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 $SINGLEMDS) = "zfs" ] &&
80 # bug number for skipped test: LU-2778 LU-2059
81 ALWAYS_EXCEPT="$ALWAYS_EXCEPT 57b 50h"
86 require_dsh_mds || exit 0
87 require_dsh_ost || exit 0
89 [ "$SLOW" = "no" ] && EXCEPT_SLOW="30a 31 45"
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 -t lustre $MGSNID:/${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" $ORIG || return 5
1065 run_test 28 "permanent parameter setting"
1068 [ "$OSTCOUNT" -lt "2" ] && skip_env "$OSTCOUNT < 2, skipping" && return
1069 setup > /dev/null 2>&1
1073 local PARAM="$FSNAME-OST0001.osc.active"
1074 local PROC_ACT="osc.$FSNAME-OST0001-osc-[^M]*.active"
1075 local PROC_UUID="osc.$FSNAME-OST0001-osc-[^M]*.ost_server_uuid"
1077 ACTV=$(lctl get_param -n $PROC_ACT)
1079 set_conf_param_and_check client \
1080 "lctl get_param -n $PROC_ACT" "$PARAM" $DEAC || return 2
1081 # also check ost_server_uuid status
1082 RESULT=$(lctl get_param -n $PROC_UUID | grep DEACTIV)
1083 if [ -z "$RESULT" ]; then
1084 echo "Live client not deactivated: $(lctl get_param -n $PROC_UUID)"
1087 echo "Live client success: got $RESULT"
1091 for num in $(seq $MDSCOUNT); do
1092 local mdtosc=$(get_mdtosc_proc_path mds${num} $FSNAME-OST0001)
1093 local MPROC="osc.$mdtosc.active"
1098 RESULT=$(do_facet mds${num} " lctl get_param -n $MPROC")
1099 [ ${PIPESTATUS[0]} = 0 ] || error "Can't read $MPROC"
1100 if [ $RESULT -eq $DEAC ]; then
1101 echo -n "MDT deactivated also after"
1102 echo "$WAIT sec (got $RESULT)"
1106 if [ $WAIT -eq $MAX ]; then
1107 echo -n "MDT not deactivated: wanted $DEAC"
1111 echo "Waiting $(($MAX - $WAIT))secs for MDT deactivated"
1114 # test new client starts deactivated
1115 umount_client $MOUNT || return 200
1117 RESULT=$(lctl get_param -n $PROC_UUID | grep DEACTIV | grep NEW)
1118 if [ -z "$RESULT" ]; then
1119 echo "New client not deactivated from start: $(lctl get_param -n $PROC_UUID)"
1122 echo "New client success: got $RESULT"
1125 # make sure it reactivates
1126 set_conf_param_and_check client \
1127 "lctl get_param -n $PROC_ACT" "$PARAM" $ACTV || return 6
1129 umount_client $MOUNT
1132 #writeconf to remove all ost2 traces for subsequent tests
1133 writeconf_or_reformat
1135 run_test 29 "permanently remove an OST"
1140 echo Big config llog
1141 TEST="lctl get_param -n llite.$FSNAME-*.max_read_ahead_whole_mb"
1143 LIST=(1 2 3 4 5 4 3 2 1 2 3 4 5 4 3 2 1 2 3 4 5)
1144 for i in ${LIST[@]}; do
1145 set_conf_param_and_check client "$TEST" \
1146 "$FSNAME.llite.max_read_ahead_whole_mb" $i || return 3
1148 # make sure client restart still works
1149 umount_client $MOUNT
1150 mount_client $MOUNT || return 4
1151 [ "$($TEST)" -ne "$i" ] && error "Param didn't stick across restart $($TEST) != $i"
1154 echo Erase parameter setting
1155 do_facet mgs "$LCTL conf_param -d $FSNAME.llite.max_read_ahead_whole_mb" || return 6
1156 umount_client $MOUNT
1157 mount_client $MOUNT || return 6
1159 echo "deleted (default) value=$FINAL, orig=$ORIG"
1160 # assumes this parameter started at the default value
1161 [ "$FINAL" -eq "$ORIG" ] || fail "Deleted value=$FINAL, orig=$ORIG"
1165 run_test 30a "Big config llog and conf_param deletion"
1170 # Make a fake nid. Use the OST nid, and add 20 to the least significant
1171 # numerical part of it. Hopefully that's not already a failover address for
1173 OSTNID=$(do_facet ost1 "$LCTL get_param nis" | tail -1 | awk '{print $1}')
1174 ORIGVAL=$(echo $OSTNID | egrep -oi "[0-9]*@")
1175 NEWVAL=$((($(echo $ORIGVAL | egrep -oi "[0-9]*") + 20) % 256))
1176 NEW=$(echo $OSTNID | sed "s/$ORIGVAL/$NEWVAL@/")
1177 echo "Using fake nid $NEW"
1179 TEST="$LCTL get_param -n osc.$FSNAME-OST0000-osc-[^M]*.import | grep failover_nids | sed -n 's/.*\($NEW\).*/\1/p'"
1180 set_conf_param_and_check client "$TEST" \
1181 "$FSNAME-OST0000.failover.node" $NEW ||
1182 error "didn't add failover nid $NEW"
1183 NIDS=$($LCTL get_param -n osc.$FSNAME-OST0000-osc-[^M]*.import | grep failover_nids)
1185 NIDCOUNT=$(($(echo "$NIDS" | wc -w) - 1))
1186 echo "should have 2 failover nids: $NIDCOUNT"
1187 [ $NIDCOUNT -eq 2 ] || error "Failover nid not added"
1188 do_facet mgs "$LCTL conf_param -d $FSNAME-OST0000.failover.node" || error "conf_param delete failed"
1189 umount_client $MOUNT
1190 mount_client $MOUNT || return 3
1192 NIDS=$($LCTL get_param -n osc.$FSNAME-OST0000-osc-[^M]*.import | grep failover_nids)
1194 NIDCOUNT=$(($(echo "$NIDS" | wc -w) - 1))
1195 echo "only 1 final nid should remain: $NIDCOUNT"
1196 [ $NIDCOUNT -eq 1 ] || error "Failover nids not removed"
1200 run_test 30b "Remove failover nids"
1202 test_31() { # bug 10734
1203 # ipaddr must not exist
1204 mount -t lustre 4.3.2.1@tcp:/lustre $MOUNT || true
1207 run_test 31 "Connect to non-existent node (shouldn't crash)"
1211 T32_BLIMIT=20480 # Kbytes
1215 # This is not really a test but a tool to create new disk
1216 # image tarballs for the upgrade tests.
1218 # Disk image tarballs should be created on single-node
1219 # clusters by running this test with default configurations
1220 # plus a few mandatory environment settings that are verified
1221 # at the beginning of the test.
1223 test_32newtarball() {
1227 local tmp=$TMP/t32_image_create
1229 if [ $FSNAME != t32fs -o $MDSCOUNT -ne 1 -o \
1230 \( -z "$MDSDEV" -a -z "$MDSDEV1" \) -o $OSTCOUNT -ne 1 -o \
1231 -z "$OSTDEV1" ]; then
1232 error "Needs FSNAME=t32fs MDSCOUNT=1 MDSDEV1=<nonexistent_file>" \
1233 "(or MDSDEV, in the case of b1_8) OSTCOUNT=1" \
1234 "OSTDEV1=<nonexistent_file>"
1238 echo "Found stale $tmp"
1243 tar cf - -C $src . | tar xf - -C $tmp/src
1244 dd if=/dev/zero of=$tmp/src/t32_qf_old bs=1M \
1245 count=$(($T32_BLIMIT / 1024 / 2))
1246 chown $T32_QID.$T32_QID $tmp/src/t32_qf_old
1252 [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.3.50) ] &&
1253 $LFS quotacheck -ug /mnt/$FSNAME
1254 $LFS setquota -u $T32_QID -b 0 -B $T32_BLIMIT -i 0 -I $T32_ILIMIT \
1257 tar cf - -C $tmp/src . | tar xf - -C /mnt/$FSNAME
1264 ls -Rni --time-style=+%s >$tmp/img/list
1265 find . ! -name .lustre -type f -exec sha1sum {} \; |
1266 sort -k 2 >$tmp/img/sha1sums
1268 $LCTL get_param -n version | head -n 1 |
1269 sed -e 's/^lustre: *//' >$tmp/img/commit
1271 [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.3.50) ] &&
1272 $LFS quotaon -ug /mnt/$FSNAME
1273 $LFS quota -u $T32_QID -v /mnt/$FSNAME
1274 $LFS quota -v -u $T32_QID /mnt/$FSNAME |
1275 awk 'BEGIN { num='1' } { if ($1 == "'/mnt/$FSNAME'") \
1276 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1277 | tr -d "*" > $tmp/img/bspace
1278 $LFS quota -v -u $T32_QID /mnt/$FSNAME |
1279 awk 'BEGIN { num='5' } { if ($1 == "'/mnt/$FSNAME'") \
1280 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1281 | tr -d "*" > $tmp/img/ispace
1286 find -type f -exec sha1sum {} \; | sort -k 2 >$tmp/sha1sums.src
1289 if ! diff -u $tmp/sha1sums.src $tmp/img/sha1sums; then
1290 echo "Data verification failed"
1293 uname -r >$tmp/img/kernel
1294 uname -m >$tmp/img/arch
1296 mv ${MDSDEV1:-$MDSDEV} $tmp/img
1297 mv $OSTDEV1 $tmp/img
1299 version=$(sed -e 's/\(^[0-9]\+\.[0-9]\+\)\(.*$\)/\1/' $tmp/img/commit |
1300 sed -e 's/\./_/g') # E.g., "1.8.7" -> "1_8"
1303 tar cjvf $dst/disk$version-$(facet_fstype $SINGLEMDS).tar.bz2 -S *
1308 #run_test 32newtarball "Create a new test_32 disk image tarball for this version"
1311 # The list of applicable tarballs is returned via the caller's
1312 # variable "tarballs".
1315 local node=$(facet_active_host $SINGLEMDS)
1316 local r="do_node $node"
1318 if [ "$CLIENTONLY" ]; then
1319 skip "Client-only testing"
1323 if ! $r which $TUNEFS; then
1324 skip_env "tunefs.lustre required on $node"
1328 local IMGTYPE=$(facet_fstype $SINGLEMDS)
1330 tarballs=$($r find $RLUSTRE/tests -maxdepth 1 -name \'disk*-$IMGTYPE.tar.bz2\')
1332 if [ -z "$tarballs" ]; then
1333 skip "No applicable tarballs found"
1338 t32_test_cleanup() {
1342 if $shall_cleanup_lustre; then
1343 umount $tmp/mnt/lustre || rc=$?
1345 if $shall_cleanup_mdt; then
1346 $r umount -d $tmp/mnt/mdt || rc=$?
1348 if $shall_cleanup_mdt1; then
1349 $r umount -d $tmp/mnt/mdt1 || rc=$?
1351 if $shall_cleanup_ost; then
1352 $r umount -d $tmp/mnt/ost || rc=$?
1360 t32_bits_per_long() {
1362 # Yes, this is not meant to be perfect.
1372 t32_reload_modules() {
1374 local all_removed=false
1377 while ((i < 20)); do
1378 echo "Unloading modules on $node: Attempt $i"
1379 do_rpc_nodes $node $LUSTRE_RMMOD $(facet_fstype $SINGLEMDS) &&
1381 do_rpc_nodes $node check_mem_leak || return 1
1382 if $all_removed; then
1383 do_rpc_nodes $node load_modules
1389 echo "Unloading modules on $node: Given up"
1393 t32_wait_til_devices_gone() {
1398 echo wait for devices to go
1399 while ((i < 20)); do
1400 devices=$(do_rpc_nodes $node $LCTL device_list | wc -l)
1401 ((devices == 0)) && return 0
1405 echo "waiting for devices on $node: Given up"
1409 t32_verify_quota() {
1413 local fstype=$(facet_fstype $SINGLEMDS)
1417 $LFS quota -u $T32_QID -v $mnt
1419 qval=$($LFS quota -v -u $T32_QID $mnt |
1420 awk 'BEGIN { num='1' } { if ($1 == "'$mnt'") \
1421 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1423 [ $qval -eq $img_bspace ] || {
1424 echo "bspace, act:$qval, exp:$img_bspace"
1428 qval=$($LFS quota -v -u $T32_QID $mnt |
1429 awk 'BEGIN { num='5' } { if ($1 == "'$mnt'") \
1430 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1432 [ $qval -eq $img_ispace ] || {
1433 echo "ispace, act:$qval, exp:$img_ispace"
1437 qval=$($LFS quota -v -u $T32_QID $mnt |
1438 awk 'BEGIN { num='3' } { if ($1 == "'$mnt'") \
1439 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1441 [ $qval -eq $T32_BLIMIT ] || {
1442 echo "blimit, act:$qval, exp:$T32_BLIMIT"
1446 qval=$($LFS quota -v -u $T32_QID $mnt |
1447 awk 'BEGIN { num='7' } { if ($1 == "'$mnt'") \
1448 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1450 [ $qval -eq $T32_ILIMIT ] || {
1451 echo "ilimit, act:$qval, exp:$T32_ILIMIT"
1455 do_node $node $LCTL conf_param $fsname.quota.mdt=ug
1456 cmd="$LCTL get_param -n osd-$fstype.$fsname-MDT0000"
1457 cmd=$cmd.quota_slave.enabled
1458 wait_update $node "$cmd" "ug" || {
1459 echo "Enable mdt quota failed"
1463 do_node $node $LCTL conf_param $fsname.quota.ost=ug
1464 cmd="$LCTL get_param -n osd-$fstype.$fsname-OST0000"
1465 cmd=$cmd.quota_slave.enabled
1466 wait_update $node "$cmd" "ug" || {
1467 echo "Enable ost quota failed"
1472 runas -u $T32_QID -g $T32_QID dd if=/dev/zero of=$mnt/t32_qf_new \
1473 bs=1M count=$(($T32_BLIMIT / 1024)) oflag=sync && {
1474 echo "Write succeed, but expect -EDQUOT"
1477 rm -f $mnt/t32_qf_new
1479 runas -u $T32_QID -g $T32_QID createmany -m $mnt/t32_qf_ \
1481 echo "Create succeed, but expect -EDQUOT"
1484 unlinkmany $mnt/t32_qf_ $T32_ILIMIT
1492 local dne_upgrade=${dne_upgrade:-"no"}
1493 local ff_convert=${ff_convert:-"no"}
1494 local shall_cleanup_mdt=false
1495 local shall_cleanup_mdt1=false
1496 local shall_cleanup_ost=false
1497 local shall_cleanup_lustre=false
1498 local node=$(facet_active_host $SINGLEMDS)
1499 local r="do_node $node"
1500 local node2=$(facet_active_host mds2)
1508 local nid=$($r $LCTL list_nids | head -1)
1514 local fstype=$(facet_fstype $SINGLEMDS)
1516 trap 'trap - RETURN; t32_test_cleanup' RETURN
1518 mkdir -p $tmp/mnt/lustre
1519 $r mkdir -p $tmp/mnt/{mdt,ost}
1520 $r tar xjvf $tarball -S -C $tmp || {
1521 error_noexit "Unpacking the disk image tarball"
1524 img_commit=$($r cat $tmp/commit)
1525 img_kernel=$($r cat $tmp/kernel)
1526 img_arch=$($r cat $tmp/arch)
1527 img_bspace=$($r cat $tmp/bspace)
1528 img_ispace=$($r cat $tmp/ispace)
1529 echo "Upgrading from $(basename $tarball), created with:"
1530 echo " Commit: $img_commit"
1531 echo " Kernel: $img_kernel"
1532 echo " Arch: $img_arch"
1534 local version=$(version_code $img_commit)
1535 [[ $version -gt $(version_code 2.4.0) ]] && ff_convert="no"
1537 $r $LCTL set_param debug="$PTLDEBUG"
1539 $r $TUNEFS --dryrun $tmp/mdt || {
1540 error_noexit "tunefs.lustre before mounting the MDT"
1543 if [ "$writeconf" ]; then
1544 mopts=loop,writeconf
1545 if [ $fstype == "ldiskfs" ]; then
1546 $r $TUNEFS --quota $tmp/mdt || {
1547 error_noexit "Enable mdt quota feature"
1552 if [ -n "$($LCTL list_nids | grep -v '\(tcp\|lo\)[[:digit:]]*$')" ]; then
1553 [[ $(lustre_version_code mgs) -ge $(version_code 2.3.59) ]] ||
1554 { skip "LU-2200: Cannot run over Inifiniband w/o lctl replace_nids "
1555 "(Need MGS version at least 2.3.59)"; return 0; }
1557 local osthost=$(facet_active_host ost1)
1558 local ostnid=$(do_node $osthost $LCTL list_nids | head -1)
1560 $r mount -t lustre -o loop,nosvc $tmp/mdt $tmp/mnt/mdt
1561 $r lctl replace_nids $fsname-OST0000 $ostnid
1562 $r lctl replace_nids $fsname-MDT0000 $nid
1563 $r umount $tmp/mnt/mdt
1566 mopts=loop,exclude=$fsname-OST0000
1569 t32_wait_til_devices_gone $node
1571 $r mount -t lustre -o $mopts $tmp/mdt $tmp/mnt/mdt || {
1572 error_noexit "Mounting the MDT"
1575 shall_cleanup_mdt=true
1577 if [ "$dne_upgrade" != "no" ]; then
1578 local fs2mdsdev=$(mdsdevname 1_2)
1579 local fs2mdsvdev=$(mdsvdevname 1_2)
1581 echo "mkfs new MDT on ${fs2mdsdev}...."
