2 # -*- mode: Bash; tab-width: 4; indent-tabs-mode: t; -*-
3 # vim:autoindent:shiftwidth=4:tabstop=4:
5 # FIXME - there is no reason to use all of these different
6 # return codes, espcially when most of them are mapped to something
7 # else anyway. The combination of test number and return code
8 # figure out what failed.
14 # bug number for skipped test:
16 ALWAYS_EXCEPT="$CONF_SANITY_EXCEPT"
17 # UPDATE THE COMMENT ABOVE WITH BUG NUMBERS WHEN CHANGING ALWAYS_EXCEPT!
19 if [ "$FAILURE_MODE" = "HARD" ]; then
20 CONFIG_EXCEPTIONS="24a " && \
21 echo "Except the tests: $CONFIG_EXCEPTIONS for FAILURE_MODE=$FAILURE_MODE, bug 23573" && \
22 ALWAYS_EXCEPT="$ALWAYS_EXCEPT $CONFIG_EXCEPTIONS"
26 PATH=$PWD/$SRCDIR:$SRCDIR:$SRCDIR/../utils:$PATH
28 PTLDEBUG=${PTLDEBUG:--1}
30 LUSTRE=${LUSTRE:-`dirname $0`/..}
31 RLUSTRE=${RLUSTRE:-$LUSTRE}
33 . $LUSTRE/tests/test-framework.sh
36 # use small MDS + OST size to speed formatting time
37 # do not use too small MDSSIZE/OSTSIZE, which affect the default jouranl size
40 . ${CONFIG:=$LUSTRE/tests/cfg/$NAME.sh}
42 if ! combined_mgs_mds; then
43 # bug number for skipped test: 23954
44 ALWAYS_EXCEPT="$ALWAYS_EXCEPT 24b"
47 # STORED_MDSSIZE is used in test_18
48 if [ -n "$MDSSIZE" ]; then
49 STORED_MDSSIZE=$MDSSIZE
55 require_dsh_mds || exit 0
56 require_dsh_ost || exit 0
58 [ "$SLOW" = "no" ] && EXCEPT_SLOW="30a 31 45"
73 # who knows if/where $TUNEFS is installed? Better reformat if it fails...
74 do_facet ${facet} "$TUNEFS --quiet --writeconf $dev" ||
75 { echo "tunefs failed, reformatting instead" && reformat_and_config && return 1; }
82 # if writeconf fails anywhere, we reformat everything
83 writeconf1 mds `mdsdevname 1` || return 0
84 writeconf1 ost1 `ostdevname 1` || return 0
85 writeconf1 ost2 `ostdevname 2` || return 0
89 # The MGS must be started before the OSTs for a new fs, so start
90 # and stop to generate the startup logs.
93 wait_osc_import_state mds ost FULL
98 reformat_and_config() {
100 if ! combined_mgs_mds ; then
108 start mgs $MGSDEV $MGS_MOUNT_OPTS
112 local facet=$SINGLEMDS
113 # we can not use MDSDEV1 here because SINGLEMDS could be set not to mds1 only
114 local num=$(echo $facet | tr -d "mds")
115 local dev=$(mdsdevname $num)
116 echo "start mds service on `facet_active_host $facet`"
117 start $facet ${dev} $MDS_MOUNT_OPTS $@ || return 94
121 if ! combined_mgs_mds ; then
128 echo "stop mds service on `facet_active_host $SINGLEMDS`"
129 # These tests all use non-failover stop
130 stop $SINGLEMDS -f || return 97
134 echo "stop mgs service on `facet_active_host mgs`"
135 # These tests all use non-failover stop
136 stop mgs -f || return 97
140 echo "start ost1 service on `facet_active_host ost1`"
141 start ost1 `ostdevname 1` $OST_MOUNT_OPTS $@ || return 95
145 echo "stop ost1 service on `facet_active_host ost1`"
146 # These tests all use non-failover stop
147 stop ost1 -f || return 98
151 echo "start ost2 service on `facet_active_host ost2`"
152 start ost2 `ostdevname 2` $OST_MOUNT_OPTS $@ || return 92
156 echo "stop ost2 service on `facet_active_host ost2`"
157 # These tests all use non-failover stop
158 stop ost2 -f || return 93
163 echo "mount $FSNAME on ${MOUNTPATH}....."
164 zconf_mount `hostname` $MOUNTPATH || return 96
168 local SAVEMOUNTOPT=$MOUNTOPT
169 MOUNTOPT="remount,$1"
171 echo "remount '$1' lustre on ${MOUNTPATH}....."
172 zconf_mount `hostname` $MOUNTPATH || return 96
173 MOUNTOPT=$SAVEMOUNTOPT
178 echo "umount lustre on ${MOUNTPATH}....."
179 zconf_umount `hostname` $MOUNTPATH || return 97
182 manual_umount_client(){
185 echo "manual umount lustre on ${MOUNT}...."
186 do_facet client "umount -d ${FORCE} $MOUNT"
192 start_mds || error "MDT start failed"
193 start_ost || error "OST start failed"
194 mount_client $MOUNT || error "client start failed"
198 if ! combined_mgs_mds ; then
207 unload_modules_conf () {
208 if combined_mgs_mds || ! local_mode; then
209 unload_modules || return 1
214 stop_ost || return 202
215 stop_mds || return 201
216 unload_modules_conf || return 203
220 umount_client $MOUNT || return 200
221 cleanup_nocli || return $?
225 do_facet client "cp /etc/passwd $DIR/a" || return 71
226 do_facet client "rm $DIR/a" || return 72
227 # make sure lustre is actually mounted (touch will block,
228 # but grep won't, so do it after)
229 do_facet client "grep $MOUNT' ' /proc/mounts > /dev/null" || return 73
230 echo "setup single mount lustre success"
234 do_facet client "touch $DIR/a" || return 71
235 do_facet client "rm $DIR/a" || return 72
236 do_facet client "touch $DIR2/a" || return 73
237 do_facet client "rm $DIR2/a" || return 74
238 echo "setup double mount lustre success"
243 if [ "$ONLY" == "setup" ]; then
248 if [ "$ONLY" == "cleanup" ]; then
255 #create single point mountpoint
261 check_mount || return 41
264 run_test 0 "single mount setup"
267 start_mds || error "MDT start failed"
269 echo "start ost second time..."
270 start_ost && error "2nd OST start should fail"
271 mount_client $MOUNT || error "client start failed"
272 check_mount || return 42
275 run_test 1 "start up ost twice (should return errors)"
279 echo "start mds second time.."
280 start_mds && error "2nd MDT start should fail"
283 check_mount || return 43
286 run_test 2 "start up mds twice (should return err)"
290 #mount.lustre returns an error if already in mtab
291 mount_client $MOUNT && error "2nd client mount should fail"
292 check_mount || return 44
295 run_test 3 "mount client twice (should return err)"
299 touch $DIR/$tfile || return 85
303 # ok for ost to fail shutdown
304 if [ 202 -ne $eno ]; then
309 run_test 4 "force cleanup ost, then cleanup"
311 test_5a() { # was test_5
313 touch $DIR/$tfile || return 1
314 fuser -m -v $MOUNT && echo "$MOUNT is in use by user space process."
316 stop_mds -f || return 2
318 # cleanup may return an error from the failed
319 # disconnects; for now I'll consider this successful
320 # if all the modules have unloaded.
324 echo "killing umount"
325 kill -TERM $UMOUNT_PID
326 echo "waiting for umount to finish"
328 if grep " $MOUNT " /proc/mounts; then
329 echo "test 5: /proc/mounts after failed umount"
333 echo "killing umount"
334 kill -TERM $UMOUNT_PID
335 echo "waiting for umount to finish"
337 grep " $MOUNT " /proc/mounts && echo "test 5: /proc/mounts after second umount" && return 11
341 # stop_mds is a no-op here, and should not fail
342 cleanup_nocli || return $?
343 # df may have lingering entry
345 # mtab may have lingering entry
349 while [ "$WAIT" -ne "$MAX_WAIT" ]; do
351 grep -q $MOUNT" " /etc/mtab || break
352 echo "Waiting /etc/mtab updated ... "
353 WAIT=$(( WAIT + sleep))
355 [ "$WAIT" -eq "$MAX_WAIT" ] && error "/etc/mtab is not updated in $WAIT secs"
356 echo "/etc/mtab updated in $WAIT secs"
358 run_test 5a "force cleanup mds, then cleanup"
366 grep " $MOUNT " /etc/mtab && \
367 error false "unexpected entry in mtab before mount" && return 10
371 if ! combined_mgs_mds ; then
372 trap cleanup_5b EXIT ERR
377 [ -d $MOUNT ] || mkdir -p $MOUNT
378 mount_client $MOUNT && rc=1
379 grep " $MOUNT " /etc/mtab && \
380 error "$MOUNT entry in mtab after failed mount" && rc=11
382 # stop_mds is a no-op here, and should not fail
383 cleanup_nocli || rc=$?
384 if ! combined_mgs_mds ; then
389 run_test 5b "Try to start a client with no MGS (should return errs)"
392 grep " $MOUNT " /etc/mtab && \
393 error false "unexpected entry in mtab before mount" && return 10
398 [ -d $MOUNT ] || mkdir -p $MOUNT
399 local oldfs="${FSNAME}"
400 FSNAME="wrong.${FSNAME}"
401 mount_client $MOUNT || :
403 grep " $MOUNT " /etc/mtab && \
404 error "$MOUNT entry in mtab after failed mount" && rc=11
406 cleanup_nocli || rc=$?
