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
197 stop_ost || return 202
198 stop_mds || return 201
199 unload_modules || return 203
203 umount_client $MOUNT || return 200
204 cleanup_nocli || return $?
208 do_facet client "cp /etc/passwd $DIR/a" || return 71
209 do_facet client "rm $DIR/a" || return 72
210 # make sure lustre is actually mounted (touch will block,
211 # but grep won't, so do it after)
212 do_facet client "grep $MOUNT' ' /proc/mounts > /dev/null" || return 73
213 echo "setup single mount lustre success"
217 do_facet client "touch $DIR/a" || return 71
218 do_facet client "rm $DIR/a" || return 72
219 do_facet client "touch $DIR2/a" || return 73
220 do_facet client "rm $DIR2/a" || return 74
221 echo "setup double mount lustre success"
226 if [ "$ONLY" == "setup" ]; then
231 if [ "$ONLY" == "cleanup" ]; then
238 #create single point mountpoint
244 check_mount || return 41
247 run_test 0 "single mount setup"
250 start_mds || error "MDT start failed"
252 echo "start ost second time..."
253 start_ost && error "2nd OST start should fail"
254 mount_client $MOUNT || error "client start failed"
255 check_mount || return 42
258 run_test 1 "start up ost twice (should return errors)"
262 echo "start mds second time.."
263 start_mds && error "2nd MDT start should fail"
266 check_mount || return 43
269 run_test 2 "start up mds twice (should return err)"
273 #mount.lustre returns an error if already in mtab
274 mount_client $MOUNT && error "2nd client mount should fail"
275 check_mount || return 44
278 run_test 3 "mount client twice (should return err)"
282 touch $DIR/$tfile || return 85
286 # ok for ost to fail shutdown
287 if [ 202 -ne $eno ]; then
292 run_test 4 "force cleanup ost, then cleanup"
294 test_5a() { # was test_5
296 touch $DIR/$tfile || return 1
297 fuser -m -v $MOUNT && echo "$MOUNT is in use by user space process."
299 stop_mds -f || return 2
301 # cleanup may return an error from the failed
302 # disconnects; for now I'll consider this successful
303 # if all the modules have unloaded.
307 echo "killing umount"
308 kill -TERM $UMOUNT_PID
309 echo "waiting for umount to finish"
311 if grep " $MOUNT " /proc/mounts; then
312 echo "test 5: /proc/mounts after failed umount"
316 echo "killing umount"
317 kill -TERM $UMOUNT_PID
318 echo "waiting for umount to finish"
320 grep " $MOUNT " /proc/mounts && echo "test 5: /proc/mounts after second umount" && return 11
324 # stop_mds is a no-op here, and should not fail
325 cleanup_nocli || return $?
326 # df may have lingering entry
328 # mtab may have lingering entry
332 while [ "$WAIT" -ne "$MAX_WAIT" ]; do
334 grep -q $MOUNT" " /etc/mtab || break
335 echo "Waiting /etc/mtab updated ... "
336 WAIT=$(( WAIT + sleep))
338 [ "$WAIT" -eq "$MAX_WAIT" ] && error "/etc/mtab is not updated in $WAIT secs"
339 echo "/etc/mtab updated in $WAIT secs"
341 run_test 5a "force cleanup mds, then cleanup"
345 [ -d $MOUNT ] || mkdir -p $MOUNT
346 grep " $MOUNT " /etc/mtab && echo "test 5b: mtab before mount" && return 10
347 mount_client $MOUNT && return 1
348 grep " $MOUNT " /etc/mtab && echo "test 5b: mtab after failed mount" && return 11
350 # stop_mds is a no-op here, and should not fail
351 cleanup_nocli || return $?
354 run_test 5b "mds down, cleanup after failed mount (bug 2712) (should return errs)"
359 [ -d $MOUNT ] || mkdir -p $MOUNT
360 grep " $MOUNT " /etc/mtab && echo "test 5c: mtab before mount" && return 10
361 local oldfs="${FSNAME}"
362 FSNAME="wrong.${FSNAME}"
363 mount_client $MOUNT || :
365 grep " $MOUNT " /etc/mtab && echo "test 5c: mtab after failed mount" && return 11
367 cleanup_nocli || return $?
369 run_test 5c "cleanup after failed mount (bug 2712) (should return errs)"
375 grep " $MOUNT " /etc/mtab && echo "test 5d: mtab before mount" && return 10
376 mount_client $MOUNT || return 1
378 grep " $MOUNT " /etc/mtab && echo "test 5d: mtab after unmount" && return 11
381 run_test 5d "mount with ost down"
387 #define OBD_FAIL_PTLRPC_DELAY_SEND 0x506
388 do_facet client "lctl set_param fail_loc=0x80000506"
389 grep " $MOUNT " /etc/mtab && echo "test 5e: mtab before mount" && return 10
390 mount_client $MOUNT || echo "mount failed (not fatal)"
392 grep " $MOUNT " /etc/mtab && echo "test 5e: mtab after unmount" && return 11
395 run_test 5e "delayed connect, don't crash (bug 10268)"
400 mount_client ${MOUNT} || return 87
401 touch $DIR/a || return 86
404 run_test 6 "manual umount, then mount again"
409 cleanup_nocli || return $?
411 run_test 7 "manual umount, then cleanup"
416 check_mount2 || return 45
417 umount_client $MOUNT2
420 run_test 8 "double mount setup"
425 do_facet ost1 lctl set_param debug=\'inode trace\' || return 1
426 do_facet ost1 lctl set_param subsystem_debug=\'mds ost\' || return 1
428 CHECK_PTLDEBUG="`do_facet ost1 lctl get_param -n debug`"
429 if [ "$CHECK_PTLDEBUG" ] && { \
430 [ "$CHECK_PTLDEBUG" = "trace inode warning error emerg console" ] ||
431 [ "$CHECK_PTLDEBUG" = "trace inode" ]; }; then
432 echo "lnet.debug success"
434 echo "lnet.debug: want 'trace inode', have '$CHECK_PTLDEBUG'"
437 CHECK_SUBSYS="`do_facet ost1 lctl get_param -n subsystem_debug`"
438 if [ "$CHECK_SUBSYS" ] && [ "$CHECK_SUBSYS" = "mds ost" ]; then
439 echo "lnet.subsystem_debug success"
441 echo "lnet.subsystem_debug: want 'mds ost', have '$CHECK_SUBSYS'"
444 stop_ost || return $?
446 run_test 9 "test ptldebug and subsystem for mkfs"
449 # Test 16 was to "verify that lustre will correct the mode of OBJECTS".
450 # But with new MDS stack we don't care about the mode of local objects
451 # anymore, so this test is removed. See bug 22944 for more details.
455 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
457 if [ ! -e "$MDSDEV" ]; then
458 echo "no $MDSDEV existing, so mount Lustre to create one"
460 check_mount || return 41
464 echo "Remove mds config log"
465 do_facet $SINGLEMDS "$DEBUGFS -w -R 'unlink CONFIGS/$FSNAME-MDT0000' $MDSDEV || return \$?" || return $?
468 start_mds && return 42
471 run_test 17 "Verify failed mds_postsetup won't fail assertion (2936) (should return errs)"
474 [ "$FSTYPE" != "ldiskfs" ] && skip "not needed for FSTYPE=$FSTYPE" && return
476 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
481 # check if current MDSSIZE is large enough
482 [ $MDSSIZE -ge $MIN ] && OK=1 && myMDSSIZE=$MDSSIZE && \
483 log "use MDSSIZE=$MDSSIZE"
485 # check if the global config has a large enough MDSSIZE
486 [ -z "$OK" -a ! -z "$STORED_MDSSIZE" ] && [ $STORED_MDSSIZE -ge $MIN ] && \
487 OK=1 && myMDSSIZE=$STORED_MDSSIZE && \
488 log "use STORED_MDSSIZE=$STORED_MDSSIZE"
490 # check if the block device is large enough
491 [ -z "$OK" -a -b $MDSDEV ] && \
492 [ "$(dd if=$MDSDEV of=/dev/null bs=1k count=1 skip=$MIN 2>&1 |
493 awk '($3 == "in") { print $1 }')" = "1+0" ] && OK=1 && \
494 myMDSSIZE=$MIN && log "use device $MDSDEV with MIN=$MIN"
496 # check if a loopback device has enough space for fs metadata (5%)
497 [ -z "$OK" ] && [ -f $MDSDEV -o ! -e $MDSDEV ] &&
498 SPACE=$(df -P $(dirname $MDSDEV) |
499 awk '($1 != "Filesystem") {print $4}') &&
500 [ $SPACE -gt $((MIN / 20)) ] && OK=1 && myMDSSIZE=$MIN && \
501 log "use file $MDSDEV with MIN=$MIN"
503 [ -z "$OK" ] && skip_env "$MDSDEV too small for ${MIN}kB MDS" && return
506 echo "mount mds with large journal..."
507 local OLD_MDS_MKFS_OPTS=$MDS_MKFS_OPTS
509 MDS_MKFS_OPTS="--mgs --mdt --fsname=$FSNAME --device-size=$myMDSSIZE --param sys.timeout=$TIMEOUT $MDSOPT"
512 echo "mount lustre system..."
514 check_mount || return 41
516 echo "check journal size..."
