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
55 require_dsh_mds || exit 0
56 require_dsh_ost || exit 0
58 [ "$SLOW" = "no" ] && EXCEPT_SLOW="30a 31 45"
73 # who knows if/where $TUNEFS is installed? Better reformat if it fails...
74 do_facet ${facet} "$TUNEFS --quiet --writeconf $dev" ||
75 { echo "tunefs failed, reformatting instead" && reformat_and_config && return 1; }
82 # if writeconf fails anywhere, we reformat everything
83 writeconf1 mds `mdsdevname 1` || return 0
84 writeconf1 ost1 `ostdevname 1` || return 0
85 writeconf1 ost2 `ostdevname 2` || return 0
89 # The MGS must be started before the OSTs for a new fs, so start
90 # and stop to generate the startup logs.
93 wait_osc_import_state mds ost FULL
98 reformat_and_config() {
100 if ! combined_mgs_mds ; then
108 start mgs $MGSDEV $MGS_MOUNT_OPTS
112 local facet=$SINGLEMDS
113 # we can not use MDSDEV1 here because SINGLEMDS could be set not to mds1 only
114 local num=$(echo $facet | tr -d "mds")
115 local dev=$(mdsdevname $num)
116 echo "start mds service on `facet_active_host $facet`"
117 start $facet ${dev} $MDS_MOUNT_OPTS $@ || return 94
121 if ! combined_mgs_mds ; then
128 echo "stop mds service on `facet_active_host $SINGLEMDS`"
129 # These tests all use non-failover stop
130 stop $SINGLEMDS -f || return 97
134 echo "start ost1 service on `facet_active_host ost1`"
135 start ost1 `ostdevname 1` $OST_MOUNT_OPTS $@ || return 95
139 echo "stop ost1 service on `facet_active_host ost1`"
140 # These tests all use non-failover stop
141 stop ost1 -f || return 98
145 echo "start ost2 service on `facet_active_host ost2`"
146 start ost2 `ostdevname 2` $OST_MOUNT_OPTS $@ || return 92
150 echo "stop ost2 service on `facet_active_host ost2`"
151 # These tests all use non-failover stop
152 stop ost2 -f || return 93
157 echo "mount $FSNAME on ${MOUNTPATH}....."
158 zconf_mount `hostname` $MOUNTPATH || return 96
162 local SAVEMOUNTOPT=$MOUNTOPT
163 MOUNTOPT="remount,$1"
165 echo "remount '$1' lustre on ${MOUNTPATH}....."
166 zconf_mount `hostname` $MOUNTPATH || return 96
167 MOUNTOPT=$SAVEMOUNTOPT
172 echo "umount lustre on ${MOUNTPATH}....."
173 zconf_umount `hostname` $MOUNTPATH || return 97
176 manual_umount_client(){
179 echo "manual umount lustre on ${MOUNT}...."
180 do_facet client "umount -d ${FORCE} $MOUNT"
186 start_mds || error "MDT start failed"
187 start_ost || error "OST start failed"
188 mount_client $MOUNT || error "client start failed"
192 if ! combined_mgs_mds ; then
201 unload_modules_conf () {
202 if combined_mgs_mds || ! local_mode; then
203 unload_modules || return 1
208 stop_ost || return 202
209 stop_mds || return 201
210 unload_modules_conf || return 203
214 umount_client $MOUNT || return 200
215 cleanup_nocli || return $?
219 do_facet client "cp /etc/passwd $DIR/a" || return 71
220 do_facet client "rm $DIR/a" || return 72
221 # make sure lustre is actually mounted (touch will block,
222 # but grep won't, so do it after)
223 do_facet client "grep $MOUNT' ' /proc/mounts > /dev/null" || return 73
224 echo "setup single mount lustre success"
228 do_facet client "touch $DIR/a" || return 71
229 do_facet client "rm $DIR/a" || return 72
230 do_facet client "touch $DIR2/a" || return 73
231 do_facet client "rm $DIR2/a" || return 74
232 echo "setup double mount lustre success"
237 if [ "$ONLY" == "setup" ]; then
242 if [ "$ONLY" == "cleanup" ]; then
249 #create single point mountpoint
255 check_mount || return 41
258 run_test 0 "single mount setup"
261 start_mds || error "MDT start failed"
263 echo "start ost second time..."
264 start_ost && error "2nd OST start should fail"
265 mount_client $MOUNT || error "client start failed"
266 check_mount || return 42
269 run_test 1 "start up ost twice (should return errors)"
273 echo "start mds second time.."
274 start_mds && error "2nd MDT start should fail"
277 check_mount || return 43
280 run_test 2 "start up mds twice (should return err)"
284 #mount.lustre returns an error if already in mtab
285 mount_client $MOUNT && error "2nd client mount should fail"
286 check_mount || return 44
289 run_test 3 "mount client twice (should return err)"
293 touch $DIR/$tfile || return 85
297 # ok for ost to fail shutdown
298 if [ 202 -ne $eno ]; then
303 run_test 4 "force cleanup ost, then cleanup"
305 test_5a() { # was test_5
307 touch $DIR/$tfile || return 1
308 fuser -m -v $MOUNT && echo "$MOUNT is in use by user space process."
310 stop_mds -f || return 2
312 # cleanup may return an error from the failed
313 # disconnects; for now I'll consider this successful
314 # if all the modules have unloaded.
318 echo "killing umount"
319 kill -TERM $UMOUNT_PID
320 echo "waiting for umount to finish"
322 if grep " $MOUNT " /proc/mounts; then
323 echo "test 5: /proc/mounts after failed umount"
327 echo "killing umount"
328 kill -TERM $UMOUNT_PID
329 echo "waiting for umount to finish"
331 grep " $MOUNT " /proc/mounts && echo "test 5: /proc/mounts after second umount" && return 11
335 # stop_mds is a no-op here, and should not fail
336 cleanup_nocli || return $?
337 # df may have lingering entry
339 # mtab may have lingering entry
343 while [ "$WAIT" -ne "$MAX_WAIT" ]; do
345 grep -q $MOUNT" " /etc/mtab || break
346 echo "Waiting /etc/mtab updated ... "
347 WAIT=$(( WAIT + sleep))
349 [ "$WAIT" -eq "$MAX_WAIT" ] && error "/etc/mtab is not updated in $WAIT secs"
350 echo "/etc/mtab updated in $WAIT secs"
352 run_test 5a "force cleanup mds, then cleanup"
360 grep " $MOUNT " /etc/mtab && \
361 error false "unexpected entry in mtab before mount" && return 10
365 if ! combined_mgs_mds ; then
366 trap cleanup_5b EXIT ERR
371 [ -d $MOUNT ] || mkdir -p $MOUNT
372 mount_client $MOUNT && rc=1
373 grep " $MOUNT " /etc/mtab && \
374 error "$MOUNT entry in mtab after failed mount" && rc=11
376 # stop_mds is a no-op here, and should not fail
377 cleanup_nocli || rc=$?
378 if ! combined_mgs_mds ; then
383 run_test 5b "Try to start a client with no MGS (should return errs)"
386 grep " $MOUNT " /etc/mtab && \
387 error false "unexpected entry in mtab before mount" && return 10
392 [ -d $MOUNT ] || mkdir -p $MOUNT
393 local oldfs="${FSNAME}"
394 FSNAME="wrong.${FSNAME}"
395 mount_client $MOUNT || :
397 grep " $MOUNT " /etc/mtab && \
398 error "$MOUNT entry in mtab after failed mount" && rc=11
400 cleanup_nocli || rc=$?
403 run_test 5c "cleanup after failed mount (bug 2712) (should return errs)"
406 grep " $MOUNT " /etc/mtab && \
407 error false "unexpected entry in mtab before mount" && return 10
413 mount_client $MOUNT || rc=1
415 grep " $MOUNT " /etc/mtab && \
416 error "$MOUNT entry in mtab after unmount" && rc=11
419 run_test 5d "mount with ost down"
422 grep " $MOUNT " /etc/mtab && \
423 error false "unexpected entry in mtab before mount" && return 10
429 #define OBD_FAIL_PTLRPC_DELAY_SEND 0x506
430 do_facet client "lctl set_param fail_loc=0x80000506"
431 mount_client $MOUNT || echo "mount failed (not fatal)"
433 grep " $MOUNT " /etc/mtab && \
434 error "$MOUNT entry in mtab after unmount" && rc=11
437 run_test 5e "delayed connect, don't crash (bug 10268)"
440 if combined_mgs_mds ; then
441 skip "combined mgs and mds"
445 grep " $MOUNT " /etc/mtab && \
446 error false "unexpected entry in mtab before mount" && return 10
450 [ -d $MOUNT ] || mkdir -p $MOUNT
451 mount_client $MOUNT &
453 echo client_mount pid is $pid
457 if ! ps -f -p $pid >/dev/null; then
460 grep " $MOUNT " /etc/mtab && echo "test 5f: mtab after mount"
461 error "mount returns $rc, expected to hang"
470 # mount should succeed after start mds
473 [ $rc -eq 0 ] || error "mount returned $rc"
474 grep " $MOUNT " /etc/mtab && echo "test 5f: mtab after mount"
478 run_test 5f "mds down, cleanup after failed mount (bug 2712)"
483 mount_client ${MOUNT} || return 87
484 touch $DIR/a || return 86
487 run_test 6 "manual umount, then mount again"
492 cleanup_nocli || return $?
494 run_test 7 "manual umount, then cleanup"
499 check_mount2 || return 45
500 umount_client $MOUNT2
503 run_test 8 "double mount setup"
508 do_facet ost1 lctl set_param debug=\'inode trace\' || return 1
509 do_facet ost1 lctl set_param subsystem_debug=\'mds ost\' || return 1
511 CHECK_PTLDEBUG="`do_facet ost1 lctl get_param -n debug`"
512 if [ "$CHECK_PTLDEBUG" ] && { \
513 [ "$CHECK_PTLDEBUG" = "trace inode warning error emerg console" ] ||
514 [ "$CHECK_PTLDEBUG" = "trace inode" ]; }; then
515 echo "lnet.debug success"
517 echo "lnet.debug: want 'trace inode', have '$CHECK_PTLDEBUG'"
520 CHECK_SUBSYS="`do_facet ost1 lctl get_param -n subsystem_debug`"
521 if [ "$CHECK_SUBSYS" ] && [ "$CHECK_SUBSYS" = "mds ost" ]; then
522 echo "lnet.subsystem_debug success"
524 echo "lnet.subsystem_debug: want 'mds ost', have '$CHECK_SUBSYS'"
527 stop_ost || return $?
529 run_test 9 "test ptldebug and subsystem for mkfs"
537 do_facet $facet "test -b $dev" || rc=1
538 if [[ "$size" ]]; then
539 local in=$(do_facet $facet "dd if=$dev of=/dev/null bs=1k count=1 skip=$size 2>&1" |\
540 awk '($3 == "in") { print $1 }')
541 [[ $in = "1+0" ]] || rc=1
547 # Test 16 was to "verify that lustre will correct the mode of OBJECTS".
