.lustre returns an error if already in mtab
331 mount_client $MOUNT && error "2nd client mount should fail"
332 check_mount || return 44
335 run_test 3 "mount client twice (should return err)"
339 touch $DIR/$tfile || return 85
343 # ok for ost to fail shutdown
344 if [ 202 -ne $eno ]; then
349 run_test 4 "force cleanup ost, then cleanup"
351 test_5a() { # was test_5
353 touch $DIR/$tfile || return 1
354 fuser -m -v $MOUNT && echo "$MOUNT is in use by user space process."
356 stop_mds -f || return 2
358 # cleanup may return an error from the failed
359 # disconnects; for now I'll consider this successful
360 # if all the modules have unloaded.
364 echo "killing umount"
365 kill -TERM $UMOUNT_PID
366 echo "waiting for umount to finish"
368 if grep " $MOUNT " /proc/mounts; then
369 echo "test 5: /proc/mounts after failed umount"
373 echo "killing umount"
374 kill -TERM $UMOUNT_PID
375 echo "waiting for umount to finish"
377 grep " $MOUNT " /proc/mounts && echo "test 5: /proc/mounts after second umount" && return 11
381 # stop_mds is a no-op here, and should not fail
382 cleanup_nocli || return $?
383 # df may have lingering entry
385 # mtab may have lingering entry
389 while [ "$WAIT" -ne "$MAX_WAIT" ]; do
391 grep -q $MOUNT" " /etc/mtab || break
392 echo "Waiting /etc/mtab updated ... "
393 WAIT=$(( WAIT + sleep))
395 [ "$WAIT" -eq "$MAX_WAIT" ] && error "/etc/mtab is not updated in $WAIT secs"
396 echo "/etc/mtab updated in $WAIT secs"
398 run_test 5a "force cleanup mds, then cleanup"
406 grep " $MOUNT " /etc/mtab && \
407 error false "unexpected entry in mtab before mount" && return 10
411 if ! combined_mgs_mds ; then
412 trap cleanup_5b EXIT ERR
417 [ -d $MOUNT ] || mkdir -p $MOUNT
418 mount_client $MOUNT && rc=1
419 grep " $MOUNT " /etc/mtab && \
420 error "$MOUNT entry in mtab after failed mount" && rc=11
422 # stop_mds is a no-op here, and should not fail
423 cleanup_nocli || rc=$?
424 if ! combined_mgs_mds ; then
429 run_test 5b "Try to start a client with no MGS (should return errs)"
432 grep " $MOUNT " /etc/mtab && \
433 error false "unexpected entry in mtab before mount" && return 10
438 [ -d $MOUNT ] || mkdir -p $MOUNT
439 local oldfs="${FSNAME}"
440 FSNAME="wrong.${FSNAME}"
441 mount_client $MOUNT || :
443 grep " $MOUNT " /etc/mtab && \
444 error "$MOUNT entry in mtab after failed mount" && rc=11
446 cleanup_nocli || rc=$?
449 run_test 5c "cleanup after failed mount (bug 2712) (should return errs)"
452 grep " $MOUNT " /etc/mtab && \
453 error false "unexpected entry in mtab before mount" && return 10
455 [ "$(facet_fstype ost1)" = "zfs" ] &&
456 skip "LU-2059: no local config for ZFS OSTs" && return
462 mount_client $MOUNT || rc=1
464 grep " $MOUNT " /etc/mtab && \
465 error "$MOUNT entry in mtab after unmount" && rc=11
468 run_test 5d "mount with ost down"
471 grep " $MOUNT " /etc/mtab && \
472 error false "unexpected entry in mtab before mount" && return 10
478 #define OBD_FAIL_PTLRPC_DELAY_SEND 0x506
479 do_facet client "lctl set_param fail_loc=0x80000506"
480 mount_client $MOUNT || echo "mount failed (not fatal)"
482 grep " $MOUNT " /etc/mtab && \
483 error "$MOUNT entry in mtab after unmount" && rc=11
486 run_test 5e "delayed connect, don't crash (bug 10268)"
489 if combined_mgs_mds ; then
490 skip "combined mgs and mds"
494 grep " $MOUNT " /etc/mtab && \
495 error false "unexpected entry in mtab before mount" && return 10
499 [ -d $MOUNT ] || mkdir -p $MOUNT
500 mount_client $MOUNT &
502 echo client_mount pid is $pid
506 if ! ps -f -p $pid >/dev/null; then
509 grep " $MOUNT " /etc/mtab && echo "test 5f: mtab after mount"
510 error "mount returns $rc, expected to hang"
519 # mount should succeed after start mds
522 [ $rc -eq 0 ] || error "mount returned $rc"
523 grep " $MOUNT " /etc/mtab && echo "test 5f: mtab after mount"
527 run_test 5f "mds down, cleanup after failed mount (bug 2712)"
532 mount_client ${MOUNT} || return 87
533 touch $DIR/a || return 86
536 run_test 6 "manual umount, then mount again"
541 cleanup_nocli || return $?
543 run_test 7 "manual umount, then cleanup"
548 check_mount2 || return 45
549 umount_client $MOUNT2
552 run_test 8 "double mount setup"
557 do_facet ost1 lctl set_param debug=\'inode trace\' || return 1
558 do_facet ost1 lctl set_param subsystem_debug=\'mds ost\' || return 1
560 CHECK_PTLDEBUG="`do_facet ost1 lctl get_param -n debug`"
561 if [ "$CHECK_PTLDEBUG" ] && { \
562 [ "$CHECK_PTLDEBUG" = "trace inode warning error emerg console" ] ||
563 [ "$CHECK_PTLDEBUG" = "trace inode" ]; }; then
564 echo "lnet.debug success"
566 echo "lnet.debug: want 'trace inode', have '$CHECK_PTLDEBUG'"
569 CHECK_SUBSYS="`do_facet ost1 lctl get_param -n subsystem_debug`"
570 if [ "$CHECK_SUBSYS" ] && [ "$CHECK_SUBSYS" = "mds ost" ]; then
571 echo "lnet.subsystem_debug success"
573 echo "lnet.subsystem_debug: want 'mds ost', have '$CHECK_SUBSYS'"
576 stop_ost || return $?
578 run_test 9 "test ptldebug and subsystem for mkfs"
586 do_facet $facet "test -b $dev" || rc=1
587 if [[ "$size" ]]; then
588 local in=$(do_facet $facet "dd if=$dev of=/dev/null bs=1k count=1 skip=$size 2>&1" |\
589 awk '($3 == "in") { print $1 }')
590 [[ $in = "1+0" ]] || rc=1
596 # Test 16 was to "verify that lustre will correct the mode of OBJECTS".
597 # But with new MDS stack we don't care about the mode of local objects
598 # anymore, so this test is removed. See bug 22944 for more details.
602 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
603 skip "Only applicable to ldiskfs-based MDTs"
608 check_mount || return 41
611 echo "Remove mds config log"
612 if ! combined_mgs_mds ; then
616 do_facet mgs "$DEBUGFS -w -R 'unlink CONFIGS/$FSNAME-MDT0000' \
617 $(mgsdevname) || return \$?" || return $?
619 if ! combined_mgs_mds ; then
624 start_mds && return 42
627 run_test 17 "Verify failed mds_postsetup won't fail assertion (2936) (should return errs)"
630 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
631 skip "Only applicable to ldiskfs-based MDTs"
635 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
640 # check if current MDSSIZE is large enough
641 [ $MDSSIZE -ge $MIN ] && OK=1 && myMDSSIZE=$MDSSIZE && \
642 log "use MDSSIZE=$MDSSIZE"
644 # check if the global config has a large enough MDSSIZE
645 [ -z "$OK" -a ! -z "$STORED_MDSSIZE" ] && [ $STORED_MDSSIZE -ge $MIN ] && \
646 OK=1 && myMDSSIZE=$STORED_MDSSIZE && \
647 log "use STORED_MDSSIZE=$STORED_MDSSIZE"
649 # check if the block device is large enough
650 is_blkdev $SINGLEMDS $MDSDEV $MIN
651 local large_enough=$?
652 if [ -n "$OK" ]; then
653 [ $large_enough -ne 0 ] && OK=""
655 [ $large_enough -eq 0 ] && OK=1 && myMDSSIZE=$MIN &&
656 log "use device $MDSDEV with MIN=$MIN"
659 # check if a loopback device has enough space for fs metadata (5%)
661 if [ -z "$OK" ]; then
662 local SPACE=$(do_facet $SINGLEMDS "[ -f $MDSDEV -o ! -e $MDSDEV ] && df -P \\\$(dirname $MDSDEV)" |
663 awk '($1 != "Filesystem") {print $4}')
664 ! [ -z "$SPACE" ] && [ $SPACE -gt $((MIN / 20)) ] && \
665 OK=1 && myMDSSIZE=$MIN && \
666 log "use file $MDSDEV with MIN=$MIN"
669 [ -z "$OK" ] && skip_env "$MDSDEV too small for ${MIN}kB MDS" && return
672 echo "mount mds with large journal..."
674 local OLD_MDSSIZE=$MDSSIZE
678 echo "mount lustre system..."
680 check_mount || return 41
682 echo "check journal size..."
683 local FOUNDSIZE=$(do_facet $SINGLEMDS "$DEBUGFS -c -R 'stat <8>' $MDSDEV" | awk '/Size: / { print $NF; exit;}')
684 if [ $FOUNDSIZE -gt $((32 * 1024 * 1024)) ]; then
685 log "Success: mkfs creates large journals. Size: $((FOUNDSIZE >> 20))M"
687 error "expected journal size > 32M, found $((FOUNDSIZE >> 20))M"
695 run_test 18 "check mkfs creates large journals"
698 start_mds || return 1
699 stop_mds -f || return 2
701 run_test 19a "start/stop MDS without OSTs"
704 [ "$(facet_fstype ost1)" = "zfs" ] &&
705 skip "LU-2059: no local config for ZFS OSTs" && return
707 start_ost || return 1
708 stop_ost -f || return 2
710 run_test 19b "start/stop OSTs without MDS"
713 # first format the ost/mdt
717 check_mount || return 43
719 remount_client ro $MOUNT || return 44
720 touch $DIR/$tfile && echo "$DIR/$tfile created incorrectly" && return 45
721 [ -e $DIR/$tfile ] && echo "$DIR/$tfile exists incorrectly" && return 46
722 remount_client rw $MOUNT || return 47
724 [ ! -f $DIR/$tfile ] && echo "$DIR/$tfile missing" && return 48
725 MCNT=`grep -c $MOUNT /etc/mtab`
726 [ "$MCNT" -ne 1 ] && echo "$MOUNT in /etc/mtab $MCNT times" && return 49
731 run_test 20 "remount ro,rw mounts work and doesn't break /etc/mtab"
736 wait_osc_import_state mds ost FULL
740 run_test 21a "start mds before ost, stop ost first"
743 [ "$(facet_fstype ost1)" = "zfs" ] &&
744 skip "LU-2059: no local config for ZFS OSTs" && return
748 wait_osc_import_state mds ost FULL
752 run_test 21b "start ost before mds, stop mds first"
758 wait_osc_import_state mds ost2 FULL
762 #writeconf to remove all ost2 traces for subsequent tests
763 writeconf_or_reformat
765 run_test 21c "start mds between two osts, stop mds last"
768 if combined_mgs_mds ; then
769 skip "need separate mgs device" && return 0
779 wait_osc_import_state mds ost2 FULL
785 #writeconf to remove all ost2 traces for subsequent tests
786 writeconf_or_reformat
789 run_test 21d "start mgs then ost and then mds"
794 echo Client mount with ost in logs, but none running
796 # wait until mds connected to ost and open client connection
797 wait_osc_import_state mds ost FULL
800 # check_mount will block trying to contact ost
801 mcreate $DIR/$tfile || return 40
802 rm -f $DIR/$tfile || return 42
806 echo Client mount with a running ost
809 # if gss enabled, wait full time to let connection from
810 # mds to ost be established, due to the mismatch between
811 # initial connect timeout and gss context negotiation timeout.
812 # This perhaps could be remove after AT landed.
813 echo "sleep $((TIMEOUT + TIMEOUT + TIMEOUT))s"
814 sleep $((TIMEOUT + TIMEOUT + TIMEOUT))
817 wait_osc_import_state mds ost FULL
818 wait_osc_import_state client ost FULL
819 check_mount || return 41
824 run_test 22 "start a client before osts (should return errs)"
826 test_23a() { # was test_23
830 # force down client so that recovering mds waits for reconnect
831 local running=$(grep -c $MOUNT /proc/mounts) || true
832 if [ $running -ne 0 ]; then
833 echo "Stopping client $MOUNT (opts: -f)"
837 # enter recovery on mds
839 # try to start a new client
840 mount_client $MOUNT &
842 MOUNT_PID=$(ps -ef | grep "t lustre" | grep -v grep | awk '{print $2}')
843 MOUNT_LUSTRE_PID=`ps -ef | grep mount.lustre | grep -v grep | awk '{print $2}'`
844 echo mount pid is ${MOUNT_PID}, mount.lustre pid is ${MOUNT_LUSTRE_PID}
846 ps --ppid $MOUNT_LUSTRE_PID
847 echo "waiting for mount to finish"
849 # "ctrl-c" sends SIGINT but it usually (in script) does not work on child process
850 # SIGTERM works but it does not spread to offspring processses
851 kill -s TERM $MOUNT_PID
852 kill -s TERM $MOUNT_LUSTRE_PID
853 # we can not wait $MOUNT_PID because it is not a child of this shell
859 while [ "$WAIT" -lt "$MAX_WAIT" ]; do
861 PID1=$(ps -ef | awk '{print $2}' | grep -w $MOUNT_PID)
862 PID2=$(ps -ef | awk '{print $2}' | grep -w $MOUNT_LUSTRE_PID)
865 [ -z "$PID1" -a -z "$PID2" ] && break
866 echo "waiting for mount to finish ... "
867 WAIT=$(( WAIT + sleep))
869 if [ "$WAIT" -eq "$MAX_WAIT" ]; then
870 error "MOUNT_PID $MOUNT_PID and "\
871 "MOUNT_LUSTRE_PID $MOUNT_LUSTRE_PID still not killed in $WAIT secs"
874 stop_mds || error "stopping MDSes failed"
875 stop_ost || error "stopping OSSes failed"
877 run_test 23a "interrupt client during recovery mount delay"
882 test_23b() { # was test_23
885 # Simulate -EINTR during mount OBD_FAIL_LDLM_CLOSE_THREAD
886 lctl set_param fail_loc=0x80000313
890 run_test 23b "Simulate -EINTR during mount"
892 fs2mds_HOST=$mds_HOST
893 fs2ost_HOST=$ost_HOST
895 MDSDEV1_2=$fs2mds_DEV
896 OSTDEV1_2=$fs2ost_DEV
897 OSTDEV2_2=$fs3ost_DEV
901 echo "umount $MOUNT2 ..."
902 umount $MOUNT2 || true
903 echo "stopping fs2mds ..."
904 stop fs2mds -f || true
905 echo "stopping fs2ost ..."
906 stop fs2ost -f || true
910 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
912 if [ -z "$fs2ost_DEV" -o -z "$fs2mds_DEV" ]; then
913 is_blkdev $SINGLEMDS $MDSDEV && \
914 skip_env "mixed loopback and real device not working" && return
917 [ -n "$ost1_HOST" ] && fs2ost_HOST=$ost1_HOST
919 local fs2mdsdev=$(mdsdevname 1_2)
920 local fs2ostdev=$(ostdevname 1_2)
921 local fs2mdsvdev=$(mdsvdevname 1_2)
922 local fs2ostvdev=$(ostvdevname 1_2)
924 # test 8-char fsname as well
925 local FSNAME2=test1234
927 add fs2mds $(mkfs_opts mds1 ${fs2mdsdev} ) --nomgs --mgsnode=$MGSNID \
928 --fsname=${FSNAME2} --reformat $fs2mdsdev $fs2mdsvdev || exit 10
930 add fs2ost $(mkfs_opts ost1 ${fs2ostdev}) --fsname=${FSNAME2} \
931 --reformat $fs2ostdev $fs2ostvdev || exit 10
934 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS && trap cleanup_24a EXIT INT
935 start fs2ost $fs2ostdev $OST_MOUNT_OPTS
937 $MOUNT_CMD $MGSNID:/${FSNAME2} $MOUNT2 || return 1
939 check_mount || return 2
940 # files written on 1 should not show up on 2
941 cp /etc/passwd $DIR/$tfile
943 [ -e $MOUNT2/$tfile ] && error "File bleed" && return 7
946 cp /etc/passwd $MOUNT2/b || return 3
947 rm $MOUNT2/b || return 4
948 # 2 is actually mounted
949 grep $MOUNT2' ' /proc/mounts > /dev/null || return 5
951 facet_failover fs2mds
952 facet_failover fs2ost
955 # the MDS must remain up until last MDT
957 MDS=$(do_facet $SINGLEMDS "lctl get_param -n devices" | awk '($3 ~ "mdt" && $4 ~ "MDT") { print $4 }' | head -1)
958 [ -z "$MDS" ] && error "No MDT" && return 8
960 cleanup_nocli || return 6
962 run_test 24a "Multiple MDTs on a single node"
965 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
967 if [ -z "$fs2mds_DEV" ]; then
968 local dev=${SINGLEMDS}_dev
970 is_blkdev $SINGLEMDS $MDSDEV && \
971 skip_env "mixed loopback and real device not working" && return
974 local fs2mdsdev=$(mdsdevname 1_2)
975 local fs2mdsvdev=$(mdsvdevname 1_2)
977 add fs2mds $(mkfs_opts mds1 ${fs2mdsdev} ) --mgs --fsname=${FSNAME}2 \
978 --reformat $fs2mdsdev $fs2mdsvdev || exit 10
980 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS && return 2
983 run_test 24b "Multiple MGSs on a single node (should return err)"
987 check_mount || return 2
988 local MODULES=$($LCTL modules | awk '{ print $2 }')
989 rmmod $MODULES 2>/dev/null || true
992 run_test 25 "Verify modules are referenced"
996 # we need modules before mount for sysctl, so make sure...
