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/$NAME.sh}
42 # STORED_MDSSIZE is used in test_18
43 if [ -n "$MDSSIZE" ]; then
44 STORED_MDSSIZE=$MDSSIZE
50 require_dsh_mds || exit 0
51 require_dsh_ost || exit 0
53 [ "$SLOW" = "no" ] && EXCEPT_SLOW="30a 31 45"
68 # who knows if/where $TUNEFS is installed? Better reformat if it fails...
69 do_facet ${facet} "$TUNEFS --quiet --writeconf $dev" ||
70 { echo "tunefs failed, reformatting instead" && reformat_and_config && return 1; }
77 # if writeconf fails anywhere, we reformat everything
78 writeconf1 mds `mdsdevname 1` || return 0
79 writeconf1 ost1 `ostdevname 1` || return 0
80 writeconf1 ost2 `ostdevname 2` || return 0
84 # The MGS must be started before the OSTs for a new fs, so start
85 # and stop to generate the startup logs.
88 wait_osc_import_state mds ost FULL
93 reformat_and_config() {
95 if ! combined_mgs_mds ; then
103 start mgs $MGSDEV $MGS_MOUNT_OPTS
107 local facet=$SINGLEMDS
108 # we can not use MDSDEV1 here because SINGLEMDS could be set not to mds1 only
109 local num=$(echo $facet | tr -d "mds")
110 local dev=$(mdsdevname $num)
111 echo "start mds service on `facet_active_host $facet`"
112 start $facet ${dev} $MDS_MOUNT_OPTS $@ || return 94
116 if ! combined_mgs_mds ; then
123 echo "stop mds service on `facet_active_host $SINGLEMDS`"
124 # These tests all use non-failover stop
125 stop $SINGLEMDS -f || return 97
129 echo "start ost1 service on `facet_active_host ost1`"
130 start ost1 `ostdevname 1` $OST_MOUNT_OPTS $@ || return 95
134 echo "stop ost1 service on `facet_active_host ost1`"
135 # These tests all use non-failover stop
136 stop ost1 -f || return 98
140 echo "start ost2 service on `facet_active_host ost2`"
141 start ost2 `ostdevname 2` $OST_MOUNT_OPTS $@ || return 92
145 echo "stop ost2 service on `facet_active_host ost2`"
146 # These tests all use non-failover stop
147 stop ost2 -f || return 93
152 echo "mount $FSNAME on ${MOUNTPATH}....."
153 zconf_mount `hostname` $MOUNTPATH || return 96
157 local SAVEMOUNTOPT=$MOUNTOPT
158 MOUNTOPT="remount,$1"
160 echo "remount '$1' lustre on ${MOUNTPATH}....."
161 zconf_mount `hostname` $MOUNTPATH || return 96
162 MOUNTOPT=$SAVEMOUNTOPT
167 echo "umount lustre on ${MOUNTPATH}....."
168 zconf_umount `hostname` $MOUNTPATH || return 97
171 manual_umount_client(){
174 echo "manual umount lustre on ${MOUNT}...."
175 do_facet client "umount -d ${FORCE} $MOUNT"
181 start_mds || error "MDT start failed"
182 start_ost || error "OST start failed"
183 mount_client $MOUNT || error "client start failed"
187 if ! combined_mgs_mds ; then
196 unload_modules_conf () {
197 if combined_mgs_mds || ! local_mode; then
198 unload_modules || return 1
203 stop_ost || return 202
204 stop_mds || return 201
205 unload_modules_conf || return 203
209 umount_client $MOUNT || return 200
210 cleanup_nocli || return $?
214 do_facet client "cp /etc/passwd $DIR/a" || return 71
215 do_facet client "rm $DIR/a" || return 72
216 # make sure lustre is actually mounted (touch will block,
217 # but grep won't, so do it after)
218 do_facet client "grep $MOUNT' ' /proc/mounts > /dev/null" || return 73
219 echo "setup single mount lustre success"
223 do_facet client "touch $DIR/a" || return 71
224 do_facet client "rm $DIR/a" || return 72
225 do_facet client "touch $DIR2/a" || return 73
226 do_facet client "rm $DIR2/a" || return 74
227 echo "setup double mount lustre success"
232 if [ "$ONLY" == "setup" ]; then
237 if [ "$ONLY" == "cleanup" ]; then
244 #create single point mountpoint
250 check_mount || return 41
253 run_test 0 "single mount setup"
256 start_mds || error "MDT start failed"
258 echo "start ost second time..."
259 start_ost && error "2nd OST start should fail"
260 mount_client $MOUNT || error "client start failed"
261 check_mount || return 42
264 run_test 1 "start up ost twice (should return errors)"
268 echo "start mds second time.."
269 start_mds && error "2nd MDT start should fail"
272 check_mount || return 43
275 run_test 2 "start up mds twice (should return err)"
279 #mount.lustre returns an error if already in mtab
280 mount_client $MOUNT && error "2nd client mount should fail"
281 check_mount || return 44
284 run_test 3 "mount client twice (should return err)"
288 touch $DIR/$tfile || return 85
292 # ok for ost to fail shutdown
293 if [ 202 -ne $eno ]; then
298 run_test 4 "force cleanup ost, then cleanup"
300 test_5a() { # was test_5
302 touch $DIR/$tfile || return 1
303 fuser -m -v $MOUNT && echo "$MOUNT is in use by user space process."
305 stop_mds -f || return 2
307 # cleanup may return an error from the failed
308 # disconnects; for now I'll consider this successful
309 # if all the modules have unloaded.
313 echo "killing umount"
314 kill -TERM $UMOUNT_PID
315 echo "waiting for umount to finish"
317 if grep " $MOUNT " /proc/mounts; then
318 echo "test 5: /proc/mounts after failed umount"
322 echo "killing umount"
323 kill -TERM $UMOUNT_PID
324 echo "waiting for umount to finish"
326 grep " $MOUNT " /proc/mounts && echo "test 5: /proc/mounts after second umount" && return 11
330 # stop_mds is a no-op here, and should not fail
331 cleanup_nocli || return $?
332 # df may have lingering entry
334 # mtab may have lingering entry
338 while [ "$WAIT" -ne "$MAX_WAIT" ]; do
340 grep -q $MOUNT" " /etc/mtab || break
341 echo "Waiting /etc/mtab updated ... "
342 WAIT=$(( WAIT + sleep))
344 [ "$WAIT" -eq "$MAX_WAIT" ] && error "/etc/mtab is not updated in $WAIT secs"
345 echo "/etc/mtab updated in $WAIT secs"
347 run_test 5a "force cleanup mds, then cleanup"
355 grep " $MOUNT " /etc/mtab && \
356 error false "unexpected entry in mtab before mount" && return 10
360 if ! combined_mgs_mds ; then
361 trap cleanup_5b EXIT ERR
366 [ -d $MOUNT ] || mkdir -p $MOUNT
367 mount_client $MOUNT && rc=1
368 grep " $MOUNT " /etc/mtab && \
369 error "$MOUNT entry in mtab after failed mount" && rc=11
371 # stop_mds is a no-op here, and should not fail
372 cleanup_nocli || rc=$?
373 if ! combined_mgs_mds ; then
378 run_test 5b "Try to start a client with no MGS (should return errs)"
381 grep " $MOUNT " /etc/mtab && \
382 error false "unexpected entry in mtab before mount" && return 10
387 [ -d $MOUNT ] || mkdir -p $MOUNT
388 local oldfs="${FSNAME}"
389 FSNAME="wrong.${FSNAME}"
390 mount_client $MOUNT || :
392 grep " $MOUNT " /etc/mtab && \
393 error "$MOUNT entry in mtab after failed mount" && rc=11
395 cleanup_nocli || rc=$?
398 run_test 5c "cleanup after failed mount (bug 2712) (should return errs)"
401 grep " $MOUNT " /etc/mtab && \
402 error false "unexpected entry in mtab before mount" && return 10
408 mount_client $MOUNT || rc=1
410 grep " $MOUNT " /etc/mtab && \
411 error "$MOUNT entry in mtab after unmount" && rc=11
414 run_test 5d "mount with ost down"
417 grep " $MOUNT " /etc/mtab && \
418 error false "unexpected entry in mtab before mount" && return 10
424 #define OBD_FAIL_PTLRPC_DELAY_SEND 0x506
425 do_facet client "lctl set_param fail_loc=0x80000506"
426 mount_client $MOUNT || echo "mount failed (not fatal)"
428 grep " $MOUNT " /etc/mtab && \
429 error "$MOUNT entry in mtab after unmount" && rc=11
432 run_test 5e "delayed connect, don't crash (bug 10268)"
435 if combined_mgs_mds ; then
436 skip "combined mgs and mds"
440 grep " $MOUNT " /etc/mtab && \
441 error false "unexpected entry in mtab before mount" && return 10
445 [ -d $MOUNT ] || mkdir -p $MOUNT
446 mount_client $MOUNT &
448 echo client_mount pid is $pid
452 if ! ps -f -p $pid >/dev/null; then
455 grep " $MOUNT " /etc/mtab && echo "test 5f: mtab after mount"
456 error "mount returns $rc, expected to hang"
465 # mount should succeed after start mds
468 [ $rc -eq 0 ] || error "mount returned $rc"
469 grep " $MOUNT " /etc/mtab && echo "test 5f: mtab after mount"
473 run_test 5f "mds down, cleanup after failed mount (bug 2712)"
478 mount_client ${MOUNT} || return 87
479 touch $DIR/a || return 86
482 run_test 6 "manual umount, then mount again"
487 cleanup_nocli || return $?
489 run_test 7 "manual umount, then cleanup"
494 check_mount2 || return 45
495 umount_client $MOUNT2
498 run_test 8 "double mount setup"
503 do_facet ost1 lctl set_param debug=\'inode trace\' || return 1
504 do_facet ost1 lctl set_param subsystem_debug=\'mds ost\' || return 1
506 CHECK_PTLDEBUG="`do_facet ost1 lctl get_param -n debug`"
507 if [ "$CHECK_PTLDEBUG" ] && { \
508 [ "$CHECK_PTLDEBUG" = "trace inode warning error emerg console" ] ||
509 [ "$CHECK_PTLDEBUG" = "trace inode" ]; }; then
510 echo "lnet.debug success"
512 echo "lnet.debug: want 'trace inode', have '$CHECK_PTLDEBUG'"
515 CHECK_SUBSYS="`do_facet ost1 lctl get_param -n subsystem_debug`"
516 if [ "$CHECK_SUBSYS" ] && [ "$CHECK_SUBSYS" = "mds ost" ]; then
517 echo "lnet.subsystem_debug success"
519 echo "lnet.subsystem_debug: want 'mds ost', have '$CHECK_SUBSYS'"
522 stop_ost || return $?
524 run_test 9 "test ptldebug and subsystem for mkfs"
532 do_facet $facet "test -b $dev" || rc=1
533 if [[ "$size" ]]; then
534 local in=$(do_facet $facet "dd if=$dev of=/dev/null bs=1k count=1 skip=$size 2>&1" |\
535 awk '($3 == "in") { print $1 }')
536 [[ $in = "1+0" ]] || rc=1
542 # Test 16 was to "verify that lustre will correct the mode of OBJECTS".