548 # But with new MDS stack we don't care about the mode of local objects
549 # anymore, so this test is removed. See bug 22944 for more details.
554 check_mount || return 41
557 echo "Remove mds config log"
558 if ! combined_mgs_mds ; then
562 do_facet mgs "$DEBUGFS -w -R 'unlink CONFIGS/$FSNAME-MDT0000' $MGSDEV || return \$?" || return $?
564 if ! combined_mgs_mds ; then
569 start_mds && return 42
572 run_test 17 "Verify failed mds_postsetup won't fail assertion (2936) (should return errs)"
575 [ "$FSTYPE" != "ldiskfs" ] && skip "not needed for FSTYPE=$FSTYPE" && return
577 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
582 # check if current MDSSIZE is large enough
583 [ $MDSSIZE -ge $MIN ] && OK=1 && myMDSSIZE=$MDSSIZE && \
584 log "use MDSSIZE=$MDSSIZE"
586 # check if the global config has a large enough MDSSIZE
587 [ -z "$OK" -a ! -z "$STORED_MDSSIZE" ] && [ $STORED_MDSSIZE -ge $MIN ] && \
588 OK=1 && myMDSSIZE=$STORED_MDSSIZE && \
589 log "use STORED_MDSSIZE=$STORED_MDSSIZE"
591 # check if the block device is large enough
592 [ -z "$OK" ] && $(is_blkdev $SINGLEMDS $MDSDEV $MIN) && OK=1 &&
593 myMDSSIZE=$MIN && log "use device $MDSDEV with MIN=$MIN"
595 # check if a loopback device has enough space for fs metadata (5%)
597 if [ -z "$OK" ]; then
598 local SPACE=$(do_facet $SINGLEMDS "[ -f $MDSDEV -o ! -e $MDSDEV ] && df -P \\\$(dirname $MDSDEV)" |
599 awk '($1 != "Filesystem") {print $4}')
600 ! [ -z "$SPACE" ] && [ $SPACE -gt $((MIN / 20)) ] && \
601 OK=1 && myMDSSIZE=$MIN && \
602 log "use file $MDSDEV with MIN=$MIN"
605 [ -z "$OK" ] && skip_env "$MDSDEV too small for ${MIN}kB MDS" && return
608 echo "mount mds with large journal..."
609 local OLD_MDS_MKFS_OPTS=$MDS_MKFS_OPTS
611 local opts="--mdt --fsname=$FSNAME --device-size=$myMDSSIZE --param sys.timeout=$TIMEOUT $MDSOPT"
613 if combined_mgs_mds ; then
614 MDS_MKFS_OPTS="--mgs $opts"
616 MDS_MKFS_OPTS="--mgsnode=$MGSNID $opts"
620 echo "mount lustre system..."
622 check_mount || return 41
624 echo "check journal size..."
625 local FOUNDSIZE=$(do_facet $SINGLEMDS "$DEBUGFS -c -R 'stat <8>' $MDSDEV" | awk '/Size: / { print $NF; exit;}')
626 if [ $FOUNDSIZE -gt $((32 * 1024 * 1024)) ]; then
627 log "Success: mkfs creates large journals. Size: $((FOUNDSIZE >> 20))M"
629 error "expected journal size > 32M, found $((FOUNDSIZE >> 20))M"
634 MDS_MKFS_OPTS=$OLD_MDS_MKFS_OPTS
637 run_test 18 "check mkfs creates large journals"
640 start_mds || return 1
641 stop_mds -f || return 2
643 run_test 19a "start/stop MDS without OSTs"
646 start_ost || return 1
647 stop_ost -f || return 2
649 run_test 19b "start/stop OSTs without MDS"
652 # first format the ost/mdt
656 check_mount || return 43
658 remount_client ro $MOUNT || return 44
659 touch $DIR/$tfile && echo "$DIR/$tfile created incorrectly" && return 45
660 [ -e $DIR/$tfile ] && echo "$DIR/$tfile exists incorrectly" && return 46
661 remount_client rw $MOUNT || return 47
663 [ ! -f $DIR/$tfile ] && echo "$DIR/$tfile missing" && return 48
664 MCNT=`grep -c $MOUNT /etc/mtab`
665 [ "$MCNT" -ne 1 ] && echo "$MOUNT in /etc/mtab $MCNT times" && return 49
670 run_test 20 "remount ro,rw mounts work and doesn't break /etc/mtab"
675 wait_osc_import_state mds ost FULL
679 run_test 21a "start mds before ost, stop ost first"
684 wait_osc_import_state mds ost FULL
688 run_test 21b "start ost before mds, stop mds first"
694 wait_osc_import_state mds ost2 FULL
698 #writeconf to remove all ost2 traces for subsequent tests
701 run_test 21c "start mds between two osts, stop mds last"
706 echo Client mount with ost in logs, but none running
708 # wait until mds connected to ost and open client connection
709 wait_osc_import_state mds ost FULL
712 # check_mount will block trying to contact ost
713 mcreate $DIR/$tfile || return 40
714 rm -f $DIR/$tfile || return 42
718 echo Client mount with a running ost
721 # if gss enabled, wait full time to let connection from
722 # mds to ost be established, due to the mismatch between
723 # initial connect timeout and gss context negotiation timeout.
724 # This perhaps could be remove after AT landed.
725 echo "sleep $((TIMEOUT + TIMEOUT + TIMEOUT))s"
726 sleep $((TIMEOUT + TIMEOUT + TIMEOUT))
729 check_mount || return 41
734 run_test 22 "start a client before osts (should return errs)"
736 test_23a() { # was test_23
740 # force down client so that recovering mds waits for reconnect
741 local running=$(grep -c $MOUNT /proc/mounts) || true
742 if [ $running -ne 0 ]; then
743 echo "Stopping client $MOUNT (opts: -f)"
747 # enter recovery on mds
749 # try to start a new client
750 mount_client $MOUNT &
752 MOUNT_PID=$(ps -ef | grep "t lustre" | grep -v grep | awk '{print $2}')
753 MOUNT_LUSTRE_PID=`ps -ef | grep mount.lustre | grep -v grep | awk '{print $2}'`
754 echo mount pid is ${MOUNT_PID}, mount.lustre pid is ${MOUNT_LUSTRE_PID}
756 ps --ppid $MOUNT_LUSTRE_PID
757 # FIXME why o why can't I kill these? Manual "ctrl-c" works...
758 kill -TERM $MOUNT_LUSTRE_PID
759 echo "waiting for mount to finish"
761 # we can not wait $MOUNT_PID because it is not a child of this shell
767 while [ "$WAIT" -lt "$MAX_WAIT" ]; do
769 PID1=$(ps -ef | awk '{print $2}' | grep -w $MOUNT_PID)
770 PID2=$(ps -ef | awk '{print $2}' | grep -w $MOUNT_LUSTRE_PID)
773 [ -z "$PID1" -a -z "$PID2" ] && break
774 echo "waiting for mount to finish ... "
775 WAIT=$(( WAIT + sleep))
777 if [ "$WAIT" -eq "$MAX_WAIT" ]; then
778 error "MOUNT_PID $MOUNT_PID and "\
779 "MOUNT_LUSTRE_PID $MOUNT_LUSTRE_PID still not killed in $WAIT secs"
785 run_test 23a "interrupt client during recovery mount delay"
790 test_23b() { # was test_23
793 # Simulate -EINTR during mount OBD_FAIL_LDLM_CLOSE_THREAD
794 lctl set_param fail_loc=0x80000313
798 run_test 23b "Simulate -EINTR during mount"
800 fs2mds_HOST=$mds_HOST
801 fs2ost_HOST=$ost_HOST
805 echo "umount $MOUNT2 ..."
806 umount $MOUNT2 || true
807 echo "stopping fs2mds ..."
808 stop fs2mds -f || true
809 echo "stopping fs2ost ..."
810 stop fs2ost -f || true
814 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
816 if [ -z "$fs2ost_DEV" -o -z "$fs2mds_DEV" ]; then
817 is_blkdev $SINGLEMDS $MDSDEV && \
818 skip_env "mixed loopback and real device not working" && return
821 [ -n "$ost1_HOST" ] && fs2ost_HOST=$ost1_HOST
823 local fs2mdsdev=${fs2mds_DEV:-${MDSDEV}_2}
824 local fs2ostdev=${fs2ost_DEV:-$(ostdevname 1)_2}
826 # test 8-char fsname as well
827 local FSNAME2=test1234
828 add fs2mds $MDS_MKFS_OPTS --fsname=${FSNAME2} --nomgs --mgsnode=$MGSNID --reformat $fs2mdsdev || exit 10
830 add fs2ost $OST_MKFS_OPTS --fsname=${FSNAME2} --reformat $fs2ostdev || exit 10
833 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS && trap cleanup_24a EXIT INT
834 start fs2ost $fs2ostdev $OST_MOUNT_OPTS
836 mount -t lustre $MGSNID:/${FSNAME2} $MOUNT2 || return 1
838 check_mount || return 2
839 # files written on 1 should not show up on 2
840 cp /etc/passwd $DIR/$tfile
842 [ -e $MOUNT2/$tfile ] && error "File bleed" && return 7
845 cp /etc/passwd $MOUNT2/b || return 3
846 rm $MOUNT2/b || return 4
847 # 2 is actually mounted
848 grep $MOUNT2' ' /proc/mounts > /dev/null || return 5
850 facet_failover fs2mds
851 facet_failover fs2ost
854 # the MDS must remain up until last MDT
856 MDS=$(do_facet $SINGLEMDS "lctl get_param -n devices" | awk '($3 ~ "mdt" && $4 ~ "MDT") { print $4 }' | head -1)
857 [ -z "$MDS" ] && error "No MDT" && return 8
859 cleanup_nocli || return 6
861 run_test 24a "Multiple MDTs on a single node"
864 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
866 if [ -z "$fs2mds_DEV" ]; then
867 local dev=${SINGLEMDS}_dev
869 is_blkdev $SINGLEMDS $MDSDEV && \
870 skip_env "mixed loopback and real device not working" && return
873 local fs2mdsdev=${fs2mds_DEV:-${MDSDEV}_2}
875 add fs2mds $MDS_MKFS_OPTS --fsname=${FSNAME}2 --mgs --reformat $fs2mdsdev || exit 10
877 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS && return 2
880 run_test 24b "Multiple MGSs on a single node (should return err)"
884 check_mount || return 2
885 local MODULES=$($LCTL modules | awk '{ print $2 }')
886 rmmod $MODULES 2>/dev/null || true
889 run_test 25 "Verify modules are referenced"
893 # we need modules before mount for sysctl, so make sure...
894 do_facet $SINGLEMDS "lsmod | grep -q lustre || modprobe lustre"
895 #define OBD_FAIL_MDS_FS_SETUP 0x135
896 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000135"
897 start_mds && echo MDS started && return 1
898 lctl get_param -n devices
899 DEVS=$(lctl get_param -n devices | egrep -v MG | wc -l)
900 [ $DEVS -gt 0 ] && return 2
901 unload_modules_conf || return $?
