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 $node"
1328 if [ -n "$($LCTL list_nids | grep -v '\(tcp\|lo\)[[:digit:]]*$')" ]; then
1329 skip "LU-2200: Test cannot run over Infiniband"
1333 local IMGTYPE=$(facet_fstype $SINGLEMDS)
1335 tarballs=$($r find $RLUSTRE/tests -maxdepth 1 -name \'disk*-$IMGTYPE.tar.bz2\')
1337 if [ -z "$tarballs" ]; then
1338 skip "No applicable tarballs found"
1343 t32_test_cleanup() {
1347 if $shall_cleanup_lustre; then
1348 umount $tmp/mnt/lustre || rc=$?
1350 if $shall_cleanup_mdt; then
1351 $r umount -d $tmp/mnt/mdt || rc=$?
1353 if $shall_cleanup_mdt1; then
1354 $r umount -d $tmp/mnt/mdt1 || rc=$?
1356 if $shall_cleanup_ost; then
1357 $r umount -d $tmp/mnt/ost || rc=$?
1365 t32_bits_per_long() {
1367 # Yes, this is not meant to be perfect.
1377 t32_reload_modules() {
1379 local all_removed=false
1382 while ((i < 20)); do
1383 echo "Unloading modules on $node: Attempt $i"
1384 do_rpc_nodes $node $LUSTRE_RMMOD $(facet_fstype $SINGLEMDS) &&
1386 do_rpc_nodes $node check_mem_leak || return 1
1387 if $all_removed; then
1388 do_rpc_nodes $node load_modules
1394 echo "Unloading modules on $node: Given up"
1398 t32_wait_til_devices_gone() {
1403 echo wait for devices to go
1404 while ((i < 20)); do
1405 devices=$(do_rpc_nodes $node $LCTL device_list | wc -l)
1406 ((devices == 0)) && return 0
1410 echo "waiting for devices on $node: Given up"
1414 t32_verify_quota() {
1418 local fstype=$(facet_fstype $SINGLEMDS)
1422 $LFS quota -u $T32_QID -v $mnt
1424 qval=$($LFS quota -v -u $T32_QID $mnt |
1425 awk 'BEGIN { num='1' } { if ($1 == "'$mnt'") \
1426 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1428 [ $qval -eq $img_bspace ] || {
1429 echo "bspace, act:$qval, exp:$img_bspace"
1433 qval=$($LFS quota -v -u $T32_QID $mnt |
1434 awk 'BEGIN { num='5' } { if ($1 == "'$mnt'") \
1435 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1437 [ $qval -eq $img_ispace ] || {
1438 echo "ispace, act:$qval, exp:$img_ispace"
1442 qval=$($LFS quota -v -u $T32_QID $mnt |
1443 awk 'BEGIN { num='3' } { if ($1 == "'$mnt'") \
1444 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1446 [ $qval -eq $T32_BLIMIT ] || {
1447 echo "blimit, act:$qval, exp:$T32_BLIMIT"
1451 qval=$($LFS quota -v -u $T32_QID $mnt |
1452 awk 'BEGIN { num='7' } { if ($1 == "'$mnt'") \
1453 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1455 [ $qval -eq $T32_ILIMIT ] || {
1456 echo "ilimit, act:$qval, exp:$T32_ILIMIT"
1460 do_node $node $LCTL conf_param $fsname.quota.mdt=ug
1461 cmd="$LCTL get_param -n osd-$fstype.$fsname-MDT0000"
1462 cmd=$cmd.quota_slave.enabled
1463 wait_update $node "$cmd" "ug" || {
1464 echo "Enable mdt quota failed"
1468 do_node $node $LCTL conf_param $fsname.quota.ost=ug
1469 cmd="$LCTL get_param -n osd-$fstype.$fsname-OST0000"
1470 cmd=$cmd.quota_slave.enabled
1471 wait_update $node "$cmd" "ug" || {
1472 echo "Enable ost quota failed"
1477 runas -u $T32_QID -g $T32_QID dd if=/dev/zero of=$mnt/t32_qf_new \
1478 bs=1M count=$(($T32_BLIMIT / 1024)) oflag=sync && {
1479 echo "Write succeed, but expect -EDQUOT"
1482 rm -f $mnt/t32_qf_new
1484 runas -u $T32_QID -g $T32_QID createmany -m $mnt/t32_qf_ \
1486 echo "Create succeed, but expect -EDQUOT"
1489 unlinkmany $mnt/t32_qf_ $T32_ILIMIT
1497 local dne_upgrade=${dne_upgrade:-"no"}
1498 local shall_cleanup_mdt=false
1499 local shall_cleanup_mdt1=false
1500 local shall_cleanup_ost=false
1501 local shall_cleanup_lustre=false
1502 local node=$(facet_active_host $SINGLEMDS)
1503 local r="do_node $node"
1504 local node2=$(facet_active_host mds2)
1512 local nid=$($r $LCTL list_nids | head -1)
1518 local fstype=$(facet_fstype $SINGLEMDS)
1520 trap 'trap - RETURN; t32_test_cleanup' RETURN
1522 mkdir -p $tmp/mnt/lustre
1523 $r mkdir -p $tmp/mnt/{mdt,ost}
1524 $r tar xjvf $tarball -S -C $tmp || {
1525 error_noexit "Unpacking the disk image tarball"
1528 img_commit=$($r cat $tmp/commit)
1529 img_kernel=$($r cat $tmp/kernel)
1530 img_arch=$($r cat $tmp/arch)
1531 img_bspace=$($r cat $tmp/bspace)
1532 img_ispace=$($r cat $tmp/ispace)
1533 echo "Upgrading from $(basename $tarball), created with:"
1534 echo " Commit: $img_commit"
1535 echo " Kernel: $img_kernel"
1536 echo " Arch: $img_arch"
1538 $r $LCTL set_param debug="$PTLDEBUG"
1540 $r $TUNEFS --dryrun $tmp/mdt || {
1541 error_noexit "tunefs.lustre before mounting the MDT"
1544 if [ "$writeconf" ]; then
1545 mopts=loop,writeconf
1546 if [ $fstype == "ldiskfs" ]; then
1547 $r $TUNEFS --quota $tmp/mdt || {
1548 error_noexit "Enable mdt quota feature"
1553 mopts=loop,exclude=$fsname-OST0000
1556 t32_wait_til_devices_gone $node
1558 $r mount -t lustre -o $mopts $tmp/mdt $tmp/mnt/mdt || {
1559 error_noexit "Mounting the MDT"
1562 shall_cleanup_mdt=true
1564 if [ "$dne_upgrade" != "no" ]; then
1565 local fs2mdsdev=$(mdsdevname 1_2)
1566 local fs2mdsvdev=$(mdsvdevname 1_2)
1568 echo "mkfs new MDT on ${fs2mdsdev}...."
1569 if [ $(facet_fstype mds1) == ldiskfs ]; then
1570 mkfsoptions="--mkfsoptions=\\\"-J size=8\\\""
1573 add fs2mds $(mkfs_opts mds2 $fs2mdsdev $fsname) --reformat \
1574 $mkfsoptions $fs2mdsdev $fs2mdsvdev > /dev/null || {
1575 error_noexit "Mkfs new MDT failed"
1579 $r $TUNEFS --dryrun $fs2mdsdev || {
1580 error_noexit "tunefs.lustre before mounting the MDT"
1584 echo "mount new MDT....$fs2mdsdev"
1585 $r mkdir -p $tmp/mnt/mdt1
1586 $r mount -t lustre -o $mopts $fs2mdsdev $tmp/mnt/mdt1 || {
1587 error_noexit "mount mdt1 failed"
1590 shall_cleanup_mdt1=true
1593 uuid=$($r $LCTL get_param -n mdt.$fsname-MDT0000.uuid) || {
1594 error_noexit "Getting MDT UUID"
1597 if [ "$uuid" != $fsname-MDT0000_UUID ]; then
1598 error_noexit "Unexpected MDT UUID: \"$uuid\""
1602 $r $TUNEFS --dryrun $tmp/ost || {
1603 error_noexit "tunefs.lustre before mounting the OST"
1606 if [ "$writeconf" ]; then
1607 mopts=loop,mgsnode=$nid,$writeconf
1608 if [ $fstype == "ldiskfs" ]; then
1609 $r $TUNEFS --quota $tmp/ost || {
1610 error_noexit "Enable ost quota feature"
1615 mopts=loop,mgsnode=$nid
1617 $r mount -t lustre -o $mopts $tmp/ost $tmp/mnt/ost || {
1618 error_noexit "Mounting the OST"
1621 shall_cleanup_ost=true
1623 uuid=$($r $LCTL get_param -n obdfilter.$fsname-OST0000.uuid) || {
1624 error_noexit "Getting OST UUID"
1627 if [ "$uuid" != $fsname-OST0000_UUID ]; then
1628 error_noexit "Unexpected OST UUID: \"$uuid\""
1632 $r $LCTL conf_param $fsname-OST0000.osc.max_dirty_mb=15 || {
1633 error_noexit "Setting \"max_dirty_mb\""
1636 $r $LCTL conf_param $fsname-OST0000.failover.node=$nid || {
1637 error_noexit "Setting OST \"failover.node\""
1640 $r $LCTL conf_param $fsname-MDT0000.mdc.max_rpcs_in_flight=9 || {
1641 error_noexit "Setting \"max_rpcs_in_flight\""
1644 $r $LCTL conf_param $fsname-MDT0000.failover.node=$nid || {
1645 error_noexit "Setting MDT \"failover.node\""
1648 $r $LCTL pool_new $fsname.interop || {
1649 error_noexit "Setting \"interop\""
1652 $r $LCTL conf_param $fsname-MDT0000.lov.stripesize=4M || {
1653 error_noexit "Setting \"lov.stripesize\""
1657 if [ "$dne_upgrade" != "no" ]; then
1658 $r $LCTL conf_param \
1659 $fsname-MDT0001.mdc.max_rpcs_in_flight=9 || {
1660 error_noexit "Setting MDT1 \"max_rpcs_in_flight\""
1663 $r $LCTL conf_param $fsname-MDT0001.failover.node=$nid || {
1664 error_noexit "Setting MDT1 \"failover.node\""
1667 $r $LCTL conf_param $fsname-MDT0001.lov.stripesize=4M || {
1668 error_noexit "Setting MDT1 \"lov.stripesize\""
1674 if [ "$writeconf" ]; then
1675 mount -t lustre $nid:/$fsname $tmp/mnt/lustre || {
1676 error_noexit "Mounting the client"
1679 shall_cleanup_lustre=true
1680 $LCTL set_param debug="$PTLDEBUG"
1682 t32_verify_quota $node $fsname $tmp/mnt/lustre || {
1683 error_noexit "verify quota failed"
1687 if [ "$dne_upgrade" != "no" ]; then
1688 $LFS mkdir -i 1 $tmp/mnt/lustre/remote_dir || {
1689 error_noexit "set remote dir failed"
1693 pushd $tmp/mnt/lustre
1694 tar -cf - . --exclude=./remote_dir |
1695 tar -xvf - -C remote_dir 1>/dev/null || {
1696 error_noexit "cp to remote dir failed"
1702 dd if=/dev/zero of=$tmp/mnt/lustre/tmp_file bs=10k count=10 || {
1703 error_noexit "dd failed"
1706 rm -rf $tmp/mnt/lustre/tmp_file || {
1707 error_noexit "rm failed"
1711 if $r test -f $tmp/sha1sums; then
1712 # LU-2393 - do both sorts on same node to ensure locale
1714 $r cat $tmp/sha1sums | sort -k 2 >$tmp/sha1sums.orig
1715 if [ "$dne_upgrade" != "no" ]; then
1716 pushd $tmp/mnt/lustre/remote_dir
1718 pushd $tmp/mnt/lustre
1721 find ! -name .lustre -type f -exec sha1sum {} \; |
1722 sort -k 2 >$tmp/sha1sums || {
1723 error_noexit "sha1sum"
1727 if ! diff -ub $tmp/sha1sums.orig $tmp/sha1sums; then
1728 error_noexit "sha1sum verification failed"
1732 echo "sha1sum verification skipped"
1735 if [ "$dne_upgrade" != "no" ]; then
1736 rm -rf $tmp/mnt/lustre/remote_dir || {
1737 error_noexit "remove remote dir failed"
1742 if $r test -f $tmp/list; then
1744 # There is not a Test Framework API to copy files to or
1745 # from a remote node.