517 local FOUNDSIZE=`do_facet mds "$DEBUGFS -c -R 'stat <8>' $MDSDEV" | awk '/Size: / { print $NF; exit;}'`
518 if [ $FOUNDSIZE -gt $((32 * 1024 * 1024)) ]; then
519 log "Success: mkfs creates large journals. Size: $((FOUNDSIZE >> 20))M"
521 error "expected journal size > 32M, found $((FOUNDSIZE >> 20))M"
526 MDS_MKFS_OPTS=$OLD_MDS_MKFS_OPTS
529 run_test 18 "check mkfs creates large journals"
532 start_mds || return 1
533 stop_mds -f || return 2
535 run_test 19a "start/stop MDS without OSTs"
538 start_ost || return 1
539 stop_ost -f || return 2
541 run_test 19b "start/stop OSTs without MDS"
544 # first format the ost/mdt
548 check_mount || return 43
550 remount_client ro $MOUNT || return 44
551 touch $DIR/$tfile && echo "$DIR/$tfile created incorrectly" && return 45
552 [ -e $DIR/$tfile ] && echo "$DIR/$tfile exists incorrectly" && return 46
553 remount_client rw $MOUNT || return 47
555 [ ! -f $DIR/$tfile ] && echo "$DIR/$tfile missing" && return 48
556 MCNT=`grep -c $MOUNT /etc/mtab`
557 [ "$MCNT" -ne 1 ] && echo "$MOUNT in /etc/mtab $MCNT times" && return 49
562 run_test 20 "remount ro,rw mounts work and doesn't break /etc/mtab"
567 wait_osc_import_state mds ost FULL
571 run_test 21a "start mds before ost, stop ost first"
576 wait_osc_import_state mds ost FULL
580 run_test 21b "start ost before mds, stop mds first"
586 wait_osc_import_state mds ost2 FULL
590 #writeconf to remove all ost2 traces for subsequent tests
593 run_test 21c "start mds between two osts, stop mds last"
598 echo Client mount with ost in logs, but none running
600 # wait until mds connected to ost and open client connection
601 wait_osc_import_state mds ost FULL
604 # check_mount will block trying to contact ost
605 mcreate $DIR/$tfile || return 40
606 rm -f $DIR/$tfile || return 42
610 echo Client mount with a running ost
613 # if gss enabled, wait full time to let connection from
614 # mds to ost be established, due to the mismatch between
615 # initial connect timeout and gss context negotiation timeout.
616 # This perhaps could be remove after AT landed.
617 echo "sleep $((TIMEOUT + TIMEOUT + TIMEOUT))s"
618 sleep $((TIMEOUT + TIMEOUT + TIMEOUT))
621 check_mount || return 41
626 run_test 22 "start a client before osts (should return errs)"
628 test_23a() { # was test_23
632 # force down client so that recovering mds waits for reconnect
633 local running=$(grep -c $MOUNT /proc/mounts) || true
634 if [ $running -ne 0 ]; then
635 echo "Stopping client $MOUNT (opts: -f)"
639 # enter recovery on mds
641 # try to start a new client
642 mount_client $MOUNT &
644 MOUNT_PID=$(ps -ef | grep "t lustre" | grep -v grep | awk '{print $2}')
645 MOUNT_LUSTRE_PID=`ps -ef | grep mount.lustre | grep -v grep | awk '{print $2}'`
646 echo mount pid is ${MOUNT_PID}, mount.lustre pid is ${MOUNT_LUSTRE_PID}
648 ps --ppid $MOUNT_LUSTRE_PID
649 # FIXME why o why can't I kill these? Manual "ctrl-c" works...
650 kill -TERM $MOUNT_LUSTRE_PID
651 echo "waiting for mount to finish"
653 # we can not wait $MOUNT_PID because it is not a child of this shell
659 while [ "$WAIT" -lt "$MAX_WAIT" ]; do
661 PID1=$(ps -ef | awk '{print $2}' | grep -w $MOUNT_PID)
662 PID2=$(ps -ef | awk '{print $2}' | grep -w $MOUNT_LUSTRE_PID)
665 [ -z "$PID1" -a -z "$PID2" ] && break
666 echo "waiting for mount to finish ... "
667 WAIT=$(( WAIT + sleep))
669 if [ "$WAIT" -eq "$MAX_WAIT" ]; then
670 error "MOUNT_PID $MOUNT_PID and "\
671 "MOUNT_LUSTRE_PID $MOUNT_LUSTRE_PID still not killed in $WAIT secs"
677 run_test 23a "interrupt client during recovery mount delay"
682 test_23b() { # was test_23
685 # Simulate -EINTR during mount OBD_FAIL_LDLM_CLOSE_THREAD
686 lctl set_param fail_loc=0x80000313
690 run_test 23b "Simulate -EINTR during mount"
692 fs2mds_HOST=$mds_HOST
693 fs2ost_HOST=$ost_HOST
697 echo "umount $MOUNT2 ..."
698 umount $MOUNT2 || true
699 echo "stopping fs2mds ..."
700 stop fs2mds -f || true
701 echo "stopping fs2ost ..."
702 stop fs2ost -f || true
710 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
712 [ -n "$ost1_HOST" ] && fs2ost_HOST=$ost1_HOST
713 if [ -z "$fs2ost_DEV" -o -z "$fs2mds_DEV" ]; then
714 do_facet $SINGLEMDS [ -b "$MDSDEV" ] && \
715 skip_env "mixed loopback and real device not working" && return
718 local fs2mdsdev=${fs2mds_DEV:-${MDSDEV}_2}
719 local fs2ostdev=${fs2ost_DEV:-$(ostdevname 1)_2}
721 # test 8-char fsname as well
722 local FSNAME2=test1234
723 add fs2mds $MDS_MKFS_OPTS --fsname=${FSNAME2} --nomgs --mgsnode=$MGSNID --reformat $fs2mdsdev || exit 10
725 add fs2ost $OST_MKFS_OPTS --fsname=${FSNAME2} --reformat $fs2ostdev || exit 10
728 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS && trap cleanup_24a EXIT INT
729 start fs2ost $fs2ostdev $OST_MOUNT_OPTS
731 mount -t lustre $MGSNID:/${FSNAME2} $MOUNT2 || return 1
733 check_mount || return 2
734 # files written on 1 should not show up on 2
735 cp /etc/passwd $DIR/$tfile
737 [ -e $MOUNT2/$tfile ] && error "File bleed" && return 7
740 cp /etc/passwd $MOUNT2/b || return 3
741 rm $MOUNT2/b || return 4
742 # 2 is actually mounted
743 grep $MOUNT2' ' /proc/mounts > /dev/null || return 5
745 facet_failover fs2mds
746 facet_failover fs2ost
749 # the MDS must remain up until last MDT
751 MDS=$(do_facet $SINGLEMDS "lctl get_param -n devices" | awk '($3 ~ "mdt" && $4 ~ "MDT") { print $4 }' | head -1)
752 [ -z "$MDS" ] && error "No MDT" && return 8
754 cleanup_nocli || return 6
756 run_test 24a "Multiple MDTs on a single node"
759 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
761 if [ -z "$fs2mds_DEV" ]; then
762 do_facet $SINGLEMDS [ -b "$MDSDEV" ] && \
763 skip_env "mixed loopback and real device not working" && return
766 local fs2mdsdev=${fs2mds_DEV:-${MDSDEV}_2}
768 add fs2mds $MDS_MKFS_OPTS --fsname=${FSNAME}2 --mgs --reformat $fs2mdsdev || exit 10
770 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS && return 2
773 run_test 24b "Multiple MGSs on a single node (should return err)"
777 check_mount || return 2
778 local MODULES=$($LCTL modules | awk '{ print $2 }')
779 rmmod $MODULES 2>/dev/null || true
782 run_test 25 "Verify modules are referenced"
786 # we need modules before mount for sysctl, so make sure...
787 do_facet $SINGLEMDS "lsmod | grep -q lustre || modprobe lustre"
788 #define OBD_FAIL_MDS_FS_SETUP 0x135
789 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000135"
790 start_mds && echo MDS started && return 1
791 lctl get_param -n devices
792 DEVS=$(lctl get_param -n devices | wc -l)
793 [ $DEVS -gt 0 ] && return 2
794 unload_modules || return 203
796 run_test 26 "MDT startup failure cleans LOV (should return errs)"
802 local ORIG=$(do_facet $myfacet "$TEST")
803 if [ $# -gt 3 ]; then
809 echo "Setting $PARAM from $ORIG to $FINAL"
810 do_facet $SINGLEMDS "$LCTL conf_param $PARAM='$FINAL'" || error conf_param failed
812 wait_update $(facet_host $myfacet) "$TEST" "$FINAL" || error check failed!