903 run_test 26 "MDT startup failure cleans LOV (should return errs)"
909 local ORIG=$(do_facet $myfacet "$TEST")
910 if [ $# -gt 3 ]; then
916 echo "Setting $PARAM from $ORIG to $FINAL"
917 do_facet $SINGLEMDS "$LCTL conf_param $PARAM='$FINAL'" || error conf_param failed
919 wait_update $(facet_host $myfacet) "$TEST" "$FINAL" || error check failed!
923 start_ost || return 1
924 start_mds || return 2
925 echo "Requeue thread should have started: "
926 ps -e | grep ll_cfg_requeue
927 set_and_check ost1 "lctl get_param -n obdfilter.$FSNAME-OST0000.client_cache_seconds" "$FSNAME-OST0000.ost.client_cache_seconds" || return 3
930 run_test 27a "Reacquire MGS lock if OST started first"
935 local device=$(do_facet $SINGLEMDS "lctl get_param -n devices" | awk '($3 ~ "mdt" && $4 ~ "MDT") { print $4 }')
937 facet_failover $SINGLEMDS
938 set_and_check $SINGLEMDS "lctl get_param -n mdt.$device.identity_acquire_expire" "$device.mdt.identity_acquire_expire" || return 3
939 set_and_check client "lctl get_param -n mdc.$device-mdc-*.max_rpcs_in_flight" "$device.mdc.max_rpcs_in_flight" || return 4
943 run_test 27b "Reacquire MGS lock after failover"
947 TEST="lctl get_param -n llite.$FSNAME-*.max_read_ahead_whole_mb"
948 PARAM="$FSNAME.llite.max_read_ahead_whole_mb"
951 set_and_check client "$TEST" "$PARAM" $FINAL || return 3
952 FINAL=$(($FINAL + 1))
953 set_and_check client "$TEST" "$PARAM" $FINAL || return 4
954 umount_client $MOUNT || return 200
957 if [ $RESULT -ne $FINAL ]; then
958 echo "New config not seen: wanted $FINAL got $RESULT"
961 echo "New config success: got $RESULT"
963 set_and_check client "$TEST" "$PARAM" $ORIG || return 5
966 run_test 28 "permanent parameter setting"
969 [ "$OSTCOUNT" -lt "2" ] && skip_env "$OSTCOUNT < 2, skipping" && return
970 setup > /dev/null 2>&1
974 local PARAM="$FSNAME-OST0001.osc.active"
975 local PROC_ACT="osc.$FSNAME-OST0001-osc-[^M]*.active"
976 local PROC_UUID="osc.$FSNAME-OST0001-osc-[^M]*.ost_server_uuid"
978 ACTV=$(lctl get_param -n $PROC_ACT)
980 set_and_check client "lctl get_param -n $PROC_ACT" "$PARAM" $DEAC || return 2
981 # also check ost_server_uuid status
982 RESULT=$(lctl get_param -n $PROC_UUID | grep DEACTIV)
983 if [ -z "$RESULT" ]; then
984 echo "Live client not deactivated: $(lctl get_param -n $PROC_UUID)"
987 echo "Live client success: got $RESULT"
991 local mdtosc=$(get_mdtosc_proc_path $SINGLEMDS $FSNAME-OST0001)
992 mdtosc=${mdtosc/-MDT*/-MDT\*}
993 local MPROC="osc.$mdtosc.active"
998 RESULT=`do_facet $SINGLEMDS " lctl get_param -n $MPROC"`
999 [ ${PIPESTATUS[0]} = 0 ] || error "Can't read $MPROC"
1000 if [ $RESULT -eq $DEAC ]; then
1001 echo "MDT deactivated also after $WAIT sec (got $RESULT)"
1005 if [ $WAIT -eq $MAX ]; then
1006 echo "MDT not deactivated: wanted $DEAC got $RESULT"
1009 echo "Waiting $(($MAX - $WAIT)) secs for MDT deactivated"
1012 # test new client starts deactivated
1013 umount_client $MOUNT || return 200
1015 RESULT=$(lctl get_param -n $PROC_UUID | grep DEACTIV | grep NEW)
1016 if [ -z "$RESULT" ]; then
1017 echo "New client not deactivated from start: $(lctl get_param -n $PROC_UUID)"
1020 echo "New client success: got $RESULT"
1023 # make sure it reactivates
1024 set_and_check client "lctl get_param -n $PROC_ACT" "$PARAM" $ACTV || return 6
1026 umount_client $MOUNT
1029 #writeconf to remove all ost2 traces for subsequent tests
1032 run_test 29 "permanently remove an OST"
1037 echo Big config llog
1038 TEST="lctl get_param -n llite.$FSNAME-*.max_read_ahead_whole_mb"
1040 LIST=(1 2 3 4 5 4 3 2 1 2 3 4 5 4 3 2 1 2 3 4 5)
1041 for i in ${LIST[@]}; do
1042 set_and_check client "$TEST" "$FSNAME.llite.max_read_ahead_whole_mb" $i || return 3
1044 # make sure client restart still works
1045 umount_client $MOUNT
1046 mount_client $MOUNT || return 4
1047 [ "$($TEST)" -ne "$i" ] && error "Param didn't stick across restart $($TEST) != $i"
1050 echo Erase parameter setting
1051 do_facet mgs "$LCTL conf_param -d $FSNAME.llite.max_read_ahead_whole_mb" || return 6
1052 umount_client $MOUNT
1053 mount_client $MOUNT || return 6
1055 echo "deleted (default) value=$FINAL, orig=$ORIG"
1056 # assumes this parameter started at the default value
1057 [ "$FINAL" -eq "$ORIG" ] || fail "Deleted value=$FINAL, orig=$ORIG"
1061 run_test 30a "Big config llog and conf_param deletion"
1066 # Make a fake nid. Use the OST nid, and add 20 to the least significant
1067 # numerical part of it. Hopefully that's not already a failover address for
1069 OSTNID=$(do_facet ost1 "$LCTL get_param nis" | tail -1 | awk '{print $1}')
1070 ORIGVAL=$(echo $OSTNID | egrep -oi "[0-9]*@")
1071 NEWVAL=$((($(echo $ORIGVAL | egrep -oi "[0-9]*") + 20) % 256))
1072 NEW=$(echo $OSTNID | sed "s/$ORIGVAL/$NEWVAL@/")
1073 echo "Using fake nid $NEW"
1075 TEST="$LCTL get_param -n osc.$FSNAME-OST0000-osc-[^M]*.import | grep failover_nids | sed -n 's/.*\($NEW\).*/\1/p'"
1076 set_and_check client "$TEST" "$FSNAME-OST0000.failover.node" $NEW || error "didn't add failover nid $NEW"
1077 NIDS=$($LCTL get_param -n osc.$FSNAME-OST0000-osc-[^M]*.import | grep failover_nids)
1079 NIDCOUNT=$(($(echo "$NIDS" | wc -w) - 1))
1080 echo "should have 2 failover nids: $NIDCOUNT"
1081 [ $NIDCOUNT -eq 2 ] || error "Failover nid not added"
1082 do_facet mgs "$LCTL conf_param -d $FSNAME-OST0000.failover.node" || error "conf_param delete failed"
1083 umount_client $MOUNT
1084 mount_client $MOUNT || return 3
1086 NIDS=$($LCTL get_param -n osc.$FSNAME-OST0000-osc-[^M]*.import | grep failover_nids)
1088 NIDCOUNT=$(($(echo "$NIDS" | wc -w) - 1))
1089 echo "only 1 final nid should remain: $NIDCOUNT"
1090 [ $NIDCOUNT -eq 1 ] || error "Failover nids not removed"
1094 run_test 30b "Remove failover nids"
1096 test_31() { # bug 10734
1097 # ipaddr must not exist
1098 mount -t lustre 4.3.2.1@tcp:/lustre $MOUNT || true
1101 run_test 31 "Connect to non-existent node (shouldn't crash)"
1103 # Use these start32/stop32 fn instead of t-f start/stop fn,
1104 # for local devices, to skip global facet vars init
1108 echo "Stopping local ${MOUNT%/*}/${facet} (opts:$@)"
1109 umount -d $@ ${MOUNT%/*}/${facet}
1118 mkdir -p ${MOUNT%/*}/${facet}
1120 echo "Starting local ${facet}: $@ $device ${MOUNT%/*}/${facet}"
1121 mount -t lustre $@ ${device} ${MOUNT%/*}/${facet}
1123 if [ $RC -ne 0 ]; then
1124 echo "mount -t lustre $@ ${device} ${MOUNT%/*}/${facet}"
1125 echo "Start of ${device} of local ${facet} failed ${RC}"
1131 cleanup_nocli32 () {
1139 echo "Cleanup test_32 umount $MOUNT ..."
1140 umount -f $MOUNT || true
1141 echo "Cleanup local mds ost1 ..."