543 # But with new MDS stack we don't care about the mode of local objects
544 # anymore, so this test is removed. See bug 22944 for more details.
549 check_mount || return 41
552 echo "Remove mds config log"
553 if ! combined_mgs_mds ; then
557 do_facet mgs "$DEBUGFS -w -R 'unlink CONFIGS/$FSNAME-MDT0000' $MGSDEV || return \$?" || return $?
559 if ! combined_mgs_mds ; then
564 start_mds && return 42
567 run_test 17 "Verify failed mds_postsetup won't fail assertion (2936) (should return errs)"
570 [ "$FSTYPE" != "ldiskfs" ] && skip "not needed for FSTYPE=$FSTYPE" && return
572 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
577 # check if current MDSSIZE is large enough
578 [ $MDSSIZE -ge $MIN ] && OK=1 && myMDSSIZE=$MDSSIZE && \
579 log "use MDSSIZE=$MDSSIZE"
581 # check if the global config has a large enough MDSSIZE
582 [ -z "$OK" -a ! -z "$STORED_MDSSIZE" ] && [ $STORED_MDSSIZE -ge $MIN ] && \
583 OK=1 && myMDSSIZE=$STORED_MDSSIZE && \
584 log "use STORED_MDSSIZE=$STORED_MDSSIZE"
586 # check if the block device is large enough
587 [ -z "$OK" ] && $(is_blkdev $SINGLEMDS $MDSDEV $MIN) && OK=1 &&
588 myMDSSIZE=$MIN && log "use device $MDSDEV with MIN=$MIN"
590 # check if a loopback device has enough space for fs metadata (5%)
592 if [ -z "$OK" ]; then
593 local SPACE=$(do_facet $SINGLEMDS "[ -f $MDSDEV -o ! -e $MDSDEV ] && df -P \\\$(dirname $MDSDEV)" |
594 awk '($1 != "Filesystem") {print $4}')
595 ! [ -z "$SPACE" ] && [ $SPACE -gt $((MIN / 20)) ] && \
596 OK=1 && myMDSSIZE=$MIN && \
597 log "use file $MDSDEV with MIN=$MIN"
600 [ -z "$OK" ] && skip_env "$MDSDEV too small for ${MIN}kB MDS" && return
603 echo "mount mds with large journal..."
604 local OLD_MDS_MKFS_OPTS=$MDS_MKFS_OPTS
606 local opts="--mdt --fsname=$FSNAME --device-size=$myMDSSIZE --param sys.timeout=$TIMEOUT $MDSOPT"
608 if combined_mgs_mds ; then
609 MDS_MKFS_OPTS="--mgs $opts"
611 MDS_MKFS_OPTS="--mgsnode=$MGSNID $opts"
615 echo "mount lustre system..."
617 check_mount || return 41
619 echo "check journal size..."
620 local FOUNDSIZE=$(do_facet $SINGLEMDS "$DEBUGFS -c -R 'stat <8>' $MDSDEV" | awk '/Size: / { print $NF; exit;}')
621 if [ $FOUNDSIZE -gt $((32 * 1024 * 1024)) ]; then
622 log "Success: mkfs creates large journals. Size: $((FOUNDSIZE >> 20))M"
624 error "expected journal size > 32M, found $((FOUNDSIZE >> 20))M"
629 MDS_MKFS_OPTS=$OLD_MDS_MKFS_OPTS
632 run_test 18 "check mkfs creates large journals"
635 start_mds || return 1
636 stop_mds -f || return 2
638 run_test 19a "start/stop MDS without OSTs"
641 start_ost || return 1
642 stop_ost -f || return 2
644 run_test 19b "start/stop OSTs without MDS"
647 # first format the ost/mdt
651 check_mount || return 43
653 remount_client ro $MOUNT || return 44
654 touch $DIR/$tfile && echo "$DIR/$tfile created incorrectly" && return 45
655 [ -e $DIR/$tfile ] && echo "$DIR/$tfile exists incorrectly" && return 46
656 remount_client rw $MOUNT || return 47
658 [ ! -f $DIR/$tfile ] && echo "$DIR/$tfile missing" && return 48
659 MCNT=`grep -c $MOUNT /etc/mtab`
660 [ "$MCNT" -ne 1 ] && echo "$MOUNT in /etc/mtab $MCNT times" && return 49
665 run_test 20 "remount ro,rw mounts work and doesn't break /etc/mtab"
670 wait_osc_import_state mds ost FULL
674 run_test 21a "start mds before ost, stop ost first"
679 wait_osc_import_state mds ost FULL
683 run_test 21b "start ost before mds, stop mds first"
689 wait_osc_import_state mds ost2 FULL
693 #writeconf to remove all ost2 traces for subsequent tests
696 run_test 21c "start mds between two osts, stop mds last"
701 echo Client mount with ost in logs, but none running
703 # wait until mds connected to ost and open client connection
704 wait_osc_import_state mds ost FULL
707 # check_mount will block trying to contact ost
708 mcreate $DIR/$tfile || return 40
709 rm -f $DIR/$tfile || return 42
713 echo Client mount with a running ost
716 # if gss enabled, wait full time to let connection from
717 # mds to ost be established, due to the mismatch between
718 # initial connect timeout and gss context negotiation timeout.
719 # This perhaps could be remove after AT landed.
720 echo "sleep $((TIMEOUT + TIMEOUT + TIMEOUT))s"
721 sleep $((TIMEOUT + TIMEOUT + TIMEOUT))
724 check_mount || return 41
729 run_test 22 "start a client before osts (should return errs)"
731 test_23a() { # was test_23
735 # force down client so that recovering mds waits for reconnect
736 local running=$(grep -c $MOUNT /proc/mounts) || true
737 if [ $running -ne 0 ]; then
738 echo "Stopping client $MOUNT (opts: -f)"
742 # enter recovery on mds
744 # try to start a new client
745 mount_client $MOUNT &
747 MOUNT_PID=$(ps -ef | grep "t lustre" | grep -v grep | awk '{print $2}')
748 MOUNT_LUSTRE_PID=`ps -ef | grep mount.lustre | grep -v grep | awk '{print $2}'`
749 echo mount pid is ${MOUNT_PID}, mount.lustre pid is ${MOUNT_LUSTRE_PID}
751 ps --ppid $MOUNT_LUSTRE_PID
752 # FIXME why o why can't I kill these? Manual "ctrl-c" works...
753 kill -TERM $MOUNT_LUSTRE_PID
754 echo "waiting for mount to finish"
756 # we can not wait $MOUNT_PID because it is not a child of this shell
762 while [ "$WAIT" -lt "$MAX_WAIT" ]; do
764 PID1=$(ps -ef | awk '{print $2}' | grep -w $MOUNT_PID)
765 PID2=$(ps -ef | awk '{print $2}' | grep -w $MOUNT_LUSTRE_PID)
768 [ -z "$PID1" -a -z "$PID2" ] && break
769 echo "waiting for mount to finish ... "
770 WAIT=$(( WAIT + sleep))
772 if [ "$WAIT" -eq "$MAX_WAIT" ]; then
773 error "MOUNT_PID $MOUNT_PID and "\
774 "MOUNT_LUSTRE_PID $MOUNT_LUSTRE_PID still not killed in $WAIT secs"
780 run_test 23a "interrupt client during recovery mount delay"
785 test_23b() { # was test_23
788 # Simulate -EINTR during mount OBD_FAIL_LDLM_CLOSE_THREAD
789 lctl set_param fail_loc=0x80000313
793 run_test 23b "Simulate -EINTR during mount"
795 fs2mds_HOST=$mds_HOST
796 fs2ost_HOST=$ost_HOST
800 echo "umount $MOUNT2 ..."
801 umount $MOUNT2 || true
802 echo "stopping fs2mds ..."
803 stop fs2mds -f || true
804 echo "stopping fs2ost ..."
805 stop fs2ost -f || true
809 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
811 if [ -z "$fs2ost_DEV" -o -z "$fs2mds_DEV" ]; then
812 is_blkdev $SINGLEMDS $MDSDEV && \
813 skip_env "mixed loopback and real device not working" && return
816 [ -n "$ost1_HOST" ] && fs2ost_HOST=$ost1_HOST
818 local fs2mdsdev=${fs2mds_DEV:-${MDSDEV}_2}
819 local fs2ostdev=${fs2ost_DEV:-$(ostdevname 1)_2}
821 # test 8-char fsname as well
822 local FSNAME2=test1234
823 add fs2mds $MDS_MKFS_OPTS --fsname=${FSNAME2} --nomgs --mgsnode=$MGSNID --reformat $fs2mdsdev || exit 10
825 add fs2ost $OST_MKFS_OPTS --fsname=${FSNAME2} --reformat $fs2ostdev || exit 10
828 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS && trap cleanup_24a EXIT INT
829 start fs2ost $fs2ostdev $OST_MOUNT_OPTS
831 mount -t lustre $MGSNID:/${FSNAME2} $MOUNT2 || return 1
833 check_mount || return 2
834 # files written on 1 should not show up on 2
835 cp /etc/passwd $DIR/$tfile
837 [ -e $MOUNT2/$tfile ] && error "File bleed" && return 7
840 cp /etc/passwd $MOUNT2/b || return 3
841 rm $MOUNT2/b || return 4
842 # 2 is actually mounted
843 grep $MOUNT2' ' /proc/mounts > /dev/null || return 5
845 facet_failover fs2mds
846 facet_failover fs2ost
849 # the MDS must remain up until last MDT
851 MDS=$(do_facet $SINGLEMDS "lctl get_param -n devices" | awk '($3 ~ "mdt" && $4 ~ "MDT") { print $4 }' | head -1)
852 [ -z "$MDS" ] && error "No MDT" && return 8
854 cleanup_nocli || return 6
856 run_test 24a "Multiple MDTs on a single node"
859 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
861 if [ -z "$fs2mds_DEV" ]; then
862 local dev=${SINGLEMDS}_dev
864 is_blkdev $SINGLEMDS $MDSDEV && \
865 skip_env "mixed loopback and real device not working" && return
868 local fs2mdsdev=${fs2mds_DEV:-${MDSDEV}_2}
870 add fs2mds $MDS_MKFS_OPTS --fsname=${FSNAME}2 --mgs --reformat $fs2mdsdev || exit 10
872 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS && return 2
875 run_test 24b "Multiple MGSs on a single node (should return err)"
879 check_mount || return 2
880 local MODULES=$($LCTL modules | awk '{ print $2 }')
881 rmmod $MODULES 2>/dev/null || true
884 run_test 25 "Verify modules are referenced"
888 # we need modules before mount for sysctl, so make sure...