997 do_facet $SINGLEMDS "lsmod | grep -q lustre || modprobe lustre"
998 #define OBD_FAIL_MDS_FS_SETUP 0x135
999 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000135"
1000 start_mds && echo MDS started && return 1
1001 lctl get_param -n devices
1002 DEVS=$(lctl get_param -n devices | egrep -v MG | wc -l)
1003 [ $DEVS -gt 0 ] && return 2
1004 # start mds to drop writeconf setting
1005 start_mds || return 3
1006 stop_mds || return 4
1007 unload_modules_conf || return $?
1009 run_test 26 "MDT startup failure cleans LOV (should return errs)"
1012 [ "$(facet_fstype ost1)" = "zfs" ] &&
1013 skip "LU-2059: no local config for ZFS OSTs" && return
1015 start_ost || return 1
1016 start_mds || return 2
1017 echo "Requeue thread should have started: "
1018 ps -e | grep ll_cfg_requeue
1019 set_conf_param_and_check ost1 \
1020 "lctl get_param -n obdfilter.$FSNAME-OST0000.client_cache_seconds" \
1021 "$FSNAME-OST0000.ost.client_cache_seconds" || return 3
1024 run_test 27a "Reacquire MGS lock if OST started first"
1029 local device=$(do_facet $SINGLEMDS "lctl get_param -n devices" |
1030 awk '($3 ~ "mdt" && $4 ~ "MDT0000") { print $4 }')
1032 facet_failover $SINGLEMDS
1033 set_conf_param_and_check $SINGLEMDS \
1034 "lctl get_param -n mdt.$device.identity_acquire_expire" \
1035 "$device.mdt.identity_acquire_expire" || return 3
1036 set_conf_param_and_check client \
1037 "lctl get_param -n mdc.$device-mdc-*.max_rpcs_in_flight"\
1038 "$device.mdc.max_rpcs_in_flight" || return 4
1042 run_test 27b "Reacquire MGS lock after failover"
1046 TEST="lctl get_param -n llite.$FSNAME-*.max_read_ahead_whole_mb"
1047 PARAM="$FSNAME.llite.max_read_ahead_whole_mb"
1049 FINAL=$(($ORIG + 1))
1050 set_conf_param_and_check client "$TEST" "$PARAM" $FINAL || return 3
1051 FINAL=$(($FINAL + 1))
1052 set_conf_param_and_check client "$TEST" "$PARAM" $FINAL || return 4
1053 umount_client $MOUNT || return 200
1056 if [ $RESULT -ne $FINAL ]; then
1057 echo "New config not seen: wanted $FINAL got $RESULT"
1060 echo "New config success: got $RESULT"
1062 set_conf_param_and_check client "$TEST" "$PARAM" $ORIG || return 5
1065 run_test 28 "permanent parameter setting"
1067 test_28a() { # LU-4221
1068 [ $(lustre_version_code ost1) -eq $(version_code 2.5.0) -o \
1069 $(lustre_version_code ost1) -ge $(version_code 2.5.52) ] ||
1070 { skip "Need OST version >= 2.5.52 or = 2.5.0" &&
1072 [ "$(facet_fstype ost1)" = "zfs" ] &&
1073 skip "LU-4221: no such proc params for ZFS OSTs" && return
1080 local device="$FSNAME-OST0000"
1084 # In this test we will set three kinds of proc parameters with
1086 # 1. the ones moved from the OFD to the OSD, and only their
1087 # symlinks kept in obdfilter
1088 # 2. non-symlink ones in the OFD
1089 # 3. non-symlink ones in the OSD
1092 # prepare a symlink parameter in the OFD
1093 name="writethrough_cache_enable"
1094 param="$device.ost.$name"
1095 cmd="$LCTL get_param -n obdfilter.$device.$name"
1097 # conf_param the symlink parameter in the OFD
1098 old=$(do_facet ost1 $cmd)
1099 new=$(((old + 1) % 2))
1100 set_conf_param_and_check ost1 "$cmd" "$param" $new ||
1101 error "lctl conf_param $device.ost.$param=$new failed"
1103 # conf_param the target parameter in the OSD
1104 param="$device.osd.$name"
1105 cmd="$LCTL get_param -n osd-*.$device.$name"
1106 set_conf_param_and_check ost1 "$cmd" "$param" $old ||
1107 error "lctl conf_param $device.osd.$param=$old failed"
1110 # prepare a non-symlink parameter in the OFD
1111 name="client_cache_seconds"
1112 param="$device.ost.$name"
1113 cmd="$LCTL get_param -n obdfilter.$device.$name"
1115 # conf_param the parameter in the OFD
1116 old=$(do_facet ost1 $cmd)
1118 set_conf_param_and_check ost1 "$cmd" "$param" $new ||
1119 error "lctl conf_param $device.ost.$param=$new failed"
1120 set_conf_param_and_check ost1 "$cmd" "$param" $old ||
1121 error "lctl conf_param $device.ost.$param=$old failed"
1124 # prepare a non-symlink parameter in the OSD
1126 param="$device.osd.$name"
1127 cmd="$LCTL get_param -n osd-*.$device.$name"
1129 # conf_param the parameter in the OSD
1130 old=$(do_facet ost1 $cmd)
1131 new=$(((old + 1) % 2))
1132 set_conf_param_and_check ost1 "$cmd" "$param" $new ||
1133 error "lctl conf_param $device.osd.$param=$new failed"
1134 set_conf_param_and_check ost1 "$cmd" "$param" $old ||
1135 error "lctl conf_param $device.osd.$param=$old failed"
1139 run_test 28a "set symlink parameters permanently with conf_param"
1142 [ "$OSTCOUNT" -lt "2" ] && skip_env "$OSTCOUNT < 2, skipping" && return
1143 setup > /dev/null 2>&1
1147 local PARAM="$FSNAME-OST0001.osc.active"
1148 local PROC_ACT="osc.$FSNAME-OST0001-osc-[^M]*.active"
1149 local PROC_UUID="osc.$FSNAME-OST0001-osc-[^M]*.ost_server_uuid"
1151 ACTV=$(lctl get_param -n $PROC_ACT)
1153 set_conf_param_and_check client \
1154 "lctl get_param -n $PROC_ACT" "$PARAM" $DEAC || return 2
1155 # also check ost_server_uuid status
1156 RESULT=$(lctl get_param -n $PROC_UUID | grep DEACTIV)
1157 if [ -z "$RESULT" ]; then
1158 echo "Live client not deactivated: $(lctl get_param -n $PROC_UUID)"
1161 echo "Live client success: got $RESULT"
1165 for num in $(seq $MDSCOUNT); do
1166 local mdtosc=$(get_mdtosc_proc_path mds${num} $FSNAME-OST0001)
1167 local MPROC="osc.$mdtosc.active"
1172 RESULT=$(do_facet mds${num} " lctl get_param -n $MPROC")
1173 [ ${PIPESTATUS[0]} = 0 ] || error "Can't read $MPROC"
1174 if [ $RESULT -eq $DEAC ]; then
1175 echo -n "MDT deactivated also after"
1176 echo "$WAIT sec (got $RESULT)"
1180 if [ $WAIT -eq $MAX ]; then
1181 echo -n "MDT not deactivated: wanted $DEAC"
1185 echo "Waiting $(($MAX - $WAIT))secs for MDT deactivated"
1188 # test new client starts deactivated
1189 umount_client $MOUNT || return 200
1191 RESULT=$(lctl get_param -n $PROC_UUID | grep DEACTIV | grep NEW)
1192 if [ -z "$RESULT" ]; then
1193 echo "New client not deactivated from start: $(lctl get_param -n $PROC_UUID)"
1196 echo "New client success: got $RESULT"
1199 # make sure it reactivates
1200 set_conf_param_and_check client \
1201 "lctl get_param -n $PROC_ACT" "$PARAM" $ACTV || return 6
1203 umount_client $MOUNT
1206 #writeconf to remove all ost2 traces for subsequent tests
1207 writeconf_or_reformat
1209 run_test 29 "permanently remove an OST"
1214 echo Big config llog
1215 TEST="lctl get_param -n llite.$FSNAME-*.max_read_ahead_whole_mb"
1217 LIST=(1 2 3 4 5 4 3 2 1 2 3 4 5 4 3 2 1 2 3 4 5)
1218 for i in ${LIST[@]}; do
1219 set_conf_param_and_check client "$TEST" \
1220 "$FSNAME.llite.max_read_ahead_whole_mb" $i || return 3
1222 # make sure client restart still works
1223 umount_client $MOUNT
1224 mount_client $MOUNT || return 4
1225 [ "$($TEST)" -ne "$i" ] && error "Param didn't stick across restart $($TEST) != $i"
1228 echo Erase parameter setting
1229 do_facet mgs "$LCTL conf_param -d $FSNAME.llite.max_read_ahead_whole_mb" || return 6
1230 umount_client $MOUNT
1231 mount_client $MOUNT || return 6
1233 echo "deleted (default) value=$FINAL, orig=$ORIG"
1234 # assumes this parameter started at the default value
1235 [ "$FINAL" -eq "$ORIG" ] || fail "Deleted value=$FINAL, orig=$ORIG"
1239 run_test 30a "Big config llog and conf_param deletion"
1244 # Make a fake nid. Use the OST nid, and add 20 to the least significant
1245 # numerical part of it. Hopefully that's not already a failover address for
1247 OSTNID=$(do_facet ost1 "$LCTL get_param nis" | tail -1 | awk '{print $1}')
1248 ORIGVAL=$(echo $OSTNID | egrep -oi "[0-9]*@")
1249 NEWVAL=$((($(echo $ORIGVAL | egrep -oi "[0-9]*") + 20) % 256))
1250 NEW=$(echo $OSTNID | sed "s/$ORIGVAL/$NEWVAL@/")
1251 echo "Using fake nid $NEW"
1253 TEST="$LCTL get_param -n osc.$FSNAME-OST0000-osc-[^M]*.import |
1254 grep failover_nids | sed -n 's/.*\($NEW\).*/\1/p'"
1255 set_conf_param_and_check client "$TEST" \
1256 "$FSNAME-OST0000.failover.node" $NEW ||
1257 error "didn't add failover nid $NEW"
1258 NIDS=$($LCTL get_param -n osc.$FSNAME-OST0000-osc-[^M]*.import |
1261 NIDCOUNT=$(($(echo "$NIDS" | wc -w) - 3))
1262 echo "should have 2 failover nids: $NIDCOUNT"
1263 [ $NIDCOUNT -eq 2 ] || error "Failover nid not added"
1264 do_facet mgs "$LCTL conf_param -d $FSNAME-OST0000.failover.node" ||
1265 error "conf_param delete failed"
1266 umount_client $MOUNT
1267 mount_client $MOUNT || return 3
1269 NIDS=$($LCTL get_param -n osc.$FSNAME-OST0000-osc-[^M]*.import |
1272 NIDCOUNT=$(($(echo "$NIDS" | wc -w) - 3))
1273 echo "only 1 final nid should remain: $NIDCOUNT"
1274 [ $NIDCOUNT -eq 1 ] || error "Failover nids not removed"
1278 run_test 30b "Remove failover nids"
1280 test_31() { # bug 10734
1281 # ipaddr must not exist
1282 $MOUNT_CMD 4.3.2.1@tcp:/lustre $MOUNT || true
1285 run_test 31 "Connect to non-existent node (shouldn't crash)"
1289 T32_BLIMIT=20480 # Kbytes
1293 # This is not really a test but a tool to create new disk
1294 # image tarballs for the upgrade tests.
1296 # Disk image tarballs should be created on single-node
1297 # clusters by running this test with default configurations
1298 # plus a few mandatory environment settings that are verified
1299 # at the beginning of the test.
1301 test_32newtarball() {
1305 local tmp=$TMP/t32_image_create
1307 if [ $FSNAME != t32fs -o $MDSCOUNT -ne 1 -o \
1308 \( -z "$MDSDEV" -a -z "$MDSDEV1" \) -o $OSTCOUNT -ne 1 -o \
1309 -z "$OSTDEV1" ]; then
1310 error "Needs FSNAME=t32fs MDSCOUNT=1 MDSDEV1=<nonexistent_file>" \
1311 "(or MDSDEV, in the case of b1_8) OSTCOUNT=1" \
1312 "OSTDEV1=<nonexistent_file>"
1316 echo "Found stale $tmp"
1321 tar cf - -C $src . | tar xf - -C $tmp/src
1322 dd if=/dev/zero of=$tmp/src/t32_qf_old bs=1M \
1323 count=$(($T32_BLIMIT / 1024 / 2))
1324 chown $T32_QID.$T32_QID $tmp/src/t32_qf_old
1326 # format ost with comma-separated NIDs to verify LU-4460
1327 local failnid="$(h2$NETTYPE 1.2.3.4),$(h2$NETTYPE 4.3.2.1)"
1328 MGSNID="$MGSNID,$MGSNID" OSTOPT="--failnode=$failnid" formatall
1332 [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.3.50) ] &&
1333 $LFS quotacheck -ug /mnt/$FSNAME
1334 $LFS setquota -u $T32_QID -b 0 -B $T32_BLIMIT -i 0 -I $T32_ILIMIT \
1337 tar cf - -C $tmp/src . | tar xf - -C /mnt/$FSNAME
1344 ls -Rni --time-style=+%s >$tmp/img/list
1345 find . ! -name .lustre -type f -exec sha1sum {} \; |
1346 sort -k 2 >$tmp/img/sha1sums
1348 $LCTL get_param -n version | head -n 1 |
1349 sed -e 's/^lustre: *//' >$tmp/img/commit
1351 [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.3.50) ] &&
1352 $LFS quotaon -ug /mnt/$FSNAME
1353 $LFS quota -u $T32_QID -v /mnt/$FSNAME
1354 $LFS quota -v -u $T32_QID /mnt/$FSNAME |
1355 awk 'BEGIN { num='1' } { if ($1 == "'/mnt/$FSNAME'") \
1356 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1357 | tr -d "*" > $tmp/img/bspace
1358 $LFS quota -v -u $T32_QID /mnt/$FSNAME |
1359 awk 'BEGIN { num='5' } { if ($1 == "'/mnt/$FSNAME'") \
1360 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1361 | tr -d "*" > $tmp/img/ispace
1366 find -type f -exec sha1sum {} \; | sort -k 2 >$tmp/sha1sums.src
1369 if ! diff -u $tmp/sha1sums.src $tmp/img/sha1sums; then
1370 echo "Data verification failed"
1373 uname -r >$tmp/img/kernel
1374 uname -m >$tmp/img/arch
1376 mv ${MDSDEV1:-$MDSDEV} $tmp/img
1377 mv $OSTDEV1 $tmp/img
1379 version=$(sed -e 's/\(^[0-9]\+\.[0-9]\+\)\(.*$\)/\1/' $tmp/img/commit |
1380 sed -e 's/\./_/g') # E.g., "1.8.7" -> "1_8"
1383 tar cjvf $dst/disk$version-$(facet_fstype $SINGLEMDS).tar.bz2 -S *
1388 #run_test 32newtarball "Create a new test_32 disk image tarball for this version"
1391 # The list of applicable tarballs is returned via the caller's
1392 # variable "tarballs".
1395 local node=$(facet_active_host $SINGLEMDS)
1396 local r="do_node $node"
1398 if [ "$CLIENTONLY" ]; then
1399 skip "Client-only testing"
1403 if ! $r which $TUNEFS; then
1404 skip_env "tunefs.lustre required on $node"
1408 local IMGTYPE=$(facet_fstype $SINGLEMDS)
1410 tarballs=$($r find $RLUSTRE/tests -maxdepth 1 -name \'disk*-$IMGTYPE.tar.bz2\')
1412 if [ -z "$tarballs" ]; then
1413 skip "No applicable tarballs found"
1418 t32_test_cleanup() {
1420 local fstype=$(facet_fstype $SINGLEMDS)
1423 if $shall_cleanup_lustre; then
1424 umount $tmp/mnt/lustre || rc=$?
1426 if $shall_cleanup_mdt; then
1427 $r umount -d $tmp/mnt/mdt || rc=$?
1429 if $shall_cleanup_mdt1; then
1430 $r umount -d $tmp/mnt/mdt1 || rc=$?
1432 if $shall_cleanup_ost; then
1433 $r umount -d $tmp/mnt/ost || rc=$?
1438 if [ $fstype == "zfs" ]; then
1439 $r $ZPOOL destroy t32fs-mdt1 || rc=$?