1747 # LU-2393 - do both sorts on same node to ensure locale
1749 $r cat $tmp/list | sort -k 6 >$tmp/list.orig
1750 pushd $tmp/mnt/lustre
1751 ls -Rni --time-style=+%s | sort -k 6 >$tmp/list || {
1757 # 32-bit and 64-bit clients use different algorithms to
1758 # convert FIDs into inode numbers. Hence, remove the inode
1759 # numbers from the lists, if the original list was created
1760 # on an architecture with different number of bits per
1763 if [ $(t32_bits_per_long $(uname -m)) != \
1764 $(t32_bits_per_long $img_arch) ]; then
1765 echo "Different number of bits per \"long\" from the disk image"
1766 for list in list.orig list; do
1767 sed -i -e 's/^[0-9]\+[ \t]\+//' $tmp/$list
1770 if ! diff -ub $tmp/list.orig $tmp/list; then
1771 error_noexit "list verification failed"
1775 echo "list verification skipped"
1779 # When adding new data verification tests, please check for
1780 # the presence of the required reference files first, like
1781 # the "sha1sums" and "list" tests above, to avoid the need to
1782 # regenerate every image for each test addition.
1785 nrpcs_orig=$($LCTL get_param \
1786 -n mdc.*MDT0000*.max_rpcs_in_flight) || {
1787 error_noexit "Getting \"max_rpcs_in_flight\""
1790 nrpcs=$((nrpcs_orig + 5))
1791 $r $LCTL conf_param $fsname-MDT0000.mdc.max_rpcs_in_flight=$nrpcs || {
1792 error_noexit "Changing \"max_rpcs_in_flight\""
1795 wait_update $HOSTNAME "$LCTL get_param \
1796 -n mdc.*MDT0000*.max_rpcs_in_flight" $nrpcs || {
1797 error_noexit "Verifying \"max_rpcs_in_flight\""
1801 umount $tmp/mnt/lustre || {
1802 error_noexit "Unmounting the client"
1805 shall_cleanup_lustre=false
1807 if [ "$dne_upgrade" != "no" ]; then
1808 $r umount -d $tmp/mnt/mdt1 || {
1809 error_noexit "Unmounting the MDT2"
1812 shall_cleanup_mdt1=false
1815 $r umount -d $tmp/mnt/mdt || {
1816 error_noexit "Unmounting the MDT"
1819 shall_cleanup_mdt=false
1821 $r umount -d $tmp/mnt/ost || {
1822 error_noexit "Unmounting the OST"
1825 shall_cleanup_ost=false
1827 t32_reload_modules $node || {
1828 error_noexit "Reloading modules"
1832 # mount a second time to make sure we didnt leave upgrade flag on
1833 $r $TUNEFS --dryrun $tmp/mdt || {
1834 error_noexit "tunefs.lustre before remounting the MDT"
1837 $r mount -t lustre -o loop,exclude=$fsname-OST0000 $tmp/mdt \
1839 error_noexit "Remounting the MDT"
1842 shall_cleanup_mdt=true
1852 for tarball in $tarballs; do
1853 t32_test $tarball || rc=$?
1857 run_test 32a "Upgrade (not live)"
1865 for tarball in $tarballs; do
1866 t32_test $tarball writeconf || rc=$?
1870 run_test 32b "Upgrade with writeconf"
1877 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
1879 for tarball in $tarballs; do
1880 dne_upgrade=yes t32_test $tarball writeconf || rc=$?
1884 run_test 32c "dne upgrade test"
1886 test_33a() { # bug 12333, was test_33
1888 local FSNAME2=test-123
1889 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
1892 [ -n "$ost1_HOST" ] && fs2ost_HOST=$ost1_HOST
1894 if [ -z "$fs2ost_DEV" -o -z "$fs2mds_DEV" ]; then
1895 local dev=${SINGLEMDS}_dev
1896 local MDSDEV=${!dev}
1897 is_blkdev $SINGLEMDS $MDSDEV && \
1898 skip_env "mixed loopback and real device not working" && return
1901 local fs2mdsdev=$(mdsdevname 1_2)
1902 local fs2ostdev=$(ostdevname 1_2)
1903 local fs2mdsvdev=$(mdsvdevname 1_2)
1904 local fs2ostvdev=$(ostvdevname 1_2)
1906 if [ $(facet_fstype mds1) == ldiskfs ]; then
1907 mkfsoptions="--mkfsoptions=\\\"-J size=8\\\"" # See bug 17931.
1910 add fs2mds $(mkfs_opts mds1 ${fs2mdsdev}) --mgs --fsname=${FSNAME2} \
1911 --reformat $mkfsoptions $fs2mdsdev $fs2mdsvdev || exit 10
1912 add fs2ost $(mkfs_opts ost1 ${fs2ostdev}) --mgsnode=$MGSNID \
1913 --fsname=${FSNAME2} --index=8191 --reformat $fs2ostdev \
1914 $fs2ostvdev || exit 10
1916 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS && trap cleanup_24a EXIT INT
1917 start fs2ost $fs2ostdev $OST_MOUNT_OPTS
1918 do_facet $SINGLEMDS "$LCTL conf_param $FSNAME2.sys.timeout=200" || rc=1
1920 mount -t lustre $MGSNID:/${FSNAME2} $MOUNT2 || rc=2
1923 cp /etc/hosts $MOUNT2/ || rc=3
1924 $LFS getstripe $MOUNT2/hosts
1929 cleanup_nocli || rc=6
1932 run_test 33a "Mount ost with a large index number"
1934 test_33b() { # was test_34
1937 do_facet client dd if=/dev/zero of=$MOUNT/24 bs=1024k count=1
1938 # Drop lock cancelation reply during umount
1939 #define OBD_FAIL_LDLM_CANCEL_NET 0x304
1940 do_facet client lctl set_param fail_loc=0x80000304
1941 #lctl set_param debug=-1
1942 umount_client $MOUNT
1945 run_test 33b "Drop cancel during umount"
1949 do_facet client "sh runmultiop_bg_pause $DIR/file O_c"
1950 manual_umount_client
1952 do_facet client killall -USR1 multiop
1953 if [ $rc -eq 0 ]; then
1954 error "umount not fail!"
1959 run_test 34a "umount with opened file should be fail"
1964 touch $DIR/$tfile || return 1
1965 stop_mds --force || return 2
1967 manual_umount_client --force
1969 if [ $rc -ne 0 ]; then
1970 error "mtab after failed umount - rc $rc"
1976 run_test 34b "force umount with failed mds should be normal"
1980 touch $DIR/$tfile || return 1
1981 stop_ost --force || return 2
1983 manual_umount_client --force
1985 if [ $rc -ne 0 ]; then
1986 error "mtab after failed umount - rc $rc"
1992 run_test 34c "force umount with failed ost should be normal"
1994 test_35a() { # bug 12459
1997 DBG_SAVE="`lctl get_param -n debug`"
1998 lctl set_param debug="ha"
2000 log "Set up a fake failnode for the MDS"
2002 local device=$(do_facet $SINGLEMDS "lctl get_param -n devices" | awk '($3 ~ "mdt" && $4 ~ "MDT") { print $4 }' | head -1)
2003 do_facet mgs "$LCTL conf_param ${device}.failover.node=" \
2004 "$(h2$NETTYPE $FAKENID)" || return 4
2006 log "Wait for RECONNECT_INTERVAL seconds (10s)"
2009 MSG="conf-sanity.sh test_35a `date +%F%kh%Mm%Ss`"
2012 log "Stopping the MDT: $device"
2013 stop_mdt 1 || return 5
2015 df $MOUNT > /dev/null 2>&1 &
2017 log "Restarting the MDT: $device"
2018 start_mdt 1 || return 6
2019 log "Wait for df ($DFPID) ... "
2022 lctl set_param debug="$DBG_SAVE"
2024 # retrieve from the log the first server that the client tried to
2025 # contact after the connection loss
2026 $LCTL dk $TMP/lustre-log-$TESTNAME.log
2027 NEXTCONN=`awk "/${MSG}/ {start = 1;}
2028 /import_select_connection.*$device-mdc.* using connection/ {
2030 if (\\\$NF ~ /$FAKENID/)
2036 }" $TMP/lustre-log-$TESTNAME.log`
2037 [ "$NEXTCONN" != "0" ] && log "The client didn't try to reconnect to the last active server (tried ${NEXTCONN} instead)" && return 7
2039 # remove nid settings
2040 writeconf_or_reformat
2042 run_test 35a "Reconnect to the last active server first"
2044 test_35b() { # bug 18674
2045 remote_mds || { skip "local MDS" && return 0; }
2049 $LCTL set_param debug="ha"
2051 MSG="conf-sanity.sh test_35b `date +%F%kh%Mm%Ss`"
2054 log "Set up a fake failnode for the MDS"
2056 local device=$(do_facet $SINGLEMDS "$LCTL get_param -n devices" | \
2057 awk '($3 ~ "mdt" && $4 ~ "MDT") { print $4 }' | head -1)
2058 do_facet mgs "$LCTL conf_param ${device}.failover.node=" \
2059 "$(h2$NETTYPE $FAKENID)" || return 1
2061 local at_max_saved=0
2062 # adaptive timeouts may prevent seeing the issue
2063 if at_is_enabled; then
2064 at_max_saved=$(at_max_get mds)
2065 at_max_set 0 mds client
2068 mkdir -p $MOUNT/$tdir
2070 log "Injecting EBUSY on MDS"
2071 # Setting OBD_FAIL_MDS_RESEND=0x136
2072 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0x80000136" || return 2
2074 $LCTL set_param mdc.${FSNAME}*.stats=clear
2076 log "Creating a test file and stat it"
2077 touch $MOUNT/$tdir/$tfile
2078 stat $MOUNT/$tdir/$tfile
2080 log "Stop injecting EBUSY on MDS"
2081 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0" || return 3
2082 rm -f $MOUNT/$tdir/$tfile
2085 # restore adaptive timeout
2086 [ $at_max_saved -ne 0 ] && at_max_set $at_max_saved mds client
2088 $LCTL dk $TMP/lustre-log-$TESTNAME.log
2090 CONNCNT=`$LCTL get_param mdc.${FSNAME}*.stats | awk '/mds_connect/{print $2}'`
2092 # retrieve from the log if the client has ever tried to
2093 # contact the fake server after the loss of connection
2094 FAILCONN=`awk "BEGIN {ret = 0;}
2095 /import_select_connection.*${FSNAME}-MDT0000-mdc.* using connection/ {
2097 if (\\\$NF ~ /$FAKENID/) {
2102 END {print ret}" $TMP/lustre-log-$TESTNAME.log`
2104 [ "$FAILCONN" == "0" ] && \
2105 log "ERROR: The client reconnection has not been triggered" && \
2107 [ "$FAILCONN" == "2" ] && \
2108 log "ERROR: The client tried to reconnect to the failover server while the primary was busy" && \
2112 # When OBD_FAIL_MDS_RESEND is hit, we sleep for 2 * obd_timeout
2113 # Reconnects are supposed to be rate limited to one every 5s
2114 [ $CONNCNT -gt $((2 * $TIMEOUT / 5 + 1)) ] && \
2115 log "ERROR: Too many reconnects $CONNCNT" && \
2119 # remove nid settings
2120 writeconf_or_reformat
2122 run_test 35b "Continue reconnection retries, if the active server is busy"
2125 [ $OSTCOUNT -lt 2 ] && skip_env "skipping test for single OST" && return
2127 [ "$ost_HOST" = "`hostname`" -o "$ost1_HOST" = "`hostname`" ] || \
2128 { skip "remote OST" && return 0; }
2131 local FSNAME2=test1234
2132 local fs3ost_HOST=$ost_HOST
2133 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
2135 [ -n "$ost1_HOST" ] && fs2ost_HOST=$ost1_HOST && fs3ost_HOST=$ost1_HOST
2137 if [ -z "$fs2ost_DEV" -o -z "$fs2mds_DEV" -o -z "$fs3ost_DEV" ]; then
2138 is_blkdev $SINGLEMDS $MDSDEV && \
2139 skip_env "mixed loopback and real device not working" && return
2142 local fs2mdsdev=$(mdsdevname 1_2)
2143 local fs2ostdev=$(ostdevname 1_2)
2144 local fs3ostdev=$(ostdevname 2_2)
2145 local fs2mdsvdev=$(mdsvdevname 1_2)
2146 local fs2ostvdev=$(ostvdevname 1_2)
2147 local fs3ostvdev=$(ostvdevname 2_2)
2149 add fs2mds $(mkfs_opts mds1 ${fs2mdsdev}) --mgs --fsname=${FSNAME2} \
2150 --reformat $fs2mdsdev $fs2mdsvdev || exit 10
2151 # XXX after we support non 4K disk blocksize in ldiskfs, specify a
2152 # different one than the default value here.