889 do_facet $SINGLEMDS "lsmod | grep -q lustre || modprobe lustre"
890 #define OBD_FAIL_MDS_FS_SETUP 0x135
891 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000135"
892 start_mds && echo MDS started && return 1
893 lctl get_param -n devices
894 DEVS=$(lctl get_param -n devices | egrep -v MG | wc -l)
895 [ $DEVS -gt 0 ] && return 2
896 unload_modules_conf || return $?
898 run_test 26 "MDT startup failure cleans LOV (should return errs)"
904 local ORIG=$(do_facet $myfacet "$TEST")
905 if [ $# -gt 3 ]; then
911 echo "Setting $PARAM from $ORIG to $FINAL"
912 do_facet $SINGLEMDS "$LCTL conf_param $PARAM='$FINAL'" || error conf_param failed
914 wait_update $(facet_host $myfacet) "$TEST" "$FINAL" || error check failed!
918 start_ost || return 1
919 start_mds || return 2
920 echo "Requeue thread should have started: "
921 ps -e | grep ll_cfg_requeue
922 set_and_check ost1 "lctl get_param -n obdfilter.$FSNAME-OST0000.client_cache_seconds" "$FSNAME-OST0000.ost.client_cache_seconds" || return 3
925 run_test 27a "Reacquire MGS lock if OST started first"
930 local device=$(do_facet $SINGLEMDS "lctl get_param -n devices" | awk '($3 ~ "mdt" && $4 ~ "MDT") { print $4 }')
932 facet_failover $SINGLEMDS
933 set_and_check $SINGLEMDS "lctl get_param -n mdt.$device.identity_acquire_expire" "$device.mdt.identity_acquire_expire" || return 3
934 set_and_check client "lctl get_param -n mdc.$device-mdc-*.max_rpcs_in_flight" "$device.mdc.max_rpcs_in_flight" || return 4
938 run_test 27b "Reacquire MGS lock after failover"
942 TEST="lctl get_param -n llite.$FSNAME-*.max_read_ahead_whole_mb"
943 PARAM="$FSNAME.llite.max_read_ahead_whole_mb"
946 set_and_check client "$TEST" "$PARAM" $FINAL || return 3
947 FINAL=$(($FINAL + 1))
948 set_and_check client "$TEST" "$PARAM" $FINAL || return 4
949 umount_client $MOUNT || return 200
952 if [ $RESULT -ne $FINAL ]; then
953 echo "New config not seen: wanted $FINAL got $RESULT"
956 echo "New config success: got $RESULT"
958 set_and_check client "$TEST" "$PARAM" $ORIG || return 5
961 run_test 28 "permanent parameter setting"
964 [ "$OSTCOUNT" -lt "2" ] && skip_env "$OSTCOUNT < 2, skipping" && return
965 setup > /dev/null 2>&1
969 local PARAM="$FSNAME-OST0001.osc.active"
970 local PROC_ACT="osc.$FSNAME-OST0001-osc-[^M]*.active"
971 local PROC_UUID="osc.$FSNAME-OST0001-osc-[^M]*.ost_server_uuid"
973 ACTV=$(lctl get_param -n $PROC_ACT)
975 set_and_check client "lctl get_param -n $PROC_ACT" "$PARAM" $DEAC || return 2
976 # also check ost_server_uuid status
977 RESULT=$(lctl get_param -n $PROC_UUID | grep DEACTIV)
978 if [ -z "$RESULT" ]; then
979 echo "Live client not deactivated: $(lctl get_param -n $PROC_UUID)"
982 echo "Live client success: got $RESULT"
986 local mdtosc=$(get_mdtosc_proc_path $SINGLEMDS $FSNAME-OST0001)
987 mdtosc=${mdtosc/-MDT*/-MDT\*}
988 local MPROC="osc.$mdtosc.active"
993 RESULT=`do_facet $SINGLEMDS " lctl get_param -n $MPROC"`
994 [ ${PIPESTATUS[0]} = 0 ] || error "Can't read $MPROC"
995 if [ $RESULT -eq $DEAC ]; then
996 echo "MDT deactivated also after $WAIT sec (got $RESULT)"
1000 if [ $WAIT -eq $MAX ]; then
1001 echo "MDT not deactivated: wanted $DEAC got $RESULT"
1004 echo "Waiting $(($MAX - $WAIT)) secs for MDT deactivated"
1007 # test new client starts deactivated
1008 umount_client $MOUNT || return 200
1010 RESULT=$(lctl get_param -n $PROC_UUID | grep DEACTIV | grep NEW)
1011 if [ -z "$RESULT" ]; then
1012 echo "New client not deactivated from start: $(lctl get_param -n $PROC_UUID)"
1015 echo "New client success: got $RESULT"
1018 # make sure it reactivates
1019 set_and_check client "lctl get_param -n $PROC_ACT" "$PARAM" $ACTV || return 6
1021 umount_client $MOUNT
1024 #writeconf to remove all ost2 traces for subsequent tests
1027 run_test 29 "permanently remove an OST"
1032 echo Big config llog
1033 TEST="lctl get_param -n llite.$FSNAME-*.max_read_ahead_whole_mb"
1035 LIST=(1 2 3 4 5 4 3 2 1 2 3 4 5 4 3 2 1 2 3 4 5)
1036 for i in ${LIST[@]}; do
1037 set_and_check client "$TEST" "$FSNAME.llite.max_read_ahead_whole_mb" $i || return 3
1039 # make sure client restart still works
1040 umount_client $MOUNT
1041 mount_client $MOUNT || return 4
1042 [ "$($TEST)" -ne "$i" ] && error "Param didn't stick across restart $($TEST) != $i"
1045 echo Erase parameter setting
1046 do_facet mgs "$LCTL conf_param -d $FSNAME.llite.max_read_ahead_whole_mb" || return 6
1047 umount_client $MOUNT
1048 mount_client $MOUNT || return 6
1050 echo "deleted (default) value=$FINAL, orig=$ORIG"
1051 # assumes this parameter started at the default value
1052 [ "$FINAL" -eq "$ORIG" ] || fail "Deleted value=$FINAL, orig=$ORIG"
1056 run_test 30a "Big config llog and conf_param deletion"
1061 # Make a fake nid. Use the OST nid, and add 20 to the least significant
1062 # numerical part of it. Hopefully that's not already a failover address for
1064 OSTNID=$(do_facet ost1 "$LCTL get_param nis" | tail -1 | awk '{print $1}')
1065 ORIGVAL=$(echo $OSTNID | egrep -oi "[0-9]*@")
1066 NEWVAL=$((($(echo $ORIGVAL | egrep -oi "[0-9]*") + 20) % 256))
1067 NEW=$(echo $OSTNID | sed "s/$ORIGVAL/$NEWVAL@/")
1068 echo "Using fake nid $NEW"
1070 TEST="$LCTL get_param -n osc.$FSNAME-OST0000-osc-[^M]*.import | grep failover_nids | sed -n 's/.*\($NEW\).*/\1/p'"
1071 set_and_check client "$TEST" "$FSNAME-OST0000.failover.node" $NEW || error "didn't add failover nid $NEW"
1072 NIDS=$($LCTL get_param -n osc.$FSNAME-OST0000-osc-[^M]*.import | grep failover_nids)
1074 NIDCOUNT=$(($(echo "$NIDS" | wc -w) - 1))
1075 echo "should have 2 failover nids: $NIDCOUNT"
1076 [ $NIDCOUNT -eq 2 ] || error "Failover nid not added"
1077 do_facet mgs "$LCTL conf_param -d $FSNAME-OST0000.failover.node" || error "conf_param delete failed"
1078 umount_client $MOUNT
1079 mount_client $MOUNT || return 3
1081 NIDS=$($LCTL get_param -n osc.$FSNAME-OST0000-osc-[^M]*.import | grep failover_nids)
1083 NIDCOUNT=$(($(echo "$NIDS" | wc -w) - 1))
1084 echo "only 1 final nid should remain: $NIDCOUNT"
1085 [ $NIDCOUNT -eq 1 ] || error "Failover nids not removed"
1089 run_test 30b "Remove failover nids"
1091 test_31() { # bug 10734
1092 # ipaddr must not exist
1093 mount -t lustre 4.3.2.1@tcp:/lustre $MOUNT || true
1096 run_test 31 "Connect to non-existent node (shouldn't crash)"
1098 # Use these start32/stop32 fn instead of t-f start/stop fn,
1099 # for local devices, to skip global facet vars init
1103 echo "Stopping local ${MOUNT%/*}/${facet} (opts:$@)"
1104 umount -d $@ ${MOUNT%/*}/${facet}
1113 mkdir -p ${MOUNT%/*}/${facet}
1115 echo "Starting local ${facet}: $@ $device ${MOUNT%/*}/${facet}"
1116 mount -t lustre $@ ${device} ${MOUNT%/*}/${facet}
1118 if [ $RC -ne 0 ]; then
1119 echo "mount -t lustre $@ ${device} ${MOUNT%/*}/${facet}"
1120 echo "Start of ${device} of local ${facet} failed ${RC}"
1126 cleanup_nocli32 () {
1134 echo "Cleanup test_32 umount $MOUNT ..."
1135 umount -f $MOUNT || true
1136 echo "Cleanup local mds ost1 ..."