1440 $r $ZPOOL destroy t32fs-ost1 || rc=$?
1445 t32_bits_per_long() {
1447 # Yes, this is not meant to be perfect.
1457 t32_reload_modules() {
1459 local all_removed=false
1462 while ((i < 20)); do
1463 echo "Unloading modules on $node: Attempt $i"
1464 do_rpc_nodes $node $LUSTRE_RMMOD $(facet_fstype $SINGLEMDS) &&
1466 do_rpc_nodes $node check_mem_leak || return 1
1467 if $all_removed; then
1468 do_rpc_nodes $node load_modules
1474 echo "Unloading modules on $node: Given up"
1478 t32_wait_til_devices_gone() {
1484 echo wait for devices to go
1485 while ((i < 20)); do
1486 devices=$(do_rpc_nodes $node $LCTL device_list | wc -l)
1487 loops=$(do_rpc_nodes $node losetup -a | grep -c t32)
1488 ((devices == 0 && loops == 0)) && return 0
1492 echo "waiting for dev on $node: dev $devices loop $loops given up"
1493 do_rpc_nodes $node "losetup -a"
1494 do_rpc_nodes $node "$LCTL devices_list"
1498 t32_verify_quota() {
1502 local fstype=$(facet_fstype $SINGLEMDS)
1506 $LFS quota -u $T32_QID -v $mnt
1508 qval=$($LFS quota -v -u $T32_QID $mnt |
1509 awk 'BEGIN { num='1' } { if ($1 == "'$mnt'") \
1510 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1512 [ $qval -eq $img_bspace ] || {
1513 echo "bspace, act:$qval, exp:$img_bspace"
1517 qval=$($LFS quota -v -u $T32_QID $mnt |
1518 awk 'BEGIN { num='5' } { if ($1 == "'$mnt'") \
1519 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1521 [ $qval -eq $img_ispace ] || {
1522 echo "ispace, act:$qval, exp:$img_ispace"
1526 qval=$($LFS quota -v -u $T32_QID $mnt |
1527 awk 'BEGIN { num='3' } { if ($1 == "'$mnt'") \
1528 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1530 [ $qval -eq $T32_BLIMIT ] || {
1531 echo "blimit, act:$qval, exp:$T32_BLIMIT"
1535 qval=$($LFS quota -v -u $T32_QID $mnt |
1536 awk 'BEGIN { num='7' } { if ($1 == "'$mnt'") \
1537 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1539 [ $qval -eq $T32_ILIMIT ] || {
1540 echo "ilimit, act:$qval, exp:$T32_ILIMIT"
1544 do_node $node $LCTL conf_param $fsname.quota.mdt=ug
1545 cmd="$LCTL get_param -n osd-$fstype.$fsname-MDT0000"
1546 cmd=$cmd.quota_slave.enabled
1547 wait_update $node "$cmd" "ug" || {
1548 echo "Enable mdt quota failed"
1552 do_node $node $LCTL conf_param $fsname.quota.ost=ug
1553 cmd="$LCTL get_param -n osd-$fstype.$fsname-OST0000"
1554 cmd=$cmd.quota_slave.enabled
1555 wait_update $node "$cmd" "ug" || {
1556 echo "Enable ost quota failed"
1561 runas -u $T32_QID -g $T32_QID dd if=/dev/zero of=$mnt/t32_qf_new \
1562 bs=1M count=$(($T32_BLIMIT / 1024)) oflag=sync && {
1563 echo "Write succeed, but expect -EDQUOT"
1566 rm -f $mnt/t32_qf_new
1568 runas -u $T32_QID -g $T32_QID createmany -m $mnt/t32_qf_ \
1570 echo "Create succeed, but expect -EDQUOT"
1573 unlinkmany $mnt/t32_qf_ $T32_ILIMIT
1581 local dne_upgrade=${dne_upgrade:-"no"}
1582 local ff_convert=${ff_convert:-"no"}
1583 local shall_cleanup_mdt=false
1584 local shall_cleanup_mdt1=false
1585 local shall_cleanup_ost=false
1586 local shall_cleanup_lustre=false
1587 local node=$(facet_active_host $SINGLEMDS)
1588 local r="do_node $node"
1589 local node2=$(facet_active_host mds2)
1597 local nid=$($r $LCTL list_nids | head -1)
1603 local fstype=$(facet_fstype $SINGLEMDS)
1604 local mdt_dev=$tmp/mdt
1605 local ost_dev=$tmp/ost
1607 trap 'trap - RETURN; t32_test_cleanup' RETURN
1609 mkdir -p $tmp/mnt/lustre
1610 $r mkdir -p $tmp/mnt/{mdt,ost}
1611 $r tar xjvf $tarball -S -C $tmp || {
1612 error_noexit "Unpacking the disk image tarball"
1615 img_commit=$($r cat $tmp/commit)
1616 img_kernel=$($r cat $tmp/kernel)
1617 img_arch=$($r cat $tmp/arch)
1618 img_bspace=$($r cat $tmp/bspace)
1619 img_ispace=$($r cat $tmp/ispace)
1620 echo "Upgrading from $(basename $tarball), created with:"
1621 echo " Commit: $img_commit"
1622 echo " Kernel: $img_kernel"
1623 echo " Arch: $img_arch"
1624 echo "OST version: $(get_lustre_version ost1)"
1626 # The convertion can be made only when both of the following
1627 # conditions are satisfied:
1628 # - client img version < 2.3.64
1629 # - ost server version >= 2.5
1630 [ $(version_code $img_commit) -ge $(version_code 2.3.64) -o \
1631 $(lustre_version_code ost1) -lt $(version_code 2.5.0) ] &&
1634 if [ $fstype == "zfs" ]; then
1636 $r $ZPOOL import -f -d $tmp t32fs-mdt1
1637 $r $ZPOOL import -f -d $tmp t32fs-ost1
1638 mdt_dev=t32fs-mdt1/mdt1
1639 ost_dev=t32fs-ost1/ost1
1640 wait_update_facet $SINGLEMDS "$ZPOOL list |
1641 awk '/^t32fs-mdt1/ { print \\\$1 }'" "t32fs-mdt1" || {
1642 error_noexit "import zfs pool failed"
1647 $r $LCTL set_param debug="$PTLDEBUG"
1649 $r $TUNEFS --dryrun $mdt_dev || {
1651 error_noexit "tunefs.lustre before mounting the MDT"
1654 if [ "$writeconf" ]; then
1656 if [ $fstype == "ldiskfs" ]; then
1658 $r $TUNEFS --quota $mdt_dev || {
1660 error_noexit "Enable mdt quota feature"
1665 if [ -n "$($LCTL list_nids | grep -v '\(tcp\|lo\)[[:digit:]]*$')" ]; then
1666 [[ $(lustre_version_code mgs) -ge $(version_code 2.3.59) ]] ||
1667 { skip "LU-2200: Cannot run over Inifiniband w/o lctl replace_nids "
1668 "(Need MGS version at least 2.3.59)"; return 0; }
1670 local osthost=$(facet_active_host ost1)
1671 local ostnid=$(do_node $osthost $LCTL list_nids | head -1)
1674 if [ $fstype == "ldiskfs" ]; then
1677 $r $MOUNT_CMD -o $mopts $mdt_dev $tmp/mnt/mdt
1678 $r lctl replace_nids $fsname-OST0000 $ostnid
1679 $r lctl replace_nids $fsname-MDT0000 $nid
1680 $r umount -d $tmp/mnt/mdt
1683 mopts=exclude=$fsname-OST0000
1684 if [ $fstype == "ldiskfs" ]; then
1689 t32_wait_til_devices_gone $node
1691 $r $MOUNT_CMD -o $mopts $mdt_dev $tmp/mnt/mdt || {
1693 error_noexit "Mounting the MDT"
1696 shall_cleanup_mdt=true
1698 if [ "$dne_upgrade" != "no" ]; then
1699 local fs2mdsdev=$(mdsdevname 1_2)
1700 local fs2mdsvdev=$(mdsvdevname 1_2)
1702 echo "mkfs new MDT on ${fs2mdsdev}...."
1703 if [ $(facet_fstype mds1) == ldiskfs ]; then
1704 mkfsoptions="--mkfsoptions=\\\"-J size=8\\\""
1707 add fs2mds $(mkfs_opts mds2 $fs2mdsdev $fsname) --reformat \
1708 $mkfsoptions $fs2mdsdev $fs2mdsvdev > /dev/null || {
1709 error_noexit "Mkfs new MDT failed"
1713 $r $TUNEFS --dryrun $fs2mdsdev || {
1714 error_noexit "tunefs.lustre before mounting the MDT"
1718 echo "mount new MDT....$fs2mdsdev"
1719 $r mkdir -p $tmp/mnt/mdt1
1720 $r $MOUNT_CMD -o $mopts $fs2mdsdev $tmp/mnt/mdt1 || {
1721 error_noexit "mount mdt1 failed"
1724 shall_cleanup_mdt1=true
1727 uuid=$($r $LCTL get_param -n mdt.$fsname-MDT0000.uuid) || {
1728 error_noexit "Getting MDT UUID"
1731 if [ "$uuid" != $fsname-MDT0000_UUID ]; then
1732 error_noexit "Unexpected MDT UUID: \"$uuid\""
1736 $r $TUNEFS --dryrun $ost_dev || {
1737 error_noexit "tunefs.lustre before mounting the OST"
1740 if [ "$writeconf" ]; then
1741 mopts=mgsnode=$nid,$writeconf
1742 if [ $fstype == "ldiskfs" ]; then
1744 $r $TUNEFS --quota $ost_dev || {
1746 error_noexit "Enable ost quota feature"
1752 if [ $fstype == "ldiskfs" ]; then
1756 $r $MOUNT_CMD -o $mopts $ost_dev $tmp/mnt/ost || {
1757 error_noexit "Mounting the OST"
1760 shall_cleanup_ost=true
1762 uuid=$($r $LCTL get_param -n obdfilter.$fsname-OST0000.uuid) || {
1763 error_noexit "Getting OST UUID"
1766 if [ "$uuid" != $fsname-OST0000_UUID ]; then
1767 error_noexit "Unexpected OST UUID: \"$uuid\""
1771 $r $LCTL conf_param $fsname-OST0000.osc.max_dirty_mb=15 || {
1772 error_noexit "Setting \"max_dirty_mb\""
1775 $r $LCTL conf_param $fsname-OST0000.failover.node=$nid || {
1776 error_noexit "Setting OST \"failover.node\""
1779 $r $LCTL conf_param $fsname-MDT0000.mdc.max_rpcs_in_flight=9 || {
1780 error_noexit "Setting \"max_rpcs_in_flight\""
1783 $r $LCTL conf_param $fsname-MDT0000.failover.node=$nid || {
1784 error_noexit "Setting MDT \"failover.node\""
1787 $r $LCTL pool_new $fsname.interop || {
1788 error_noexit "Setting \"interop\""
1791 $r $LCTL conf_param $fsname-MDT0000.lov.stripesize=4M || {
1792 error_noexit "Setting \"lov.stripesize\""
1796 if [ "$ff_convert" != "no" -a $(facet_fstype ost1) == "ldiskfs" ]; then
1797 $r $LCTL lfsck_start -M $fsname-OST0000 || {
1798 error_noexit "Start OI scrub on OST0"
1802 # The oi_scrub should be on ost1, but for test_32(),
1803 # all on the SINGLEMDS.
1804 wait_update_facet $SINGLEMDS "$LCTL get_param -n \
1805 osd-ldiskfs.$fsname-OST0000.oi_scrub |
1806 awk '/^status/ { print \\\$2 }'" "completed" 30 || {
1807 error_noexit "Failed to get the expected 'completed'"
1811 local UPDATED=$($r $LCTL get_param -n \
1812 osd-ldiskfs.$fsname-OST0000.oi_scrub |
1813 awk '/^updated/ { print $2 }')
1814 [ $UPDATED -ge 1 ] || {
1815 error_noexit "Only $UPDATED objects have been converted"
1820 if [ "$dne_upgrade" != "no" ]; then
1821 $r $LCTL conf_param \
1822 $fsname-MDT0001.mdc.max_rpcs_in_flight=9 || {
1823 error_noexit "Setting MDT1 \"max_rpcs_in_flight\""
1826 $r $LCTL conf_param $fsname-MDT0001.failover.node=$nid || {
1827 error_noexit "Setting MDT1 \"failover.node\""
1830 $r $LCTL conf_param $fsname-MDT0001.lov.stripesize=4M || {
1831 error_noexit "Setting MDT1 \"lov.stripesize\""
1837 if [ "$writeconf" ]; then
1838 $MOUNT_CMD $nid:/$fsname $tmp/mnt/lustre || {
1839 error_noexit "Mounting the client"
1842 shall_cleanup_lustre=true
1843 $LCTL set_param debug="$PTLDEBUG"
1845 t32_verify_quota $node $fsname $tmp/mnt/lustre || {
1846 error_noexit "verify quota failed"
1850 if [ "$dne_upgrade" != "no" ]; then
1851 $LFS mkdir -i 1 $tmp/mnt/lustre/remote_dir || {
1852 error_noexit "set remote dir failed"
1856 pushd $tmp/mnt/lustre
1857 tar -cf - . --exclude=./remote_dir |
1858 tar -xvf - -C remote_dir 1>/dev/null || {
1859 error_noexit "cp to remote dir failed"
1865 dd if=/dev/zero of=$tmp/mnt/lustre/tmp_file bs=10k count=10 || {
1866 error_noexit "dd failed"
1869 rm -rf $tmp/mnt/lustre/tmp_file || {
1870 error_noexit "rm failed"
1874 if $r test -f $tmp/sha1sums; then
1875 # LU-2393 - do both sorts on same node to ensure locale
1877 $r cat $tmp/sha1sums | sort -k 2 >$tmp/sha1sums.orig
1878 if [ "$dne_upgrade" != "no" ]; then
1879 pushd $tmp/mnt/lustre/remote_dir
1881 pushd $tmp/mnt/lustre
1884 find ! -name .lustre -type f -exec sha1sum {} \; |
1885 sort -k 2 >$tmp/sha1sums || {
1886 error_noexit "sha1sum"
1890 if ! diff -ub $tmp/sha1sums.orig $tmp/sha1sums; then
1891 error_noexit "sha1sum verification failed"
1895 echo "sha1sum verification skipped"
1898 if [ "$dne_upgrade" != "no" ]; then
1899 rm -rf $tmp/mnt/lustre/remote_dir || {
1900 error_noexit "remove remote dir failed"
1905 if $r test -f $tmp/list; then
1907 # There is not a Test Framework API to copy files to or
1908 # from a remote node.
1910 # LU-2393 - do both sorts on same node to ensure locale
1912 $r cat $tmp/list | sort -k 6 >$tmp/list.orig
1913 pushd $tmp/mnt/lustre
1914 ls -Rni --time-style=+%s | sort -k 6 >$tmp/list || {
1920 # 32-bit and 64-bit clients use different algorithms to
1921 # convert FIDs into inode numbers. Hence, remove the inode
1922 # numbers from the lists, if the original list was created
1923 # on an architecture with different number of bits per
1926 if [ $(t32_bits_per_long $(uname -m)) != \
1927 $(t32_bits_per_long $img_arch) ]; then
1928 echo "Different number of bits per \"long\" from the disk image"
1929 for list in list.orig list; do
1930 sed -i -e 's/^[0-9]\+[ \t]\+//' $tmp/$list
1933 if ! diff -ub $tmp/list.orig $tmp/list; then
1934 error_noexit "list verification failed"
1938 echo "list verification skipped"
1942 # When adding new data verification tests, please check for
1943 # the presence of the required reference files first, like
1944 # the "sha1sums" and "list" tests above, to avoid the need to
1945 # regenerate every image for each test addition.
1948 nrpcs_orig=$($LCTL get_param \
1949 -n mdc.*MDT0000*.max_rpcs_in_flight) || {
1950 error_noexit "Getting \"max_rpcs_in_flight\""
1953 nrpcs=$((nrpcs_orig + 5))
1954 $r $LCTL conf_param $fsname-MDT0000.mdc.max_rpcs_in_flight=$nrpcs || {
1955 error_noexit "Changing \"max_rpcs_in_flight\""
1958 wait_update $HOSTNAME "$LCTL get_param \
1959 -n mdc.*MDT0000*.max_rpcs_in_flight" $nrpcs || {
1960 error_noexit "Verifying \"max_rpcs_in_flight\""
1964 umount $tmp/mnt/lustre || {
1965 error_noexit "Unmounting the client"
1968 shall_cleanup_lustre=false
1970 if [ "$dne_upgrade" != "no" ]; then
1971 $r umount -d $tmp/mnt/mdt1 || {
1972 error_noexit "Unmounting the MDT2"
1975 shall_cleanup_mdt1=false
1978 $r umount -d $tmp/mnt/mdt || {
1979 error_noexit "Unmounting the MDT"
1982 shall_cleanup_mdt=false
1984 $r umount -d $tmp/mnt/ost || {
1985 error_noexit "Unmounting the OST"
1988 shall_cleanup_ost=false
1990 t32_reload_modules $node || {
1991 error_noexit "Reloading modules"
1995 # mount a second time to make sure we didnt leave upgrade flag on
1996 $r $TUNEFS --dryrun $mdt_dev || {
1998 error_noexit "tunefs.lustre before remounting the MDT"
2002 mopts=exclude=$fsname-OST0000
2003 if [ $fstype == "ldiskfs" ]; then
2006 $r $MOUNT_CMD -o $mopts $mdt_dev $tmp/mnt/mdt || {
2007 error_noexit "Remounting the MDT"
2010 shall_cleanup_mdt=true
2020 for tarball in $tarballs; do
2021 t32_test $tarball || let "rc += $?"
