2 # -*- mode: Bash; tab-width: 4; indent-tabs-mode: t; -*-
3 # vim:autoindent:shiftwidth=4:tabstop=4:
5 # FIXME - there is no reason to use all of these different
6 # return codes, espcially when most of them are mapped to something
7 # else anyway. The combination of test number and return code
8 # figure out what failed.
14 # bug number for skipped test:
16 ALWAYS_EXCEPT="$CONF_SANITY_EXCEPT"
17 # UPDATE THE COMMENT ABOVE WITH BUG NUMBERS WHEN CHANGING ALWAYS_EXCEPT!
19 if [ "$FAILURE_MODE" = "HARD" ]; then
20 CONFIG_EXCEPTIONS="24a " && \
21 echo "Except the tests: $CONFIG_EXCEPTIONS for FAILURE_MODE=$FAILURE_MODE, bug 23573" && \
22 ALWAYS_EXCEPT="$ALWAYS_EXCEPT $CONFIG_EXCEPTIONS"
26 PATH=$PWD/$SRCDIR:$SRCDIR:$SRCDIR/../utils:$PATH
28 PTLDEBUG=${PTLDEBUG:--1}
30 LUSTRE=${LUSTRE:-`dirname $0`/..}
31 RLUSTRE=${RLUSTRE:-$LUSTRE}
33 . $LUSTRE/tests/test-framework.sh
36 # use small MDS + OST size to speed formatting time
37 # do not use too small MDSSIZE/OSTSIZE, which affect the default jouranl size
40 . ${CONFIG:=$LUSTRE/tests/cfg/$NAME.sh}
42 if ! combined_mgs_mds; then
43 # bug number for skipped test: 23954
44 ALWAYS_EXCEPT="$ALWAYS_EXCEPT 24b"
47 # STORED_MDSSIZE is used in test_18
48 if [ -n "$MDSSIZE" ]; then
49 STORED_MDSSIZE=$MDSSIZE
52 # pass "-E lazy_itable_init" to mke2fs to speed up the formatting time
53 for facet in MGS MDS OST; do
54 opts=${facet}_MKFS_OPTS
55 if [[ ${!opts} != *lazy_itable_init* ]]; then
56 eval SAVED_${facet}_MKFS_OPTS=\"${!opts}\"
57 eval ${facet}_MKFS_OPTS=\"${!opts} \
58 --mkfsoptions='\\\"-E lazy_itable_init\\\"'\"
65 require_dsh_mds || exit 0
66 require_dsh_ost || exit 0
68 [ "$SLOW" = "no" ] && EXCEPT_SLOW="30a 31 45"
83 # who knows if/where $TUNEFS is installed? Better reformat if it fails...
84 do_facet ${facet} "$TUNEFS --quiet --writeconf $dev" ||
85 { echo "tunefs failed, reformatting instead" && reformat_and_config && return 1; }
92 # if writeconf fails anywhere, we reformat everything
93 writeconf1 mds `mdsdevname 1` || return 0
94 writeconf1 ost1 `ostdevname 1` || return 0
95 writeconf1 ost2 `ostdevname 2` || return 0
99 # The MGS must be started before the OSTs for a new fs, so start
100 # and stop to generate the startup logs.
103 wait_osc_import_state mds ost FULL
108 reformat_and_config() {
110 if ! combined_mgs_mds ; then
118 start mgs $MGSDEV $MGS_MOUNT_OPTS
122 local facet=$SINGLEMDS
123 # we can not use MDSDEV1 here because SINGLEMDS could be set not to mds1 only
124 local num=$(echo $facet | tr -d "mds")
125 local dev=$(mdsdevname $num)
126 echo "start mds service on `facet_active_host $facet`"
127 start $facet ${dev} $MDS_MOUNT_OPTS $@ || return 94
131 if ! combined_mgs_mds ; then
138 echo "stop mds service on `facet_active_host $SINGLEMDS`"
139 # These tests all use non-failover stop
140 stop $SINGLEMDS -f || return 97
144 echo "stop mgs service on `facet_active_host mgs`"
145 # These tests all use non-failover stop
146 stop mgs -f || return 97
150 echo "start ost1 service on `facet_active_host ost1`"
151 start ost1 `ostdevname 1` $OST_MOUNT_OPTS $@ || return 95
155 echo "stop ost1 service on `facet_active_host ost1`"
156 # These tests all use non-failover stop
157 stop ost1 -f || return 98
161 echo "start ost2 service on `facet_active_host ost2`"
162 start ost2 `ostdevname 2` $OST_MOUNT_OPTS $@ || return 92
166 echo "stop ost2 service on `facet_active_host ost2`"
167 # These tests all use non-failover stop
168 stop ost2 -f || return 93
173 echo "mount $FSNAME on ${MOUNTPATH}....."
174 zconf_mount `hostname` $MOUNTPATH || return 96
178 local mountopt="-o remount,$1"
180 echo "remount '$1' lustre on ${MOUNTPATH}....."
181 zconf_mount `hostname` $MOUNTPATH "$mountopt" || return 96
186 echo "umount lustre on ${MOUNTPATH}....."
187 zconf_umount `hostname` $MOUNTPATH || return 97
190 manual_umount_client(){
193 echo "manual umount lustre on ${MOUNT}...."
194 do_facet client "umount -d ${FORCE} $MOUNT"
200 start_mds || error "MDT start failed"
201 start_ost || error "OST start failed"
202 mount_client $MOUNT || error "client start failed"
203 client_up || error "client_up failed"
207 if ! combined_mgs_mds ; then
216 unload_modules_conf () {
217 if combined_mgs_mds || ! local_mode; then
218 unload_modules || return 1
223 stop_ost || return 202
224 stop_mds || return 201
225 unload_modules_conf || return 203
229 umount_client $MOUNT || return 200
230 cleanup_nocli || return $?
234 do_facet client "cp /etc/passwd $DIR/a" || return 71
235 do_facet client "rm $DIR/a" || return 72
236 # make sure lustre is actually mounted (touch will block,
237 # but grep won't, so do it after)
238 do_facet client "grep $MOUNT' ' /proc/mounts > /dev/null" || return 73
239 echo "setup single mount lustre success"
243 do_facet client "touch $DIR/a" || return 71
244 do_facet client "rm $DIR/a" || return 72
245 do_facet client "touch $DIR2/a" || return 73
246 do_facet client "rm $DIR2/a" || return 74
247 echo "setup double mount lustre success"
252 if [ "$ONLY" == "setup" ]; then
257 if [ "$ONLY" == "cleanup" ]; then
264 #create single point mountpoint
270 check_mount || return 41
273 run_test 0 "single mount setup"
276 start_mds || error "MDT start failed"
278 echo "start ost second time..."
279 start_ost && error "2nd OST start should fail"
280 mount_client $MOUNT || error "client start failed"
281 check_mount || return 42
284 run_test 1 "start up ost twice (should return errors)"
288 echo "start mds second time.."
289 start_mds && error "2nd MDT start should fail"
292 check_mount || return 43
295 run_test 2 "start up mds twice (should return err)"
299 #mount.lustre returns an error if already in mtab
300 mount_client $MOUNT && error "2nd client mount should fail"
301 check_mount || return 44
304 run_test 3 "mount client twice (should return err)"
308 touch $DIR/$tfile || return 85
312 # ok for ost to fail shutdown
313 if [ 202 -ne $eno ]; then
318 run_test 4 "force cleanup ost, then cleanup"
320 test_5a() { # was test_5
322 touch $DIR/$tfile || return 1
323 fuser -m -v $MOUNT && echo "$MOUNT is in use by user space process."
325 stop_mds -f || return 2
327 # cleanup may return an error from the failed
328 # disconnects; for now I'll consider this successful
329 # if all the modules have unloaded.
333 echo "killing umount"
334 kill -TERM $UMOUNT_PID
335 echo "waiting for umount to finish"
337 if grep " $MOUNT " /proc/mounts; then
338 echo "test 5: /proc/mounts after failed umount"
342 echo "killing umount"
343 kill -TERM $UMOUNT_PID
344 echo "waiting for umount to finish"
346 grep " $MOUNT " /proc/mounts && echo "test 5: /proc/mounts after second umount" && return 11
350 # stop_mds is a no-op here, and should not fail
351 cleanup_nocli || return $?
352 # df may have lingering entry
354 # mtab may have lingering entry
358 while [ "$WAIT" -ne "$MAX_WAIT" ]; do
360 grep -q $MOUNT" " /etc/mtab || break
361 echo "Waiting /etc/mtab updated ... "
362 WAIT=$(( WAIT + sleep))
364 [ "$WAIT" -eq "$MAX_WAIT" ] && error "/etc/mtab is not updated in $WAIT secs"
365 echo "/etc/mtab updated in $WAIT secs"
367 run_test 5a "force cleanup mds, then cleanup"
375 grep " $MOUNT " /etc/mtab && \
376 error false "unexpected entry in mtab before mount" && return 10
380 if ! combined_mgs_mds ; then
381 trap cleanup_5b EXIT ERR
386 [ -d $MOUNT ] || mkdir -p $MOUNT
387 mount_client $MOUNT && rc=1
388 grep " $MOUNT " /etc/mtab && \
389 error "$MOUNT entry in mtab after failed mount" && rc=11
391 # stop_mds is a no-op here, and should not fail
392 cleanup_nocli || rc=$?
393 if ! combined_mgs_mds ; then
398 run_test 5b "Try to start a client with no MGS (should return errs)"
401 grep " $MOUNT " /etc/mtab && \
402 error false "unexpected entry in mtab before mount" && return 10
407 [ -d $MOUNT ] || mkdir -p $MOUNT
408 local oldfs="${FSNAME}"
409 FSNAME="wrong.${FSNAME}"
410 mount_client $MOUNT || :
412 grep " $MOUNT " /etc/mtab && \
413 error "$MOUNT entry in mtab after failed mount" && rc=11
415 cleanup_nocli || rc=$?
418 run_test 5c "cleanup after failed mount (bug 2712) (should return errs)"
421 grep " $MOUNT " /etc/mtab && \
422 error false "unexpected entry in mtab before mount" && return 10
428 mount_client $MOUNT || rc=1
430 grep " $MOUNT " /etc/mtab && \
431 error "$MOUNT entry in mtab after unmount" && rc=11
434 run_test 5d "mount with ost down"
437 grep " $MOUNT " /etc/mtab && \
438 error false "unexpected entry in mtab before mount" && return 10
444 #define OBD_FAIL_PTLRPC_DELAY_SEND 0x506
445 do_facet client "lctl set_param fail_loc=0x80000506"
446 mount_client $MOUNT || echo "mount failed (not fatal)"
448 grep " $MOUNT " /etc/mtab && \
449 error "$MOUNT entry in mtab after unmount" && rc=11
452 run_test 5e "delayed connect, don't crash (bug 10268)"
455 if combined_mgs_mds ; then
456 skip "combined mgs and mds"
460 grep " $MOUNT " /etc/mtab && \
461 error false "unexpected entry in mtab before mount" && return 10
465 [ -d $MOUNT ] || mkdir -p $MOUNT
466 mount_client $MOUNT &
468 echo client_mount pid is $pid
472 if ! ps -f -p $pid >/dev/null; then
475 grep " $MOUNT " /etc/mtab && echo "test 5f: mtab after mount"
476 error "mount returns $rc, expected to hang"
485 # mount should succeed after start mds
488 [ $rc -eq 0 ] || error "mount returned $rc"
489 grep " $MOUNT " /etc/mtab && echo "test 5f: mtab after mount"
493 run_test 5f "mds down, cleanup after failed mount (bug 2712)"
498 mount_client ${MOUNT} || return 87
499 touch $DIR/a || return 86
502 run_test 6 "manual umount, then mount again"
507 cleanup_nocli || return $?