1138 combined_mgs_mds || start_mgs
1143 client_only && skip "client only testing" && return 0
1144 [ "$NETTYPE" = "tcp" ] || { skip "NETTYPE != tcp" && return 0; }
1145 [ -z "$TUNEFS" ] && skip_env "No tunefs" && return 0
1147 local DISK1_8=$LUSTRE/tests/disk1_8.tar.bz2
1148 [ ! -r $DISK1_8 ] && skip_env "Cannot find $DISK1_8" && return 0
1149 local tmpdir=$TMP/conf32a
1151 tar xjvf $DISK1_8 -C $tmpdir || \
1152 { skip_env "Cannot untar $DISK1_8" && return 0; }
1155 $LCTL set_param debug=$PTLDEBUG
1157 $TUNEFS $tmpdir/mds || error "tunefs failed"
1159 combined_mgs_mds || stop mgs
1161 # nids are wrong, so client wont work, but server should start
1162 start32 mds1 $tmpdir/mds "-o loop,exclude=lustre-OST0000" && \
1163 trap cleanup_32 EXIT INT || return 3
1165 local UUID=$($LCTL get_param -n mdt.lustre-MDT0000.uuid)
1167 [ "$UUID" == "lustre-MDT0000_UUID" ] || error "UUID is wrong: $UUID"
1169 $TUNEFS --mgsnode=$HOSTNAME $tmpdir/ost1 || error "tunefs failed"
1170 start32 ost1 $tmpdir/ost1 "-o loop" || return 5
1171 UUID=$($LCTL get_param -n obdfilter.lustre-OST0000.uuid)
1173 [ "$UUID" == "lustre-OST0000_UUID" ] || error "UUID is wrong: $UUID"
1175 local NID=$($LCTL list_nids | head -1)
1177 echo "OSC changes should succeed:"
1178 $LCTL conf_param lustre-OST0000.osc.max_dirty_mb=15 || return 7
1179 $LCTL conf_param lustre-OST0000.failover.node=$NID || return 8
1182 echo "MDC changes should succeed:"
1183 $LCTL conf_param lustre-MDT0000.mdc.max_rpcs_in_flight=9 || return 9
1184 $LCTL conf_param lustre-MDT0000.failover.node=$NID || return 10
1187 echo "LOV changes should succeed:"
1188 $LCTL pool_new lustre.interop || return 11
1189 $LCTL conf_param lustre-MDT0000.lov.stripesize=4M || return 12
1194 # mount a second time to make sure we didnt leave upgrade flag on
1196 $TUNEFS --dryrun $tmpdir/mds || error "tunefs failed"
1198 combined_mgs_mds || stop mgs
1200 start32 mds1 $tmpdir/mds "-o loop,exclude=lustre-OST0000" && \
1201 trap cleanup_32 EXIT INT || return 12
1205 rm -rf $tmpdir || true # true is only for TMP on NFS
1207 run_test 32a "Upgrade from 1.8 (not live)"
1210 client_only && skip "client only testing" && return 0
1211 [ "$NETTYPE" = "tcp" ] || { skip "NETTYPE != tcp" && return 0; }
1212 [ -z "$TUNEFS" ] && skip_env "No tunefs" && return
1214 local DISK1_8=$LUSTRE/tests/disk1_8.tar.bz2
1215 [ ! -r $DISK1_8 ] && skip_env "Cannot find $DISK1_8" && return 0
1216 local tmpdir=$TMP/conf32b
1218 tar xjvf $DISK1_8 -C $tmpdir || \
1219 { skip_env "Cannot untar $DISK1_8" && return ; }
1222 $LCTL set_param debug="config"
1223 local NEWNAME=lustre
1225 # writeconf will cause servers to register with their current nids
1226 $TUNEFS --writeconf --fsname=$NEWNAME $tmpdir/mds || error "tunefs failed"
1227 combined_mgs_mds || stop mgs
1229 start32 mds1 $tmpdir/mds "-o loop" && \
1230 trap cleanup_32 EXIT INT || return 3
1232 local UUID=$($LCTL get_param -n mdt.${NEWNAME}-MDT0000.uuid)
1234 [ "$UUID" == "${NEWNAME}-MDT0000_UUID" ] || error "UUID is wrong: $UUID"
1236 $TUNEFS --mgsnode=$HOSTNAME --writeconf --fsname=$NEWNAME $tmpdir/ost1 ||\
1237 error "tunefs failed"
1238 start32 ost1 $tmpdir/ost1 "-o loop" || return 5
1239 UUID=$($LCTL get_param -n obdfilter.${NEWNAME}-OST0000.uuid)
1241 [ "$UUID" == "${NEWNAME}-OST0000_UUID" ] || error "UUID is wrong: $UUID"
1243 local NID=$($LCTL list_nids | head -1)
1245 echo "OSC changes should succeed:"
1246 $LCTL conf_param ${NEWNAME}-OST0000.osc.max_dirty_mb=15 || return 7
1247 $LCTL conf_param ${NEWNAME}-OST0000.failover.node=$NID || return 8
1250 echo "MDC changes should succeed:"
1251 $LCTL conf_param ${NEWNAME}-MDT0000.mdc.max_rpcs_in_flight=9 || return 9
1252 $LCTL conf_param ${NEWNAME}-MDT0000.failover.node=$NID || return 10
1255 echo "LOV changes should succeed:"
1256 $LCTL pool_new ${NEWNAME}.interop || return 11
1257 $LCTL conf_param ${NEWNAME}-MDT0000.lov.stripesize=4M || return 12
1260 # MDT and OST should have registered with new nids, so we should have
1261 # a fully-functioning client
1262 echo "Check client and old fs contents"
1264 local device=`h2$NETTYPE $HOSTNAME`:/$NEWNAME
1265 echo "Starting local client: $HOSTNAME: $device $MOUNT"
1266 mount -t lustre $device $MOUNT || return 1
1268 local old=$($LCTL get_param -n mdc.*.max_rpcs_in_flight)
1269 local new=$((old + 5))
1270 $LCTL conf_param ${NEWNAME}-MDT0000.mdc.max_rpcs_in_flight=$new
1271 wait_update $HOSTNAME "$LCTL get_param -n mdc.*.max_rpcs_in_flight" $new || return 11
1273 [ "$(cksum $MOUNT/passwd | cut -d' ' -f 1,2)" == "94306271 1478" ] || return 12
1278 rm -rf $tmpdir || true # true is only for TMP on NFS
1280 run_test 32b "Upgrade from 1.8 with writeconf"
1282 test_33a() { # bug 12333, was test_33
1284 local FSNAME2=test-123
1285 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
1287 [ -n "$ost1_HOST" ] && fs2ost_HOST=$ost1_HOST
1289 if [ -z "$fs2ost_DEV" -o -z "$fs2mds_DEV" ]; then
1290 local dev=${SINGLEMDS}_dev
1291 local MDSDEV=${!dev}
1292 is_blkdev $SINGLEMDS $MDSDEV && \
1293 skip_env "mixed loopback and real device not working" && return
1296 combined_mgs_mds || mkfs_opts="$mkfs_opts --nomgs"
1298 local fs2mdsdev=${fs2mds_DEV:-${MDSDEV}_2}
1299 local fs2ostdev=${fs2ost_DEV:-$(ostdevname 1)_2}
1300 add fs2mds $MDS_MKFS_OPTS --mkfsoptions='\"-J size=8\"' --fsname=${FSNAME2} --reformat $fs2mdsdev || exit 10
1301 add fs2ost $OST_MKFS_OPTS --fsname=${FSNAME2} --index=8191 --mgsnode=$MGSNID --reformat $fs2ostdev || exit 10
1303 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS && trap cleanup_24a EXIT INT
1304 start fs2ost $fs2ostdev $OST_MOUNT_OPTS
1305 do_facet $SINGLEMDS "$LCTL conf_param $FSNAME2.sys.timeout=200" || rc=1
1307 mount -t lustre $MGSNID:/${FSNAME2} $MOUNT2 || rc=2
1310 cp /etc/hosts $MOUNT2/ || rc=3
1311 $LFS getstripe $MOUNT2/hosts
1316 rm -rf $MOUNT2 $fs2mdsdev $fs2ostdev
1317 cleanup_nocli || rc=6
1320 run_test 33a "Mount ost with a large index number"
1322 test_33b() { # was test_34
1325 do_facet client dd if=/dev/zero of=$MOUNT/24 bs=1024k count=1
1326 # Drop lock cancelation reply during umount
1327 #define OBD_FAIL_LDLM_CANCEL 0x304
1328 do_facet client lctl set_param fail_loc=0x80000304
1329 #lctl set_param debug=-1
1330 umount_client $MOUNT
1333 run_test 33b "Drop cancel during umount"
1337 do_facet client "sh runmultiop_bg_pause $DIR/file O_c"
1338 manual_umount_client
1340 do_facet client killall -USR1 multiop
1341 if [ $rc -eq 0 ]; then
1342 error "umount not fail!"
1347 run_test 34a "umount with opened file should be fail"
1352 touch $DIR/$tfile || return 1
1353 stop_mds --force || return 2
1355 manual_umount_client --force
1357 if [ $rc -ne 0 ]; then
1358 error "mtab after failed umount - rc $rc"
1364 run_test 34b "force umount with failed mds should be normal"
1368 touch $DIR/$tfile || return 1
1369 stop_ost --force || return 2
1371 manual_umount_client --force
1373 if [ $rc -ne 0 ]; then
1374 error "mtab after failed umount - rc $rc"
1380 run_test 34c "force umount with failed ost should be normal"
1382 test_35a() { # bug 12459
1385 DBG_SAVE="`lctl get_param -n debug`"
1386 lctl set_param debug="ha"
1388 log "Set up a fake failnode for the MDS"
1390 local device=$(do_facet $SINGLEMDS "lctl get_param -n devices" | awk '($3 ~ "mdt" && $4 ~ "MDT") { print $4 }' | head -1)
1391 do_facet $SINGLEMDS $LCTL conf_param ${device}.failover.node=$FAKENID || return 4
1393 log "Wait for RECONNECT_INTERVAL seconds (10s)"
1396 MSG="conf-sanity.sh test_35a `date +%F%kh%Mm%Ss`"
1399 log "Stopping the MDT:"
1400 stop_mds || return 5
1402 df $MOUNT > /dev/null 2>&1 &
1404 log "Restarting the MDT:"
1405 start_mds || return 6
1406 log "Wait for df ($DFPID) ... "
1409 lctl set_param debug="$DBG_SAVE"
1411 # retrieve from the log the first server that the client tried to
1412 # contact after the connection loss
1413 $LCTL dk $TMP/lustre-log-$TESTNAME.log
1414 NEXTCONN=`awk "/${MSG}/ {start = 1;}
1415 /import_select_connection.*$device-mdc.* using connection/ {
1417 if (\\\$NF ~ /$FAKENID/)
1423 }" $TMP/lustre-log-$TESTNAME.log`
1424 [ "$NEXTCONN" != "0" ] && log "The client didn't try to reconnect to the last active server (tried ${NEXTCONN} instead)" && return 7
1426 # remove nid settings
1429 run_test 35a "Reconnect to the last active server first"
1431 test_35b() { # bug 18674
1432 remote_mds || { skip "local MDS" && return 0; }
1436 $LCTL set_param debug="ha"
1438 MSG="conf-sanity.sh test_35b `date +%F%kh%Mm%Ss`"
1441 log "Set up a fake failnode for the MDS"
1443 local device=$(do_facet $SINGLEMDS "$LCTL get_param -n devices" | \
1444 awk '($3 ~ "mdt" && $4 ~ "MDT") { print $4 }' | head -1)
1445 do_facet $SINGLEMDS "$LCTL conf_param ${device}.failover.node=$FAKENID" || \
1448 local at_max_saved=0