2153 add fs2ost $(mkfs_opts ost1 ${fs2ostdev}) --mgsnode=$MGSNID \
2154 --fsname=${FSNAME2} --reformat $fs2ostdev $fs2ostvdev || exit 10
2155 add fs3ost $(mkfs_opts ost1 ${fs3ostdev}) --mgsnode=$MGSNID \
2156 --fsname=${FSNAME2} --reformat $fs3ostdev $fs3ostvdev || exit 10
2158 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS
2159 start fs2ost $fs2ostdev $OST_MOUNT_OPTS
2160 start fs3ost $fs3ostdev $OST_MOUNT_OPTS
2162 mount -t lustre $MGSNID:/${FSNAME2} $MOUNT2 || return 1
2164 sleep 5 # until 11778 fixed
2166 dd if=/dev/zero of=$MOUNT2/$tfile bs=1M count=7 || return 2
2168 BKTOTAL=`lctl get_param -n obdfilter.*.kbytestotal | awk 'BEGIN{total=0}; {total+=$1}; END{print total}'`
2169 BKFREE=`lctl get_param -n obdfilter.*.kbytesfree | awk 'BEGIN{free=0}; {free+=$1}; END{print free}'`
2170 BKAVAIL=`lctl get_param -n obdfilter.*.kbytesavail | awk 'BEGIN{avail=0}; {avail+=$1}; END{print avail}'`
2171 STRING=`df -P $MOUNT2 | tail -n 1 | awk '{print $2","$3","$4}'`
2172 DFTOTAL=`echo $STRING | cut -d, -f1`
2173 DFUSED=`echo $STRING | cut -d, -f2`
2174 DFAVAIL=`echo $STRING | cut -d, -f3`
2175 DFFREE=$(($DFTOTAL - $DFUSED))
2177 ALLOWANCE=$((64 * $OSTCOUNT))
2179 if [ $DFTOTAL -lt $(($BKTOTAL - $ALLOWANCE)) ] ||
2180 [ $DFTOTAL -gt $(($BKTOTAL + $ALLOWANCE)) ] ; then
2181 echo "**** FAIL: df total($DFTOTAL) mismatch OST total($BKTOTAL)"
2184 if [ $DFFREE -lt $(($BKFREE - $ALLOWANCE)) ] ||
2185 [ $DFFREE -gt $(($BKFREE + $ALLOWANCE)) ] ; then
2186 echo "**** FAIL: df free($DFFREE) mismatch OST free($BKFREE)"
2189 if [ $DFAVAIL -lt $(($BKAVAIL - $ALLOWANCE)) ] ||
2190 [ $DFAVAIL -gt $(($BKAVAIL + $ALLOWANCE)) ] ; then
2191 echo "**** FAIL: df avail($DFAVAIL) mismatch OST avail($BKAVAIL)"
2196 stop fs3ost -f || return 200
2197 stop fs2ost -f || return 201
2198 stop fs2mds -f || return 202
2199 unload_modules_conf || return 203
2202 run_test 36 "df report consistency on OSTs with different block size"
2205 local mntpt=$(facet_mntpt $SINGLEMDS)
2206 local mdsdev=$(mdsdevname ${SINGLEMDS//mds/})
2207 local mdsdev_sym="$TMP/sym_mdt.img"
2208 local opts=$MDS_MOUNT_OPTS
2211 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
2212 skip "Currently only applicable to ldiskfs-based MDTs"
2216 echo "MDS : $mdsdev"
2217 echo "SYMLINK : $mdsdev_sym"
2218 do_facet $SINGLEMDS rm -f $mdsdev_sym
2220 do_facet $SINGLEMDS ln -s $mdsdev $mdsdev_sym
2222 echo "mount symlink device - $mdsdev_sym"
2224 if ! do_facet $SINGLEMDS test -b $mdsdev; then
2225 opts=$(csa_add "$opts" -o loop)
2227 mount_op=$(do_facet $SINGLEMDS mount -v -t lustre $opts \
2228 $mdsdev_sym $mntpt 2>&1)
2231 echo mount_op=$mount_op
2233 do_facet $SINGLEMDS "umount -d $mntpt && rm -f $mdsdev_sym"
2235 if $(echo $mount_op | grep -q "unable to set tunable"); then
2236 error "set tunables failed for symlink device"
2239 [ $rc -eq 0 ] || error "mount symlink $mdsdev_sym failed! rc=$rc"
2243 run_test 37 "verify set tunables works for symlink device"
2245 test_38() { # bug 14222
2246 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
2247 skip "Only applicable to ldiskfs-based MDTs"
2255 FILES=`find $SRC -type f -mtime +1 | head -n $COUNT`
2256 log "copying $(echo $FILES | wc -w) files to $DIR/$tdir"
2258 tar cf - $FILES | tar xf - -C $DIR/$tdir || \
2259 error "copying $SRC to $DIR/$tdir"
2261 umount_client $MOUNT
2263 log "rename lov_objid file on MDS"
2264 rm -f $TMP/lov_objid.orig
2266 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
2267 do_facet $SINGLEMDS "$DEBUGFS -c -R \\\"dump lov_objid $TMP/lov_objid.orig\\\" $MDSDEV"
2268 do_facet $SINGLEMDS "$DEBUGFS -w -R \\\"rm lov_objid\\\" $MDSDEV"
2270 do_facet $SINGLEMDS "od -Ax -td8 $TMP/lov_objid.orig"
2271 # check create in mds_lov_connect
2275 [ $V ] && log "verifying $DIR/$tdir/$f"
2276 diff -q $f $DIR/$tdir/$f || ERROR=y
2278 do_facet $SINGLEMDS "$DEBUGFS -c -R \\\"dump lov_objid $TMP/lov_objid.new\\\" $MDSDEV"
2279 do_facet $SINGLEMDS "od -Ax -td8 $TMP/lov_objid.new"
2280 [ "$ERROR" = "y" ] && error "old and new files are different after connect" || true
2282 # check it's updates in sync
2283 umount_client $MOUNT
2286 do_facet $SINGLEMDS dd if=/dev/zero of=$TMP/lov_objid.clear bs=4096 count=1
2287 do_facet $SINGLEMDS "$DEBUGFS -w -R \\\"rm lov_objid\\\" $MDSDEV"
2288 do_facet $SINGLEMDS "$DEBUGFS -w -R \\\"write $TMP/lov_objid.clear lov_objid\\\" $MDSDEV "
2293 [ $V ] && log "verifying $DIR/$tdir/$f"
2294 diff -q $f $DIR/$tdir/$f || ERROR=y
2296 do_facet $SINGLEMDS "$DEBUGFS -c -R \\\"dump lov_objid $TMP/lov_objid.new1\\\" $MDSDEV"
2297 do_facet $SINGLEMDS "od -Ax -td8 $TMP/lov_objid.new1"
2298 umount_client $MOUNT
2300 [ "$ERROR" = "y" ] && error "old and new files are different after sync" || true
2302 log "files compared the same"
2305 run_test 38 "MDS recreates missing lov_objid file from OST data"
2311 perl $SRCDIR/leak_finder.pl $TMP/debug 2>&1 | egrep '*** Leak:' &&
2312 error "memory leak detected" || true
2314 run_test 39 "leak_finder recognizes both LUSTRE and LNET malloc messages"
2316 test_40() { # bug 15759
2318 #define OBD_FAIL_TGT_TOOMANY_THREADS 0x706
2319 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0x80000706"
2323 run_test 40 "race during service thread startup"
2325 test_41a() { #bug 14134
2326 if [ $(facet_fstype $SINGLEMDS) == ldiskfs ] &&
2327 ! do_facet $SINGLEMDS test -b $(mdsdevname 1); then
2328 skip "Loop devices does not work with nosvc option"
2333 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
2335 start $SINGLEMDS $MDSDEV $MDS_MOUNT_OPTS -o nosvc -n
2336 start ost1 `ostdevname 1` $OST_MOUNT_OPTS
2337 start $SINGLEMDS $MDSDEV $MDS_MOUNT_OPTS -o nomgs,force
2339 mount_client $MOUNT || return 1
2342 echo "blah blah" > $MOUNT/$tfile
2345 umount_client $MOUNT
2346 stop ost1 -f || return 201
2347 stop_mds -f || return 202
2348 stop_mds -f || return 203
2349 unload_modules_conf || return 204
2352 run_test 41a "mount mds with --nosvc and --nomgs"
2355 if [ $(facet_fstype $SINGLEMDS) == ldiskfs ] &&
2356 ! do_facet $SINGLEMDS test -b $(mdsdevname 1); then
2357 skip "Loop devices does not work with nosvc option"
2361 ! combined_mgs_mds && skip "needs combined mgs device" && return 0
2365 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
2367 start $SINGLEMDS $MDSDEV $MDS_MOUNT_OPTS -o nosvc -n
2369 start $SINGLEMDS $MDSDEV $MDS_MOUNT_OPTS -o nomgs,force
2371 mount_client $MOUNT || return 1
2374 echo "blah blah" > $MOUNT/$tfile
2375 cat $MOUNT/$tfile || return 200
2377 umount_client $MOUNT
2378 stop_ost || return 201
2379 stop_mds -f || return 202
2380 stop_mds -f || return 203
2383 run_test 41b "mount mds with --nosvc and --nomgs on first mount"
2385 test_42() { #bug 14693
2387 check_mount || error "client was not mounted"
2389 do_facet mgs $LCTL conf_param $FSNAME.llite.some_wrong_param=10
2390 umount_client $MOUNT ||
2391 error "unmounting client failed with invalid llite param"
2392 mount_client $MOUNT ||
2393 error "mounting client failed with invalid llite param"
2395 do_facet mgs $LCTL conf_param $FSNAME.sys.some_wrong_param=20
2396 cleanup || error "stopping $FSNAME failed with invalid sys param"
2399 check_mount || "client was not mounted with invalid sys param"
2400 cleanup || error "stopping $FSNAME failed with invalid sys param"
2403 run_test 42 "allow client/server mount/unmount with invalid config param"
2406 [ $UID -ne 0 -o $RUNAS_ID -eq 0 ] && skip_env "run as root"
2408 chmod ugo+x $DIR || error "chmod 0 failed"
2409 set_conf_param_and_check mds \
2410 "lctl get_param -n mdt.$FSNAME-MDT0000.root_squash" \
2411 "$FSNAME.mdt.root_squash" \
2413 set_conf_param_and_check mds \
2414 "lctl get_param -n mdt.$FSNAME-MDT0000.nosquash_nids" \
2415 "$FSNAME.mdt.nosquash_nids" \
2419 # create set of test files
2421 echo "111" > $DIR/$tfile-userfile || error "write 1 failed"
2422 chmod go-rw $DIR/$tfile-userfile || error "chmod 1 failed"
2423 chown $RUNAS_ID.$RUNAS_ID $DIR/$tfile-userfile || error "chown failed"
2425 echo "222" > $DIR/$tfile-rootfile || error "write 2 failed"
2426 chmod go-rw $DIR/$tfile-rootfile || error "chmod 2 faield"