816 start_ost || return 1
817 start_mds || return 2
818 echo "Requeue thread should have started: "
819 ps -e | grep ll_cfg_requeue
820 set_and_check ost1 "lctl get_param -n obdfilter.$FSNAME-OST0000.client_cache_seconds" "$FSNAME-OST0000.ost.client_cache_seconds" || return 3
823 run_test 27a "Reacquire MGS lock if OST started first"
828 local device=$(do_facet $SINGLEMDS "lctl get_param -n devices" | awk '($3 ~ "mdt" && $4 ~ "MDT") { print $4 }')
830 facet_failover $SINGLEMDS
831 set_and_check $SINGLEMDS "lctl get_param -n mdt.$device.identity_acquire_expire" "$device.mdt.identity_acquire_expire" || return 3
832 set_and_check client "lctl get_param -n mdc.$device-mdc-*.max_rpcs_in_flight" "$device.mdc.max_rpcs_in_flight" || return 4
836 run_test 27b "Reacquire MGS lock after failover"
840 TEST="lctl get_param -n llite.$FSNAME-*.max_read_ahead_whole_mb"
841 PARAM="llite.$FSNAME.max_read_ahead_whole_mb"
844 set_and_check client "$TEST" "$PARAM" $FINAL || return 3
845 FINAL=$(($FINAL + 1))
846 set_and_check client "$TEST" "$PARAM" $FINAL || return 4
847 umount_client $MOUNT || return 200
850 if [ $RESULT -ne $FINAL ]; then
851 echo "New config not seen: wanted $FINAL got $RESULT"
854 echo "New config success: got $RESULT"
856 set_and_check client "$TEST" "$PARAM" $ORIG || return 5
859 run_test 28a "permanent parameter setting"
864 set_and_check $NODE "$LCTL get_param -n $1*.$2 | head -1" "$1.$2" "$3" || \
865 error "conf_param $1.$2 failed"
871 do_facet mgs "$LCTL conf_param foo=1 2>/dev/null" && \
872 error "Bad format should fail"
873 do_facet mgs "$LCTL conf_param osc.notanfs-OST0000.active=0 2>/dev/null" && \
874 error "Setting on unknown fs should fail"
875 do_facet mgs "$LCTL conf_param osc.$FSNAME-OST00000.active=0 2>/dev/null" && \
876 error "Bad target name should fail"
878 check_28b mds mdt.$FSNAME-MDT0000 capa_timeout 1500
879 check_28b mds mdt.$FSNAME-MDT* identity_expire 150
880 check_28b mds mdd.$FSNAME-MDT0000 atime_diff 15
881 check_28b mds mdd.$FSNAME-MDT* sync_permission 0
882 check_28b ost1 obdfilter.$FSNAME-OST0000 client_cache_seconds 15
883 check_28b ost1 obdfilter.$FSNAME-OST* client_cache_count 15
884 check_28b mds lov.$FSNAME-MDT0000 qos_maxage "15 Sec"
885 check_28b mds lov.$FSNAME-MDT0000 qos_prio_free "15%"
886 check_28b client mdc.$FSNAME-MDT0000 max_rpcs_in_flight 15
887 check_28b client osc.$FSNAME-OST0000 active 0
888 check_28b client osc.$FSNAME-OST0000 active 1
889 check_28b client osc.$FSNAME-OST0000 max_dirty_mb 15
890 check_28b client llite.$FSNAME max_read_ahead_mb 15
891 set_and_check client "$LCTL get_param -n at_max" "sys.$FSNAME.at_max" 1500 || \
892 error "conf_param sys.fsname.at_max failed"
895 run_test 28b "permanent parameter setting, set_param syntax"
898 [ "$OSTCOUNT" -lt "2" ] && skip_env "$OSTCOUNT < 2, skipping" && return
899 setup > /dev/null 2>&1
903 local PARAM="osc.$FSNAME-OST0001.active"
904 local PROC_ACT="osc.$FSNAME-OST0001-osc-[^M]*.active"
905 local PROC_UUID="osc.$FSNAME-OST0001-osc-[^M]*.ost_server_uuid"
907 ACTV=$(lctl get_param -n $PROC_ACT)
909 set_and_check client "lctl get_param -n $PROC_ACT" "$PARAM" $DEAC || return 2
910 # also check ost_server_uuid status
911 RESULT=$(lctl get_param -n $PROC_UUID | grep DEACTIV)
912 if [ -z "$RESULT" ]; then
913 echo "Live client not deactivated: $(lctl get_param -n $PROC_UUID)"
916 echo "Live client success: got $RESULT"
920 local mdtosc=$(get_mdtosc_proc_path $SINGLEMDS $FSNAME-OST0001)
921 mdtosc=${mdtosc/-MDT*/-MDT\*}
922 local MPROC="osc.$mdtosc.active"
927 RESULT=`do_facet $SINGLEMDS " lctl get_param -n $MPROC"`
928 [ ${PIPESTATUS[0]} = 0 ] || error "Can't read $MPROC"
929 if [ $RESULT -eq $DEAC ]; then
930 echo "MDT deactivated also after $WAIT sec (got $RESULT)"
934 if [ $WAIT -eq $MAX ]; then
935 echo "MDT not deactivated: wanted $DEAC got $RESULT"
938 echo "Waiting $(($MAX - $WAIT)) secs for MDT deactivated"
941 # test new client starts deactivated
942 umount_client $MOUNT || return 200
944 RESULT=$(lctl get_param -n $PROC_UUID | grep DEACTIV | grep NEW)
945 if [ -z "$RESULT" ]; then
946 echo "New client not deactivated from start: $(lctl get_param -n $PROC_UUID)"
949 echo "New client success: got $RESULT"
952 # make sure it reactivates
953 set_and_check client "lctl get_param -n $PROC_ACT" "$PARAM" $ACTV || return 6
958 #writeconf to remove all ost2 traces for subsequent tests
961 run_test 29 "permanently remove an OST"
967 TEST="lctl get_param -n llite.$FSNAME-*.max_read_ahead_whole_mb"
969 LIST=(1 2 3 4 5 4 3 2 1 2 3 4 5 4 3 2 1 2 3 4 5)
970 for i in ${LIST[@]}; do
971 set_and_check client "$TEST" "llite.$FSNAME.max_read_ahead_whole_mb" $i || return 3
973 # make sure client restart still works
975 mount_client $MOUNT || return 4
976 [ "$($TEST)" -ne "$i" ] && error "Param didn't stick across restart $($TEST) != $i"
979 echo Erase parameter setting
980 do_facet mgs "$LCTL conf_param -d llite.$FSNAME.max_read_ahead_whole_mb" || return 6
982 mount_client $MOUNT || return 6
984 echo "deleted (default) value=$FINAL, orig=$ORIG"
985 # assumes this parameter started at the default value
986 [ "$FINAL" -eq "$ORIG" ] || fail "Deleted value=$FINAL, orig=$ORIG"
990 run_test 30a "Big config llog and conf_param deletion"
995 # Make a fake nid. Use the OST nid, and add 20 to the least significant
996 # numerical part of it. Hopefully that's not already a failover address for
998 OSTNID=$(do_facet ost1 "$LCTL get_param nis" | tail -1 | awk '{print $1}')
999 ORIGVAL=$(echo $OSTNID | egrep -oi "[0-9]*@")
1000 NEWVAL=$((($(echo $ORIGVAL | egrep -oi "[0-9]*") + 20) % 256))
1001 NEW=$(echo $OSTNID | sed "s/$ORIGVAL/$NEWVAL@/")
1002 echo "Using fake nid $NEW"
1004 TEST="$LCTL get_param -n osc.$FSNAME-OST0000-osc-[^M]*.import | grep failover_nids | sed -n 's/.*\($NEW\).*/\1/p'"
1005 set_and_check client "$TEST" "osc.$FSNAME-OST0000.failover.node" $NEW || error "didn't add failover nid $NEW"
1006 NIDS=$($LCTL get_param -n osc.$FSNAME-OST0000-osc-[^M]*.import | grep failover_nids)
1008 NIDCOUNT=$(($(echo "$NIDS" | wc -w) - 1))
1009 echo "should have 2 failover nids: $NIDCOUNT"
1010 [ $NIDCOUNT -eq 2 ] || error "Failover nid not added"
1011 do_facet mgs "$LCTL conf_param -d osc.$FSNAME-OST0000.failover.node" || error "conf_param delete failed"
1012 umount_client $MOUNT
1013 mount_client $MOUNT || return 3
1015 NIDS=$($LCTL get_param -n osc.$FSNAME-OST0000-osc-[^M]*.import | grep failover_nids)
1017 NIDCOUNT=$(($(echo "$NIDS" | wc -w) - 1))
1018 echo "only 1 final nid should remain: $NIDCOUNT"
1019 [ $NIDCOUNT -eq 1 ] || error "Failover nids not removed"
1023 run_test 30b "Remove failover nids"
1025 test_31() { # bug 10734
1026 # ipaddr must not exist
1027 mount -t lustre 4.3.2.1@tcp:/lustre $MOUNT || true
1030 run_test 31 "Connect to non-existent node (shouldn't crash)"
1032 # Use these start32/stop32 fn instead of t-f start/stop fn,
1033 # for local devices, to skip global facet vars init
1037 echo "Stopping local ${MOUNT%/*}/${facet} (opts:$@)"
1038 umount -d $@ ${MOUNT%/*}/${facet}
1047 mkdir -p ${MOUNT%/*}/${facet}
1049 echo "Starting local ${facet}: $@ $device ${MOUNT%/*}/${facet}"
1050 mount -t lustre $@ ${device} ${MOUNT%/*}/${facet}
1052 if [ $RC -ne 0 ]; then
1053 echo "mount -t lustre $@ ${device} ${MOUNT%/*}/${facet}"
1054 echo "Start of ${device} of local ${facet} failed ${RC}"
1060 cleanup_nocli32 () {
1068 echo "Cleanup test_32 umount $MOUNT ..."
1069 umount -f $MOUNT || true
1070 echo "Cleanup local mds ost1 ..."
1076 client_only && skip "client only testing" && return 0
1077 [ "$NETTYPE" = "tcp" ] || { skip "NETTYPE != tcp" && return 0; }
1078 [ -z "$TUNEFS" ] && skip_env "No tunefs" && return 0
1080 local DISK1_8=$LUSTRE/tests/disk1_8.tar.bz2
1081 [ ! -r $DISK1_8 ] && skip_env "Cannot find $DISK1_8" && return 0
1082 local tmpdir=$TMP/conf32a
1084 tar xjvf $DISK1_8 -C $tmpdir || \
1085 { skip_env "Cannot untar $DISK1_8" && return 0; }
1088 $LCTL set_param debug=$PTLDEBUG
1090 $TUNEFS $tmpdir/mds || error "tunefs failed"
1092 # nids are wrong, so client wont work, but server should start
1093 start32 mds1 $tmpdir/mds "-o loop,exclude=lustre-OST0000" && \
1094 trap cleanup_32 EXIT INT || return 3