1449 # adaptive timeouts may prevent seeing the issue
1450 if at_is_enabled; then
1451 at_max_saved=$(at_max_get mds)
1452 at_max_set 0 mds client
1455 mkdir -p $MOUNT/testdir
1456 touch $MOUNT/testdir/test
1458 log "Injecting EBUSY on MDS"
1459 # Setting OBD_FAIL_MDS_RESEND=0x136
1460 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0x80000136" || return 2
1462 log "Stat on a test file"
1463 stat $MOUNT/testdir/test
1465 log "Stop injecting EBUSY on MDS"
1466 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0" || return 3
1467 rm -f $MOUNT/testdir/test
1470 # restore adaptive timeout
1471 [ $at_max_saved -ne 0 ] && at_max_set $at_max_saved mds client
1473 $LCTL dk $TMP/lustre-log-$TESTNAME.log
1475 # retrieve from the log if the client has ever tried to
1476 # contact the fake server after the loss of connection
1477 FAILCONN=`awk "BEGIN {ret = 0;}
1478 /import_select_connection.*${FSNAME}-MDT0000-mdc.* using connection/ {
1480 if (\\\$NF ~ /$FAKENID/) {
1485 END {print ret}" $TMP/lustre-log-$TESTNAME.log`
1487 [ "$FAILCONN" == "0" ] && \
1488 log "ERROR: The client reconnection has not been triggered" && \
1490 [ "$FAILCONN" == "2" ] && \
1491 log "ERROR: The client tried to reconnect to the failover server while the primary was busy" && \
1495 # remove nid settings
1498 run_test 35b "Continue reconnection retries, if the active server is busy"
1501 [ $OSTCOUNT -lt 2 ] && skip_env "skipping test for single OST" && return
1503 [ "$ost_HOST" = "`hostname`" -o "$ost1_HOST" = "`hostname`" ] || \
1504 { skip "remote OST" && return 0; }
1507 local FSNAME2=test1234
1508 local fs3ost_HOST=$ost_HOST
1509 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
1511 [ -n "$ost1_HOST" ] && fs2ost_HOST=$ost1_HOST && fs3ost_HOST=$ost1_HOST
1513 if [ -z "$fs2ost_DEV" -o -z "$fs2mds_DEV" -o -z "$fs3ost_DEV" ]; then
1514 is_blkdev $SINGLEMDS $MDSDEV && \
1515 skip_env "mixed loopback and real device not working" && return
1518 local fs2mdsdev=${fs2mds_DEV:-${MDSDEV}_2}
1519 local fs2ostdev=${fs2ost_DEV:-$(ostdevname 1)_2}
1520 local fs3ostdev=${fs3ost_DEV:-$(ostdevname 2)_2}
1521 add fs2mds $MDS_MKFS_OPTS --fsname=${FSNAME2} --reformat $fs2mdsdev || exit 10
1522 # XXX after we support non 4K disk blocksize, change following --mkfsoptions with
1524 add fs2ost $OST_MKFS_OPTS --mkfsoptions='-b4096' --fsname=${FSNAME2} --mgsnode=$MGSNID --reformat $fs2ostdev || exit 10
1525 add fs3ost $OST_MKFS_OPTS --mkfsoptions='-b4096' --fsname=${FSNAME2} --mgsnode=$MGSNID --reformat $fs3ostdev || exit 10
1527 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS
1528 start fs2ost $fs2ostdev $OST_MOUNT_OPTS
1529 start fs3ost $fs3ostdev $OST_MOUNT_OPTS
1531 mount -t lustre $MGSNID:/${FSNAME2} $MOUNT2 || return 1
1533 sleep 5 # until 11778 fixed
1535 dd if=/dev/zero of=$MOUNT2/$tfile bs=1M count=7 || return 2
1537 BKTOTAL=`lctl get_param -n obdfilter.*.kbytestotal | awk 'BEGIN{total=0}; {total+=$1}; END{print total}'`
1538 BKFREE=`lctl get_param -n obdfilter.*.kbytesfree | awk 'BEGIN{free=0}; {free+=$1}; END{print free}'`
1539 BKAVAIL=`lctl get_param -n obdfilter.*.kbytesavail | awk 'BEGIN{avail=0}; {avail+=$1}; END{print avail}'`
1540 STRING=`df -P $MOUNT2 | tail -n 1 | awk '{print $2","$3","$4}'`
1541 DFTOTAL=`echo $STRING | cut -d, -f1`
1542 DFUSED=`echo $STRING | cut -d, -f2`
1543 DFAVAIL=`echo $STRING | cut -d, -f3`
1544 DFFREE=$(($DFTOTAL - $DFUSED))
1546 ALLOWANCE=$((64 * $OSTCOUNT))
1548 if [ $DFTOTAL -lt $(($BKTOTAL - $ALLOWANCE)) ] ||
1549 [ $DFTOTAL -gt $(($BKTOTAL + $ALLOWANCE)) ] ; then
1550 echo "**** FAIL: df total($DFTOTAL) mismatch OST total($BKTOTAL)"
1553 if [ $DFFREE -lt $(($BKFREE - $ALLOWANCE)) ] ||
1554 [ $DFFREE -gt $(($BKFREE + $ALLOWANCE)) ] ; then
1555 echo "**** FAIL: df free($DFFREE) mismatch OST free($BKFREE)"
1558 if [ $DFAVAIL -lt $(($BKAVAIL - $ALLOWANCE)) ] ||
1559 [ $DFAVAIL -gt $(($BKAVAIL + $ALLOWANCE)) ] ; then
1560 echo "**** FAIL: df avail($DFAVAIL) mismatch OST avail($BKAVAIL)"
1565 stop fs3ost -f || return 200
1566 stop fs2ost -f || return 201
1567 stop fs2mds -f || return 202
1568 rm -rf $MOUNT2 $fs2mdsdev $fs2ostdev $fs3ostdev
1569 unload_modules_conf || return 203
1572 run_test 36 "df report consistency on OSTs with different block size"
1575 local mntpt=$(facet_mntpt $SINGLEMDS)
1576 local mdsdev=$(mdsdevname ${SINGLEMDS//mds/})
1577 local mdsdev_sym="$TMP/sym_mdt.img"
1579 echo "MDS : $mdsdev"
1580 echo "SYMLINK : $mdsdev_sym"
1581 do_facet $SINGLEMDS rm -f $mdsdev_sym
1583 do_facet $SINGLEMDS ln -s $mdsdev $mdsdev_sym
1585 echo "mount symlink device - $mdsdev_sym"
1588 mount_op=$(do_facet $SINGLEMDS mount -v -t lustre $MDS_MOUNT_OPTS $mdsdev_sym $mntpt 2>&1 )
1591 echo mount_op=$mount_op
1593 do_facet $SINGLEMDS "umount -d $mntpt && rm -f $mdsdev_sym"
1595 if $(echo $mount_op | grep -q "unable to set tunable"); then
1596 error "set tunables failed for symlink device"
1599 [ $rc -eq 0 ] || error "mount symlink $mdsdev_sym failed! rc=$rc"
1603 run_test 37 "verify set tunables works for symlink device"
1605 test_38() { # bug 14222
1610 FILES=`find $SRC -type f -mtime +1 | head -n $COUNT`
1611 log "copying $(echo $FILES | wc -w) files to $DIR/$tdir"
1613 tar cf - $FILES | tar xf - -C $DIR/$tdir || \
1614 error "copying $SRC to $DIR/$tdir"
1616 umount_client $MOUNT
1618 log "rename lov_objid file on MDS"
1619 rm -f $TMP/lov_objid.orig
1621 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
1622 do_facet $SINGLEMDS "$DEBUGFS -c -R \\\"dump lov_objid $TMP/lov_objid.orig\\\" $MDSDEV"
1623 do_facet $SINGLEMDS "$DEBUGFS -w -R \\\"rm lov_objid\\\" $MDSDEV"
1625 do_facet $SINGLEMDS "od -Ax -td8 $TMP/lov_objid.orig"
1626 # check create in mds_lov_connect
1630 [ $V ] && log "verifying $DIR/$tdir/$f"
1631 diff -q $f $DIR/$tdir/$f || ERROR=y
1633 do_facet $SINGLEMDS "$DEBUGFS -c -R \\\"dump lov_objid $TMP/lov_objid.new\\\" $MDSDEV"
1634 do_facet $SINGLEMDS "od -Ax -td8 $TMP/lov_objid.new"
1635 [ "$ERROR" = "y" ] && error "old and new files are different after connect" || true
1637 # check it's updates in sync
1638 umount_client $MOUNT
1641 do_facet $SINGLEMDS dd if=/dev/zero of=$TMP/lov_objid.clear bs=4096 count=1
1642 do_facet $SINGLEMDS "$DEBUGFS -w -R \\\"rm lov_objid\\\" $MDSDEV"
1643 do_facet $SINGLEMDS "$DEBUGFS -w -R \\\"write $TMP/lov_objid.clear lov_objid\\\" $MDSDEV "
1648 [ $V ] && log "verifying $DIR/$tdir/$f"
1649 diff -q $f $DIR/$tdir/$f || ERROR=y
1651 do_facet $SINGLEMDS "$DEBUGFS -c -R \\\"dump lov_objid $TMP/lov_objid.new1\\\" $MDSDEV"
1652 do_facet $SINGLEMDS "od -Ax -td8 $TMP/lov_objid.new1"
1653 umount_client $MOUNT
1655 [ "$ERROR" = "y" ] && error "old and new files are different after sync" || true
1657 log "files compared the same"
1660 run_test 38 "MDS recreates missing lov_objid file from OST data"
1666 perl $SRCDIR/leak_finder.pl $TMP/debug 2>&1 | egrep '*** Leak:' &&
1667 error "memory leak detected" || true
1669 run_test 39 "leak_finder recognizes both LUSTRE and LNET malloc messages"
1671 test_40() { # bug 15759
1673 #define OBD_FAIL_TGT_TOOMANY_THREADS 0x706
1674 do_facet $SINGLEMDS "sysctl -w lustre.fail_loc=0x80000706"
1678 run_test 40 "race during service thread startup"
1680 test_41() { #bug 14134
1682 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
1684 start $SINGLEMDS $MDSDEV $MDS_MOUNT_OPTS -o nosvc -n
1685 start ost1 `ostdevname 1` $OST_MOUNT_OPTS
1686 start $SINGLEMDS $MDSDEV $MDS_MOUNT_OPTS -o nomgs,force
1688 mount_client $MOUNT || return 1
1691 echo "blah blah" > $MOUNT/$tfile
1694 umount_client $MOUNT
1695 stop ost1 -f || return 201
1696 stop_mds -f || return 202
1697 stop_mds -f || return 203
1698 unload_modules_conf || return 204
1701 run_test 41 "mount mds with --nosvc and --nomgs"
1703 test_42() { #bug 14693
1705 check_mount || return 2
1706 do_facet mgs $LCTL conf_param lustre.llite.some_wrong_param=10
1707 umount_client $MOUNT
1708 mount_client $MOUNT || return 1
1712 run_test 42 "invalid config param should not prevent client from mounting"
1715 [ $UID -ne 0 -o $RUNAS_ID -eq 0 ] && skip_env "run as root"
1717 chmod ugo+x $DIR || error "chmod 0 failed"
1719 "lctl get_param -n mdt.$FSNAME-MDT0000.root_squash" \
1720 "$FSNAME.mdt.root_squash" \
1723 "lctl get_param -n mdt.$FSNAME-MDT0000.nosquash_nids" \
1724 "$FSNAME.mdt.nosquash_nids" \
1728 # create set of test files
1730 echo "111" > $DIR/$tfile-userfile || error "write 1 failed"
1731 chmod go-rw $DIR/$tfile-userfile || error "chmod 1 failed"
1732 chown $RUNAS_ID.$RUNAS_ID $DIR/$tfile-userfile || error "chown failed"
1734 echo "222" > $DIR/$tfile-rootfile || error "write 2 failed"
1735 chmod go-rw $DIR/$tfile-rootfile || error "chmod 2 faield"
1737 mkdir $DIR/$tdir-rootdir -p || error "mkdir failed"