509 run_test 7 "manual umount, then cleanup"
514 check_mount2 || return 45
515 umount_client $MOUNT2
518 run_test 8 "double mount setup"
523 do_facet ost1 lctl set_param debug=\'inode trace\' || return 1
524 do_facet ost1 lctl set_param subsystem_debug=\'mds ost\' || return 1
526 CHECK_PTLDEBUG="`do_facet ost1 lctl get_param -n debug`"
527 if [ "$CHECK_PTLDEBUG" ] && { \
528 [ "$CHECK_PTLDEBUG" = "trace inode warning error emerg console" ] ||
529 [ "$CHECK_PTLDEBUG" = "trace inode" ]; }; then
530 echo "lnet.debug success"
532 echo "lnet.debug: want 'trace inode', have '$CHECK_PTLDEBUG'"
535 CHECK_SUBSYS="`do_facet ost1 lctl get_param -n subsystem_debug`"
536 if [ "$CHECK_SUBSYS" ] && [ "$CHECK_SUBSYS" = "mds ost" ]; then
537 echo "lnet.subsystem_debug success"
539 echo "lnet.subsystem_debug: want 'mds ost', have '$CHECK_SUBSYS'"
542 stop_ost || return $?
544 run_test 9 "test ptldebug and subsystem for mkfs"
552 do_facet $facet "test -b $dev" || rc=1
553 if [[ "$size" ]]; then
554 local in=$(do_facet $facet "dd if=$dev of=/dev/null bs=1k count=1 skip=$size 2>&1" |\
555 awk '($3 == "in") { print $1 }')
556 [[ $in = "1+0" ]] || rc=1
562 # Test 16 was to "verify that lustre will correct the mode of OBJECTS".
563 # But with new MDS stack we don't care about the mode of local objects
564 # anymore, so this test is removed. See bug 22944 for more details.
569 check_mount || return 41
572 echo "Remove mds config log"
573 if ! combined_mgs_mds ; then
577 do_facet mgs "$DEBUGFS -w -R 'unlink CONFIGS/$FSNAME-MDT0000' $MGSDEV || return \$?" || return $?
579 if ! combined_mgs_mds ; then
584 start_mds && return 42
587 run_test 17 "Verify failed mds_postsetup won't fail assertion (2936) (should return errs)"
590 [ "$FSTYPE" != "ldiskfs" ] && skip "not needed for FSTYPE=$FSTYPE" && return
592 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
597 # check if current MDSSIZE is large enough
598 [ $MDSSIZE -ge $MIN ] && OK=1 && myMDSSIZE=$MDSSIZE && \
599 log "use MDSSIZE=$MDSSIZE"
601 # check if the global config has a large enough MDSSIZE
602 [ -z "$OK" -a ! -z "$STORED_MDSSIZE" ] && [ $STORED_MDSSIZE -ge $MIN ] && \
603 OK=1 && myMDSSIZE=$STORED_MDSSIZE && \
604 log "use STORED_MDSSIZE=$STORED_MDSSIZE"
606 # check if the block device is large enough
607 [ -z "$OK" ] && $(is_blkdev $SINGLEMDS $MDSDEV $MIN) && OK=1 &&
608 myMDSSIZE=$MIN && log "use device $MDSDEV with MIN=$MIN"
610 # check if a loopback device has enough space for fs metadata (5%)
612 if [ -z "$OK" ]; then
613 local SPACE=$(do_facet $SINGLEMDS "[ -f $MDSDEV -o ! -e $MDSDEV ] && df -P \\\$(dirname $MDSDEV)" |
614 awk '($1 != "Filesystem") {print $4}')
615 ! [ -z "$SPACE" ] && [ $SPACE -gt $((MIN / 20)) ] && \
616 OK=1 && myMDSSIZE=$MIN && \
617 log "use file $MDSDEV with MIN=$MIN"
620 [ -z "$OK" ] && skip_env "$MDSDEV too small for ${MIN}kB MDS" && return
623 echo "mount mds with large journal..."
624 local OLD_MDS_MKFS_OPTS=$MDS_MKFS_OPTS
626 local opts="--mdt --fsname=$FSNAME --device-size=$myMDSSIZE --param sys.timeout=$TIMEOUT $MDSOPT"
628 if combined_mgs_mds ; then
629 MDS_MKFS_OPTS="--mgs $opts"
631 MDS_MKFS_OPTS="--mgsnode=$MGSNID $opts"
635 echo "mount lustre system..."
637 check_mount || return 41
639 echo "check journal size..."
640 local FOUNDSIZE=$(do_facet $SINGLEMDS "$DEBUGFS -c -R 'stat <8>' $MDSDEV" | awk '/Size: / { print $NF; exit;}')
641 if [ $FOUNDSIZE -gt $((32 * 1024 * 1024)) ]; then
642 log "Success: mkfs creates large journals. Size: $((FOUNDSIZE >> 20))M"
644 error "expected journal size > 32M, found $((FOUNDSIZE >> 20))M"
649 MDS_MKFS_OPTS=$OLD_MDS_MKFS_OPTS
652 run_test 18 "check mkfs creates large journals"
655 start_mds || return 1
656 stop_mds -f || return 2
658 run_test 19a "start/stop MDS without OSTs"
661 start_ost || return 1
662 stop_ost -f || return 2
664 run_test 19b "start/stop OSTs without MDS"
667 # first format the ost/mdt
671 check_mount || return 43
673 remount_client ro $MOUNT || return 44
674 touch $DIR/$tfile && echo "$DIR/$tfile created incorrectly" && return 45
675 [ -e $DIR/$tfile ] && echo "$DIR/$tfile exists incorrectly" && return 46
676 remount_client rw $MOUNT || return 47
678 [ ! -f $DIR/$tfile ] && echo "$DIR/$tfile missing" && return 48
679 MCNT=`grep -c $MOUNT /etc/mtab`
680 [ "$MCNT" -ne 1 ] && echo "$MOUNT in /etc/mtab $MCNT times" && return 49
685 run_test 20 "remount ro,rw mounts work and doesn't break /etc/mtab"
690 wait_osc_import_state mds ost FULL
694 run_test 21a "start mds before ost, stop ost first"
699 wait_osc_import_state mds ost FULL
703 run_test 21b "start ost before mds, stop mds first"
709 wait_osc_import_state mds ost2 FULL
713 #writeconf to remove all ost2 traces for subsequent tests
716 run_test 21c "start mds between two osts, stop mds last"
719 if combined_mgs_mds ; then
720 skip "need separate mgs device" && return 0
730 wait_osc_import_state mds ost2 FULL
736 #writeconf to remove all ost2 traces for subsequent tests
740 run_test 21d "start mgs then ost and then mds"
745 echo Client mount with ost in logs, but none running
747 # wait until mds connected to ost and open client connection
748 wait_osc_import_state mds ost FULL
751 # check_mount will block trying to contact ost
752 mcreate $DIR/$tfile || return 40
753 rm -f $DIR/$tfile || return 42
757 echo Client mount with a running ost
760 # if gss enabled, wait full time to let connection from
761 # mds to ost be established, due to the mismatch between
762 # initial connect timeout and gss context negotiation timeout.
763 # This perhaps could be remove after AT landed.
764 echo "sleep $((TIMEOUT + TIMEOUT + TIMEOUT))s"
765 sleep $((TIMEOUT + TIMEOUT + TIMEOUT))
768 wait_osc_import_state mds ost FULL
769 wait_osc_import_state client ost FULL
770 check_mount || return 41
775 run_test 22 "start a client before osts (should return errs)"
777 test_23a() { # was test_23
781 # force down client so that recovering mds waits for reconnect
782 local running=$(grep -c $MOUNT /proc/mounts) || true
783 if [ $running -ne 0 ]; then
784 echo "Stopping client $MOUNT (opts: -f)"
788 # enter recovery on mds
790 # try to start a new client
791 mount_client $MOUNT &
793 MOUNT_PID=$(ps -ef | grep "t lustre" | grep -v grep | awk '{print $2}')
794 MOUNT_LUSTRE_PID=`ps -ef | grep mount.lustre | grep -v grep | awk '{print $2}'`
795 echo mount pid is ${MOUNT_PID}, mount.lustre pid is ${MOUNT_LUSTRE_PID}
797 ps --ppid $MOUNT_LUSTRE_PID
798 echo "waiting for mount to finish"
800 # "ctrl-c" sends SIGINT but it usually (in script) does not work on child process
801 # SIGTERM works but it does not spread to offspring processses
802 kill -s TERM $MOUNT_PID
803 kill -s TERM $MOUNT_LUSTRE_PID
804 # we can not wait $MOUNT_PID because it is not a child of this shell
810 while [ "$WAIT" -lt "$MAX_WAIT" ]; do
812 PID1=$(ps -ef | awk '{print $2}' | grep -w $MOUNT_PID)
813 PID2=$(ps -ef | awk '{print $2}' | grep -w $MOUNT_LUSTRE_PID)
816 [ -z "$PID1" -a -z "$PID2" ] && break
817 echo "waiting for mount to finish ... "
818 WAIT=$(( WAIT + sleep))
820 if [ "$WAIT" -eq "$MAX_WAIT" ]; then
821 error "MOUNT_PID $MOUNT_PID and "\
822 "MOUNT_LUSTRE_PID $MOUNT_LUSTRE_PID still not killed in $WAIT secs"
828 run_test 23a "interrupt client during recovery mount delay"
833 test_23b() { # was test_23
836 # Simulate -EINTR during mount OBD_FAIL_LDLM_CLOSE_THREAD
837 lctl set_param fail_loc=0x80000313
841 run_test 23b "Simulate -EINTR during mount"
843 fs2mds_HOST=$mds_HOST
844 fs2ost_HOST=$ost_HOST
848 echo "umount $MOUNT2 ..."
849 umount $MOUNT2 || true
850 echo "stopping fs2mds ..."
851 stop fs2mds -f || true
852 echo "stopping fs2ost ..."
853 stop fs2ost -f || true
857 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
859 if [ -z "$fs2ost_DEV" -o -z "$fs2mds_DEV" ]; then
860 is_blkdev $SINGLEMDS $MDSDEV && \
861 skip_env "mixed loopback and real device not working" && return
864 [ -n "$ost1_HOST" ] && fs2ost_HOST=$ost1_HOST
866 local fs2mdsdev=${fs2mds_DEV:-${MDSDEV}_2}
867 local fs2ostdev=${fs2ost_DEV:-$(ostdevname 1)_2}
869 # test 8-char fsname as well
870 local FSNAME2=test1234
871 add fs2mds $MDS_MKFS_OPTS --fsname=${FSNAME2} --nomgs --mgsnode=$MGSNID --reformat $fs2mdsdev || exit 10
873 add fs2ost $OST_MKFS_OPTS --fsname=${FSNAME2} --reformat $fs2ostdev || exit 10
876 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS && trap cleanup_24a EXIT INT
877 start fs2ost $fs2ostdev $OST_MOUNT_OPTS
879 mount -t lustre $MGSNID:/${FSNAME2} $MOUNT2 || return 1
881 check_mount || return 2
882 # files written on 1 should not show up on 2
883 cp /etc/passwd $DIR/$tfile
885 [ -e $MOUNT2/$tfile ] && error "File bleed" && return 7
888 cp /etc/passwd $MOUNT2/b || return 3
889 rm $MOUNT2/b || return 4
890 # 2 is actually mounted
891 grep $MOUNT2' ' /proc/mounts > /dev/null || return 5
893 facet_failover fs2mds
894 facet_failover fs2ost
897 # the MDS must remain up until last MDT
899 MDS=$(do_facet $SINGLEMDS "lctl get_param -n devices" | awk '($3 ~ "mdt" && $4 ~ "MDT") { print $4 }' | head -1)
900 [ -z "$MDS" ] && error "No MDT" && return 8
902 cleanup_nocli || return 6
904 run_test 24a "Multiple MDTs on a single node"
907 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
909 if [ -z "$fs2mds_DEV" ]; then
910 local dev=${SINGLEMDS}_dev
912 is_blkdev $SINGLEMDS $MDSDEV && \
913 skip_env "mixed loopback and real device not working" && return
916 local fs2mdsdev=${fs2mds_DEV:-${MDSDEV}_2}
918 add fs2mds $MDS_MKFS_OPTS --fsname=${FSNAME}2 --mgs --reformat $fs2mdsdev || exit 10
920 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS && return 2
923 run_test 24b "Multiple MGSs on a single node (should return err)"
927 check_mount || return 2
928 local MODULES=$($LCTL modules | awk '{ print $2 }')
929 rmmod $MODULES 2>/dev/null || true
932 run_test 25 "Verify modules are referenced"
936 # we need modules before mount for sysctl, so make sure...