2025 run_test 32a "Upgrade (not live)"
2033 for tarball in $tarballs; do
2034 t32_test $tarball writeconf || let "rc += $?"
2038 run_test 32b "Upgrade with writeconf"
2045 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
2047 for tarball in $tarballs; do
2048 dne_upgrade=yes t32_test $tarball writeconf || rc=$?
2052 run_test 32c "dne upgrade test"
2060 for tarball in $tarballs; do
2061 ff_convert=yes t32_test $tarball || rc=$?
2065 run_test 32d "convert ff test"
2067 test_33a() { # bug 12333, was test_33
2069 local FSNAME2=test-123
2070 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
2073 [ -n "$ost1_HOST" ] && fs2ost_HOST=$ost1_HOST
2075 if [ -z "$fs2ost_DEV" -o -z "$fs2mds_DEV" ]; then
2076 local dev=${SINGLEMDS}_dev
2077 local MDSDEV=${!dev}
2078 is_blkdev $SINGLEMDS $MDSDEV && \
2079 skip_env "mixed loopback and real device not working" && return
2082 local fs2mdsdev=$(mdsdevname 1_2)
2083 local fs2ostdev=$(ostdevname 1_2)
2084 local fs2mdsvdev=$(mdsvdevname 1_2)
2085 local fs2ostvdev=$(ostvdevname 1_2)
2087 if [ $(facet_fstype mds1) == ldiskfs ]; then
2088 mkfsoptions="--mkfsoptions=\\\"-J size=8\\\"" # See bug 17931.
2091 add fs2mds $(mkfs_opts mds1 ${fs2mdsdev}) --mgs --fsname=${FSNAME2} \
2092 --reformat $mkfsoptions $fs2mdsdev $fs2mdsvdev || exit 10
2093 add fs2ost $(mkfs_opts ost1 ${fs2ostdev}) --mgsnode=$MGSNID \
2094 --fsname=${FSNAME2} --index=8191 --reformat $fs2ostdev \
2095 $fs2ostvdev || exit 10
2097 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS && trap cleanup_24a EXIT INT
2098 start fs2ost $fs2ostdev $OST_MOUNT_OPTS
2099 do_facet $SINGLEMDS "$LCTL conf_param $FSNAME2.sys.timeout=200" || rc=1
2101 $MOUNT_CMD $MGSNID:/${FSNAME2} $MOUNT2 || rc=2
2104 cp /etc/hosts $MOUNT2/ || rc=3
2105 $LFS getstripe $MOUNT2/hosts
2110 cleanup_nocli || rc=6
2113 run_test 33a "Mount ost with a large index number"
2115 test_33b() { # was test_34
2118 do_facet client dd if=/dev/zero of=$MOUNT/24 bs=1024k count=1
2119 # Drop lock cancelation reply during umount
2120 #define OBD_FAIL_LDLM_CANCEL_NET 0x304
2121 do_facet client lctl set_param fail_loc=0x80000304
2122 #lctl set_param debug=-1
2123 umount_client $MOUNT
2126 run_test 33b "Drop cancel during umount"
2130 do_facet client "sh runmultiop_bg_pause $DIR/file O_c"
2131 manual_umount_client
2133 do_facet client killall -USR1 multiop
2134 if [ $rc -eq 0 ]; then
2135 error "umount not fail!"
2140 run_test 34a "umount with opened file should be fail"
2145 touch $DIR/$tfile || return 1
2146 stop_mds --force || return 2
2148 manual_umount_client --force
2150 if [ $rc -ne 0 ]; then
2151 error "mtab after failed umount - rc $rc"
2157 run_test 34b "force umount with failed mds should be normal"
2161 touch $DIR/$tfile || return 1
2162 stop_ost --force || return 2
2164 manual_umount_client --force
2166 if [ $rc -ne 0 ]; then
2167 error "mtab after failed umount - rc $rc"
2173 run_test 34c "force umount with failed ost should be normal"
2175 test_35a() { # bug 12459
2178 DBG_SAVE="`lctl get_param -n debug`"
2179 lctl set_param debug="ha"
2181 log "Set up a fake failnode for the MDS"
2183 local device=$(do_facet $SINGLEMDS "lctl get_param -n devices" |
2184 awk '($3 ~ "mdt" && $4 ~ "MDT") { print $4 }' | head -1)
2185 do_facet mgs "$LCTL conf_param \
2186 ${device}.failover.node=$(h2$NETTYPE $FAKENID)" || return 4
2188 log "Wait for RECONNECT_INTERVAL seconds (10s)"
2191 MSG="conf-sanity.sh test_35a `date +%F%kh%Mm%Ss`"
2194 log "Stopping the MDT: $device"
2195 stop_mdt 1 || return 5
2197 df $MOUNT > /dev/null 2>&1 &
2199 log "Restarting the MDT: $device"
2200 start_mdt 1 || return 6
2201 log "Wait for df ($DFPID) ... "
2204 lctl set_param debug="$DBG_SAVE"
2206 # retrieve from the log the first server that the client tried to
2207 # contact after the connection loss
2208 $LCTL dk $TMP/lustre-log-$TESTNAME.log
2209 NEXTCONN=`awk "/${MSG}/ {start = 1;}
2210 /import_select_connection.*$device-mdc.* using connection/ {
2212 if (\\\$NF ~ /$FAKENID/)
2218 }" $TMP/lustre-log-$TESTNAME.log`
2219 [ "$NEXTCONN" != "0" ] && log "The client didn't try to reconnect to the last active server (tried ${NEXTCONN} instead)" && return 7
2221 # remove nid settings
2222 writeconf_or_reformat
2224 run_test 35a "Reconnect to the last active server first"
2226 test_35b() { # bug 18674
2227 remote_mds || { skip "local MDS" && return 0; }
2231 $LCTL set_param debug="ha"
2233 MSG="conf-sanity.sh test_35b `date +%F%kh%Mm%Ss`"
2236 log "Set up a fake failnode for the MDS"
2238 local device=$(do_facet $SINGLEMDS "$LCTL get_param -n devices" |
2239 awk '($3 ~ "mdt" && $4 ~ "MDT") { print $4 }' | head -1)
2240 do_facet mgs "$LCTL conf_param \
2241 ${device}.failover.node=$(h2$NETTYPE $FAKENID)" || return 1
2243 local at_max_saved=0
2244 # adaptive timeouts may prevent seeing the issue
2245 if at_is_enabled; then
2246 at_max_saved=$(at_max_get mds)
2247 at_max_set 0 mds client
2250 mkdir -p $MOUNT/$tdir
2252 log "Injecting EBUSY on MDS"
2253 # Setting OBD_FAIL_MDS_RESEND=0x136
2254 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0x80000136" || return 2
2256 $LCTL set_param mdc.${FSNAME}*.stats=clear
2258 log "Creating a test file and stat it"
2259 touch $MOUNT/$tdir/$tfile
2260 stat $MOUNT/$tdir/$tfile
2262 log "Stop injecting EBUSY on MDS"
2263 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0" || return 3
2264 rm -f $MOUNT/$tdir/$tfile
2267 # restore adaptive timeout
2268 [ $at_max_saved -ne 0 ] && at_max_set $at_max_saved mds client
2270 $LCTL dk $TMP/lustre-log-$TESTNAME.log
2272 CONNCNT=`$LCTL get_param mdc.${FSNAME}*.stats | awk '/mds_connect/{print $2}'`
2274 # retrieve from the log if the client has ever tried to
2275 # contact the fake server after the loss of connection
2276 FAILCONN=`awk "BEGIN {ret = 0;}
2277 /import_select_connection.*${FSNAME}-MDT0000-mdc.* using connection/ {
2279 if (\\\$NF ~ /$FAKENID/) {
2284 END {print ret}" $TMP/lustre-log-$TESTNAME.log`
2286 [ "$FAILCONN" == "0" ] && \
2287 log "ERROR: The client reconnection has not been triggered" && \
2289 [ "$FAILCONN" == "2" ] && \
2290 log "ERROR: The client tried to reconnect to the failover server while the primary was busy" && \
2294 # When OBD_FAIL_MDS_RESEND is hit, we sleep for 2 * obd_timeout
2295 # Reconnects are supposed to be rate limited to one every 5s
2296 [ $CONNCNT -gt $((2 * $TIMEOUT / 5 + 1)) ] && \
2297 log "ERROR: Too many reconnects $CONNCNT" && \
2301 # remove nid settings
2302 writeconf_or_reformat
2304 run_test 35b "Continue reconnection retries, if the active server is busy"
2307 [ $OSTCOUNT -lt 2 ] && skip_env "skipping test for single OST" && return
2309 [ "$ost_HOST" = "`hostname`" -o "$ost1_HOST" = "`hostname`" ] || \
2310 { skip "remote OST" && return 0; }
2313 local FSNAME2=test1234
2314 local fs3ost_HOST=$ost_HOST
2315 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
2317 [ -n "$ost1_HOST" ] && fs2ost_HOST=$ost1_HOST && fs3ost_HOST=$ost1_HOST
2319 if [ -z "$fs2ost_DEV" -o -z "$fs2mds_DEV" -o -z "$fs3ost_DEV" ]; then
2320 is_blkdev $SINGLEMDS $MDSDEV && \
2321 skip_env "mixed loopback and real device not working" && return
2324 local fs2mdsdev=$(mdsdevname 1_2)
2325 local fs2ostdev=$(ostdevname 1_2)
2326 local fs3ostdev=$(ostdevname 2_2)
2327 local fs2mdsvdev=$(mdsvdevname 1_2)
2328 local fs2ostvdev=$(ostvdevname 1_2)
2329 local fs3ostvdev=$(ostvdevname 2_2)
2331 add fs2mds $(mkfs_opts mds1 ${fs2mdsdev}) --mgs --fsname=${FSNAME2} \
2332 --reformat $fs2mdsdev $fs2mdsvdev || exit 10
2333 # XXX after we support non 4K disk blocksize in ldiskfs, specify a
2334 # different one than the default value here.
2335 add fs2ost $(mkfs_opts ost1 ${fs2ostdev}) --mgsnode=$MGSNID \
2336 --fsname=${FSNAME2} --reformat $fs2ostdev $fs2ostvdev || exit 10
2337 add fs3ost $(mkfs_opts ost1 ${fs3ostdev}) --mgsnode=$MGSNID \
2338 --fsname=${FSNAME2} --reformat $fs3ostdev $fs3ostvdev || exit 10
2340 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS
2341 start fs2ost $fs2ostdev $OST_MOUNT_OPTS
2342 start fs3ost $fs3ostdev $OST_MOUNT_OPTS
2344 $MOUNT_CMD $MGSNID:/${FSNAME2} $MOUNT2 || return 1
2346 sleep 5 # until 11778 fixed
2348 dd if=/dev/zero of=$MOUNT2/$tfile bs=1M count=7 || return 2
2350 BKTOTAL=`lctl get_param -n obdfilter.*.kbytestotal | awk 'BEGIN{total=0}; {total+=$1}; END{print total}'`
2351 BKFREE=`lctl get_param -n obdfilter.*.kbytesfree | awk 'BEGIN{free=0}; {free+=$1}; END{print free}'`
2352 BKAVAIL=`lctl get_param -n obdfilter.*.kbytesavail | awk 'BEGIN{avail=0}; {avail+=$1}; END{print avail}'`
2353 STRING=`df -P $MOUNT2 | tail -n 1 | awk '{print $2","$3","$4}'`
2354 DFTOTAL=`echo $STRING | cut -d, -f1`
2355 DFUSED=`echo $STRING | cut -d, -f2`
2356 DFAVAIL=`echo $STRING | cut -d, -f3`
2357 DFFREE=$(($DFTOTAL - $DFUSED))
2359 ALLOWANCE=$((64 * $OSTCOUNT))
2361 if [ $DFTOTAL -lt $(($BKTOTAL - $ALLOWANCE)) ] ||
2362 [ $DFTOTAL -gt $(($BKTOTAL + $ALLOWANCE)) ] ; then
2363 echo "**** FAIL: df total($DFTOTAL) mismatch OST total($BKTOTAL)"
2366 if [ $DFFREE -lt $(($BKFREE - $ALLOWANCE)) ] ||
2367 [ $DFFREE -gt $(($BKFREE + $ALLOWANCE)) ] ; then
2368 echo "**** FAIL: df free($DFFREE) mismatch OST free($BKFREE)"
2371 if [ $DFAVAIL -lt $(($BKAVAIL - $ALLOWANCE)) ] ||
2372 [ $DFAVAIL -gt $(($BKAVAIL + $ALLOWANCE)) ] ; then
2373 echo "**** FAIL: df avail($DFAVAIL) mismatch OST avail($BKAVAIL)"
2378 stop fs3ost -f || return 200
2379 stop fs2ost -f || return 201
2380 stop fs2mds -f || return 202
2381 unload_modules_conf || return 203
2384 run_test 36 "df report consistency on OSTs with different block size"
2387 local mntpt=$(facet_mntpt $SINGLEMDS)
2388 local mdsdev=$(mdsdevname ${SINGLEMDS//mds/})
2389 local mdsdev_sym="$TMP/sym_mdt.img"
2390 local opts=$MDS_MOUNT_OPTS
2393 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
2394 skip "Currently only applicable to ldiskfs-based MDTs"
2398 echo "MDS : $mdsdev"
2399 echo "SYMLINK : $mdsdev_sym"
2400 do_facet $SINGLEMDS rm -f $mdsdev_sym
2402 do_facet $SINGLEMDS ln -s $mdsdev $mdsdev_sym
2404 echo "mount symlink device - $mdsdev_sym"
2406 if ! do_facet $SINGLEMDS test -b $mdsdev; then
2407 opts=$(csa_add "$opts" -o loop)
2409 mount_op=$(do_facet $SINGLEMDS mount -v -t lustre $opts \
2410 $mdsdev_sym $mntpt 2>&1)
2413 echo mount_op=$mount_op
2415 do_facet $SINGLEMDS "umount -d $mntpt && rm -f $mdsdev_sym"
2417 if $(echo $mount_op | grep -q "unable to set tunable"); then
2418 error "set tunables failed for symlink device"
2421 [ $rc -eq 0 ] || error "mount symlink $mdsdev_sym failed! rc=$rc"
2425 run_test 37 "verify set tunables works for symlink device"
2427 test_38() { # bug 14222
2428 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
2429 skip "Only applicable to ldiskfs-based MDTs"
2437 FILES=`find $SRC -type f -mtime +1 | head -n $COUNT`
2438 log "copying $(echo $FILES | wc -w) files to $DIR/$tdir"
2440 tar cf - $FILES | tar xf - -C $DIR/$tdir || \
2441 error "copying $SRC to $DIR/$tdir"
2443 umount_client $MOUNT
2445 log "rename lov_objid file on MDS"
2446 rm -f $TMP/lov_objid.orig
2448 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
2449 do_facet $SINGLEMDS "$DEBUGFS -c -R \\\"dump lov_objid $TMP/lov_objid.orig\\\" $MDSDEV"
2450 do_facet $SINGLEMDS "$DEBUGFS -w -R \\\"rm lov_objid\\\" $MDSDEV"
2452 do_facet $SINGLEMDS "od -Ax -td8 $TMP/lov_objid.orig"
2453 # check create in mds_lov_connect
2457 [ $V ] && log "verifying $DIR/$tdir/$f"
2458 diff -q $f $DIR/$tdir/$f || ERROR=y
2460 do_facet $SINGLEMDS "$DEBUGFS -c -R \\\"dump lov_objid $TMP/lov_objid.new\\\" $MDSDEV"
2461 do_facet $SINGLEMDS "od -Ax -td8 $TMP/lov_objid.new"
2462 [ "$ERROR" = "y" ] && error "old and new files are different after connect" || true
2464 # check it's updates in sync
2465 umount_client $MOUNT
2468 do_facet $SINGLEMDS dd if=/dev/zero of=$TMP/lov_objid.clear bs=4096 count=1
2469 do_facet $SINGLEMDS "$DEBUGFS -w -R \\\"rm lov_objid\\\" $MDSDEV"
2470 do_facet $SINGLEMDS "$DEBUGFS -w -R \\\"write $TMP/lov_objid.clear lov_objid\\\" $MDSDEV "
2475 [ $V ] && log "verifying $DIR/$tdir/$f"
2476 diff -q $f $DIR/$tdir/$f || ERROR=y
2478 do_facet $SINGLEMDS "$DEBUGFS -c -R \\\"dump lov_objid $TMP/lov_objid.new1\\\" $MDSDEV"
2479 do_facet $SINGLEMDS "od -Ax -td8 $TMP/lov_objid.new1"
2480 umount_client $MOUNT
2482 [ "$ERROR" = "y" ] && 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"
2590 chmod ugo+x $DIR || error "chmod 0 failed"
2591 set_conf_param_and_check mds \
2592 "lctl get_param -n mdt.$FSNAME-MDT0000.root_squash" \
2593 "$FSNAME.mdt.root_squash" \
2595 set_conf_param_and_check mds \
2596 "lctl get_param -n mdt.$FSNAME-MDT0000.nosquash_nids" \
2597 "$FSNAME.mdt.nosquash_nids" \
2601 # create set of test files
2603 echo "111" > $DIR/$tfile-userfile || error "write 1 failed"
2604 chmod go-rw $DIR/$tfile-userfile || error "chmod 1 failed"
2605 chown $RUNAS_ID.$RUNAS_ID $DIR/$tfile-userfile || error "chown failed"
2607 echo "222" > $DIR/$tfile-rootfile || error "write 2 failed"
2608 chmod go-rw $DIR/$tfile-rootfile || error "chmod 2 faield"
2610 mkdir $DIR/$tdir-rootdir -p || error "mkdir failed"
2611 chmod go-rwx $DIR/$tdir-rootdir || error "chmod 3 failed"
2612 touch $DIR/$tdir-rootdir/tfile-1 || error "touch failed"
2615 # check root_squash:
2616 # set root squash UID:GID to RUNAS_ID
2617 # root should be able to access only files owned by RUNAS_ID
2619 set_conf_param_and_check mds \
2620 "lctl get_param -n mdt.$FSNAME-MDT0000.root_squash" \
2621 "$FSNAME.mdt.root_squash" \
2622 "$RUNAS_ID:$RUNAS_ID"
2624 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-userfile)
2625 dd if=$DIR/$tfile-userfile 1>/dev/null 2>/dev/null || \
2626 error "$ST: root read permission is denied"
2627 echo "$ST: root read permission is granted - ok"
2630 dd conv=notrunc if=$DIR/$tfile-userfile 1>/dev/null 2>/dev/null || \
2631 error "$ST: root write permission is denied"
2632 echo "$ST: root write permission is granted - ok"
2634 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-rootfile)
2635 dd if=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null && \
2636 error "$ST: root read permission is granted"
2637 echo "$ST: root read permission is denied - ok"
2640 dd conv=notrunc of=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null && \
2641 error "$ST: root write permission is granted"
2642 echo "$ST: root write permission is denied - ok"
2644 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tdir-rootdir)
2645 rm $DIR/$tdir-rootdir/tfile-1 1>/dev/null 2>/dev/null && \
2646 error "$ST: root unlink permission is granted"
2647 echo "$ST: root unlink permission is denied - ok"
2649 touch $DIR/tdir-rootdir/tfile-2 1>/dev/null 2>/dev/null && \
2650 error "$ST: root create permission is granted"
2651 echo "$ST: root create permission is denied - ok"
2654 # check nosquash_nids:
2655 # put client's NID into nosquash_nids list,
2656 # root should be able to access root file after that
2658 local NIDLIST=$(lctl list_nids all | tr '\n' ' ')
2659 NIDLIST="2@elan $NIDLIST 192.168.0.[2,10]@tcp"
2660 NIDLIST=$(echo $NIDLIST | tr -s ' ' ' ')
2661 set_conf_param_and_check mds \
2662 "lctl get_param -n mdt.$FSNAME-MDT0000.nosquash_nids" \
2663 "$FSNAME-MDTall.mdt.nosquash_nids" \
2666 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-rootfile)
2667 dd if=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null || \
2668 error "$ST: root read permission is denied"
2669 echo "$ST: root read permission is granted - ok"
2672 dd conv=notrunc of=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null || \
2673 error "$ST: root write permission is denied"
2674 echo "$ST: root write permission is granted - ok"
2676 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tdir-rootdir)
2677 rm $DIR/$tdir-rootdir/tfile-1 || \
2678 error "$ST: root unlink permission is denied"
2679 echo "$ST: root unlink permission is granted - ok"
2680 touch $DIR/$tdir-rootdir/tfile-2 || \
2681 error "$ST: root create permission is denied"
2682 echo "$ST: root create permission is granted - ok"
2686 run_test 43 "check root_squash and nosquash_nids"
2688 umount_client $MOUNT
2693 check_mount || return 2
2694 UUID=$($LCTL get_param llite.${FSNAME}*.uuid | cut -d= -f2)
2696 UUIDS=$(do_facet $SINGLEMDS "$LCTL get_param mdt.${FSNAME}*.exports.*.uuid")
2697 for VAL in $UUIDS; do
2698 NID=$(echo $VAL | cut -d= -f1)
2699 CLUUID=$(echo $VAL | cut -d= -f2)
2700 [ "$UUID" = "$CLUUID" ] && STATS_FOUND=yes && break
2702 [ "$STATS_FOUND" = "no" ] && error "stats not found for client"
2706 run_test 44 "mounted client proc entry exists"
2710 check_mount || return 2
2715 #define OBD_FAIL_PTLRPC_LONG_UNLINK 0x50f
2716 do_facet client "lctl set_param fail_loc=0x50f"
2719 manual_umount_client --force || return 3
2720 do_facet client "lctl set_param fail_loc=0x0"
2722 mount_client $MOUNT || return 4
2726 run_test 45 "long unlink handling in ptlrpcd"
2733 umount_client $MOUNT2 || rc=$?