1143 combined_mgs_mds || start_mgs
1148 client_only && skip "client only testing" && return 0
1149 [ "$NETTYPE" = "tcp" ] || { skip "NETTYPE != tcp" && return 0; }
1150 [ -z "$TUNEFS" ] && skip_env "No tunefs" && return 0
1152 local DISK1_8=$LUSTRE/tests/disk1_8.tar.bz2
1153 [ ! -r $DISK1_8 ] && skip_env "Cannot find $DISK1_8" && return 0
1154 local tmpdir=$TMP/conf32a
1156 tar xjvf $DISK1_8 -C $tmpdir || \
1157 { skip_env "Cannot untar $DISK1_8" && return 0; }
1160 $LCTL set_param debug=$PTLDEBUG
1162 $TUNEFS $tmpdir/mds || error "tunefs failed"
1164 combined_mgs_mds || stop mgs
1166 # nids are wrong, so client wont work, but server should start
1167 start32 mds1 $tmpdir/mds "-o loop,exclude=lustre-OST0000" && \
1168 trap cleanup_32 EXIT INT || return 3
1170 local UUID=$($LCTL get_param -n mdt.lustre-MDT0000.uuid)
1172 [ "$UUID" == "lustre-MDT0000_UUID" ] || error "UUID is wrong: $UUID"
1174 $TUNEFS --mgsnode=$HOSTNAME $tmpdir/ost1 || error "tunefs failed"
1175 start32 ost1 $tmpdir/ost1 "-o loop" || return 5
1176 UUID=$($LCTL get_param -n obdfilter.lustre-OST0000.uuid)
1178 [ "$UUID" == "lustre-OST0000_UUID" ] || error "UUID is wrong: $UUID"
1180 local NID=$($LCTL list_nids | head -1)
1182 echo "OSC changes should succeed:"
1183 $LCTL conf_param lustre-OST0000.osc.max_dirty_mb=15 || return 7
1184 $LCTL conf_param lustre-OST0000.failover.node=$NID || return 8
1187 echo "MDC changes should succeed:"
1188 $LCTL conf_param lustre-MDT0000.mdc.max_rpcs_in_flight=9 || return 9
1189 $LCTL conf_param lustre-MDT0000.failover.node=$NID || return 10
1192 echo "LOV changes should succeed:"
1193 $LCTL pool_new lustre.interop || return 11
1194 $LCTL conf_param lustre-MDT0000.lov.stripesize=4M || return 12
1199 # mount a second time to make sure we didnt leave upgrade flag on
1201 $TUNEFS --dryrun $tmpdir/mds || error "tunefs failed"
1203 combined_mgs_mds || stop mgs
1205 start32 mds1 $tmpdir/mds "-o loop,exclude=lustre-OST0000" && \
1206 trap cleanup_32 EXIT INT || return 12
1210 rm -rf $tmpdir || true # true is only for TMP on NFS
1212 run_test 32a "Upgrade from 1.8 (not live)"
1215 client_only && skip "client only testing" && return 0
1216 [ "$NETTYPE" = "tcp" ] || { skip "NETTYPE != tcp" && return 0; }
1217 [ -z "$TUNEFS" ] && skip_env "No tunefs" && return
1219 local DISK1_8=$LUSTRE/tests/disk1_8.tar.bz2
1220 [ ! -r $DISK1_8 ] && skip_env "Cannot find $DISK1_8" && return 0
1221 local tmpdir=$TMP/conf32b
1223 tar xjvf $DISK1_8 -C $tmpdir || \
1224 { skip_env "Cannot untar $DISK1_8" && return ; }
1227 $LCTL set_param debug="config"
1228 local NEWNAME=lustre
1230 # writeconf will cause servers to register with their current nids
1231 $TUNEFS --writeconf --fsname=$NEWNAME $tmpdir/mds || error "tunefs failed"
1232 combined_mgs_mds || stop mgs
1234 start32 mds1 $tmpdir/mds "-o loop" && \
1235 trap cleanup_32 EXIT INT || return 3
1237 local UUID=$($LCTL get_param -n mdt.${NEWNAME}-MDT0000.uuid)
1239 [ "$UUID" == "${NEWNAME}-MDT0000_UUID" ] || error "UUID is wrong: $UUID"
1241 $TUNEFS --mgsnode=$HOSTNAME --writeconf --fsname=$NEWNAME $tmpdir/ost1 ||\
1242 error "tunefs failed"
1243 start32 ost1 $tmpdir/ost1 "-o loop" || return 5
1244 UUID=$($LCTL get_param -n obdfilter.${NEWNAME}-OST0000.uuid)
1246 [ "$UUID" == "${NEWNAME}-OST0000_UUID" ] || error "UUID is wrong: $UUID"
1248 local NID=$($LCTL list_nids | head -1)
1250 echo "OSC changes should succeed:"
1251 $LCTL conf_param ${NEWNAME}-OST0000.osc.max_dirty_mb=15 || return 7
1252 $LCTL conf_param ${NEWNAME}-OST0000.failover.node=$NID || return 8
1255 echo "MDC changes should succeed:"
1256 $LCTL conf_param ${NEWNAME}-MDT0000.mdc.max_rpcs_in_flight=9 || return 9
1257 $LCTL conf_param ${NEWNAME}-MDT0000.failover.node=$NID || return 10
1260 echo "LOV changes should succeed:"
1261 $LCTL pool_new ${NEWNAME}.interop || return 11
1262 $LCTL conf_param ${NEWNAME}-MDT0000.lov.stripesize=4M || return 12
1265 # MDT and OST should have registered with new nids, so we should have
1266 # a fully-functioning client
1267 echo "Check client and old fs contents"
1269 local device=`h2$NETTYPE $HOSTNAME`:/$NEWNAME
1270 echo "Starting local client: $HOSTNAME: $device $MOUNT"
1271 mount -t lustre $device $MOUNT || return 1
1273 local old=$($LCTL get_param -n mdc.*.max_rpcs_in_flight)
1274 local new=$((old + 5))
1275 $LCTL conf_param ${NEWNAME}-MDT0000.mdc.max_rpcs_in_flight=$new
1276 wait_update $HOSTNAME "$LCTL get_param -n mdc.*.max_rpcs_in_flight" $new || return 11
1278 [ "$(cksum $MOUNT/passwd | cut -d' ' -f 1,2)" == "94306271 1478" ] || return 12
1283 rm -rf $tmpdir || true # true is only for TMP on NFS
1285 run_test 32b "Upgrade from 1.8 with writeconf"
1287 test_33a() { # bug 12333, was test_33
1289 local FSNAME2=test-123
1290 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
1292 [ -n "$ost1_HOST" ] && fs2ost_HOST=$ost1_HOST
1294 if [ -z "$fs2ost_DEV" -o -z "$fs2mds_DEV" ]; then
1295 local dev=${SINGLEMDS}_dev
1296 local MDSDEV=${!dev}
1297 is_blkdev $SINGLEMDS $MDSDEV && \
1298 skip_env "mixed loopback and real device not working" && return
1301 combined_mgs_mds || mkfs_opts="$mkfs_opts --nomgs"
1303 local fs2mdsdev=${fs2mds_DEV:-${MDSDEV}_2}
1304 local fs2ostdev=${fs2ost_DEV:-$(ostdevname 1)_2}
1305 add fs2mds $MDS_MKFS_OPTS --mkfsoptions='\"-J size=8\"' --fsname=${FSNAME2} --reformat $fs2mdsdev || exit 10
1306 add fs2ost $OST_MKFS_OPTS --fsname=${FSNAME2} --index=8191 --mgsnode=$MGSNID --reformat $fs2ostdev || exit 10
1308 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS && trap cleanup_24a EXIT INT
1309 start fs2ost $fs2ostdev $OST_MOUNT_OPTS
1310 do_facet $SINGLEMDS "$LCTL conf_param $FSNAME2.sys.timeout=200" || rc=1
1312 mount -t lustre $MGSNID:/${FSNAME2} $MOUNT2 || rc=2
1315 cp /etc/hosts $MOUNT2/ || rc=3
1316 $LFS getstripe $MOUNT2/hosts
1321 rm -rf $MOUNT2 $fs2mdsdev $fs2ostdev
1322 cleanup_nocli || rc=6
1325 run_test 33a "Mount ost with a large index number"
1327 test_33b() { # was test_34
1330 do_facet client dd if=/dev/zero of=$MOUNT/24 bs=1024k count=1
1331 # Drop lock cancelation reply during umount
1332 #define OBD_FAIL_LDLM_CANCEL 0x304
1333 do_facet client lctl set_param fail_loc=0x80000304
1334 #lctl set_param debug=-1
1335 umount_client $MOUNT
1338 run_test 33b "Drop cancel during umount"
1342 do_facet client "sh runmultiop_bg_pause $DIR/file O_c"
1343 manual_umount_client
1345 do_facet client killall -USR1 multiop
1346 if [ $rc -eq 0 ]; then
1347 error "umount not fail!"
1352 run_test 34a "umount with opened file should be fail"
1357 touch $DIR/$tfile || return 1
1358 stop_mds --force || return 2
1360 manual_umount_client --force
1362 if [ $rc -ne 0 ]; then
1363 error "mtab after failed umount - rc $rc"
1369 run_test 34b "force umount with failed mds should be normal"
1373 touch $DIR/$tfile || return 1
1374 stop_ost --force || return 2
1376 manual_umount_client --force
1378 if [ $rc -ne 0 ]; then
1379 error "mtab after failed umount - rc $rc"
1385 run_test 34c "force umount with failed ost should be normal"
1387 test_35a() { # bug 12459
1390 DBG_SAVE="`lctl get_param -n debug`"
1391 lctl set_param debug="ha"
1393 log "Set up a fake failnode for the MDS"
1395 local device=$(do_facet $SINGLEMDS "lctl get_param -n devices" | awk '($3 ~ "mdt" && $4 ~ "MDT") { print $4 }' | head -1)
1396 do_facet $SINGLEMDS $LCTL conf_param ${device}.failover.node=$FAKENID || return 4
1398 log "Wait for RECONNECT_INTERVAL seconds (10s)"
1401 MSG="conf-sanity.sh test_35a `date +%F%kh%Mm%Ss`"
1404 log "Stopping the MDT:"
1405 stop_mds || return 5
1407 df $MOUNT > /dev/null 2>&1 &
1409 log "Restarting the MDT:"
1410 start_mds || return 6
1411 log "Wait for df ($DFPID) ... "
1414 lctl set_param debug="$DBG_SAVE"
1416 # retrieve from the log the first server that the client tried to
1417 # contact after the connection loss
1418 $LCTL dk $TMP/lustre-log-$TESTNAME.log
1419 NEXTCONN=`awk "/${MSG}/ {start = 1;}
1420 /import_select_connection.*$device-mdc.* using connection/ {
1422 if (\\\$NF ~ /$FAKENID/)
1428 }" $TMP/lustre-log-$TESTNAME.log`
1429 [ "$NEXTCONN" != "0" ] && log "The client didn't try to reconnect to the last active server (tried ${NEXTCONN} instead)" && return 7
1431 # remove nid settings
1434 run_test 35a "Reconnect to the last active server first"
1436 test_35b() { # bug 18674
1437 remote_mds || { skip "local MDS" && return 0; }
1441 $LCTL set_param debug="ha"
1443 MSG="conf-sanity.sh test_35b `date +%F%kh%Mm%Ss`"
1446 log "Set up a fake failnode for the MDS"
1448 local device=$(do_facet $SINGLEMDS "$LCTL get_param -n devices" | \
1449 awk '($3 ~ "mdt" && $4 ~ "MDT") { print $4 }' | head -1)
1450 do_facet $SINGLEMDS "$LCTL conf_param ${device}.failover.node=$FAKENID" || \
1453 local at_max_saved=0