937 do_facet $SINGLEMDS "lsmod | grep -q lustre || modprobe lustre"
938 #define OBD_FAIL_MDS_FS_SETUP 0x135
939 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000135"
940 start_mds && echo MDS started && return 1
941 lctl get_param -n devices
942 DEVS=$(lctl get_param -n devices | egrep -v MG | wc -l)
943 [ $DEVS -gt 0 ] && return 2
944 unload_modules_conf || return $?
946 run_test 26 "MDT startup failure cleans LOV (should return errs)"
952 local ORIG=$(do_facet $myfacet "$TEST")
953 if [ $# -gt 3 ]; then
959 echo "Setting $PARAM from $ORIG to $FINAL"
960 do_facet $SINGLEMDS "$LCTL conf_param $PARAM='$FINAL'" || error conf_param failed
962 wait_update $(facet_host $myfacet) "$TEST" "$FINAL" || error check failed!
966 start_ost || return 1
967 start_mds || return 2
968 echo "Requeue thread should have started: "
969 ps -e | grep ll_cfg_requeue
970 set_and_check ost1 "lctl get_param -n obdfilter.$FSNAME-OST0000.client_cache_seconds" "$FSNAME-OST0000.ost.client_cache_seconds" || return 3
973 run_test 27a "Reacquire MGS lock if OST started first"
978 local device=$(do_facet $SINGLEMDS "lctl get_param -n devices" | awk '($3 ~ "mdt" && $4 ~ "MDT") { print $4 }')
980 facet_failover $SINGLEMDS
981 set_and_check $SINGLEMDS "lctl get_param -n mdt.$device.identity_acquire_expire" "$device.mdt.identity_acquire_expire" || return 3
982 set_and_check client "lctl get_param -n mdc.$device-mdc-*.max_rpcs_in_flight" "$device.mdc.max_rpcs_in_flight" || return 4
986 run_test 27b "Reacquire MGS lock after failover"
990 TEST="lctl get_param -n llite.$FSNAME-*.max_read_ahead_whole_mb"
991 PARAM="$FSNAME.llite.max_read_ahead_whole_mb"
994 set_and_check client "$TEST" "$PARAM" $FINAL || return 3
995 FINAL=$(($FINAL + 1))
996 set_and_check client "$TEST" "$PARAM" $FINAL || return 4
997 umount_client $MOUNT || return 200
1000 if [ $RESULT -ne $FINAL ]; then
1001 echo "New config not seen: wanted $FINAL got $RESULT"
1004 echo "New config success: got $RESULT"
1006 set_and_check client "$TEST" "$PARAM" $ORIG || return 5
1009 run_test 28 "permanent parameter setting"
1012 [ "$OSTCOUNT" -lt "2" ] && skip_env "$OSTCOUNT < 2, skipping" && return
1013 setup > /dev/null 2>&1
1017 local PARAM="$FSNAME-OST0001.osc.active"
1018 local PROC_ACT="osc.$FSNAME-OST0001-osc-[^M]*.active"
1019 local PROC_UUID="osc.$FSNAME-OST0001-osc-[^M]*.ost_server_uuid"
1021 ACTV=$(lctl get_param -n $PROC_ACT)
1023 set_and_check client "lctl get_param -n $PROC_ACT" "$PARAM" $DEAC || return 2
1024 # also check ost_server_uuid status
1025 RESULT=$(lctl get_param -n $PROC_UUID | grep DEACTIV)
1026 if [ -z "$RESULT" ]; then
1027 echo "Live client not deactivated: $(lctl get_param -n $PROC_UUID)"
1030 echo "Live client success: got $RESULT"
1034 local mdtosc=$(get_mdtosc_proc_path $SINGLEMDS $FSNAME-OST0001)
1035 mdtosc=${mdtosc/-MDT*/-MDT\*}
1036 local MPROC="osc.$mdtosc.active"
1041 RESULT=`do_facet $SINGLEMDS " lctl get_param -n $MPROC"`
1042 [ ${PIPESTATUS[0]} = 0 ] || error "Can't read $MPROC"
1043 if [ $RESULT -eq $DEAC ]; then
1044 echo "MDT deactivated also after $WAIT sec (got $RESULT)"
1048 if [ $WAIT -eq $MAX ]; then
1049 echo "MDT not deactivated: wanted $DEAC got $RESULT"
1052 echo "Waiting $(($MAX - $WAIT)) secs for MDT deactivated"
1055 # quotacheck should not fail immediately after deactivate
1056 [ -n "$ENABLE_QUOTA" ] && { $LFS quotacheck -ug $MOUNT || error "quotacheck has failed" ; }
1058 # test new client starts deactivated
1059 umount_client $MOUNT || return 200
1061 RESULT=$(lctl get_param -n $PROC_UUID | grep DEACTIV | grep NEW)
1062 if [ -z "$RESULT" ]; then
1063 echo "New client not deactivated from start: $(lctl get_param -n $PROC_UUID)"
1066 echo "New client success: got $RESULT"
1069 # quotacheck should not fail after umount/mount operation
1070 [ -n "$ENABLE_QUOTA" ] && { $LFS quotacheck -ug $MOUNT || error "quotacheck has failed" ; }
1072 # make sure it reactivates
1073 set_and_check client "lctl get_param -n $PROC_ACT" "$PARAM" $ACTV || return 6
1075 umount_client $MOUNT
1078 #writeconf to remove all ost2 traces for subsequent tests
1081 run_test 29 "permanently remove an OST"
1086 echo Big config llog
1087 TEST="lctl get_param -n llite.$FSNAME-*.max_read_ahead_whole_mb"
1089 LIST=(1 2 3 4 5 4 3 2 1 2 3 4 5 4 3 2 1 2 3 4 5)
1090 for i in ${LIST[@]}; do
1091 set_and_check client "$TEST" "$FSNAME.llite.max_read_ahead_whole_mb" $i || return 3
1093 # make sure client restart still works
1094 umount_client $MOUNT
1095 mount_client $MOUNT || return 4
1096 [ "$($TEST)" -ne "$i" ] && error "Param didn't stick across restart $($TEST) != $i"
1099 echo Erase parameter setting
1100 do_facet mgs "$LCTL conf_param -d $FSNAME.llite.max_read_ahead_whole_mb" || return 6
1101 umount_client $MOUNT
1102 mount_client $MOUNT || return 6
1104 echo "deleted (default) value=$FINAL, orig=$ORIG"
1105 # assumes this parameter started at the default value
1106 [ "$FINAL" -eq "$ORIG" ] || fail "Deleted value=$FINAL, orig=$ORIG"
1110 run_test 30a "Big config llog and conf_param deletion"
1115 # Make a fake nid. Use the OST nid, and add 20 to the least significant
1116 # numerical part of it. Hopefully that's not already a failover address for
1118 OSTNID=$(do_facet ost1 "$LCTL get_param nis" | tail -1 | awk '{print $1}')
1119 ORIGVAL=$(echo $OSTNID | egrep -oi "[0-9]*@")
1120 NEWVAL=$((($(echo $ORIGVAL | egrep -oi "[0-9]*") + 20) % 256))
1121 NEW=$(echo $OSTNID | sed "s/$ORIGVAL/$NEWVAL@/")
1122 echo "Using fake nid $NEW"
1124 TEST="$LCTL get_param -n osc.$FSNAME-OST0000-osc-[^M]*.import | grep failover_nids | sed -n 's/.*\($NEW\).*/\1/p'"
1125 set_and_check client "$TEST" "$FSNAME-OST0000.failover.node" $NEW || error "didn't add failover nid $NEW"
1126 NIDS=$($LCTL get_param -n osc.$FSNAME-OST0000-osc-[^M]*.import | grep failover_nids)
1128 NIDCOUNT=$(($(echo "$NIDS" | wc -w) - 1))
1129 echo "should have 2 failover nids: $NIDCOUNT"
1130 [ $NIDCOUNT -eq 2 ] || error "Failover nid not added"
1131 do_facet mgs "$LCTL conf_param -d $FSNAME-OST0000.failover.node" || error "conf_param delete failed"
1132 umount_client $MOUNT
1133 mount_client $MOUNT || return 3
1135 NIDS=$($LCTL get_param -n osc.$FSNAME-OST0000-osc-[^M]*.import | grep failover_nids)
1137 NIDCOUNT=$(($(echo "$NIDS" | wc -w) - 1))
1138 echo "only 1 final nid should remain: $NIDCOUNT"
1139 [ $NIDCOUNT -eq 1 ] || error "Failover nids not removed"
1143 run_test 30b "Remove failover nids"
1145 test_31() { # bug 10734
1146 # ipaddr must not exist
1147 mount -t lustre 4.3.2.1@tcp:/lustre $MOUNT || true
1150 run_test 31 "Connect to non-existent node (shouldn't crash)"
1152 # Use these start32/stop32 fn instead of t-f start/stop fn,
1153 # for local devices, to skip global facet vars init
1157 echo "Stopping local ${MOUNT%/*}/${facet} (opts:$@)"
1158 umount -d $@ ${MOUNT%/*}/${facet}
1167 mkdir -p ${MOUNT%/*}/${facet}
1169 echo "Starting local ${facet}: $@ $device ${MOUNT%/*}/${facet}"
1170 mount -t lustre $@ ${device} ${MOUNT%/*}/${facet}
1172 if [ $RC -ne 0 ]; then
1173 echo "mount -t lustre $@ ${device} ${MOUNT%/*}/${facet}"
1174 echo "Start of ${device} of local ${facet} failed ${RC}"
1180 cleanup_nocli32 () {
1188 echo "Cleanup test_32 umount $MOUNT ..."
1189 umount -f $MOUNT || true
1190 echo "Cleanup local mds ost1 ..."