2734 umount_client $MOUNT || rc=$?
2735 while [ $count -gt 0 ]; do
2736 stop ost${count} -f || rc=$?
2740 cleanup_nocli || rc=$?
2741 #writeconf to remove all ost2 traces for subsequent tests
2742 writeconf_or_reformat
2747 echo "Testing with $OSTCOUNT OSTs"
2749 start_mds || return 1
2750 #first client should see only one ost
2751 start_ost || return 2
2752 wait_osc_import_state mds ost FULL
2754 mount_client $MOUNT || return 3
2755 trap "cleanup_46a $OSTCOUNT" EXIT ERR
2758 for (( i=2; i<=$OSTCOUNT; i++ )); do
2759 start ost$i `ostdevname $i` $OST_MOUNT_OPTS || return $((i+2))
2762 # wait until osts in sync
2763 for (( i=2; i<=$OSTCOUNT; i++ )); do
2764 wait_osc_import_state mds ost$i FULL
2765 wait_osc_import_state client ost$i FULL
2768 #second client see all ost's
2770 mount_client $MOUNT2 || return 8
2771 $LFS setstripe -c -1 $MOUNT2 || return 9
2772 $LFS getstripe $MOUNT2 || return 10
2774 echo "ok" > $MOUNT2/widestripe
2775 $LFS getstripe $MOUNT2/widestripe || return 11
2776 # fill acl buffer for avoid expand lsm to them
2777 awk -F : '{if (FNR < 25) { print "u:"$1":rwx" }}' /etc/passwd | while read acl; do
2778 setfacl -m $acl $MOUNT2/widestripe
2782 stat $MOUNT/widestripe || return 12
2784 cleanup_46a $OSTCOUNT || { echo "cleanup_46a failed!" && return 13; }
2787 run_test 46a "handle ost additional - wide striped file"
2792 check_mount || return 2
2793 $LCTL set_param ldlm.namespaces.$FSNAME-*-*-*.lru_size=100
2797 for ns in $($LCTL get_param ldlm.namespaces.$FSNAME-*-*-*.lru_size); do
2798 if echo $ns | grep "MDT[[:digit:]]*"; then
2801 lrs=$(echo $ns | sed 's/.*lru_size=//')
2802 lru_size[count]=$lrs
2807 facet_failover $SINGLEMDS
2808 client_up || return 3
2811 for ns in $($LCTL get_param ldlm.namespaces.$FSNAME-*-*-*.lru_size); do
2812 if echo $ns | grep "MDT[[:digit:]]*"; then
2815 lrs=$(echo $ns | sed 's/.*lru_size=//')
2816 if ! test "$lrs" -eq "${lru_size[count]}"; then
2817 n=$(echo $ns | sed -e 's/ldlm.namespaces.//' -e 's/.lru_size=.*//')
2818 error "$n has lost lru_size: $lrs vs. ${lru_size[count]}"
2826 run_test 47 "server restart does not make client loss lru_resize settings"
2831 # reformat after this test is needed - if test will failed
2832 # we will have unkillable file at FS
2836 test_48() { # bug 17636
2839 check_mount || return 2
2841 $LFS setstripe -c -1 $MOUNT || return 9
2842 $LFS getstripe $MOUNT || return 10
2844 echo "ok" > $MOUNT/widestripe
2845 $LFS getstripe $MOUNT/widestripe || return 11
2847 trap cleanup_48 EXIT ERR
2849 # fill acl buffer for avoid expand lsm to them
2850 getent passwd | awk -F : '{ print "u:"$1":rwx" }' | while read acl; do
2851 setfacl -m $acl $MOUNT/widestripe
2854 stat $MOUNT/widestripe || return 12
2859 run_test 48 "too many acls on file"
2861 # check PARAM_SYS_LDLM_TIMEOUT option of MKFS.LUSTRE
2862 test_49() { # bug 17710
2863 local timeout_orig=$TIMEOUT
2864 local ldlm_timeout_orig=$LDLM_TIMEOUT
2865 local LOCAL_TIMEOUT=20
2867 LDLM_TIMEOUT=$LOCAL_TIMEOUT
2868 TIMEOUT=$LOCAL_TIMEOUT
2872 check_mount || return 1
2874 echo "check ldlm_timout..."
2875 LDLM_MDS="`do_facet $SINGLEMDS lctl get_param -n ldlm_timeout`"
2876 LDLM_OST1="`do_facet ost1 lctl get_param -n ldlm_timeout`"
2877 LDLM_CLIENT="`do_facet client lctl get_param -n ldlm_timeout`"
2879 if [ $LDLM_MDS -ne $LDLM_OST1 ] || [ $LDLM_MDS -ne $LDLM_CLIENT ]; then
2880 error "Different LDLM_TIMEOUT:$LDLM_MDS $LDLM_OST1 $LDLM_CLIENT"
2883 if [ $LDLM_MDS -ne $((LOCAL_TIMEOUT / 3)) ]; then
2884 error "LDLM_TIMEOUT($LDLM_MDS) is not correct"
2887 umount_client $MOUNT
2888 stop_ost || return 2
2889 stop_mds || return 3
2891 LDLM_TIMEOUT=$((LOCAL_TIMEOUT - 1))
2895 check_mount || return 7
2897 LDLM_MDS="`do_facet $SINGLEMDS lctl get_param -n ldlm_timeout`"
2898 LDLM_OST1="`do_facet ost1 lctl get_param -n ldlm_timeout`"
2899 LDLM_CLIENT="`do_facet client lctl get_param -n ldlm_timeout`"
2901 if [ $LDLM_MDS -ne $LDLM_OST1 ] || [ $LDLM_MDS -ne $LDLM_CLIENT ]; then
2902 error "Different LDLM_TIMEOUT:$LDLM_MDS $LDLM_OST1 $LDLM_CLIENT"
2905 if [ $LDLM_MDS -ne $((LOCAL_TIMEOUT - 1)) ]; then
2906 error "LDLM_TIMEOUT($LDLM_MDS) is not correct"
2909 cleanup || return $?
2911 LDLM_TIMEOUT=$ldlm_timeout_orig
2912 TIMEOUT=$timeout_orig
2914 run_test 49 "check PARAM_SYS_LDLM_TIMEOUT option of MKFS.LUSTRE"
2917 # Test both statfs and lfs df and fail if either one fails
2918 multiop_bg_pause $1 f_
2921 killall -USR1 multiop
2922 [ $RC1 -ne 0 ] && log "lazystatfs multiop failed"
2923 wait $PID || { RC1=$?; log "multiop return error "; }
2930 if [ $RC2 -eq 0 ]; then
2932 log "lazystatfs df failed"
2936 [[ $RC1 -ne 0 || $RC2 -eq 0 ]] && RC=1
2942 lctl set_param llite.$FSNAME-*.lazystatfs=1
2945 lazystatfs $MOUNT || error "lazystatfs failed but no down servers"
2947 cleanup || return $?
2949 run_test 50a "lazystatfs all servers available =========================="
2953 lctl set_param llite.$FSNAME-*.lazystatfs=1
2956 # Wait for client to detect down OST
2957 stop_ost || error "Unable to stop OST1"
2958 wait_osc_import_state mds ost DISCONN
2960 lazystatfs $MOUNT || error "lazystatfs should don't have returned EIO"
2962 umount_client $MOUNT || error "Unable to unmount client"
2963 stop_mds || error "Unable to stop MDS"
2965 run_test 50b "lazystatfs all servers down =========================="
2968 start_mds || error "Unable to start MDS"
2969 start_ost || error "Unable to start OST1"
2970 start_ost2 || error "Unable to start OST2"
2971 mount_client $MOUNT || error "Unable to mount client"
2972 lctl set_param llite.$FSNAME-*.lazystatfs=1
2975 # Wait for client to detect down OST
2976 stop_ost || error "Unable to stop OST1"
2977 wait_osc_import_state mds ost DISCONN
2978 lazystatfs $MOUNT || error "lazystatfs failed with one down server"
2980 umount_client $MOUNT || error "Unable to unmount client"
2981 stop_ost2 || error "Unable to stop OST2"
2982 stop_mds || error "Unable to stop MDS"
2983 #writeconf to remove all ost2 traces for subsequent tests
2984 writeconf_or_reformat
2986 run_test 50c "lazystatfs one server down =========================="
2989 start_mds || error "Unable to start MDS"
2990 start_ost || error "Unable to start OST1"
2991 start_ost2 || error "Unable to start OST2"
2992 mount_client $MOUNT || error "Unable to mount client"
2993 lctl set_param llite.$FSNAME-*.lazystatfs=1
2996 # Issue the statfs during the window where the client still
2997 # belives the OST to be available but it is in fact down.
2998 # No failure just a statfs which hangs for a timeout interval.