1096 local UUID=$($LCTL get_param -n mdt.lustre-MDT0000.uuid)
1098 [ "$UUID" == "lustre-MDT0000_UUID" ] || error "UUID is wrong: $UUID"
1100 $TUNEFS --mgsnode=$HOSTNAME $tmpdir/ost1 || error "tunefs failed"
1101 start32 ost1 $tmpdir/ost1 "-o loop" || return 5
1102 UUID=$($LCTL get_param -n obdfilter.lustre-OST0000.uuid)
1104 [ "$UUID" == "lustre-OST0000_UUID" ] || error "UUID is wrong: $UUID"
1106 local NID=$($LCTL list_nids | head -1)
1108 echo "OSC changes should succeed:"
1109 $LCTL conf_param lustre-OST0000.osc.max_dirty_mb=15 || return 7
1110 $LCTL conf_param lustre-OST0000.failover.node=$NID || return 8
1113 echo "MDC changes should succeed:"
1114 $LCTL conf_param lustre-MDT0000.mdc.max_rpcs_in_flight=9 || return 9
1115 $LCTL conf_param lustre-MDT0000.failover.node=$NID || return 10
1118 echo "LOV changes should succeed:"
1119 $LCTL pool_new lustre.interop || return 11
1120 $LCTL conf_param lustre-MDT0000.lov.stripesize=4M || return 12
1125 # mount a second time to make sure we didnt leave upgrade flag on
1127 $TUNEFS --dryrun $tmpdir/mds || error "tunefs failed"
1128 start32 mds1 $tmpdir/mds "-o loop,exclude=lustre-OST0000" && \
1129 trap cleanup_32 EXIT INT || return 12
1133 rm -rf $tmpdir || true # true is only for TMP on NFS
1135 run_test 32a "Upgrade from 1.8 (not live)"
1138 client_only && skip "client only testing" && return 0
1139 [ "$NETTYPE" = "tcp" ] || { skip "NETTYPE != tcp" && return 0; }
1140 [ -z "$TUNEFS" ] && skip_env "No tunefs" && return
1142 local DISK1_8=$LUSTRE/tests/disk1_8.tar.bz2
1143 [ ! -r $DISK1_8 ] && skip_env "Cannot find $DISK1_8" && return 0
1144 local tmpdir=$TMP/conf32b
1146 tar xjvf $DISK1_8 -C $tmpdir || \
1147 { skip_env "Cannot untar $DISK1_8" && return ; }
1150 $LCTL set_param debug="config"
1151 local NEWNAME=lustre
1153 # writeconf will cause servers to register with their current nids
1154 $TUNEFS --writeconf --fsname=$NEWNAME $tmpdir/mds || error "tunefs failed"
1155 start32 mds1 $tmpdir/mds "-o loop" && \
1156 trap cleanup_32 EXIT INT || return 3
1158 local UUID=$($LCTL get_param -n mdt.${NEWNAME}-MDT0000.uuid)
1160 [ "$UUID" == "${NEWNAME}-MDT0000_UUID" ] || error "UUID is wrong: $UUID"
1162 $TUNEFS --mgsnode=$HOSTNAME --writeconf --fsname=$NEWNAME $tmpdir/ost1 ||\
1163 error "tunefs failed"
1164 start32 ost1 $tmpdir/ost1 "-o loop" || return 5
1165 UUID=$($LCTL get_param -n obdfilter.${NEWNAME}-OST0000.uuid)
1167 [ "$UUID" == "${NEWNAME}-OST0000_UUID" ] || error "UUID is wrong: $UUID"
1169 local NID=$($LCTL list_nids | head -1)
1171 echo "OSC changes should succeed:"
1172 $LCTL conf_param ${NEWNAME}-OST0000.osc.max_dirty_mb=15 || return 7
1173 $LCTL conf_param ${NEWNAME}-OST0000.failover.node=$NID || return 8
1176 echo "MDC changes should succeed:"
1177 $LCTL conf_param ${NEWNAME}-MDT0000.mdc.max_rpcs_in_flight=9 || return 9
1178 $LCTL conf_param ${NEWNAME}-MDT0000.failover.node=$NID || return 10
1181 echo "LOV changes should succeed:"
1182 $LCTL pool_new ${NEWNAME}.interop || return 11
1183 $LCTL conf_param ${NEWNAME}-MDT0000.lov.stripesize=4M || return 12
1186 # MDT and OST should have registered with new nids, so we should have
1187 # a fully-functioning client
1188 echo "Check client and old fs contents"
1190 local device=`h2$NETTYPE $HOSTNAME`:/$NEWNAME
1191 echo "Starting local client: $HOSTNAME: $device $MOUNT"
1192 mount -t lustre $device $MOUNT || return 1
1194 local old=$($LCTL get_param -n mdc.*.max_rpcs_in_flight)
1195 local new=$((old + 5))
1196 $LCTL conf_param ${NEWNAME}-MDT0000.mdc.max_rpcs_in_flight=$new
1197 wait_update $HOSTNAME "$LCTL get_param -n mdc.*.max_rpcs_in_flight" $new || return 11
1199 [ "$(cksum $MOUNT/passwd | cut -d' ' -f 1,2)" == "94306271 1478" ] || return 12
1204 rm -rf $tmpdir || true # true is only for TMP on NFS
1206 run_test 32b "Upgrade from 1.8 with writeconf"
1208 test_33a() { # bug 12333, was test_33
1210 local FSNAME2=test-123
1211 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
1213 [ -n "$ost1_HOST" ] && fs2ost_HOST=$ost1_HOST
1215 if [ -z "$fs2ost_DEV" -o -z "$fs2mds_DEV" ]; then
1216 do_facet $SINGLEMDS [ -b "$MDSDEV" ] && \
1217 skip_env "mixed loopback and real device not working" && return
1220 local fs2mdsdev=${fs2mds_DEV:-${MDSDEV}_2}
1221 local fs2ostdev=${fs2ost_DEV:-$(ostdevname 1)_2}
1222 add fs2mds $MDS_MKFS_OPTS --mkfsoptions='\"-J size=8\"' --fsname=${FSNAME2} --reformat $fs2mdsdev || exit 10
1223 add fs2ost $OST_MKFS_OPTS --fsname=${FSNAME2} --index=8191 --mgsnode=$MGSNID --reformat $fs2ostdev || exit 10
1225 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS && trap cleanup_24a EXIT INT
1226 start fs2ost $fs2ostdev $OST_MOUNT_OPTS
1227 do_facet mgs "$LCTL conf_param sys.$FSNAME2.timeout=200" || rc=1
1229 mount -t lustre $MGSNID:/${FSNAME2} $MOUNT2 || rc=2
1232 cp /etc/hosts $MOUNT2/ || rc=3
1233 $LFS getstripe $MOUNT2/hosts
1238 rm -rf $MOUNT2 $fs2mdsdev $fs2ostdev
1239 cleanup_nocli || rc=6
1242 run_test 33a "Mount ost with a large index number"
1244 test_33b() { # was test_34
1247 do_facet client dd if=/dev/zero of=$MOUNT/24 bs=1024k count=1
1248 # Drop lock cancelation reply during umount
1249 #define OBD_FAIL_LDLM_CANCEL 0x304
1250 do_facet client lctl set_param fail_loc=0x80000304
1251 #lctl set_param debug=-1
1252 umount_client $MOUNT
1255 run_test 33b "Drop cancel during umount"
1259 do_facet client "sh runmultiop_bg_pause $DIR/file O_c"
1260 manual_umount_client
1262 do_facet client killall -USR1 multiop
1263 if [ $rc -eq 0 ]; then
1264 error "umount not fail!"
1269 run_test 34a "umount with opened file should fail"
1274 touch $DIR/$tfile || return 1
1275 stop_mds --force || return 2
1277 manual_umount_client --force
1279 if [ $rc -ne 0 ]; then
1280 error "mtab after failed umount - rc $rc"
1286 run_test 34b "force umount with failed mds should be normal"
1290 touch $DIR/$tfile || return 1
1291 stop_ost --force || return 2
1293 manual_umount_client --force
1295 if [ $rc -ne 0 ]; then
1296 error "mtab after failed umount - rc $rc"
1302 run_test 34c "force umount with failed ost should be normal"
1304 test_35a() { # bug 12459
1307 DBG_SAVE="`lctl get_param -n debug`"
1308 lctl set_param debug="ha"
1310 log "Set up a fake failnode for the MDS"
1312 local device=$(do_facet $SINGLEMDS "lctl get_param -n devices" | awk '($3 ~ "mdt" && $4 ~ "MDT") { print $4 }' | head -1)
1313 do_facet $SINGLEMDS $LCTL conf_param mdc.${device}.failover.node=$FAKENID || return 4
1315 log "Wait for RECONNECT_INTERVAL seconds (10s)"
1318 MSG="conf-sanity.sh test_35a `date +%F%kh%Mm%Ss`"
1321 log "Stopping the MDT:"
1322 stop_mds || return 5
1324 df $MOUNT > /dev/null 2>&1 &
1326 log "Restarting the MDT:"
1327 start_mds || return 6
1328 log "Wait for df ($DFPID) ... "
1331 lctl set_param debug="$DBG_SAVE"
1333 # retrieve from the log the first server that the client tried to
1334 # contact after the connection loss
1335 $LCTL dk $TMP/lustre-log-$TESTNAME.log
1336 NEXTCONN=`awk "/${MSG}/ {start = 1;}
1337 /import_select_connection.*$device-mdc.* using connection/ {
1339 if (\\\$NF ~ /$FAKENID/)
1345 }" $TMP/lustre-log-$TESTNAME.log`
1346 [ "$NEXTCONN" != "0" ] && log "The client didn't try to reconnect to the last active server (tried ${NEXTCONN} instead)" && return 7
1348 # remove nid settings
1351 run_test 35a "Reconnect to the last active server first"
1353 test_35b() { # bug 18674
1354 remote_mds || { skip "local MDS" && return 0; }
1358 $LCTL set_param debug="ha"
1360 MSG="conf-sanity.sh test_35b `date +%F%kh%Mm%Ss`"
1363 log "Set up a fake failnode for the MDS"
1365 local device=$(do_facet mds "$LCTL get_param -n devices" | \
1366 awk '($3 ~ "mdt" && $4 ~ "MDT") { print $4 }' | head -1)
1367 do_facet mds "$LCTL conf_param mdc.${device}.failover.node=$FAKENID" || \
1370 local at_max_saved=0
1371 # adaptive timeouts may prevent seeing the issue
1372 if at_is_enabled; then
1373 at_max_saved=$(at_max_get mds)
1374 at_max_set 0 mds client
1377 mkdir -p $MOUNT/testdir
1378 touch $MOUNT/testdir/test
1380 log "Injecting EBUSY on MDS"
1381 # Setting OBD_FAIL_MDS_RESEND=0x136
1382 do_facet mds "$LCTL set_param fail_loc=0x80000136" || return 2
1384 log "Stat on a test file"