2428 mkdir $DIR/$tdir-rootdir -p || error "mkdir failed"
2429 chmod go-rwx $DIR/$tdir-rootdir || error "chmod 3 failed"
2430 touch $DIR/$tdir-rootdir/tfile-1 || error "touch failed"
2433 # check root_squash:
2434 # set root squash UID:GID to RUNAS_ID
2435 # root should be able to access only files owned by RUNAS_ID
2437 set_conf_param_and_check mds \
2438 "lctl get_param -n mdt.$FSNAME-MDT0000.root_squash" \
2439 "$FSNAME.mdt.root_squash" \
2440 "$RUNAS_ID:$RUNAS_ID"
2442 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-userfile)
2443 dd if=$DIR/$tfile-userfile 1>/dev/null 2>/dev/null || \
2444 error "$ST: root read permission is denied"
2445 echo "$ST: root read permission is granted - ok"
2448 dd conv=notrunc if=$DIR/$tfile-userfile 1>/dev/null 2>/dev/null || \
2449 error "$ST: root write permission is denied"
2450 echo "$ST: root write permission is granted - ok"
2452 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-rootfile)
2453 dd if=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null && \
2454 error "$ST: root read permission is granted"
2455 echo "$ST: root read permission is denied - ok"
2458 dd conv=notrunc of=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null && \
2459 error "$ST: root write permission is granted"
2460 echo "$ST: root write permission is denied - ok"
2462 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tdir-rootdir)
2463 rm $DIR/$tdir-rootdir/tfile-1 1>/dev/null 2>/dev/null && \
2464 error "$ST: root unlink permission is granted"
2465 echo "$ST: root unlink permission is denied - ok"
2467 touch $DIR/tdir-rootdir/tfile-2 1>/dev/null 2>/dev/null && \
2468 error "$ST: root create permission is granted"
2469 echo "$ST: root create permission is denied - ok"
2472 # check nosquash_nids:
2473 # put client's NID into nosquash_nids list,
2474 # root should be able to access root file after that
2476 local NIDLIST=$(lctl list_nids all | tr '\n' ' ')
2477 NIDLIST="2@elan $NIDLIST 192.168.0.[2,10]@tcp"
2478 NIDLIST=$(echo $NIDLIST | tr -s ' ' ' ')
2479 set_conf_param_and_check mds \
2480 "lctl get_param -n mdt.$FSNAME-MDT0000.nosquash_nids" \
2481 "$FSNAME-MDTall.mdt.nosquash_nids" \
2484 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-rootfile)
2485 dd if=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null || \
2486 error "$ST: root read permission is denied"
2487 echo "$ST: root read permission is granted - ok"
2490 dd conv=notrunc of=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null || \
2491 error "$ST: root write permission is denied"
2492 echo "$ST: root write permission is granted - ok"
2494 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tdir-rootdir)
2495 rm $DIR/$tdir-rootdir/tfile-1 || \
2496 error "$ST: root unlink permission is denied"
2497 echo "$ST: root unlink permission is granted - ok"
2498 touch $DIR/$tdir-rootdir/tfile-2 || \
2499 error "$ST: root create permission is denied"
2500 echo "$ST: root create permission is granted - ok"
2504 run_test 43 "check root_squash and nosquash_nids"
2506 umount_client $MOUNT
2511 check_mount || return 2
2512 UUID=$($LCTL get_param llite.${FSNAME}*.uuid | cut -d= -f2)
2514 UUIDS=$(do_facet $SINGLEMDS "$LCTL get_param mdt.${FSNAME}*.exports.*.uuid")
2515 for VAL in $UUIDS; do
2516 NID=$(echo $VAL | cut -d= -f1)
2517 CLUUID=$(echo $VAL | cut -d= -f2)
2518 [ "$UUID" = "$CLUUID" ] && STATS_FOUND=yes && break
2520 [ "$STATS_FOUND" = "no" ] && error "stats not found for client"
2524 run_test 44 "mounted client proc entry exists"
2528 check_mount || return 2
2533 #define OBD_FAIL_PTLRPC_LONG_UNLINK 0x50f
2534 do_facet client "lctl set_param fail_loc=0x50f"
2537 manual_umount_client --force || return 3
2538 do_facet client "lctl set_param fail_loc=0x0"
2540 mount_client $MOUNT || return 4
2544 run_test 45 "long unlink handling in ptlrpcd"
2551 umount_client $MOUNT2 || rc=$?
2552 umount_client $MOUNT || rc=$?
2553 while [ $count -gt 0 ]; do
2554 stop ost${count} -f || rc=$?
2558 cleanup_nocli || rc=$?
2559 #writeconf to remove all ost2 traces for subsequent tests
2560 writeconf_or_reformat
2565 echo "Testing with $OSTCOUNT OSTs"
2567 start_mds || return 1
2568 #first client should see only one ost
2569 start_ost || return 2
2570 wait_osc_import_state mds ost FULL
2572 mount_client $MOUNT || return 3
2573 trap "cleanup_46a $OSTCOUNT" EXIT ERR
2576 for (( i=2; i<=$OSTCOUNT; i++ )); do
2577 start ost$i `ostdevname $i` $OST_MOUNT_OPTS || return $((i+2))
2580 # wait until osts in sync
2581 for (( i=2; i<=$OSTCOUNT; i++ )); do
2582 wait_osc_import_state mds ost$i FULL
2583 wait_osc_import_state client ost$i FULL
2586 #second client see all ost's
2588 mount_client $MOUNT2 || return 8
2589 $LFS setstripe -c -1 $MOUNT2 || return 9
2590 $LFS getstripe $MOUNT2 || return 10
2592 echo "ok" > $MOUNT2/widestripe
2593 $LFS getstripe $MOUNT2/widestripe || return 11
2594 # fill acl buffer for avoid expand lsm to them
2595 awk -F : '{if (FNR < 25) { print "u:"$1":rwx" }}' /etc/passwd | while read acl; do
2596 setfacl -m $acl $MOUNT2/widestripe
2600 stat $MOUNT/widestripe || return 12
2602 cleanup_46a $OSTCOUNT || { echo "cleanup_46a failed!" && return 13; }
2605 run_test 46a "handle ost additional - wide striped file"
2610 check_mount || return 2
2611 $LCTL set_param ldlm.namespaces.$FSNAME-*-*-*.lru_size=100
2615 for ns in $($LCTL get_param ldlm.namespaces.$FSNAME-*-*-*.lru_size); do
2616 if echo $ns | grep "MDT[[:digit:]]*"; then
2619 lrs=$(echo $ns | sed 's/.*lru_size=//')
2620 lru_size[count]=$lrs
2625 facet_failover $SINGLEMDS
2626 client_up || return 3
2629 for ns in $($LCTL get_param ldlm.namespaces.$FSNAME-*-*-*.lru_size); do
2630 if echo $ns | grep "MDT[[:digit:]]*"; then
2633 lrs=$(echo $ns | sed 's/.*lru_size=//')
2634 if ! test "$lrs" -eq "${lru_size[count]}"; then
2635 n=$(echo $ns | sed -e 's/ldlm.namespaces.//' -e 's/.lru_size=.*//')
2636 error "$n has lost lru_size: $lrs vs. ${lru_size[count]}"
2644 run_test 47 "server restart does not make client loss lru_resize settings"
2649 # reformat after this test is needed - if test will failed
2650 # we will have unkillable file at FS
2654 test_48() { # bug 17636
2657 check_mount || return 2
2659 $LFS setstripe -c -1 $MOUNT || return 9
2660 $LFS getstripe $MOUNT || return 10
2662 echo "ok" > $MOUNT/widestripe
2663 $LFS getstripe $MOUNT/widestripe || return 11
2665 trap cleanup_48 EXIT ERR
2667 # fill acl buffer for avoid expand lsm to them
2668 getent passwd | awk -F : '{ print "u:"$1":rwx" }' | while read acl; do
2669 setfacl -m $acl $MOUNT/widestripe
2672 stat $MOUNT/widestripe || return 12
2677 run_test 48 "too many acls on file"
2679 # check PARAM_SYS_LDLM_TIMEOUT option of MKFS.LUSTRE
2680 test_49() { # bug 17710
2681 local timeout_orig=$TIMEOUT
2682 local ldlm_timeout_orig=$LDLM_TIMEOUT
2683 local LOCAL_TIMEOUT=20
2685 LDLM_TIMEOUT=$LOCAL_TIMEOUT
2686 TIMEOUT=$LOCAL_TIMEOUT
2690 check_mount || return 1
2692 echo "check ldlm_timout..."
2693 LDLM_MDS="`do_facet $SINGLEMDS lctl get_param -n ldlm_timeout`"
2694 LDLM_OST1="`do_facet ost1 lctl get_param -n ldlm_timeout`"
2695 LDLM_CLIENT="`do_facet client lctl get_param -n ldlm_timeout`"
2697 if [ $LDLM_MDS -ne $LDLM_OST1 ] || [ $LDLM_MDS -ne $LDLM_CLIENT ]; then
2698 error "Different LDLM_TIMEOUT:$LDLM_MDS $LDLM_OST1 $LDLM_CLIENT"
2701 if [ $LDLM_MDS -ne $((LOCAL_TIMEOUT / 3)) ]; then
2702 error "LDLM_TIMEOUT($LDLM_MDS) is not correct"
2705 umount_client $MOUNT
2706 stop_ost || return 2
2707 stop_mds || return 3
2709 LDLM_TIMEOUT=$((LOCAL_TIMEOUT - 1))
2713 check_mount || return 7
2715 LDLM_MDS="`do_facet $SINGLEMDS lctl get_param -n ldlm_timeout`"
2716 LDLM_OST1="`do_facet ost1 lctl get_param -n ldlm_timeout`"
2717 LDLM_CLIENT="`do_facet client lctl get_param -n ldlm_timeout`"
2719 if [ $LDLM_MDS -ne $LDLM_OST1 ] || [ $LDLM_MDS -ne $LDLM_CLIENT ]; then
2720 error "Different LDLM_TIMEOUT:$LDLM_MDS $LDLM_OST1 $LDLM_CLIENT"
2723 if [ $LDLM_MDS -ne $((LOCAL_TIMEOUT - 1)) ]; then
2724 error "LDLM_TIMEOUT($LDLM_MDS) is not correct"
2727 cleanup || return $?