409 run_test 5c "cleanup after failed mount (bug 2712) (should return errs)"
412 grep " $MOUNT " /etc/mtab && \
413 error false "unexpected entry in mtab before mount" && return 10
419 mount_client $MOUNT || rc=1
421 grep " $MOUNT " /etc/mtab && \
422 error "$MOUNT entry in mtab after unmount" && rc=11
425 run_test 5d "mount with ost down"
428 grep " $MOUNT " /etc/mtab && \
429 error false "unexpected entry in mtab before mount" && return 10
435 #define OBD_FAIL_PTLRPC_DELAY_SEND 0x506
436 do_facet client "lctl set_param fail_loc=0x80000506"
437 mount_client $MOUNT || echo "mount failed (not fatal)"
439 grep " $MOUNT " /etc/mtab && \
440 error "$MOUNT entry in mtab after unmount" && rc=11
443 run_test 5e "delayed connect, don't crash (bug 10268)"
446 if combined_mgs_mds ; then
447 skip "combined mgs and mds"
451 grep " $MOUNT " /etc/mtab && \
452 error false "unexpected entry in mtab before mount" && return 10
456 [ -d $MOUNT ] || mkdir -p $MOUNT
457 mount_client $MOUNT &
459 echo client_mount pid is $pid
463 if ! ps -f -p $pid >/dev/null; then
466 grep " $MOUNT " /etc/mtab && echo "test 5f: mtab after mount"
467 error "mount returns $rc, expected to hang"
476 # mount should succeed after start mds
479 [ $rc -eq 0 ] || error "mount returned $rc"
480 grep " $MOUNT " /etc/mtab && echo "test 5f: mtab after mount"
484 run_test 5f "mds down, cleanup after failed mount (bug 2712)"
489 mount_client ${MOUNT} || return 87
490 touch $DIR/a || return 86
493 run_test 6 "manual umount, then mount again"
498 cleanup_nocli || return $?
500 run_test 7 "manual umount, then cleanup"
505 check_mount2 || return 45
506 umount_client $MOUNT2
509 run_test 8 "double mount setup"
514 do_facet ost1 lctl set_param debug=\'inode trace\' || return 1
515 do_facet ost1 lctl set_param subsystem_debug=\'mds ost\' || return 1
517 CHECK_PTLDEBUG="`do_facet ost1 lctl get_param -n debug`"
518 if [ "$CHECK_PTLDEBUG" ] && { \
519 [ "$CHECK_PTLDEBUG" = "trace inode warning error emerg console" ] ||
520 [ "$CHECK_PTLDEBUG" = "trace inode" ]; }; then
521 echo "lnet.debug success"
523 echo "lnet.debug: want 'trace inode', have '$CHECK_PTLDEBUG'"
526 CHECK_SUBSYS="`do_facet ost1 lctl get_param -n subsystem_debug`"
527 if [ "$CHECK_SUBSYS" ] && [ "$CHECK_SUBSYS" = "mds ost" ]; then
528 echo "lnet.subsystem_debug success"
530 echo "lnet.subsystem_debug: want 'mds ost', have '$CHECK_SUBSYS'"
533 stop_ost || return $?
535 run_test 9 "test ptldebug and subsystem for mkfs"
543 do_facet $facet "test -b $dev" || rc=1
544 if [[ "$size" ]]; then
545 local in=$(do_facet $facet "dd if=$dev of=/dev/null bs=1k count=1 skip=$size 2>&1" |\
546 awk '($3 == "in") { print $1 }')
547 [[ $in = "1+0" ]] || rc=1
553 # Test 16 was to "verify that lustre will correct the mode of OBJECTS".
554 # But with new MDS stack we don't care about the mode of local objects
555 # anymore, so this test is removed. See bug 22944 for more details.
560 check_mount || return 41
563 echo "Remove mds config log"
564 if ! combined_mgs_mds ; then
568 do_facet mgs "$DEBUGFS -w -R 'unlink CONFIGS/$FSNAME-MDT0000' $MGSDEV || return \$?" || return $?
570 if ! combined_mgs_mds ; then
575 start_mds && return 42
578 run_test 17 "Verify failed mds_postsetup won't fail assertion (2936) (should return errs)"
581 [ "$FSTYPE" != "ldiskfs" ] && skip "not needed for FSTYPE=$FSTYPE" && return
583 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
588 # check if current MDSSIZE is large enough
589 [ $MDSSIZE -ge $MIN ] && OK=1 && myMDSSIZE=$MDSSIZE && \
590 log "use MDSSIZE=$MDSSIZE"
592 # check if the global config has a large enough MDSSIZE
593 [ -z "$OK" -a ! -z "$STORED_MDSSIZE" ] && [ $STORED_MDSSIZE -ge $MIN ] && \
594 OK=1 && myMDSSIZE=$STORED_MDSSIZE && \
595 log "use STORED_MDSSIZE=$STORED_MDSSIZE"
597 # check if the block device is large enough
598 [ -z "$OK" ] && $(is_blkdev $SINGLEMDS $MDSDEV $MIN) && OK=1 &&
599 myMDSSIZE=$MIN && log "use device $MDSDEV with MIN=$MIN"
601 # check if a loopback device has enough space for fs metadata (5%)
603 if [ -z "$OK" ]; then
604 local SPACE=$(do_facet $SINGLEMDS "[ -f $MDSDEV -o ! -e $MDSDEV ] && df -P \\\$(dirname $MDSDEV)" |
605 awk '($1 != "Filesystem") {print $4}')
606 ! [ -z "$SPACE" ] && [ $SPACE -gt $((MIN / 20)) ] && \
607 OK=1 && myMDSSIZE=$MIN && \
608 log "use file $MDSDEV with MIN=$MIN"
611 [ -z "$OK" ] && skip_env "$MDSDEV too small for ${MIN}kB MDS" && return
614 echo "mount mds with large journal..."
615 local OLD_MDS_MKFS_OPTS=$MDS_MKFS_OPTS
617 local opts="--mdt --fsname=$FSNAME --device-size=$myMDSSIZE --param sys.timeout=$TIMEOUT $MDSOPT"
619 if combined_mgs_mds ; then
620 MDS_MKFS_OPTS="--mgs $opts"
622 MDS_MKFS_OPTS="--mgsnode=$MGSNID $opts"
626 echo "mount lustre system..."
628 check_mount || return 41
630 echo "check journal size..."
631 local FOUNDSIZE=$(do_facet $SINGLEMDS "$DEBUGFS -c -R 'stat <8>' $MDSDEV" | awk '/Size: / { print $NF; exit;}')
632 if [ $FOUNDSIZE -gt $((32 * 1024 * 1024)) ]; then
633 log "Success: mkfs creates large journals. Size: $((FOUNDSIZE >> 20))M"
635 error "expected journal size > 32M, found $((FOUNDSIZE >> 20))M"
640 MDS_MKFS_OPTS=$OLD_MDS_MKFS_OPTS
643 run_test 18 "check mkfs creates large journals"
646 start_mds || return 1
647 stop_mds -f || return 2
649 run_test 19a "start/stop MDS without OSTs"
652 start_ost || return 1
653 stop_ost -f || return 2
655 run_test 19b "start/stop OSTs without MDS"
658 # first format the ost/mdt
662 check_mount || return 43
664 remount_client ro $MOUNT || return 44
665 touch $DIR/$tfile && echo "$DIR/$tfile created incorrectly" && return 45
666 [ -e $DIR/$tfile ] && echo "$DIR/$tfile exists incorrectly" && return 46
667 remount_client rw $MOUNT || return 47
669 [ ! -f $DIR/$tfile ] && echo "$DIR/$tfile missing" && return 48
670 MCNT=`grep -c $MOUNT /etc/mtab`
671 [ "$MCNT" -ne 1 ] && echo "$MOUNT in /etc/mtab $MCNT times" && return 49
676 run_test 20 "remount ro,rw mounts work and doesn't break /etc/mtab"
681 wait_osc_import_state mds ost FULL
685 run_test 21a "start mds before ost, stop ost first"
690 wait_osc_import_state mds ost FULL
694 run_test 21b "start ost before mds, stop mds first"
700 wait_osc_import_state mds ost2 FULL
704 #writeconf to remove all ost2 traces for subsequent tests
707 run_test 21c "start mds between two osts, stop mds last"
710 if combined_mgs_mds ; then
711 skip "need separate mgs device" && return 0
721 wait_osc_import_state mds ost2 FULL
728 run_test 21d "start mgs then ost and then mds"
733 echo Client mount with ost in logs, but none running
735 # wait until mds connected to ost and open client connection
736 wait_osc_import_state mds ost FULL
739 # check_mount will block trying to contact ost
740 mcreate $DIR/$tfile || return 40
741 rm -f $DIR/$tfile || return 42
745 echo Client mount with a running ost
748 # if gss enabled, wait full time to let connection from
749 # mds to ost be established, due to the mismatch between
750 # initial connect timeout and gss context negotiation timeout.
751 # This perhaps could be remove after AT landed.