1385 stat $MOUNT/testdir/test
1387 log "Stop injecting EBUSY on MDS"
1388 do_facet mds "$LCTL set_param fail_loc=0" || return 3
1389 rm -f $MOUNT/testdir/test
1392 # restore adaptive timeout
1393 [ $at_max_saved -ne 0 ] && at_max_set $at_max_saved mds client
1395 $LCTL dk $TMP/lustre-log-$TESTNAME.log
1397 # retrieve from the log if the client has ever tried to
1398 # contact the fake server after the loss of connection
1399 FAILCONN=`awk "BEGIN {ret = 0;}
1400 /import_select_connection.*${FSNAME}-MDT0000-mdc.* using connection/ {
1402 if (\\\$NF ~ /$FAKENID/) {
1407 END {print ret}" $TMP/lustre-log-$TESTNAME.log`
1409 [ "$FAILCONN" == "0" ] && \
1410 log "ERROR: The client reconnection has not been triggered" && \
1412 [ "$FAILCONN" == "2" ] && \
1413 log "ERROR: The client tried to reconnect to the failover server while the primary was busy" && \
1417 # remove nid settings
1420 run_test 35b "Continue reconnection retries, if the active server is busy"
1424 local FSNAME2=test1234
1425 local fs3ost_HOST=$ost_HOST
1426 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
1428 [ -n "$ost1_HOST" ] && fs2ost_HOST=$ost1_HOST && fs3ost_HOST=$ost1_HOST
1431 if [ -z "$fs2ost_DEV" -o -z "$fs2mds_DEV" -o -z "$fs3ost_DEV" ]; then
1432 do_facet $SINGLEMDS [ -b "$MDSDEV" ] && \
1433 skip_env "mixed loopback and real device not working" && return
1435 [ $OSTCOUNT -lt 2 ] && skip_env "skipping test for single OST" && return
1437 [ "$ost_HOST" = "`hostname`" -o "$ost1_HOST" = "`hostname`" ] || \
1438 { skip "remote OST" && return 0; }
1440 local fs2mdsdev=${fs2mds_DEV:-${MDSDEV}_2}
1441 local fs2ostdev=${fs2ost_DEV:-$(ostdevname 1)_2}
1442 local fs3ostdev=${fs3ost_DEV:-$(ostdevname 2)_2}
1443 add fs2mds $MDS_MKFS_OPTS --fsname=${FSNAME2} --reformat $fs2mdsdev || exit 10
1444 # XXX after we support non 4K disk blocksize, change following --mkfsoptions with
1446 add fs2ost $OST_MKFS_OPTS --mkfsoptions='-b4096' --fsname=${FSNAME2} --mgsnode=$MGSNID --reformat $fs2ostdev || exit 10
1447 add fs3ost $OST_MKFS_OPTS --mkfsoptions='-b4096' --fsname=${FSNAME2} --mgsnode=$MGSNID --reformat $fs3ostdev || exit 10
1449 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS
1450 start fs2ost $fs2ostdev $OST_MOUNT_OPTS
1451 start fs3ost $fs3ostdev $OST_MOUNT_OPTS
1453 mount -t lustre $MGSNID:/${FSNAME2} $MOUNT2 || return 1
1455 sleep 5 # until 11778 fixed
1457 dd if=/dev/zero of=$MOUNT2/$tfile bs=1M count=7 || return 2
1459 BKTOTAL=`lctl get_param -n obdfilter.*.kbytestotal | awk 'BEGIN{total=0}; {total+=$1}; END{print total}'`
1460 BKFREE=`lctl get_param -n obdfilter.*.kbytesfree | awk 'BEGIN{free=0}; {free+=$1}; END{print free}'`
1461 BKAVAIL=`lctl get_param -n obdfilter.*.kbytesavail | awk 'BEGIN{avail=0}; {avail+=$1}; END{print avail}'`
1462 STRING=`df -P $MOUNT2 | tail -n 1 | awk '{print $2","$3","$4}'`
1463 DFTOTAL=`echo $STRING | cut -d, -f1`
1464 DFUSED=`echo $STRING | cut -d, -f2`
1465 DFAVAIL=`echo $STRING | cut -d, -f3`
1466 DFFREE=$(($DFTOTAL - $DFUSED))
1468 ALLOWANCE=$((64 * $OSTCOUNT))
1470 if [ $DFTOTAL -lt $(($BKTOTAL - $ALLOWANCE)) ] ||
1471 [ $DFTOTAL -gt $(($BKTOTAL + $ALLOWANCE)) ] ; then
1472 echo "**** FAIL: df total($DFTOTAL) mismatch OST total($BKTOTAL)"
1475 if [ $DFFREE -lt $(($BKFREE - $ALLOWANCE)) ] ||
1476 [ $DFFREE -gt $(($BKFREE + $ALLOWANCE)) ] ; then
1477 echo "**** FAIL: df free($DFFREE) mismatch OST free($BKFREE)"
1480 if [ $DFAVAIL -lt $(($BKAVAIL - $ALLOWANCE)) ] ||
1481 [ $DFAVAIL -gt $(($BKAVAIL + $ALLOWANCE)) ] ; then
1482 echo "**** FAIL: df avail($DFAVAIL) mismatch OST avail($BKAVAIL)"
1487 stop fs3ost -f || return 200
1488 stop fs2ost -f || return 201
1489 stop fs2mds -f || return 202
1490 rm -rf $MOUNT2 $fs2mdsdev $fs2ostdev $fs3ostdev
1491 unload_modules || return 203
1494 run_test 36 "df report consistency on OSTs with different block size"
1497 client_only && skip "client only testing" && return 0
1498 LOCAL_MDSDEV="$TMP/mdt.img"
1499 SYM_MDSDEV="$TMP/sym_mdt.img"
1501 echo "MDS : $LOCAL_MDSDEV"
1502 echo "SYMLINK : $SYM_MDSDEV"
1506 mkfs.lustre --reformat --fsname=lustre --mdt --mgs --device-size=9000 $LOCAL_MDSDEV ||
1507 error "mkfs.lustre $LOCAL_MDSDEV failed"
1508 ln -s $LOCAL_MDSDEV $SYM_MDSDEV
1510 echo "mount symlink device - $SYM_MDSDEV"
1512 mount_op=`mount -v -t lustre -o loop $SYM_MDSDEV ${MOUNT%/*}/mds 2>&1 | grep "unable to set tunable"`
1513 umount -d ${MOUNT%/*}/mds
1514 rm -f $LOCAL_MDSDEV $SYM_MDSDEV
1516 if [ -n "$mount_op" ]; then
1517 error "**** FAIL: set tunables failed for symlink device"
1521 run_test 37 "verify set tunables works for symlink device"
1523 test_38() { # bug 14222
1528 FILES=`find $SRC -type f -mtime +1 | head -n $COUNT`
1529 log "copying $(echo $FILES | wc -w) files to $DIR/$tdir"
1531 tar cf - $FILES | tar xf - -C $DIR/$tdir || \
1532 error "copying $SRC to $DIR/$tdir"
1534 umount_client $MOUNT
1536 log "rename lov_objid file on MDS"
1537 rm -f $TMP/lov_objid.orig
1539 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
1540 do_facet $SINGLEMDS "$DEBUGFS -c -R \\\"dump lov_objid $TMP/lov_objid.orig\\\" $MDSDEV"
1541 do_facet $SINGLEMDS "$DEBUGFS -w -R \\\"rm lov_objid\\\" $MDSDEV"
1543 do_facet $SINGLEMDS "od -Ax -td8 $TMP/lov_objid.orig"
1544 # check create in mds_lov_connect
1548 [ $V ] && log "verifying $DIR/$tdir/$f"
1549 diff -q $f $DIR/$tdir/$f || ERROR=y
1551 do_facet $SINGLEMDS "$DEBUGFS -c -R \\\"dump lov_objid $TMP/lov_objid.new\\\" $MDSDEV"
1552 do_facet $SINGLEMDS "od -Ax -td8 $TMP/lov_objid.new"
1553 [ "$ERROR" = "y" ] && error "old and new files are different after connect" || true
1555 # check it's updates in sync
1556 umount_client $MOUNT
1559 do_facet $SINGLEMDS dd if=/dev/zero of=$TMP/lov_objid.clear bs=4096 count=1
1560 do_facet $SINGLEMDS "$DEBUGFS -w -R \\\"rm lov_objid\\\" $MDSDEV"
1561 do_facet $SINGLEMDS "$DEBUGFS -w -R \\\"write $TMP/lov_objid.clear lov_objid\\\" $MDSDEV "
1566 [ $V ] && log "verifying $DIR/$tdir/$f"
1567 diff -q $f $DIR/$tdir/$f || ERROR=y
1569 do_facet $SINGLEMDS "$DEBUGFS -c -R \\\"dump lov_objid $TMP/lov_objid.new1\\\" $MDSDEV"
1570 do_facet $SINGLEMDS "od -Ax -td8 $TMP/lov_objid.new1"
1571 umount_client $MOUNT
1573 [ "$ERROR" = "y" ] && error "old and new files are different after sync" || true
1575 log "files compared the same"
1578 run_test 38 "MDS recreates missing lov_objid file from OST data"
1584 perl $SRCDIR/leak_finder.pl $TMP/debug 2>&1 | egrep '*** Leak:' &&
1585 error "memory leak detected" || true
1587 run_test 39 "leak_finder recognizes both LUSTRE and LNET malloc messages"
1589 test_40() { # bug 15759
1591 #define OBD_FAIL_TGT_TOOMANY_THREADS 0x706
1592 do_facet $SINGLEMDS "sysctl -w lustre.fail_loc=0x80000706"
1596 run_test 40 "race during service thread startup"
1598 test_41() { #bug 14134
1600 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
1602 start $SINGLEMDS $MDSDEV $MDS_MOUNT_OPTS -o nosvc -n
1603 start ost1 `ostdevname 1` $OST_MOUNT_OPTS
1604 start $SINGLEMDS $MDSDEV $MDS_MOUNT_OPTS -o nomgs,force
1606 mount_client $MOUNT || return 1
1609 echo "blah blah" > $MOUNT/$tfile
1612 umount_client $MOUNT
1613 stop ost1 -f || return 201
1614 stop_mds -f || return 202
1615 stop_mds -f || return 203
1616 unload_modules || return 204
1619 run_test 41 "mount mds with --nosvc and --nomgs"
1621 test_42() { #bug 14693
1623 check_mount || return 2
1624 do_facet mgs $LCTL conf_param llite.$FSNAME.some_wrong_param=10
1625 umount_client $MOUNT
1626 mount_client $MOUNT || return 1
1630 run_test 42 "invalid config param should not prevent client from mounting"
1633 [ $UID -ne 0 -o $RUNAS_ID -eq 0 ] && skip_env "run as root"
1635 chmod ugo+x $DIR || error "chmod 0 failed"
1637 "lctl get_param -n mdt.$FSNAME-MDT0000.root_squash" \
1638 "mdt.$FSNAME-MDT*.root_squash" \
1641 "lctl get_param -n mdt.$FSNAME-MDT0000.nosquash_nids" \
1642 "mdt.$FSNAME-MDT*.nosquash_nids" \
1646 # create set of test files
1648 echo "111" > $DIR/$tfile-userfile || error "write 1 failed"
1649 chmod go-rw $DIR/$tfile-userfile || error "chmod 1 failed"
1650 chown $RUNAS_ID.$RUNAS_ID $DIR/$tfile-userfile || error "chown failed"
1652 echo "222" > $DIR/$tfile-rootfile || error "write 2 failed"
1653 chmod go-rw $DIR/$tfile-rootfile || error "chmod 2 faield"