1192 combined_mgs_mds || start_mgs
1197 client_only && skip "client only testing" && return 0
1198 [ "$NETTYPE" = "tcp" ] || { skip "NETTYPE != tcp" && return 0; }
1199 [ -z "$TUNEFS" ] && skip_env "No tunefs" && return 0
1201 local DISK1_8=$LUSTRE/tests/disk1_8.tar.bz2
1202 [ ! -r $DISK1_8 ] && skip_env "Cannot find $DISK1_8" && return 0
1203 local tmpdir=$TMP/conf32a
1205 tar xjvf $DISK1_8 -C $tmpdir || \
1206 { skip_env "Cannot untar $DISK1_8" && return 0; }
1209 $LCTL set_param debug=$PTLDEBUG
1211 $TUNEFS $tmpdir/mds || error "tunefs failed"
1213 combined_mgs_mds || stop mgs
1215 # nids are wrong, so client wont work, but server should start
1216 start32 mds1 $tmpdir/mds "-o loop,exclude=lustre-OST0000" && \
1217 trap cleanup_32 EXIT INT || return 3
1219 local UUID=$($LCTL get_param -n mdt.lustre-MDT0000.uuid)
1221 [ "$UUID" == "lustre-MDT0000_UUID" ] || error "UUID is wrong: $UUID"
1223 $TUNEFS --mgsnode=$HOSTNAME $tmpdir/ost1 || error "tunefs failed"
1224 start32 ost1 $tmpdir/ost1 "-o loop" || return 5
1225 UUID=$($LCTL get_param -n obdfilter.lustre-OST0000.uuid)
1227 [ "$UUID" == "lustre-OST0000_UUID" ] || error "UUID is wrong: $UUID"
1229 local NID=$($LCTL list_nids | head -1)
1231 echo "OSC changes should succeed:"
1232 $LCTL conf_param lustre-OST0000.osc.max_dirty_mb=15 || return 7
1233 $LCTL conf_param lustre-OST0000.failover.node=$NID || return 8
1236 echo "MDC changes should succeed:"
1237 $LCTL conf_param lustre-MDT0000.mdc.max_rpcs_in_flight=9 || return 9
1238 $LCTL conf_param lustre-MDT0000.failover.node=$NID || return 10
1241 echo "LOV changes should succeed:"
1242 $LCTL pool_new lustre.interop || return 11
1243 $LCTL conf_param lustre-MDT0000.lov.stripesize=4M || return 12
1248 # mount a second time to make sure we didnt leave upgrade flag on
1250 $TUNEFS --dryrun $tmpdir/mds || error "tunefs failed"
1252 combined_mgs_mds || stop mgs
1254 start32 mds1 $tmpdir/mds "-o loop,exclude=lustre-OST0000" && \
1255 trap cleanup_32 EXIT INT || return 12
1259 rm -rf $tmpdir || true # true is only for TMP on NFS
1261 run_test 32a "Upgrade from 1.8 (not live)"
1264 client_only && skip "client only testing" && return 0
1265 [ "$NETTYPE" = "tcp" ] || { skip "NETTYPE != tcp" && return 0; }
1266 [ -z "$TUNEFS" ] && skip_env "No tunefs" && return
1268 local DISK1_8=$LUSTRE/tests/disk1_8.tar.bz2
1269 [ ! -r $DISK1_8 ] && skip_env "Cannot find $DISK1_8" && return 0
1270 local tmpdir=$TMP/conf32b
1272 tar xjvf $DISK1_8 -C $tmpdir || \
1273 { skip_env "Cannot untar $DISK1_8" && return ; }
1276 $LCTL set_param debug="config"
1277 local NEWNAME=lustre
1279 # writeconf will cause servers to register with their current nids
1280 $TUNEFS --writeconf --fsname=$NEWNAME $tmpdir/mds || error "tunefs failed"
1281 combined_mgs_mds || stop mgs
1283 start32 mds1 $tmpdir/mds "-o loop" && \
1284 trap cleanup_32 EXIT INT || return 3
1286 local UUID=$($LCTL get_param -n mdt.${NEWNAME}-MDT0000.uuid)
1288 [ "$UUID" == "${NEWNAME}-MDT0000_UUID" ] || error "UUID is wrong: $UUID"
1290 $TUNEFS --mgsnode=$HOSTNAME --writeconf --fsname=$NEWNAME $tmpdir/ost1 ||\
1291 error "tunefs failed"
1292 start32 ost1 $tmpdir/ost1 "-o loop" || return 5
1293 UUID=$($LCTL get_param -n obdfilter.${NEWNAME}-OST0000.uuid)
1295 [ "$UUID" == "${NEWNAME}-OST0000_UUID" ] || error "UUID is wrong: $UUID"
1297 local NID=$($LCTL list_nids | head -1)
1299 echo "OSC changes should succeed:"
1300 $LCTL conf_param ${NEWNAME}-OST0000.osc.max_dirty_mb=15 || return 7
1301 $LCTL conf_param ${NEWNAME}-OST0000.failover.node=$NID || return 8
1304 echo "MDC changes should succeed:"
1305 $LCTL conf_param ${NEWNAME}-MDT0000.mdc.max_rpcs_in_flight=9 || return 9
1306 $LCTL conf_param ${NEWNAME}-MDT0000.failover.node=$NID || return 10
1309 echo "LOV changes should succeed:"
1310 $LCTL pool_new ${NEWNAME}.interop || return 11
1311 $LCTL conf_param ${NEWNAME}-MDT0000.lov.stripesize=4M || return 12
1314 # MDT and OST should have registered with new nids, so we should have
1315 # a fully-functioning client
1316 echo "Check client and old fs contents"
1318 local device=`h2$NETTYPE $HOSTNAME`:/$NEWNAME
1319 echo "Starting local client: $HOSTNAME: $device $MOUNT"
1320 mount -t lustre $device $MOUNT || return 1
1322 local old=$($LCTL get_param -n mdc.*.max_rpcs_in_flight)
1323 local new=$((old + 5))
1324 $LCTL conf_param ${NEWNAME}-MDT0000.mdc.max_rpcs_in_flight=$new
1325 wait_update $HOSTNAME "$LCTL get_param -n mdc.*.max_rpcs_in_flight" $new || return 11
1327 [ "$(cksum $MOUNT/passwd | cut -d' ' -f 1,2)" == "94306271 1478" ] || return 12
1332 rm -rf $tmpdir || true # true is only for TMP on NFS
1334 run_test 32b "Upgrade from 1.8 with writeconf"
1336 test_33a() { # bug 12333, was test_33
1338 local FSNAME2=test-123
1339 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
1341 [ -n "$ost1_HOST" ] && fs2ost_HOST=$ost1_HOST
1343 if [ -z "$fs2ost_DEV" -o -z "$fs2mds_DEV" ]; then
1344 local dev=${SINGLEMDS}_dev
1345 local MDSDEV=${!dev}
1346 is_blkdev $SINGLEMDS $MDSDEV && \
1347 skip_env "mixed loopback and real device not working" && return
1350 combined_mgs_mds || mkfs_opts="$mkfs_opts --nomgs"
1352 local fs2mdsdev=${fs2mds_DEV:-${MDSDEV}_2}
1353 local fs2ostdev=${fs2ost_DEV:-$(ostdevname 1)_2}
1354 add fs2mds $MDS_MKFS_OPTS --mkfsoptions='\"-J size=8\"' --fsname=${FSNAME2} --reformat $fs2mdsdev || exit 10
1355 add fs2ost $OST_MKFS_OPTS --fsname=${FSNAME2} --index=8191 --mgsnode=$MGSNID --reformat $fs2ostdev || exit 10
1357 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS && trap cleanup_24a EXIT INT
1358 start fs2ost $fs2ostdev $OST_MOUNT_OPTS
1359 do_facet $SINGLEMDS "$LCTL conf_param $FSNAME2.sys.timeout=200" || rc=1
1361 mount -t lustre $MGSNID:/${FSNAME2} $MOUNT2 || rc=2
1364 cp /etc/hosts $MOUNT2/ || rc=3
1365 $LFS getstripe $MOUNT2/hosts
1370 rm -rf $MOUNT2 $fs2mdsdev $fs2ostdev
1371 cleanup_nocli || rc=6
1374 run_test 33a "Mount ost with a large index number"
1376 test_33b() { # was test_34
1379 do_facet client dd if=/dev/zero of=$MOUNT/24 bs=1024k count=1
1380 # Drop lock cancelation reply during umount
1381 #define OBD_FAIL_LDLM_CANCEL 0x304
1382 do_facet client lctl set_param fail_loc=0x80000304
1383 #lctl set_param debug=-1
1384 umount_client $MOUNT
1387 run_test 33b "Drop cancel during umount"
1391 do_facet client "sh runmultiop_bg_pause $DIR/file O_c"
1392 manual_umount_client
1394 do_facet client killall -USR1 multiop
1395 if [ $rc -eq 0 ]; then
1396 error "umount not fail!"
1401 run_test 34a "umount with opened file should be fail"
1406 touch $DIR/$tfile || return 1
1407 stop_mds --force || return 2
1409 manual_umount_client --force
1411 if [ $rc -ne 0 ]; then
1412 error "mtab after failed umount - rc $rc"
1418 run_test 34b "force umount with failed mds should be normal"
1422 touch $DIR/$tfile || return 1
1423 stop_ost --force || return 2
1425 manual_umount_client --force
1427 if [ $rc -ne 0 ]; then
1428 error "mtab after failed umount - rc $rc"
1434 run_test 34c "force umount with failed ost should be normal"
1436 test_35a() { # bug 12459
1439 DBG_SAVE="`lctl get_param -n debug`"
1440 lctl set_param debug="ha"
1442 log "Set up a fake failnode for the MDS"
1444 local device=$(do_facet $SINGLEMDS "lctl get_param -n devices" | awk '($3 ~ "mdt" && $4 ~ "MDT") { print $4 }' | head -1)
1445 do_facet $SINGLEMDS $LCTL conf_param ${device}.failover.node=$FAKENID || return 4
1447 log "Wait for RECONNECT_INTERVAL seconds (10s)"
1450 MSG="conf-sanity.sh test_35a `date +%F%kh%Mm%Ss`"
1453 log "Stopping the MDT:"
1454 stop_mds || return 5
1456 df $MOUNT > /dev/null 2>&1 &
1458 log "Restarting the MDT:"
1459 start_mds || return 6
1460 log "Wait for df ($DFPID) ... "
1463 lctl set_param debug="$DBG_SAVE"
1465 # retrieve from the log the first server that the client tried to
1466 # contact after the connection loss
1467 $LCTL dk $TMP/lustre-log-$TESTNAME.log
1468 NEXTCONN=`awk "/${MSG}/ {start = 1;}
1469 /import_select_connection.*$device-mdc.* using connection/ {
1471 if (\\\$NF ~ /$FAKENID/)
1477 }" $TMP/lustre-log-$TESTNAME.log`
1478 [ "$NEXTCONN" != "0" ] && log "The client didn't try to reconnect to the last active server (tried ${NEXTCONN} instead)" && return 7
1480 # remove nid settings
1483 run_test 35a "Reconnect to the last active server first"
1485 test_35b() { # bug 18674
1486 remote_mds || { skip "local MDS" && return 0; }
1490 $LCTL set_param debug="ha"
1492 MSG="conf-sanity.sh test_35b `date +%F%kh%Mm%Ss`"
1495 log "Set up a fake failnode for the MDS"
1497 local device=$(do_facet $SINGLEMDS "$LCTL get_param -n devices" | \
1498 awk '($3 ~ "mdt" && $4 ~ "MDT") { print $4 }' | head -1)
1499 do_facet $SINGLEMDS "$LCTL conf_param ${device}.failover.node=$FAKENID" || \
1502 local at_max_saved=0
1503 # adaptive timeouts may prevent seeing the issue
1504 if at_is_enabled; then
1505 at_max_saved=$(at_max_get mds)
1506 at_max_set 0 mds client
1509 mkdir -p $MOUNT/$tdir
1511 log "Injecting EBUSY on MDS"
1512 # Setting OBD_FAIL_MDS_RESEND=0x136
1513 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0x80000136" || return 2
1515 $LCTL set_param mdc.${FSNAME}*.stats=clear