2729 LDLM_TIMEOUT=$ldlm_timeout_orig
2730 TIMEOUT=$timeout_orig
2732 run_test 49 "check PARAM_SYS_LDLM_TIMEOUT option of MKFS.LUSTRE"
2735 # Test both statfs and lfs df and fail if either one fails
2736 multiop_bg_pause $1 f_
2739 killall -USR1 multiop
2740 [ $RC1 -ne 0 ] && log "lazystatfs multiop failed"
2741 wait $PID || { RC1=$?; log "multiop return error "; }
2748 if [ $RC2 -eq 0 ]; then
2750 log "lazystatfs df failed"
2754 [[ $RC1 -ne 0 || $RC2 -eq 0 ]] && RC=1
2760 lctl set_param llite.$FSNAME-*.lazystatfs=1
2763 lazystatfs $MOUNT || error "lazystatfs failed but no down servers"
2765 cleanup || return $?
2767 run_test 50a "lazystatfs all servers available =========================="
2771 lctl set_param llite.$FSNAME-*.lazystatfs=1
2774 # Wait for client to detect down OST
2775 stop_ost || error "Unable to stop OST1"
2776 wait_osc_import_state mds ost DISCONN
2778 lazystatfs $MOUNT || error "lazystatfs should don't have returned EIO"
2780 umount_client $MOUNT || error "Unable to unmount client"
2781 stop_mds || error "Unable to stop MDS"
2783 run_test 50b "lazystatfs all servers down =========================="
2786 start_mds || error "Unable to start MDS"
2787 start_ost || error "Unable to start OST1"
2788 start_ost2 || error "Unable to start OST2"
2789 mount_client $MOUNT || error "Unable to mount client"
2790 lctl set_param llite.$FSNAME-*.lazystatfs=1
2793 # Wait for client to detect down OST
2794 stop_ost || error "Unable to stop OST1"
2795 wait_osc_import_state mds ost DISCONN
2796 lazystatfs $MOUNT || error "lazystatfs failed with one down server"
2798 umount_client $MOUNT || error "Unable to unmount client"
2799 stop_ost2 || error "Unable to stop OST2"
2800 stop_mds || error "Unable to stop MDS"
2801 #writeconf to remove all ost2 traces for subsequent tests
2802 writeconf_or_reformat
2804 run_test 50c "lazystatfs one server down =========================="
2807 start_mds || error "Unable to start MDS"
2808 start_ost || error "Unable to start OST1"
2809 start_ost2 || error "Unable to start OST2"
2810 mount_client $MOUNT || error "Unable to mount client"
2811 lctl set_param llite.$FSNAME-*.lazystatfs=1
2814 # Issue the statfs during the window where the client still
2815 # belives the OST to be available but it is in fact down.
2816 # No failure just a statfs which hangs for a timeout interval.
2817 stop_ost || error "Unable to stop OST1"
2818 lazystatfs $MOUNT || error "lazystatfs failed with one down server"
2820 umount_client $MOUNT || error "Unable to unmount client"
2821 stop_ost2 || error "Unable to stop OST2"
2822 stop_mds || error "Unable to stop MDS"
2823 #writeconf to remove all ost2 traces for subsequent tests
2824 writeconf_or_reformat
2826 run_test 50d "lazystatfs client/server conn race =========================="
2833 start_mds || return 1
2834 #first client should see only one ost
2835 start_ost || return 2
2836 wait_osc_import_state mds ost FULL
2838 # Wait for client to detect down OST
2839 stop_ost || error "Unable to stop OST1"
2840 wait_osc_import_state mds ost DISCONN
2842 mount_client $MOUNT || error "Unable to mount client"
2843 lctl set_param llite.$FSNAME-*.lazystatfs=0
2845 multiop_bg_pause $MOUNT _f
2849 if [ $RC1 -ne 0 ]; then
2850 log "multiop failed $RC1"
2853 sleep $(( $TIMEOUT+1 ))
2855 [ $? -ne 0 ] && error "process isn't sleep"
2856 start_ost || error "Unable to start OST1"
2857 wait $pid || error "statfs failed"
2860 umount_client $MOUNT || error "Unable to unmount client"
2861 stop_ost || error "Unable to stop OST1"
2862 stop_mds || error "Unable to stop MDS"
2864 run_test 50e "normal statfs all servers down =========================="
2869 CONN_PROC="osc.$FSNAME-OST0001-osc-[M]*.ost_server_uuid"
2871 start_mds || error "Unable to start mds"
2872 #first client should see only one ost
2873 start_ost || error "Unable to start OST1"
2874 wait_osc_import_state mds ost FULL
2876 start_ost2 || error "Unable to start OST2"
2877 wait_osc_import_state mds ost2 FULL
2879 # Wait for client to detect down OST
2880 stop_ost2 || error "Unable to stop OST2"
2882 wait_osc_import_state mds ost2 DISCONN
2883 mount_client $MOUNT || error "Unable to mount client"
2884 lctl set_param llite.$FSNAME-*.lazystatfs=0
2886 multiop_bg_pause $MOUNT _f
2890 if [ $RC1 -ne 0 ]; then
2891 log "lazystatfs multiop failed $RC1"
2894 sleep $(( $TIMEOUT+1 ))
2896 [ $? -ne 0 ] && error "process isn't sleep"
2897 start_ost2 || error "Unable to start OST2"
2898 wait $pid || error "statfs failed"
2899 stop_ost2 || error "Unable to stop OST2"
2902 umount_client $MOUNT || error "Unable to unmount client"
2903 stop_ost || error "Unable to stop OST1"
2904 stop_mds || error "Unable to stop MDS"
2905 #writeconf to remove all ost2 traces for subsequent tests
2906 writeconf_or_reformat
2908 run_test 50f "normal statfs one server in down =========================="
2911 [ "$OSTCOUNT" -lt "2" ] && skip_env "$OSTCOUNT < 2, skipping" && return
2913 start_ost2 || error "Unable to start OST2"
2914 wait_osc_import_state mds ost2 FULL
2915 wait_osc_import_state client ost2 FULL
2917 local PARAM="${FSNAME}-OST0001.osc.active"
2919 $LFS setstripe -c -1 $DIR/$tfile || error "Unable to lfs setstripe"
2920 do_facet mgs $LCTL conf_param $PARAM=0 || error "Unable to deactivate OST"
2922 umount_client $MOUNT || error "Unable to unmount client"
2923 mount_client $MOUNT || error "Unable to mount client"
2924 # This df should not cause a panic
2927 do_facet mgs $LCTL conf_param $PARAM=1 || error "Unable to activate OST"
2929 umount_client $MOUNT || error "Unable to unmount client"
2930 stop_ost2 || error "Unable to stop OST2"
2931 stop_ost || error "Unable to stop OST1"
2932 stop_mds || error "Unable to stop MDS"
2933 #writeconf to remove all ost2 traces for subsequent tests
2934 writeconf_or_reformat
2936 run_test 50g "deactivated OST should not cause panic====================="
2940 # prepare MDT/OST, make OSC inactive for OST1
2941 [ "$OSTCOUNT" -lt "2" ] && skip_env "$OSTCOUNT < 2, skipping" && return
2942 do_facet ost1 "$TUNEFS --param osc.active=0 `ostdevname 1`" ||
2943 error "tunefs OST1 failed"
2944 start_mds || error "Unable to start MDT"
2945 start_ost || error "Unable to start OST1"
2946 start_ost2 || error "Unable to start OST2"
2947 mount_client $MOUNT || error "client start failed"
2951 # activatate OSC for OST1
2952 local TEST="$LCTL get_param -n osc.${FSNAME}-OST0000-osc-[!M]*.active"
2953 set_conf_param_and_check client \
2954 "$TEST" "${FSNAME}-OST0000.osc.active" 1 ||
2955 error "Unable to activate OST1"
2957 mkdir -p $DIR/$tdir/2
2958 $LFS setstripe -c -1 -i 0 $DIR/$tdir/2
2959 sleep 1 && echo "create a file after OST1 is activated"
2961 createmany -o $DIR/$tdir/2/$tfile-%d 1
2963 # check OSC import is working
2964 stat $DIR/$tdir/2/* >/dev/null 2>&1 ||
2965 error "some OSC imports are still not connected"
2968 umount_client $MOUNT || error "Unable to umount client"
2969 stop_ost2 || error "Unable to stop OST2"
2972 run_test 50h "LU-642: activate deactivated OST ==="
2975 local LOCAL_TIMEOUT=20
2979 check_mount || return 1
2982 $LFS setstripe -c -1 $MOUNT/d1
2983 #define OBD_FAIL_MDS_REINT_DELAY 0x142
2984 do_facet $SINGLEMDS "lctl set_param fail_loc=0x142"
2985 touch $MOUNT/d1/f1 &
2988 start_ost2 || return 2
2990 stop_ost2 || return 3
2992 #writeconf to remove all ost2 traces for subsequent tests
2993 writeconf_or_reformat
2995 run_test 51 "Verify that mdt_reint handles RMF_MDT_MD correctly when an OST is added"
3004 do_node $node mkdir -p $dest
3005 [ $? -eq 0 ] || { error "Unable to create directory"; return 1; }
3007 do_node $node 'tar cf - '$@' | tar xf - -C '$dest';
3008 [ \"\${PIPESTATUS[*]}\" = \"0 0\" ] || exit 1'
3009 [ $? -eq 0 ] || { error "Unable to tar files"; return 2; }
3011 do_node $node 'getfattr -d -m "[a-z]*\\." '$@' > '$xattrs
3012 [ $? -eq 0 ] || { error "Unable to read xattrs"; return 3; }
3022 local backup2=${TMP}/backup2
3024 do_node $node mkdir -p $backup2
3025 [ $? -eq 0 ] || { error "Unable to create directory"; return 1; }
3027 do_node $node 'tar cf - '$@' | tar xf - -C '$backup2';
3028 [ \"\${PIPESTATUS[*]}\" = \"0 0\" ] || exit 1'
3029 [ $? -eq 0 ] || { error "Unable to tar files to diff"; return 2; }
3031 do_node $node "diff -rq $backup $backup2"
3032 [ $? -eq 0 ] || { error "contents differ"; return 3; }
3034 local xattrs2=${TMP}/xattrs2
3035 do_node $node 'getfattr -d -m "[a-z]*\\." '$@' > '$xattrs2
3036 [ $? -eq 0 ] || { error "Unable to read xattrs to diff"; return 4; }
3038 do_node $node "diff $xattrs $xattrs2"
3039 [ $? -eq 0 ] || { error "xattrs differ"; return 5; }
3041 do_node $node "rm -rf $backup2 $xattrs2"
3042 [ $? -eq 0 ] || { error "Unable to delete temporary files"; return 6; }
3046 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
3047 skip "Only applicable to ldiskfs-based MDTs"
3052 [ $? -eq 0 ] || { error "Unable to start MDS"; return 1; }
3054 [ $? -eq 0 ] || { error "Unable to start OST1"; return 2; }
3056 [ $? -eq 0 ] || { error "Unable to mount client"; return 3; }
3059 local ost1mnt=$(facet_mntpt ost1)
3060 local ost1node=$(facet_active_host ost1)
3061 local ost1tmp=$TMP/conf52
3065 [ $? -eq 0 ] || { error "Unable to create tdir"; return 4; }
3066 touch $TMP/modified_first
3067 [ $? -eq 0 ] || { error "Unable to create temporary file"; return 5; }
3068 local mtime=$(stat -c %Y $TMP/modified_first)
3069 do_node $ost1node "mkdir -p $ost1tmp && touch -m -d @$mtime $ost1tmp/modified_first"
3071 [ $? -eq 0 ] || { error "Unable to create temporary file"; return 6; }
3074 $LFS setstripe -c -1 -S 1M $DIR/$tdir
3075 [ $? -eq 0 ] || { error "lfs setstripe failed"; return 7; }
3077 for (( i=0; i < nrfiles; i++ )); do
3078 multiop $DIR/$tdir/$tfile-$i Ow1048576w1048576w524288c
3079 [ $? -eq 0 ] || { error "multiop failed"; return 8; }
3085 echo backup files to $TMP/files
3086 local files=$(find $DIR/$tdir -type f -newer $TMP/modified_first)
3087 copy_files_xattrs `hostname` $TMP/files $TMP/file_xattrs $files
3088 [ $? -eq 0 ] || { error "Unable to copy files"; return 9; }
3090 umount_client $MOUNT
3091 [ $? -eq 0 ] || { error "Unable to umount client"; return 10; }
3093 [ $? -eq 0 ] || { error "Unable to stop ost1"; return 11; }
3095 echo mount ost1 as ldiskfs
3096 do_node $ost1node mkdir -p $ost1mnt
3097 [ $? -eq 0 ] || { error "Unable to create $ost1mnt"; return 23; }
3098 if ! do_node $ost1node test -b $ost1_dev; then
3101 do_node $ost1node mount -t $(facet_fstype ost1) $loop $ost1_dev \
3103 [ $? -eq 0 ] || { error "Unable to mount ost1 as ldiskfs"; return 12; }
3106 echo backup objects to $ost1tmp/objects
3107 local objects=$(do_node $ost1node 'find '$ost1mnt'/O/[0-9]* -type f'\
3108 '-size +0 -newer '$ost1tmp'/modified_first -regex ".*\/[0-9]+"')
3109 copy_files_xattrs $ost1node $ost1tmp/objects $ost1tmp/object_xattrs \
3111 [ $? -eq 0 ] || { error "Unable to copy objects"; return 13; }
3113 # move objects to lost+found
3114 do_node $ost1node 'mv '$objects' '${ost1mnt}'/lost+found'
3115 [ $? -eq 0 ] || { error "Unable to move objects"; return 14; }
3118 do_node $ost1node "ll_recover_lost_found_objs -d $ost1mnt/lost+found"
3119 [ $? -eq 0 ] || { error "ll_recover_lost_found_objs failed"; return 15; }
3121 # compare restored objects against saved ones
3122 diff_files_xattrs $ost1node $ost1tmp/objects $ost1tmp/object_xattrs $objects
3123 [ $? -eq 0 ] || { error "Unable to diff objects"; return 16; }
3125 do_node $ost1node "umount $ost1mnt"
3126 [ $? -eq 0 ] || { error "Unable to umount ost1 as ldiskfs"; return 17; }
3129 [ $? -eq 0 ] || { error "Unable to start ost1"; return 18; }
3131 [ $? -eq 0 ] || { error "Unable to mount client"; return 19; }
3134 diff_files_xattrs `hostname` $TMP/files $TMP/file_xattrs $files
3135 [ $? -eq 0 ] || { error "Unable to diff files"; return 20; }
3137 rm -rf $TMP/files $TMP/file_xattrs
3138 [ $? -eq 0 ] || { error "Unable to delete temporary files"; return 21; }
3139 do_node $ost1node "rm -rf $ost1tmp"
3140 [ $? -eq 0 ] || { error "Unable to delete temporary files"; return 22; }
3143 run_test 52 "check recovering objects from lost+found"
3145 # Checks threads_min/max/started for some service
3147 # Arguments: service name (OST or MDT), facet (e.g., ost1, $SINGLEMDS), and a
3148 # parameter pattern prefix like 'ost.*.ost'.