1655 mkdir $DIR/$tdir-rootdir -p || error "mkdir failed"
1656 chmod go-rwx $DIR/$tdir-rootdir || error "chmod 3 failed"
1657 touch $DIR/$tdir-rootdir/tfile-1 || error "touch failed"
1660 # check root_squash:
1661 # set root squash UID:GID to RUNAS_ID
1662 # root should be able to access only files owned by RUNAS_ID
1665 "lctl get_param -n mdt.$FSNAME-MDT0000.root_squash" \
1666 "mdt.$FSNAME-MDT*.root_squash" \
1667 "$RUNAS_ID:$RUNAS_ID"
1669 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-userfile)
1670 dd if=$DIR/$tfile-userfile 1>/dev/null 2>/dev/null || \
1671 error "$ST: root read permission is denied"
1672 echo "$ST: root read permission is granted - ok"
1675 dd conv=notrunc if=$DIR/$tfile-userfile 1>/dev/null 2>/dev/null || \
1676 error "$ST: root write permission is denied"
1677 echo "$ST: root write permission is granted - ok"
1679 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-rootfile)
1680 dd if=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null && \
1681 error "$ST: root read permission is granted"
1682 echo "$ST: root read permission is denied - ok"
1685 dd conv=notrunc of=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null && \
1686 error "$ST: root write permission is granted"
1687 echo "$ST: root write permission is denied - ok"
1689 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tdir-rootdir)
1690 rm $DIR/$tdir-rootdir/tfile-1 1>/dev/null 2>/dev/null && \
1691 error "$ST: root unlink permission is granted"
1692 echo "$ST: root unlink permission is denied - ok"
1694 touch $DIR/tdir-rootdir/tfile-2 1>/dev/null 2>/dev/null && \
1695 error "$ST: root create permission is granted"
1696 echo "$ST: root create permission is denied - ok"
1699 # check nosquash_nids:
1700 # put client's NID into nosquash_nids list,
1701 # root should be able to access root file after that
1703 local NIDLIST=$(lctl list_nids all | tr '\n' ' ')
1704 NIDLIST="2@elan $NIDLIST 192.168.0.[2,10]@tcp"
1705 NIDLIST=$(echo $NIDLIST | tr -s ' ' ' ')
1707 "lctl get_param -n mdt.$FSNAME-MDT0000.nosquash_nids" \
1708 "mdt.$FSNAME-MDT*.nosquash_nids" \
1711 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-rootfile)
1712 dd if=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null || \
1713 error "$ST: root read permission is denied"
1714 echo "$ST: root read permission is granted - ok"
1717 dd conv=notrunc of=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null || \
1718 error "$ST: root write permission is denied"
1719 echo "$ST: root write permission is granted - ok"
1721 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tdir-rootdir)
1722 rm $DIR/$tdir-rootdir/tfile-1 || \
1723 error "$ST: root unlink permission is denied"
1724 echo "$ST: root unlink permission is granted - ok"
1725 touch $DIR/$tdir-rootdir/tfile-2 || \
1726 error "$ST: root create permission is denied"
1727 echo "$ST: root create permission is granted - ok"
1731 run_test 43 "check root_squash and nosquash_nids"
1733 umount_client $MOUNT
1738 check_mount || return 2
1739 UUID=$($LCTL get_param llite.${FSNAME}*.uuid | cut -d= -f2)
1741 UUIDS=$(do_facet mds "$LCTL get_param mdt.${FSNAME}*.exports.*.uuid")
1742 for VAL in $UUIDS; do
1743 NID=$(echo $VAL | cut -d= -f1)
1744 CLUUID=$(echo $VAL | cut -d= -f2)
1745 [ "$UUID" = "$CLUUID" ] && STATS_FOUND=yes && break
1747 [ "$STATS_FOUND" = "no" ] && error "stats not found for client"
1751 run_test 44 "mounted client proc entry exists"
1755 check_mount || return 2
1760 #define OBD_FAIL_PTLRPC_LONG_UNLINK 0x50f
1761 do_facet client "lctl set_param fail_loc=0x50f"
1764 manual_umount_client --force || return 3
1765 do_facet client "lctl set_param fail_loc=0x0"
1767 mount_client $MOUNT || return 4
1771 run_test 45 "long unlink handling in ptlrpcd"
1778 umount_client $MOUNT2 || rc=$?
1779 umount_client $MOUNT || rc=$?
1780 while [ $count -gt 0 ]; do
1781 stop ost${count} -f || rc=$?
1785 cleanup_nocli || rc=$?
1786 #writeconf to remove all ost2 traces for subsequent tests
1792 echo "Testing with $OSTCOUNT OSTs"
1794 start_mds || return 1
1795 #first client should see only one ost
1796 start_ost || return 2
1797 wait_osc_import_state mds ost FULL
1799 mount_client $MOUNT || return 3
1800 trap "cleanup_46a $OSTCOUNT" EXIT ERR
1803 for (( i=2; i<=$OSTCOUNT; i++ )); do
1804 start ost$i `ostdevname $i` $OST_MOUNT_OPTS || return $((i+2))
1807 # wait until osts in sync
1808 for (( i=2; i<=$OSTCOUNT; i++ )); do
1809 wait_osc_import_state mds ost$i FULL
1813 #second client see all ost's
1815 mount_client $MOUNT2 || return 8
1816 $LFS setstripe $MOUNT2 -c -1 || return 9
1817 $LFS getstripe $MOUNT2 || return 10
1819 echo "ok" > $MOUNT2/widestripe
1820 $LFS getstripe $MOUNT2/widestripe || return 11
1821 # fill acl buffer for avoid expand lsm to them
1822 awk -F : '{if (FNR < 25) { print "u:"$1":rwx" }}' /etc/passwd | while read acl; do
1823 setfacl -m $acl $MOUNT2/widestripe
1827 stat $MOUNT/widestripe || return 12
1829 cleanup_46a $OSTCOUNT || { echo "cleanup_46a failed!" && return 13; }
1832 run_test 46a "handle ost additional - wide striped file"
1837 check_mount || return 2
1838 $LCTL set_param ldlm.namespaces.$FSNAME-*-*-*.lru_size=100
1842 for ns in $($LCTL get_param ldlm.namespaces.$FSNAME-*-*-*.lru_size); do
1843 if echo $ns | grep "MDT[[:digit:]]*"; then
1846 lrs=$(echo $ns | sed 's/.*lru_size=//')
1847 lru_size[count]=$lrs
1852 facet_failover $SINGLEMDS
1853 client_up || return 3
1856 for ns in $($LCTL get_param ldlm.namespaces.$FSNAME-*-*-*.lru_size); do
1857 if echo $ns | grep "MDT[[:digit:]]*"; then
1860 lrs=$(echo $ns | sed 's/.*lru_size=//')
1861 if ! test "$lrs" -eq "${lru_size[count]}"; then
1862 n=$(echo $ns | sed -e 's/ldlm.namespaces.//' -e 's/.lru_size=.*//')
1863 error "$n has lost lru_size: $lrs vs. ${lru_size[count]}"
1871 run_test 47 "server restart does not make client loss lru_resize settings"
1876 # reformat after this test is needed - if test will failed
1877 # we will have unkillable file at FS
1881 test_48() { # bug 17636
1884 check_mount || return 2
1886 $LFS setstripe $MOUNT -c -1 || return 9
1887 $LFS getstripe $MOUNT || return 10
1889 echo "ok" > $MOUNT/widestripe
1890 $LFS getstripe $MOUNT/widestripe || return 11
1892 trap cleanup_48 EXIT ERR
1894 # fill acl buffer for avoid expand lsm to them
1895 getent passwd | awk -F : '{ print "u:"$1":rwx" }' | while read acl; do
1896 setfacl -m $acl $MOUNT/widestripe
1899 stat $MOUNT/widestripe || return 12
1904 run_test 48 "too many acls on file"
1906 # check PARAM_SYS_LDLM_TIMEOUT option of MKFS.LUSTRE
1907 test_49() { # bug 17710
1908 local OLD_MDS_MKFS_OPTS=$MDS_MKFS_OPTS
1909 local OLD_OST_MKFS_OPTS=$OST_MKFS_OPTS
1910 local LOCAL_TIMEOUT=20
1913 OST_MKFS_OPTS="--ost --fsname=$FSNAME --device-size=$OSTSIZE --mgsnode=$MGSNID --param sys.timeout=$LOCAL_TIMEOUT --param sys.ldlm_timeout=$LOCAL_TIMEOUT $MKFSOPT $OSTOPT"
1917 check_mount || return 1
1919 echo "check ldlm_timout..."
1920 LDLM_MDS="`do_facet mds lctl get_param -n ldlm_timeout`"
1921 LDLM_OST1="`do_facet ost1 lctl get_param -n ldlm_timeout`"
1922 LDLM_CLIENT="`do_facet client lctl get_param -n ldlm_timeout`"
1924 if [ $LDLM_MDS -ne $LDLM_OST1 ] || [ $LDLM_MDS -ne $LDLM_CLIENT ]; then
1925 error "Different LDLM_TIMEOUT:$LDLM_MDS $LDLM_OST1 $LDLM_CLIENT"
1928 if [ $LDLM_MDS -ne $((LOCAL_TIMEOUT / 3)) ]; then
1929 error "LDLM_TIMEOUT($LDLM_MDS) is not correct"
1932 umount_client $MOUNT
1933 stop_ost || return 2
1934 stop_mds || return 3
1936 OST_MKFS_OPTS="--ost --fsname=$FSNAME --device-size=$OSTSIZE --mgsnode=$MGSNID --param sys.timeout=$LOCAL_TIMEOUT --param sys.ldlm_timeout=$((LOCAL_TIMEOUT - 1)) $MKFSOPT $OSTOPT"
1940 check_mount || return 7
1942 LDLM_MDS="`do_facet mds lctl get_param -n ldlm_timeout`"
1943 LDLM_OST1="`do_facet ost1 lctl get_param -n ldlm_timeout`"
1944 LDLM_CLIENT="`do_facet client lctl get_param -n ldlm_timeout`"
1946 if [ $LDLM_MDS -ne $LDLM_OST1 ] || [ $LDLM_MDS -ne $LDLM_CLIENT ]; then
1947 error "Different LDLM_TIMEOUT:$LDLM_MDS $LDLM_OST1 $LDLM_CLIENT"
1950 if [ $LDLM_MDS -ne $((LOCAL_TIMEOUT - 1)) ]; then
1951 error "LDLM_TIMEOUT($LDLM_MDS) is not correct"
1954 cleanup || return $?