752 echo "sleep $((TIMEOUT + TIMEOUT + TIMEOUT))s"
753 sleep $((TIMEOUT + TIMEOUT + TIMEOUT))
756 check_mount || return 41
761 run_test 22 "start a client before osts (should return errs)"
763 test_23a() { # was test_23
767 # force down client so that recovering mds waits for reconnect
768 local running=$(grep -c $MOUNT /proc/mounts) || true
769 if [ $running -ne 0 ]; then
770 echo "Stopping client $MOUNT (opts: -f)"
774 # enter recovery on mds
776 # try to start a new client
777 mount_client $MOUNT &
779 MOUNT_PID=$(ps -ef | grep "t lustre" | grep -v grep | awk '{print $2}')
780 MOUNT_LUSTRE_PID=`ps -ef | grep mount.lustre | grep -v grep | awk '{print $2}'`
781 echo mount pid is ${MOUNT_PID}, mount.lustre pid is ${MOUNT_LUSTRE_PID}
783 ps --ppid $MOUNT_LUSTRE_PID
784 # FIXME why o why can't I kill these? Manual "ctrl-c" works...
785 kill -TERM $MOUNT_LUSTRE_PID
786 echo "waiting for mount to finish"
788 # we can not wait $MOUNT_PID because it is not a child of this shell
794 while [ "$WAIT" -lt "$MAX_WAIT" ]; do
796 PID1=$(ps -ef | awk '{print $2}' | grep -w $MOUNT_PID)
797 PID2=$(ps -ef | awk '{print $2}' | grep -w $MOUNT_LUSTRE_PID)
800 [ -z "$PID1" -a -z "$PID2" ] && break
801 echo "waiting for mount to finish ... "
802 WAIT=$(( WAIT + sleep))
804 if [ "$WAIT" -eq "$MAX_WAIT" ]; then
805 error "MOUNT_PID $MOUNT_PID and "\
806 "MOUNT_LUSTRE_PID $MOUNT_LUSTRE_PID still not killed in $WAIT secs"
812 run_test 23a "interrupt client during recovery mount delay"
817 test_23b() { # was test_23
820 # Simulate -EINTR during mount OBD_FAIL_LDLM_CLOSE_THREAD
821 lctl set_param fail_loc=0x80000313
825 run_test 23b "Simulate -EINTR during mount"
827 fs2mds_HOST=$mds_HOST
828 fs2ost_HOST=$ost_HOST
832 echo "umount $MOUNT2 ..."
833 umount $MOUNT2 || true
834 echo "stopping fs2mds ..."
835 stop fs2mds -f || true
836 echo "stopping fs2ost ..."
837 stop fs2ost -f || true
841 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
843 if [ -z "$fs2ost_DEV" -o -z "$fs2mds_DEV" ]; then
844 is_blkdev $SINGLEMDS $MDSDEV && \
845 skip_env "mixed loopback and real device not working" && return
848 [ -n "$ost1_HOST" ] && fs2ost_HOST=$ost1_HOST
850 local fs2mdsdev=${fs2mds_DEV:-${MDSDEV}_2}
851 local fs2ostdev=${fs2ost_DEV:-$(ostdevname 1)_2}
853 # test 8-char fsname as well
854 local FSNAME2=test1234
855 add fs2mds $MDS_MKFS_OPTS --fsname=${FSNAME2} --nomgs --mgsnode=$MGSNID --reformat $fs2mdsdev || exit 10
857 add fs2ost $OST_MKFS_OPTS --fsname=${FSNAME2} --reformat $fs2ostdev || exit 10
860 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS && trap cleanup_24a EXIT INT
861 start fs2ost $fs2ostdev $OST_MOUNT_OPTS
863 mount -t lustre $MGSNID:/${FSNAME2} $MOUNT2 || return 1
865 check_mount || return 2
866 # files written on 1 should not show up on 2
867 cp /etc/passwd $DIR/$tfile
869 [ -e $MOUNT2/$tfile ] && error "File bleed" && return 7
872 cp /etc/passwd $MOUNT2/b || return 3
873 rm $MOUNT2/b || return 4
874 # 2 is actually mounted
875 grep $MOUNT2' ' /proc/mounts > /dev/null || return 5
877 facet_failover fs2mds
878 facet_failover fs2ost
881 # the MDS must remain up until last MDT
883 MDS=$(do_facet $SINGLEMDS "lctl get_param -n devices" | awk '($3 ~ "mdt" && $4 ~ "MDT") { print $4 }' | head -1)
884 [ -z "$MDS" ] && error "No MDT" && return 8
886 cleanup_nocli || return 6
888 run_test 24a "Multiple MDTs on a single node"
891 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
893 if [ -z "$fs2mds_DEV" ]; then
894 local dev=${SINGLEMDS}_dev
896 is_blkdev $SINGLEMDS $MDSDEV && \
897 skip_env "mixed loopback and real device not working" && return
900 local fs2mdsdev=${fs2mds_DEV:-${MDSDEV}_2}
902 add fs2mds $MDS_MKFS_OPTS --fsname=${FSNAME}2 --mgs --reformat $fs2mdsdev || exit 10
904 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS && return 2
907 run_test 24b "Multiple MGSs on a single node (should return err)"
911 check_mount || return 2
912 local MODULES=$($LCTL modules | awk '{ print $2 }')
913 rmmod $MODULES 2>/dev/null || true
916 run_test 25 "Verify modules are referenced"
920 # we need modules before mount for sysctl, so make sure...
921 do_facet $SINGLEMDS "lsmod | grep -q lustre || modprobe lustre"
922 #define OBD_FAIL_MDS_FS_SETUP 0x135
923 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000135"
924 start_mds && echo MDS started && return 1
925 lctl get_param -n devices
926 DEVS=$(lctl get_param -n devices | egrep -v MG | wc -l)
927 [ $DEVS -gt 0 ] && return 2
928 unload_modules_conf || return $?
930 run_test 26 "MDT startup failure cleans LOV (should return errs)"
936 local ORIG=$(do_facet $myfacet "$TEST")
937 if [ $# -gt 3 ]; then
943 echo "Setting $PARAM from $ORIG to $FINAL"
944 do_facet $SINGLEMDS "$LCTL conf_param $PARAM='$FINAL'" || error conf_param failed
946 wait_update $(facet_host $myfacet) "$TEST" "$FINAL" || error check failed!
950 start_ost || return 1
951 start_mds || return 2
952 echo "Requeue thread should have started: "
953 ps -e | grep ll_cfg_requeue
954 set_and_check ost1 "lctl get_param -n obdfilter.$FSNAME-OST0000.client_cache_seconds" "$FSNAME-OST0000.ost.client_cache_seconds" || return 3
957 run_test 27a "Reacquire MGS lock if OST started first"
962 local device=$(do_facet $SINGLEMDS "lctl get_param -n devices" | awk '($3 ~ "mdt" && $4 ~ "MDT") { print $4 }')
964 facet_failover $SINGLEMDS
965 set_and_check $SINGLEMDS "lctl get_param -n mdt.$device.identity_acquire_expire" "$device.mdt.identity_acquire_expire" || return 3
966 set_and_check client "lctl get_param -n mdc.$device-mdc-*.max_rpcs_in_flight" "$device.mdc.max_rpcs_in_flight" || return 4
970 run_test 27b "Reacquire MGS lock after failover"
974 TEST="lctl get_param -n llite.$FSNAME-*.max_read_ahead_whole_mb"
975 PARAM="$FSNAME.llite.max_read_ahead_whole_mb"
978 set_and_check client "$TEST" "$PARAM" $FINAL || return 3
979 FINAL=$(($FINAL + 1))
980 set_and_check client "$TEST" "$PARAM" $FINAL || return 4
981 umount_client $MOUNT || return 200
984 if [ $RESULT -ne $FINAL ]; then
985 echo "New config not seen: wanted $FINAL got $RESULT"
988 echo "New config success: got $RESULT"
990 set_and_check client "$TEST" "$PARAM" $ORIG || return 5
993 run_test 28 "permanent parameter setting"
996 [ "$OSTCOUNT" -lt "2" ] && skip_env "$OSTCOUNT < 2, skipping" && return
997 setup > /dev/null 2>&1
1001 local PARAM="$FSNAME-OST0001.osc.active"
1002 local PROC_ACT="osc.$FSNAME-OST0001-osc-[^M]*.active"
1003 local PROC_UUID="osc.$FSNAME-OST0001-osc-[^M]*.ost_server_uuid"
1005 ACTV=$(lctl get_param -n $PROC_ACT)
1007 set_and_check client "lctl get_param -n $PROC_ACT" "$PARAM" $DEAC || return 2
1008 # also check ost_server_uuid status
1009 RESULT=$(lctl get_param -n $PROC_UUID | grep DEACTIV)
1010 if [ -z "$RESULT" ]; then
1011 echo "Live client not deactivated: $(lctl get_param -n $PROC_UUID)"
1014 echo "Live client success: got $RESULT"
1018 local mdtosc=$(get_mdtosc_proc_path $SINGLEMDS $FSNAME-OST0001)
1019 mdtosc=${mdtosc/-MDT*/-MDT\*}
1020 local MPROC="osc.$mdtosc.active"
1025 RESULT=`do_facet $SINGLEMDS " lctl get_param -n $MPROC"`
1026 [ ${PIPESTATUS[0]} = 0 ] || error "Can't read $MPROC"
1027 if [ $RESULT -eq $DEAC ]; then
1028 echo "MDT deactivated also after $WAIT sec (got $RESULT)"