1582 if [ $(facet_fstype mds1) == ldiskfs ]; then
1583 mkfsoptions="--mkfsoptions=\\\"-J size=8\\\""
1586 add fs2mds $(mkfs_opts mds2 $fs2mdsdev $fsname) --reformat \
1587 $mkfsoptions $fs2mdsdev $fs2mdsvdev > /dev/null || {
1588 error_noexit "Mkfs new MDT failed"
1592 $r $TUNEFS --dryrun $fs2mdsdev || {
1593 error_noexit "tunefs.lustre before mounting the MDT"
1597 echo "mount new MDT....$fs2mdsdev"
1598 $r mkdir -p $tmp/mnt/mdt1
1599 $r mount -t lustre -o $mopts $fs2mdsdev $tmp/mnt/mdt1 || {
1600 error_noexit "mount mdt1 failed"
1603 shall_cleanup_mdt1=true
1606 uuid=$($r $LCTL get_param -n mdt.$fsname-MDT0000.uuid) || {
1607 error_noexit "Getting MDT UUID"
1610 if [ "$uuid" != $fsname-MDT0000_UUID ]; then
1611 error_noexit "Unexpected MDT UUID: \"$uuid\""
1615 $r $TUNEFS --dryrun $tmp/ost || {
1616 error_noexit "tunefs.lustre before mounting the OST"
1619 if [ "$writeconf" ]; then
1620 mopts=loop,mgsnode=$nid,$writeconf
1621 if [ $fstype == "ldiskfs" ]; then
1622 $r $TUNEFS --quota $tmp/ost || {
1623 error_noexit "Enable ost quota feature"
1628 mopts=loop,mgsnode=$nid
1630 $r mount -t lustre -o $mopts $tmp/ost $tmp/mnt/ost || {
1631 error_noexit "Mounting the OST"
1634 shall_cleanup_ost=true
1636 uuid=$($r $LCTL get_param -n obdfilter.$fsname-OST0000.uuid) || {
1637 error_noexit "Getting OST UUID"
1640 if [ "$uuid" != $fsname-OST0000_UUID ]; then
1641 error_noexit "Unexpected OST UUID: \"$uuid\""
1645 $r $LCTL conf_param $fsname-OST0000.osc.max_dirty_mb=15 || {
1646 error_noexit "Setting \"max_dirty_mb\""
1649 $r $LCTL conf_param $fsname-OST0000.failover.node=$nid || {
1650 error_noexit "Setting OST \"failover.node\""
1653 $r $LCTL conf_param $fsname-MDT0000.mdc.max_rpcs_in_flight=9 || {
1654 error_noexit "Setting \"max_rpcs_in_flight\""
1657 $r $LCTL conf_param $fsname-MDT0000.failover.node=$nid || {
1658 error_noexit "Setting MDT \"failover.node\""
1661 $r $LCTL pool_new $fsname.interop || {
1662 error_noexit "Setting \"interop\""
1665 $r $LCTL conf_param $fsname-MDT0000.lov.stripesize=4M || {
1666 error_noexit "Setting \"lov.stripesize\""
1670 if [ "$ff_convert" != "no" -a $(facet_fstype ost1) == "ldiskfs" ]; then
1671 $r $LCTL lfsck_start -M $fsname-OST0000 || {
1672 error_noexit "Start OI scrub on OST0"
1676 # The oi_scrub should be on ost1, but for test_32(),
1677 # all on the SINGLEMDS.
1678 wait_update_facet $SINGLEMDS "$LCTL get_param -n \
1679 osd-ldiskfs.$fsname-OST0000.oi_scrub |
1680 awk '/^status/ { print \\\$2 }'" "completed" 30 || {
1681 error_noexit "Failed to get the expected 'completed'"
1685 local UPDATED=$($r $LCTL get_param -n \
1686 osd-ldiskfs.$fsname-OST0000.oi_scrub |
1687 awk '/^updated/ { print $2 }')
1688 [ $UPDATED -ge 1 ] || {
1689 error_noexit "Only $UPDATED objects have been converted"
1694 if [ "$dne_upgrade" != "no" ]; then
1695 $r $LCTL conf_param \
1696 $fsname-MDT0001.mdc.max_rpcs_in_flight=9 || {
1697 error_noexit "Setting MDT1 \"max_rpcs_in_flight\""
1700 $r $LCTL conf_param $fsname-MDT0001.failover.node=$nid || {
1701 error_noexit "Setting MDT1 \"failover.node\""
1704 $r $LCTL conf_param $fsname-MDT0001.lov.stripesize=4M || {
1705 error_noexit "Setting MDT1 \"lov.stripesize\""
1711 if [ "$writeconf" ]; then
1712 mount -t lustre $nid:/$fsname $tmp/mnt/lustre || {
1713 error_noexit "Mounting the client"
1716 shall_cleanup_lustre=true
1717 $LCTL set_param debug="$PTLDEBUG"
1719 t32_verify_quota $node $fsname $tmp/mnt/lustre || {
1720 error_noexit "verify quota failed"
1724 if [ "$dne_upgrade" != "no" ]; then
1725 $LFS mkdir -i 1 $tmp/mnt/lustre/remote_dir || {
1726 error_noexit "set remote dir failed"
1730 pushd $tmp/mnt/lustre
1731 tar -cf - . --exclude=./remote_dir |
1732 tar -xvf - -C remote_dir 1>/dev/null || {
1733 error_noexit "cp to remote dir failed"
1739 dd if=/dev/zero of=$tmp/mnt/lustre/tmp_file bs=10k count=10 || {
1740 error_noexit "dd failed"
1743 rm -rf $tmp/mnt/lustre/tmp_file || {
1744 error_noexit "rm failed"
1748 if $r test -f $tmp/sha1sums; then
1749 # LU-2393 - do both sorts on same node to ensure locale
1751 $r cat $tmp/sha1sums | sort -k 2 >$tmp/sha1sums.orig
1752 if [ "$dne_upgrade" != "no" ]; then
1753 pushd $tmp/mnt/lustre/remote_dir
1755 pushd $tmp/mnt/lustre
1758 find ! -name .lustre -type f -exec sha1sum {} \; |
1759 sort -k 2 >$tmp/sha1sums || {
1760 error_noexit "sha1sum"
1764 if ! diff -ub $tmp/sha1sums.orig $tmp/sha1sums; then
1765 error_noexit "sha1sum verification failed"
1769 echo "sha1sum verification skipped"
1772 if [ "$dne_upgrade" != "no" ]; then
1773 rm -rf $tmp/mnt/lustre/remote_dir || {
1774 error_noexit "remove remote dir failed"
1779 if $r test -f $tmp/list; then
1781 # There is not a Test Framework API to copy files to or
1782 # from a remote node.
1784 # LU-2393 - do both sorts on same node to ensure locale
1786 $r cat $tmp/list | sort -k 6 >$tmp/list.orig
1787 pushd $tmp/mnt/lustre
1788 ls -Rni --time-style=+%s | sort -k 6 >$tmp/list || {
1794 # 32-bit and 64-bit clients use different algorithms to
1795 # convert FIDs into inode numbers. Hence, remove the inode
1796 # numbers from the lists, if the original list was created
1797 # on an architecture with different number of bits per
1800 if [ $(t32_bits_per_long $(uname -m)) != \
1801 $(t32_bits_per_long $img_arch) ]; then
1802 echo "Different number of bits per \"long\" from the disk image"
1803 for list in list.orig list; do
1804 sed -i -e 's/^[0-9]\+[ \t]\+//' $tmp/$list
1807 if ! diff -ub $tmp/list.orig $tmp/list; then
1808 error_noexit "list verification failed"
1812 echo "list verification skipped"
1816 # When adding new data verification tests, please check for
1817 # the presence of the required reference files first, like
1818 # the "sha1sums" and "list" tests above, to avoid the need to
1819 # regenerate every image for each test addition.
1822 nrpcs_orig=$($LCTL get_param \
1823 -n mdc.*MDT0000*.max_rpcs_in_flight) || {
1824 error_noexit "Getting \"max_rpcs_in_flight\""
1827 nrpcs=$((nrpcs_orig + 5))
1828 $r $LCTL conf_param $fsname-MDT0000.mdc.max_rpcs_in_flight=$nrpcs || {
1829 error_noexit "Changing \"max_rpcs_in_flight\""
1832 wait_update $HOSTNAME "$LCTL get_param \
1833 -n mdc.*MDT0000*.max_rpcs_in_flight" $nrpcs || {
1834 error_noexit "Verifying \"max_rpcs_in_flight\""
1838 umount $tmp/mnt/lustre || {
1839 error_noexit "Unmounting the client"
1842 shall_cleanup_lustre=false
1844 if [ "$dne_upgrade" != "no" ]; then
1845 $r umount -d $tmp/mnt/mdt1 || {
1846 error_noexit "Unmounting the MDT2"
1849 shall_cleanup_mdt1=false
1852 $r umount -d $tmp/mnt/mdt || {
1853 error_noexit "Unmounting the MDT"
1856 shall_cleanup_mdt=false
1858 $r umount -d $tmp/mnt/ost || {
1859 error_noexit "Unmounting the OST"
1862 shall_cleanup_ost=false
1864 t32_reload_modules $node || {
1865 error_noexit "Reloading modules"
1869 # mount a second time to make sure we didnt leave upgrade flag on
1870 $r $TUNEFS --dryrun $tmp/mdt || {
1871 error_noexit "tunefs.lustre before remounting the MDT"
1874 $r mount -t lustre -o loop,exclude=$fsname-OST0000 $tmp/mdt \
1876 error_noexit "Remounting the MDT"
1879 shall_cleanup_mdt=true
1889 for tarball in $tarballs; do
1890 t32_test $tarball || let "rc += $?"
1894 run_test 32a "Upgrade (not live)"
1902 for tarball in $tarballs; do
1903 t32_test $tarball writeconf || let "rc += $?"
1907 run_test 32b "Upgrade with writeconf"
1914 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
1916 for tarball in $tarballs; do
1917 dne_upgrade=yes t32_test $tarball writeconf || rc=$?
1921 run_test 32c "dne upgrade test"
1929 for tarball in $tarballs; do
1930 ff_convert=yes t32_test $tarball || rc=$?
1934 run_test 32d "convert ff test"
1936 test_33a() { # bug 12333, was test_33
1938 local FSNAME2=test-123
1939 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
1942 [ -n "$ost1_HOST" ] && fs2ost_HOST=$ost1_HOST
1944 if [ -z "$fs2ost_DEV" -o -z "$fs2mds_DEV" ]; then
1945 local dev=${SINGLEMDS}_dev
1946 local MDSDEV=${!dev}
1947 is_blkdev $SINGLEMDS $MDSDEV && \
1948 skip_env "mixed loopback and real device not working" && return
1951 local fs2mdsdev=$(mdsdevname 1_2)
1952 local fs2ostdev=$(ostdevname 1_2)
1953 local fs2mdsvdev=$(mdsvdevname 1_2)
1954 local fs2ostvdev=$(ostvdevname 1_2)
1956 if [ $(facet_fstype mds1) == ldiskfs ]; then
1957 mkfsoptions="--mkfsoptions=\\\"-J size=8\\\"" # See bug 17931.