3161 local msg="Insane $modname thread counts"
3162 local ncpts=$(check_cpt_number $facet)
3166 check_mount || return 41
3168 # We need to expand $parampat, but it may match multiple parameters, so
3169 # we'll pick the first one
3170 if ! paramp=$(do_facet $facet "lctl get_param -N ${parampat}.threads_min"|head -1); then
3171 error "Couldn't expand ${parampat}.threads_min parameter name"
3175 # Remove the .threads_min part
3176 paramp=${paramp%.threads_min}
3178 # Check for sanity in defaults
3179 tmin=$(do_facet $facet "lctl get_param -n ${paramp}.threads_min" || echo 0)
3180 tmax=$(do_facet $facet "lctl get_param -n ${paramp}.threads_max" || echo 0)
3181 tstarted=$(do_facet $facet "lctl get_param -n ${paramp}.threads_started" || echo 0)
3182 lassert 23 "$msg (PDSH problems?)" '(($tstarted && $tmin && $tmax))' || return $?
3183 lassert 24 "$msg" '(($tstarted >= $tmin && $tstarted <= $tmax ))' || return $?
3184 nthrs=$(expr $tmax - $tmin)
3185 if [ $nthrs -lt $ncpts ]; then
3191 [ $tmin -eq $tmax -a $tmin -eq $tstarted ] &&
3192 skip_env "module parameter forced $facet thread count" &&
3193 tmin=3 && tmax=$((3 * tmax))
3195 # Check that we can change min/max
3196 do_facet $facet "lctl set_param ${paramp}.threads_min=$((tmin + nthrs))"
3197 do_facet $facet "lctl set_param ${paramp}.threads_max=$((tmax - nthrs))"
3198 tmin2=$(do_facet $facet "lctl get_param -n ${paramp}.threads_min" || echo 0)
3199 tmax2=$(do_facet $facet "lctl get_param -n ${paramp}.threads_max" || echo 0)
3200 lassert 25 "$msg" '(($tmin2 == ($tmin + $nthrs) && $tmax2 == ($tmax - $nthrs)))' || return $?
3202 # Check that we can set min/max to the same value
3203 tmin=$(do_facet $facet "lctl get_param -n ${paramp}.threads_min" || echo 0)
3204 do_facet $facet "lctl set_param ${paramp}.threads_max=$tmin"
3205 tmin2=$(do_facet $facet "lctl get_param -n ${paramp}.threads_min" || echo 0)
3206 tmax2=$(do_facet $facet "lctl get_param -n ${paramp}.threads_max" || echo 0)
3207 lassert 26 "$msg" '(($tmin2 == $tmin && $tmax2 == $tmin))' || return $?
3209 # Check that we can't set max < min
3210 do_facet $facet "lctl set_param ${paramp}.threads_max=$((tmin - 1))"
3211 tmin2=$(do_facet $facet "lctl get_param -n ${paramp}.threads_min" || echo 0)
3212 tmax2=$(do_facet $facet "lctl get_param -n ${paramp}.threads_max" || echo 0)
3213 lassert 27 "$msg" '(($tmin2 <= $tmax2))' || return $?
3215 # We need to ensure that we get the module options desired; to do this
3216 # we set LOAD_MODULES_REMOTE=true and we call setmodopts below.
3217 LOAD_MODULES_REMOTE=true
3220 local newvalue="${opts}=$(expr $basethr \* $ncpts)"
3221 setmodopts -a $modname "$newvalue" oldvalue
3225 check_mount || return 41
3227 # Restore previous setting of MODOPTS_*
3228 setmodopts $modname "$oldvalue"
3230 # Check that $opts took
3231 tmin=$(do_facet $facet "lctl get_param -n ${paramp}.threads_min")
3232 tmax=$(do_facet $facet "lctl get_param -n ${paramp}.threads_max")
3233 tstarted=$(do_facet $facet "lctl get_param -n ${paramp}.threads_started")
3234 lassert 28 "$msg" '(($tstarted == $tmin && $tstarted == $tmax ))' || return $?
3243 thread_sanity OST ost1 'ost.*.ost' 'oss_num_threads' '16'
3246 run_test 53a "check OSS thread count params"
3250 local mds=$(do_facet $SINGLEMDS "lctl get_param -N mds.*.*.threads_max \
3252 if [ -z "$mds" ]; then
3253 #running this on an old MDT
3254 thread_sanity MDT $SINGLEMDS 'mdt.*.*.' 'mdt_num_threads' 16
3256 thread_sanity MDT $SINGLEMDS 'mds.*.*.' 'mds_num_threads' 16
3260 run_test 53b "check MDS thread count params"
3263 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
3264 skip "Only applicable to ldiskfs-based MDTs"
3268 do_rpc_nodes $(facet_host ost1) run_llverdev $(ostdevname 1) -p
3269 [ $? -eq 0 ] || error "llverdev failed!"
3272 run_test 54a "test llverdev and partial verify of device"
3275 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
3276 skip "Only applicable to ldiskfs-based MDTs"
3281 run_llverfs $MOUNT -p
3282 [ $? -eq 0 ] || error "llverfs failed!"
3285 run_test 54b "test llverfs and partial verify of filesystem"
3289 local max_ost_index=$1
3290 echo -n $(((max_ost_index + 1) * 8))
3294 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
3295 skip "Only applicable to ldiskfs-based MDTs"
3299 local mdsdev=$(mdsdevname 1)
3300 local mdsvdev=$(mdsvdevname 1)
3304 add mds1 $(mkfs_opts mds1 ${mdsdev}) --reformat $mdsdev \
3306 add ost1 $(mkfs_opts ost1 $(ostdevname 1)) --index=$i \
3307 --reformat $(ostdevname 1) $(ostvdevname 1)
3313 echo checking size of lov_objid for ost index $i
3314 LOV_OBJID_SIZE=$(do_facet mds1 "$DEBUGFS -R 'stat lov_objid' $mdsdev 2>/dev/null" | grep ^User | awk '{print $6}')
3315 if [ "$LOV_OBJID_SIZE" != $(lov_objid_size $i) ]; then
3316 error "lov_objid size has to be $(lov_objid_size $i), not $LOV_OBJID_SIZE"
3318 echo ok, lov_objid size is correct: $LOV_OBJID_SIZE
3325 run_test 55 "check lov_objid size"
3328 local mds_journal_size_orig=$MDSJOURNALSIZE
3332 for num in $(seq 1 $MDSCOUNT); do
3333 add mds${num} $(mkfs_opts mds${num} $(mdsdevname $num)) \
3334 --reformat $(mdsdevname $num) $(mdsvdevname $num)
3336 add ost1 $(mkfs_opts ost1 $(ostdevname 1)) --index=1000 --reformat \
3337 $(ostdevname 1) $(ostvdevname 1)
3338 add ost2 $(mkfs_opts ost2 $(ostdevname 2)) --index=10000 --reformat \
3339 $(ostdevname 2) $(ostvdevname 2)
3343 start_ost2 || error "Unable to start second ost"
3344 mount_client $MOUNT || error "Unable to mount client"
3348 MDSJOURNALSIZE=$mds_journal_size_orig
3351 run_test 56 "check big indexes"
3353 test_57a() { # bug 22656
3354 local NID=$(do_facet ost1 "$LCTL get_param nis" | tail -1 | awk '{print $1}')
3355 writeconf_or_reformat
3356 do_facet ost1 "$TUNEFS --failnode=$NID `ostdevname 1`" || error "tunefs failed"
3358 start_ost && error "OST registration from failnode should fail"
3361 run_test 57a "initial registration from failnode should fail (should return errs)"
3364 local NID=$(do_facet ost1 "$LCTL get_param nis" | tail -1 | awk '{print $1}')
3365 writeconf_or_reformat
3366 do_facet ost1 "$TUNEFS --servicenode=$NID `ostdevname 1`" || error "tunefs failed"
3368 start_ost || error "OST registration from servicenode should not fail"
3371 run_test 57b "initial registration from servicenode should not fail"
3374 do_facet mgs $LCTL get_param mgs.MGS.live.$FSNAME | grep OST | wc -l
3377 test_58() { # bug 22658
3378 if [ $(facet_fstype mds) != ldiskfs ]; then
3379 skip "Only applicable to ldiskfs-based MDTs"
3384 createmany -o $DIR/$tdir/$tfile-%d 100
3385 # make sure that OSTs do not cancel llog cookies before we unmount the MDS
3386 #define OBD_FAIL_OBD_LOG_CANCEL_NET 0x601
3387 do_facet $SINGLEMDS "lctl set_param fail_loc=0x601"
3388 unlinkmany $DIR/$tdir/$tfile-%d 100
3391 local MNTDIR=$(facet_mntpt $SINGLEMDS)
3392 local devname=$(mdsdevname ${SINGLEMDS//mds/})
3394 if ! do_facet $SINGLEMDS "test -b $devname"; then
3398 # remove all files from the OBJECTS dir
3399 do_facet $SINGLEMDS "mount -t ldiskfs $opts $devname $MNTDIR"
3400 do_facet $SINGLEMDS "find $MNTDIR/O/1/d* -type f -delete"
3401 do_facet $SINGLEMDS "umount $MNTDIR"
3402 # restart MDS with missing llog files
3404 do_facet mds "lctl set_param fail_loc=0"
3407 run_test 58 "missing llog files must not prevent MDT from mounting"
3410 start_mgsmds >> /dev/null
3411 local C1=$(count_osts)
3412 if [ $C1 -eq 0 ]; then
3413 start_ost >> /dev/null
3417 echo "original ost count: $C1 (expect > 0)"
3418 [ $C1 -gt 0 ] || error "No OSTs in $FSNAME log"
3419 start_mgsmds -o writeconf >> /dev/null || error "MDT start failed"
3420 local C2=$(count_osts)
3421 echo "after mdt writeconf count: $C2 (expect 0)"
3422 [ $C2 -gt 0 ] && error "MDT writeconf should erase OST logs"
3423 echo "OST start without writeconf should fail:"
3424 start_ost >> /dev/null && error "OST start without writeconf didn't fail"
3425 echo "OST start with writeconf should succeed:"
3426 start_ost -o writeconf >> /dev/null || error "OST1 start failed"
3427 local C3=$(count_osts)
3428 echo "after ost writeconf count: $C3 (expect 1)"
3429 [ $C3 -eq 1 ] || error "new OST writeconf should add:"
3430 start_ost2 -o writeconf >> /dev/null || error "OST2 start failed"
3431 local C4=$(count_osts)
3432 echo "after ost2 writeconf count: $C4 (expect 2)"
3433 [ $C4 -eq 2 ] || error "OST2 writeconf should add log"
3434 stop_ost2 >> /dev/null
3435 cleanup_nocli >> /dev/null
3436 #writeconf to remove all ost2 traces for subsequent tests