1517 log "Creating a test file and stat it"
1518 touch $MOUNT/$tdir/$tfile
1519 stat $MOUNT/$tdir/$tfile
1521 log "Stop injecting EBUSY on MDS"
1522 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0" || return 3
1523 rm -f $MOUNT/$tdir/$tfile
1526 # restore adaptive timeout
1527 [ $at_max_saved -ne 0 ] && at_max_set $at_max_saved mds client
1529 $LCTL dk $TMP/lustre-log-$TESTNAME.log
1531 CONNCNT=`$LCTL get_param mdc.${FSNAME}*.stats | awk '/mds_connect/{print $2}'`
1533 # retrieve from the log if the client has ever tried to
1534 # contact the fake server after the loss of connection
1535 FAILCONN=`awk "BEGIN {ret = 0;}
1536 /import_select_connection.*${FSNAME}-MDT0000-mdc.* using connection/ {
1538 if (\\\$NF ~ /$FAKENID/) {
1543 END {print ret}" $TMP/lustre-log-$TESTNAME.log`
1545 [ "$FAILCONN" == "0" ] && \
1546 log "ERROR: The client reconnection has not been triggered" && \
1548 [ "$FAILCONN" == "2" ] && \
1549 log "ERROR: The client tried to reconnect to the failover server while the primary was busy" && \
1553 # When OBD_FAIL_MDS_RESEND is hit, we sleep for 2 * obd_timeout
1554 # Reconnects are supposed to be rate limited to one every 5s
1555 [ $CONNCNT -gt $((2 * $TIMEOUT / 5 + 1)) ] && \
1556 log "ERROR: Too many reconnects $CONNCNT" && \
1560 # remove nid settings
1563 run_test 35b "Continue reconnection retries, if the active server is busy"
1566 [ $OSTCOUNT -lt 2 ] && skip_env "skipping test for single OST" && return
1568 [ "$ost_HOST" = "`hostname`" -o "$ost1_HOST" = "`hostname`" ] || \
1569 { skip "remote OST" && return 0; }
1572 local FSNAME2=test1234
1573 local fs3ost_HOST=$ost_HOST
1574 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
1576 [ -n "$ost1_HOST" ] && fs2ost_HOST=$ost1_HOST && fs3ost_HOST=$ost1_HOST
1578 if [ -z "$fs2ost_DEV" -o -z "$fs2mds_DEV" -o -z "$fs3ost_DEV" ]; then
1579 is_blkdev $SINGLEMDS $MDSDEV && \
1580 skip_env "mixed loopback and real device not working" && return
1583 local fs2mdsdev=${fs2mds_DEV:-${MDSDEV}_2}
1584 local fs2ostdev=${fs2ost_DEV:-$(ostdevname 1)_2}
1585 local fs3ostdev=${fs3ost_DEV:-$(ostdevname 2)_2}
1586 add fs2mds $MDS_MKFS_OPTS --fsname=${FSNAME2} --reformat $fs2mdsdev || exit 10
1587 # XXX after we support non 4K disk blocksize, change following --mkfsoptions with
1589 add fs2ost $OST_MKFS_OPTS --mkfsoptions='-b4096' --fsname=${FSNAME2} --mgsnode=$MGSNID --reformat $fs2ostdev || exit 10
1590 add fs3ost $OST_MKFS_OPTS --mkfsoptions='-b4096' --fsname=${FSNAME2} --mgsnode=$MGSNID --reformat $fs3ostdev || exit 10
1592 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS
1593 start fs2ost $fs2ostdev $OST_MOUNT_OPTS
1594 start fs3ost $fs3ostdev $OST_MOUNT_OPTS
1596 mount -t lustre $MGSNID:/${FSNAME2} $MOUNT2 || return 1
1598 sleep 5 # until 11778 fixed
1600 dd if=/dev/zero of=$MOUNT2/$tfile bs=1M count=7 || return 2
1602 BKTOTAL=`lctl get_param -n obdfilter.*.kbytestotal | awk 'BEGIN{total=0}; {total+=$1}; END{print total}'`
1603 BKFREE=`lctl get_param -n obdfilter.*.kbytesfree | awk 'BEGIN{free=0}; {free+=$1}; END{print free}'`
1604 BKAVAIL=`lctl get_param -n obdfilter.*.kbytesavail | awk 'BEGIN{avail=0}; {avail+=$1}; END{print avail}'`
1605 STRING=`df -P $MOUNT2 | tail -n 1 | awk '{print $2","$3","$4}'`
1606 DFTOTAL=`echo $STRING | cut -d, -f1`
1607 DFUSED=`echo $STRING | cut -d, -f2`
1608 DFAVAIL=`echo $STRING | cut -d, -f3`
1609 DFFREE=$(($DFTOTAL - $DFUSED))
1611 ALLOWANCE=$((64 * $OSTCOUNT))
1613 if [ $DFTOTAL -lt $(($BKTOTAL - $ALLOWANCE)) ] ||
1614 [ $DFTOTAL -gt $(($BKTOTAL + $ALLOWANCE)) ] ; then
1615 echo "**** FAIL: df total($DFTOTAL) mismatch OST total($BKTOTAL)"
1618 if [ $DFFREE -lt $(($BKFREE - $ALLOWANCE)) ] ||
1619 [ $DFFREE -gt $(($BKFREE + $ALLOWANCE)) ] ; then
1620 echo "**** FAIL: df free($DFFREE) mismatch OST free($BKFREE)"
1623 if [ $DFAVAIL -lt $(($BKAVAIL - $ALLOWANCE)) ] ||
1624 [ $DFAVAIL -gt $(($BKAVAIL + $ALLOWANCE)) ] ; then
1625 echo "**** FAIL: df avail($DFAVAIL) mismatch OST avail($BKAVAIL)"
1630 stop fs3ost -f || return 200
1631 stop fs2ost -f || return 201
1632 stop fs2mds -f || return 202
1633 rm -rf $MOUNT2 $fs2mdsdev $fs2ostdev $fs3ostdev
1634 unload_modules_conf || return 203
1637 run_test 36 "df report consistency on OSTs with different block size"
1640 local mntpt=$(facet_mntpt $SINGLEMDS)
1641 local mdsdev=$(mdsdevname ${SINGLEMDS//mds/})
1642 local mdsdev_sym="$TMP/sym_mdt.img"
1644 echo "MDS : $mdsdev"
1645 echo "SYMLINK : $mdsdev_sym"
1646 do_facet $SINGLEMDS rm -f $mdsdev_sym
1648 do_facet $SINGLEMDS ln -s $mdsdev $mdsdev_sym
1650 echo "mount symlink device - $mdsdev_sym"
1653 mount_op=$(do_facet $SINGLEMDS mount -v -t lustre $MDS_MOUNT_OPTS $mdsdev_sym $mntpt 2>&1 )
1656 echo mount_op=$mount_op
1658 do_facet $SINGLEMDS "umount -d $mntpt && rm -f $mdsdev_sym"
1660 if $(echo $mount_op | grep -q "unable to set tunable"); then
1661 error "set tunables failed for symlink device"
1664 [ $rc -eq 0 ] || error "mount symlink $mdsdev_sym failed! rc=$rc"
1668 run_test 37 "verify set tunables works for symlink device"
1670 test_38() { # bug 14222
1675 FILES=`find $SRC -type f -mtime +1 | head -n $COUNT`
1676 log "copying $(echo $FILES | wc -w) files to $DIR/$tdir"
1678 tar cf - $FILES | tar xf - -C $DIR/$tdir || \
1679 error "copying $SRC to $DIR/$tdir"
1681 umount_client $MOUNT
1683 log "rename lov_objid file on MDS"
1684 rm -f $TMP/lov_objid.orig
1686 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
1687 do_facet $SINGLEMDS "$DEBUGFS -c -R \\\"dump lov_objid $TMP/lov_objid.orig\\\" $MDSDEV"
1688 do_facet $SINGLEMDS "$DEBUGFS -w -R \\\"rm lov_objid\\\" $MDSDEV"
1690 do_facet $SINGLEMDS "od -Ax -td8 $TMP/lov_objid.orig"
1691 # check create in mds_lov_connect
1695 [ $V ] && log "verifying $DIR/$tdir/$f"
1696 diff -q $f $DIR/$tdir/$f || ERROR=y
1698 do_facet $SINGLEMDS "$DEBUGFS -c -R \\\"dump lov_objid $TMP/lov_objid.new\\\" $MDSDEV"
1699 do_facet $SINGLEMDS "od -Ax -td8 $TMP/lov_objid.new"
1700 [ "$ERROR" = "y" ] && error "old and new files are different after connect" || true
1702 # check it's updates in sync
1703 umount_client $MOUNT
1706 do_facet $SINGLEMDS dd if=/dev/zero of=$TMP/lov_objid.clear bs=4096 count=1
1707 do_facet $SINGLEMDS "$DEBUGFS -w -R \\\"rm lov_objid\\\" $MDSDEV"
1708 do_facet $SINGLEMDS "$DEBUGFS -w -R \\\"write $TMP/lov_objid.clear lov_objid\\\" $MDSDEV "
1713 [ $V ] && log "verifying $DIR/$tdir/$f"
1714 diff -q $f $DIR/$tdir/$f || ERROR=y
1716 do_facet $SINGLEMDS "$DEBUGFS -c -R \\\"dump lov_objid $TMP/lov_objid.new1\\\" $MDSDEV"
1717 do_facet $SINGLEMDS "od -Ax -td8 $TMP/lov_objid.new1"
1718 umount_client $MOUNT
1720 [ "$ERROR" = "y" ] && error "old and new files are different after sync" || true
1722 log "files compared the same"
1725 run_test 38 "MDS recreates missing lov_objid file from OST data"
1731 perl $SRCDIR/leak_finder.pl $TMP/debug 2>&1 | egrep '*** Leak:' &&
1732 error "memory leak detected" || true
1734 run_test 39 "leak_finder recognizes both LUSTRE and LNET malloc messages"
1736 test_40() { # bug 15759
1738 #define OBD_FAIL_TGT_TOOMANY_THREADS 0x706
1739 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0x80000706"
1743 run_test 40 "race during service thread startup"
1745 test_41a() { #bug 14134
1746 echo $MDS_MOUNT_OPTS | grep "loop" && skip " loop devices does not work with nosvc option" && return
1749 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
1751 start $SINGLEMDS $MDSDEV $MDS_MOUNT_OPTS -o nosvc -n
1752 start ost1 `ostdevname 1` $OST_MOUNT_OPTS
1753 start $SINGLEMDS $MDSDEV $MDS_MOUNT_OPTS -o nomgs,force
1755 mount_client $MOUNT || return 1
1758 echo "blah blah" > $MOUNT/$tfile
1761 umount_client $MOUNT
1762 stop ost1 -f || return 201
1763 stop_mds -f || return 202
1764 stop_mds -f || return 203
1765 unload_modules_conf || return 204
1768 run_test 41a "mount mds with --nosvc and --nomgs"
1771 echo $MDS_MOUNT_OPTS | grep "loop" && skip " loop devices does not work with nosvc option" && return
1773 ! combined_mgs_mds && skip "needs combined mgs device" && return 0
1777 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
1779 start $SINGLEMDS $MDSDEV $MDS_MOUNT_OPTS -o nosvc -n
1781 start $SINGLEMDS $MDSDEV $MDS_MOUNT_OPTS -o nomgs,force
1783 mount_client $MOUNT || return 1
1786 echo "blah blah" > $MOUNT/$tfile
1787 cat $MOUNT/$tfile || return 200
1789 umount_client $MOUNT
1790 stop_ost || return 201
1791 stop_mds -f || return 202
1792 stop_mds -f || return 203
1795 run_test 41b "mount mds with --nosvc and --nomgs on first mount"
1797 test_42() { #bug 14693
1799 check_mount || return 2
1800 do_facet mgs $LCTL conf_param lustre.llite.some_wrong_param=10
1801 umount_client $MOUNT
1802 mount_client $MOUNT || return 1
1806 run_test 42 "invalid config param should not prevent client from mounting"
1809 [ $UID -ne 0 -o $RUNAS_ID -eq 0 ] && skip_env "run as root"
1811 chmod ugo+x $DIR || error "chmod 0 failed"
1813 "lctl get_param -n mdt.$FSNAME-MDT0000.root_squash" \
1814 "$FSNAME.mdt.root_squash" \
1817 "lctl get_param -n mdt.$FSNAME-MDT0000.nosquash_nids" \
1818 "$FSNAME.mdt.nosquash_nids" \
1822 # create set of test files
1824 echo "111" > $DIR/$tfile-userfile || error "write 1 failed"