2999 stop_ost || error "Unable to stop OST1"
3000 lazystatfs $MOUNT || error "lazystatfs failed with one down server"
3002 umount_client $MOUNT || error "Unable to unmount client"
3003 stop_ost2 || error "Unable to stop OST2"
3004 stop_mds || error "Unable to stop MDS"
3005 #writeconf to remove all ost2 traces for subsequent tests
3006 writeconf_or_reformat
3008 run_test 50d "lazystatfs client/server conn race =========================="
3015 start_mds || return 1
3016 #first client should see only one ost
3017 start_ost || return 2
3018 wait_osc_import_state mds ost FULL
3020 # Wait for client to detect down OST
3021 stop_ost || error "Unable to stop OST1"
3022 wait_osc_import_state mds ost DISCONN
3024 mount_client $MOUNT || error "Unable to mount client"
3025 lctl set_param llite.$FSNAME-*.lazystatfs=0
3027 multiop_bg_pause $MOUNT _f
3031 if [ $RC1 -ne 0 ]; then
3032 log "multiop failed $RC1"
3035 sleep $(( $TIMEOUT+1 ))
3037 [ $? -ne 0 ] && error "process isn't sleep"
3038 start_ost || error "Unable to start OST1"
3039 wait $pid || error "statfs failed"
3042 umount_client $MOUNT || error "Unable to unmount client"
3043 stop_ost || error "Unable to stop OST1"
3044 stop_mds || error "Unable to stop MDS"
3046 run_test 50e "normal statfs all servers down =========================="
3051 CONN_PROC="osc.$FSNAME-OST0001-osc-[M]*.ost_server_uuid"
3053 start_mds || error "Unable to start mds"
3054 #first client should see only one ost
3055 start_ost || error "Unable to start OST1"
3056 wait_osc_import_state mds ost FULL
3058 start_ost2 || error "Unable to start OST2"
3059 wait_osc_import_state mds ost2 FULL
3061 # Wait for client to detect down OST
3062 stop_ost2 || error "Unable to stop OST2"
3064 wait_osc_import_state mds ost2 DISCONN
3065 mount_client $MOUNT || error "Unable to mount client"
3066 lctl set_param llite.$FSNAME-*.lazystatfs=0
3068 multiop_bg_pause $MOUNT _f
3072 if [ $RC1 -ne 0 ]; then
3073 log "lazystatfs multiop failed $RC1"
3076 sleep $(( $TIMEOUT+1 ))
3078 [ $? -ne 0 ] && error "process isn't sleep"
3079 start_ost2 || error "Unable to start OST2"
3080 wait $pid || error "statfs failed"
3081 stop_ost2 || error "Unable to stop OST2"
3084 umount_client $MOUNT || error "Unable to unmount client"
3085 stop_ost || error "Unable to stop OST1"
3086 stop_mds || error "Unable to stop MDS"
3087 #writeconf to remove all ost2 traces for subsequent tests
3088 writeconf_or_reformat
3090 run_test 50f "normal statfs one server in down =========================="
3093 [ "$OSTCOUNT" -lt "2" ] && skip_env "$OSTCOUNT < 2, skipping" && return
3095 start_ost2 || error "Unable to start OST2"
3096 wait_osc_import_state mds ost2 FULL
3097 wait_osc_import_state client ost2 FULL
3099 local PARAM="${FSNAME}-OST0001.osc.active"
3101 $LFS setstripe -c -1 $DIR/$tfile || error "Unable to lfs setstripe"
3102 do_facet mgs $LCTL conf_param $PARAM=0 || error "Unable to deactivate OST"
3104 umount_client $MOUNT || error "Unable to unmount client"
3105 mount_client $MOUNT || error "Unable to mount client"
3106 # This df should not cause a panic
3109 do_facet mgs $LCTL conf_param $PARAM=1 || error "Unable to activate OST"
3111 umount_client $MOUNT || error "Unable to unmount client"
3112 stop_ost2 || error "Unable to stop OST2"
3113 stop_ost || error "Unable to stop OST1"
3114 stop_mds || error "Unable to stop MDS"
3115 #writeconf to remove all ost2 traces for subsequent tests
3116 writeconf_or_reformat
3118 run_test 50g "deactivated OST should not cause panic====================="
3122 # prepare MDT/OST, make OSC inactive for OST1
3123 [ "$OSTCOUNT" -lt "2" ] && skip_env "$OSTCOUNT < 2, skipping" && return
3124 do_facet ost1 "$TUNEFS --param osc.active=0 `ostdevname 1`" ||
3125 error "tunefs OST1 failed"
3126 start_mds || error "Unable to start MDT"
3127 start_ost || error "Unable to start OST1"
3128 start_ost2 || error "Unable to start OST2"
3129 mount_client $MOUNT || error "client start failed"
3133 # activatate OSC for OST1
3134 local TEST="$LCTL get_param -n osc.${FSNAME}-OST0000-osc-[!M]*.active"
3135 set_conf_param_and_check client \
3136 "$TEST" "${FSNAME}-OST0000.osc.active" 1 ||
3137 error "Unable to activate OST1"
3139 mkdir -p $DIR/$tdir/2
3140 $LFS setstripe -c -1 -i 0 $DIR/$tdir/2
3141 sleep 1 && echo "create a file after OST1 is activated"
3143 createmany -o $DIR/$tdir/2/$tfile-%d 1
3145 # check OSC import is working
3146 stat $DIR/$tdir/2/* >/dev/null 2>&1 ||
3147 error "some OSC imports are still not connected"
3150 umount_client $MOUNT || error "Unable to umount client"
3151 stop_ost2 || error "Unable to stop OST2"
3154 run_test 50h "LU-642: activate deactivated OST ==="
3157 local LOCAL_TIMEOUT=20
3161 check_mount || return 1
3164 $LFS setstripe -c -1 $MOUNT/d1
3165 #define OBD_FAIL_MDS_REINT_DELAY 0x142
3166 do_facet $SINGLEMDS "lctl set_param fail_loc=0x142"
3167 touch $MOUNT/d1/f1 &
3170 start_ost2 || return 2
3172 stop_ost2 || return 3
3174 #writeconf to remove all ost2 traces for subsequent tests
3175 writeconf_or_reformat
3177 run_test 51 "Verify that mdt_reint handles RMF_MDT_MD correctly when an OST is added"
3186 do_node $node mkdir -p $dest
3187 [ $? -eq 0 ] || { error "Unable to create directory"; return 1; }
3189 do_node $node 'tar cf - '$@' | tar xf - -C '$dest';
3190 [ \"\${PIPESTATUS[*]}\" = \"0 0\" ] || exit 1'
3191 [ $? -eq 0 ] || { error "Unable to tar files"; return 2; }
3193 do_node $node 'getfattr -d -m "[a-z]*\\." '$@' > '$xattrs
3194 [ $? -eq 0 ] || { error "Unable to read xattrs"; return 3; }
3204 local backup2=${TMP}/backup2
3206 do_node $node mkdir -p $backup2
3207 [ $? -eq 0 ] || { error "Unable to create directory"; return 1; }
3209 do_node $node 'tar cf - '$@' | tar xf - -C '$backup2';
3210 [ \"\${PIPESTATUS[*]}\" = \"0 0\" ] || exit 1'
3211 [ $? -eq 0 ] || { error "Unable to tar files to diff"; return 2; }
3213 do_node $node "diff -rq $backup $backup2"
3214 [ $? -eq 0 ] || { error "contents differ"; return 3; }
3216 local xattrs2=${TMP}/xattrs2
3217 do_node $node 'getfattr -d -m "[a-z]*\\." '$@' > '$xattrs2
3218 [ $? -eq 0 ] || { error "Unable to read xattrs to diff"; return 4; }
3220 do_node $node "diff $xattrs $xattrs2"
3221 [ $? -eq 0 ] || { error "xattrs differ"; return 5; }
3223 do_node $node "rm -rf $backup2 $xattrs2"
3224 [ $? -eq 0 ] || { error "Unable to delete temporary files"; return 6; }
3228 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
3229 skip "Only applicable to ldiskfs-based MDTs"
3234 [ $? -eq 0 ] || { error "Unable to start MDS"; return 1; }
3236 [ $? -eq 0 ] || { error "Unable to start OST1"; return 2; }
3238 [ $? -eq 0 ] || { error "Unable to mount client"; return 3; }
3241 local ost1mnt=$(facet_mntpt ost1)
3242 local ost1node=$(facet_active_host ost1)
3243 local ost1tmp=$TMP/conf52
3247 [ $? -eq 0 ] || { error "Unable to create tdir"; return 4; }
3248 touch $TMP/modified_first
3249 [ $? -eq 0 ] || { error "Unable to create temporary file"; return 5; }
3250 local mtime=$(stat -c %Y $TMP/modified_first)
3251 do_node $ost1node "mkdir -p $ost1tmp && touch -m -d @$mtime $ost1tmp/modified_first"
3253 [ $? -eq 0 ] || { error "Unable to create temporary file"; return 6; }
3256 $LFS setstripe -c -1 -S 1M $DIR/$tdir
3257 [ $? -eq 0 ] || { error "lfs setstripe failed"; return 7; }
3259 for (( i=0; i < nrfiles; i++ )); do
3260 multiop $DIR/$tdir/$tfile-$i Ow1048576w1048576w524288c
3261 [ $? -eq 0 ] || { error "multiop failed"; return 8; }
3267 echo backup files to $TMP/files
3268 local files=$(find $DIR/$tdir -type f -newer $TMP/modified_first)
3269 copy_files_xattrs `hostname` $TMP/files $TMP/file_xattrs $files
3270 [ $? -eq 0 ] || { error "Unable to copy files"; return 9; }
3272 umount_client $MOUNT
3273 [ $? -eq 0 ] || { error "Unable to umount client"; return 10; }
3275 [ $? -eq 0 ] || { error "Unable to stop ost1"; return 11; }
3277 echo mount ost1 as ldiskfs
3278 do_node $ost1node mkdir -p $ost1mnt
3279 [ $? -eq 0 ] || { error "Unable to create $ost1mnt"; return 23; }
3280 if ! do_node $ost1node test -b $ost1_dev; then
3283 do_node $ost1node mount -t $(facet_fstype ost1) $loop $ost1_dev \
3285 [ $? -eq 0 ] || { error "Unable to mount ost1 as ldiskfs"; return 12; }
3288 echo backup objects to $ost1tmp/objects
3289 local objects=$(do_node $ost1node 'find '$ost1mnt'/O/[0-9]* -type f'\
3290 '-size +0 -newer '$ost1tmp'/modified_first -regex ".*\/[0-9]+"')
3291 copy_files_xattrs $ost1node $ost1tmp/objects $ost1tmp/object_xattrs \
3293 [ $? -eq 0 ] || { error "Unable to copy objects"; return 13; }
3295 # move objects to lost+found
3296 do_node $ost1node 'mv '$objects' '${ost1mnt}'/lost+found'
3297 [ $? -eq 0 ] || { error "Unable to move objects"; return 14; }
3300 do_node $ost1node "ll_recover_lost_found_objs -d $ost1mnt/lost+found"
3301 [ $? -eq 0 ] || { error "ll_recover_lost_found_objs failed"; return 15; }
3303 # compare restored objects against saved ones
3304 diff_files_xattrs $ost1node $ost1tmp/objects $ost1tmp/object_xattrs $objects
3305 [ $? -eq 0 ] || { error "Unable to diff objects"; return 16; }
3307 do_node $ost1node "umount $ost1mnt"
3308 [ $? -eq 0 ] || { error "Unable to umount ost1 as ldiskfs"; return 17; }
3311 [ $? -eq 0 ] || { error "Unable to start ost1"; return 18; }
3313 [ $? -eq 0 ] || { error "Unable to mount client"; return 19; }
3316 diff_files_xattrs `hostname` $TMP/files $TMP/file_xattrs $files
3317 [ $? -eq 0 ] || { error "Unable to diff files"; return 20; }
3319 rm -rf $TMP/files $TMP/file_xattrs
3320 [ $? -eq 0 ] || { error "Unable to delete temporary files"; return 21; }
3321 do_node $ost1node "rm -rf $ost1tmp"
3322 [ $? -eq 0 ] || { error "Unable to delete temporary files"; return 22; }
3325 run_test 52 "check recovering objects from lost+found"
3327 # Checks threads_min/max/started for some service
3329 # Arguments: service name (OST or MDT), facet (e.g., ost1, $SINGLEMDS), and a
3330 # parameter pattern prefix like 'ost.*.ost'.
3343 local msg="Insane $modname thread counts"
3344 local ncpts=$(check_cpt_number $facet)
3348 check_mount || return 41
3350 # We need to expand $parampat, but it may match multiple parameters, so
3351 # we'll pick the first one
3352 if ! paramp=$(do_facet $facet "lctl get_param -N ${parampat}.threads_min"|head -1); then
3353 error "Couldn't expand ${parampat}.threads_min parameter name"
3357 # Remove the .threads_min part
3358 paramp=${paramp%.threads_min}
3360 # Check for sanity in defaults
3361 tmin=$(do_facet $facet "lctl get_param -n ${paramp}.threads_min" || echo 0)
3362 tmax=$(do_facet $facet "lctl get_param -n ${paramp}.threads_max" || echo 0)
3363 tstarted=$(do_facet $facet "lctl get_param -n ${paramp}.threads_started" || echo 0)
3364 lassert 23 "$msg (PDSH problems?)" '(($tstarted && $tmin && $tmax))' || return $?
3365 lassert 24 "$msg" '(($tstarted >= $tmin && $tstarted <= $tmax ))' || return $?
3366 nthrs=$(expr $tmax - $tmin)
3367 if [ $nthrs -lt $ncpts ]; then
3373 [ $tmin -eq $tmax -a $tmin -eq $tstarted ] &&
3374 skip_env "module parameter forced $facet thread count" &&
3375 tmin=3 && tmax=$((3 * tmax))
3377 # Check that we can change min/max
3378 do_facet $facet "lctl set_param ${paramp}.threads_min=$((tmin + nthrs))"
3379 do_facet $facet "lctl set_param ${paramp}.threads_max=$((tmax - nthrs))"
3380 tmin2=$(do_facet $facet "lctl get_param -n ${paramp}.threads_min" || echo 0)
3381 tmax2=$(do_facet $facet "lctl get_param -n ${paramp}.threads_max" || echo 0)
3382 lassert 25 "$msg" '(($tmin2 == ($tmin + $nthrs) && $tmax2 == ($tmax - $nthrs)))' || return $?
3384 # Check that we can set min/max to the same value
3385 tmin=$(do_facet $facet "lctl get_param -n ${paramp}.threads_min" || echo 0)
3386 do_facet $facet "lctl set_param ${paramp}.threads_max=$tmin"
3387 tmin2=$(do_facet $facet "lctl get_param -n ${paramp}.threads_min" || echo 0)
3388 tmax2=$(do_facet $facet "lctl get_param -n ${paramp}.threads_max" || echo 0)
3389 lassert 26 "$msg" '(($tmin2 == $tmin && $tmax2 == $tmin))' || return $?
3391 # Check that we can't set max < min
3392 do_facet $facet "lctl set_param ${paramp}.threads_max=$((tmin - 1))"
3393 tmin2=$(do_facet $facet "lctl get_param -n ${paramp}.threads_min" || echo 0)
3394 tmax2=$(do_facet $facet "lctl get_param -n ${paramp}.threads_max" || echo 0)
3395 lassert 27 "$msg" '(($tmin2 <= $tmax2))' || return $?
3397 # We need to ensure that we get the module options desired; to do this
3398 # we set LOAD_MODULES_REMOTE=true and we call setmodopts below.
3399 LOAD_MODULES_REMOTE=true
3402 local newvalue="${opts}=$(expr $basethr \* $ncpts)"
3403 setmodopts -a $modname "$newvalue" oldvalue
3407 check_mount || return 41
3409 # Restore previous setting of MODOPTS_*
3410 setmodopts $modname "$oldvalue"
3412 # Check that $opts took
3413 tmin=$(do_facet $facet "lctl get_param -n ${paramp}.threads_min")
3414 tmax=$(do_facet $facet "lctl get_param -n ${paramp}.threads_max")
3415 tstarted=$(do_facet $facet "lctl get_param -n ${paramp}.threads_started")
3416 lassert 28 "$msg" '(($tstarted >= $tmin && $tstarted <= $tmax ))' || return $?
3425 thread_sanity OST ost1 'ost.*.ost' 'oss_num_threads' '16'
3428 run_test 53a "check OSS thread count params"
3432 local mds=$(do_facet $SINGLEMDS "lctl get_param -N mds.*.*.threads_max \
3434 if [ -z "$mds" ]; then
3435 #running this on an old MDT
3436 thread_sanity MDT $SINGLEMDS 'mdt.*.*.' 'mdt_num_threads' 16
3438 thread_sanity MDT $SINGLEMDS 'mds.*.*.' 'mds_num_threads' 16
3442 run_test 53b "check MDS thread count params"
3445 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
3446 skip "Only applicable to ldiskfs-based MDTs"
3450 do_rpc_nodes $(facet_host ost1) run_llverdev $(ostdevname 1) -p
3451 [ $? -eq 0 ] || error "llverdev failed!"
3454 run_test 54a "test llverdev and partial verify of device"
3457 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
3458 skip "Only applicable to ldiskfs-based MDTs"
3463 run_llverfs $MOUNT -p
3464 [ $? -eq 0 ] || error "llverfs failed!"