1960 add fs2mds $(mkfs_opts mds1 ${fs2mdsdev}) --mgs --fsname=${FSNAME2} \
1961 --reformat $mkfsoptions $fs2mdsdev $fs2mdsvdev || exit 10
1962 add fs2ost $(mkfs_opts ost1 ${fs2ostdev}) --mgsnode=$MGSNID \
1963 --fsname=${FSNAME2} --index=8191 --reformat $fs2ostdev \
1964 $fs2ostvdev || exit 10
1966 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS && trap cleanup_24a EXIT INT
1967 start fs2ost $fs2ostdev $OST_MOUNT_OPTS
1968 do_facet $SINGLEMDS "$LCTL conf_param $FSNAME2.sys.timeout=200" || rc=1
1970 mount -t lustre $MGSNID:/${FSNAME2} $MOUNT2 || rc=2
1973 cp /etc/hosts $MOUNT2/ || rc=3
1974 $LFS getstripe $MOUNT2/hosts
1979 cleanup_nocli || rc=6
1982 run_test 33a "Mount ost with a large index number"
1984 test_33b() { # was test_34
1987 do_facet client dd if=/dev/zero of=$MOUNT/24 bs=1024k count=1
1988 # Drop lock cancelation reply during umount
1989 #define OBD_FAIL_LDLM_CANCEL_NET 0x304
1990 do_facet client lctl set_param fail_loc=0x80000304
1991 #lctl set_param debug=-1
1992 umount_client $MOUNT
1995 run_test 33b "Drop cancel during umount"
1999 do_facet client "sh runmultiop_bg_pause $DIR/file O_c"
2000 manual_umount_client
2002 do_facet client killall -USR1 multiop
2003 if [ $rc -eq 0 ]; then
2004 error "umount not fail!"
2009 run_test 34a "umount with opened file should be fail"
2014 touch $DIR/$tfile || return 1
2015 stop_mds --force || return 2
2017 manual_umount_client --force
2019 if [ $rc -ne 0 ]; then
2020 error "mtab after failed umount - rc $rc"
2026 run_test 34b "force umount with failed mds should be normal"
2030 touch $DIR/$tfile || return 1
2031 stop_ost --force || return 2
2033 manual_umount_client --force
2035 if [ $rc -ne 0 ]; then
2036 error "mtab after failed umount - rc $rc"
2042 run_test 34c "force umount with failed ost should be normal"
2044 test_35a() { # bug 12459
2047 DBG_SAVE="`lctl get_param -n debug`"
2048 lctl set_param debug="ha"
2050 log "Set up a fake failnode for the MDS"
2052 local device=$(do_facet $SINGLEMDS "lctl get_param -n devices" |
2053 awk '($3 ~ "mdt" && $4 ~ "MDT") { print $4 }' | head -1)
2054 do_facet mgs "$LCTL conf_param \
2055 ${device}.failover.node=$(h2$NETTYPE $FAKENID)" || return 4
2057 log "Wait for RECONNECT_INTERVAL seconds (10s)"
2060 MSG="conf-sanity.sh test_35a `date +%F%kh%Mm%Ss`"
2063 log "Stopping the MDT: $device"
2064 stop_mdt 1 || return 5
2066 df $MOUNT > /dev/null 2>&1 &
2068 log "Restarting the MDT: $device"
2069 start_mdt 1 || return 6
2070 log "Wait for df ($DFPID) ... "
2073 lctl set_param debug="$DBG_SAVE"
2075 # retrieve from the log the first server that the client tried to
2076 # contact after the connection loss
2077 $LCTL dk $TMP/lustre-log-$TESTNAME.log
2078 NEXTCONN=`awk "/${MSG}/ {start = 1;}
2079 /import_select_connection.*$device-mdc.* using connection/ {
2081 if (\\\$NF ~ /$FAKENID/)
2087 }" $TMP/lustre-log-$TESTNAME.log`
2088 [ "$NEXTCONN" != "0" ] && log "The client didn't try to reconnect to the last active server (tried ${NEXTCONN} instead)" && return 7
2090 # remove nid settings
2091 writeconf_or_reformat
2093 run_test 35a "Reconnect to the last active server first"
2095 test_35b() { # bug 18674
2096 remote_mds || { skip "local MDS" && return 0; }
2100 $LCTL set_param debug="ha"
2102 MSG="conf-sanity.sh test_35b `date +%F%kh%Mm%Ss`"
2105 log "Set up a fake failnode for the MDS"
2107 local device=$(do_facet $SINGLEMDS "$LCTL get_param -n devices" |
2108 awk '($3 ~ "mdt" && $4 ~ "MDT") { print $4 }' | head -1)
2109 do_facet mgs "$LCTL conf_param \
2110 ${device}.failover.node=$(h2$NETTYPE $FAKENID)" || return 1
2112 local at_max_saved=0
2113 # adaptive timeouts may prevent seeing the issue
2114 if at_is_enabled; then
2115 at_max_saved=$(at_max_get mds)
2116 at_max_set 0 mds client
2119 mkdir -p $MOUNT/$tdir
2121 log "Injecting EBUSY on MDS"
2122 # Setting OBD_FAIL_MDS_RESEND=0x136
2123 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0x80000136" || return 2
2125 $LCTL set_param mdc.${FSNAME}*.stats=clear
2127 log "Creating a test file and stat it"
2128 touch $MOUNT/$tdir/$tfile
2129 stat $MOUNT/$tdir/$tfile
2131 log "Stop injecting EBUSY on MDS"
2132 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0" || return 3
2133 rm -f $MOUNT/$tdir/$tfile
2136 # restore adaptive timeout
2137 [ $at_max_saved -ne 0 ] && at_max_set $at_max_saved mds client
2139 $LCTL dk $TMP/lustre-log-$TESTNAME.log
2141 CONNCNT=`$LCTL get_param mdc.${FSNAME}*.stats | awk '/mds_connect/{print $2}'`
2143 # retrieve from the log if the client has ever tried to
2144 # contact the fake server after the loss of connection
2145 FAILCONN=`awk "BEGIN {ret = 0;}
2146 /import_select_connection.*${FSNAME}-MDT0000-mdc.* using connection/ {
2148 if (\\\$NF ~ /$FAKENID/) {
2153 END {print ret}" $TMP/lustre-log-$TESTNAME.log`
2155 [ "$FAILCONN" == "0" ] && \
2156 log "ERROR: The client reconnection has not been triggered" && \
2158 [ "$FAILCONN" == "2" ] && \
2159 log "ERROR: The client tried to reconnect to the failover server while the primary was busy" && \
2163 # When OBD_FAIL_MDS_RESEND is hit, we sleep for 2 * obd_timeout
2164 # Reconnects are supposed to be rate limited to one every 5s
2165 [ $CONNCNT -gt $((2 * $TIMEOUT / 5 + 1)) ] && \
2166 log "ERROR: Too many reconnects $CONNCNT" && \
2170 # remove nid settings
2171 writeconf_or_reformat
2173 run_test 35b "Continue reconnection retries, if the active server is busy"
2176 [ $OSTCOUNT -lt 2 ] && skip_env "skipping test for single OST" && return
2178 [ "$ost_HOST" = "`hostname`" -o "$ost1_HOST" = "`hostname`" ] || \
2179 { skip "remote OST" && return 0; }
2182 local FSNAME2=test1234
2183 local fs3ost_HOST=$ost_HOST
2184 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
2186 [ -n "$ost1_HOST" ] && fs2ost_HOST=$ost1_HOST && fs3ost_HOST=$ost1_HOST
2188 if [ -z "$fs2ost_DEV" -o -z "$fs2mds_DEV" -o -z "$fs3ost_DEV" ]; then
2189 is_blkdev $SINGLEMDS $MDSDEV && \
2190 skip_env "mixed loopback and real device not working" && return
2193 local fs2mdsdev=$(mdsdevname 1_2)
2194 local fs2ostdev=$(ostdevname 1_2)
2195 local fs3ostdev=$(ostdevname 2_2)
2196 local fs2mdsvdev=$(mdsvdevname 1_2)
2197 local fs2ostvdev=$(ostvdevname 1_2)
2198 local fs3ostvdev=$(ostvdevname 2_2)
2200 add fs2mds $(mkfs_opts mds1 ${fs2mdsdev}) --mgs --fsname=${FSNAME2} \
2201 --reformat $fs2mdsdev $fs2mdsvdev || exit 10
2202 # XXX after we support non 4K disk blocksize in ldiskfs, specify a
2203 # different one than the default value here.
2204 add fs2ost $(mkfs_opts ost1 ${fs2ostdev}) --mgsnode=$MGSNID \
2205 --fsname=${FSNAME2} --reformat $fs2ostdev $fs2ostvdev || exit 10
2206 add fs3ost $(mkfs_opts ost1 ${fs3ostdev}) --mgsnode=$MGSNID \
2207 --fsname=${FSNAME2} --reformat $fs3ostdev $fs3ostvdev || exit 10
2209 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS
2210 start fs2ost $fs2ostdev $OST_MOUNT_OPTS
2211 start fs3ost $fs3ostdev $OST_MOUNT_OPTS
2213 mount -t lustre $MGSNID:/${FSNAME2} $MOUNT2 || return 1
2215 sleep 5 # until 11778 fixed
2217 dd if=/dev/zero of=$MOUNT2/$tfile bs=1M count=7 || return 2
2219 BKTOTAL=`lctl get_param -n obdfilter.*.kbytestotal | awk 'BEGIN{total=0}; {total+=$1}; END{print total}'`
2220 BKFREE=`lctl get_param -n obdfilter.*.kbytesfree | awk 'BEGIN{free=0}; {free+=$1}; END{print free}'`
2221 BKAVAIL=`lctl get_param -n obdfilter.*.kbytesavail | awk 'BEGIN{avail=0}; {avail+=$1}; END{print avail}'`
2222 STRING=`df -P $MOUNT2 | tail -n 1 | awk '{print $2","$3","$4}'`
2223 DFTOTAL=`echo $STRING | cut -d, -f1`
2224 DFUSED=`echo $STRING | cut -d, -f2`
2225 DFAVAIL=`echo $STRING | cut -d, -f3`
2226 DFFREE=$(($DFTOTAL - $DFUSED))
2228 ALLOWANCE=$((64 * $OSTCOUNT))
2230 if [ $DFTOTAL -lt $(($BKTOTAL - $ALLOWANCE)) ] ||
2231 [ $DFTOTAL -gt $(($BKTOTAL + $ALLOWANCE)) ] ; then
2232 echo "**** FAIL: df total($DFTOTAL) mismatch OST total($BKTOTAL)"
2235 if [ $DFFREE -lt $(($BKFREE - $ALLOWANCE)) ] ||
2236 [ $DFFREE -gt $(($BKFREE + $ALLOWANCE)) ] ; then
2237 echo "**** FAIL: df free($DFFREE) mismatch OST free($BKFREE)"
2240 if [ $DFAVAIL -lt $(($BKAVAIL - $ALLOWANCE)) ] ||
2241 [ $DFAVAIL -gt $(($BKAVAIL + $ALLOWANCE)) ] ; then
2242 echo "**** FAIL: df avail($DFAVAIL) mismatch OST avail($BKAVAIL)"
2247 stop fs3ost -f || return 200
2248 stop fs2ost -f || return 201
2249 stop fs2mds -f || return 202
2250 unload_modules_conf || return 203
2253 run_test 36 "df report consistency on OSTs with different block size"
2256 local mntpt=$(facet_mntpt $SINGLEMDS)
2257 local mdsdev=$(mdsdevname ${SINGLEMDS//mds/})
2258 local mdsdev_sym="$TMP/sym_mdt.img"
2259 local opts=$MDS_MOUNT_OPTS
2262 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
2263 skip "Currently only applicable to ldiskfs-based MDTs"
2267 echo "MDS : $mdsdev"
2268 echo "SYMLINK : $mdsdev_sym"
2269 do_facet $SINGLEMDS rm -f $mdsdev_sym
2271 do_facet $SINGLEMDS ln -s $mdsdev $mdsdev_sym
2273 echo "mount symlink device - $mdsdev_sym"
2275 if ! do_facet $SINGLEMDS test -b $mdsdev; then
2276 opts=$(csa_add "$opts" -o loop)
2278 mount_op=$(do_facet $SINGLEMDS mount -v -t lustre $opts \
2279 $mdsdev_sym $mntpt 2>&1)
2282 echo mount_op=$mount_op
2284 do_facet $SINGLEMDS "umount -d $mntpt && rm -f $mdsdev_sym"
2286 if $(echo $mount_op | grep -q "unable to set tunable"); then
2287 error "set tunables failed for symlink device"
2290 [ $rc -eq 0 ] || error "mount symlink $mdsdev_sym failed! rc=$rc"
2294 run_test 37 "verify set tunables works for symlink device"
2296 test_38() { # bug 14222
2297 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
2298 skip "Only applicable to ldiskfs-based MDTs"
2306 FILES=`find $SRC -type f -mtime +1 | head -n $COUNT`
2307 log "copying $(echo $FILES | wc -w) files to $DIR/$tdir"
2309 tar cf - $FILES | tar xf - -C $DIR/$tdir || \
2310 error "copying $SRC to $DIR/$tdir"
2312 umount_client $MOUNT
2314 log "rename lov_objid file on MDS"
2315 rm -f $TMP/lov_objid.orig
2317 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
2318 do_facet $SINGLEMDS "$DEBUGFS -c -R \\\"dump lov_objid $TMP/lov_objid.orig\\\" $MDSDEV"
2319 do_facet $SINGLEMDS "$DEBUGFS -w -R \\\"rm lov_objid\\\" $MDSDEV"
2321 do_facet $SINGLEMDS "od -Ax -td8 $TMP/lov_objid.orig"
2322 # check create in mds_lov_connect
2326 [ $V ] && log "verifying $DIR/$tdir/$f"
2327 diff -q $f $DIR/$tdir/$f || ERROR=y
2329 do_facet $SINGLEMDS "$DEBUGFS -c -R \\\"dump lov_objid $TMP/lov_objid.new\\\" $MDSDEV"
2330 do_facet $SINGLEMDS "od -Ax -td8 $TMP/lov_objid.new"
2331 [ "$ERROR" = "y" ] && error "old and new files are different after connect" || true
2333 # check it's updates in sync
2334 umount_client $MOUNT
2337 do_facet $SINGLEMDS dd if=/dev/zero of=$TMP/lov_objid.clear bs=4096 count=1
2338 do_facet $SINGLEMDS "$DEBUGFS -w -R \\\"rm lov_objid\\\" $MDSDEV"
2339 do_facet $SINGLEMDS "$DEBUGFS -w -R \\\"write $TMP/lov_objid.clear lov_objid\\\" $MDSDEV "
2344 [ $V ] && log "verifying $DIR/$tdir/$f"
2345 diff -q $f $DIR/$tdir/$f || ERROR=y
2347 do_facet $SINGLEMDS "$DEBUGFS -c -R \\\"dump lov_objid $TMP/lov_objid.new1\\\" $MDSDEV"
2348 do_facet $SINGLEMDS "od -Ax -td8 $TMP/lov_objid.new1"
2349 umount_client $MOUNT
2351 [ "$ERROR" = "y" ] && error "old and new files are different after sync" || true
2353 log "files compared the same"
2356 run_test 38 "MDS recreates missing lov_objid file from OST data"