1825 chmod go-rw $DIR/$tfile-userfile || error "chmod 1 failed"
1826 chown $RUNAS_ID.$RUNAS_ID $DIR/$tfile-userfile || error "chown failed"
1828 echo "222" > $DIR/$tfile-rootfile || error "write 2 failed"
1829 chmod go-rw $DIR/$tfile-rootfile || error "chmod 2 faield"
1831 mkdir $DIR/$tdir-rootdir -p || error "mkdir failed"
1832 chmod go-rwx $DIR/$tdir-rootdir || error "chmod 3 failed"
1833 touch $DIR/$tdir-rootdir/tfile-1 || error "touch failed"
1836 # check root_squash:
1837 # set root squash UID:GID to RUNAS_ID
1838 # root should be able to access only files owned by RUNAS_ID
1841 "lctl get_param -n mdt.$FSNAME-MDT0000.root_squash" \
1842 "$FSNAME.mdt.root_squash" \
1843 "$RUNAS_ID:$RUNAS_ID"
1845 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-userfile)
1846 dd if=$DIR/$tfile-userfile 1>/dev/null 2>/dev/null || \
1847 error "$ST: root read permission is denied"
1848 echo "$ST: root read permission is granted - ok"
1851 dd conv=notrunc if=$DIR/$tfile-userfile 1>/dev/null 2>/dev/null || \
1852 error "$ST: root write permission is denied"
1853 echo "$ST: root write permission is granted - ok"
1855 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-rootfile)
1856 dd if=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null && \
1857 error "$ST: root read permission is granted"
1858 echo "$ST: root read permission is denied - ok"
1861 dd conv=notrunc of=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null && \
1862 error "$ST: root write permission is granted"
1863 echo "$ST: root write permission is denied - ok"
1865 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tdir-rootdir)
1866 rm $DIR/$tdir-rootdir/tfile-1 1>/dev/null 2>/dev/null && \
1867 error "$ST: root unlink permission is granted"
1868 echo "$ST: root unlink permission is denied - ok"
1870 touch $DIR/tdir-rootdir/tfile-2 1>/dev/null 2>/dev/null && \
1871 error "$ST: root create permission is granted"
1872 echo "$ST: root create permission is denied - ok"
1875 # check nosquash_nids:
1876 # put client's NID into nosquash_nids list,
1877 # root should be able to access root file after that
1879 local NIDLIST=$(lctl list_nids all | tr '\n' ' ')
1880 NIDLIST="2@elan $NIDLIST 192.168.0.[2,10]@tcp"
1881 NIDLIST=$(echo $NIDLIST | tr -s ' ' ' ')
1883 "lctl get_param -n mdt.$FSNAME-MDT0000.nosquash_nids" \
1884 "$FSNAME-MDTall.mdt.nosquash_nids" \
1887 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-rootfile)
1888 dd if=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null || \
1889 error "$ST: root read permission is denied"
1890 echo "$ST: root read permission is granted - ok"
1893 dd conv=notrunc of=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null || \
1894 error "$ST: root write permission is denied"
1895 echo "$ST: root write permission is granted - ok"
1897 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tdir-rootdir)
1898 rm $DIR/$tdir-rootdir/tfile-1 || \
1899 error "$ST: root unlink permission is denied"
1900 echo "$ST: root unlink permission is granted - ok"
1901 touch $DIR/$tdir-rootdir/tfile-2 || \
1902 error "$ST: root create permission is denied"
1903 echo "$ST: root create permission is granted - ok"
1907 run_test 43 "check root_squash and nosquash_nids"
1909 umount_client $MOUNT
1914 check_mount || return 2
1915 UUID=$($LCTL get_param llite.${FSNAME}*.uuid | cut -d= -f2)
1917 UUIDS=$(do_facet $SINGLEMDS "$LCTL get_param mdt.${FSNAME}*.exports.*.uuid")
1918 for VAL in $UUIDS; do
1919 NID=$(echo $VAL | cut -d= -f1)
1920 CLUUID=$(echo $VAL | cut -d= -f2)
1921 [ "$UUID" = "$CLUUID" ] && STATS_FOUND=yes && break
1923 [ "$STATS_FOUND" = "no" ] && error "stats not found for client"
1927 run_test 44 "mounted client proc entry exists"
1931 check_mount || return 2
1936 #define OBD_FAIL_PTLRPC_LONG_UNLINK 0x50f
1937 do_facet client "lctl set_param fail_loc=0x50f"
1940 manual_umount_client --force || return 3
1941 do_facet client "lctl set_param fail_loc=0x0"
1943 mount_client $MOUNT || return 4
1947 run_test 45 "long unlink handling in ptlrpcd"
1954 umount_client $MOUNT2 || rc=$?
1955 umount_client $MOUNT || rc=$?
1956 while [ $count -gt 0 ]; do
1957 stop ost${count} -f || rc=$?
1961 cleanup_nocli || rc=$?
1962 #writeconf to remove all ost2 traces for subsequent tests
1968 echo "Testing with $OSTCOUNT OSTs"
1970 start_mds || return 1
1971 #first client should see only one ost
1972 start_ost || return 2
1973 wait_osc_import_state mds ost FULL
1975 mount_client $MOUNT || return 3
1976 trap "cleanup_46a $OSTCOUNT" EXIT ERR
1979 for (( i=2; i<=$OSTCOUNT; i++ )); do
1980 start ost$i `ostdevname $i` $OST_MOUNT_OPTS || return $((i+2))
1983 # wait until osts in sync
1984 for (( i=2; i<=$OSTCOUNT; i++ )); do
1985 wait_osc_import_state mds ost$i FULL
1986 wait_osc_import_state client ost$i FULL
1989 #second client see all ost's
1991 mount_client $MOUNT2 || return 8
1992 $LFS setstripe $MOUNT2 -c -1 || return 9
1993 $LFS getstripe $MOUNT2 || return 10
1995 echo "ok" > $MOUNT2/widestripe
1996 $LFS getstripe $MOUNT2/widestripe || return 11
1997 # fill acl buffer for avoid expand lsm to them
1998 awk -F : '{if (FNR < 25) { print "u:"$1":rwx" }}' /etc/passwd | while read acl; do
1999 setfacl -m $acl $MOUNT2/widestripe
2003 stat $MOUNT/widestripe || return 12
2005 cleanup_46a $OSTCOUNT || { echo "cleanup_46a failed!" && return 13; }
2008 run_test 46a "handle ost additional - wide striped file"
2013 check_mount || return 2
2014 $LCTL set_param ldlm.namespaces.$FSNAME-*-*-*.lru_size=100
2018 for ns in $($LCTL get_param ldlm.namespaces.$FSNAME-*-*-*.lru_size); do
2019 if echo $ns | grep "MDT[[:digit:]]*"; then
2022 lrs=$(echo $ns | sed 's/.*lru_size=//')
2023 lru_size[count]=$lrs
2028 facet_failover $SINGLEMDS
2029 client_up || return 3
2032 for ns in $($LCTL get_param ldlm.namespaces.$FSNAME-*-*-*.lru_size); do
2033 if echo $ns | grep "MDT[[:digit:]]*"; then
2036 lrs=$(echo $ns | sed 's/.*lru_size=//')
2037 if ! test "$lrs" -eq "${lru_size[count]}"; then
2038 n=$(echo $ns | sed -e 's/ldlm.namespaces.//' -e 's/.lru_size=.*//')
2039 error "$n has lost lru_size: $lrs vs. ${lru_size[count]}"
2047 run_test 47 "server restart does not make client loss lru_resize settings"
2052 # reformat after this test is needed - if test will failed
2053 # we will have unkillable file at FS
2057 test_48() { # bug 17636
2060 check_mount || return 2
2062 $LFS setstripe $MOUNT -c -1 || return 9
2063 $LFS getstripe $MOUNT || return 10
2065 echo "ok" > $MOUNT/widestripe
2066 $LFS getstripe $MOUNT/widestripe || return 11
2068 trap cleanup_48 EXIT ERR
2070 # fill acl buffer for avoid expand lsm to them
2071 getent passwd | awk -F : '{ print "u:"$1":rwx" }' | while read acl; do
2072 setfacl -m $acl $MOUNT/widestripe
2075 stat $MOUNT/widestripe || return 12
2080 run_test 48 "too many acls on file"
2082 # check PARAM_SYS_LDLM_TIMEOUT option of MKFS.LUSTRE
2083 test_49() { # bug 17710
2084 local OLD_MDS_MKFS_OPTS=$MDS_MKFS_OPTS
2085 local OLD_OST_MKFS_OPTS=$OST_MKFS_OPTS
2086 local LOCAL_TIMEOUT=20
2089 OST_MKFS_OPTS="--ost --fsname=$FSNAME --device-size=$OSTSIZE --mgsnode=$MGSNID --param sys.timeout=$LOCAL_TIMEOUT --param sys.ldlm_timeout=$LOCAL_TIMEOUT $MKFSOPT $OSTOPT"
2093 check_mount || return 1
2095 echo "check ldlm_timout..."
2096 LDLM_MDS="`do_facet $SINGLEMDS lctl get_param -n ldlm_timeout`"
2097 LDLM_OST1="`do_facet ost1 lctl get_param -n ldlm_timeout`"
2098 LDLM_CLIENT="`do_facet client lctl get_param -n ldlm_timeout`"
2100 if [ $LDLM_MDS -ne $LDLM_OST1 ] || [ $LDLM_MDS -ne $LDLM_CLIENT ]; then
2101 error "Different LDLM_TIMEOUT:$LDLM_MDS $LDLM_OST1 $LDLM_CLIENT"
2104 if [ $LDLM_MDS -ne $((LOCAL_TIMEOUT / 3)) ]; then
2105 error "LDLM_TIMEOUT($LDLM_MDS) is not correct"
2108 umount_client $MOUNT
2109 stop_ost || return 2
2110 stop_mds || return 3
2112 OST_MKFS_OPTS="--ost --fsname=$FSNAME --device-size=$OSTSIZE --mgsnode=$MGSNID --param sys.timeout=$LOCAL_TIMEOUT --param sys.ldlm_timeout=$((LOCAL_TIMEOUT - 1)) $MKFSOPT $OSTOPT"
2116 check_mount || return 7
2118 LDLM_MDS="`do_facet $SINGLEMDS lctl get_param -n ldlm_timeout`"
2119 LDLM_OST1="`do_facet ost1 lctl get_param -n ldlm_timeout`"
2120 LDLM_CLIENT="`do_facet client lctl get_param -n ldlm_timeout`"
2122 if [ $LDLM_MDS -ne $LDLM_OST1 ] || [ $LDLM_MDS -ne $LDLM_CLIENT ]; then
2123 error "Different LDLM_TIMEOUT:$LDLM_MDS $LDLM_OST1 $LDLM_CLIENT"
2126 if [ $LDLM_MDS -ne $((LOCAL_TIMEOUT - 1)) ]; then
2127 error "LDLM_TIMEOUT($LDLM_MDS) is not correct"
2130 cleanup || return $?
2132 MDS_MKFS_OPTS=$OLD_MDS_MKFS_OPTS
2133 OST_MKFS_OPTS=$OLD_OST_MKFS_OPTS
2135 run_test 49 "check PARAM_SYS_LDLM_TIMEOUT option of MKFS.LUSTRE"
2138 # Test both statfs and lfs df and fail if either one fails
2139 multiop_bg_pause $1 f_
2142 killall -USR1 multiop
2143 [ $RC1 -ne 0 ] && log "lazystatfs multiop failed"
2144 wait $PID || { RC1=$?; log "multiop return error "; }
2151 if [ $RC2 -eq 0 ]; then
2153 log "lazystatfs df failed"
2157 [[ $RC1 -ne 0 || $RC2 -eq 0 ]] && RC=1
2163 lctl set_param llite.$FSNAME-*.lazystatfs=1
2166 lazystatfs $MOUNT || error "lazystatfs failed but no down servers"
2168 cleanup || return $?