1032 if [ $WAIT -eq $MAX ]; then
1033 echo "MDT not deactivated: wanted $DEAC got $RESULT"
1036 echo "Waiting $(($MAX - $WAIT)) secs for MDT deactivated"
1039 # test new client starts deactivated
1040 umount_client $MOUNT || return 200
1042 RESULT=$(lctl get_param -n $PROC_UUID | grep DEACTIV | grep NEW)
1043 if [ -z "$RESULT" ]; then
1044 echo "New client not deactivated from start: $(lctl get_param -n $PROC_UUID)"
1047 echo "New client success: got $RESULT"
1050 # make sure it reactivates
1051 set_and_check client "lctl get_param -n $PROC_ACT" "$PARAM" $ACTV || return 6
1053 umount_client $MOUNT
1056 #writeconf to remove all ost2 traces for subsequent tests
1059 run_test 29 "permanently remove an OST"
1064 echo Big config llog
1065 TEST="lctl get_param -n llite.$FSNAME-*.max_read_ahead_whole_mb"
1067 LIST=(1 2 3 4 5 4 3 2 1 2 3 4 5 4 3 2 1 2 3 4 5)
1068 for i in ${LIST[@]}; do
1069 set_and_check client "$TEST" "$FSNAME.llite.max_read_ahead_whole_mb" $i || return 3
1071 # make sure client restart still works
1072 umount_client $MOUNT
1073 mount_client $MOUNT || return 4
1074 [ "$($TEST)" -ne "$i" ] && error "Param didn't stick across restart $($TEST) != $i"
1077 echo Erase parameter setting
1078 do_facet mgs "$LCTL conf_param -d $FSNAME.llite.max_read_ahead_whole_mb" || return 6
1079 umount_client $MOUNT
1080 mount_client $MOUNT || return 6
1082 echo "deleted (default) value=$FINAL, orig=$ORIG"
1083 # assumes this parameter started at the default value
1084 [ "$FINAL" -eq "$ORIG" ] || fail "Deleted value=$FINAL, orig=$ORIG"
1088 run_test 30a "Big config llog and conf_param deletion"
1093 # Make a fake nid. Use the OST nid, and add 20 to the least significant
1094 # numerical part of it. Hopefully that's not already a failover address for
1096 OSTNID=$(do_facet ost1 "$LCTL get_param nis" | tail -1 | awk '{print $1}')
1097 ORIGVAL=$(echo $OSTNID | egrep -oi "[0-9]*@")
1098 NEWVAL=$((($(echo $ORIGVAL | egrep -oi "[0-9]*") + 20) % 256))
1099 NEW=$(echo $OSTNID | sed "s/$ORIGVAL/$NEWVAL@/")
1100 echo "Using fake nid $NEW"
1102 TEST="$LCTL get_param -n osc.$FSNAME-OST0000-osc-[^M]*.import | grep failover_nids | sed -n 's/.*\($NEW\).*/\1/p'"
1103 set_and_check client "$TEST" "$FSNAME-OST0000.failover.node" $NEW || error "didn't add failover nid $NEW"
1104 NIDS=$($LCTL get_param -n osc.$FSNAME-OST0000-osc-[^M]*.import | grep failover_nids)
1106 NIDCOUNT=$(($(echo "$NIDS" | wc -w) - 1))
1107 echo "should have 2 failover nids: $NIDCOUNT"
1108 [ $NIDCOUNT -eq 2 ] || error "Failover nid not added"
1109 do_facet mgs "$LCTL conf_param -d $FSNAME-OST0000.failover.node" || error "conf_param delete failed"
1110 umount_client $MOUNT
1111 mount_client $MOUNT || return 3
1113 NIDS=$($LCTL get_param -n osc.$FSNAME-OST0000-osc-[^M]*.import | grep failover_nids)
1115 NIDCOUNT=$(($(echo "$NIDS" | wc -w) - 1))
1116 echo "only 1 final nid should remain: $NIDCOUNT"
1117 [ $NIDCOUNT -eq 1 ] || error "Failover nids not removed"
1121 run_test 30b "Remove failover nids"
1123 test_31() { # bug 10734
1124 # ipaddr must not exist
1125 mount -t lustre 4.3.2.1@tcp:/lustre $MOUNT || true
1128 run_test 31 "Connect to non-existent node (shouldn't crash)"
1130 # Use these start32/stop32 fn instead of t-f start/stop fn,
1131 # for local devices, to skip global facet vars init
1135 echo "Stopping local ${MOUNT%/*}/${facet} (opts:$@)"
1136 umount -d $@ ${MOUNT%/*}/${facet}
1145 mkdir -p ${MOUNT%/*}/${facet}
1147 echo "Starting local ${facet}: $@ $device ${MOUNT%/*}/${facet}"
1148 mount -t lustre $@ ${device} ${MOUNT%/*}/${facet}
1150 if [ $RC -ne 0 ]; then
1151 echo "mount -t lustre $@ ${device} ${MOUNT%/*}/${facet}"
1152 echo "Start of ${device} of local ${facet} failed ${RC}"
1158 cleanup_nocli32 () {
1166 echo "Cleanup test_32 umount $MOUNT ..."
1167 umount -f $MOUNT || true
1168 echo "Cleanup local mds ost1 ..."
1170 combined_mgs_mds || start_mgs
1175 client_only && skip "client only testing" && return 0
1176 [ "$NETTYPE" = "tcp" ] || { skip "NETTYPE != tcp" && return 0; }
1177 [ -z "$TUNEFS" ] && skip_env "No tunefs" && return 0
1179 local DISK1_8=$LUSTRE/tests/disk1_8.tar.bz2
1180 [ ! -r $DISK1_8 ] && skip_env "Cannot find $DISK1_8" && return 0
1181 local tmpdir=$TMP/conf32a
1183 tar xjvf $DISK1_8 -C $tmpdir || \
1184 { skip_env "Cannot untar $DISK1_8" && return 0; }
1187 $LCTL set_param debug=$PTLDEBUG
1189 $TUNEFS $tmpdir/mds || error "tunefs failed"
1191 combined_mgs_mds || stop mgs
1193 # nids are wrong, so client wont work, but server should start
1194 start32 mds1 $tmpdir/mds "-o loop,exclude=lustre-OST0000" && \
1195 trap cleanup_32 EXIT INT || return 3
1197 local UUID=$($LCTL get_param -n mdt.lustre-MDT0000.uuid)
1199 [ "$UUID" == "lustre-MDT0000_UUID" ] || error "UUID is wrong: $UUID"
1201 $TUNEFS --mgsnode=$HOSTNAME $tmpdir/ost1 || error "tunefs failed"
1202 start32 ost1 $tmpdir/ost1 "-o loop" || return 5
1203 UUID=$($LCTL get_param -n obdfilter.lustre-OST0000.uuid)
1205 [ "$UUID" == "lustre-OST0000_UUID" ] || error "UUID is wrong: $UUID"
1207 local NID=$($LCTL list_nids | head -1)
1209 echo "OSC changes should succeed:"
1210 $LCTL conf_param lustre-OST0000.osc.max_dirty_mb=15 || return 7
1211 $LCTL conf_param lustre-OST0000.failover.node=$NID || return 8
1214 echo "MDC changes should succeed:"
1215 $LCTL conf_param lustre-MDT0000.mdc.max_rpcs_in_flight=9 || return 9
1216 $LCTL conf_param lustre-MDT0000.failover.node=$NID || return 10
1219 echo "LOV changes should succeed:"
1220 $LCTL pool_new lustre.interop || return 11
1221 $LCTL conf_param lustre-MDT0000.lov.stripesize=4M || return 12
1226 # mount a second time to make sure we didnt leave upgrade flag on
1228 $TUNEFS --dryrun $tmpdir/mds || error "tunefs failed"
1230 combined_mgs_mds || stop mgs
1232 start32 mds1 $tmpdir/mds "-o loop,exclude=lustre-OST0000" && \
1233 trap cleanup_32 EXIT INT || return 12
1237 rm -rf $tmpdir || true # true is only for TMP on NFS
1239 run_test 32a "Upgrade from 1.8 (not live)"
1242 client_only && skip "client only testing" && return 0
1243 [ "$NETTYPE" = "tcp" ] || { skip "NETTYPE != tcp" && return 0; }
1244 [ -z "$TUNEFS" ] && skip_env "No tunefs" && return
1246 local DISK1_8=$LUSTRE/tests/disk1_8.tar.bz2
1247 [ ! -r $DISK1_8 ] && skip_env "Cannot find $DISK1_8" && return 0
1248 local tmpdir=$TMP/conf32b
1250 tar xjvf $DISK1_8 -C $tmpdir || \
1251 { skip_env "Cannot untar $DISK1_8" && return ; }
1254 $LCTL set_param debug="config"
1255 local NEWNAME=lustre
1257 # writeconf will cause servers to register with their current nids
1258 $TUNEFS --writeconf --fsname=$NEWNAME $tmpdir/mds || error "tunefs failed"
1259 combined_mgs_mds || stop mgs
1261 start32 mds1 $tmpdir/mds "-o loop" && \
1262 trap cleanup_32 EXIT INT || return 3
1264 local UUID=$($LCTL get_param -n mdt.${NEWNAME}-MDT0000.uuid)
1266 [ "$UUID" == "${NEWNAME}-MDT0000_UUID" ] || error "UUID is wrong: $UUID"
1268 $TUNEFS --mgsnode=$HOSTNAME --writeconf --fsname=$NEWNAME $tmpdir/ost1 ||\
1269 error "tunefs failed"
1270 start32 ost1 $tmpdir/ost1 "-o loop" || return 5
1271 UUID=$($LCTL get_param -n obdfilter.${NEWNAME}-OST0000.uuid)
1273 [ "$UUID" == "${NEWNAME}-OST0000_UUID" ] || error "UUID is wrong: $UUID"
1275 local NID=$($LCTL list_nids | head -1)