2362 perl $SRCDIR/leak_finder.pl $TMP/debug 2>&1 | egrep '*** Leak:' &&
2363 error "memory leak detected" || true
2365 run_test 39 "leak_finder recognizes both LUSTRE and LNET malloc messages"
2367 test_40() { # bug 15759
2369 #define OBD_FAIL_TGT_TOOMANY_THREADS 0x706
2370 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0x80000706"
2374 run_test 40 "race during service thread startup"
2376 test_41a() { #bug 14134
2377 if [ $(facet_fstype $SINGLEMDS) == ldiskfs ] &&
2378 ! do_facet $SINGLEMDS test -b $(mdsdevname 1); then
2379 skip "Loop devices does not work with nosvc option"
2384 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
2386 start $SINGLEMDS $MDSDEV $MDS_MOUNT_OPTS -o nosvc -n
2387 start ost1 `ostdevname 1` $OST_MOUNT_OPTS
2388 start $SINGLEMDS $MDSDEV $MDS_MOUNT_OPTS -o nomgs,force
2390 mount_client $MOUNT || return 1
2393 echo "blah blah" > $MOUNT/$tfile
2396 umount_client $MOUNT
2397 stop ost1 -f || return 201
2398 stop_mds -f || return 202
2399 stop_mds -f || return 203
2400 unload_modules_conf || return 204
2403 run_test 41a "mount mds with --nosvc and --nomgs"
2406 if [ $(facet_fstype $SINGLEMDS) == ldiskfs ] &&
2407 ! do_facet $SINGLEMDS test -b $(mdsdevname 1); then
2408 skip "Loop devices does not work with nosvc option"
2412 ! combined_mgs_mds && skip "needs combined mgs device" && return 0
2416 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
2418 start $SINGLEMDS $MDSDEV $MDS_MOUNT_OPTS -o nosvc -n
2420 start $SINGLEMDS $MDSDEV $MDS_MOUNT_OPTS -o nomgs,force
2422 mount_client $MOUNT || return 1
2425 echo "blah blah" > $MOUNT/$tfile
2426 cat $MOUNT/$tfile || return 200
2428 umount_client $MOUNT
2429 stop_ost || return 201
2430 stop_mds -f || return 202
2431 stop_mds -f || return 203
2434 run_test 41b "mount mds with --nosvc and --nomgs on first mount"
2436 test_42() { #bug 14693
2438 check_mount || error "client was not mounted"
2440 do_facet mgs $LCTL conf_param $FSNAME.llite.some_wrong_param=10
2441 umount_client $MOUNT ||
2442 error "unmounting client failed with invalid llite param"
2443 mount_client $MOUNT ||
2444 error "mounting client failed with invalid llite param"
2446 do_facet mgs $LCTL conf_param $FSNAME.sys.some_wrong_param=20
2447 cleanup || error "stopping $FSNAME failed with invalid sys param"
2450 check_mount || "client was not mounted with invalid sys param"
2451 cleanup || error "stopping $FSNAME failed with invalid sys param"
2454 run_test 42 "allow client/server mount/unmount with invalid config param"
2457 [ $UID -ne 0 -o $RUNAS_ID -eq 0 ] && skip_env "run as root"
2459 chmod ugo+x $DIR || error "chmod 0 failed"
2460 set_conf_param_and_check mds \
2461 "lctl get_param -n mdt.$FSNAME-MDT0000.root_squash" \
2462 "$FSNAME.mdt.root_squash" \
2464 set_conf_param_and_check mds \
2465 "lctl get_param -n mdt.$FSNAME-MDT0000.nosquash_nids" \
2466 "$FSNAME.mdt.nosquash_nids" \
2470 # create set of test files
2472 echo "111" > $DIR/$tfile-userfile || error "write 1 failed"
2473 chmod go-rw $DIR/$tfile-userfile || error "chmod 1 failed"
2474 chown $RUNAS_ID.$RUNAS_ID $DIR/$tfile-userfile || error "chown failed"
2476 echo "222" > $DIR/$tfile-rootfile || error "write 2 failed"
2477 chmod go-rw $DIR/$tfile-rootfile || error "chmod 2 faield"
2479 mkdir $DIR/$tdir-rootdir -p || error "mkdir failed"
2480 chmod go-rwx $DIR/$tdir-rootdir || error "chmod 3 failed"
2481 touch $DIR/$tdir-rootdir/tfile-1 || error "touch failed"
2484 # check root_squash:
2485 # set root squash UID:GID to RUNAS_ID
2486 # root should be able to access only files owned by RUNAS_ID
2488 set_conf_param_and_check mds \
2489 "lctl get_param -n mdt.$FSNAME-MDT0000.root_squash" \
2490 "$FSNAME.mdt.root_squash" \
2491 "$RUNAS_ID:$RUNAS_ID"
2493 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-userfile)
2494 dd if=$DIR/$tfile-userfile 1>/dev/null 2>/dev/null || \
2495 error "$ST: root read permission is denied"
2496 echo "$ST: root read permission is granted - ok"
2499 dd conv=notrunc if=$DIR/$tfile-userfile 1>/dev/null 2>/dev/null || \
2500 error "$ST: root write permission is denied"
2501 echo "$ST: root write permission is granted - ok"
2503 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-rootfile)
2504 dd if=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null && \
2505 error "$ST: root read permission is granted"
2506 echo "$ST: root read permission is denied - ok"
2509 dd conv=notrunc of=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null && \
2510 error "$ST: root write permission is granted"
2511 echo "$ST: root write permission is denied - ok"
2513 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tdir-rootdir)
2514 rm $DIR/$tdir-rootdir/tfile-1 1>/dev/null 2>/dev/null && \
2515 error "$ST: root unlink permission is granted"
2516 echo "$ST: root unlink permission is denied - ok"
2518 touch $DIR/tdir-rootdir/tfile-2 1>/dev/null 2>/dev/null && \
2519 error "$ST: root create permission is granted"
2520 echo "$ST: root create permission is denied - ok"
2523 # check nosquash_nids:
2524 # put client's NID into nosquash_nids list,
2525 # root should be able to access root file after that
2527 local NIDLIST=$(lctl list_nids all | tr '\n' ' ')
2528 NIDLIST="2@elan $NIDLIST 192.168.0.[2,10]@tcp"
2529 NIDLIST=$(echo $NIDLIST | tr -s ' ' ' ')
2530 set_conf_param_and_check mds \
2531 "lctl get_param -n mdt.$FSNAME-MDT0000.nosquash_nids" \
2532 "$FSNAME-MDTall.mdt.nosquash_nids" \
2535 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-rootfile)
2536 dd if=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null || \
2537 error "$ST: root read permission is denied"
2538 echo "$ST: root read permission is granted - ok"
2541 dd conv=notrunc of=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null || \
2542 error "$ST: root write permission is denied"
2543 echo "$ST: root write permission is granted - ok"
2545 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tdir-rootdir)
2546 rm $DIR/$tdir-rootdir/tfile-1 || \
2547 error "$ST: root unlink permission is denied"
2548 echo "$ST: root unlink permission is granted - ok"
2549 touch $DIR/$tdir-rootdir/tfile-2 || \
2550 error "$ST: root create permission is denied"
2551 echo "$ST: root create permission is granted - ok"
2555 run_test 43 "check root_squash and nosquash_nids"
2557 umount_client $MOUNT
2562 check_mount || return 2
2563 UUID=$($LCTL get_param llite.${FSNAME}*.uuid | cut -d= -f2)
2565 UUIDS=$(do_facet $SINGLEMDS "$LCTL get_param mdt.${FSNAME}*.exports.*.uuid")
2566 for VAL in $UUIDS; do
2567 NID=$(echo $VAL | cut -d= -f1)
2568 CLUUID=$(echo $VAL | cut -d= -f2)
2569 [ "$UUID" = "$CLUUID" ] && STATS_FOUND=yes && break
2571 [ "$STATS_FOUND" = "no" ] && error "stats not found for client"
2575 run_test 44 "mounted client proc entry exists"
2579 check_mount || return 2
2584 #define OBD_FAIL_PTLRPC_LONG_UNLINK 0x50f
2585 do_facet client "lctl set_param fail_loc=0x50f"
2588 manual_umount_client --force || return 3
2589 do_facet client "lctl set_param fail_loc=0x0"
2591 mount_client $MOUNT || return 4
2595 run_test 45 "long unlink handling in ptlrpcd"
2602 umount_client $MOUNT2 || rc=$?
2603 umount_client $MOUNT || rc=$?
2604 while [ $count -gt 0 ]; do
2605 stop ost${count} -f || rc=$?
2609 cleanup_nocli || rc=$?
2610 #writeconf to remove all ost2 traces for subsequent tests
2611 writeconf_or_reformat
2616 echo "Testing with $OSTCOUNT OSTs"
2618 start_mds || return 1
2619 #first client should see only one ost
2620 start_ost || return 2
2621 wait_osc_import_state mds ost FULL
2623 mount_client $MOUNT || return 3
2624 trap "cleanup_46a $OSTCOUNT" EXIT ERR
2627 for (( i=2; i<=$OSTCOUNT; i++ )); do
2628 start ost$i `ostdevname $i` $OST_MOUNT_OPTS || return $((i+2))
2631 # wait until osts in sync
2632 for (( i=2; i<=$OSTCOUNT; i++ )); do
2633 wait_osc_import_state mds ost$i FULL
2634 wait_osc_import_state client ost$i FULL
2637 #second client see all ost's
2639 mount_client $MOUNT2 || return 8
2640 $LFS setstripe -c -1 $MOUNT2 || return 9
2641 $LFS getstripe $MOUNT2 || return 10
2643 echo "ok" > $MOUNT2/widestripe
2644 $LFS getstripe $MOUNT2/widestripe || return 11
2645 # fill acl buffer for avoid expand lsm to them
2646 awk -F : '{if (FNR < 25) { print "u:"$1":rwx" }}' /etc/passwd | while read acl; do
2647 setfacl -m $acl $MOUNT2/widestripe
2651 stat $MOUNT/widestripe || return 12
2653 cleanup_46a $OSTCOUNT || { echo "cleanup_46a failed!" && return 13; }
2656 run_test 46a "handle ost additional - wide striped file"
2661 check_mount || return 2
2662 $LCTL set_param ldlm.namespaces.$FSNAME-*-*-*.lru_size=100
2666 for ns in $($LCTL get_param ldlm.namespaces.$FSNAME-*-*-*.lru_size); do
2667 if echo $ns | grep "MDT[[:digit:]]*"; then
2670 lrs=$(echo $ns | sed 's/.*lru_size=//')
2671 lru_size[count]=$lrs
2676 facet_failover $SINGLEMDS
2677 client_up || return 3
2680 for ns in $($LCTL get_param ldlm.namespaces.$FSNAME-*-*-*.lru_size); do
2681 if echo $ns | grep "MDT[[:digit:]]*"; then
2684 lrs=$(echo $ns | sed 's/.*lru_size=//')
2685 if ! test "$lrs" -eq "${lru_size[count]}"; then
2686 n=$(echo $ns | sed -e 's/ldlm.namespaces.//' -e 's/.lru_size=.*//')
2687 error "$n has lost lru_size: $lrs vs. ${lru_size[count]}"
2695 run_test 47 "server restart does not make client loss lru_resize settings"
2700 # reformat after this test is needed - if test will failed
2701 # we will have unkillable file at FS
2705 test_48() { # bug 17636
2708 check_mount || return 2
2710 $LFS setstripe -c -1 $MOUNT || return 9
2711 $LFS getstripe $MOUNT || return 10
2713 echo "ok" > $MOUNT/widestripe
2714 $LFS getstripe $MOUNT/widestripe || return 11
2716 trap cleanup_48 EXIT ERR
2718 # fill acl buffer for avoid expand lsm to them
2719 getent passwd | awk -F : '{ print "u:"$1":rwx" }' | while read acl; do
2720 setfacl -m $acl $MOUNT/widestripe
2723 stat $MOUNT/widestripe || return 12
2728 run_test 48 "too many acls on file"
2730 # check PARAM_SYS_LDLM_TIMEOUT option of MKFS.LUSTRE
2731 test_49() { # bug 17710
2732 local timeout_orig=$TIMEOUT
2733 local ldlm_timeout_orig=$LDLM_TIMEOUT
2734 local LOCAL_TIMEOUT=20
2736 LDLM_TIMEOUT=$LOCAL_TIMEOUT
2737 TIMEOUT=$LOCAL_TIMEOUT
2741 check_mount || return 1
2743 echo "check ldlm_timout..."
2744 LDLM_MDS="`do_facet $SINGLEMDS lctl get_param -n ldlm_timeout`"
2745 LDLM_OST1="`do_facet ost1 lctl get_param -n ldlm_timeout`"
2746 LDLM_CLIENT="`do_facet client lctl get_param -n ldlm_timeout`"
2748 if [ $LDLM_MDS -ne $LDLM_OST1 ] || [ $LDLM_MDS -ne $LDLM_CLIENT ]; then
2749 error "Different LDLM_TIMEOUT:$LDLM_MDS $LDLM_OST1 $LDLM_CLIENT"
2752 if [ $LDLM_MDS -ne $((LOCAL_TIMEOUT / 3)) ]; then
2753 error "LDLM_TIMEOUT($LDLM_MDS) is not correct"
2756 umount_client $MOUNT
2757 stop_ost || return 2
2758 stop_mds || return 3
2760 LDLM_TIMEOUT=$((LOCAL_TIMEOUT - 1))
2764 check_mount || return 7
2766 LDLM_MDS="`do_facet $SINGLEMDS lctl get_param -n ldlm_timeout`"
2767 LDLM_OST1="`do_facet ost1 lctl get_param -n ldlm_timeout`"
2768 LDLM_CLIENT="`do_facet client lctl get_param -n ldlm_timeout`"
2770 if [ $LDLM_MDS -ne $LDLM_OST1 ] || [ $LDLM_MDS -ne $LDLM_CLIENT ]; then
2771 error "Different LDLM_TIMEOUT:$LDLM_MDS $LDLM_OST1 $LDLM_CLIENT"
2774 if [ $LDLM_MDS -ne $((LOCAL_TIMEOUT - 1)) ]; then
2775 error "LDLM_TIMEOUT($LDLM_MDS) is not correct"
2778 cleanup || return $?