3437 writeconf_or_reformat
3439 run_test 59 "writeconf mount option"
3441 test_60() { # LU-471
3444 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
3445 skip "Only applicable to ldiskfs-based MDTs"
3449 for num in $(seq $MDSCOUNT); do
3450 add mds${num} $(mkfs_opts mds${num} $(mdsdevname $num)) \
3451 --mkfsoptions='\" -E stride=64 -O ^uninit_bg\"' \
3452 --reformat $(mdsdevname $num) $(mdsvdevname $num) ||
3456 dump=$(do_facet $SINGLEMDS dumpe2fs $(mdsdevname 1))
3458 [ $rc -eq 0 ] || error "dumpe2fs $(mdsdevname 1) failed"
3460 # MDT default has dirdata feature
3461 echo $dump | grep dirdata > /dev/null || error "dirdata is not set"
3462 # we disable uninit_bg feature
3463 echo $dump | grep uninit_bg > /dev/null && error "uninit_bg is set"
3464 # we set stride extended options
3465 echo $dump | grep stride > /dev/null || error "stride is not set"
3468 run_test 60 "check mkfs.lustre --mkfsoptions -E -O options setting"
3471 local reformat=false
3473 [ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.1.53) ] ||
3474 { skip "Need MDS version at least 2.1.53"; return 0; }
3476 if [ $(facet_fstype $SINGLEMDS) == ldiskfs ] &&
3477 ! large_xattr_enabled; then
3479 LDISKFS_MKFS_OPTS+=" -O large_xattr"
3481 for num in $(seq $MDSCOUNT); do
3482 add mds${num} $(mkfs_opts mds$num $(mdsdevname $num)) \
3483 --reformat $(mdsdevname $num) $(mdsvdevname $num) ||
3484 error "add mds $num failed"
3488 setup_noconfig || error "setting up the filesystem failed"
3489 client_up || error "starting client failed"
3491 local file=$DIR/$tfile
3494 local large_value="$(generate_string $(max_xattr_size))"
3495 local small_value="bar"
3497 local name="trusted.big"
3498 log "save large xattr $name on $file"
3499 setfattr -n $name -v $large_value $file ||
3500 error "saving $name on $file failed"
3502 local new_value=$(get_xattr_value $name $file)
3503 [[ "$new_value" != "$large_value" ]] &&
3504 error "$name different after saving"
3506 log "shrink value of $name on $file"
3507 setfattr -n $name -v $small_value $file ||
3508 error "shrinking value of $name on $file failed"
3510 new_value=$(get_xattr_value $name $file)
3511 [[ "$new_value" != "$small_value" ]] &&
3512 error "$name different after shrinking"
3514 log "grow value of $name on $file"
3515 setfattr -n $name -v $large_value $file ||
3516 error "growing value of $name on $file failed"
3518 new_value=$(get_xattr_value $name $file)
3519 [[ "$new_value" != "$large_value" ]] &&
3520 error "$name different after growing"
3522 log "check value of $name on $file after remounting MDS"
3524 new_value=$(get_xattr_value $name $file)
3525 [[ "$new_value" != "$large_value" ]] &&
3526 error "$name different after remounting MDS"
3528 log "remove large xattr $name from $file"
3529 setfattr -x $name $file || error "removing $name from $file failed"
3534 LDISKFS_MKFS_OPTS=${LDISKFS_MKFS_OPTS% -O large_xattr}
3538 run_test 61 "large xattr"
3541 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
3542 skip "Only applicable to ldiskfs-based MDTs"
3547 local mdsdev=$(mdsdevname 1)
3548 local ostdev=$(ostdevname 1)
3550 [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.2.51) ]] ||
3551 { skip "Need MDS version at least 2.2.51"; return 0; }
3553 echo "disable journal for mds"
3554 do_facet mds tune2fs -O ^has_journal $mdsdev || error "tune2fs failed"
3555 start_mds && error "MDT start should fail"
3556 echo "disable journal for ost"
3557 do_facet ost1 tune2fs -O ^has_journal $ostdev || error "tune2fs failed"
3558 start_ost && error "OST start should fail"
3559 cleanup || return $?
3562 run_test 62 "start with disabled journal"
3565 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
3566 skip "Only applicable to ldiskfs-based MDTs"
3570 local inode_slab=$(do_facet $SINGLEMDS \
3571 "awk '/ldiskfs_inode_cache/ { print \\\$5 }' /proc/slabinfo")
3572 if [ -z "$inode_slab" ]; then
3573 skip "ldiskfs module has not been loaded"
3577 echo "$inode_slab ldisk inodes per page"
3578 [ "$inode_slab" -ge "3" ] ||
3579 error "ldisk inode size is too big, $inode_slab objs per page"
3582 run_test 63 "Verify each page can at least hold 3 ldisk inodes"
3587 start_ost2 || error "Unable to start second ost"
3588 mount_client $MOUNT || error "Unable to mount client"
3589 stop_ost2 || error "Unable to stop second ost"
3591 $LFS df --lazy || error "lfs df failed"
3592 cleanup || return $?
3593 #writeconf to remove all ost2 traces for subsequent tests
3594 writeconf_or_reformat
3596 run_test 64 "check lfs df --lazy "
3598 test_65() { # LU-2237
3599 # Currently, the test is only valid for ldiskfs backend
3600 [ "$(facet_fstype $SINGLEMDS)" != "ldiskfs" ] &&
3601 skip "non-ldiskfs backend" && return
3603 local devname=$(mdsdevname ${SINGLEMDS//mds/})
3604 local brpt=$(facet_mntpt brpt)
3607 if ! do_facet $SINGLEMDS "test -b $devname"; then
3612 local obj=$(do_facet $SINGLEMDS \
3613 "$DEBUGFS -c -R \\\"stat last_rcvd\\\" $devname" |
3615 if [ -z "$obj" ]; then
3616 # The MDT may be just re-formatted, mount the MDT for the
3617 # first time to guarantee the "last_rcvd" file is there.
3618 start_mds || error "fail to mount the MDS for the first time"
3622 # remove the "last_rcvd" file
3623 do_facet $SINGLEMDS "mkdir -p $brpt"
3624 do_facet $SINGLEMDS \
3625 "mount -t $(facet_fstype $SINGLEMDS) $opts $devname $brpt"
3626 do_facet $SINGLEMDS "rm -f ${brpt}/last_rcvd"
3627 do_facet $SINGLEMDS "umount $brpt"
3629 # restart MDS, the "last_rcvd" file should be recreated.
3630 start_mds || error "fail to restart the MDS"
3632 obj=$(do_facet $SINGLEMDS \
3633 "$DEBUGFS -c -R \\\"stat last_rcvd\\\" $devname" | grep Inode)
3634 [ -n "$obj" ] || error "fail to re-create the last_rcvd"
3636 run_test 65 "re-create the lost last_rcvd file when server mount"
3639 [[ $(lustre_version_code mgs) -ge $(version_code 2.3.59) ]] ||
3640 { skip "Need MGS version at least 2.3.59"; return 0; }
3643 local OST1_NID=$(do_facet ost1 $LCTL list_nids | head -1)
3644 local MDS_NID=$(do_facet $SINGLEMDS $LCTL list_nids | head -1)
3646 echo "replace_nids should fail if MDS, OSTs and clients are UP"
3647 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 $OST1_NID &&
3648 error "replace_nids fail"
3650 umount_client $MOUNT || error "unmounting client failed"
3651 echo "replace_nids should fail if MDS and OSTs are UP"
3652 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 $OST1_NID &&
3653 error "replace_nids fail"
3656 echo "replace_nids should fail if MDS is UP"
3657 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 $OST1_NID &&
3658 error "replace_nids fail"
3660 stop_mds || error "stopping mds failed"
3662 if combined_mgs_mds; then
3663 start_mdt 1 "-o nosvc" ||
3664 error "starting mds with nosvc option failed"
3667 echo "command should accept two parameters"
3668 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 &&
3669 error "command should accept two params"
3671 echo "correct device name should be passed"
3672 do_facet mgs $LCTL replace_nids $FSNAME-WRONG0000 $OST1_NID &&
3673 error "wrong devname"
3675 echo "wrong nids list should not destroy the system"
3676 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 "wrong nids list" &&
3679 echo "replace OST nid"
3680 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 $OST1_NID ||
3681 error "replace nids failed"
3683 echo "command should accept two parameters"
3684 do_facet mgs $LCTL replace_nids $FSNAME-MDT0000 &&
3685 error "command should accept two params"
3687 echo "wrong nids list should not destroy the system"
3688 do_facet mgs $LCTL replace_nids $FSNAME-MDT0000 "wrong nids list" &&
3691 echo "replace MDS nid"
3692 do_facet mgs $LCTL replace_nids $FSNAME-MDT0000 $MDS_NID ||
3693 error "replace nids failed"
3695 if ! combined_mgs_mds ; then
3702 check_mount || error "error after nid replace"
3703 cleanup || error "cleanup failed"
3706 run_test 66 "replace nids"
3708 test_67() { #LU-2950
3709 local legacy="$TMP/legacy_lnet_config"
3710 local new="$TMP/new_routes_test"
3711 local out="$TMP/config_out_file"
3712 local verify="$TMP/conv_verify"
3713 local verify_conf="$TMP/conf_verify"
3715 # Create the legacy file that will be run through the
3716 # lustre_routes_conversion script
3717 cat <<- LEGACY_LNET_CONFIG > $legacy
3718 tcp1 23 192.168.213.1@tcp:1; tcp5 34 193.30.4.3@tcp:4;
3719 tcp2 54 10.1.3.2@tcp;
3720 tcp3 10.3.4.3@tcp:3;
3724 # Create the verification file to verify the output of
3725 # lustre_routes_conversion script against.