1454 # adaptive timeouts may prevent seeing the issue
1455 if at_is_enabled; then
1456 at_max_saved=$(at_max_get mds)
1457 at_max_set 0 mds client
1460 mkdir -p $MOUNT/testdir
1461 touch $MOUNT/testdir/test
1463 log "Injecting EBUSY on MDS"
1464 # Setting OBD_FAIL_MDS_RESEND=0x136
1465 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0x80000136" || return 2
1467 log "Stat on a test file"
1468 stat $MOUNT/testdir/test
1470 log "Stop injecting EBUSY on MDS"
1471 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0" || return 3
1472 rm -f $MOUNT/testdir/test
1475 # restore adaptive timeout
1476 [ $at_max_saved -ne 0 ] && at_max_set $at_max_saved mds client
1478 $LCTL dk $TMP/lustre-log-$TESTNAME.log
1480 # retrieve from the log if the client has ever tried to
1481 # contact the fake server after the loss of connection
1482 FAILCONN=`awk "BEGIN {ret = 0;}
1483 /import_select_connection.*${FSNAME}-MDT0000-mdc.* using connection/ {
1485 if (\\\$NF ~ /$FAKENID/) {
1490 END {print ret}" $TMP/lustre-log-$TESTNAME.log`
1492 [ "$FAILCONN" == "0" ] && \
1493 log "ERROR: The client reconnection has not been triggered" && \
1495 [ "$FAILCONN" == "2" ] && \
1496 log "ERROR: The client tried to reconnect to the failover server while the primary was busy" && \
1500 # remove nid settings
1503 run_test 35b "Continue reconnection retries, if the active server is busy"
1506 [ $OSTCOUNT -lt 2 ] && skip_env "skipping test for single OST" && return
1508 [ "$ost_HOST" = "`hostname`" -o "$ost1_HOST" = "`hostname`" ] || \
1509 { skip "remote OST" && return 0; }
1512 local FSNAME2=test1234
1513 local fs3ost_HOST=$ost_HOST
1514 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
1516 [ -n "$ost1_HOST" ] && fs2ost_HOST=$ost1_HOST && fs3ost_HOST=$ost1_HOST
1518 if [ -z "$fs2ost_DEV" -o -z "$fs2mds_DEV" -o -z "$fs3ost_DEV" ]; then
1519 is_blkdev $SINGLEMDS $MDSDEV && \
1520 skip_env "mixed loopback and real device not working" && return
1523 local fs2mdsdev=${fs2mds_DEV:-${MDSDEV}_2}
1524 local fs2ostdev=${fs2ost_DEV:-$(ostdevname 1)_2}
1525 local fs3ostdev=${fs3ost_DEV:-$(ostdevname 2)_2}
1526 add fs2mds $MDS_MKFS_OPTS --fsname=${FSNAME2} --reformat $fs2mdsdev || exit 10
1527 # XXX after we support non 4K disk blocksize, change following --mkfsoptions with
1529 add fs2ost $OST_MKFS_OPTS --mkfsoptions='-b4096' --fsname=${FSNAME2} --mgsnode=$MGSNID --reformat $fs2ostdev || exit 10
1530 add fs3ost $OST_MKFS_OPTS --mkfsoptions='-b4096' --fsname=${FSNAME2} --mgsnode=$MGSNID --reformat $fs3ostdev || exit 10
1532 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS
1533 start fs2ost $fs2ostdev $OST_MOUNT_OPTS
1534 start fs3ost $fs3ostdev $OST_MOUNT_OPTS
1536 mount -t lustre $MGSNID:/${FSNAME2} $MOUNT2 || return 1
1538 sleep 5 # until 11778 fixed
1540 dd if=/dev/zero of=$MOUNT2/$tfile bs=1M count=7 || return 2
1542 BKTOTAL=`lctl get_param -n obdfilter.*.kbytestotal | awk 'BEGIN{total=0}; {total+=$1}; END{print total}'`
1543 BKFREE=`lctl get_param -n obdfilter.*.kbytesfree | awk 'BEGIN{free=0}; {free+=$1}; END{print free}'`
1544 BKAVAIL=`lctl get_param -n obdfilter.*.kbytesavail | awk 'BEGIN{avail=0}; {avail+=$1}; END{print avail}'`
1545 STRING=`df -P $MOUNT2 | tail -n 1 | awk '{print $2","$3","$4}'`
1546 DFTOTAL=`echo $STRING | cut -d, -f1`
1547 DFUSED=`echo $STRING | cut -d, -f2`
1548 DFAVAIL=`echo $STRING | cut -d, -f3`
1549 DFFREE=$(($DFTOTAL - $DFUSED))
1551 ALLOWANCE=$((64 * $OSTCOUNT))
1553 if [ $DFTOTAL -lt $(($BKTOTAL - $ALLOWANCE)) ] ||
1554 [ $DFTOTAL -gt $(($BKTOTAL + $ALLOWANCE)) ] ; then
1555 echo "**** FAIL: df total($DFTOTAL) mismatch OST total($BKTOTAL)"
1558 if [ $DFFREE -lt $(($BKFREE - $ALLOWANCE)) ] ||
1559 [ $DFFREE -gt $(($BKFREE + $ALLOWANCE)) ] ; then
1560 echo "**** FAIL: df free($DFFREE) mismatch OST free($BKFREE)"
1563 if [ $DFAVAIL -lt $(($BKAVAIL - $ALLOWANCE)) ] ||
1564 [ $DFAVAIL -gt $(($BKAVAIL + $ALLOWANCE)) ] ; then
1565 echo "**** FAIL: df avail($DFAVAIL) mismatch OST avail($BKAVAIL)"
1570 stop fs3ost -f || return 200
1571 stop fs2ost -f || return 201
1572 stop fs2mds -f || return 202
1573 rm -rf $MOUNT2 $fs2mdsdev $fs2ostdev $fs3ostdev
1574 unload_modules_conf || return 203
1577 run_test 36 "df report consistency on OSTs with different block size"
1580 local mntpt=$(facet_mntpt $SINGLEMDS)
1581 local mdsdev=$(mdsdevname ${SINGLEMDS//mds/})
1582 local mdsdev_sym="$TMP/sym_mdt.img"
1584 echo "MDS : $mdsdev"
1585 echo "SYMLINK : $mdsdev_sym"
1586 do_facet $SINGLEMDS rm -f $mdsdev_sym
1588 do_facet $SINGLEMDS ln -s $mdsdev $mdsdev_sym
1590 echo "mount symlink device - $mdsdev_sym"
1593 mount_op=$(do_facet $SINGLEMDS mount -v -t lustre $MDS_MOUNT_OPTS $mdsdev_sym $mntpt 2>&1 )
1596 echo mount_op=$mount_op
1598 do_facet $SINGLEMDS "umount -d $mntpt && rm -f $mdsdev_sym"
1600 if $(echo $mount_op | grep -q "unable to set tunable"); then
1601 error "set tunables failed for symlink device"
1604 [ $rc -eq 0 ] || error "mount symlink $mdsdev_sym failed! rc=$rc"
1608 run_test 37 "verify set tunables works for symlink device"
1610 test_38() { # bug 14222
1615 FILES=`find $SRC -type f -mtime +1 | head -n $COUNT`
1616 log "copying $(echo $FILES | wc -w) files to $DIR/$tdir"
1618 tar cf - $FILES | tar xf - -C $DIR/$tdir || \
1619 error "copying $SRC to $DIR/$tdir"
1621 umount_client $MOUNT
1623 log "rename lov_objid file on MDS"
1624 rm -f $TMP/lov_objid.orig
1626 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
1627 do_facet $SINGLEMDS "$DEBUGFS -c -R \\\"dump lov_objid $TMP/lov_objid.orig\\\" $MDSDEV"
1628 do_facet $SINGLEMDS "$DEBUGFS -w -R \\\"rm lov_objid\\\" $MDSDEV"
1630 do_facet $SINGLEMDS "od -Ax -td8 $TMP/lov_objid.orig"
1631 # check create in mds_lov_connect
1635 [ $V ] && log "verifying $DIR/$tdir/$f"
1636 diff -q $f $DIR/$tdir/$f || ERROR=y
1638 do_facet $SINGLEMDS "$DEBUGFS -c -R \\\"dump lov_objid $TMP/lov_objid.new\\\" $MDSDEV"
1639 do_facet $SINGLEMDS "od -Ax -td8 $TMP/lov_objid.new"
1640 [ "$ERROR" = "y" ] && error "old and new files are different after connect" || true
1642 # check it's updates in sync
1643 umount_client $MOUNT
1646 do_facet $SINGLEMDS dd if=/dev/zero of=$TMP/lov_objid.clear bs=4096 count=1
1647 do_facet $SINGLEMDS "$DEBUGFS -w -R \\\"rm lov_objid\\\" $MDSDEV"
1648 do_facet $SINGLEMDS "$DEBUGFS -w -R \\\"write $TMP/lov_objid.clear lov_objid\\\" $MDSDEV "
1653 [ $V ] && log "verifying $DIR/$tdir/$f"
1654 diff -q $f $DIR/$tdir/$f || ERROR=y
1656 do_facet $SINGLEMDS "$DEBUGFS -c -R \\\"dump lov_objid $TMP/lov_objid.new1\\\" $MDSDEV"
1657 do_facet $SINGLEMDS "od -Ax -td8 $TMP/lov_objid.new1"
1658 umount_client $MOUNT
1660 [ "$ERROR" = "y" ] && error "old and new files are different after sync" || true
1662 log "files compared the same"
1665 run_test 38 "MDS recreates missing lov_objid file from OST data"
1671 perl $SRCDIR/leak_finder.pl $TMP/debug 2>&1 | egrep '*** Leak:' &&
1672 error "memory leak detected" || true
1674 run_test 39 "leak_finder recognizes both LUSTRE and LNET malloc messages"
1676 test_40() { # bug 15759
1678 #define OBD_FAIL_TGT_TOOMANY_THREADS 0x706
1679 do_facet $SINGLEMDS "sysctl -w lustre.fail_loc=0x80000706"
1683 run_test 40 "race during service thread startup"
1685 test_41() { #bug 14134
1687 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
1689 start $SINGLEMDS $MDSDEV $MDS_MOUNT_OPTS -o nosvc -n
1690 start ost1 `ostdevname 1` $OST_MOUNT_OPTS
1691 start $SINGLEMDS $MDSDEV $MDS_MOUNT_OPTS -o nomgs,force
1693 mount_client $MOUNT || return 1
1696 echo "blah blah" > $MOUNT/$tfile
1699 umount_client $MOUNT
1700 stop ost1 -f || return 201
1701 stop_mds -f || return 202
1702 stop_mds -f || return 203
1703 unload_modules_conf || return 204
1706 run_test 41 "mount mds with --nosvc and --nomgs"
1708 test_42() { #bug 14693
1710 check_mount || return 2
1711 do_facet mgs $LCTL conf_param lustre.llite.some_wrong_param=10
1712 umount_client $MOUNT
1713 mount_client $MOUNT || return 1
1717 run_test 42 "invalid config param should not prevent client from mounting"
1720 [ $UID -ne 0 -o $RUNAS_ID -eq 0 ] && skip_env "run as root"
1722 chmod ugo+x $DIR || error "chmod 0 failed"
1724 "lctl get_param -n mdt.$FSNAME-MDT0000.root_squash" \
1725 "$FSNAME.mdt.root_squash" \
1728 "lctl get_param -n mdt.$FSNAME-MDT0000.nosquash_nids" \
1729 "$FSNAME.mdt.nosquash_nids" \
1733 # create set of test files
1735 echo "111" > $DIR/$tfile-userfile || error "write 1 failed"
1736 chmod go-rw $DIR/$tfile-userfile || error "chmod 1 failed"
1737 chown $RUNAS_ID.$RUNAS_ID $DIR/$tfile-userfile || error "chown failed"
1739 echo "222" > $DIR/$tfile-rootfile || error "write 2 failed"
1740 chmod go-rw $DIR/$tfile-rootfile || error "chmod 2 faield"
1742 mkdir $DIR/$tdir-rootdir -p || error "mkdir failed"