1738 chmod go-rwx $DIR/$tdir-rootdir || error "chmod 3 failed"
1739 touch $DIR/$tdir-rootdir/tfile-1 || error "touch failed"
1742 # check root_squash:
1743 # set root squash UID:GID to RUNAS_ID
1744 # root should be able to access only files owned by RUNAS_ID
1747 "lctl get_param -n mdt.$FSNAME-MDT0000.root_squash" \
1748 "$FSNAME.mdt.root_squash" \
1749 "$RUNAS_ID:$RUNAS_ID"
1751 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-userfile)
1752 dd if=$DIR/$tfile-userfile 1>/dev/null 2>/dev/null || \
1753 error "$ST: root read permission is denied"
1754 echo "$ST: root read permission is granted - ok"
1757 dd conv=notrunc if=$DIR/$tfile-userfile 1>/dev/null 2>/dev/null || \
1758 error "$ST: root write permission is denied"
1759 echo "$ST: root write permission is granted - ok"
1761 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-rootfile)
1762 dd if=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null && \
1763 error "$ST: root read permission is granted"
1764 echo "$ST: root read permission is denied - ok"
1767 dd conv=notrunc of=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null && \
1768 error "$ST: root write permission is granted"
1769 echo "$ST: root write permission is denied - ok"
1771 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tdir-rootdir)
1772 rm $DIR/$tdir-rootdir/tfile-1 1>/dev/null 2>/dev/null && \
1773 error "$ST: root unlink permission is granted"
1774 echo "$ST: root unlink permission is denied - ok"
1776 touch $DIR/tdir-rootdir/tfile-2 1>/dev/null 2>/dev/null && \
1777 error "$ST: root create permission is granted"
1778 echo "$ST: root create permission is denied - ok"
1781 # check nosquash_nids:
1782 # put client's NID into nosquash_nids list,
1783 # root should be able to access root file after that
1785 local NIDLIST=$(lctl list_nids all | tr '\n' ' ')
1786 NIDLIST="2@elan $NIDLIST 192.168.0.[2,10]@tcp"
1787 NIDLIST=$(echo $NIDLIST | tr -s ' ' ' ')
1789 "lctl get_param -n mdt.$FSNAME-MDT0000.nosquash_nids" \
1790 "$FSNAME-MDTall.mdt.nosquash_nids" \
1793 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-rootfile)
1794 dd if=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null || \
1795 error "$ST: root read permission is denied"
1796 echo "$ST: root read permission is granted - ok"
1799 dd conv=notrunc of=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null || \
1800 error "$ST: root write permission is denied"
1801 echo "$ST: root write permission is granted - ok"
1803 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tdir-rootdir)
1804 rm $DIR/$tdir-rootdir/tfile-1 || \
1805 error "$ST: root unlink permission is denied"
1806 echo "$ST: root unlink permission is granted - ok"
1807 touch $DIR/$tdir-rootdir/tfile-2 || \
1808 error "$ST: root create permission is denied"
1809 echo "$ST: root create permission is granted - ok"
1813 run_test 43 "check root_squash and nosquash_nids"
1815 umount_client $MOUNT
1820 check_mount || return 2
1821 UUID=$($LCTL get_param llite.${FSNAME}*.uuid | cut -d= -f2)
1823 UUIDS=$(do_facet $SINGLEMDS "$LCTL get_param mdt.${FSNAME}*.exports.*.uuid")
1824 for VAL in $UUIDS; do
1825 NID=$(echo $VAL | cut -d= -f1)
1826 CLUUID=$(echo $VAL | cut -d= -f2)
1827 [ "$UUID" = "$CLUUID" ] && STATS_FOUND=yes && break
1829 [ "$STATS_FOUND" = "no" ] && error "stats not found for client"
1833 run_test 44 "mounted client proc entry exists"
1837 check_mount || return 2
1842 #define OBD_FAIL_PTLRPC_LONG_UNLINK 0x50f
1843 do_facet client "lctl set_param fail_loc=0x50f"
1846 manual_umount_client --force || return 3
1847 do_facet client "lctl set_param fail_loc=0x0"
1849 mount_client $MOUNT || return 4
1853 run_test 45 "long unlink handling in ptlrpcd"
1860 umount_client $MOUNT2 || rc=$?
1861 umount_client $MOUNT || rc=$?
1862 while [ $count -gt 0 ]; do
1863 stop ost${count} -f || rc=$?
1867 cleanup_nocli || rc=$?
1868 #writeconf to remove all ost2 traces for subsequent tests
1874 echo "Testing with $OSTCOUNT OSTs"
1876 start_mds || return 1
1877 #first client should see only one ost
1878 start_ost || return 2
1879 wait_osc_import_state mds ost FULL
1881 mount_client $MOUNT || return 3
1882 trap "cleanup_46a $OSTCOUNT" EXIT ERR
1885 for (( i=2; i<=$OSTCOUNT; i++ )); do
1886 start ost$i `ostdevname $i` $OST_MOUNT_OPTS || return $((i+2))
1889 # wait until osts in sync
1890 for (( i=2; i<=$OSTCOUNT; i++ )); do
1891 wait_osc_import_state mds ost$i FULL
1895 #second client see all ost's
1897 mount_client $MOUNT2 || return 8
1898 $LFS setstripe $MOUNT2 -c -1 || return 9
1899 $LFS getstripe $MOUNT2 || return 10
1901 echo "ok" > $MOUNT2/widestripe
1902 $LFS getstripe $MOUNT2/widestripe || return 11
1903 # fill acl buffer for avoid expand lsm to them
1904 awk -F : '{if (FNR < 25) { print "u:"$1":rwx" }}' /etc/passwd | while read acl; do
1905 setfacl -m $acl $MOUNT2/widestripe
1909 stat $MOUNT/widestripe || return 12
1911 cleanup_46a $OSTCOUNT || { echo "cleanup_46a failed!" && return 13; }
1914 run_test 46a "handle ost additional - wide striped file"
1919 check_mount || return 2
1920 $LCTL set_param ldlm.namespaces.$FSNAME-*-*-*.lru_size=100
1924 for ns in $($LCTL get_param ldlm.namespaces.$FSNAME-*-*-*.lru_size); do
1925 if echo $ns | grep "MDT[[:digit:]]*"; then
1928 lrs=$(echo $ns | sed 's/.*lru_size=//')
1929 lru_size[count]=$lrs
1934 facet_failover $SINGLEMDS
1935 client_up || return 3
1938 for ns in $($LCTL get_param ldlm.namespaces.$FSNAME-*-*-*.lru_size); do
1939 if echo $ns | grep "MDT[[:digit:]]*"; then
1942 lrs=$(echo $ns | sed 's/.*lru_size=//')
1943 if ! test "$lrs" -eq "${lru_size[count]}"; then
1944 n=$(echo $ns | sed -e 's/ldlm.namespaces.//' -e 's/.lru_size=.*//')
1945 error "$n has lost lru_size: $lrs vs. ${lru_size[count]}"
1953 run_test 47 "server restart does not make client loss lru_resize settings"
1958 # reformat after this test is needed - if test will failed
1959 # we will have unkillable file at FS
1963 test_48() { # bug 17636
1966 check_mount || return 2
1968 $LFS setstripe $MOUNT -c -1 || return 9
1969 $LFS getstripe $MOUNT || return 10
1971 echo "ok" > $MOUNT/widestripe
1972 $LFS getstripe $MOUNT/widestripe || return 11
1974 trap cleanup_48 EXIT ERR
1976 # fill acl buffer for avoid expand lsm to them
1977 getent passwd | awk -F : '{ print "u:"$1":rwx" }' | while read acl; do
1978 setfacl -m $acl $MOUNT/widestripe
1981 stat $MOUNT/widestripe || return 12
1986 run_test 48 "too many acls on file"
1988 # check PARAM_SYS_LDLM_TIMEOUT option of MKFS.LUSTRE
1989 test_49() { # bug 17710
1990 local OLD_MDS_MKFS_OPTS=$MDS_MKFS_OPTS
1991 local OLD_OST_MKFS_OPTS=$OST_MKFS_OPTS
1992 local LOCAL_TIMEOUT=20
1995 OST_MKFS_OPTS="--ost --fsname=$FSNAME --device-size=$OSTSIZE --mgsnode=$MGSNID --param sys.timeout=$LOCAL_TIMEOUT --param sys.ldlm_timeout=$LOCAL_TIMEOUT $MKFSOPT $OSTOPT"
1999 check_mount || return 1
2001 echo "check ldlm_timout..."
2002 LDLM_MDS="`do_facet $SINGLEMDS lctl get_param -n ldlm_timeout`"
2003 LDLM_OST1="`do_facet ost1 lctl get_param -n ldlm_timeout`"
2004 LDLM_CLIENT="`do_facet client lctl get_param -n ldlm_timeout`"
2006 if [ $LDLM_MDS -ne $LDLM_OST1 ] || [ $LDLM_MDS -ne $LDLM_CLIENT ]; then
2007 error "Different LDLM_TIMEOUT:$LDLM_MDS $LDLM_OST1 $LDLM_CLIENT"
2010 if [ $LDLM_MDS -ne $((LOCAL_TIMEOUT / 3)) ]; then
2011 error "LDLM_TIMEOUT($LDLM_MDS) is not correct"
2014 umount_client $MOUNT
2015 stop_ost || return 2
2016 stop_mds || return 3
2018 OST_MKFS_OPTS="--ost --fsname=$FSNAME --device-size=$OSTSIZE --mgsnode=$MGSNID --param sys.timeout=$LOCAL_TIMEOUT --param sys.ldlm_timeout=$((LOCAL_TIMEOUT - 1)) $MKFSOPT $OSTOPT"
2022 check_mount || return 7
2024 LDLM_MDS="`do_facet $SINGLEMDS lctl get_param -n ldlm_timeout`"
2025 LDLM_OST1="`do_facet ost1 lctl get_param -n ldlm_timeout`"
2026 LDLM_CLIENT="`do_facet client lctl get_param -n ldlm_timeout`"
2028 if [ $LDLM_MDS -ne $LDLM_OST1 ] || [ $LDLM_MDS -ne $LDLM_CLIENT ]; then
2029 error "Different LDLM_TIMEOUT:$LDLM_MDS $LDLM_OST1 $LDLM_CLIENT"
2032 if [ $LDLM_MDS -ne $((LOCAL_TIMEOUT - 1)) ]; then
2033 error "LDLM_TIMEOUT($LDLM_MDS) is not correct"
2036 cleanup || return $?
2038 MDS_MKFS_OPTS=$OLD_MDS_MKFS_OPTS
2039 OST_MKFS_OPTS=$OLD_OST_MKFS_OPTS
2041 run_test 49 "check PARAM_SYS_LDLM_TIMEOUT option of MKFS.LUSTRE"
2044 # Test both statfs and lfs df and fail if either one fails
2045 multiop_bg_pause $1 f_
2048 killall -USR1 multiop
2049 [ $RC1 -ne 0 ] && log "lazystatfs multiop failed"
2050 wait $PID || { RC1=$?; log "multiop return error "; }
2057 if [ $RC2 -eq 0 ]; then
2059 log "lazystatfs df failed"
2063 [[ $RC1 -ne 0 || $RC2 -eq 0 ]] && RC=1
2069 lctl set_param llite.$FSNAME-*.lazystatfs=1
2072 lazystatfs $MOUNT || error "lazystatfs failed but no down servers"
2074 cleanup || return $?