2780 LDLM_TIMEOUT=$ldlm_timeout_orig
2781 TIMEOUT=$timeout_orig
2783 run_test 49 "check PARAM_SYS_LDLM_TIMEOUT option of MKFS.LUSTRE"
2786 # Test both statfs and lfs df and fail if either one fails
2787 multiop_bg_pause $1 f_
2790 killall -USR1 multiop
2791 [ $RC1 -ne 0 ] && log "lazystatfs multiop failed"
2792 wait $PID || { RC1=$?; log "multiop return error "; }
2799 if [ $RC2 -eq 0 ]; then
2801 log "lazystatfs df failed"
2805 [[ $RC1 -ne 0 || $RC2 -eq 0 ]] && RC=1
2811 lctl set_param llite.$FSNAME-*.lazystatfs=1
2814 lazystatfs $MOUNT || error "lazystatfs failed but no down servers"
2816 cleanup || return $?
2818 run_test 50a "lazystatfs all servers available =========================="
2822 lctl set_param llite.$FSNAME-*.lazystatfs=1
2825 # Wait for client to detect down OST
2826 stop_ost || error "Unable to stop OST1"
2827 wait_osc_import_state mds ost DISCONN
2829 lazystatfs $MOUNT || error "lazystatfs should don't have returned EIO"
2831 umount_client $MOUNT || error "Unable to unmount client"
2832 stop_mds || error "Unable to stop MDS"
2834 run_test 50b "lazystatfs all servers down =========================="
2837 start_mds || error "Unable to start MDS"
2838 start_ost || error "Unable to start OST1"
2839 start_ost2 || error "Unable to start OST2"
2840 mount_client $MOUNT || error "Unable to mount client"
2841 lctl set_param llite.$FSNAME-*.lazystatfs=1
2844 # Wait for client to detect down OST
2845 stop_ost || error "Unable to stop OST1"
2846 wait_osc_import_state mds ost DISCONN
2847 lazystatfs $MOUNT || error "lazystatfs failed with one down server"
2849 umount_client $MOUNT || error "Unable to unmount client"
2850 stop_ost2 || error "Unable to stop OST2"
2851 stop_mds || error "Unable to stop MDS"
2852 #writeconf to remove all ost2 traces for subsequent tests
2853 writeconf_or_reformat
2855 run_test 50c "lazystatfs one server down =========================="
2858 start_mds || error "Unable to start MDS"
2859 start_ost || error "Unable to start OST1"
2860 start_ost2 || error "Unable to start OST2"
2861 mount_client $MOUNT || error "Unable to mount client"
2862 lctl set_param llite.$FSNAME-*.lazystatfs=1
2865 # Issue the statfs during the window where the client still
2866 # belives the OST to be available but it is in fact down.
2867 # No failure just a statfs which hangs for a timeout interval.
2868 stop_ost || error "Unable to stop OST1"
2869 lazystatfs $MOUNT || error "lazystatfs failed with one down server"
2871 umount_client $MOUNT || error "Unable to unmount client"
2872 stop_ost2 || error "Unable to stop OST2"
2873 stop_mds || error "Unable to stop MDS"
2874 #writeconf to remove all ost2 traces for subsequent tests
2875 writeconf_or_reformat
2877 run_test 50d "lazystatfs client/server conn race =========================="
2884 start_mds || return 1
2885 #first client should see only one ost
2886 start_ost || return 2
2887 wait_osc_import_state mds ost FULL
2889 # Wait for client to detect down OST
2890 stop_ost || error "Unable to stop OST1"
2891 wait_osc_import_state mds ost DISCONN
2893 mount_client $MOUNT || error "Unable to mount client"
2894 lctl set_param llite.$FSNAME-*.lazystatfs=0
2896 multiop_bg_pause $MOUNT _f
2900 if [ $RC1 -ne 0 ]; then
2901 log "multiop failed $RC1"
2904 sleep $(( $TIMEOUT+1 ))
2906 [ $? -ne 0 ] && error "process isn't sleep"
2907 start_ost || error "Unable to start OST1"
2908 wait $pid || error "statfs failed"
2911 umount_client $MOUNT || error "Unable to unmount client"
2912 stop_ost || error "Unable to stop OST1"
2913 stop_mds || error "Unable to stop MDS"
2915 run_test 50e "normal statfs all servers down =========================="
2920 CONN_PROC="osc.$FSNAME-OST0001-osc-[M]*.ost_server_uuid"
2922 start_mds || error "Unable to start mds"
2923 #first client should see only one ost
2924 start_ost || error "Unable to start OST1"
2925 wait_osc_import_state mds ost FULL
2927 start_ost2 || error "Unable to start OST2"
2928 wait_osc_import_state mds ost2 FULL
2930 # Wait for client to detect down OST
2931 stop_ost2 || error "Unable to stop OST2"
2933 wait_osc_import_state mds ost2 DISCONN
2934 mount_client $MOUNT || error "Unable to mount client"
2935 lctl set_param llite.$FSNAME-*.lazystatfs=0
2937 multiop_bg_pause $MOUNT _f
2941 if [ $RC1 -ne 0 ]; then
2942 log "lazystatfs multiop failed $RC1"
2945 sleep $(( $TIMEOUT+1 ))
2947 [ $? -ne 0 ] && error "process isn't sleep"
2948 start_ost2 || error "Unable to start OST2"
2949 wait $pid || error "statfs failed"
2950 stop_ost2 || error "Unable to stop OST2"
2953 umount_client $MOUNT || error "Unable to unmount client"
2954 stop_ost || error "Unable to stop OST1"
2955 stop_mds || error "Unable to stop MDS"
2956 #writeconf to remove all ost2 traces for subsequent tests
2957 writeconf_or_reformat
2959 run_test 50f "normal statfs one server in down =========================="
2962 [ "$OSTCOUNT" -lt "2" ] && skip_env "$OSTCOUNT < 2, skipping" && return
2964 start_ost2 || error "Unable to start OST2"
2965 wait_osc_import_state mds ost2 FULL
2966 wait_osc_import_state client ost2 FULL
2968 local PARAM="${FSNAME}-OST0001.osc.active"
2970 $LFS setstripe -c -1 $DIR/$tfile || error "Unable to lfs setstripe"
2971 do_facet mgs $LCTL conf_param $PARAM=0 || error "Unable to deactivate OST"
2973 umount_client $MOUNT || error "Unable to unmount client"
2974 mount_client $MOUNT || error "Unable to mount client"
2975 # This df should not cause a panic
2978 do_facet mgs $LCTL conf_param $PARAM=1 || error "Unable to activate OST"
2980 umount_client $MOUNT || error "Unable to unmount client"
2981 stop_ost2 || error "Unable to stop OST2"
2982 stop_ost || error "Unable to stop OST1"
2983 stop_mds || error "Unable to stop MDS"
2984 #writeconf to remove all ost2 traces for subsequent tests
2985 writeconf_or_reformat
2987 run_test 50g "deactivated OST should not cause panic====================="
2991 # prepare MDT/OST, make OSC inactive for OST1
2992 [ "$OSTCOUNT" -lt "2" ] && skip_env "$OSTCOUNT < 2, skipping" && return
2993 do_facet ost1 "$TUNEFS --param osc.active=0 `ostdevname 1`" ||
2994 error "tunefs OST1 failed"
2995 start_mds || error "Unable to start MDT"
2996 start_ost || error "Unable to start OST1"
2997 start_ost2 || error "Unable to start OST2"
2998 mount_client $MOUNT || error "client start failed"
3002 # activatate OSC for OST1
3003 local TEST="$LCTL get_param -n osc.${FSNAME}-OST0000-osc-[!M]*.active"
3004 set_conf_param_and_check client \
3005 "$TEST" "${FSNAME}-OST0000.osc.active" 1 ||
3006 error "Unable to activate OST1"
3008 mkdir -p $DIR/$tdir/2
3009 $LFS setstripe -c -1 -i 0 $DIR/$tdir/2
3010 sleep 1 && echo "create a file after OST1 is activated"
3012 createmany -o $DIR/$tdir/2/$tfile-%d 1
3014 # check OSC import is working
3015 stat $DIR/$tdir/2/* >/dev/null 2>&1 ||
3016 error "some OSC imports are still not connected"
3019 umount_client $MOUNT || error "Unable to umount client"
3020 stop_ost2 || error "Unable to stop OST2"
3023 run_test 50h "LU-642: activate deactivated OST ==="
3026 local LOCAL_TIMEOUT=20
3030 check_mount || return 1
3033 $LFS setstripe -c -1 $MOUNT/d1
3034 #define OBD_FAIL_MDS_REINT_DELAY 0x142
3035 do_facet $SINGLEMDS "lctl set_param fail_loc=0x142"
3036 touch $MOUNT/d1/f1 &
3039 start_ost2 || return 2
3041 stop_ost2 || return 3
3043 #writeconf to remove all ost2 traces for subsequent tests
3044 writeconf_or_reformat
3046 run_test 51 "Verify that mdt_reint handles RMF_MDT_MD correctly when an OST is added"
3055 do_node $node mkdir -p $dest
3056 [ $? -eq 0 ] || { error "Unable to create directory"; return 1; }
3058 do_node $node 'tar cf - '$@' | tar xf - -C '$dest';
3059 [ \"\${PIPESTATUS[*]}\" = \"0 0\" ] || exit 1'
3060 [ $? -eq 0 ] || { error "Unable to tar files"; return 2; }
3062 do_node $node 'getfattr -d -m "[a-z]*\\." '$@' > '$xattrs
3063 [ $? -eq 0 ] || { error "Unable to read xattrs"; return 3; }
3073 local backup2=${TMP}/backup2
3075 do_node $node mkdir -p $backup2
3076 [ $? -eq 0 ] || { error "Unable to create directory"; return 1; }
3078 do_node $node 'tar cf - '$@' | tar xf - -C '$backup2';
3079 [ \"\${PIPESTATUS[*]}\" = \"0 0\" ] || exit 1'
3080 [ $? -eq 0 ] || { error "Unable to tar files to diff"; return 2; }
3082 do_node $node "diff -rq $backup $backup2"
3083 [ $? -eq 0 ] || { error "contents differ"; return 3; }
3085 local xattrs2=${TMP}/xattrs2
3086 do_node $node 'getfattr -d -m "[a-z]*\\." '$@' > '$xattrs2
3087 [ $? -eq 0 ] || { error "Unable to read xattrs to diff"; return 4; }
3089 do_node $node "diff $xattrs $xattrs2"
3090 [ $? -eq 0 ] || { error "xattrs differ"; return 5; }
3092 do_node $node "rm -rf $backup2 $xattrs2"
3093 [ $? -eq 0 ] || { error "Unable to delete temporary files"; return 6; }
3097 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
3098 skip "Only applicable to ldiskfs-based MDTs"
3103 [ $? -eq 0 ] || { error "Unable to start MDS"; return 1; }
3105 [ $? -eq 0 ] || { error "Unable to start OST1"; return 2; }
3107 [ $? -eq 0 ] || { error "Unable to mount client"; return 3; }
3110 local ost1mnt=$(facet_mntpt ost1)
3111 local ost1node=$(facet_active_host ost1)
3112 local ost1tmp=$TMP/conf52
3116 [ $? -eq 0 ] || { error "Unable to create tdir"; return 4; }
3117 touch $TMP/modified_first
3118 [ $? -eq 0 ] || { error "Unable to create temporary file"; return 5; }
3119 local mtime=$(stat -c %Y $TMP/modified_first)
3120 do_node $ost1node "mkdir -p $ost1tmp && touch -m -d @$mtime $ost1tmp/modified_first"
3122 [ $? -eq 0 ] || { error "Unable to create temporary file"; return 6; }
3125 $LFS setstripe -c -1 -S 1M $DIR/$tdir
3126 [ $? -eq 0 ] || { error "lfs setstripe failed"; return 7; }
3128 for (( i=0; i < nrfiles; i++ )); do
3129 multiop $DIR/$tdir/$tfile-$i Ow1048576w1048576w524288c
3130 [ $? -eq 0 ] || { error "multiop failed"; return 8; }
3136 echo backup files to $TMP/files
3137 local files=$(find $DIR/$tdir -type f -newer $TMP/modified_first)
3138 copy_files_xattrs `hostname` $TMP/files $TMP/file_xattrs $files
3139 [ $? -eq 0 ] || { error "Unable to copy files"; return 9; }
3141 umount_client $MOUNT
3142 [ $? -eq 0 ] || { error "Unable to umount client"; return 10; }
3144 [ $? -eq 0 ] || { error "Unable to stop ost1"; return 11; }
3146 echo mount ost1 as ldiskfs
3147 do_node $ost1node mkdir -p $ost1mnt
3148 [ $? -eq 0 ] || { error "Unable to create $ost1mnt"; return 23; }
3149 if ! do_node $ost1node test -b $ost1_dev; then
3152 do_node $ost1node mount -t $(facet_fstype ost1) $loop $ost1_dev \
3154 [ $? -eq 0 ] || { error "Unable to mount ost1 as ldiskfs"; return 12; }
3157 echo backup objects to $ost1tmp/objects
3158 local objects=$(do_node $ost1node 'find '$ost1mnt'/O/[0-9]* -type f'\
3159 '-size +0 -newer '$ost1tmp'/modified_first -regex ".*\/[0-9]+"')
3160 copy_files_xattrs $ost1node $ost1tmp/objects $ost1tmp/object_xattrs \
3162 [ $? -eq 0 ] || { error "Unable to copy objects"; return 13; }
3164 # move objects to lost+found
3165 do_node $ost1node 'mv '$objects' '${ost1mnt}'/lost+found'
3166 [ $? -eq 0 ] || { error "Unable to move objects"; return 14; }
3169 do_node $ost1node "ll_recover_lost_found_objs -d $ost1mnt/lost+found"
3170 [ $? -eq 0 ] || { error "ll_recover_lost_found_objs failed"; return 15; }
3172 # compare restored objects against saved ones
3173 diff_files_xattrs $ost1node $ost1tmp/objects $ost1tmp/object_xattrs $objects
3174 [ $? -eq 0 ] || { error "Unable to diff objects"; return 16; }
3176 do_node $ost1node "umount $ost1mnt"
3177 [ $? -eq 0 ] || { error "Unable to umount ost1 as ldiskfs"; return 17; }
3180 [ $? -eq 0 ] || { error "Unable to start ost1"; return 18; }
3182 [ $? -eq 0 ] || { error "Unable to mount client"; return 19; }
3185 diff_files_xattrs `hostname` $TMP/files $TMP/file_xattrs $files
3186 [ $? -eq 0 ] || { error "Unable to diff files"; return 20; }
3188 rm -rf $TMP/files $TMP/file_xattrs
3189 [ $? -eq 0 ] || { error "Unable to delete temporary files"; return 21; }
3190 do_node $ost1node "rm -rf $ost1tmp"
3191 [ $? -eq 0 ] || { error "Unable to delete temporary files"; return 22; }
3194 run_test 52 "check recovering objects from lost+found"
3196 # Checks threads_min/max/started for some service
3198 # Arguments: service name (OST or MDT), facet (e.g., ost1, $SINGLEMDS), and a
3199 # parameter pattern prefix like 'ost.*.ost'.