1956 MDS_MKFS_OPTS=$OLD_MDS_MKFS_OPTS
1957 OST_MKFS_OPTS=$OLD_OST_MKFS_OPTS
1959 run_test 49 "check PARAM_SYS_LDLM_TIMEOUT option of MKFS.LUSTRE"
1962 # Test both statfs and lfs df and fail if either one fails
1963 multiop_bg_pause $1 f_
1966 killall -USR1 multiop
1967 [ $RC1 -ne 0 ] && log "lazystatfs multiop failed"
1968 wait $PID || { RC1=$?; log "multiop return error "; }
1975 if [ $RC2 -eq 0 ]; then
1977 log "lazystatfs df failed"
1981 [[ $RC1 -ne 0 || $RC2 -eq 0 ]] && RC=1
1987 lctl set_param llite.$FSNAME-*.lazystatfs=1
1990 lazystatfs $MOUNT || error "lazystatfs failed but no down servers"
1992 cleanup || return $?
1994 run_test 50a "lazystatfs all servers available =========================="
1998 lctl set_param llite.$FSNAME-*.lazystatfs=1
2001 # Wait for client to detect down OST
2002 stop_ost || error "Unable to stop OST1"
2003 wait_osc_import_state mds ost DISCONN
2005 lazystatfs $MOUNT || error "lazystatfs should don't have returned EIO"
2007 umount_client $MOUNT || error "Unable to unmount client"
2008 stop_mds || error "Unable to stop MDS"
2010 run_test 50b "lazystatfs all servers down =========================="
2013 start_mds || error "Unable to start MDS"
2014 start_ost || error "Unable to start OST1"
2015 start_ost2 || error "Unable to start OST2"
2016 mount_client $MOUNT || error "Unable to mount client"
2017 lctl set_param llite.$FSNAME-*.lazystatfs=1
2020 # Wait for client to detect down OST
2021 stop_ost || error "Unable to stop OST1"
2022 wait_osc_import_state mds ost DISCONN
2023 lazystatfs $MOUNT || error "lazystatfs failed with one down server"
2025 umount_client $MOUNT || error "Unable to unmount client"
2026 stop_ost2 || error "Unable to stop OST2"
2027 stop_mds || error "Unable to stop MDS"
2029 run_test 50c "lazystatfs one server down =========================="
2032 start_mds || error "Unable to start MDS"
2033 start_ost || error "Unable to start OST1"
2034 start_ost2 || error "Unable to start OST2"
2035 mount_client $MOUNT || error "Unable to mount client"
2036 lctl set_param llite.$FSNAME-*.lazystatfs=1
2039 # Issue the statfs during the window where the client still
2040 # belives the OST to be available but it is in fact down.
2041 # No failure just a statfs which hangs for a timeout interval.
2042 stop_ost || error "Unable to stop OST1"
2043 lazystatfs $MOUNT || error "lazystatfs failed with one down server"
2045 umount_client $MOUNT || error "Unable to unmount client"
2046 stop_ost2 || error "Unable to stop OST2"
2047 stop_mds || error "Unable to stop MDS"
2049 run_test 50d "lazystatfs client/server conn race =========================="
2056 start_mds || return 1
2057 #first client should see only one ost
2058 start_ost || return 2
2059 wait_osc_import_state mds ost FULL
2061 # Wait for client to detect down OST
2062 stop_ost || error "Unable to stop OST1"
2063 wait_osc_import_state mds ost DISCONN
2065 mount_client $MOUNT || error "Unable to mount client"
2066 lctl set_param llite.$FSNAME-*.lazystatfs=0
2068 multiop_bg_pause $MOUNT _f
2072 if [ $RC1 -ne 0 ]; then
2073 log "multiop failed $RC1"
2076 sleep $(( $TIMEOUT+1 ))
2078 [ $? -ne 0 ] && error "process isn't sleep"
2079 start_ost || error "Unable to start OST1"
2080 wait $pid || error "statfs failed"
2083 umount_client $MOUNT || error "Unable to unmount client"
2084 stop_ost || error "Unable to stop OST1"
2085 stop_mds || error "Unable to stop MDS"
2087 run_test 50e "normal statfs all servers down =========================="
2092 CONN_PROC="osc.$FSNAME-OST0001-osc-[M]*.ost_server_uuid"
2094 start_mds || error "Unable to start mds"
2095 #first client should see only one ost
2096 start_ost || error "Unable to start OST1"
2097 wait_osc_import_state mds ost FULL
2099 start_ost2 || error "Unable to start OST2"
2100 wait_osc_import_state mds ost2 FULL
2102 # Wait for client to detect down OST
2103 stop_ost2 || error "Unable to stop OST2"
2105 wait_osc_import_state mds ost2 DISCONN
2106 mount_client $MOUNT || error "Unable to mount client"
2107 lctl set_param llite.$FSNAME-*.lazystatfs=0
2109 multiop_bg_pause $MOUNT _f
2113 if [ $RC1 -ne 0 ]; then
2114 log "lazystatfs multiop failed $RC1"
2117 sleep $(( $TIMEOUT+1 ))
2119 [ $? -ne 0 ] && error "process isn't sleep"
2120 start_ost2 || error "Unable to start OST2"
2121 wait $pid || error "statfs failed"
2122 stop_ost2 || error "Unable to stop OST2"
2125 umount_client $MOUNT || error "Unable to unmount client"
2126 stop_ost || error "Unable to stop OST1"
2127 stop_mds || error "Unable to stop MDS"
2128 #writeconf to remove all ost2 traces for subsequent tests
2131 run_test 50f "normal statfs one server in down =========================="
2134 [ "$OSTCOUNT" -lt "2" ] && skip_env "$OSTCOUNT < 2, skipping" && return
2136 start_ost2 || error "Unable to start OST2"
2138 local PARAM="${FSNAME}-OST0001.osc.active"
2140 $LFS setstripe -c -1 $DIR/$tfile || error "Unable to lfs setstripe"
2141 do_facet mgs $LCTL conf_param $PARAM=0 || error "Unable to deactivate OST"
2143 umount_client $MOUNT || error "Unable to unmount client"
2144 mount_client $MOUNT || error "Unable to mount client"
2145 # This df should not cause a panic
2148 do_facet mgs $LCTL conf_param $PARAM=1 || error "Unable to activate OST"
2150 umount_client $MOUNT || error "Unable to unmount client"
2151 stop_ost2 || error "Unable to stop OST2"
2152 stop_ost || error "Unable to stop OST1"
2153 stop_mds || error "Unable to stop MDS"
2154 #writeconf to remove all ost2 traces for subsequent tests
2157 run_test 50g "deactivated OST should not cause panic====================="
2160 local LOCAL_TIMEOUT=20
2164 check_mount || return 1
2167 $LFS setstripe -c -1 $MOUNT/d1
2168 #define OBD_FAIL_MDS_REINT_DELAY 0x142
2169 do_facet $SINGLEMDS "lctl set_param fail_loc=0x142"
2170 touch $MOUNT/d1/f1 &
2173 start_ost2 || return 2
2175 stop_ost2 || return 3
2178 run_test 51 "Verify that mdt_reint handles RMF_MDT_MD correctly when an OST is added"
2187 do_node $node mkdir -p $dest
2188 [ $? -eq 0 ] || { error "Unable to create directory"; return 1; }
2190 do_node $node 'tar cf - '$@' | tar xf - -C '$dest';
2191 [ \"\${PIPESTATUS[*]}\" = \"0 0\" ] || exit 1'
2192 [ $? -eq 0 ] || { error "Unable to tar files"; return 2; }
2194 do_node $node 'getfattr -d -m "[a-z]*\\." '$@' > '$xattrs
2195 [ $? -eq 0 ] || { error "Unable to read xattrs"; return 3; }
2205 local backup2=${TMP}/backup2
2207 do_node $node mkdir -p $backup2
2208 [ $? -eq 0 ] || { error "Unable to create directory"; return 1; }
2210 do_node $node 'tar cf - '$@' | tar xf - -C '$backup2';
2211 [ \"\${PIPESTATUS[*]}\" = \"0 0\" ] || exit 1'
2212 [ $? -eq 0 ] || { error "Unable to tar files to diff"; return 2; }
2214 do_node $node "diff -rq $backup $backup2"
2215 [ $? -eq 0 ] || { error "contents differ"; return 3; }
2217 local xattrs2=${TMP}/xattrs2
2218 do_node $node 'getfattr -d -m "[a-z]*\\." '$@' > '$xattrs2
2219 [ $? -eq 0 ] || { error "Unable to read xattrs to diff"; return 4; }
2221 do_node $node "diff $xattrs $xattrs2"
2222 [ $? -eq 0 ] || { error "xattrs differ"; return 5; }
2224 do_node $node "rm -rf $backup2 $xattrs2"
2225 [ $? -eq 0 ] || { error "Unable to delete temporary files"; return 6; }
2230 [ $? -eq 0 ] || { error "Unable to start MDS"; return 1; }
2232 [ $? -eq 0 ] || { error "Unable to start OST1"; return 2; }
2234 [ $? -eq 0 ] || { error "Unable to mount client"; return 3; }
2237 local ost1mnt=${MOUNT%/*}/ost1
2238 local ost1node=$(facet_active_host ost1)
2239 local ost1tmp=$TMP/conf52
2242 [ $? -eq 0 ] || { error "Unable to create tdir"; return 4; }
2243 touch $TMP/modified_first
2244 [ $? -eq 0 ] || { error "Unable to create temporary file"; return 5; }
2245 do_node $ost1node "mkdir -p $ost1tmp && touch $ost1tmp/modified_first"
2246 [ $? -eq 0 ] || { error "Unable to create temporary file"; return 6; }
2249 $LFS setstripe $DIR/$tdir -c -1 -s 1M
2250 [ $? -eq 0 ] || { error "lfs setstripe failed"; return 7; }
2252 for (( i=0; i < nrfiles; i++ )); do
2253 multiop $DIR/$tdir/$tfile-$i Ow1048576w1048576w524288c
2254 [ $? -eq 0 ] || { error "multiop failed"; return 8; }
2260 echo backup files to $TMP/files
2261 local files=$(find $DIR/$tdir -type f -newer $TMP/modified_first)
2262 copy_files_xattrs `hostname` $TMP/files $TMP/file_xattrs $files
2263 [ $? -eq 0 ] || { error "Unable to copy files"; return 9; }
2265 umount_client $MOUNT
2266 [ $? -eq 0 ] || { error "Unable to umount client"; return 10; }
2268 [ $? -eq 0 ] || { error "Unable to stop ost1"; return 11; }
2270 echo mount ost1 as ldiskfs
2271 do_node $ost1node mount -t $FSTYPE $ost1_dev $ost1mnt $OST_MOUNT_OPTS
2272 [ $? -eq 0 ] || { error "Unable to mount ost1 as ldiskfs"; return 12; }
2275 echo backup objects to $ost1tmp/objects
2276 local objects=$(do_node $ost1node 'find '$ost1mnt'/O/0 -type f -size +0'\
2277 '-newer '$ost1tmp'/modified_first -regex ".*\/[0-9]+"')
2278 copy_files_xattrs $ost1node $ost1tmp/objects $ost1tmp/object_xattrs $objects
2279 [ $? -eq 0 ] || { error "Unable to copy objects"; return 13; }
2281 # move objects to lost+found
2282 do_node $ost1node 'mv '$objects' '${ost1mnt}'/lost+found'
2283 [ $? -eq 0 ] || { error "Unable to move objects"; return 14; }
2286 do_node $ost1node "ll_recover_lost_found_objs -d $ost1mnt/lost+found"
2287 [ $? -eq 0 ] || { error "ll_recover_lost_found_objs failed"; return 15; }
2289 # compare restored objects against saved ones
2290 diff_files_xattrs $ost1node $ost1tmp/objects $ost1tmp/object_xattrs $objects
2291 [ $? -eq 0 ] || { error "Unable to diff objects"; return 16; }
2293 do_node $ost1node "umount $ost1_dev"
2294 [ $? -eq 0 ] || { error "Unable to umount ost1 as ldiskfs"; return 17; }
2297 [ $? -eq 0 ] || { error "Unable to start ost1"; return 18; }
2299 [ $? -eq 0 ] || { error "Unable to mount client"; return 19; }
2302 diff_files_xattrs `hostname` $TMP/files $TMP/file_xattrs $files
2303 [ $? -eq 0 ] || { error "Unable to diff files"; return 20; }
2305 rm -rf $TMP/files $TMP/file_xattrs
2306 [ $? -eq 0 ] || { error "Unable to delete temporary files"; return 21; }
2307 do_node $ost1node "rm -rf $ost1tmp"
2308 [ $? -eq 0 ] || { error "Unable to delete temporary files"; return 22; }
2311 run_test 52 "check recovering objects from lost+found"
2313 # Checks threads_min/max/started for some service
2315 # Arguments: service name (OST or MDT), facet (e.g., ost1, $SINGLEMDS), and a
2316 # parameter pattern prefix like 'ost.*.ost'.