3467 run_test 54b "test llverfs and partial verify of filesystem"
3471 local max_ost_index=$1
3472 echo -n $(((max_ost_index + 1) * 8))
3476 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
3477 skip "Only applicable to ldiskfs-based MDTs"
3481 local mdsdev=$(mdsdevname 1)
3482 local mdsvdev=$(mdsvdevname 1)
3486 add mds1 $(mkfs_opts mds1 ${mdsdev}) --reformat $mdsdev \
3488 add ost1 $(mkfs_opts ost1 $(ostdevname 1)) --index=$i \
3489 --reformat $(ostdevname 1) $(ostvdevname 1)
3495 echo checking size of lov_objid for ost index $i
3496 LOV_OBJID_SIZE=$(do_facet mds1 "$DEBUGFS -R 'stat lov_objid' $mdsdev 2>/dev/null" | grep ^User | awk '{print $6}')
3497 if [ "$LOV_OBJID_SIZE" != $(lov_objid_size $i) ]; then
3498 error "lov_objid size has to be $(lov_objid_size $i), not $LOV_OBJID_SIZE"
3500 echo ok, lov_objid size is correct: $LOV_OBJID_SIZE
3507 run_test 55 "check lov_objid size"
3510 local mds_journal_size_orig=$MDSJOURNALSIZE
3514 for num in $(seq 1 $MDSCOUNT); do
3515 add mds${num} $(mkfs_opts mds${num} $(mdsdevname $num)) \
3516 --reformat $(mdsdevname $num) $(mdsvdevname $num)
3518 add ost1 $(mkfs_opts ost1 $(ostdevname 1)) --index=1000 --reformat \
3519 $(ostdevname 1) $(ostvdevname 1)
3520 add ost2 $(mkfs_opts ost2 $(ostdevname 2)) --index=10000 --reformat \
3521 $(ostdevname 2) $(ostvdevname 2)
3525 start_ost2 || error "Unable to start second ost"
3526 mount_client $MOUNT || error "Unable to mount client"
3530 MDSJOURNALSIZE=$mds_journal_size_orig
3533 run_test 56 "check big indexes"
3535 test_57a() { # bug 22656
3536 local NID=$(do_facet ost1 "$LCTL get_param nis" | tail -1 | awk '{print $1}')
3537 writeconf_or_reformat
3538 [ $(facet_fstype ost1) == zfs ] && import_zpool ost1
3539 do_facet ost1 "$TUNEFS --failnode=$NID `ostdevname 1`" || error "tunefs failed"
3541 start_ost && error "OST registration from failnode should fail"
3544 run_test 57a "initial registration from failnode should fail (should return errs)"
3547 local NID=$(do_facet ost1 "$LCTL get_param nis" | tail -1 | awk '{print $1}')
3548 writeconf_or_reformat
3549 [ $(facet_fstype ost1) == zfs ] && import_zpool ost1
3550 do_facet ost1 "$TUNEFS --servicenode=$NID `ostdevname 1`" || error "tunefs failed"
3552 start_ost || error "OST registration from servicenode should not fail"
3555 run_test 57b "initial registration from servicenode should not fail"
3558 do_facet mgs $LCTL get_param mgs.MGS.live.$FSNAME | grep OST | wc -l
3561 test_58() { # bug 22658
3562 if [ $(facet_fstype mds) != ldiskfs ]; then
3563 skip "Only applicable to ldiskfs-based MDTs"
3568 createmany -o $DIR/$tdir/$tfile-%d 100
3569 # make sure that OSTs do not cancel llog cookies before we unmount the MDS
3570 #define OBD_FAIL_OBD_LOG_CANCEL_NET 0x601
3571 do_facet $SINGLEMDS "lctl set_param fail_loc=0x601"
3572 unlinkmany $DIR/$tdir/$tfile-%d 100
3575 local MNTDIR=$(facet_mntpt $SINGLEMDS)
3576 local devname=$(mdsdevname ${SINGLEMDS//mds/})
3578 if ! do_facet $SINGLEMDS "test -b $devname"; then
3582 # remove all files from the OBJECTS dir
3583 do_facet $SINGLEMDS "mount -t ldiskfs $opts $devname $MNTDIR"
3584 do_facet $SINGLEMDS "find $MNTDIR/O/1/d* -type f -delete"
3585 do_facet $SINGLEMDS "umount -d $MNTDIR"
3586 # restart MDS with missing llog files
3588 do_facet mds "lctl set_param fail_loc=0"
3591 run_test 58 "missing llog files must not prevent MDT from mounting"
3594 start_mgsmds >> /dev/null
3595 local C1=$(count_osts)
3596 if [ $C1 -eq 0 ]; then
3597 start_ost >> /dev/null
3601 echo "original ost count: $C1 (expect > 0)"
3602 [ $C1 -gt 0 ] || error "No OSTs in $FSNAME log"
3603 start_mgsmds -o writeconf >> /dev/null || error "MDT start failed"
3604 local C2=$(count_osts)
3605 echo "after mdt writeconf count: $C2 (expect 0)"
3606 [ $C2 -gt 0 ] && error "MDT writeconf should erase OST logs"
3607 echo "OST start without writeconf should fail:"
3608 start_ost >> /dev/null && error "OST start without writeconf didn't fail"
3609 echo "OST start with writeconf should succeed:"
3610 start_ost -o writeconf >> /dev/null || error "OST1 start failed"
3611 local C3=$(count_osts)
3612 echo "after ost writeconf count: $C3 (expect 1)"
3613 [ $C3 -eq 1 ] || error "new OST writeconf should add:"
3614 start_ost2 -o writeconf >> /dev/null || error "OST2 start failed"
3615 local C4=$(count_osts)
3616 echo "after ost2 writeconf count: $C4 (expect 2)"
3617 [ $C4 -eq 2 ] || error "OST2 writeconf should add log"
3618 stop_ost2 >> /dev/null
3619 cleanup_nocli >> /dev/null
3620 #writeconf to remove all ost2 traces for subsequent tests
3621 writeconf_or_reformat
3623 run_test 59 "writeconf mount option"
3625 test_60() { # LU-471
3628 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
3629 skip "Only applicable to ldiskfs-based MDTs"
3633 for num in $(seq $MDSCOUNT); do
3634 add mds${num} $(mkfs_opts mds${num} $(mdsdevname $num)) \
3635 --mkfsoptions='\" -E stride=64 -O ^uninit_bg\"' \
3636 --reformat $(mdsdevname $num) $(mdsvdevname $num) ||
3640 dump=$(do_facet $SINGLEMDS dumpe2fs $(mdsdevname 1))
3642 [ $rc -eq 0 ] || error "dumpe2fs $(mdsdevname 1) failed"
3644 # MDT default has dirdata feature
3645 echo $dump | grep dirdata > /dev/null || error "dirdata is not set"
3646 # we disable uninit_bg feature
3647 echo $dump | grep uninit_bg > /dev/null && error "uninit_bg is set"
3648 # we set stride extended options
3649 echo $dump | grep stride > /dev/null || error "stride is not set"
3652 run_test 60 "check mkfs.lustre --mkfsoptions -E -O options setting"
3655 local reformat=false
3657 [ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.1.53) ] ||
3658 { skip "Need MDS version at least 2.1.53"; return 0; }
3660 if [ $(facet_fstype $SINGLEMDS) == ldiskfs ] &&
3661 ! large_xattr_enabled; then
3663 LDISKFS_MKFS_OPTS+=" -O large_xattr"
3665 for num in $(seq $MDSCOUNT); do
3666 add mds${num} $(mkfs_opts mds$num $(mdsdevname $num)) \
3667 --reformat $(mdsdevname $num) $(mdsvdevname $num) ||
3668 error "add mds $num failed"
3672 setup_noconfig || error "setting up the filesystem failed"
3673 client_up || error "starting client failed"
3675 local file=$DIR/$tfile
3678 local large_value="$(generate_string $(max_xattr_size))"
3679 local small_value="bar"
3681 local name="trusted.big"
3682 log "save large xattr $name on $file"
3683 setfattr -n $name -v $large_value $file ||
3684 error "saving $name on $file failed"
3686 local new_value=$(get_xattr_value $name $file)
3687 [[ "$new_value" != "$large_value" ]] &&
3688 error "$name different after saving"
3690 log "shrink value of $name on $file"
3691 setfattr -n $name -v $small_value $file ||
3692 error "shrinking value of $name on $file failed"
3694 new_value=$(get_xattr_value $name $file)
3695 [[ "$new_value" != "$small_value" ]] &&
3696 error "$name different after shrinking"
3698 log "grow value of $name on $file"
3699 setfattr -n $name -v $large_value $file ||
3700 error "growing value of $name on $file failed"
3702 new_value=$(get_xattr_value $name $file)
3703 [[ "$new_value" != "$large_value" ]] &&
3704 error "$name different after growing"
3706 log "check value of $name on $file after remounting MDS"
3708 new_value=$(get_xattr_value $name $file)
3709 [[ "$new_value" != "$large_value" ]] &&
3710 error "$name different after remounting MDS"
3712 log "remove large xattr $name from $file"
3713 setfattr -x $name $file || error "removing $name from $file failed"
3718 LDISKFS_MKFS_OPTS=${LDISKFS_MKFS_OPTS% -O large_xattr}
3722 run_test 61 "large xattr"
3725 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
3726 skip "Only applicable to ldiskfs-based MDTs"
3731 local mdsdev=$(mdsdevname 1)
3732 local ostdev=$(ostdevname 1)
3734 [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.2.51) ]] ||
3735 { skip "Need MDS version at least 2.2.51"; return 0; }
3737 echo "disable journal for mds"
3738 do_facet mds tune2fs -O ^has_journal $mdsdev || error "tune2fs failed"
3739 start_mds && error "MDT start should fail"
3740 echo "disable journal for ost"
3741 do_facet ost1 tune2fs -O ^has_journal $ostdev || error "tune2fs failed"
3742 start_ost && error "OST start should fail"
3743 cleanup || return $?
3746 run_test 62 "start with disabled journal"
3749 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
3750 skip "Only applicable to ldiskfs-based MDTs"
3754 local inode_slab=$(do_facet $SINGLEMDS \
3755 "awk '/ldiskfs_inode_cache/ { print \\\$5 }' /proc/slabinfo")
3756 if [ -z "$inode_slab" ]; then
3757 skip "ldiskfs module has not been loaded"
3761 echo "$inode_slab ldisk inodes per page"
3762 [ "$inode_slab" -ge "3" ] ||
3763 error "ldisk inode size is too big, $inode_slab objs per page"
3766 run_test 63 "Verify each page can at least hold 3 ldisk inodes"
3771 start_ost2 || error "Unable to start second ost"
3772 mount_client $MOUNT || error "Unable to mount client"
3773 stop_ost2 || error "Unable to stop second ost"
3775 $LFS df --lazy || error "lfs df failed"
3776 cleanup || return $?
3777 #writeconf to remove all ost2 traces for subsequent tests
3778 writeconf_or_reformat
3780 run_test 64 "check lfs df --lazy "
3782 test_65() { # LU-2237
3783 # Currently, the test is only valid for ldiskfs backend
3784 [ "$(facet_fstype $SINGLEMDS)" != "ldiskfs" ] &&
3785 skip "non-ldiskfs backend" && return
3787 local devname=$(mdsdevname ${SINGLEMDS//mds/})
3788 local brpt=$(facet_mntpt brpt)
3791 if ! do_facet $SINGLEMDS "test -b $devname"; then
3796 local obj=$(do_facet $SINGLEMDS \
3797 "$DEBUGFS -c -R \\\"stat last_rcvd\\\" $devname" |
3799 if [ -z "$obj" ]; then
3800 # The MDT may be just re-formatted, mount the MDT for the
3801 # first time to guarantee the "last_rcvd" file is there.
3802 start_mds || error "fail to mount the MDS for the first time"
3806 # remove the "last_rcvd" file
3807 do_facet $SINGLEMDS "mkdir -p $brpt"
3808 do_facet $SINGLEMDS \
3809 "mount -t $(facet_fstype $SINGLEMDS) $opts $devname $brpt"
3810 do_facet $SINGLEMDS "rm -f ${brpt}/last_rcvd"
3811 do_facet $SINGLEMDS "umount -d $brpt"
3813 # restart MDS, the "last_rcvd" file should be recreated.
3814 start_mds || error "fail to restart the MDS"
3816 obj=$(do_facet $SINGLEMDS \
3817 "$DEBUGFS -c -R \\\"stat last_rcvd\\\" $devname" | grep Inode)
3818 [ -n "$obj" ] || error "fail to re-create the last_rcvd"
3820 run_test 65 "re-create the lost last_rcvd file when server mount"
3823 [[ $(lustre_version_code mgs) -ge $(version_code 2.3.59) ]] ||
3824 { skip "Need MGS version at least 2.3.59"; return 0; }
3827 local OST1_NID=$(do_facet ost1 $LCTL list_nids | head -1)
3828 local MDS_NID=$(do_facet $SINGLEMDS $LCTL list_nids | head -1)
3830 echo "replace_nids should fail if MDS, OSTs and clients are UP"
3831 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 $OST1_NID &&
3832 error "replace_nids fail"
3834 umount_client $MOUNT || error "unmounting client failed"
3835 echo "replace_nids should fail if MDS and OSTs are UP"
3836 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 $OST1_NID &&
3837 error "replace_nids fail"
3840 echo "replace_nids should fail if MDS is UP"
3841 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 $OST1_NID &&
3842 error "replace_nids fail"
3844 stop_mds || error "stopping mds failed"
3846 if combined_mgs_mds; then
3847 start_mdt 1 "-o nosvc" ||
3848 error "starting mds with nosvc option failed"
3851 echo "command should accept two parameters"
3852 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 &&
3853 error "command should accept two params"
3855 echo "correct device name should be passed"
3856 do_facet mgs $LCTL replace_nids $FSNAME-WRONG0000 $OST1_NID &&
3857 error "wrong devname"
3859 echo "wrong nids list should not destroy the system"
3860 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 "wrong nids list" &&
3863 echo "replace OST nid"
3864 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 $OST1_NID ||
3865 error "replace nids failed"
3867 echo "command should accept two parameters"
3868 do_facet mgs $LCTL replace_nids $FSNAME-MDT0000 &&
3869 error "command should accept two params"
3871 echo "wrong nids list should not destroy the system"
3872 do_facet mgs $LCTL replace_nids $FSNAME-MDT0000 "wrong nids list" &&
3875 echo "replace MDS nid"
3876 do_facet mgs $LCTL replace_nids $FSNAME-MDT0000 $MDS_NID ||
3877 error "replace nids failed"
3879 if ! combined_mgs_mds ; then
3886 check_mount || error "error after nid replace"
3887 cleanup || error "cleanup failed"
3890 run_test 66 "replace nids"
3892 test_67() { #LU-2950
3893 local legacy="$TMP/legacy_lnet_config"
3894 local new="$TMP/new_routes_test"
3895 local out="$TMP/config_out_file"
3896 local verify="$TMP/conv_verify"
3897 local verify_conf="$TMP/conf_verify"
3899 # Create the legacy file that will be run through the
3900 # lustre_routes_conversion script
3901 cat <<- LEGACY_LNET_CONFIG > $legacy
3902 tcp1 23 192.168.213.1@tcp:1; tcp5 34 193.30.4.3@tcp:4;
3903 tcp2 54 10.1.3.2@tcp;
3904 tcp3 10.3.4.3@tcp:3;
3908 # Create the verification file to verify the output of
3909 # lustre_routes_conversion script against.
3910 cat <<- VERIFY_LNET_CONFIG > $verify
3911 tcp1: { gateway: 192.168.213.1@tcp, hop: 23, priority: 1 }
3912 tcp5: { gateway: 193.30.4.3@tcp, hop: 34, priority: 4 }
3913 tcp2: { gateway: 10.1.3.2@tcp, hop: 54 }
3914 tcp3: { gateway: 10.3.4.3@tcp, priority: 3 }
3915 tcp4: { gateway: 10.3.3.4@tcp }
3918 # Create the verification file to verify the output of
3919 # lustre_routes_config script against
3920 cat <<- VERIFY_LNET_CONFIG > $verify_conf
3921 lctl --net tcp1 add_route 192.168.213.1@tcp 23 1
3922 lctl --net tcp5 add_route 193.30.4.3@tcp 34 4
3923 lctl --net tcp2 add_route 10.1.3.2@tcp 54 4
3924 lctl --net tcp3 add_route 10.3.4.3@tcp 1 3
3925 lctl --net tcp4 add_route 10.3.3.4@tcp 1 3
3928 lustre_routes_conversion $legacy $new > /dev/null
3929 if [ -f $new ]; then
3930 # verify the conversion output
3931 cmp -s $new $verify > /dev/null
3932 if [ $? -eq 1 ]; then
3933 error "routes conversion failed"
3936 lustre_routes_config --dry-run --verbose $new > $out
3937 # check that the script succeeded
3938 cmp -s $out $verify_conf > /dev/null
3939 if [ $? -eq 1 ]; then
3940 error "routes config failed"
3943 error "routes conversion test failed"
3945 # remove generated files
3946 rm -f $new $legacy $verify $verify_conf $out
3948 run_test 67 "test routes conversion and configuration"
3956 [ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.4.53) ] ||
3957 { skip "Need MDS version at least 2.4.53"; return 0; }
3959 umount_client $MOUNT || error "umount client failed"
3961 start_mdt 1 || error "MDT start failed"
3964 # START-END - the sequences we'll be reserving
3965 START=$(do_facet $SINGLEMDS \
3966 lctl get_param -n seq.ctl*.space | awk -F'[[ ]' '{print $2}')
3967 END=$((START + (1 << 30)))
3968 do_facet $SINGLEMDS \
3969 lctl set_param seq.ctl*.fldb="[$START-$END\):0:mdt"
3971 # reset the sequences MDT0000 has already assigned
3972 do_facet $SINGLEMDS \
3973 lctl set_param seq.srv*MDT0000.space=clear
3975 # remount to let the client allocate new sequence
3976 mount_client $MOUNT || error "mount client failed"
3979 do_facet $SINGLEMDS \
3980 lctl get_param seq.srv*MDT0000.space
3981 $LFS path2fid $DIR/$tfile
3983 local old_ifs="$IFS"
3985 fid=($($LFS path2fid $DIR/$tfile))
3989 if [[ $seq < $END ]]; then
3990 error "used reserved sequence $seq?"
3992 cleanup || return $?
3994 run_test 68 "be able to reserve specific sequences in FLDB"
3997 local server_version=$(lustre_version_code $SINGLEMDS)
3999 [[ $server_version -lt $(version_code 2.4.2) ]] &&
4000 skip "Need MDS version at least 2.4.2" && return
4002 [[ $server_version -ge $(version_code 2.4.50) ]] &&
4003 [[ $server_version -lt $(version_code 2.5.0) ]] &&
4004 skip "Need MDS version at least 2.5.0" && return
4008 # use OST0000 since it probably has the most creations
4009 local OSTNAME=$(ostname_from_index 0)
4010 local mdtosc_proc1=$(get_mdtosc_proc_path mds1 $OSTNAME)
4011 local last_id=$(do_facet mds1 lctl get_param -n \
4012 osc.$mdtosc_proc1.prealloc_last_id)
4014 # Want to have OST LAST_ID over 1.5 * OST_MAX_PRECREATE to
4015 # verify that the LAST_ID recovery is working properly. If
4016 # not, then the OST will refuse to allow the MDS connect
4017 # because the LAST_ID value is too different from the MDS
4018 #define OST_MAX_PRECREATE=20000
4019 local num_create=$((20000 * 5))
4022 $LFS setstripe -i 0 $DIR/$tdir
4023 createmany -o $DIR/$tdir/$tfile- $num_create ||
4024 error "createmany: failed to create $num_create files: $?"
4025 # delete all of the files with objects on OST0 so the
4026 # filesystem is not inconsistent later on
4027 $LFS find $MOUNT --ost 0 | xargs rm
4029 stop_ost || error "OST0 stop failure"
4030 add ost1 $(mkfs_opts ost1 $(ostdevname 1)) --reformat --replace \
4031 $(ostdevname 1) $(ostvdevname 1) ||
4032 error "reformat and replace $ostdev failed"
4033 start_ost || error "OST0 restart failure"
4034 wait_osc_import_state mds ost FULL
4036 touch $DIR/$tdir/$tfile-last || error "create file after reformat"
4037 local idx=$($LFS getstripe -i $DIR/$tdir/$tfile-last)
4038 [ $idx -ne 0 ] && error "$DIR/$tdir/$tfile-last on $idx not 0" || true
4042 run_test 69 "replace an OST with the same index"
4045 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
4050 start_mdt 1 || error "MDT0 start fail"
4052 start_ost || error "OST0 start fail"
4054 start_mdt 2 || error "MDT1 start fail"
4056 mount_client $MOUNT || error "mount client fails"
4058 mkdir -p $DIR/$tdir || error "create dir fail"
4060 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
4061 error "create remote dir fail"
4063 rm -rf $DIR/$tdir || error "delete dir fail"
4064 cleanup || return $?