2076 run_test 50a "lazystatfs all servers available =========================="
2080 lctl set_param llite.$FSNAME-*.lazystatfs=1
2083 # Wait for client to detect down OST
2084 stop_ost || error "Unable to stop OST1"
2085 wait_osc_import_state mds ost DISCONN
2087 lazystatfs $MOUNT || error "lazystatfs should don't have returned EIO"
2089 umount_client $MOUNT || error "Unable to unmount client"
2090 stop_mds || error "Unable to stop MDS"
2092 run_test 50b "lazystatfs all servers down =========================="
2095 start_mds || error "Unable to start MDS"
2096 start_ost || error "Unable to start OST1"
2097 start_ost2 || error "Unable to start OST2"
2098 mount_client $MOUNT || error "Unable to mount client"
2099 lctl set_param llite.$FSNAME-*.lazystatfs=1
2102 # Wait for client to detect down OST
2103 stop_ost || error "Unable to stop OST1"
2104 wait_osc_import_state mds ost DISCONN
2105 lazystatfs $MOUNT || error "lazystatfs failed with one down server"
2107 umount_client $MOUNT || error "Unable to unmount client"
2108 stop_ost2 || error "Unable to stop OST2"
2109 stop_mds || error "Unable to stop MDS"
2111 run_test 50c "lazystatfs one server down =========================="
2114 start_mds || error "Unable to start MDS"
2115 start_ost || error "Unable to start OST1"
2116 start_ost2 || error "Unable to start OST2"
2117 mount_client $MOUNT || error "Unable to mount client"
2118 lctl set_param llite.$FSNAME-*.lazystatfs=1
2121 # Issue the statfs during the window where the client still
2122 # belives the OST to be available but it is in fact down.
2123 # No failure just a statfs which hangs for a timeout interval.
2124 stop_ost || error "Unable to stop OST1"
2125 lazystatfs $MOUNT || error "lazystatfs failed with one down server"
2127 umount_client $MOUNT || error "Unable to unmount client"
2128 stop_ost2 || error "Unable to stop OST2"
2129 stop_mds || error "Unable to stop MDS"
2131 run_test 50d "lazystatfs client/server conn race =========================="
2138 start_mds || return 1
2139 #first client should see only one ost
2140 start_ost || return 2
2141 wait_osc_import_state mds ost FULL
2143 # Wait for client to detect down OST
2144 stop_ost || error "Unable to stop OST1"
2145 wait_osc_import_state mds ost DISCONN
2147 mount_client $MOUNT || error "Unable to mount client"
2148 lctl set_param llite.$FSNAME-*.lazystatfs=0
2150 multiop_bg_pause $MOUNT _f
2154 if [ $RC1 -ne 0 ]; then
2155 log "multiop failed $RC1"
2158 sleep $(( $TIMEOUT+1 ))
2160 [ $? -ne 0 ] && error "process isn't sleep"
2161 start_ost || error "Unable to start OST1"
2162 wait $pid || error "statfs failed"
2165 umount_client $MOUNT || error "Unable to unmount client"
2166 stop_ost || error "Unable to stop OST1"
2167 stop_mds || error "Unable to stop MDS"
2169 run_test 50e "normal statfs all servers down =========================="
2174 CONN_PROC="osc.$FSNAME-OST0001-osc-[M]*.ost_server_uuid"
2176 start_mds || error "Unable to start mds"
2177 #first client should see only one ost
2178 start_ost || error "Unable to start OST1"
2179 wait_osc_import_state mds ost FULL
2181 start_ost2 || error "Unable to start OST2"
2182 wait_osc_import_state mds ost2 FULL
2184 # Wait for client to detect down OST
2185 stop_ost2 || error "Unable to stop OST2"
2187 wait_osc_import_state mds ost2 DISCONN
2188 mount_client $MOUNT || error "Unable to mount client"
2189 lctl set_param llite.$FSNAME-*.lazystatfs=0
2191 multiop_bg_pause $MOUNT _f
2195 if [ $RC1 -ne 0 ]; then
2196 log "lazystatfs multiop failed $RC1"
2199 sleep $(( $TIMEOUT+1 ))
2201 [ $? -ne 0 ] && error "process isn't sleep"
2202 start_ost2 || error "Unable to start OST2"
2203 wait $pid || error "statfs failed"
2204 stop_ost2 || error "Unable to stop OST2"
2207 umount_client $MOUNT || error "Unable to unmount client"
2208 stop_ost || error "Unable to stop OST1"
2209 stop_mds || error "Unable to stop MDS"
2210 #writeconf to remove all ost2 traces for subsequent tests
2213 run_test 50f "normal statfs one server in down =========================="
2216 [ "$OSTCOUNT" -lt "2" ] && skip_env "$OSTCOUNT < 2, skipping" && return
2218 start_ost2 || error "Unable to start OST2"
2219 wait_osc_import_state mds ost2 FULL
2220 wait_osc_import_state client ost2 FULL
2222 local PARAM="${FSNAME}-OST0001.osc.active"
2224 $LFS setstripe -c -1 $DIR/$tfile || error "Unable to lfs setstripe"
2225 do_facet mgs $LCTL conf_param $PARAM=0 || error "Unable to deactivate OST"
2227 umount_client $MOUNT || error "Unable to unmount client"
2228 mount_client $MOUNT || error "Unable to mount client"
2229 # This df should not cause a panic
2232 do_facet mgs $LCTL conf_param $PARAM=1 || error "Unable to activate OST"
2234 umount_client $MOUNT || error "Unable to unmount client"
2235 stop_ost2 || error "Unable to stop OST2"
2236 stop_ost || error "Unable to stop OST1"
2237 stop_mds || error "Unable to stop MDS"
2238 #writeconf to remove all ost2 traces for subsequent tests
2241 run_test 50g "deactivated OST should not cause panic====================="
2244 local LOCAL_TIMEOUT=20
2248 check_mount || return 1
2251 $LFS setstripe -c -1 $MOUNT/d1
2252 #define OBD_FAIL_MDS_REINT_DELAY 0x142
2253 do_facet $SINGLEMDS "lctl set_param fail_loc=0x142"
2254 touch $MOUNT/d1/f1 &
2257 start_ost2 || return 2
2259 stop_ost2 || return 3
2262 run_test 51 "Verify that mdt_reint handles RMF_MDT_MD correctly when an OST is added"
2271 do_node $node mkdir -p $dest
2272 [ $? -eq 0 ] || { error "Unable to create directory"; return 1; }
2274 do_node $node 'tar cf - '$@' | tar xf - -C '$dest';
2275 [ \"\${PIPESTATUS[*]}\" = \"0 0\" ] || exit 1'
2276 [ $? -eq 0 ] || { error "Unable to tar files"; return 2; }
2278 do_node $node 'getfattr -d -m "[a-z]*\\." '$@' > '$xattrs
2279 [ $? -eq 0 ] || { error "Unable to read xattrs"; return 3; }
2289 local backup2=${TMP}/backup2
2291 do_node $node mkdir -p $backup2
2292 [ $? -eq 0 ] || { error "Unable to create directory"; return 1; }
2294 do_node $node 'tar cf - '$@' | tar xf - -C '$backup2';
2295 [ \"\${PIPESTATUS[*]}\" = \"0 0\" ] || exit 1'
2296 [ $? -eq 0 ] || { error "Unable to tar files to diff"; return 2; }
2298 do_node $node "diff -rq $backup $backup2"
2299 [ $? -eq 0 ] || { error "contents differ"; return 3; }
2301 local xattrs2=${TMP}/xattrs2
2302 do_node $node 'getfattr -d -m "[a-z]*\\." '$@' > '$xattrs2
2303 [ $? -eq 0 ] || { error "Unable to read xattrs to diff"; return 4; }
2305 do_node $node "diff $xattrs $xattrs2"
2306 [ $? -eq 0 ] || { error "xattrs differ"; return 5; }
2308 do_node $node "rm -rf $backup2 $xattrs2"
2309 [ $? -eq 0 ] || { error "Unable to delete temporary files"; return 6; }
2314 [ $? -eq 0 ] || { error "Unable to start MDS"; return 1; }
2316 [ $? -eq 0 ] || { error "Unable to start OST1"; return 2; }
2318 [ $? -eq 0 ] || { error "Unable to mount client"; return 3; }
2321 local ost1mnt=${MOUNT%/*}/ost1
2322 local ost1node=$(facet_active_host ost1)
2323 local ost1tmp=$TMP/conf52
2326 [ $? -eq 0 ] || { error "Unable to create tdir"; return 4; }
2327 touch $TMP/modified_first
2328 [ $? -eq 0 ] || { error "Unable to create temporary file"; return 5; }
2329 local mtime=$(stat -c %Y $TMP/modified_first)
2330 do_node $ost1node "mkdir -p $ost1tmp && touch -m -d @$mtime $ost1tmp/modified_first"
2332 [ $? -eq 0 ] || { error "Unable to create temporary file"; return 6; }
2335 $LFS setstripe $DIR/$tdir -c -1 -s 1M
2336 [ $? -eq 0 ] || { error "lfs setstripe failed"; return 7; }
2338 for (( i=0; i < nrfiles; i++ )); do
2339 multiop $DIR/$tdir/$tfile-$i Ow1048576w1048576w524288c
2340 [ $? -eq 0 ] || { error "multiop failed"; return 8; }
2346 echo backup files to $TMP/files
2347 local files=$(find $DIR/$tdir -type f -newer $TMP/modified_first)
2348 copy_files_xattrs `hostname` $TMP/files $TMP/file_xattrs $files
2349 [ $? -eq 0 ] || { error "Unable to copy files"; return 9; }
2351 umount_client $MOUNT
2352 [ $? -eq 0 ] || { error "Unable to umount client"; return 10; }
2354 [ $? -eq 0 ] || { error "Unable to stop ost1"; return 11; }
2356 echo mount ost1 as ldiskfs
2357 do_node $ost1node mount -t $FSTYPE $ost1_dev $ost1mnt $OST_MOUNT_OPTS
2358 [ $? -eq 0 ] || { error "Unable to mount ost1 as ldiskfs"; return 12; }
2361 echo backup objects to $ost1tmp/objects
2362 local objects=$(do_node $ost1node 'find '$ost1mnt'/O/0 -type f -size +0'\
2363 '-newer '$ost1tmp'/modified_first -regex ".*\/[0-9]+"')
2364 copy_files_xattrs $ost1node $ost1tmp/objects $ost1tmp/object_xattrs $objects
2365 [ $? -eq 0 ] || { error "Unable to copy objects"; return 13; }
2367 # move objects to lost+found
2368 do_node $ost1node 'mv '$objects' '${ost1mnt}'/lost+found'
2369 [ $? -eq 0 ] || { error "Unable to move objects"; return 14; }
2372 do_node $ost1node "ll_recover_lost_found_objs -d $ost1mnt/lost+found"
2373 [ $? -eq 0 ] || { error "ll_recover_lost_found_objs failed"; return 15; }
2375 # compare restored objects against saved ones
2376 diff_files_xattrs $ost1node $ost1tmp/objects $ost1tmp/object_xattrs $objects
2377 [ $? -eq 0 ] || { error "Unable to diff objects"; return 16; }
2379 do_node $ost1node "umount $ost1_dev"
2380 [ $? -eq 0 ] || { error "Unable to umount ost1 as ldiskfs"; return 17; }
2383 [ $? -eq 0 ] || { error "Unable to start ost1"; return 18; }
2385 [ $? -eq 0 ] || { error "Unable to mount client"; return 19; }
2388 diff_files_xattrs `hostname` $TMP/files $TMP/file_xattrs $files
2389 [ $? -eq 0 ] || { error "Unable to diff files"; return 20; }
2391 rm -rf $TMP/files $TMP/file_xattrs
2392 [ $? -eq 0 ] || { error "Unable to delete temporary files"; return 21; }
2393 do_node $ost1node "rm -rf $ost1tmp"
2394 [ $? -eq 0 ] || { error "Unable to delete temporary files"; return 22; }
2397 run_test 52 "check recovering objects from lost+found"
2399 # Checks threads_min/max/started for some service
2401 # Arguments: service name (OST or MDT), facet (e.g., ost1, $SINGLEMDS), and a
2402 # parameter pattern prefix like 'ost.*.ost'.