1743 chmod go-rwx $DIR/$tdir-rootdir || error "chmod 3 failed"
1744 touch $DIR/$tdir-rootdir/tfile-1 || error "touch failed"
1747 # check root_squash:
1748 # set root squash UID:GID to RUNAS_ID
1749 # root should be able to access only files owned by RUNAS_ID
1752 "lctl get_param -n mdt.$FSNAME-MDT0000.root_squash" \
1753 "$FSNAME.mdt.root_squash" \
1754 "$RUNAS_ID:$RUNAS_ID"
1756 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-userfile)
1757 dd if=$DIR/$tfile-userfile 1>/dev/null 2>/dev/null || \
1758 error "$ST: root read permission is denied"
1759 echo "$ST: root read permission is granted - ok"
1762 dd conv=notrunc if=$DIR/$tfile-userfile 1>/dev/null 2>/dev/null || \
1763 error "$ST: root write permission is denied"
1764 echo "$ST: root write permission is granted - ok"
1766 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-rootfile)
1767 dd if=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null && \
1768 error "$ST: root read permission is granted"
1769 echo "$ST: root read permission is denied - ok"
1772 dd conv=notrunc of=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null && \
1773 error "$ST: root write permission is granted"
1774 echo "$ST: root write permission is denied - ok"
1776 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tdir-rootdir)
1777 rm $DIR/$tdir-rootdir/tfile-1 1>/dev/null 2>/dev/null && \
1778 error "$ST: root unlink permission is granted"
1779 echo "$ST: root unlink permission is denied - ok"
1781 touch $DIR/tdir-rootdir/tfile-2 1>/dev/null 2>/dev/null && \
1782 error "$ST: root create permission is granted"
1783 echo "$ST: root create permission is denied - ok"
1786 # check nosquash_nids:
1787 # put client's NID into nosquash_nids list,
1788 # root should be able to access root file after that
1790 local NIDLIST=$(lctl list_nids all | tr '\n' ' ')
1791 NIDLIST="2@elan $NIDLIST 192.168.0.[2,10]@tcp"
1792 NIDLIST=$(echo $NIDLIST | tr -s ' ' ' ')
1794 "lctl get_param -n mdt.$FSNAME-MDT0000.nosquash_nids" \
1795 "$FSNAME-MDTall.mdt.nosquash_nids" \
1798 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-rootfile)
1799 dd if=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null || \
1800 error "$ST: root read permission is denied"
1801 echo "$ST: root read permission is granted - ok"
1804 dd conv=notrunc of=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null || \
1805 error "$ST: root write permission is denied"
1806 echo "$ST: root write permission is granted - ok"
1808 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tdir-rootdir)
1809 rm $DIR/$tdir-rootdir/tfile-1 || \
1810 error "$ST: root unlink permission is denied"
1811 echo "$ST: root unlink permission is granted - ok"
1812 touch $DIR/$tdir-rootdir/tfile-2 || \
1813 error "$ST: root create permission is denied"
1814 echo "$ST: root create permission is granted - ok"
1818 run_test 43 "check root_squash and nosquash_nids"
1820 umount_client $MOUNT
1825 check_mount || return 2
1826 UUID=$($LCTL get_param llite.${FSNAME}*.uuid | cut -d= -f2)
1828 UUIDS=$(do_facet $SINGLEMDS "$LCTL get_param mdt.${FSNAME}*.exports.*.uuid")
1829 for VAL in $UUIDS; do
1830 NID=$(echo $VAL | cut -d= -f1)
1831 CLUUID=$(echo $VAL | cut -d= -f2)
1832 [ "$UUID" = "$CLUUID" ] && STATS_FOUND=yes && break
1834 [ "$STATS_FOUND" = "no" ] && error "stats not found for client"
1838 run_test 44 "mounted client proc entry exists"
1842 check_mount || return 2
1847 #define OBD_FAIL_PTLRPC_LONG_UNLINK 0x50f
1848 do_facet client "lctl set_param fail_loc=0x50f"
1851 manual_umount_client --force || return 3
1852 do_facet client "lctl set_param fail_loc=0x0"
1854 mount_client $MOUNT || return 4
1858 run_test 45 "long unlink handling in ptlrpcd"
1865 umount_client $MOUNT2 || rc=$?
1866 umount_client $MOUNT || rc=$?
1867 while [ $count -gt 0 ]; do
1868 stop ost${count} -f || rc=$?
1872 cleanup_nocli || rc=$?
1873 #writeconf to remove all ost2 traces for subsequent tests
1879 echo "Testing with $OSTCOUNT OSTs"
1881 start_mds || return 1
1882 #first client should see only one ost
1883 start_ost || return 2
1884 wait_osc_import_state mds ost FULL
1886 mount_client $MOUNT || return 3
1887 trap "cleanup_46a $OSTCOUNT" EXIT ERR
1890 for (( i=2; i<=$OSTCOUNT; i++ )); do
1891 start ost$i `ostdevname $i` $OST_MOUNT_OPTS || return $((i+2))
1894 # wait until osts in sync
1895 for (( i=2; i<=$OSTCOUNT; i++ )); do
1896 wait_osc_import_state mds ost$i FULL
1900 #second client see all ost's
1902 mount_client $MOUNT2 || return 8
1903 $LFS setstripe $MOUNT2 -c -1 || return 9
1904 $LFS getstripe $MOUNT2 || return 10
1906 echo "ok" > $MOUNT2/widestripe
1907 $LFS getstripe $MOUNT2/widestripe || return 11
1908 # fill acl buffer for avoid expand lsm to them
1909 awk -F : '{if (FNR < 25) { print "u:"$1":rwx" }}' /etc/passwd | while read acl; do
1910 setfacl -m $acl $MOUNT2/widestripe
1914 stat $MOUNT/widestripe || return 12
1916 cleanup_46a $OSTCOUNT || { echo "cleanup_46a failed!" && return 13; }
1919 run_test 46a "handle ost additional - wide striped file"
1924 check_mount || return 2
1925 $LCTL set_param ldlm.namespaces.$FSNAME-*-*-*.lru_size=100
1929 for ns in $($LCTL get_param ldlm.namespaces.$FSNAME-*-*-*.lru_size); do
1930 if echo $ns | grep "MDT[[:digit:]]*"; then
1933 lrs=$(echo $ns | sed 's/.*lru_size=//')
1934 lru_size[count]=$lrs
1939 facet_failover $SINGLEMDS
1940 client_up || return 3
1943 for ns in $($LCTL get_param ldlm.namespaces.$FSNAME-*-*-*.lru_size); do
1944 if echo $ns | grep "MDT[[:digit:]]*"; then
1947 lrs=$(echo $ns | sed 's/.*lru_size=//')
1948 if ! test "$lrs" -eq "${lru_size[count]}"; then
1949 n=$(echo $ns | sed -e 's/ldlm.namespaces.//' -e 's/.lru_size=.*//')
1950 error "$n has lost lru_size: $lrs vs. ${lru_size[count]}"
1958 run_test 47 "server restart does not make client loss lru_resize settings"
1963 # reformat after this test is needed - if test will failed
1964 # we will have unkillable file at FS
1968 test_48() { # bug 17636
1971 check_mount || return 2
1973 $LFS setstripe $MOUNT -c -1 || return 9
1974 $LFS getstripe $MOUNT || return 10
1976 echo "ok" > $MOUNT/widestripe
1977 $LFS getstripe $MOUNT/widestripe || return 11
1979 trap cleanup_48 EXIT ERR
1981 # fill acl buffer for avoid expand lsm to them
1982 getent passwd | awk -F : '{ print "u:"$1":rwx" }' | while read acl; do
1983 setfacl -m $acl $MOUNT/widestripe
1986 stat $MOUNT/widestripe || return 12
1991 run_test 48 "too many acls on file"
1993 # check PARAM_SYS_LDLM_TIMEOUT option of MKFS.LUSTRE
1994 test_49() { # bug 17710
1995 local OLD_MDS_MKFS_OPTS=$MDS_MKFS_OPTS
1996 local OLD_OST_MKFS_OPTS=$OST_MKFS_OPTS
1997 local LOCAL_TIMEOUT=20
2000 OST_MKFS_OPTS="--ost --fsname=$FSNAME --device-size=$OSTSIZE --mgsnode=$MGSNID --param sys.timeout=$LOCAL_TIMEOUT --param sys.ldlm_timeout=$LOCAL_TIMEOUT $MKFSOPT $OSTOPT"
2004 check_mount || return 1
2006 echo "check ldlm_timout..."
2007 LDLM_MDS="`do_facet $SINGLEMDS lctl get_param -n ldlm_timeout`"
2008 LDLM_OST1="`do_facet ost1 lctl get_param -n ldlm_timeout`"
2009 LDLM_CLIENT="`do_facet client lctl get_param -n ldlm_timeout`"
2011 if [ $LDLM_MDS -ne $LDLM_OST1 ] || [ $LDLM_MDS -ne $LDLM_CLIENT ]; then
2012 error "Different LDLM_TIMEOUT:$LDLM_MDS $LDLM_OST1 $LDLM_CLIENT"
2015 if [ $LDLM_MDS -ne $((LOCAL_TIMEOUT / 3)) ]; then
2016 error "LDLM_TIMEOUT($LDLM_MDS) is not correct"
2019 umount_client $MOUNT
2020 stop_ost || return 2
2021 stop_mds || return 3
2023 OST_MKFS_OPTS="--ost --fsname=$FSNAME --device-size=$OSTSIZE --mgsnode=$MGSNID --param sys.timeout=$LOCAL_TIMEOUT --param sys.ldlm_timeout=$((LOCAL_TIMEOUT - 1)) $MKFSOPT $OSTOPT"
2027 check_mount || return 7
2029 LDLM_MDS="`do_facet $SINGLEMDS lctl get_param -n ldlm_timeout`"
2030 LDLM_OST1="`do_facet ost1 lctl get_param -n ldlm_timeout`"
2031 LDLM_CLIENT="`do_facet client lctl get_param -n ldlm_timeout`"
2033 if [ $LDLM_MDS -ne $LDLM_OST1 ] || [ $LDLM_MDS -ne $LDLM_CLIENT ]; then
2034 error "Different LDLM_TIMEOUT:$LDLM_MDS $LDLM_OST1 $LDLM_CLIENT"
2037 if [ $LDLM_MDS -ne $((LOCAL_TIMEOUT - 1)) ]; then
2038 error "LDLM_TIMEOUT($LDLM_MDS) is not correct"
2041 cleanup || return $?
2043 MDS_MKFS_OPTS=$OLD_MDS_MKFS_OPTS
2044 OST_MKFS_OPTS=$OLD_OST_MKFS_OPTS
2046 run_test 49 "check PARAM_SYS_LDLM_TIMEOUT option of MKFS.LUSTRE"
2049 # Test both statfs and lfs df and fail if either one fails
2050 multiop_bg_pause $1 f_
2053 killall -USR1 multiop
2054 [ $RC1 -ne 0 ] && log "lazystatfs multiop failed"
2055 wait $PID || { RC1=$?; log "multiop return error "; }
2062 if [ $RC2 -eq 0 ]; then
2064 log "lazystatfs df failed"
2068 [[ $RC1 -ne 0 || $RC2 -eq 0 ]] && RC=1
2074 lctl set_param llite.$FSNAME-*.lazystatfs=1
2077 lazystatfs $MOUNT || error "lazystatfs failed but no down servers"
2079 cleanup || return $?