4066 run_test 70a "start MDT0, then OST, then MDT1"
4069 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
4072 start_ost || error "OST0 start fail"
4074 start_mdt 1 || error "MDT0 start fail"
4075 start_mdt 2 || error "MDT1 start fail"
4077 mount_client $MOUNT || error "mount client fails"
4079 mkdir -p $DIR/$tdir || error "create dir fail"
4081 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
4082 error "create remote dir fail"
4084 rm -rf $DIR/$tdir || error "delete dir fail"
4086 cleanup || return $?
4088 run_test 70b "start OST, MDT1, MDT0"
4091 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
4094 start_mdt 1 || error "MDT0 start fail"
4095 start_mdt 2 || error "MDT1 start fail"
4096 start_ost || error "OST0 start fail"
4098 mount_client $MOUNT || error "mount client fails"
4099 stop_mdt 1 || error "MDT1 start fail"
4101 local mdc_for_mdt1=$($LCTL dl | grep MDT0000-mdc | awk '{print $4}')
4102 echo "deactivate $mdc_for_mdt1"
4103 $LCTL --device $mdc_for_mdt1 deactivate || return 1
4105 mkdir -p $DIR/$tdir && error "mkdir succeed"
4107 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir &&
4108 error "create remote dir succeed"
4110 cleanup || return $?
4112 run_test 70c "stop MDT0, mkdir fail, create remote dir fail"
4115 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
4118 start_mdt 1 || error "MDT0 start fail"
4119 start_mdt 2 || error "MDT1 start fail"
4120 start_ost || error "OST0 start fail"
4122 mount_client $MOUNT || error "mount client fails"
4124 stop_mdt 2 || error "MDT1 start fail"
4126 local mdc_for_mdt2=$($LCTL dl | grep MDT0001-mdc |
4128 echo "deactivate $mdc_for_mdt2"
4129 $LCTL --device $mdc_for_mdt2 deactivate ||
4130 error "set $mdc_for_mdt2 deactivate failed"
4132 mkdir -p $DIR/$tdir || error "mkdir fail"
4133 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir &&
4134 error "create remote dir succeed"
4136 rm -rf $DIR/$tdir || error "delete dir fail"
4138 cleanup || return $?
4140 run_test 70d "stop MDT1, mkdir succeed, create remote dir fail"
4143 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
4144 if combined_mgs_mds; then
4145 skip "needs separate MGS/MDT" && return
4149 start_mdt 1 || error "MDT0 start fail"
4150 start_ost || error "OST0 start fail"
4151 start_mdt 2 || error "MDT1 start fail"
4152 start_ost2 || error "OST1 start fail"
4154 mount_client $MOUNT || error "mount client fails"
4156 mkdir -p $DIR/$tdir || error "mkdir fail"
4157 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
4158 error "create remote dir succeed"
4160 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
4161 rm -rf $DIR/$tdir || error "delete dir fail"
4163 umount_client $MOUNT
4164 stop_mdt 1 || error "MDT0 stop fail"
4165 stop_mdt 2 || error "MDT1 stop fail"
4166 stop_ost || error "OST0 stop fail"
4167 stop_ost2 || error "OST1 stop fail"
4169 run_test 71a "start MDT0 OST0, MDT1, OST1"
4172 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
4173 if combined_mgs_mds; then
4174 skip "needs separate MGS/MDT" && return
4178 start_mdt 2 || error "MDT1 start fail"
4179 start_ost || error "OST0 start fail"
4180 start_mdt 1 || error "MDT0 start fail"
4181 start_ost2 || error "OST1 start fail"
4183 mount_client $MOUNT || error "mount client fails"
4185 mkdir -p $DIR/$tdir || error "mkdir fail"
4186 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
4187 error "create remote dir succeed"
4189 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
4190 rm -rf $DIR/$tdir || error "delete dir fail"
4192 umount_client $MOUNT
4193 stop_mdt 1 || error "MDT0 stop fail"
4194 stop_mdt 2 || error "MDT1 stop fail"
4195 stop_ost || error "OST0 stop fail"
4196 stop_ost2 || error "OST1 stop fail"
4198 run_test 71b "start MDT1, OST0, MDT0, OST1"
4201 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
4202 if combined_mgs_mds; then
4203 skip "needs separate MGS/MDT" && return
4207 start_ost || error "OST0 start fail"
4208 start_ost2 || error "OST1 start fail"
4209 start_mdt 2 || error "MDT1 start fail"
4210 start_mdt 1 || error "MDT0 start fail"
4212 mount_client $MOUNT || error "mount client fails"
4214 mkdir -p $DIR/$tdir || error "mkdir fail"
4215 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
4216 error "create remote dir succeed"
4218 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
4219 rm -rf $DIR/$tdir || error "delete dir fail"
4221 umount_client $MOUNT
4222 stop_mdt 1 || error "MDT0 stop fail"
4223 stop_mdt 2 || error "MDT1 stop fail"
4224 stop_ost || error "OST0 stop fail"
4225 stop_ost2 || error "OST1 stop fail"
4228 run_test 71c "start OST0, OST1, MDT1, MDT0"
4231 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
4232 if combined_mgs_mds; then
4233 skip "needs separate MGS/MDT" && return
4237 start_ost || error "OST0 start fail"
4238 start_mdt 2 || error "MDT0 start fail"
4239 start_mdt 1 || error "MDT0 start fail"
4240 start_ost2 || error "OST1 start fail"
4242 mount_client $MOUNT || error "mount client fails"
4244 mkdir -p $DIR/$tdir || error "mkdir fail"
4245 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
4246 error "create remote dir succeed"
4248 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
4249 rm -rf $DIR/$tdir || error "delete dir fail"
4251 umount_client $MOUNT
4252 stop_mdt 1 || error "MDT0 stop fail"
4253 stop_mdt 2 || error "MDT1 stop fail"
4254 stop_ost || error "OST0 stop fail"
4255 stop_ost2 || error "OST1 stop fail"
4258 run_test 71d "start OST0, MDT1, MDT0, OST1"
4261 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
4262 if combined_mgs_mds; then
4263 skip "needs separate MGS/MDT" && return
4267 start_ost || error "OST0 start fail"
4268 start_mdt 2 || error "MDT1 start fail"
4269 start_ost2 || error "OST1 start fail"
4270 start_mdt 1 || error "MDT0 start fail"
4272 mount_client $MOUNT || error "mount client fails"
4274 mkdir -p $DIR/$tdir || error "mkdir fail"
4275 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
4276 error "create remote dir succeed"
4278 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
4279 rm -rf $DIR/$tdir || error "delete dir fail"
4281 umount_client $MOUNT
4282 stop_mdt 1 || error "MDT0 stop fail"
4283 stop_mdt 2 || error "MDT1 stop fail"
4284 stop_ost || error "OST0 stop fail"
4285 stop_ost2 || error "OST1 stop fail"
4288 run_test 71e "start OST0, MDT1, OST1, MDT0"
4290 test_72() { #LU-2634
4291 local mdsdev=$(mdsdevname 1)
4292 local ostdev=$(ostdevname 1)
4293 local cmd="$E2FSCK -fnvd $mdsdev"
4296 [ "$(facet_fstype $SINGLEMDS)" != "ldiskfs" ] &&
4297 skip "ldiskfs only test" && return
4299 #tune MDT with "-O extents"
4301 for num in $(seq $MDSCOUNT); do
4302 add mds${num} $(mkfs_opts mds$num $(mdsdevname $num)) \
4303 --reformat $(mdsdevname $num) $(mdsvdevname $num) ||
4304 error "add mds $num failed"
4305 $TUNE2FS -O extents $(mdsdevname $num)
4308 add ost1 $(mkfs_opts ost1 $ostdev) --reformat $ostdev ||
4309 error "add $ostdev failed"
4310 start_mgsmds || error "start mds failed"
4311 start_ost || error "start ost failed"
4312 mount_client $MOUNT || error "mount client failed"
4314 #create some short symlinks
4316 createmany -o $DIR/$tdir/$tfile-%d $fn
4317 echo "create $fn short symlinks"
4318 for i in $(seq -w 1 $fn); do
4319 ln -s $DIR/$tdir/$tfile-$i $MOUNT/$tfile-$i
4324 umount_client $MOUNT || error "umount client failed"
4325 stop_mds || error "stop mds failed"
4326 stop_ost || error "stop ost failed"
4329 run_e2fsck $(facet_active_host $SINGLEMDS) $mdsdev "-n"
4331 run_test 72 "test fast symlink with extents flag enabled"
4333 test_73() { #LU-3006
4335 [ $(facet_fstype ost1) == zfs ] && import_zpool ost1
4336 do_facet ost1 "$TUNEFS --failnode=1.2.3.4@$NETTYPE $(ostdevname 1)" ||
4337 error "1st tunefs failed"
4338 start_mgsmds || error "start mds failed"
4339 start_ost || error "start ost failed"
4340 mount_client $MOUNT || error "mount client failed"
4341 lctl get_param -n osc.*OST0000-osc-[^M]*.import | grep failover_nids |
4342 grep 1.2.3.4@$NETTYPE || error "failover nids haven't changed"
4343 umount_client $MOUNT || error "umount client failed"
4347 run_test 73 "failnode to update from mountdata properly"
4349 test_74() { # LU-1606
4350 for TESTPROG in $LUSTRE_TESTS_API_DIR/*.c; do
4351 gcc -Wall -Werror $LUSTRE_TESTS_API_DIR/simple_test.c \
4353 -L$LUSTRE/utils -llustreapi ||
4354 error "client api broken"
4356 cleanup || return $?
4358 run_test 74 "Lustre client api program can compile and link"
4360 test_75() { # LU-2374
4361 [[ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.4.1) ]] &&
4362 skip "Need MDS version at least 2.4.1" && return
4365 local opts_mds="$(mkfs_opts mds1 $(mdsdevname 1)) \
4366 --reformat $(mdsdevname 1) $(mdsvdevname 1)"
4367 local opts_ost="$(mkfs_opts ost1 $(ostdevname 1)) \
4368 --reformat $(ostdevname 1) $(ostvdevname 1)"
4370 #check with default parameters
4371 add mds1 $opts_mds || error "add mds1 failed for default params"
4372 add ost1 $opts_ost || error "add ost1 failed for default params"
4374 opts_mds=$(echo $opts_mds | sed -e "s/--mdt//")
4375 opts_mds=$(echo $opts_mds |
4376 sed -e "s/--index=$index/--index=$index --mdt/")
4377 opts_ost=$(echo $opts_ost | sed -e "s/--ost//")
4378 opts_ost=$(echo $opts_ost |
4379 sed -e "s/--index=$index/--index=$index --ost/")
4381 add mds1 $opts_mds || error "add mds1 failed for new params"
4382 add ost1 $opts_ost || error "add ost1 failed for new params"
4385 run_test 75 "The order of --index should be irrelevant"
4388 [[ $(lustre_version_code mgs) -ge $(version_code 2.4.52) ]] ||
4389 { skip "Need MDS version at least 2.4.52" && return 0; }
4391 local MDMB_PARAM="osc.*.max_dirty_mb"
4392 echo "Change MGS params"
4393 local MAX_DIRTY_MB=$($LCTL get_param -n $MDMB_PARAM |
4395 echo "max_dirty_mb: $MAX_DIRTY_MB"
4396 local NEW_MAX_DIRTY_MB=$((MAX_DIRTY_MB + MAX_DIRTY_MB))
4397 echo "new_max_dirty_mb: $NEW_MAX_DIRTY_MB"
4398 do_facet mgs $LCTL set_param -P $MDMB_PARAM=$NEW_MAX_DIRTY_MB
4399 wait_update $HOSTNAME "lctl get_param -n $MDMB_PARAM |
4400 head -1" $NEW_MAX_DIRTY_MB
4401 MAX_DIRTY_MB=$($LCTL get_param -n $MDMB_PARAM | head -1)
4402 echo "$MAX_DIRTY_MB"
4403 [ $MAX_DIRTY_MB = $NEW_MAX_DIRTY_MB ] ||
4404 error "error while apply max_dirty_mb"
4406 echo "Check the value is stored after remount"
4409 wait_update $HOSTNAME "lctl get_param -n $MDMB_PARAM |
4410 head -1" $NEW_MAX_DIRTY_MB
4411 MAX_DIRTY_MB=$($LCTL get_param -n $MDMB_PARAM | head -1)
4412 [ $MAX_DIRTY_MB = $NEW_MAX_DIRTY_MB ] ||
4413 error "max_dirty_mb is not saved after remount"
4415 echo "Change OST params"
4416 CLIENT_PARAM="obdfilter.*.client_cache_count"
4417 local CLIENT_CACHE_COUNT
4418 CLIENT_CACHE_COUNT=$(do_facet ost1 $LCTL get_param -n $CLIENT_PARAM |
4420 echo "client_cache_count: $CLIENT_CACHE_COUNT"
4421 NEW_CLIENT_CACHE_COUNT=$((CLIENT_CACHE_COUNT+CLIENT_CACHE_COUNT))
4422 echo "new_client_cache_count: $NEW_CLIENT_CACHE_COUNT"
4423 do_facet mgs $LCTL set_param -P $CLIENT_PARAM=$NEW_CLIENT_CACHE_COUNT
4424 wait_update $(facet_host ost1) "lctl get_param -n $CLIENT_PARAM |
4425 head -1" $NEW_CLIENT_CACHE_COUNT
4426 CLIENT_CACHE_COUNT=$(do_facet ost1 $LCTL get_param -n $CLIENT_PARAM |
4428 echo "$CLIENT_CACHE_COUNT"
4429 [ $CLIENT_CACHE_COUNT = $NEW_CLIENT_CACHE_COUNT ] ||
4430 error "error while apply client_cache_count"
4432 echo "Check the value is stored after remount"
4435 wait_update $(facet_host ost1) "lctl get_param -n $CLIENT_PARAM |
4436 head -1" $NEW_CLIENT_CACHE_COUNT
4437 CLIENT_CACHE_COUNT=$(do_facet ost1 $LCTL get_param -n $CLIENT_PARAM |
4439 echo "$CLIENT_CACHE_COUNT"
4440 [ $CLIENT_CACHE_COUNT = $NEW_CLIENT_CACHE_COUNT ] ||
4441 error "client_cache_count is not saved after remount"
4444 run_test 76a "set permanent params set_param -P"
4446 test_76b() { # LU-4783
4447 [[ $(lustre_version_code mgs) -ge $(version_code 2.5.57) ]] ||
4448 { skip "Need MGS version at least 2.5.57" && return 0; }
4451 do_facet mgs $LCTL get_param mgs.MGS.live.params ||
4452 error "start params log failed"
4455 run_test 76b "verify params log setup correctly"
4457 test_77() { # LU-3445
4458 local server_version=$(lustre_version_code $SINGLEMDS)
4460 [[ $server_version -ge $(version_code 2.2.60) ]] &&
4461 [[ $server_version -le $(version_code 2.4.0) ]] &&
4462 skip "Need MDS version < 2.2.60 or > 2.4.0" && return
4464 if [[ -z "$fs2ost_DEV" || -z "$fs2mds_DEV" ]]; then
4465 is_blkdev $SINGLEMDS $(mdsdevname ${SINGLEMDS//mds/}) &&
4466 skip_env "mixed loopback and real device not working" && return
4469 local fs2mdsdev=$(mdsdevname 1_2)
4470 local fs2ostdev=$(ostdevname 1_2)
4471 local fs2mdsvdev=$(mdsvdevname 1_2)
4472 local fs2ostvdev=$(ostvdevname 1_2)
4473 local fsname=test1234
4475 local failnid="$(h2$NETTYPE 1.2.3.4),$(h2$NETTYPE 4.3.2.1)"
4477 add fs2mds $(mkfs_opts mds1 $fs2mdsdev) --mgs --fsname=$fsname \
4478 --reformat $fs2mdsdev $fs2mdsvdev || error "add fs2mds failed"
4479 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS && trap cleanup_24a EXIT INT ||
4480 error "start fs2mds failed"
4482 mgsnid=$(do_facet fs2mds $LCTL list_nids | xargs | tr ' ' ,)
4483 [[ $mgsnid = *,* ]] || mgsnid+=",$mgsnid"
4485 add fs2ost $(mkfs_opts ost1 $fs2ostdev) --mgsnode=$mgsnid \
4486 --failnode=$failnid --fsname=$fsname \
4487 --reformat $fs2ostdev $fs2ostvdev ||
4488 error "add fs2ost failed"
4489 start fs2ost $fs2ostdev $OST_MOUNT_OPTS || error "start fs2ost failed"
4492 $MOUNT_CMD $mgsnid:/$fsname $MOUNT2 || error "mount $MOUNT2 failed"
4493 DIR=$MOUNT2 MOUNT=$MOUNT2 check_mount || error "check $MOUNT2 failed"
4496 run_test 77 "comma-separated MGS NIDs and failover node NIDs"
4498 if ! combined_mgs_mds ; then