2170 run_test 50a "lazystatfs all servers available =========================="
2174 lctl set_param llite.$FSNAME-*.lazystatfs=1
2177 # Wait for client to detect down OST
2178 stop_ost || error "Unable to stop OST1"
2179 wait_osc_import_state mds ost DISCONN
2181 lazystatfs $MOUNT || error "lazystatfs should don't have returned EIO"
2183 umount_client $MOUNT || error "Unable to unmount client"
2184 stop_mds || error "Unable to stop MDS"
2186 run_test 50b "lazystatfs all servers down =========================="
2189 start_mds || error "Unable to start MDS"
2190 start_ost || error "Unable to start OST1"
2191 start_ost2 || error "Unable to start OST2"
2192 mount_client $MOUNT || error "Unable to mount client"
2193 lctl set_param llite.$FSNAME-*.lazystatfs=1
2196 # Wait for client to detect down OST
2197 stop_ost || error "Unable to stop OST1"
2198 wait_osc_import_state mds ost DISCONN
2199 lazystatfs $MOUNT || error "lazystatfs failed with one down server"
2201 umount_client $MOUNT || error "Unable to unmount client"
2202 stop_ost2 || error "Unable to stop OST2"
2203 stop_mds || error "Unable to stop MDS"
2204 #writeconf to remove all ost2 traces for subsequent tests
2207 run_test 50c "lazystatfs one server down =========================="
2210 start_mds || error "Unable to start MDS"
2211 start_ost || error "Unable to start OST1"
2212 start_ost2 || error "Unable to start OST2"
2213 mount_client $MOUNT || error "Unable to mount client"
2214 lctl set_param llite.$FSNAME-*.lazystatfs=1
2217 # Issue the statfs during the window where the client still
2218 # belives the OST to be available but it is in fact down.
2219 # No failure just a statfs which hangs for a timeout interval.
2220 stop_ost || error "Unable to stop OST1"
2221 lazystatfs $MOUNT || error "lazystatfs failed with one down server"
2223 umount_client $MOUNT || error "Unable to unmount client"
2224 stop_ost2 || error "Unable to stop OST2"
2225 stop_mds || error "Unable to stop MDS"
2226 #writeconf to remove all ost2 traces for subsequent tests
2229 run_test 50d "lazystatfs client/server conn race =========================="
2236 start_mds || return 1
2237 #first client should see only one ost
2238 start_ost || return 2
2239 wait_osc_import_state mds ost FULL
2241 # Wait for client to detect down OST
2242 stop_ost || error "Unable to stop OST1"
2243 wait_osc_import_state mds ost DISCONN
2245 mount_client $MOUNT || error "Unable to mount client"
2246 lctl set_param llite.$FSNAME-*.lazystatfs=0
2248 multiop_bg_pause $MOUNT _f
2252 if [ $RC1 -ne 0 ]; then
2253 log "multiop failed $RC1"
2256 sleep $(( $TIMEOUT+1 ))
2258 [ $? -ne 0 ] && error "process isn't sleep"
2259 start_ost || error "Unable to start OST1"
2260 wait $pid || error "statfs failed"
2263 umount_client $MOUNT || error "Unable to unmount client"
2264 stop_ost || error "Unable to stop OST1"
2265 stop_mds || error "Unable to stop MDS"
2267 run_test 50e "normal statfs all servers down =========================="
2272 CONN_PROC="osc.$FSNAME-OST0001-osc-[M]*.ost_server_uuid"
2274 start_mds || error "Unable to start mds"
2275 #first client should see only one ost
2276 start_ost || error "Unable to start OST1"
2277 wait_osc_import_state mds ost FULL
2279 start_ost2 || error "Unable to start OST2"
2280 wait_osc_import_state mds ost2 FULL
2282 # Wait for client to detect down OST
2283 stop_ost2 || error "Unable to stop OST2"
2285 wait_osc_import_state mds ost2 DISCONN
2286 mount_client $MOUNT || error "Unable to mount client"
2287 lctl set_param llite.$FSNAME-*.lazystatfs=0
2289 multiop_bg_pause $MOUNT _f
2293 if [ $RC1 -ne 0 ]; then
2294 log "lazystatfs multiop failed $RC1"
2297 sleep $(( $TIMEOUT+1 ))
2299 [ $? -ne 0 ] && error "process isn't sleep"
2300 start_ost2 || error "Unable to start OST2"
2301 wait $pid || error "statfs failed"
2302 stop_ost2 || error "Unable to stop OST2"
2305 umount_client $MOUNT || error "Unable to unmount client"
2306 stop_ost || error "Unable to stop OST1"
2307 stop_mds || error "Unable to stop MDS"
2308 #writeconf to remove all ost2 traces for subsequent tests
2311 run_test 50f "normal statfs one server in down =========================="
2314 [ "$OSTCOUNT" -lt "2" ] && skip_env "$OSTCOUNT < 2, skipping" && return
2316 start_ost2 || error "Unable to start OST2"
2317 wait_osc_import_state mds ost2 FULL
2318 wait_osc_import_state client ost2 FULL
2320 local PARAM="${FSNAME}-OST0001.osc.active"
2322 $LFS setstripe -c -1 $DIR/$tfile || error "Unable to lfs setstripe"
2323 do_facet mgs $LCTL conf_param $PARAM=0 || error "Unable to deactivate OST"
2325 umount_client $MOUNT || error "Unable to unmount client"
2326 mount_client $MOUNT || error "Unable to mount client"
2327 # This df should not cause a panic
2330 do_facet mgs $LCTL conf_param $PARAM=1 || error "Unable to activate OST"
2332 umount_client $MOUNT || error "Unable to unmount client"
2333 stop_ost2 || error "Unable to stop OST2"
2334 stop_ost || error "Unable to stop OST1"
2335 stop_mds || error "Unable to stop MDS"
2336 #writeconf to remove all ost2 traces for subsequent tests
2339 run_test 50g "deactivated OST should not cause panic====================="
2342 local LOCAL_TIMEOUT=20
2346 check_mount || return 1
2349 $LFS setstripe -c -1 $MOUNT/d1
2350 #define OBD_FAIL_MDS_REINT_DELAY 0x142
2351 do_facet $SINGLEMDS "lctl set_param fail_loc=0x142"
2352 touch $MOUNT/d1/f1 &
2355 start_ost2 || return 2
2357 stop_ost2 || return 3
2359 #writeconf to remove all ost2 traces for subsequent tests
2362 run_test 51 "Verify that mdt_reint handles RMF_MDT_MD correctly when an OST is added"
2371 do_node $node mkdir -p $dest
2372 [ $? -eq 0 ] || { error "Unable to create directory"; return 1; }
2374 do_node $node 'tar cf - '$@' | tar xf - -C '$dest';
2375 [ \"\${PIPESTATUS[*]}\" = \"0 0\" ] || exit 1'
2376 [ $? -eq 0 ] || { error "Unable to tar files"; return 2; }
2378 do_node $node 'getfattr -d -m "[a-z]*\\." '$@' > '$xattrs
2379 [ $? -eq 0 ] || { error "Unable to read xattrs"; return 3; }
2389 local backup2=${TMP}/backup2
2391 do_node $node mkdir -p $backup2
2392 [ $? -eq 0 ] || { error "Unable to create directory"; return 1; }
2394 do_node $node 'tar cf - '$@' | tar xf - -C '$backup2';
2395 [ \"\${PIPESTATUS[*]}\" = \"0 0\" ] || exit 1'
2396 [ $? -eq 0 ] || { error "Unable to tar files to diff"; return 2; }
2398 do_node $node "diff -rq $backup $backup2"
2399 [ $? -eq 0 ] || { error "contents differ"; return 3; }
2401 local xattrs2=${TMP}/xattrs2
2402 do_node $node 'getfattr -d -m "[a-z]*\\." '$@' > '$xattrs2
2403 [ $? -eq 0 ] || { error "Unable to read xattrs to diff"; return 4; }
2405 do_node $node "diff $xattrs $xattrs2"
2406 [ $? -eq 0 ] || { error "xattrs differ"; return 5; }
2408 do_node $node "rm -rf $backup2 $xattrs2"
2409 [ $? -eq 0 ] || { error "Unable to delete temporary files"; return 6; }
2414 [ $? -eq 0 ] || { error "Unable to start MDS"; return 1; }
2416 [ $? -eq 0 ] || { error "Unable to start OST1"; return 2; }
2418 [ $? -eq 0 ] || { error "Unable to mount client"; return 3; }
2421 local ost1mnt=${MOUNT%/*}/ost1
2422 local ost1node=$(facet_active_host ost1)
2423 local ost1tmp=$TMP/conf52
2426 [ $? -eq 0 ] || { error "Unable to create tdir"; return 4; }
2427 touch $TMP/modified_first
2428 [ $? -eq 0 ] || { error "Unable to create temporary file"; return 5; }
2429 local mtime=$(stat -c %Y $TMP/modified_first)
2430 do_node $ost1node "mkdir -p $ost1tmp && touch -m -d @$mtime $ost1tmp/modified_first"
2432 [ $? -eq 0 ] || { error "Unable to create temporary file"; return 6; }
2435 $LFS setstripe $DIR/$tdir -c -1 -s 1M
2436 [ $? -eq 0 ] || { error "lfs setstripe failed"; return 7; }
2438 for (( i=0; i < nrfiles; i++ )); do
2439 multiop $DIR/$tdir/$tfile-$i Ow1048576w1048576w524288c
2440 [ $? -eq 0 ] || { error "multiop failed"; return 8; }
2446 echo backup files to $TMP/files
2447 local files=$(find $DIR/$tdir -type f -newer $TMP/modified_first)
2448 copy_files_xattrs `hostname` $TMP/files $TMP/file_xattrs $files
2449 [ $? -eq 0 ] || { error "Unable to copy files"; return 9; }
2451 umount_client $MOUNT
2452 [ $? -eq 0 ] || { error "Unable to umount client"; return 10; }
2454 [ $? -eq 0 ] || { error "Unable to stop ost1"; return 11; }
2456 echo mount ost1 as ldiskfs
2457 do_node $ost1node mkdir -p $ost1mnt
2458 [ $? -eq 0 ] || { error "Unable to create $ost1mnt"; return 23; }
2459 do_node $ost1node mount -t $FSTYPE $ost1_dev $ost1mnt $OST_MOUNT_OPTS
2460 [ $? -eq 0 ] || { error "Unable to mount ost1 as ldiskfs"; return 12; }
2463 echo backup objects to $ost1tmp/objects
2464 local objects=$(do_node $ost1node 'find '$ost1mnt'/O/0 -type f -size +0'\
2465 '-newer '$ost1tmp'/modified_first -regex ".*\/[0-9]+"')
2466 copy_files_xattrs $ost1node $ost1tmp/objects $ost1tmp/object_xattrs $objects
2467 [ $? -eq 0 ] || { error "Unable to copy objects"; return 13; }
2469 # move objects to lost+found
2470 do_node $ost1node 'mv '$objects' '${ost1mnt}'/lost+found'
2471 [ $? -eq 0 ] || { error "Unable to move objects"; return 14; }
2474 do_node $ost1node "ll_recover_lost_found_objs -d $ost1mnt/lost+found"
2475 [ $? -eq 0 ] || { error "ll_recover_lost_found_objs failed"; return 15; }
2477 # compare restored objects against saved ones
2478 diff_files_xattrs $ost1node $ost1tmp/objects $ost1tmp/object_xattrs $objects
2479 [ $? -eq 0 ] || { error "Unable to diff objects"; return 16; }
2481 do_node $ost1node "umount $ost1mnt"
2482 [ $? -eq 0 ] || { error "Unable to umount ost1 as ldiskfs"; return 17; }
2485 [ $? -eq 0 ] || { error "Unable to start ost1"; return 18; }
2487 [ $? -eq 0 ] || { error "Unable to mount client"; return 19; }
2490 diff_files_xattrs `hostname` $TMP/files $TMP/file_xattrs $files
2491 [ $? -eq 0 ] || { error "Unable to diff files"; return 20; }
2493 rm -rf $TMP/files $TMP/file_xattrs
2494 [ $? -eq 0 ] || { error "Unable to delete temporary files"; return 21; }
2495 do_node $ost1node "rm -rf $ost1tmp"
2496 [ $? -eq 0 ] || { error "Unable to delete temporary files"; return 22; }
2499 run_test 52 "check recovering objects from lost+found"
2501 # Checks threads_min/max/started for some service
2503 # Arguments: service name (OST or MDT), facet (e.g., ost1, $SINGLEMDS), and a
2504 # parameter pattern prefix like 'ost.*.ost'.