2328 local msg="Insane $modname thread counts"
2332 check_mount || return 41
2334 # We need to expand $parampat, but it may match multiple parameters, so
2335 # we'll pick the first one
2336 paramp=$(do_facet $facet "lctl get_param -N ${parampat}.threads_min"|head -1)
2337 if [ -z "$paramp" ]; then
2338 error "Couldn't expand ${parampat}.threads_min parameter name"
2342 # Remove the .threads_min part
2343 paramp=${paramp%.threads_min}
2345 # Check for sanity in defaults
2346 tmin=$(do_facet $facet "lctl get_param -n ${paramp}.threads_min" || echo 0)
2347 tmax=$(do_facet $facet "lctl get_param -n ${paramp}.threads_max" || echo 0)
2348 tstarted=$(do_facet $facet "lctl get_param -n ${paramp}.threads_started" || echo 0)
2349 lassert 23 "$msg (PDSH problems?)" '(($tstarted && $tmin && $tmax))' || return $?
2350 lassert 24 "$msg" '(($tstarted >= $tmin && $tstarted <= tmax ))' || return $?
2352 # Check that we can lower min/max
2353 do_facet $facet "lctl set_param ${paramp}.threads_min=$((tmin - 1))"
2354 do_facet $facet "lctl set_param ${paramp}.threads_max=$((tmax - 10))"
2355 tmin2=$(do_facet $facet "lctl get_param -n ${paramp}.threads_min" || echo 0)
2356 tmax2=$(do_facet $facet "lctl get_param -n ${paramp}.threads_max" || echo 0)
2357 lassert 25 "$msg" '(($tmin2 == ($tmin - 1) && $tmax2 == ($tmax -10)))' || return $?
2359 # Check that we can set min/max to the same value
2360 do_facet $facet "lctl set_param ${paramp}.threads_min=$tmin"
2361 do_facet $facet "lctl set_param ${paramp}.threads_max=$tmin"
2362 tmin2=$(do_facet $facet "lctl get_param -n ${paramp}.threads_min" || echo 0)
2363 tmax2=$(do_facet $facet "lctl get_param -n ${paramp}.threads_max" || echo 0)
2364 lassert 26 "$msg" '(($tmin2 == $tmin && $tmax2 == $tmin))' || return $?
2366 # Check that we can't set max < min
2367 do_facet $facet "lctl set_param ${paramp}.threads_max=$((tmin - 1))"
2368 tmin2=$(do_facet $facet "lctl get_param -n ${paramp}.threads_min" || echo 0)
2369 tmax2=$(do_facet $facet "lctl get_param -n ${paramp}.threads_max" || echo 0)
2370 lassert 27 "$msg" '(($tmin <= $tmax2))' || return $?
2372 # We need to ensure that we get the module options desired; to do this
2373 # we set LOAD_MODULES_REMOTE=true and we call setmodopts below.
2374 LOAD_MODULES_REMOTE=true
2377 setmodopts -a $modname "$opts" oldvalue
2381 check_mount || return 41
2383 # Restore previous setting of MODOPTS_*
2384 setmodopts $modname "$oldvalue"
2386 # Check that $opts took
2387 tmin=$(do_facet $facet "lctl get_param -n ${paramp}.threads_min")
2388 tmax=$(do_facet $facet "lctl get_param -n ${paramp}.threads_max")
2389 tstarted=$(do_facet $facet "lctl get_param -n ${paramp}.threads_started")
2390 lassert 28 "$msg" '(($tstarted == $tmin && $tstarted == $tmax ))' || return $?
2393 # Workaround a YALA bug where YALA expects that modules will remain
2394 # loaded on the servers
2395 LOAD_MODULES_REMOTE=false
2402 thread_sanity OST ost1 'ost.*.ost' 'oss_num_threads=64'
2404 run_test 53a "check OSS thread count params"
2407 thread_sanity MDT $SINGLEMDS 'mdt.*.*.' 'mdt_num_threads=64'
2409 run_test 53b "check MDT thread count params"
2414 local partial_arg=""
2415 local size=$(df -B G $dir | tail -1 | awk '{print $2}' | sed 's/G//') # Gb
2417 # Run in partial (fast) mode if the size
2418 # of a partition > 10 GB
2419 [ $size -gt 10 ] && partial_arg="-p"
2421 llverfs $partial_arg $dir
2425 do_rpc_nodes $(facet_host ost1) run_llverdev $(ostdevname 1)
2426 [ $? -eq 0 ] || error "llverdev failed!"
2429 run_test 54a "llverdev"
2434 [ $? -eq 0 ] || error "llverfs failed!"
2437 run_test 54b "llverfs"
2441 local max_ost_index=$1
2442 echo -n $(((max_ost_index + 1) * 8))
2446 local mdsdev=$(mdsdevname 1)
2447 local ostdev=$(ostdevname 1)
2448 local saved_opts=$OST_MKFS_OPTS
2450 for i in 0 1023 2048
2452 OST_MKFS_OPTS="$saved_opts --index $i"
2460 echo checking size of lov_objid for ost index $i
2461 LOV_OBJID_SIZE=$(do_facet mds1 "$DEBUGFS -R 'stat lov_objid' $mdsdev 2>/dev/null" | grep ^User | awk '{print $6}')
2462 if [ "$LOV_OBJID_SIZE" != $(lov_objid_size $i) ]; then
2463 error "lov_objid size has to be $(lov_objid_size $i), not $LOV_OBJID_SIZE"
2465 echo ok, lov_objid size is correct: $LOV_OBJID_SIZE
2470 OST_MKFS_OPTS=$saved_opts
2473 run_test 55 "check lov_objid size"
2476 add mds1 $MDS_MKFS_OPTS --mkfsoptions='\"-J size=16\"' --reformat $(mdsdevname 1)
2477 add ost1 $OST_MKFS_OPTS --index=1000 --reformat $(ostdevname 1)
2478 add ost2 $OST_MKFS_OPTS --index=10000 --reformat $(ostdevname 2)
2482 start_ost2 || error "Unable to start second ost"
2483 mount_client $MOUNT || error "Unable to mount client"
2489 run_test 56 "check big indexes"
2491 test_57() { # bug 22656
2492 local NID=$(do_facet ost1 "$LCTL get_param nis" | tail -1 | awk '{print $1}')
2494 do_facet ost1 "$TUNEFS --failnode=$NID `ostdevname 1`" || error "tunefs failed"
2496 start_ost && error "OST registration from failnode should fail"
2500 run_test 57 "initial registration from failnode should fail (should return errs)"
2503 do_facet mgs $LCTL get_param mgs.MGS.live.$FSNAME | grep OST | wc -l
2507 start_mgsmds >> /dev/null
2508 local C1=$(count_osts)
2509 if [ $C1 -eq 0 ]; then
2510 start_ost >> /dev/null
2514 echo "original ost count: $C1 (expect > 0)"
2515 [ $C1 -gt 0 ] || error "No OSTs in $FSNAME log"
2516 start_mgsmds -o writeconf >> /dev/null || error "MDT start failed"
2517 local C2=$(count_osts)
2518 echo "after mdt writeconf count: $C2 (expect 0)"
2519 [ $C2 -gt 0 ] && error "MDT writeconf should erase OST logs"
2520 echo "OST start without writeconf should fail:"
2521 start_ost >> /dev/null && error "OST start without writeconf didn't fail"
2522 echo "OST start with writeconf should succeed:"
2523 start_ost -o writeconf >> /dev/null || error "OST1 start failed"
2524 local C3=$(count_osts)
2525 echo "after ost writeconf count: $C3 (expect 1)"
2526 [ $C3 -eq 1 ] || error "new OST writeconf should add:"
2527 start_ost2 -o writeconf >> /dev/null || error "OST2 start failed"
2528 local C4=$(count_osts)
2529 echo "after ost2 writeconf count: $C4 (expect 2)"
2530 [ $C4 -eq 2 ] || error "OST2 writeconf should add log"
2531 stop_ost2 >> /dev/null
2532 cleanup_nocli >> /dev/null
2534 run_test 59 "writeconf mount option"
2537 if ! combined_mgs_mds ; then
2541 equals_msg `basename $0`: test complete
2542 [ -f "$TESTSUITELOG" ] && cat $TESTSUITELOG && grep -q FAIL $TESTSUITELOG && exit 1 || true