3212 local msg="Insane $modname thread counts"
3213 local ncpts=$(check_cpt_number $facet)
3217 check_mount || return 41
3219 # We need to expand $parampat, but it may match multiple parameters, so
3220 # we'll pick the first one
3221 if ! paramp=$(do_facet $facet "lctl get_param -N ${parampat}.threads_min"|head -1); then
3222 error "Couldn't expand ${parampat}.threads_min parameter name"
3226 # Remove the .threads_min part
3227 paramp=${paramp%.threads_min}
3229 # Check for sanity in defaults
3230 tmin=$(do_facet $facet "lctl get_param -n ${paramp}.threads_min" || echo 0)
3231 tmax=$(do_facet $facet "lctl get_param -n ${paramp}.threads_max" || echo 0)
3232 tstarted=$(do_facet $facet "lctl get_param -n ${paramp}.threads_started" || echo 0)
3233 lassert 23 "$msg (PDSH problems?)" '(($tstarted && $tmin && $tmax))' || return $?
3234 lassert 24 "$msg" '(($tstarted >= $tmin && $tstarted <= $tmax ))' || return $?
3235 nthrs=$(expr $tmax - $tmin)
3236 if [ $nthrs -lt $ncpts ]; then
3242 [ $tmin -eq $tmax -a $tmin -eq $tstarted ] &&
3243 skip_env "module parameter forced $facet thread count" &&
3244 tmin=3 && tmax=$((3 * tmax))
3246 # Check that we can change min/max
3247 do_facet $facet "lctl set_param ${paramp}.threads_min=$((tmin + nthrs))"
3248 do_facet $facet "lctl set_param ${paramp}.threads_max=$((tmax - nthrs))"
3249 tmin2=$(do_facet $facet "lctl get_param -n ${paramp}.threads_min" || echo 0)
3250 tmax2=$(do_facet $facet "lctl get_param -n ${paramp}.threads_max" || echo 0)
3251 lassert 25 "$msg" '(($tmin2 == ($tmin + $nthrs) && $tmax2 == ($tmax - $nthrs)))' || return $?
3253 # Check that we can set min/max to the same value
3254 tmin=$(do_facet $facet "lctl get_param -n ${paramp}.threads_min" || echo 0)
3255 do_facet $facet "lctl set_param ${paramp}.threads_max=$tmin"
3256 tmin2=$(do_facet $facet "lctl get_param -n ${paramp}.threads_min" || echo 0)
3257 tmax2=$(do_facet $facet "lctl get_param -n ${paramp}.threads_max" || echo 0)
3258 lassert 26 "$msg" '(($tmin2 == $tmin && $tmax2 == $tmin))' || return $?
3260 # Check that we can't set max < min
3261 do_facet $facet "lctl set_param ${paramp}.threads_max=$((tmin - 1))"
3262 tmin2=$(do_facet $facet "lctl get_param -n ${paramp}.threads_min" || echo 0)
3263 tmax2=$(do_facet $facet "lctl get_param -n ${paramp}.threads_max" || echo 0)
3264 lassert 27 "$msg" '(($tmin2 <= $tmax2))' || return $?
3266 # We need to ensure that we get the module options desired; to do this
3267 # we set LOAD_MODULES_REMOTE=true and we call setmodopts below.
3268 LOAD_MODULES_REMOTE=true
3271 local newvalue="${opts}=$(expr $basethr \* $ncpts)"
3272 setmodopts -a $modname "$newvalue" oldvalue
3276 check_mount || return 41
3278 # Restore previous setting of MODOPTS_*
3279 setmodopts $modname "$oldvalue"
3281 # Check that $opts took
3282 tmin=$(do_facet $facet "lctl get_param -n ${paramp}.threads_min")
3283 tmax=$(do_facet $facet "lctl get_param -n ${paramp}.threads_max")
3284 tstarted=$(do_facet $facet "lctl get_param -n ${paramp}.threads_started")
3285 lassert 28 "$msg" '(($tstarted >= $tmin && $tstarted <= $tmax ))' || return $?
3294 thread_sanity OST ost1 'ost.*.ost' 'oss_num_threads' '16'
3297 run_test 53a "check OSS thread count params"
3301 local mds=$(do_facet $SINGLEMDS "lctl get_param -N mds.*.*.threads_max \
3303 if [ -z "$mds" ]; then
3304 #running this on an old MDT
3305 thread_sanity MDT $SINGLEMDS 'mdt.*.*.' 'mdt_num_threads' 16
3307 thread_sanity MDT $SINGLEMDS 'mds.*.*.' 'mds_num_threads' 16
3311 run_test 53b "check MDS thread count params"
3314 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
3315 skip "Only applicable to ldiskfs-based MDTs"
3319 do_rpc_nodes $(facet_host ost1) run_llverdev $(ostdevname 1) -p
3320 [ $? -eq 0 ] || error "llverdev failed!"
3323 run_test 54a "test llverdev and partial verify of device"
3326 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
3327 skip "Only applicable to ldiskfs-based MDTs"
3332 run_llverfs $MOUNT -p
3333 [ $? -eq 0 ] || error "llverfs failed!"
3336 run_test 54b "test llverfs and partial verify of filesystem"
3340 local max_ost_index=$1
3341 echo -n $(((max_ost_index + 1) * 8))
3345 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
3346 skip "Only applicable to ldiskfs-based MDTs"
3350 local mdsdev=$(mdsdevname 1)
3351 local mdsvdev=$(mdsvdevname 1)
3355 add mds1 $(mkfs_opts mds1 ${mdsdev}) --reformat $mdsdev \
3357 add ost1 $(mkfs_opts ost1 $(ostdevname 1)) --index=$i \
3358 --reformat $(ostdevname 1) $(ostvdevname 1)
3364 echo checking size of lov_objid for ost index $i
3365 LOV_OBJID_SIZE=$(do_facet mds1 "$DEBUGFS -R 'stat lov_objid' $mdsdev 2>/dev/null" | grep ^User | awk '{print $6}')
3366 if [ "$LOV_OBJID_SIZE" != $(lov_objid_size $i) ]; then
3367 error "lov_objid size has to be $(lov_objid_size $i), not $LOV_OBJID_SIZE"
3369 echo ok, lov_objid size is correct: $LOV_OBJID_SIZE
3376 run_test 55 "check lov_objid size"
3379 local mds_journal_size_orig=$MDSJOURNALSIZE
3383 for num in $(seq 1 $MDSCOUNT); do
3384 add mds${num} $(mkfs_opts mds${num} $(mdsdevname $num)) \
3385 --reformat $(mdsdevname $num) $(mdsvdevname $num)
3387 add ost1 $(mkfs_opts ost1 $(ostdevname 1)) --index=1000 --reformat \
3388 $(ostdevname 1) $(ostvdevname 1)
3389 add ost2 $(mkfs_opts ost2 $(ostdevname 2)) --index=10000 --reformat \
3390 $(ostdevname 2) $(ostvdevname 2)
3394 start_ost2 || error "Unable to start second ost"
3395 mount_client $MOUNT || error "Unable to mount client"
3399 MDSJOURNALSIZE=$mds_journal_size_orig
3402 run_test 56 "check big indexes"
3404 test_57a() { # bug 22656
3405 local NID=$(do_facet ost1 "$LCTL get_param nis" | tail -1 | awk '{print $1}')
3406 writeconf_or_reformat
3407 [ $(facet_fstype ost1) == zfs ] && import_zpool ost1
3408 do_facet ost1 "$TUNEFS --failnode=$NID `ostdevname 1`" || error "tunefs failed"
3410 start_ost && error "OST registration from failnode should fail"
3413 run_test 57a "initial registration from failnode should fail (should return errs)"
3416 local NID=$(do_facet ost1 "$LCTL get_param nis" | tail -1 | awk '{print $1}')
3417 writeconf_or_reformat
3418 [ $(facet_fstype ost1) == zfs ] && import_zpool ost1
3419 do_facet ost1 "$TUNEFS --servicenode=$NID `ostdevname 1`" || error "tunefs failed"
3421 start_ost || error "OST registration from servicenode should not fail"
3424 run_test 57b "initial registration from servicenode should not fail"
3427 do_facet mgs $LCTL get_param mgs.MGS.live.$FSNAME | grep OST | wc -l
3430 test_58() { # bug 22658
3431 if [ $(facet_fstype mds) != ldiskfs ]; then
3432 skip "Only applicable to ldiskfs-based MDTs"
3437 createmany -o $DIR/$tdir/$tfile-%d 100
3438 # make sure that OSTs do not cancel llog cookies before we unmount the MDS
3439 #define OBD_FAIL_OBD_LOG_CANCEL_NET 0x601
3440 do_facet $SINGLEMDS "lctl set_param fail_loc=0x601"
3441 unlinkmany $DIR/$tdir/$tfile-%d 100
3444 local MNTDIR=$(facet_mntpt $SINGLEMDS)
3445 local devname=$(mdsdevname ${SINGLEMDS//mds/})
3447 if ! do_facet $SINGLEMDS "test -b $devname"; then
3451 # remove all files from the OBJECTS dir
3452 do_facet $SINGLEMDS "mount -t ldiskfs $opts $devname $MNTDIR"
3453 do_facet $SINGLEMDS "find $MNTDIR/O/1/d* -type f -delete"
3454 do_facet $SINGLEMDS "umount $MNTDIR"
3455 # restart MDS with missing llog files
3457 do_facet mds "lctl set_param fail_loc=0"
3460 run_test 58 "missing llog files must not prevent MDT from mounting"
3463 start_mgsmds >> /dev/null
3464 local C1=$(count_osts)
3465 if [ $C1 -eq 0 ]; then
3466 start_ost >> /dev/null
3470 echo "original ost count: $C1 (expect > 0)"
3471 [ $C1 -gt 0 ] || error "No OSTs in $FSNAME log"
3472 start_mgsmds -o writeconf >> /dev/null || error "MDT start failed"
3473 local C2=$(count_osts)
3474 echo "after mdt writeconf count: $C2 (expect 0)"
3475 [ $C2 -gt 0 ] && error "MDT writeconf should erase OST logs"
3476 echo "OST start without writeconf should fail:"
3477 start_ost >> /dev/null && error "OST start without writeconf didn't fail"
3478 echo "OST start with writeconf should succeed:"
3479 start_ost -o writeconf >> /dev/null || error "OST1 start failed"
3480 local C3=$(count_osts)
3481 echo "after ost writeconf count: $C3 (expect 1)"
3482 [ $C3 -eq 1 ] || error "new OST writeconf should add:"
3483 start_ost2 -o writeconf >> /dev/null || error "OST2 start failed"
3484 local C4=$(count_osts)
3485 echo "after ost2 writeconf count: $C4 (expect 2)"
3486 [ $C4 -eq 2 ] || error "OST2 writeconf should add log"
3487 stop_ost2 >> /dev/null
3488 cleanup_nocli >> /dev/null
3489 #writeconf to remove all ost2 traces for subsequent tests
3490 writeconf_or_reformat
3492 run_test 59 "writeconf mount option"
3494 test_60() { # LU-471
3497 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
3498 skip "Only applicable to ldiskfs-based MDTs"
3502 for num in $(seq $MDSCOUNT); do
3503 add mds${num} $(mkfs_opts mds${num} $(mdsdevname $num)) \
3504 --mkfsoptions='\" -E stride=64 -O ^uninit_bg\"' \
3505 --reformat $(mdsdevname $num) $(mdsvdevname $num) ||
3509 dump=$(do_facet $SINGLEMDS dumpe2fs $(mdsdevname 1))
3511 [ $rc -eq 0 ] || error "dumpe2fs $(mdsdevname 1) failed"
3513 # MDT default has dirdata feature
3514 echo $dump | grep dirdata > /dev/null || error "dirdata is not set"
3515 # we disable uninit_bg feature
3516 echo $dump | grep uninit_bg > /dev/null && error "uninit_bg is set"
3517 # we set stride extended options
3518 echo $dump | grep stride > /dev/null || error "stride is not set"
3521 run_test 60 "check mkfs.lustre --mkfsoptions -E -O options setting"
3524 local reformat=false
3526 [ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.1.53) ] ||
3527 { skip "Need MDS version at least 2.1.53"; return 0; }
3529 if [ $(facet_fstype $SINGLEMDS) == ldiskfs ] &&
3530 ! large_xattr_enabled; then
3532 LDISKFS_MKFS_OPTS+=" -O large_xattr"
3534 for num in $(seq $MDSCOUNT); do
3535 add mds${num} $(mkfs_opts mds$num $(mdsdevname $num)) \
3536 --reformat $(mdsdevname $num) $(mdsvdevname $num) ||
3537 error "add mds $num failed"
3541 setup_noconfig || error "setting up the filesystem failed"
3542 client_up || error "starting client failed"
3544 local file=$DIR/$tfile
3547 local large_value="$(generate_string $(max_xattr_size))"
3548 local small_value="bar"
3550 local name="trusted.big"
3551 log "save large xattr $name on $file"
3552 setfattr -n $name -v $large_value $file ||
3553 error "saving $name on $file failed"
3555 local new_value=$(get_xattr_value $name $file)
3556 [[ "$new_value" != "$large_value" ]] &&
3557 error "$name different after saving"
3559 log "shrink value of $name on $file"
3560 setfattr -n $name -v $small_value $file ||
3561 error "shrinking value of $name on $file failed"
3563 new_value=$(get_xattr_value $name $file)
3564 [[ "$new_value" != "$small_value" ]] &&
3565 error "$name different after shrinking"
3567 log "grow value of $name on $file"
3568 setfattr -n $name -v $large_value $file ||
3569 error "growing value of $name on $file failed"
3571 new_value=$(get_xattr_value $name $file)
3572 [[ "$new_value" != "$large_value" ]] &&
3573 error "$name different after growing"
3575 log "check value of $name on $file after remounting MDS"
3577 new_value=$(get_xattr_value $name $file)
3578 [[ "$new_value" != "$large_value" ]] &&
3579 error "$name different after remounting MDS"
3581 log "remove large xattr $name from $file"
3582 setfattr -x $name $file || error "removing $name from $file failed"
3587 LDISKFS_MKFS_OPTS=${LDISKFS_MKFS_OPTS% -O large_xattr}
3591 run_test 61 "large xattr"
3594 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
3595 skip "Only applicable to ldiskfs-based MDTs"
3600 local mdsdev=$(mdsdevname 1)
3601 local ostdev=$(ostdevname 1)
3603 [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.2.51) ]] ||
3604 { skip "Need MDS version at least 2.2.51"; return 0; }
3606 echo "disable journal for mds"
3607 do_facet mds tune2fs -O ^has_journal $mdsdev || error "tune2fs failed"
3608 start_mds && error "MDT start should fail"
3609 echo "disable journal for ost"
3610 do_facet ost1 tune2fs -O ^has_journal $ostdev || error "tune2fs failed"
3611 start_ost && error "OST start should fail"
3612 cleanup || return $?