2414 local msg="Insane $modname thread counts"
2418 check_mount || return 41
2420 # We need to expand $parampat, but it may match multiple parameters, so
2421 # we'll pick the first one
2422 if ! paramp=$(do_facet $facet "lctl get_param -N ${parampat}.threads_min"|head -1); then
2423 error "Couldn't expand ${parampat}.threads_min parameter name"
2427 # Remove the .threads_min part
2428 paramp=${paramp%.threads_min}
2430 # Check for sanity in defaults
2431 tmin=$(do_facet $facet "lctl get_param -n ${paramp}.threads_min" || echo 0)
2432 tmax=$(do_facet $facet "lctl get_param -n ${paramp}.threads_max" || echo 0)
2433 tstarted=$(do_facet $facet "lctl get_param -n ${paramp}.threads_started" || echo 0)
2434 lassert 23 "$msg (PDSH problems?)" '(($tstarted && $tmin && $tmax))' || return $?
2435 lassert 24 "$msg" '(($tstarted >= $tmin && $tstarted <= $tmax ))' || return $?
2437 # Check that we can lower min/max
2438 do_facet $facet "lctl set_param ${paramp}.threads_min=$((tmin - 1))"
2439 do_facet $facet "lctl set_param ${paramp}.threads_max=$((tmax - 1))"
2440 tmin2=$(do_facet $facet "lctl get_param -n ${paramp}.threads_min" || echo 0)
2441 tmax2=$(do_facet $facet "lctl get_param -n ${paramp}.threads_max" || echo 0)
2442 lassert 25 "$msg" '(($tmin2 == ($tmin - 1) && $tmax2 == ($tmax -1)))' || return $?
2444 # Check that we can set min/max to the same value
2445 do_facet $facet "lctl set_param ${paramp}.threads_max=$((tmin - 1))"
2446 tmin2=$(do_facet $facet "lctl get_param -n ${paramp}.threads_min" || echo 0)
2447 tmax2=$(do_facet $facet "lctl get_param -n ${paramp}.threads_max" || echo 0)
2448 lassert 26 "$msg" '(($tmin2 == ($tmin - 1) && $tmax2 == ($tmin - 1)))' || return $?
2450 # Check that we can't set max < min
2451 do_facet $facet "lctl set_param ${paramp}.threads_max=$((tmin - 2))"
2452 tmin2=$(do_facet $facet "lctl get_param -n ${paramp}.threads_min" || echo 0)
2453 tmax2=$(do_facet $facet "lctl get_param -n ${paramp}.threads_max" || echo 0)
2454 lassert 27 "$msg" '(($tmin2 <= $tmax2))' || return $?
2456 # We need to ensure that we get the module options desired; to do this
2457 # we set LOAD_MODULES_REMOTE=true and we call setmodopts below.
2458 LOAD_MODULES_REMOTE=true
2461 setmodopts -a $modname "$opts" oldvalue
2465 check_mount || return 41
2467 # Restore previous setting of MODOPTS_*
2468 setmodopts $modname "$oldvalue"
2470 # Check that $opts took
2471 tmin=$(do_facet $facet "lctl get_param -n ${paramp}.threads_min")
2472 tmax=$(do_facet $facet "lctl get_param -n ${paramp}.threads_max")
2473 tstarted=$(do_facet $facet "lctl get_param -n ${paramp}.threads_started")
2474 lassert 28 "$msg" '(($tstarted == $tmin && $tstarted == $tmax ))' || return $?
2477 # Workaround a YALA bug where YALA expects that modules will remain
2478 # loaded on the servers
2479 LOAD_MODULES_REMOTE=false
2486 thread_sanity OST ost1 'ost.*.ost' 'oss_num_threads=64'
2488 run_test 53a "check OSS thread count params"
2491 thread_sanity MDT $SINGLEMDS 'mdt.*.*.' 'mdt_num_threads=64'
2493 run_test 53b "check MDT thread count params"
2498 local partial_arg=""
2499 local size=$(df -B G $dir | tail -1 | awk '{print $2}' | sed 's/G//') # Gb
2501 # Run in partial (fast) mode if the size
2502 # of a partition > 10 GB
2503 [ $size -gt 10 ] && partial_arg="-p"
2505 llverfs $partial_arg $dir
2509 do_rpc_nodes $(facet_host ost1) run_llverdev $(ostdevname 1)
2510 [ $? -eq 0 ] || error "llverdev failed!"
2513 run_test 54a "llverdev"
2518 [ $? -eq 0 ] || error "llverfs failed!"
2521 run_test 54b "llverfs"
2525 local max_ost_index=$1
2526 echo -n $(((max_ost_index + 1) * 8))
2530 local mdsdev=$(mdsdevname 1)
2531 local ostdev=$(ostdevname 1)
2532 local saved_opts=$OST_MKFS_OPTS
2534 for i in 0 1023 2048
2536 OST_MKFS_OPTS="$saved_opts --index $i"
2544 echo checking size of lov_objid for ost index $i
2545 LOV_OBJID_SIZE=$(do_facet mds1 "$DEBUGFS -R 'stat lov_objid' $mdsdev 2>/dev/null" | grep ^User | awk '{print $6}')
2546 if [ "$LOV_OBJID_SIZE" != $(lov_objid_size $i) ]; then
2547 error "lov_objid size has to be $(lov_objid_size $i), not $LOV_OBJID_SIZE"
2549 echo ok, lov_objid size is correct: $LOV_OBJID_SIZE
2554 OST_MKFS_OPTS=$saved_opts
2557 run_test 55 "check lov_objid size"
2560 add mds1 $MDS_MKFS_OPTS --mkfsoptions='\"-J size=16\"' --reformat $(mdsdevname 1)
2561 add ost1 $OST_MKFS_OPTS --index=1000 --reformat $(ostdevname 1)
2562 add ost2 $OST_MKFS_OPTS --index=10000 --reformat $(ostdevname 2)
2566 start_ost2 || error "Unable to start second ost"
2567 mount_client $MOUNT || error "Unable to mount client"
2570 [ -n "$ENABLE_QUOTA" ] && { $LFS quotacheck -ug $MOUNT || error "quotacheck has failed" ; }
2574 run_test 56 "check big indexes"
2576 test_57() { # bug 22656
2577 local NID=$(do_facet ost1 "$LCTL get_param nis" | tail -1 | awk '{print $1}')
2579 do_facet ost1 "$TUNEFS --failnode=$NID `ostdevname 1`" || error "tunefs failed"
2581 start_ost && error "OST registration from failnode should fail"
2585 run_test 57 "initial registration from failnode should fail (should return errs)"
2588 do_facet mgs $LCTL get_param mgs.MGS.live.$FSNAME | grep OST | wc -l
2592 start_mgsmds >> /dev/null
2593 local C1=$(count_osts)
2594 if [ $C1 -eq 0 ]; then
2595 start_ost >> /dev/null
2599 echo "original ost count: $C1 (expect > 0)"
2600 [ $C1 -gt 0 ] || error "No OSTs in $FSNAME log"
2601 start_mgsmds -o writeconf >> /dev/null || error "MDT start failed"
2602 local C2=$(count_osts)
2603 echo "after mdt writeconf count: $C2 (expect 0)"
2604 [ $C2 -gt 0 ] && error "MDT writeconf should erase OST logs"
2605 echo "OST start without writeconf should fail:"
2606 start_ost >> /dev/null && error "OST start without writeconf didn't fail"
2607 echo "OST start with writeconf should succeed:"
2608 start_ost -o writeconf >> /dev/null || error "OST1 start failed"
2609 local C3=$(count_osts)
2610 echo "after ost writeconf count: $C3 (expect 1)"
2611 [ $C3 -eq 1 ] || error "new OST writeconf should add:"
2612 start_ost2 -o writeconf >> /dev/null || error "OST2 start failed"
2613 local C4=$(count_osts)
2614 echo "after ost2 writeconf count: $C4 (expect 2)"
2615 [ $C4 -eq 2 ] || error "OST2 writeconf should add log"
2616 stop_ost2 >> /dev/null
2617 cleanup_nocli >> /dev/null
2619 run_test 59 "writeconf mount option"
2622 test_58() { # bug 22658
2623 [ "$FSTYPE" != "ldiskfs" ] && skip "not supported for $FSTYPE" && return
2626 createmany -o $DIR/$tdir/$tfile-%d 100
2627 # make sure that OSTs do not cancel llog cookies before we unmount the MDS
2628 #define OBD_FAIL_OBD_LOG_CANCEL_NET 0x601
2629 do_facet mds "lctl set_param fail_loc=0x601"
2630 unlinkmany $DIR/$tdir/$tfile-%d 100
2632 local MNTDIR=$(facet_mntpt mds)
2633 # remove all files from the OBJECTS dir
2634 do_facet mds "mount -t ldiskfs $MDSDEV $MNTDIR"
2635 do_facet mds "find $MNTDIR/OBJECTS -type f -delete"
2636 do_facet mds "umount $MNTDIR"
2637 # restart MDS with missing llog files
2639 do_facet mds "lctl set_param fail_loc=0"
2642 run_test 58 "missing llog files must not prevent MDT from mounting"
2644 if ! combined_mgs_mds ; then
2648 equals_msg `basename $0`: test complete
2649 [ -f "$TESTSUITELOG" ] && cat $TESTSUITELOG && grep -q FAIL $TESTSUITELOG && exit 1 || true