1277 echo "OSC changes should succeed:"
1278 $LCTL conf_param ${NEWNAME}-OST0000.osc.max_dirty_mb=15 || return 7
1279 $LCTL conf_param ${NEWNAME}-OST0000.failover.node=$NID || return 8
1282 echo "MDC changes should succeed:"
1283 $LCTL conf_param ${NEWNAME}-MDT0000.mdc.max_rpcs_in_flight=9 || return 9
1284 $LCTL conf_param ${NEWNAME}-MDT0000.failover.node=$NID || return 10
1287 echo "LOV changes should succeed:"
1288 $LCTL pool_new ${NEWNAME}.interop || return 11
1289 $LCTL conf_param ${NEWNAME}-MDT0000.lov.stripesize=4M || return 12
1292 # MDT and OST should have registered with new nids, so we should have
1293 # a fully-functioning client
1294 echo "Check client and old fs contents"
1296 local device=`h2$NETTYPE $HOSTNAME`:/$NEWNAME
1297 echo "Starting local client: $HOSTNAME: $device $MOUNT"
1298 mount -t lustre $device $MOUNT || return 1
1300 local old=$($LCTL get_param -n mdc.*.max_rpcs_in_flight)
1301 local new=$((old + 5))
1302 $LCTL conf_param ${NEWNAME}-MDT0000.mdc.max_rpcs_in_flight=$new
1303 wait_update $HOSTNAME "$LCTL get_param -n mdc.*.max_rpcs_in_flight" $new || return 11
1305 [ "$(cksum $MOUNT/passwd | cut -d' ' -f 1,2)" == "94306271 1478" ] || return 12
1310 rm -rf $tmpdir || true # true is only for TMP on NFS
1312 run_test 32b "Upgrade from 1.8 with writeconf"
1314 test_33a() { # bug 12333, was test_33
1316 local FSNAME2=test-123
1317 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
1319 [ -n "$ost1_HOST" ] && fs2ost_HOST=$ost1_HOST
1321 if [ -z "$fs2ost_DEV" -o -z "$fs2mds_DEV" ]; then
1322 local dev=${SINGLEMDS}_dev
1323 local MDSDEV=${!dev}
1324 is_blkdev $SINGLEMDS $MDSDEV && \
1325 skip_env "mixed loopback and real device not working" && return
1328 combined_mgs_mds || mkfs_opts="$mkfs_opts --nomgs"
1330 local fs2mdsdev=${fs2mds_DEV:-${MDSDEV}_2}
1331 local fs2ostdev=${fs2ost_DEV:-$(ostdevname 1)_2}
1332 add fs2mds $MDS_MKFS_OPTS --mkfsoptions='\"-J size=8\"' --fsname=${FSNAME2} --reformat $fs2mdsdev || exit 10
1333 add fs2ost $OST_MKFS_OPTS --fsname=${FSNAME2} --index=8191 --mgsnode=$MGSNID --reformat $fs2ostdev || exit 10
1335 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS && trap cleanup_24a EXIT INT
1336 start fs2ost $fs2ostdev $OST_MOUNT_OPTS
1337 do_facet $SINGLEMDS "$LCTL conf_param $FSNAME2.sys.timeout=200" || rc=1
1339 mount -t lustre $MGSNID:/${FSNAME2} $MOUNT2 || rc=2
1342 cp /etc/hosts $MOUNT2/ || rc=3
1343 $LFS getstripe $MOUNT2/hosts
1348 rm -rf $MOUNT2 $fs2mdsdev $fs2ostdev
1349 cleanup_nocli || rc=6
1352 run_test 33a "Mount ost with a large index number"
1354 test_33b() { # was test_34
1357 do_facet client dd if=/dev/zero of=$MOUNT/24 bs=1024k count=1
1358 # Drop lock cancelation reply during umount
1359 #define OBD_FAIL_LDLM_CANCEL 0x304
1360 do_facet client lctl set_param fail_loc=0x80000304
1361 #lctl set_param debug=-1
1362 umount_client $MOUNT
1365 run_test 33b "Drop cancel during umount"
1369 do_facet client "sh runmultiop_bg_pause $DIR/file O_c"
1370 manual_umount_client
1372 do_facet client killall -USR1 multiop
1373 if [ $rc -eq 0 ]; then
1374 error "umount not fail!"
1379 run_test 34a "umount with opened file should be fail"
1384 touch $DIR/$tfile || return 1
1385 stop_mds --force || return 2
1387 manual_umount_client --force
1389 if [ $rc -ne 0 ]; then
1390 error "mtab after failed umount - rc $rc"
1396 run_test 34b "force umount with failed mds should be normal"
1400 touch $DIR/$tfile || return 1
1401 stop_ost --force || return 2
1403 manual_umount_client --force
1405 if [ $rc -ne 0 ]; then
1406 error "mtab after failed umount - rc $rc"
1412 run_test 34c "force umount with failed ost should be normal"
1414 test_35a() { # bug 12459
1417 DBG_SAVE="`lctl get_param -n debug`"
1418 lctl set_param debug="ha"
1420 log "Set up a fake failnode for the MDS"
1422 local device=$(do_facet $SINGLEMDS "lctl get_param -n devices" | awk '($3 ~ "mdt" && $4 ~ "MDT") { print $4 }' | head -1)
1423 do_facet $SINGLEMDS $LCTL conf_param ${device}.failover.node=$FAKENID || return 4
1425 log "Wait for RECONNECT_INTERVAL seconds (10s)"
1428 MSG="conf-sanity.sh test_35a `date +%F%kh%Mm%Ss`"
1431 log "Stopping the MDT:"
1432 stop_mds || return 5
1434 df $MOUNT > /dev/null 2>&1 &
1436 log "Restarting the MDT:"
1437 start_mds || return 6
1438 log "Wait for df ($DFPID) ... "
1441 lctl set_param debug="$DBG_SAVE"
1443 # retrieve from the log the first server that the client tried to
1444 # contact after the connection loss
1445 $LCTL dk $TMP/lustre-log-$TESTNAME.log
1446 NEXTCONN=`awk "/${MSG}/ {start = 1;}
1447 /import_select_connection.*$device-mdc.* using connection/ {
1449 if (\\\$NF ~ /$FAKENID/)
1455 }" $TMP/lustre-log-$TESTNAME.log`
1456 [ "$NEXTCONN" != "0" ] && log "The client didn't try to reconnect to the last active server (tried ${NEXTCONN} instead)" && return 7
1458 # remove nid settings
1461 run_test 35a "Reconnect to the last active server first"
1463 test_35b() { # bug 18674
1464 remote_mds || { skip "local MDS" && return 0; }
1468 $LCTL set_param debug="ha"
1470 MSG="conf-sanity.sh test_35b `date +%F%kh%Mm%Ss`"
1473 log "Set up a fake failnode for the MDS"
1475 local device=$(do_facet $SINGLEMDS "$LCTL get_param -n devices" | \
1476 awk '($3 ~ "mdt" && $4 ~ "MDT") { print $4 }' | head -1)
1477 do_facet $SINGLEMDS "$LCTL conf_param ${device}.failover.node=$FAKENID" || \
1480 local at_max_saved=0
1481 # adaptive timeouts may prevent seeing the issue
1482 if at_is_enabled; then
1483 at_max_saved=$(at_max_get mds)
1484 at_max_set 0 mds client
1487 mkdir -p $MOUNT/testdir
1488 touch $MOUNT/testdir/test
1490 log "Injecting EBUSY on MDS"
1491 # Setting OBD_FAIL_MDS_RESEND=0x136
1492 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0x80000136" || return 2
1494 log "Stat on a test file"
1495 stat $MOUNT/testdir/test
1497 log "Stop injecting EBUSY on MDS"
1498 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0" || return 3
1499 rm -f $MOUNT/testdir/test
1502 # restore adaptive timeout
1503 [ $at_max_saved -ne 0 ] && at_max_set $at_max_saved mds client
1505 $LCTL dk $TMP/lustre-log-$TESTNAME.log
1507 # retrieve from the log if the client has ever tried to
1508 # contact the fake server after the loss of connection
1509 FAILCONN=`awk "BEGIN {ret = 0;}
1510 /import_select_connection.*${FSNAME}-MDT0000-mdc.* using connection/ {
1512 if (\\\$NF ~ /$FAKENID/) {
1517 END {print ret}" $TMP/lustre-log-$TESTNAME.log`
1519 [ "$FAILCONN" == "0" ] && \
1520 log "ERROR: The client reconnection has not been triggered" && \
1522 [ "$FAILCONN" == "2" ] && \
1523 log "ERROR: The client tried to reconnect to the failover server while the primary was busy" && \
1527 # remove nid settings
1530 run_test 35b "Continue reconnection retries, if the active server is busy"
1533 [ $OSTCOUNT -lt 2 ] && skip_env "skipping test for single OST" && return
1535 [ "$ost_HOST" = "`hostname`" -o "$ost1_HOST" = "`hostname`" ] || \
1536 { skip "remote OST" && return 0; }
1539 local FSNAME2=test1234
1540 local fs3ost_HOST=$ost_HOST
1541 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
1543 [ -n "$ost1_HOST" ] && fs2ost_HOST=$ost1_HOST && fs3ost_HOST=$ost1_HOST
1545 if [ -z "$fs2ost_DEV" -o -z "$fs2mds_DEV" -o -z "$fs3ost_DEV" ]; then