3615 run_test 62 "start with disabled journal"
3618 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
3619 skip "Only applicable to ldiskfs-based MDTs"
3623 local inode_slab=$(do_facet $SINGLEMDS \
3624 "awk '/ldiskfs_inode_cache/ { print \\\$5 }' /proc/slabinfo")
3625 if [ -z "$inode_slab" ]; then
3626 skip "ldiskfs module has not been loaded"
3630 echo "$inode_slab ldisk inodes per page"
3631 [ "$inode_slab" -ge "3" ] ||
3632 error "ldisk inode size is too big, $inode_slab objs per page"
3635 run_test 63 "Verify each page can at least hold 3 ldisk inodes"
3640 start_ost2 || error "Unable to start second ost"
3641 mount_client $MOUNT || error "Unable to mount client"
3642 stop_ost2 || error "Unable to stop second ost"
3644 $LFS df --lazy || error "lfs df failed"
3645 cleanup || return $?
3646 #writeconf to remove all ost2 traces for subsequent tests
3647 writeconf_or_reformat
3649 run_test 64 "check lfs df --lazy "
3651 test_65() { # LU-2237
3652 # Currently, the test is only valid for ldiskfs backend
3653 [ "$(facet_fstype $SINGLEMDS)" != "ldiskfs" ] &&
3654 skip "non-ldiskfs backend" && return
3656 local devname=$(mdsdevname ${SINGLEMDS//mds/})
3657 local brpt=$(facet_mntpt brpt)
3660 if ! do_facet $SINGLEMDS "test -b $devname"; then
3665 local obj=$(do_facet $SINGLEMDS \
3666 "$DEBUGFS -c -R \\\"stat last_rcvd\\\" $devname" |
3668 if [ -z "$obj" ]; then
3669 # The MDT may be just re-formatted, mount the MDT for the
3670 # first time to guarantee the "last_rcvd" file is there.
3671 start_mds || error "fail to mount the MDS for the first time"
3675 # remove the "last_rcvd" file
3676 do_facet $SINGLEMDS "mkdir -p $brpt"
3677 do_facet $SINGLEMDS \
3678 "mount -t $(facet_fstype $SINGLEMDS) $opts $devname $brpt"
3679 do_facet $SINGLEMDS "rm -f ${brpt}/last_rcvd"
3680 do_facet $SINGLEMDS "umount $brpt"
3682 # restart MDS, the "last_rcvd" file should be recreated.
3683 start_mds || error "fail to restart the MDS"
3685 obj=$(do_facet $SINGLEMDS \
3686 "$DEBUGFS -c -R \\\"stat last_rcvd\\\" $devname" | grep Inode)
3687 [ -n "$obj" ] || error "fail to re-create the last_rcvd"
3689 run_test 65 "re-create the lost last_rcvd file when server mount"
3692 [[ $(lustre_version_code mgs) -ge $(version_code 2.3.59) ]] ||
3693 { skip "Need MGS version at least 2.3.59"; return 0; }
3696 local OST1_NID=$(do_facet ost1 $LCTL list_nids | head -1)
3697 local MDS_NID=$(do_facet $SINGLEMDS $LCTL list_nids | head -1)
3699 echo "replace_nids should fail if MDS, OSTs and clients are UP"
3700 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 $OST1_NID &&
3701 error "replace_nids fail"
3703 umount_client $MOUNT || error "unmounting client failed"
3704 echo "replace_nids should fail if MDS and OSTs are UP"
3705 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 $OST1_NID &&
3706 error "replace_nids fail"
3709 echo "replace_nids should fail if MDS is UP"
3710 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 $OST1_NID &&
3711 error "replace_nids fail"
3713 stop_mds || error "stopping mds failed"
3715 if combined_mgs_mds; then
3716 start_mdt 1 "-o nosvc" ||
3717 error "starting mds with nosvc option failed"
3720 echo "command should accept two parameters"
3721 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 &&
3722 error "command should accept two params"
3724 echo "correct device name should be passed"
3725 do_facet mgs $LCTL replace_nids $FSNAME-WRONG0000 $OST1_NID &&
3726 error "wrong devname"
3728 echo "wrong nids list should not destroy the system"
3729 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 "wrong nids list" &&
3732 echo "replace OST nid"
3733 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 $OST1_NID ||
3734 error "replace nids failed"
3736 echo "command should accept two parameters"
3737 do_facet mgs $LCTL replace_nids $FSNAME-MDT0000 &&
3738 error "command should accept two params"
3740 echo "wrong nids list should not destroy the system"
3741 do_facet mgs $LCTL replace_nids $FSNAME-MDT0000 "wrong nids list" &&
3744 echo "replace MDS nid"
3745 do_facet mgs $LCTL replace_nids $FSNAME-MDT0000 $MDS_NID ||
3746 error "replace nids failed"
3748 if ! combined_mgs_mds ; then
3755 check_mount || error "error after nid replace"
3756 cleanup || error "cleanup failed"
3759 run_test 66 "replace nids"
3761 test_67() { #LU-2950
3762 local legacy="$TMP/legacy_lnet_config"
3763 local new="$TMP/new_routes_test"
3764 local out="$TMP/config_out_file"
3765 local verify="$TMP/conv_verify"
3766 local verify_conf="$TMP/conf_verify"
3768 # Create the legacy file that will be run through the
3769 # lustre_routes_conversion script
3770 cat <<- LEGACY_LNET_CONFIG > $legacy
3771 tcp1 23 192.168.213.1@tcp:1; tcp5 34 193.30.4.3@tcp:4;
3772 tcp2 54 10.1.3.2@tcp;
3773 tcp3 10.3.4.3@tcp:3;
3777 # Create the verification file to verify the output of
3778 # lustre_routes_conversion script against.
3779 cat <<- VERIFY_LNET_CONFIG > $verify
3780 tcp1: { gateway: 192.168.213.1@tcp, hop: 23, priority: 1 }
3781 tcp5: { gateway: 193.30.4.3@tcp, hop: 34, priority: 4 }
3782 tcp2: { gateway: 10.1.3.2@tcp, hop: 54 }
3783 tcp3: { gateway: 10.3.4.3@tcp, priority: 3 }
3784 tcp4: { gateway: 10.3.3.4@tcp }
3787 # Create the verification file to verify the output of
3788 # lustre_routes_config script against
3789 cat <<- VERIFY_LNET_CONFIG > $verify_conf
3790 lctl --net tcp1 add_route 192.168.213.1@tcp 23 1
3791 lctl --net tcp5 add_route 193.30.4.3@tcp 34 4
3792 lctl --net tcp2 add_route 10.1.3.2@tcp 54 4
3793 lctl --net tcp3 add_route 10.3.4.3@tcp 1 3
3794 lctl --net tcp4 add_route 10.3.3.4@tcp 1 3
3797 lustre_routes_conversion $legacy $new > /dev/null
3798 if [ -f $new ]; then
3799 # verify the conversion output
3800 cmp -s $new $verify > /dev/null
3801 if [ $? -eq 1 ]; then
3802 error "routes conversion failed"
3805 lustre_routes_config --dry-run --verbose $new > $out
3806 # check that the script succeeded
3807 cmp -s $out $verify_conf > /dev/null
3808 if [ $? -eq 1 ]; then
3809 error "routes config failed"
3812 error "routes conversion test failed"
3814 # remove generated files
3815 rm -f $new $legacy $verify $verify_conf $out
3817 run_test 67 "test routes conversion and configuration"
3825 [ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.4.53) ] ||
3826 { skip "Need MDS version at least 2.4.53"; return 0; }
3828 umount_client $MOUNT || error "umount client failed"
3830 start_mdt 1 || error "MDT start failed"
3833 # START-END - the sequences we'll be reserving
3834 START=$(do_facet $SINGLEMDS \
3835 lctl get_param -n seq.ctl*.space | awk -F'[[ ]' '{print $2}')
3836 END=$((START + (1 << 30)))
3837 do_facet $SINGLEMDS \
3838 lctl set_param seq.ctl*.fldb="[$START-$END\):0:mdt"
3840 # reset the sequences MDT0000 has already assigned
3841 do_facet $SINGLEMDS \
3842 lctl set_param seq.srv*MDT0000.space=clear
3844 # remount to let the client allocate new sequence
3845 mount_client $MOUNT || error "mount client failed"
3848 do_facet $SINGLEMDS \
3849 lctl get_param seq.srv*MDT0000.space
3850 $LFS path2fid $DIR/$tfile
3852 local old_ifs="$IFS"
3854 fid=($($LFS path2fid $DIR/$tfile))
3858 if [[ $seq < $END ]]; then
3859 error "used reserved sequence $seq?"
3861 cleanup || return $?
3863 run_test 68 "be able to reserve specific sequences in FLDB"
3868 # use OST0000 since it probably has the most creations
3869 local OSTNAME=$(ostname_from_index 0)
3870 local mdtosc_proc1=$(get_mdtosc_proc_path mds1 $OSTNAME)
3871 local last_id=$(do_facet mds1 lctl get_param -n \
3872 osc.$mdtosc_proc1.prealloc_last_id)
3874 # Want to have OST LAST_ID over 1.5 * OST_MAX_PRECREATE to
3875 # verify that the LAST_ID recovery is working properly. If
3876 # not, then the OST will refuse to allow the MDS connect
3877 # because the LAST_ID value is too different from the MDS
3878 #define OST_MAX_PRECREATE=20000
3879 local num_create=$((20000 * 3/2 - $last_id + 100))
3882 $LFS setstripe -i 0 $DIR/$tdir
3883 createmany $DIR/$tdir/$tfile- $num_create
3884 # delete all of the files with objects on OST0 so the
3885 # filesystem is not inconsistent later on
3886 $LFS find $MOUNT --index 0 -print0 | xargs -0 unlink
3888 stop_ost || error "OST0 stop failure"
3889 add ost1 $(mkfs_opts ost1 $ostdev) --reformat --replace $ostdev ||
3890 error "reformat and replace $ostdev failed"
3891 start_ost || error "OST0 restart failure"
3893 touch $DIR/$tdir/$tfile-last || error "create file after reformat"
3894 local idx=$($LFS getstripe -c $DIR/$tdir/$tfile-last)
3895 [ $idx -ne 0 ] && error "$DIR/$tdir/$tfile-last on $idx not 0" || true
3897 run_test 68 "replace an OST with the same index"
3900 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
3903 start_mdt 1 || error "MDT0 start fail"
3905 start_ost || error "OST0 start fail"
3907 start_mdt 2 || error "MDT1 start fail"
3909 mount_client $MOUNT || error "mount client fails"
3911 mkdir -p $DIR/$tdir || error "create dir fail"
3913 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
3914 error "create remote dir fail"
3916 rm -rf $DIR/$tdir || error "delete dir fail"
3917 cleanup || return $?
3919 run_test 70a "start MDT0, then OST, then MDT1"
3922 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
3925 start_ost || error "OST0 start fail"
3927 start_mdt 1 || error "MDT0 start fail"
3928 start_mdt 2 || error "MDT1 start fail"
3930 mount_client $MOUNT || error "mount client fails"
3932 mkdir -p $DIR/$tdir || error "create dir fail"
3934 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
3935 error "create remote dir fail"
3937 rm -rf $DIR/$tdir || error "delete dir fail"
3939 cleanup || return $?
3941 run_test 70b "start OST, MDT1, MDT0"
3944 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
3947 start_mdt 1 || error "MDT0 start fail"
3948 start_mdt 2 || error "MDT1 start fail"
3949 start_ost || error "OST0 start fail"
3951 mount_client $MOUNT || error "mount client fails"
3952 stop_mdt 1 || error "MDT1 start fail"
3954 local mdc_for_mdt1=$($LCTL dl | grep MDT0000-mdc | awk '{print $4}')
3955 echo "deactivate $mdc_for_mdt1"
3956 $LCTL --device $mdc_for_mdt1 deactivate || return 1
3958 mkdir -p $DIR/$tdir && error "mkdir succeed"
3960 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir &&
3961 error "create remote dir succeed"
3963 cleanup || return $?
3965 run_test 70c "stop MDT0, mkdir fail, create remote dir fail"
3968 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
3971 start_mdt 1 || error "MDT0 start fail"
3972 start_mdt 2 || error "MDT1 start fail"
3973 start_ost || error "OST0 start fail"
3975 mount_client $MOUNT || error "mount client fails"
3977 stop_mdt 2 || error "MDT1 start fail"
3979 local mdc_for_mdt2=$($LCTL dl | grep MDT0001-mdc |
3981 echo "deactivate $mdc_for_mdt2"
3982 $LCTL --device $mdc_for_mdt2 deactivate ||
3983 error "set $mdc_for_mdt2 deactivate failed"
3985 mkdir -p $DIR/$tdir || error "mkdir fail"
3986 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir &&
3987 error "create remote dir succeed"
3989 rm -rf $DIR/$tdir || error "delete dir fail"
3991 cleanup || return $?