2081 run_test 50a "lazystatfs all servers available =========================="
2085 lctl set_param llite.$FSNAME-*.lazystatfs=1
2088 # Wait for client to detect down OST
2089 stop_ost || error "Unable to stop OST1"
2090 wait_osc_import_state mds ost DISCONN
2092 lazystatfs $MOUNT || error "lazystatfs should don't have returned EIO"
2094 umount_client $MOUNT || error "Unable to unmount client"
2095 stop_mds || error "Unable to stop MDS"
2097 run_test 50b "lazystatfs all servers down =========================="
2100 start_mds || error "Unable to start MDS"
2101 start_ost || error "Unable to start OST1"
2102 start_ost2 || error "Unable to start OST2"
2103 mount_client $MOUNT || error "Unable to mount client"
2104 lctl set_param llite.$FSNAME-*.lazystatfs=1
2107 # Wait for client to detect down OST
2108 stop_ost || error "Unable to stop OST1"
2109 wait_osc_import_state mds ost DISCONN
2110 lazystatfs $MOUNT || error "lazystatfs failed with one down server"
2112 umount_client $MOUNT || error "Unable to unmount client"
2113 stop_ost2 || error "Unable to stop OST2"
2114 stop_mds || error "Unable to stop MDS"
2116 run_test 50c "lazystatfs one server down =========================="
2119 start_mds || error "Unable to start MDS"
2120 start_ost || error "Unable to start OST1"
2121 start_ost2 || error "Unable to start OST2"
2122 mount_client $MOUNT || error "Unable to mount client"
2123 lctl set_param llite.$FSNAME-*.lazystatfs=1
2126 # Issue the statfs during the window where the client still
2127 # belives the OST to be available but it is in fact down.
2128 # No failure just a statfs which hangs for a timeout interval.
2129 stop_ost || error "Unable to stop OST1"
2130 lazystatfs $MOUNT || error "lazystatfs failed with one down server"
2132 umount_client $MOUNT || error "Unable to unmount client"
2133 stop_ost2 || error "Unable to stop OST2"
2134 stop_mds || error "Unable to stop MDS"
2136 run_test 50d "lazystatfs client/server conn race =========================="
2143 start_mds || return 1
2144 #first client should see only one ost
2145 start_ost || return 2
2146 wait_osc_import_state mds ost FULL
2148 # Wait for client to detect down OST
2149 stop_ost || error "Unable to stop OST1"
2150 wait_osc_import_state mds ost DISCONN
2152 mount_client $MOUNT || error "Unable to mount client"
2153 lctl set_param llite.$FSNAME-*.lazystatfs=0
2155 multiop_bg_pause $MOUNT _f
2159 if [ $RC1 -ne 0 ]; then
2160 log "multiop failed $RC1"
2163 sleep $(( $TIMEOUT+1 ))
2165 [ $? -ne 0 ] && error "process isn't sleep"
2166 start_ost || error "Unable to start OST1"
2167 wait $pid || error "statfs failed"
2170 umount_client $MOUNT || error "Unable to unmount client"
2171 stop_ost || error "Unable to stop OST1"
2172 stop_mds || error "Unable to stop MDS"
2174 run_test 50e "normal statfs all servers down =========================="
2179 CONN_PROC="osc.$FSNAME-OST0001-osc-[M]*.ost_server_uuid"
2181 start_mds || error "Unable to start mds"
2182 #first client should see only one ost
2183 start_ost || error "Unable to start OST1"
2184 wait_osc_import_state mds ost FULL
2186 start_ost2 || error "Unable to start OST2"
2187 wait_osc_import_state mds ost2 FULL
2189 # Wait for client to detect down OST
2190 stop_ost2 || error "Unable to stop OST2"
2192 wait_osc_import_state mds ost2 DISCONN
2193 mount_client $MOUNT || error "Unable to mount client"
2194 lctl set_param llite.$FSNAME-*.lazystatfs=0
2196 multiop_bg_pause $MOUNT _f
2200 if [ $RC1 -ne 0 ]; then
2201 log "lazystatfs multiop failed $RC1"
2204 sleep $(( $TIMEOUT+1 ))
2206 [ $? -ne 0 ] && error "process isn't sleep"
2207 start_ost2 || error "Unable to start OST2"
2208 wait $pid || error "statfs failed"
2209 stop_ost2 || error "Unable to stop OST2"
2212 umount_client $MOUNT || error "Unable to unmount client"
2213 stop_ost || error "Unable to stop OST1"
2214 stop_mds || error "Unable to stop MDS"
2215 #writeconf to remove all ost2 traces for subsequent tests
2218 run_test 50f "normal statfs one server in down =========================="
2221 [ "$OSTCOUNT" -lt "2" ] && skip_env "$OSTCOUNT < 2, skipping" && return
2223 start_ost2 || error "Unable to start OST2"
2224 wait_osc_import_state mds ost2 FULL
2225 wait_osc_import_state client ost2 FULL
2227 local PARAM="${FSNAME}-OST0001.osc.active"
2229 $LFS setstripe -c -1 $DIR/$tfile || error "Unable to lfs setstripe"
2230 do_facet mgs $LCTL conf_param $PARAM=0 || error "Unable to deactivate OST"
2232 umount_client $MOUNT || error "Unable to unmount client"
2233 mount_client $MOUNT || error "Unable to mount client"
2234 # This df should not cause a panic
2237 do_facet mgs $LCTL conf_param $PARAM=1 || error "Unable to activate OST"
2239 umount_client $MOUNT || error "Unable to unmount client"
2240 stop_ost2 || error "Unable to stop OST2"
2241 stop_ost || error "Unable to stop OST1"
2242 stop_mds || error "Unable to stop MDS"
2243 #writeconf to remove all ost2 traces for subsequent tests
2246 run_test 50g "deactivated OST should not cause panic====================="
2249 local LOCAL_TIMEOUT=20
2253 check_mount || return 1
2256 $LFS setstripe -c -1 $MOUNT/d1
2257 #define OBD_FAIL_MDS_REINT_DELAY 0x142
2258 do_facet $SINGLEMDS "lctl set_param fail_loc=0x142"
2259 touch $MOUNT/d1/f1 &
2262 start_ost2 || return 2
2264 stop_ost2 || return 3
2267 run_test 51 "Verify that mdt_reint handles RMF_MDT_MD correctly when an OST is added"
2276 do_node $node mkdir -p $dest
2277 [ $? -eq 0 ] || { error "Unable to create directory"; return 1; }
2279 do_node $node 'tar cf - '$@' | tar xf - -C '$dest';
2280 [ \"\${PIPESTATUS[*]}\" = \"0 0\" ] || exit 1'
2281 [ $? -eq 0 ] || { error "Unable to tar files"; return 2; }
2283 do_node $node 'getfattr -d -m "[a-z]*\\." '$@' > '$xattrs
2284 [ $? -eq 0 ] || { error "Unable to read xattrs"; return 3; }
2294 local backup2=${TMP}/backup2
2296 do_node $node mkdir -p $backup2
2297 [ $? -eq 0 ] || { error "Unable to create directory"; return 1; }
2299 do_node $node 'tar cf - '$@' | tar xf - -C '$backup2';
2300 [ \"\${PIPESTATUS[*]}\" = \"0 0\" ] || exit 1'
2301 [ $? -eq 0 ] || { error "Unable to tar files to diff"; return 2; }
2303 do_node $node "diff -rq $backup $backup2"
2304 [ $? -eq 0 ] || { error "contents differ"; return 3; }
2306 local xattrs2=${TMP}/xattrs2
2307 do_node $node 'getfattr -d -m "[a-z]*\\." '$@' > '$xattrs2
2308 [ $? -eq 0 ] || { error "Unable to read xattrs to diff"; return 4; }
2310 do_node $node "diff $xattrs $xattrs2"
2311 [ $? -eq 0 ] || { error "xattrs differ"; return 5; }
2313 do_node $node "rm -rf $backup2 $xattrs2"
2314 [ $? -eq 0 ] || { error "Unable to delete temporary files"; return 6; }
2319 [ $? -eq 0 ] || { error "Unable to start MDS"; return 1; }
2321 [ $? -eq 0 ] || { error "Unable to start OST1"; return 2; }
2323 [ $? -eq 0 ] || { error "Unable to mount client"; return 3; }
2326 local ost1mnt=${MOUNT%/*}/ost1
2327 local ost1node=$(facet_active_host ost1)
2328 local ost1tmp=$TMP/conf52
2331 [ $? -eq 0 ] || { error "Unable to create tdir"; return 4; }
2332 touch $TMP/modified_first
2333 [ $? -eq 0 ] || { error "Unable to create temporary file"; return 5; }
2334 local mtime=$(stat -c %Y $TMP/modified_first)
2335 do_node $ost1node "mkdir -p $ost1tmp && touch -m -d @$mtime $ost1tmp/modified_first"
2337 [ $? -eq 0 ] || { error "Unable to create temporary file"; return 6; }
2340 $LFS setstripe $DIR/$tdir -c -1 -s 1M
2341 [ $? -eq 0 ] || { error "lfs setstripe failed"; return 7; }
2343 for (( i=0; i < nrfiles; i++ )); do
2344 multiop $DIR/$tdir/$tfile-$i Ow1048576w1048576w524288c
2345 [ $? -eq 0 ] || { error "multiop failed"; return 8; }
2351 echo backup files to $TMP/files
2352 local files=$(find $DIR/$tdir -type f -newer $TMP/modified_first)
2353 copy_files_xattrs `hostname` $TMP/files $TMP/file_xattrs $files
2354 [ $? -eq 0 ] || { error "Unable to copy files"; return 9; }
2356 umount_client $MOUNT
2357 [ $? -eq 0 ] || { error "Unable to umount client"; return 10; }
2359 [ $? -eq 0 ] || { error "Unable to stop ost1"; return 11; }
2361 echo mount ost1 as ldiskfs
2362 do_node $ost1node mount -t $FSTYPE $ost1_dev $ost1mnt $OST_MOUNT_OPTS
2363 [ $? -eq 0 ] || { error "Unable to mount ost1 as ldiskfs"; return 12; }
2366 echo backup objects to $ost1tmp/objects
2367 local objects=$(do_node $ost1node 'find '$ost1mnt'/O/0 -type f -size +0'\
2368 '-newer '$ost1tmp'/modified_first -regex ".*\/[0-9]+"')
2369 copy_files_xattrs $ost1node $ost1tmp/objects $ost1tmp/object_xattrs $objects
2370 [ $? -eq 0 ] || { error "Unable to copy objects"; return 13; }
2372 # move objects to lost+found
2373 do_node $ost1node 'mv '$objects' '${ost1mnt}'/lost+found'
2374 [ $? -eq 0 ] || { error "Unable to move objects"; return 14; }
2377 do_node $ost1node "ll_recover_lost_found_objs -d $ost1mnt/lost+found"
2378 [ $? -eq 0 ] || { error "ll_recover_lost_found_objs failed"; return 15; }
2380 # compare restored objects against saved ones
2381 diff_files_xattrs $ost1node $ost1tmp/objects $ost1tmp/object_xattrs $objects
2382 [ $? -eq 0 ] || { error "Unable to diff objects"; return 16; }
2384 do_node $ost1node "umount $ost1_dev"
2385 [ $? -eq 0 ] || { error "Unable to umount ost1 as ldiskfs"; return 17; }
2388 [ $? -eq 0 ] || { error "Unable to start ost1"; return 18; }
2390 [ $? -eq 0 ] || { error "Unable to mount client"; return 19; }
2393 diff_files_xattrs `hostname` $TMP/files $TMP/file_xattrs $files
2394 [ $? -eq 0 ] || { error "Unable to diff files"; return 20; }
2396 rm -rf $TMP/files $TMP/file_xattrs
2397 [ $? -eq 0 ] || { error "Unable to delete temporary files"; return 21; }
2398 do_node $ost1node "rm -rf $ost1tmp"
2399 [ $? -eq 0 ] || { error "Unable to delete temporary files"; return 22; }
2402 run_test 52 "check recovering objects from lost+found"
2404 # Checks threads_min/max/started for some service
2406 # Arguments: service name (OST or MDT), facet (e.g., ost1, $SINGLEMDS), and a
2407 # parameter pattern prefix like 'ost.*.ost'.