1546 is_blkdev $SINGLEMDS $MDSDEV && \
1547 skip_env "mixed loopback and real device not working" && return
1550 local fs2mdsdev=${fs2mds_DEV:-${MDSDEV}_2}
1551 local fs2ostdev=${fs2ost_DEV:-$(ostdevname 1)_2}
1552 local fs3ostdev=${fs3ost_DEV:-$(ostdevname 2)_2}
1553 add fs2mds $MDS_MKFS_OPTS --fsname=${FSNAME2} --reformat $fs2mdsdev || exit 10
1554 # XXX after we support non 4K disk blocksize, change following --mkfsoptions with
1556 add fs2ost $OST_MKFS_OPTS --mkfsoptions='-b4096' --fsname=${FSNAME2} --mgsnode=$MGSNID --reformat $fs2ostdev || exit 10
1557 add fs3ost $OST_MKFS_OPTS --mkfsoptions='-b4096' --fsname=${FSNAME2} --mgsnode=$MGSNID --reformat $fs3ostdev || exit 10
1559 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS
1560 start fs2ost $fs2ostdev $OST_MOUNT_OPTS
1561 start fs3ost $fs3ostdev $OST_MOUNT_OPTS
1563 mount -t lustre $MGSNID:/${FSNAME2} $MOUNT2 || return 1
1565 sleep 5 # until 11778 fixed
1567 dd if=/dev/zero of=$MOUNT2/$tfile bs=1M count=7 || return 2
1569 BKTOTAL=`lctl get_param -n obdfilter.*.kbytestotal | awk 'BEGIN{total=0}; {total+=$1}; END{print total}'`
1570 BKFREE=`lctl get_param -n obdfilter.*.kbytesfree | awk 'BEGIN{free=0}; {free+=$1}; END{print free}'`
1571 BKAVAIL=`lctl get_param -n obdfilter.*.kbytesavail | awk 'BEGIN{avail=0}; {avail+=$1}; END{print avail}'`
1572 STRING=`df -P $MOUNT2 | tail -n 1 | awk '{print $2","$3","$4}'`
1573 DFTOTAL=`echo $STRING | cut -d, -f1`
1574 DFUSED=`echo $STRING | cut -d, -f2`
1575 DFAVAIL=`echo $STRING | cut -d, -f3`
1576 DFFREE=$(($DFTOTAL - $DFUSED))
1578 ALLOWANCE=$((64 * $OSTCOUNT))
1580 if [ $DFTOTAL -lt $(($BKTOTAL - $ALLOWANCE)) ] ||
1581 [ $DFTOTAL -gt $(($BKTOTAL + $ALLOWANCE)) ] ; then
1582 echo "**** FAIL: df total($DFTOTAL) mismatch OST total($BKTOTAL)"
1585 if [ $DFFREE -lt $(($BKFREE - $ALLOWANCE)) ] ||
1586 [ $DFFREE -gt $(($BKFREE + $ALLOWANCE)) ] ; then
1587 echo "**** FAIL: df free($DFFREE) mismatch OST free($BKFREE)"
1590 if [ $DFAVAIL -lt $(($BKAVAIL - $ALLOWANCE)) ] ||
1591 [ $DFAVAIL -gt $(($BKAVAIL + $ALLOWANCE)) ] ; then
1592 echo "**** FAIL: df avail($DFAVAIL) mismatch OST avail($BKAVAIL)"
1597 stop fs3ost -f || return 200
1598 stop fs2ost -f || return 201
1599 stop fs2mds -f || return 202
1600 rm -rf $MOUNT2 $fs2mdsdev $fs2ostdev $fs3ostdev
1601 unload_modules_conf || return 203
1604 run_test 36 "df report consistency on OSTs with different block size"
1607 local mntpt=$(facet_mntpt $SINGLEMDS)
1608 local mdsdev=$(mdsdevname ${SINGLEMDS//mds/})
1609 local mdsdev_sym="$TMP/sym_mdt.img"
1611 echo "MDS : $mdsdev"
1612 echo "SYMLINK : $mdsdev_sym"
1613 do_facet $SINGLEMDS rm -f $mdsdev_sym
1615 do_facet $SINGLEMDS ln -s $mdsdev $mdsdev_sym
1617 echo "mount symlink device - $mdsdev_sym"
1620 mount_op=$(do_facet $SINGLEMDS mount -v -t lustre $MDS_MOUNT_OPTS $mdsdev_sym $mntpt 2>&1 )
1623 echo mount_op=$mount_op
1625 do_facet $SINGLEMDS "umount -d $mntpt && rm -f $mdsdev_sym"
1627 if $(echo $mount_op | grep -q "unable to set tunable"); then
1628 error "set tunables failed for symlink device"
1631 [ $rc -eq 0 ] || error "mount symlink $mdsdev_sym failed! rc=$rc"
1635 run_test 37 "verify set tunables works for symlink device"
1637 test_38() { # bug 14222
1642 FILES=`find $SRC -type f -mtime +1 | head -n $COUNT`
1643 log "copying $(echo $FILES | wc -w) files to $DIR/$tdir"
1645 tar cf - $FILES | tar xf - -C $DIR/$tdir || \
1646 error "copying $SRC to $DIR/$tdir"
1648 umount_client $MOUNT
1650 log "rename lov_objid file on MDS"
1651 rm -f $TMP/lov_objid.orig
1653 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
1654 do_facet $SINGLEMDS "$DEBUGFS -c -R \\\"dump lov_objid $TMP/lov_objid.orig\\\" $MDSDEV"
1655 do_facet $SINGLEMDS "$DEBUGFS -w -R \\\"rm lov_objid\\\" $MDSDEV"
1657 do_facet $SINGLEMDS "od -Ax -td8 $TMP/lov_objid.orig"
1658 # check create in mds_lov_connect
1662 [ $V ] && log "verifying $DIR/$tdir/$f"
1663 diff -q $f $DIR/$tdir/$f || ERROR=y
1665 do_facet $SINGLEMDS "$DEBUGFS -c -R \\\"dump lov_objid $TMP/lov_objid.new\\\" $MDSDEV"
1666 do_facet $SINGLEMDS "od -Ax -td8 $TMP/lov_objid.new"
1667 [ "$ERROR" = "y" ] && error "old and new files are different after connect" || true
1669 # check it's updates in sync
1670 umount_client $MOUNT
1673 do_facet $SINGLEMDS dd if=/dev/zero of=$TMP/lov_objid.clear bs=4096 count=1
1674 do_facet $SINGLEMDS "$DEBUGFS -w -R \\\"rm lov_objid\\\" $MDSDEV"
1675 do_facet $SINGLEMDS "$DEBUGFS -w -R \\\"write $TMP/lov_objid.clear lov_objid\\\" $MDSDEV "
1680 [ $V ] && log "verifying $DIR/$tdir/$f"
1681 diff -q $f $DIR/$tdir/$f || ERROR=y
1683 do_facet $SINGLEMDS "$DEBUGFS -c -R \\\"dump lov_objid $TMP/lov_objid.new1\\\" $MDSDEV"
1684 do_facet $SINGLEMDS "od -Ax -td8 $TMP/lov_objid.new1"
1685 umount_client $MOUNT
1687 [ "$ERROR" = "y" ] && error "old and new files are different after sync" || true
1689 log "files compared the same"
1692 run_test 38 "MDS recreates missing lov_objid file from OST data"
1698 perl $SRCDIR/leak_finder.pl $TMP/debug 2>&1 | egrep '*** Leak:' &&
1699 error "memory leak detected" || true
1701 run_test 39 "leak_finder recognizes both LUSTRE and LNET malloc messages"
1703 test_40() { # bug 15759
1705 #define OBD_FAIL_TGT_TOOMANY_THREADS 0x706
1706 do_facet $SINGLEMDS "sysctl -w lustre.fail_loc=0x80000706"
1710 run_test 40 "race during service thread startup"
1712 test_41a() { #bug 14134
1713 echo $MDS_MOUNT_OPTS | grep "loop" && skip " loop devices does not work with nosvc option" && return
1716 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
1718 start $SINGLEMDS $MDSDEV $MDS_MOUNT_OPTS -o nosvc -n
1719 start ost1 `ostdevname 1` $OST_MOUNT_OPTS
1720 start $SINGLEMDS $MDSDEV $MDS_MOUNT_OPTS -o nomgs,force
1722 mount_client $MOUNT || return 1
1725 echo "blah blah" > $MOUNT/$tfile
1728 umount_client $MOUNT
1729 stop ost1 -f || return 201
1730 stop_mds -f || return 202
1731 stop_mds -f || return 203
1732 unload_modules_conf || return 204
1735 run_test 41a "mount mds with --nosvc and --nomgs"
1738 echo $MDS_MOUNT_OPTS | grep "loop" && skip " loop devices does not work with nosvc option" && return
1742 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
1744 start $SINGLEMDS $MDSDEV $MDS_MOUNT_OPTS -o nosvc -n
1746 start $SINGLEMDS $MDSDEV $MDS_MOUNT_OPTS -o nomgs,force
1748 mount_client $MOUNT || return 1
1751 echo "blah blah" > $MOUNT/$tfile
1752 cat $MOUNT/$tfile || return 200
1754 umount_client $MOUNT
1755 stop_ost || return 201
1756 stop_mds -f || return 202
1757 stop_mds -f || return 203
1761 run_test 41b "mount mds with --nosvc and --nomgs on first mount"
1762 test_42() { #bug 14693
1764 check_mount || return 2
1765 do_facet mgs $LCTL conf_param lustre.llite.some_wrong_param=10
1766 umount_client $MOUNT
1767 mount_client $MOUNT || return 1
1771 run_test 42 "invalid config param should not prevent client from mounting"
1774 [ $UID -ne 0 -o $RUNAS_ID -eq 0 ] && skip_env "run as root"
1776 chmod ugo+x $DIR || error "chmod 0 failed"
1778 "lctl get_param -n mdt.$FSNAME-MDT0000.root_squash" \
1779 "$FSNAME.mdt.root_squash" \
1782 "lctl get_param -n mdt.$FSNAME-MDT0000.nosquash_nids" \
1783 "$FSNAME.mdt.nosquash_nids" \
1787 # create set of test files
1789 echo "111" > $DIR/$tfile-userfile || error "write 1 failed"