3993 run_test 70d "stop MDT1, mkdir succeed, create remote dir fail"
3996 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
3997 if combined_mgs_mds; then
3998 skip "needs separate MGS/MDT" && return
4002 start_mdt 1 || error "MDT0 start fail"
4003 start_ost || error "OST0 start fail"
4004 start_mdt 2 || error "MDT1 start fail"
4005 start_ost2 || error "OST1 start fail"
4007 mount_client $MOUNT || error "mount client fails"
4009 mkdir -p $DIR/$tdir || error "mkdir fail"
4010 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
4011 error "create remote dir succeed"
4013 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
4014 rm -rf $DIR/$tdir || error "delete dir fail"
4016 umount_client $MOUNT
4017 stop_mdt 1 || error "MDT0 stop fail"
4018 stop_mdt 2 || error "MDT1 stop fail"
4019 stop_ost || error "OST0 stop fail"
4020 stop_ost2 || error "OST1 stop fail"
4022 run_test 71a "start MDT0 OST0, MDT1, OST1"
4025 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
4026 if combined_mgs_mds; then
4027 skip "needs separate MGS/MDT" && return
4031 start_mdt 2 || error "MDT1 start fail"
4032 start_ost || error "OST0 start fail"
4033 start_mdt 1 || error "MDT0 start fail"
4034 start_ost2 || error "OST1 start fail"
4036 mount_client $MOUNT || error "mount client fails"
4038 mkdir -p $DIR/$tdir || error "mkdir fail"
4039 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
4040 error "create remote dir succeed"
4042 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
4043 rm -rf $DIR/$tdir || error "delete dir fail"
4045 umount_client $MOUNT
4046 stop_mdt 1 || error "MDT0 stop fail"
4047 stop_mdt 2 || error "MDT1 stop fail"
4048 stop_ost || error "OST0 stop fail"
4049 stop_ost2 || error "OST1 stop fail"
4051 run_test 71b "start MDT1, OST0, MDT0, OST1"
4054 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
4055 if combined_mgs_mds; then
4056 skip "needs separate MGS/MDT" && return
4060 start_ost || error "OST0 start fail"
4061 start_ost2 || error "OST1 start fail"
4062 start_mdt 2 || error "MDT1 start fail"
4063 start_mdt 1 || error "MDT0 start fail"
4065 mount_client $MOUNT || error "mount client fails"
4067 mkdir -p $DIR/$tdir || error "mkdir fail"
4068 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
4069 error "create remote dir succeed"
4071 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
4072 rm -rf $DIR/$tdir || error "delete dir fail"
4074 umount_client $MOUNT
4075 stop_mdt 1 || error "MDT0 stop fail"
4076 stop_mdt 2 || error "MDT1 stop fail"
4077 stop_ost || error "OST0 stop fail"
4078 stop_ost2 || error "OST1 stop fail"
4081 run_test 71c "start OST0, OST1, MDT1, MDT0"
4084 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
4085 if combined_mgs_mds; then
4086 skip "needs separate MGS/MDT" && return
4090 start_ost || error "OST0 start fail"
4091 start_mdt 2 || error "MDT0 start fail"
4092 start_mdt 1 || error "MDT0 start fail"
4093 start_ost2 || error "OST1 start fail"
4095 mount_client $MOUNT || error "mount client fails"
4097 mkdir -p $DIR/$tdir || error "mkdir fail"
4098 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
4099 error "create remote dir succeed"
4101 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
4102 rm -rf $DIR/$tdir || error "delete dir fail"
4104 umount_client $MOUNT
4105 stop_mdt 1 || error "MDT0 stop fail"
4106 stop_mdt 2 || error "MDT1 stop fail"
4107 stop_ost || error "OST0 stop fail"
4108 stop_ost2 || error "OST1 stop fail"
4111 run_test 71d "start OST0, MDT1, MDT0, OST1"
4114 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
4115 if combined_mgs_mds; then
4116 skip "needs separate MGS/MDT" && return
4120 start_ost || error "OST0 start fail"
4121 start_mdt 2 || error "MDT1 start fail"
4122 start_ost2 || error "OST1 start fail"
4123 start_mdt 1 || error "MDT0 start fail"
4125 mount_client $MOUNT || error "mount client fails"
4127 mkdir -p $DIR/$tdir || error "mkdir fail"
4128 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
4129 error "create remote dir succeed"
4131 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
4132 rm -rf $DIR/$tdir || error "delete dir fail"
4134 umount_client $MOUNT
4135 stop_mdt 1 || error "MDT0 stop fail"
4136 stop_mdt 2 || error "MDT1 stop fail"
4137 stop_ost || error "OST0 stop fail"
4138 stop_ost2 || error "OST1 stop fail"
4141 run_test 71e "start OST0, MDT1, OST1, MDT0"
4143 test_72() { #LU-2634
4144 local mdsdev=$(mdsdevname 1)
4145 local ostdev=$(ostdevname 1)
4146 local cmd="$E2FSCK -fnvd $mdsdev"
4149 [ "$(facet_fstype $SINGLEMDS)" != "ldiskfs" ] &&
4150 skip "ldiskfs only test" && return
4152 #tune MDT with "-O extents"
4154 for num in $(seq $MDSCOUNT); do
4155 add mds${num} $(mkfs_opts mds$num $(mdsdevname $num)) \
4156 --reformat $(mdsdevname $num) $(mdsvdevname $num) ||
4157 error "add mds $num failed"
4158 $TUNE2FS -O extents $(mdsdevname $num)
4161 add ost1 $(mkfs_opts ost1 $ostdev) --reformat $ostdev ||
4162 error "add $ostdev failed"
4163 start_mgsmds || error "start mds failed"
4164 start_ost || error "start ost failed"
4165 mount_client $MOUNT || error "mount client failed"
4167 #create some short symlinks
4169 createmany -o $DIR/$tdir/$tfile-%d $fn
4170 echo "create $fn short symlinks"
4171 for i in $(seq -w 1 $fn); do
4172 ln -s $DIR/$tdir/$tfile-$i $MOUNT/$tfile-$i
4177 umount_client $MOUNT || error "umount client failed"
4178 stop_mds || error "stop mds failed"
4179 stop_ost || error "stop ost failed"
4182 run_e2fsck $(facet_active_host $SINGLEMDS) $mdsdev "-n"
4184 run_test 72 "test fast symlink with extents flag enabled"
4186 test_73() { #LU-3006
4188 [ $(facet_fstype ost1) == zfs ] && import_zpool ost1
4189 do_facet ost1 "$TUNEFS --failnode=1.2.3.4@$NETTYPE $(ostdevname 1)" ||
4190 error "1st tunefs failed"
4191 start_mgsmds || error "start mds failed"
4192 start_ost || error "start ost failed"
4193 mount_client $MOUNT || error "mount client failed"
4194 lctl get_param -n osc.*OST0000-osc-[^M]*.import | grep failover_nids |
4195 grep 1.2.3.4@$NETTYPE || error "failover nids haven't changed"
4196 umount_client $MOUNT || error "umount client failed"
4200 run_test 73 "failnode to update from mountdata properly"
4202 test_74() { # LU-1606
4203 for TESTPROG in $LUSTRE_TESTS_API_DIR/*.c; do
4204 gcc -Wall -Werror $LUSTRE_TESTS_API_DIR/simple_test.c \
4206 -L$LUSTRE/utils -llustreapi ||
4207 error "client api broken"
4209 cleanup || return $?
4211 run_test 74 "Lustre client api program can compile and link"
4213 test_75() { # LU-2374
4214 [[ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.4.1) ]] &&
4215 skip "Need MDS version at least 2.4.1" && return
4218 local opts_mds="$(mkfs_opts mds1 $(mdsdevname 1)) \
4219 --reformat $(mdsdevname 1) $(mdsvdevname 1)"
4220 local opts_ost="$(mkfs_opts ost1 $(ostdevname 1)) \
4221 --reformat $(ostdevname 1) $(ostvdevname 1)"
4223 #check with default parameters
4224 add mds1 $opts_mds || error "add mds1 failed for default params"
4225 add ost1 $opts_ost || error "add ost1 failed for default params"
4227 opts_mds=$(echo $opts_mds | sed -e "s/--mdt//")
4228 opts_mds=$(echo $opts_mds |
4229 sed -e "s/--index=$index/--index=$index --mdt/")
4230 opts_ost=$(echo $opts_ost | sed -e "s/--ost//")
4231 opts_ost=$(echo $opts_ost |
4232 sed -e "s/--index=$index/--index=$index --ost/")
4234 add mds1 $opts_mds || error "add mds1 failed for new params"
4235 add ost1 $opts_ost || error "add ost1 failed for new params"
4238 run_test 75 "The order of --index should be irrelevant"
4241 [[ $(lustre_version_code mgs) -ge $(version_code 2.4.52) ]] ||
4242 { skip "Need MDS version at least 2.4.52" && return 0; }
4244 local MDMB_PARAM="osc.*.max_dirty_mb"
4245 echo "Change MGS params"
4246 local MAX_DIRTY_MB=$($LCTL get_param -n $MDMB_PARAM |
4248 echo "max_dirty_mb: $MAX_DIRTY_MB"
4249 local NEW_MAX_DIRTY_MB=$((MAX_DIRTY_MB + MAX_DIRTY_MB))
4250 echo "new_max_dirty_mb: $NEW_MAX_DIRTY_MB"
4251 do_facet mgs $LCTL set_param -P $MDMB_PARAM=$NEW_MAX_DIRTY_MB
4252 wait_update $HOSTNAME "lctl get_param -n $MDMB_PARAM |
4253 head -1" $NEW_MAX_DIRTY_MB
4254 MAX_DIRTY_MB=$($LCTL get_param -n $MDMB_PARAM | head -1)
4255 echo "$MAX_DIRTY_MB"
4256 [ $MAX_DIRTY_MB = $NEW_MAX_DIRTY_MB ] ||
4257 error "error while apply max_dirty_mb"
4259 echo "Check the value is stored after remount"
4262 wait_update $HOSTNAME "lctl get_param -n $MDMB_PARAM |
4263 head -1" $NEW_MAX_DIRTY_MB
4264 MAX_DIRTY_MB=$($LCTL get_param -n $MDMB_PARAM | head -1)
4265 [ $MAX_DIRTY_MB = $NEW_MAX_DIRTY_MB ] ||
4266 error "max_dirty_mb is not saved after remount"
4268 echo "Change OST params"
4269 CLIENT_PARAM="obdfilter.*.client_cache_count"
4270 local CLIENT_CACHE_COUNT
4271 CLIENT_CACHE_COUNT=$(do_facet ost1 $LCTL get_param -n $CLIENT_PARAM |
4273 echo "client_cache_count: $CLIENT_CACHE_COUNT"
4274 NEW_CLIENT_CACHE_COUNT=$((CLIENT_CACHE_COUNT+CLIENT_CACHE_COUNT))
4275 echo "new_client_cache_count: $NEW_CLIENT_CACHE_COUNT"
4276 do_facet mgs $LCTL set_param -P $CLIENT_PARAM=$NEW_CLIENT_CACHE_COUNT
4277 wait_update $(facet_host ost1) "lctl get_param -n $CLIENT_PARAM |
4278 head -1" $NEW_CLIENT_CACHE_COUNT
4279 CLIENT_CACHE_COUNT=$(do_facet ost1 $LCTL get_param -n $CLIENT_PARAM |
4281 echo "$CLIENT_CACHE_COUNT"
4282 [ $CLIENT_CACHE_COUNT = $NEW_CLIENT_CACHE_COUNT ] ||
4283 error "error while apply client_cache_count"
4285 echo "Check the value is stored after remount"
4288 wait_update $(facet_host ost1) "lctl get_param -n $CLIENT_PARAM |
4289 head -1" $NEW_CLIENT_CACHE_COUNT
4290 CLIENT_CACHE_COUNT=$(do_facet ost1 $LCTL get_param -n $CLIENT_PARAM |
4292 echo "$CLIENT_CACHE_COUNT"
4293 [ $CLIENT_CACHE_COUNT = $NEW_CLIENT_CACHE_COUNT ] ||
4294 error "client_cache_count is not saved after remount"
4297 run_test 76 "set permanent params set_param -P"
4299 test_77() { # LU-3445
4300 local server_version=$(lustre_version_code $SINGLEMDS)
4302 [[ $server_version -ge $(version_code 2.2.60) ]] &&
4303 [[ $server_version -le $(version_code 2.4.0) ]] &&
4304 skip "Need MDS version < 2.2.60 or > 2.4.0" && return
4306 if [[ -z "$fs2ost_DEV" || -z "$fs2mds_DEV" ]]; then
4307 is_blkdev $SINGLEMDS $(mdsdevname ${SINGLEMDS//mds/}) &&
4308 skip_env "mixed loopback and real device not working" && return
4311 local fs2mdsdev=$(mdsdevname 1_2)
4312 local fs2ostdev=$(ostdevname 1_2)
4313 local fs2mdsvdev=$(mdsvdevname 1_2)
4314 local fs2ostvdev=$(ostvdevname 1_2)
4315 local fsname=test1234
4317 local failnid="$(h2$NETTYPE 1.2.3.4),$(h2$NETTYPE 4.3.2.1)"
4319 add fs2mds $(mkfs_opts mds1 $fs2mdsdev) --mgs --fsname=$fsname \
4320 --reformat $fs2mdsdev $fs2mdsvdev || error "add fs2mds failed"
4321 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS && trap cleanup_24a EXIT INT ||
4322 error "start fs2mds failed"
4324 mgsnid=$(do_facet fs2mds $LCTL list_nids | xargs | tr ' ' ,)
4325 [[ $mgsnid = *,* ]] || mgsnid+=",$mgsnid"
4327 add fs2ost $(mkfs_opts ost1 $fs2ostdev) --mgsnode=$mgsnid \
4328 --failnode=$failnid --fsname=$fsname \
4329 --reformat $fs2ostdev $fs2ostvdev ||
4330 error "add fs2ost failed"
4331 start fs2ost $fs2ostdev $OST_MOUNT_OPTS || error "start fs2ost failed"
4334 mount -t lustre $mgsnid:/$fsname $MOUNT2 || error "mount $MOUNT2 failed"
4335 DIR=$MOUNT2 MOUNT=$MOUNT2 check_mount || error "check $MOUNT2 failed"
4338 run_test 77 "comma-separated MGS NIDs and failover node NIDs"
4340 if ! combined_mgs_mds ; then