2419 local msg="Insane $modname thread counts"
2423 check_mount || return 41
2425 # We need to expand $parampat, but it may match multiple parameters, so
2426 # we'll pick the first one
2427 if ! paramp=$(do_facet $facet "lctl get_param -N ${parampat}.threads_min"|head -1); then
2428 error "Couldn't expand ${parampat}.threads_min parameter name"
2432 # Remove the .threads_min part
2433 paramp=${paramp%.threads_min}
2435 # Check for sanity in defaults
2436 tmin=$(do_facet $facet "lctl get_param -n ${paramp}.threads_min" || echo 0)
2437 tmax=$(do_facet $facet "lctl get_param -n ${paramp}.threads_max" || echo 0)
2438 tstarted=$(do_facet $facet "lctl get_param -n ${paramp}.threads_started" || echo 0)
2439 lassert 23 "$msg (PDSH problems?)" '(($tstarted && $tmin && $tmax))' || return $?
2440 lassert 24 "$msg" '(($tstarted >= $tmin && $tstarted <= $tmax ))' || return $?
2442 # Check that we can lower min/max
2443 do_facet $facet "lctl set_param ${paramp}.threads_min=$((tmin - 1))"
2444 do_facet $facet "lctl set_param ${paramp}.threads_max=$((tmax - 1))"
2445 tmin2=$(do_facet $facet "lctl get_param -n ${paramp}.threads_min" || echo 0)
2446 tmax2=$(do_facet $facet "lctl get_param -n ${paramp}.threads_max" || echo 0)
2447 lassert 25 "$msg" '(($tmin2 == ($tmin - 1) && $tmax2 == ($tmax -1)))' || return $?
2449 # Check that we can set min/max to the same value
2450 do_facet $facet "lctl set_param ${paramp}.threads_max=$((tmin - 1))"
2451 tmin2=$(do_facet $facet "lctl get_param -n ${paramp}.threads_min" || echo 0)
2452 tmax2=$(do_facet $facet "lctl get_param -n ${paramp}.threads_max" || echo 0)
2453 lassert 26 "$msg" '(($tmin2 == ($tmin - 1) && $tmax2 == ($tmin - 1)))' || return $?
2455 # Check that we can't set max < min
2456 do_facet $facet "lctl set_param ${paramp}.threads_max=$((tmin - 2))"
2457 tmin2=$(do_facet $facet "lctl get_param -n ${paramp}.threads_min" || echo 0)
2458 tmax2=$(do_facet $facet "lctl get_param -n ${paramp}.threads_max" || echo 0)
2459 lassert 27 "$msg" '(($tmin2 <= $tmax2))' || return $?
2461 # We need to ensure that we get the module options desired; to do this
2462 # we set LOAD_MODULES_REMOTE=true and we call setmodopts below.
2463 LOAD_MODULES_REMOTE=true
2466 setmodopts -a $modname "$opts" oldvalue
2470 check_mount || return 41
2472 # Restore previous setting of MODOPTS_*
2473 setmodopts $modname "$oldvalue"
2475 # Check that $opts took
2476 tmin=$(do_facet $facet "lctl get_param -n ${paramp}.threads_min")
2477 tmax=$(do_facet $facet "lctl get_param -n ${paramp}.threads_max")
2478 tstarted=$(do_facet $facet "lctl get_param -n ${paramp}.threads_started")
2479 lassert 28 "$msg" '(($tstarted == $tmin && $tstarted == $tmax ))' || return $?
2482 # Workaround a YALA bug where YALA expects that modules will remain
2483 # loaded on the servers
2484 LOAD_MODULES_REMOTE=false
2491 thread_sanity OST ost1 'ost.*.ost' 'oss_num_threads=64'
2493 run_test 53a "check OSS thread count params"
2496 thread_sanity MDT $SINGLEMDS 'mdt.*.*.' 'mdt_num_threads=64'
2498 run_test 53b "check MDT thread count params"
2503 local partial_arg=""
2504 local size=$(df -B G $dir | tail -1 | awk '{print $2}' | sed 's/G//') # Gb
2506 # Run in partial (fast) mode if the size
2507 # of a partition > 10 GB
2508 [ $size -gt 10 ] && partial_arg="-p"
2510 llverfs $partial_arg $dir
2514 do_rpc_nodes $(facet_host ost1) run_llverdev $(ostdevname 1)
2515 [ $? -eq 0 ] || error "llverdev failed!"
2518 run_test 54a "llverdev"
2523 [ $? -eq 0 ] || error "llverfs failed!"
2526 run_test 54b "llverfs"
2530 local max_ost_index=$1
2531 echo -n $(((max_ost_index + 1) * 8))
2535 local mdsdev=$(mdsdevname 1)
2536 local ostdev=$(ostdevname 1)
2537 local saved_opts=$OST_MKFS_OPTS
2539 for i in 0 1023 2048
2541 OST_MKFS_OPTS="$saved_opts --index $i"
2549 echo checking size of lov_objid for ost index $i
2550 LOV_OBJID_SIZE=$(do_facet mds1 "$DEBUGFS -R 'stat lov_objid' $mdsdev 2>/dev/null" | grep ^User | awk '{print $6}')
2551 if [ "$LOV_OBJID_SIZE" != $(lov_objid_size $i) ]; then
2552 error "lov_objid size has to be $(lov_objid_size $i), not $LOV_OBJID_SIZE"
2554 echo ok, lov_objid size is correct: $LOV_OBJID_SIZE
2559 OST_MKFS_OPTS=$saved_opts
2562 run_test 55 "check lov_objid size"
2565 add mds1 $MDS_MKFS_OPTS --mkfsoptions='\"-J size=16\"' --reformat $(mdsdevname 1)
2566 add ost1 $OST_MKFS_OPTS --index=1000 --reformat $(ostdevname 1)
2567 add ost2 $OST_MKFS_OPTS --index=10000 --reformat $(ostdevname 2)
2571 start_ost2 || error "Unable to start second ost"
2572 mount_client $MOUNT || error "Unable to mount client"
2575 [ -n "$ENABLE_QUOTA" ] && { $LFS quotacheck -ug $MOUNT || error "quotacheck has failed" ; }
2579 run_test 56 "check big indexes"
2581 test_57() { # bug 22656
2582 local NID=$(do_facet ost1 "$LCTL get_param nis" | tail -1 | awk '{print $1}')
2584 do_facet ost1 "$TUNEFS --failnode=$NID `ostdevname 1`" || error "tunefs failed"
2586 start_ost && error "OST registration from failnode should fail"
2590 run_test 57 "initial registration from failnode should fail (should return errs)"
2593 do_facet mgs $LCTL get_param mgs.MGS.live.$FSNAME | grep OST | wc -l
2597 start_mgsmds >> /dev/null
2598 local C1=$(count_osts)
2599 if [ $C1 -eq 0 ]; then
2600 start_ost >> /dev/null
2604 echo "original ost count: $C1 (expect > 0)"
2605 [ $C1 -gt 0 ] || error "No OSTs in $FSNAME log"
2606 start_mgsmds -o writeconf >> /dev/null || error "MDT start failed"
2607 local C2=$(count_osts)
2608 echo "after mdt writeconf count: $C2 (expect 0)"
2609 [ $C2 -gt 0 ] && error "MDT writeconf should erase OST logs"
2610 echo "OST start without writeconf should fail:"
2611 start_ost >> /dev/null && error "OST start without writeconf didn't fail"
2612 echo "OST start with writeconf should succeed:"
2613 start_ost -o writeconf >> /dev/null || error "OST1 start failed"
2614 local C3=$(count_osts)
2615 echo "after ost writeconf count: $C3 (expect 1)"
2616 [ $C3 -eq 1 ] || error "new OST writeconf should add:"
2617 start_ost2 -o writeconf >> /dev/null || error "OST2 start failed"
2618 local C4=$(count_osts)
2619 echo "after ost2 writeconf count: $C4 (expect 2)"
2620 [ $C4 -eq 2 ] || error "OST2 writeconf should add log"
2621 stop_ost2 >> /dev/null
2622 cleanup_nocli >> /dev/null
2624 run_test 59 "writeconf mount option"
2627 test_58() { # bug 22658
2628 [ "$FSTYPE" != "ldiskfs" ] && skip "not supported for $FSTYPE" && return
2631 createmany -o $DIR/$tdir/$tfile-%d 100
2632 # make sure that OSTs do not cancel llog cookies before we unmount the MDS
2633 #define OBD_FAIL_OBD_LOG_CANCEL_NET 0x601
2634 do_facet mds "lctl set_param fail_loc=0x601"
2635 unlinkmany $DIR/$tdir/$tfile-%d 100
2637 local MNTDIR=$(facet_mntpt mds)
2638 # remove all files from the OBJECTS dir
2639 do_facet mds "mount -t ldiskfs $MDSDEV $MNTDIR"
2640 do_facet mds "find $MNTDIR/OBJECTS -type f -delete"
2641 do_facet mds "umount $MNTDIR"
2642 # restart MDS with missing llog files
2644 do_facet mds "lctl set_param fail_loc=0"
2647 run_test 58 "missing llog files must not prevent MDT from mounting"
2649 if ! combined_mgs_mds ; then
2653 equals_msg `basename $0`: test complete
2654 [ -f "$TESTSUITELOG" ] && cat $TESTSUITELOG && grep -q FAIL $TESTSUITELOG && exit 1 || true