1790 chmod go-rw $DIR/$tfile-userfile || error "chmod 1 failed"
1791 chown $RUNAS_ID.$RUNAS_ID $DIR/$tfile-userfile || error "chown failed"
1793 echo "222" > $DIR/$tfile-rootfile || error "write 2 failed"
1794 chmod go-rw $DIR/$tfile-rootfile || error "chmod 2 faield"
1796 mkdir $DIR/$tdir-rootdir -p || error "mkdir failed"
1797 chmod go-rwx $DIR/$tdir-rootdir || error "chmod 3 failed"
1798 touch $DIR/$tdir-rootdir/tfile-1 || error "touch failed"
1801 # check root_squash:
1802 # set root squash UID:GID to RUNAS_ID
1803 # root should be able to access only files owned by RUNAS_ID
1806 "lctl get_param -n mdt.$FSNAME-MDT0000.root_squash" \
1807 "$FSNAME.mdt.root_squash" \
1808 "$RUNAS_ID:$RUNAS_ID"
1810 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-userfile)
1811 dd if=$DIR/$tfile-userfile 1>/dev/null 2>/dev/null || \
1812 error "$ST: root read permission is denied"
1813 echo "$ST: root read permission is granted - ok"
1816 dd conv=notrunc if=$DIR/$tfile-userfile 1>/dev/null 2>/dev/null || \
1817 error "$ST: root write permission is denied"
1818 echo "$ST: root write permission is granted - ok"
1820 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-rootfile)
1821 dd if=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null && \
1822 error "$ST: root read permission is granted"
1823 echo "$ST: root read permission is denied - ok"
1826 dd conv=notrunc of=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null && \
1827 error "$ST: root write permission is granted"
1828 echo "$ST: root write permission is denied - ok"
1830 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tdir-rootdir)
1831 rm $DIR/$tdir-rootdir/tfile-1 1>/dev/null 2>/dev/null && \
1832 error "$ST: root unlink permission is granted"
1833 echo "$ST: root unlink permission is denied - ok"
1835 touch $DIR/tdir-rootdir/tfile-2 1>/dev/null 2>/dev/null && \
1836 error "$ST: root create permission is granted"
1837 echo "$ST: root create permission is denied - ok"
1840 # check nosquash_nids:
1841 # put client's NID into nosquash_nids list,
1842 # root should be able to access root file after that
1844 local NIDLIST=$(lctl list_nids all | tr '\n' ' ')
1845 NIDLIST="2@elan $NIDLIST 192.168.0.[2,10]@tcp"
1846 NIDLIST=$(echo $NIDLIST | tr -s ' ' ' ')
1848 "lctl get_param -n mdt.$FSNAME-MDT0000.nosquash_nids" \
1849 "$FSNAME-MDTall.mdt.nosquash_nids" \
1852 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-rootfile)
1853 dd if=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null || \
1854 error "$ST: root read permission is denied"
1855 echo "$ST: root read permission is granted - ok"
1858 dd conv=notrunc of=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null || \
1859 error "$ST: root write permission is denied"
1860 echo "$ST: root write permission is granted - ok"
1862 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tdir-rootdir)
1863 rm $DIR/$tdir-rootdir/tfile-1 || \
1864 error "$ST: root unlink permission is denied"
1865 echo "$ST: root unlink permission is granted - ok"
1866 touch $DIR/$tdir-rootdir/tfile-2 || \
1867 error "$ST: root create permission is denied"
1868 echo "$ST: root create permission is granted - ok"
1872 run_test 43 "check root_squash and nosquash_nids"
1874 umount_client $MOUNT
1879 check_mount || return 2
1880 UUID=$($LCTL get_param llite.${FSNAME}*.uuid | cut -d= -f2)
1882 UUIDS=$(do_facet $SINGLEMDS "$LCTL get_param mdt.${FSNAME}*.exports.*.uuid")
1883 for VAL in $UUIDS; do
1884 NID=$(echo $VAL | cut -d= -f1)
1885 CLUUID=$(echo $VAL | cut -d= -f2)
1886 [ "$UUID" = "$CLUUID" ] && STATS_FOUND=yes && break
1888 [ "$STATS_FOUND" = "no" ] && error "stats not found for client"
1892 run_test 44 "mounted client proc entry exists"
1896 check_mount || return 2
1901 #define OBD_FAIL_PTLRPC_LONG_UNLINK 0x50f
1902 do_facet client "lctl set_param fail_loc=0x50f"
1905 manual_umount_client --force || return 3
1906 do_facet client "lctl set_param fail_loc=0x0"
1908 mount_client $MOUNT || return 4
1912 run_test 45 "long unlink handling in ptlrpcd"
1919 umount_client $MOUNT2 || rc=$?
1920 umount_client $MOUNT || rc=$?
1921 while [ $count -gt 0 ]; do
1922 stop ost${count} -f || rc=$?
1926 cleanup_nocli || rc=$?
1927 #writeconf to remove all ost2 traces for subsequent tests
1933 echo "Testing with $OSTCOUNT OSTs"
1935 start_mds || return 1
1936 #first client should see only one ost
1937 start_ost || return 2
1938 wait_osc_import_state mds ost FULL
1940 mount_client $MOUNT || return 3
1941 trap "cleanup_46a $OSTCOUNT" EXIT ERR
1944 for (( i=2; i<=$OSTCOUNT; i++ )); do
1945 start ost$i `ostdevname $i` $OST_MOUNT_OPTS || return $((i+2))
1948 # wait until osts in sync
1949 for (( i=2; i<=$OSTCOUNT; i++ )); do
1950 wait_osc_import_state mds ost$i FULL
1954 #second client see all ost's
1956 mount_client $MOUNT2 || return 8
1957 $LFS setstripe $MOUNT2 -c -1 || return 9
1958 $LFS getstripe $MOUNT2 || return 10
1960 echo "ok" > $MOUNT2/widestripe
1961 $LFS getstripe $MOUNT2/widestripe || return 11
1962 # fill acl buffer for avoid expand lsm to them
1963 awk -F : '{if (FNR < 25) { print "u:"$1":rwx" }}' /etc/passwd | while read acl; do
1964 setfacl -m $acl $MOUNT2/widestripe
1968 stat $MOUNT/widestripe || return 12
1970 cleanup_46a $OSTCOUNT || { echo "cleanup_46a failed!" && return 13; }
1973 run_test 46a "handle ost additional - wide striped file"
1978 check_mount || return 2
1979 $LCTL set_param ldlm.namespaces.$FSNAME-*-*-*.lru_size=100
1983 for ns in $($LCTL get_param ldlm.namespaces.$FSNAME-*-*-*.lru_size); do
1984 if echo $ns | grep "MDT[[:digit:]]*"; then
1987 lrs=$(echo $ns | sed 's/.*lru_size=//')
1988 lru_size[count]=$lrs
1993 facet_failover $SINGLEMDS
1994 client_up || return 3
1997 for ns in $($LCTL get_param ldlm.namespaces.$FSNAME-*-*-*.lru_size); do
1998 if echo $ns | grep "MDT[[:digit:]]*"; then
2001 lrs=$(echo $ns | sed 's/.*lru_size=//')
2002 if ! test "$lrs" -eq "${lru_size[count]}"; then
2003 n=$(echo $ns | sed -e 's/ldlm.namespaces.//' -e 's/.lru_size=.*//')
2004 error "$n has lost lru_size: $lrs vs. ${lru_size[count]}"
2012 run_test 47 "server restart does not make client loss lru_resize settings"
2017 # reformat after this test is needed - if test will failed
2018 # we will have unkillable file at FS
2022 test_48() { # bug 17636
2025 check_mount || return 2
2027 $LFS setstripe $MOUNT -c -1 || return 9
2028 $LFS getstripe $MOUNT || return 10
2030 echo "ok" > $MOUNT/widestripe
2031 $LFS getstripe $MOUNT/widestripe || return 11
2033 trap cleanup_48 EXIT ERR
2035 # fill acl buffer for avoid expand lsm to them
2036 getent passwd | awk -F : '{ print "u:"$1":rwx" }' | while read acl; do
2037 setfacl -m $acl $MOUNT/widestripe
2040 stat $MOUNT/widestripe || return 12
2045 run_test 48 "too many acls on file"
2047 # check PARAM_SYS_LDLM_TIMEOUT option of MKFS.LUSTRE
2048 test_49() { # bug 17710
2049 local OLD_MDS_MKFS_OPTS=$MDS_MKFS_OPTS
2050 local OLD_OST_MKFS_OPTS=$OST_MKFS_OPTS
2051 local LOCAL_TIMEOUT=20
2054 OST_MKFS_OPTS="--ost --fsname=$FSNAME --device-size=$OSTSIZE --mgsnode=$MGSNID --param sys.timeout=$LOCAL_TIMEOUT --param sys.ldlm_timeout=$LOCAL_TIMEOUT $MKFSOPT $OSTOPT"
2058 check_mount || return 1
2060 echo "check ldlm_timout..."