3726 cat <<- VERIFY_LNET_CONFIG > $verify
3727 tcp1: { gateway: 192.168.213.1@tcp, hop: 23, priority: 1 }
3728 tcp5: { gateway: 193.30.4.3@tcp, hop: 34, priority: 4 }
3729 tcp2: { gateway: 10.1.3.2@tcp, hop: 54 }
3730 tcp3: { gateway: 10.3.4.3@tcp, priority: 3 }
3731 tcp4: { gateway: 10.3.3.4@tcp }
3734 # Create the verification file to verify the output of
3735 # lustre_routes_config script against
3736 cat <<- VERIFY_LNET_CONFIG > $verify_conf
3737 lctl --net tcp1 add_route 192.168.213.1@tcp 23 1
3738 lctl --net tcp5 add_route 193.30.4.3@tcp 34 4
3739 lctl --net tcp2 add_route 10.1.3.2@tcp 54 4
3740 lctl --net tcp3 add_route 10.3.4.3@tcp 1 3
3741 lctl --net tcp4 add_route 10.3.3.4@tcp 1 3
3744 lustre_routes_conversion $legacy $new > /dev/null
3745 if [ -f $new ]; then
3746 # verify the conversion output
3747 cmp -s $new $verify > /dev/null
3748 if [ $? -eq 1 ]; then
3749 error "routes conversion failed"
3752 lustre_routes_config --dry-run --verbose $new > $out
3753 # check that the script succeeded
3754 cmp -s $out $verify_conf > /dev/null
3755 if [ $? -eq 1 ]; then
3756 error "routes config failed"
3759 error "routes conversion test failed"
3761 # remove generated files
3762 rm -f $new $legacy $verify $verify_conf $out
3764 run_test 67 "test routes conversion and configuration"
3767 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
3770 start_mdt 1 || error "MDT0 start fail"
3772 start_ost || error "OST0 start fail"
3774 start_mdt 2 || error "MDT1 start fail"
3776 mount_client $MOUNT || error "mount client fails"
3778 mkdir -p $DIR/$tdir || error "create dir fail"
3780 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
3781 error "create remote dir fail"
3783 rm -rf $DIR/$tdir || error "delete dir fail"
3784 cleanup || return $?
3786 run_test 70a "start MDT0, then OST, then MDT1"
3789 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
3792 start_ost || error "OST0 start fail"
3794 start_mdt 1 || error "MDT0 start fail"
3795 start_mdt 2 || error "MDT1 start fail"
3797 mount_client $MOUNT || error "mount client fails"
3799 mkdir -p $DIR/$tdir || error "create dir fail"
3801 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
3802 error "create remote dir fail"
3804 rm -rf $DIR/$tdir || error "delete dir fail"
3806 cleanup || return $?
3808 run_test 70b "start OST, MDT1, MDT0"
3811 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
3814 start_mdt 1 || error "MDT0 start fail"
3815 start_mdt 2 || error "MDT1 start fail"
3816 start_ost || error "OST0 start fail"
3818 mount_client $MOUNT || error "mount client fails"
3819 stop_mdt 1 || error "MDT1 start fail"
3821 local mdc_for_mdt1=$($LCTL dl | grep MDT0000-mdc | awk '{print $4}')
3822 echo "deactivate $mdc_for_mdt1"
3823 $LCTL --device $mdc_for_mdt1 deactivate || return 1
3825 mkdir -p $DIR/$tdir && error "mkdir succeed"
3827 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir &&
3828 error "create remote dir succeed"
3830 cleanup || return $?
3832 run_test 70c "stop MDT0, mkdir fail, create remote dir fail"
3835 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
3838 start_mdt 1 || error "MDT0 start fail"
3839 start_mdt 2 || error "MDT1 start fail"
3840 start_ost || error "OST0 start fail"
3842 mount_client $MOUNT || error "mount client fails"
3844 stop_mdt 2 || error "MDT1 start fail"
3846 local mdc_for_mdt2=$($LCTL dl | grep MDT0001-mdc |
3848 echo "deactivate $mdc_for_mdt2"
3849 $LCTL --device $mdc_for_mdt2 deactivate ||
3850 error "set $mdc_for_mdt2 deactivate failed"
3852 mkdir -p $DIR/$tdir || error "mkdir fail"
3853 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir &&
3854 error "create remote dir succeed"
3856 rm -rf $DIR/$tdir || error "delete dir fail"
3858 cleanup || return $?
3860 run_test 70d "stop MDT1, mkdir succeed, create remote dir fail"
3863 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
3864 if combined_mgs_mds; then
3865 skip "needs separate MGS/MDT" && return
3869 start_mdt 1 || error "MDT0 start fail"
3870 start_ost || error "OST0 start fail"
3871 start_mdt 2 || error "MDT1 start fail"
3872 start_ost2 || error "OST1 start fail"
3874 mount_client $MOUNT || error "mount client fails"
3876 mkdir -p $DIR/$tdir || error "mkdir fail"
3877 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
3878 error "create remote dir succeed"
3880 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
3881 rm -rf $DIR/$tdir || error "delete dir fail"
3883 umount_client $MOUNT
3884 stop_mdt 1 || error "MDT0 stop fail"
3885 stop_mdt 2 || error "MDT1 stop fail"
3886 stop_ost || error "OST0 stop fail"
3887 stop_ost2 || error "OST1 stop fail"
3889 run_test 71a "start MDT0 OST0, MDT1, OST1"
3892 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
3893 if combined_mgs_mds; then
3894 skip "needs separate MGS/MDT" && return
3898 start_mdt 2 || error "MDT1 start fail"
3899 start_ost || error "OST0 start fail"
3900 start_mdt 1 || error "MDT0 start fail"
3901 start_ost2 || error "OST1 start fail"
3903 mount_client $MOUNT || error "mount client fails"
3905 mkdir -p $DIR/$tdir || error "mkdir fail"
3906 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
3907 error "create remote dir succeed"
3909 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
3910 rm -rf $DIR/$tdir || error "delete dir fail"
3912 umount_client $MOUNT
3913 stop_mdt 1 || error "MDT0 stop fail"
3914 stop_mdt 2 || error "MDT1 stop fail"
3915 stop_ost || error "OST0 stop fail"
3916 stop_ost2 || error "OST1 stop fail"
3918 run_test 71b "start MDT1, OST0, MDT0, OST1"
3921 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
3922 if combined_mgs_mds; then
3923 skip "needs separate MGS/MDT" && return
3927 start_ost || error "OST0 start fail"
3928 start_ost2 || error "OST1 start fail"
3929 start_mdt 2 || error "MDT1 start fail"
3930 start_mdt 1 || error "MDT0 start fail"
3932 mount_client $MOUNT || error "mount client fails"
3934 mkdir -p $DIR/$tdir || error "mkdir fail"
3935 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
3936 error "create remote dir succeed"
3938 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
3939 rm -rf $DIR/$tdir || error "delete dir fail"
3941 umount_client $MOUNT
3942 stop_mdt 1 || error "MDT0 stop fail"
3943 stop_mdt 2 || error "MDT1 stop fail"
3944 stop_ost || error "OST0 stop fail"
3945 stop_ost2 || error "OST1 stop fail"
3948 run_test 71c "start OST0, OST1, MDT1, MDT0"
3951 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
3952 if combined_mgs_mds; then
3953 skip "needs separate MGS/MDT" && return
3957 start_ost || error "OST0 start fail"
3958 start_mdt 2 || error "MDT0 start fail"
3959 start_mdt 1 || error "MDT0 start fail"
3960 start_ost2 || error "OST1 start fail"
3962 mount_client $MOUNT || error "mount client fails"
3964 mkdir -p $DIR/$tdir || error "mkdir fail"
3965 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
3966 error "create remote dir succeed"
3968 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
3969 rm -rf $DIR/$tdir || error "delete dir fail"
3971 umount_client $MOUNT
3972 stop_mdt 1 || error "MDT0 stop fail"
3973 stop_mdt 2 || error "MDT1 stop fail"
3974 stop_ost || error "OST0 stop fail"
3975 stop_ost2 || error "OST1 stop fail"
3978 run_test 71d "start OST0, MDT1, MDT0, OST1"
3981 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
3982 if combined_mgs_mds; then
3983 skip "needs separate MGS/MDT" && return
3987 start_ost || error "OST0 start fail"
3988 start_mdt 2 || error "MDT1 start fail"
3989 start_ost2 || error "OST1 start fail"
3990 start_mdt 1 || error "MDT0 start fail"
3992 mount_client $MOUNT || error "mount client fails"
3994 mkdir -p $DIR/$tdir || error "mkdir fail"
3995 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
3996 error "create remote dir succeed"
3998 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
3999 rm -rf $DIR/$tdir || error "delete dir fail"
4001 umount_client $MOUNT
4002 stop_mdt 1 || error "MDT0 stop fail"
4003 stop_mdt 2 || error "MDT1 stop fail"
4004 stop_ost || error "OST0 stop fail"
4005 stop_ost2 || error "OST1 stop fail"
4008 run_test 71e "start OST0, MDT1, OST1, MDT0"
4010 test_72() { #LU-2634
4011 local mdsdev=$(mdsdevname 1)
4012 local ostdev=$(ostdevname 1)
4013 local cmd="$E2FSCK -fnvd $mdsdev"
4016 [ "$(facet_fstype $SINGLEMDS)" != "ldiskfs" ] &&
4017 skip "ldiskfs only test" && return
4019 #tune MDT with "-O extents"
4021 for num in $(seq $MDSCOUNT); do
4022 add mds${num} $(mkfs_opts mds$num $(mdsdevname $num)) \
4023 --reformat $(mdsdevname $num) $(mdsvdevname $num) ||
4024 error "add mds $num failed"
4025 $TUNE2FS -O extents $(mdsdevname $num)
4028 add ost1 $(mkfs_opts ost1 $ostdev) --reformat $ostdev ||
4029 error "add $ostdev failed"
4030 start_mgsmds || error "start mds failed"
4031 start_ost || error "start ost failed"
4032 mount_client $MOUNT || error "mount client failed"
4034 #create some short symlinks
4036 createmany -o $DIR/$tdir/$tfile-%d $fn
4037 echo "create $fn short symlinks"
4038 for i in $(seq -w 1 $fn); do
4039 ln -s $DIR/$tdir/$tfile-$i $MOUNT/$tfile-$i
4044 umount_client $MOUNT || error "umount client failed"
4045 stop_mds || error "stop mds failed"
4046 stop_ost || error "stop ost failed"
4049 run_e2fsck $(facet_active_host $SINGLEMDS) $mdsdev "-n"
4051 run_test 72 "test fast symlink with extents flag enabled"
4053 test_73() { #LU-3006
4055 do_facet ost1 "$TUNEFS --failnode=1.2.3.4@tcp $(ostdevname 1)" ||
4056 error "1st tunefs failed"
4057 start_mgsmds || error "start mds failed"
4058 start_ost || error "start ost failed"
4059 mount_client $MOUNT || error "mount client failed"
4060 lctl get_param -n osc.*OST0000-osc-[^M]*.import | grep failover_nids |
4061 grep 1.2.3.4@tcp || error "failover nids haven't changed"
4062 umount_client $MOUNT || error "umount client failed"
4066 run_test 73 "failnode to update from mountdata properly"
4068 test_74() { # LU-1606
4069 for TESTPROG in $LUSTRE_TESTS_API_DIR/*.c; do
4070 gcc -Wall -Werror $LUSTRE_TESTS_API_DIR/simple_test.c \
4072 -L$LUSTRE/utils -llustreapi ||
4073 error "client api broken"
4075 cleanup || return $?
4077 run_test 74 "Lustre client api program can compile and link"
4079 if ! combined_mgs_mds ; then