2516 local msg="Insane $modname thread counts"
2520 check_mount || return 41
2522 # We need to expand $parampat, but it may match multiple parameters, so
2523 # we'll pick the first one
2524 if ! paramp=$(do_facet $facet "lctl get_param -N ${parampat}.threads_min"|head -1); then
2525 error "Couldn't expand ${parampat}.threads_min parameter name"
2529 # Remove the .threads_min part
2530 paramp=${paramp%.threads_min}
2532 # Check for sanity in defaults
2533 tmin=$(do_facet $facet "lctl get_param -n ${paramp}.threads_min" || echo 0)
2534 tmax=$(do_facet $facet "lctl get_param -n ${paramp}.threads_max" || echo 0)
2535 tstarted=$(do_facet $facet "lctl get_param -n ${paramp}.threads_started" || echo 0)
2536 lassert 23 "$msg (PDSH problems?)" '(($tstarted && $tmin && $tmax))' || return $?
2537 lassert 24 "$msg" '(($tstarted >= $tmin && $tstarted <= $tmax ))' || return $?
2539 # Check that we can change min/max
2540 do_facet $facet "lctl set_param ${paramp}.threads_min=$((tmin + 1))"
2541 do_facet $facet "lctl set_param ${paramp}.threads_max=$((tmax - 1))"
2542 tmin2=$(do_facet $facet "lctl get_param -n ${paramp}.threads_min" || echo 0)
2543 tmax2=$(do_facet $facet "lctl get_param -n ${paramp}.threads_max" || echo 0)
2544 lassert 25 "$msg" '(($tmin2 == ($tmin + 1) && $tmax2 == ($tmax -1)))' || return $?
2546 # Check that we can set min/max to the same value
2547 tmin=$(do_facet $facet "lctl get_param -n ${paramp}.threads_min" || echo 0)
2548 do_facet $facet "lctl set_param ${paramp}.threads_max=$tmin"
2549 tmin2=$(do_facet $facet "lctl get_param -n ${paramp}.threads_min" || echo 0)
2550 tmax2=$(do_facet $facet "lctl get_param -n ${paramp}.threads_max" || echo 0)
2551 lassert 26 "$msg" '(($tmin2 == $tmin && $tmax2 == $tmin))' || return $?
2553 # Check that we can't set max < min
2554 do_facet $facet "lctl set_param ${paramp}.threads_max=$((tmin - 1))"
2555 tmin2=$(do_facet $facet "lctl get_param -n ${paramp}.threads_min" || echo 0)
2556 tmax2=$(do_facet $facet "lctl get_param -n ${paramp}.threads_max" || echo 0)
2557 lassert 27 "$msg" '(($tmin2 <= $tmax2))' || return $?
2559 # We need to ensure that we get the module options desired; to do this
2560 # we set LOAD_MODULES_REMOTE=true and we call setmodopts below.
2561 LOAD_MODULES_REMOTE=true
2564 setmodopts -a $modname "$opts" oldvalue
2568 check_mount || return 41
2570 # Restore previous setting of MODOPTS_*
2571 setmodopts $modname "$oldvalue"
2573 # Check that $opts took
2574 tmin=$(do_facet $facet "lctl get_param -n ${paramp}.threads_min")
2575 tmax=$(do_facet $facet "lctl get_param -n ${paramp}.threads_max")
2576 tstarted=$(do_facet $facet "lctl get_param -n ${paramp}.threads_started")
2577 lassert 28 "$msg" '(($tstarted == $tmin && $tstarted == $tmax ))' || return $?
2580 # Workaround a YALA bug where YALA expects that modules will remain
2581 # loaded on the servers
2582 LOAD_MODULES_REMOTE=false
2589 thread_sanity OST ost1 'ost.*.ost' 'oss_num_threads=64'
2591 run_test 53a "check OSS thread count params"
2594 thread_sanity MDT $SINGLEMDS 'mdt.*.*.' 'mdt_num_threads=64'
2596 run_test 53b "check MDT thread count params"
2599 do_rpc_nodes $(facet_host ost1) run_llverdev $(ostdevname 1) -p
2600 [ $? -eq 0 ] || error "llverdev failed!"
2603 run_test 54a "test llverdev and partial verify of device"
2607 run_llverfs $MOUNT -p
2608 [ $? -eq 0 ] || error "llverfs failed!"
2611 run_test 54b "test llverfs and partial verify of filesystem"
2615 local max_ost_index=$1
2616 echo -n $(((max_ost_index + 1) * 8))
2620 local mdsdev=$(mdsdevname 1)
2621 local ostdev=$(ostdevname 1)
2622 local saved_opts=$OST_MKFS_OPTS
2626 OST_MKFS_OPTS="$saved_opts --index $i"
2634 echo checking size of lov_objid for ost index $i
2635 LOV_OBJID_SIZE=$(do_facet mds1 "$DEBUGFS -R 'stat lov_objid' $mdsdev 2>/dev/null" | grep ^User | awk '{print $6}')
2636 if [ "$LOV_OBJID_SIZE" != $(lov_objid_size $i) ]; then
2637 error "lov_objid size has to be $(lov_objid_size $i), not $LOV_OBJID_SIZE"
2639 echo ok, lov_objid size is correct: $LOV_OBJID_SIZE
2644 OST_MKFS_OPTS=$saved_opts
2647 run_test 55 "check lov_objid size"
2650 add mds1 $MDS_MKFS_OPTS --mkfsoptions='\"-J size=16\"' --reformat $(mdsdevname 1)
2651 add ost1 $OST_MKFS_OPTS --index=1000 --reformat $(ostdevname 1)
2652 add ost2 $OST_MKFS_OPTS --index=10000 --reformat $(ostdevname 2)
2656 start_ost2 || error "Unable to start second ost"
2657 mount_client $MOUNT || error "Unable to mount client"
2660 [ -n "$ENABLE_QUOTA" ] && { $LFS quotacheck -ug $MOUNT || error "quotacheck has failed" ; }
2664 run_test 56 "check big indexes"
2666 test_57a() { # bug 22656
2667 local NID=$(do_facet ost1 "$LCTL get_param nis" | tail -1 | awk '{print $1}')
2669 do_facet ost1 "$TUNEFS --failnode=$NID `ostdevname 1`" || error "tunefs failed"
2671 start_ost && error "OST registration from failnode should fail"
2674 run_test 57a "initial registration from failnode should fail (should return errs)"
2677 local NID=$(do_facet ost1 "$LCTL get_param nis" | tail -1 | awk '{print $1}')
2679 do_facet ost1 "$TUNEFS --servicenode=$NID `ostdevname 1`" || error "tunefs failed"
2681 start_ost || error "OST registration from servicenode should not fail"
2684 run_test 57b "initial registration from servicenode should not fail"
2687 do_facet mgs $LCTL get_param mgs.MGS.live.$FSNAME | grep OST | wc -l
2690 test_58() { # bug 22658
2691 [ "$FSTYPE" != "ldiskfs" ] && skip "not supported for $FSTYPE" && return
2694 createmany -o $DIR/$tdir/$tfile-%d 100
2695 # make sure that OSTs do not cancel llog cookies before we unmount the MDS
2696 #define OBD_FAIL_OBD_LOG_CANCEL_NET 0x601
2697 do_facet mds "lctl set_param fail_loc=0x601"
2698 unlinkmany $DIR/$tdir/$tfile-%d 100
2700 local MNTDIR=$(facet_mntpt mds)
2701 # remove all files from the OBJECTS dir
2702 do_facet mds "mount -t ldiskfs $MDSDEV $MNTDIR"
2703 do_facet mds "find $MNTDIR/OBJECTS -type f -delete"
2704 do_facet mds "umount $MNTDIR"
2705 # restart MDS with missing llog files
2707 do_facet mds "lctl set_param fail_loc=0"
2710 run_test 58 "missing llog files must not prevent MDT from mounting"
2713 start_mgsmds >> /dev/null
2714 local C1=$(count_osts)
2715 if [ $C1 -eq 0 ]; then
2716 start_ost >> /dev/null
2720 echo "original ost count: $C1 (expect > 0)"
2721 [ $C1 -gt 0 ] || error "No OSTs in $FSNAME log"
2722 start_mgsmds -o writeconf >> /dev/null || error "MDT start failed"
2723 local C2=$(count_osts)
2724 echo "after mdt writeconf count: $C2 (expect 0)"
2725 [ $C2 -gt 0 ] && error "MDT writeconf should erase OST logs"
2726 echo "OST start without writeconf should fail:"
2727 start_ost >> /dev/null && error "OST start without writeconf didn't fail"
2728 echo "OST start with writeconf should succeed:"
2729 start_ost -o writeconf >> /dev/null || error "OST1 start failed"
2730 local C3=$(count_osts)
2731 echo "after ost writeconf count: $C3 (expect 1)"
2732 [ $C3 -eq 1 ] || error "new OST writeconf should add:"
2733 start_ost2 -o writeconf >> /dev/null || error "OST2 start failed"
2734 local C4=$(count_osts)
2735 echo "after ost2 writeconf count: $C4 (expect 2)"
2736 [ $C4 -eq 2 ] || error "OST2 writeconf should add log"
2737 stop_ost2 >> /dev/null
2738 cleanup_nocli >> /dev/null
2739 #writeconf to remove all ost2 traces for subsequent tests
2742 run_test 59 "writeconf mount option"
2744 test_60() { # LU-471
2745 add mds1 $MDS_MKFS_OPTS --mkfsoptions='\" -E stride=64 -O ^uninit_bg\"' --reformat $(mdsdevname 1)
2747 dump=$(do_facet $SINGLEMDS dumpe2fs $(mdsdevname 1))
2749 [ $rc -eq 0 ] || error "dumpe2fs $(mdsdevname 1) failed"
2751 # MDT default has dirdata feature
2752 echo $dump | grep dirdata > /dev/null || error "dirdata is not set"
2753 # we disable uninit_bg feature
2754 echo $dump | grep uninit_bg > /dev/null && error "uninit_bg is set"
2755 # we set stride extended options
2756 echo $dump | grep stride > /dev/null || error "stride is not set"
2759 run_test 60 "check mkfs.lustre --mkfsoptions -E -O options setting"
2761 if ! combined_mgs_mds ; then
2767 # restore the ${facet}_MKFS_OPTS variables
2768 for facet in MGS MDS OST; do
2769 opts=SAVED_${facet}_MKFS_OPTS
2770 if [[ -n ${!opts} ]]; then
2771 eval ${facet}_MKFS_OPTS=\"${!opts}\"
2775 complete $(basename $0) $SECONDS