2061 LDLM_MDS="`do_facet $SINGLEMDS lctl get_param -n ldlm_timeout`"
2062 LDLM_OST1="`do_facet ost1 lctl get_param -n ldlm_timeout`"
2063 LDLM_CLIENT="`do_facet client lctl get_param -n ldlm_timeout`"
2065 if [ $LDLM_MDS -ne $LDLM_OST1 ] || [ $LDLM_MDS -ne $LDLM_CLIENT ]; then
2066 error "Different LDLM_TIMEOUT:$LDLM_MDS $LDLM_OST1 $LDLM_CLIENT"
2069 if [ $LDLM_MDS -ne $((LOCAL_TIMEOUT / 3)) ]; then
2070 error "LDLM_TIMEOUT($LDLM_MDS) is not correct"
2073 umount_client $MOUNT
2074 stop_ost || return 2
2075 stop_mds || return 3
2077 OST_MKFS_OPTS="--ost --fsname=$FSNAME --device-size=$OSTSIZE --mgsnode=$MGSNID --param sys.timeout=$LOCAL_TIMEOUT --param sys.ldlm_timeout=$((LOCAL_TIMEOUT - 1)) $MKFSOPT $OSTOPT"
2081 check_mount || return 7
2083 LDLM_MDS="`do_facet $SINGLEMDS lctl get_param -n ldlm_timeout`"
2084 LDLM_OST1="`do_facet ost1 lctl get_param -n ldlm_timeout`"
2085 LDLM_CLIENT="`do_facet client lctl get_param -n ldlm_timeout`"
2087 if [ $LDLM_MDS -ne $LDLM_OST1 ] || [ $LDLM_MDS -ne $LDLM_CLIENT ]; then
2088 error "Different LDLM_TIMEOUT:$LDLM_MDS $LDLM_OST1 $LDLM_CLIENT"
2091 if [ $LDLM_MDS -ne $((LOCAL_TIMEOUT - 1)) ]; then
2092 error "LDLM_TIMEOUT($LDLM_MDS) is not correct"
2095 cleanup || return $?
2097 MDS_MKFS_OPTS=$OLD_MDS_MKFS_OPTS
2098 OST_MKFS_OPTS=$OLD_OST_MKFS_OPTS
2100 run_test 49 "check PARAM_SYS_LDLM_TIMEOUT option of MKFS.LUSTRE"
2103 # Test both statfs and lfs df and fail if either one fails
2104 multiop_bg_pause $1 f_
2107 killall -USR1 multiop
2108 [ $RC1 -ne 0 ] && log "lazystatfs multiop failed"
2109 wait $PID || { RC1=$?; log "multiop return error "; }
2116 if [ $RC2 -eq 0 ]; then
2118 log "lazystatfs df failed"
2122 [[ $RC1 -ne 0 || $RC2 -eq 0 ]] && RC=1
2128 lctl set_param llite.$FSNAME-*.lazystatfs=1
2131 lazystatfs $MOUNT || error "lazystatfs failed but no down servers"
2133 cleanup || return $?
2135 run_test 50a "lazystatfs all servers available =========================="
2139 lctl set_param llite.$FSNAME-*.lazystatfs=1
2142 # Wait for client to detect down OST
2143 stop_ost || error "Unable to stop OST1"
2144 wait_osc_import_state mds ost DISCONN
2146 lazystatfs $MOUNT || error "lazystatfs should don't have returned EIO"
2148 umount_client $MOUNT || error "Unable to unmount client"
2149 stop_mds || error "Unable to stop MDS"
2151 run_test 50b "lazystatfs all servers down =========================="
2154 start_mds || error "Unable to start MDS"
2155 start_ost || error "Unable to start OST1"
2156 start_ost2 || error "Unable to start OST2"
2157 mount_client $MOUNT || error "Unable to mount client"
2158 lctl set_param llite.$FSNAME-*.lazystatfs=1
2161 # Wait for client to detect down OST
2162 stop_ost || error "Unable to stop OST1"
2163 wait_osc_import_state mds ost DISCONN
2164 lazystatfs $MOUNT || error "lazystatfs failed with one down server"
2166 umount_client $MOUNT || error "Unable to unmount client"
2167 stop_ost2 || error "Unable to stop OST2"
2168 stop_mds || error "Unable to stop MDS"
2170 run_test 50c "lazystatfs one server down =========================="
2173 start_mds || error "Unable to start MDS"
2174 start_ost || error "Unable to start OST1"
2175 start_ost2 || error "Unable to start OST2"
2176 mount_client $MOUNT || error "Unable to mount client"
2177 lctl set_param llite.$FSNAME-*.lazystatfs=1
2180 # Issue the statfs during the window where the client still
2181 # belives the OST to be available but it is in fact down.
2182 # No failure just a statfs which hangs for a timeout interval.
2183 stop_ost || error "Unable to stop OST1"
2184 lazystatfs $MOUNT || error "lazystatfs failed with one down server"
2186 umount_client $MOUNT || error "Unable to unmount client"
2187 stop_ost2 || error "Unable to stop OST2"
2188 stop_mds || error "Unable to stop MDS"
2190 run_test 50d "lazystatfs client/server conn race =========================="
2197 start_mds || return 1
2198 #first client should see only one ost
2199 start_ost || return 2
2200 wait_osc_import_state mds ost FULL
2202 # Wait for client to detect down OST
2203 stop_ost || error "Unable to stop OST1"
2204 wait_osc_import_state mds ost DISCONN
2206 mount_client $MOUNT || error "Unable to mount client"
2207 lctl set_param llite.$FSNAME-*.lazystatfs=0
2209 multiop_bg_pause $MOUNT _f
2213 if [ $RC1 -ne 0 ]; then
2214 log "multiop failed $RC1"
2217 sleep $(( $TIMEOUT+1 ))
2219 [ $? -ne 0 ] && error "process isn't sleep"
2220 start_ost || error "Unable to start OST1"
2221 wait $pid || error "statfs failed"
2224 umount_client $MOUNT || error "Unable to unmount client"
2225 stop_ost || error "Unable to stop OST1"
2226 stop_mds || error "Unable to stop MDS"
2228 run_test 50e "normal statfs all servers down =========================="
2233 CONN_PROC="osc.$FSNAME-OST0001-osc-[M]*.ost_server_uuid"
2235 start_mds || error "Unable to start mds"
2236 #first client should see only one ost
2237 start_ost || error "Unable to start OST1"
2238 wait_osc_import_state mds ost FULL
2240 start_ost2 || error "Unable to start OST2"
2241 wait_osc_import_state mds ost2 FULL
2243 # Wait for client to detect down OST
2244 stop_ost2 || error "Unable to stop OST2"
2246 wait_osc_import_state mds ost2 DISCONN
2247 mount_client $MOUNT || error "Unable to mount client"
2248 lctl set_param llite.$FSNAME-*.lazystatfs=0
2250 multiop_bg_pause $MOUNT _f
2254 if [ $RC1 -ne 0 ]; then
2255 log "lazystatfs multiop failed $RC1"
2258 sleep $(( $TIMEOUT+1 ))
2260 [ $? -ne 0 ] && error "process isn't sleep"
2261 start_ost2 || error "Unable to start OST2"
2262 wait $pid || error "statfs failed"
2263 stop_ost2 || error "Unable to stop OST2"
2266 umount_client $MOUNT || error "Unable to unmount client"
2267 stop_ost || error "Unable to stop OST1"
2268 stop_mds || error "Unable to stop MDS"
2269 #writeconf to remove all ost2 traces for subsequent tests
2272 run_test 50f "normal statfs one server in down =========================="
2275 [ "$OSTCOUNT" -lt "2" ] && skip_env "$OSTCOUNT < 2, skipping" && return
2277 start_ost2 || error "Unable to start OST2"
2278 wait_osc_import_state mds ost2 FULL
2279 wait_osc_import_state client ost2 FULL
2281 local PARAM="${FSNAME}-OST0001.osc.active"
2283 $LFS setstripe -c -1 $DIR/$tfile || error "Unable to lfs setstripe"
2284 do_facet mgs $LCTL conf_param $PARAM=0 || error "Unable to deactivate OST"
2286 umount_client $MOUNT || error "Unable to unmount client"
2287 mount_client $MOUNT || error "Unable to mount client"
2288 # This df should not cause a panic
2291 do_facet mgs $LCTL conf_param $PARAM=1 || error "Unable to activate OST"
2293 umount_client $MOUNT || error "Unable to unmount client"
2294 stop_ost2 || error "Unable to stop OST2"
2295 stop_ost || error "Unable to stop OST1"
2296 stop_mds || error "Unable to stop MDS"
2297 #writeconf to remove all ost2 traces for subsequent tests
2300 run_test 50g "deactivated OST should not cause panic====================="
2303 local LOCAL_TIMEOUT=20
2307 check_mount || return 1
2310 $LFS setstripe -c -1 $MOUNT/d1
2311 #define OBD_FAIL_MDS_REINT_DELAY 0x142
2312 do_facet $SINGLEMDS "lctl set_param fail_loc=0x142"
2313 touch $MOUNT/d1/f1 &
2316 start_ost2 || return 2
2318 stop_ost2 || return 3
2321 run_test 51 "Verify that mdt_reint handles RMF_MDT_MD correctly when an OST is added"
2330 do_node $node mkdir -p $dest
2331 [ $? -eq 0 ] || { error "Unable to create directory"; return 1; }
2333 do_node $node 'tar cf - '$@' | tar xf - -C '$dest';
2334 [ \"\${PIPESTATUS[*]}\" = \"0 0\" ] || exit 1'
2335 [ $? -eq 0 ] || { error "Unable to tar files"; return 2; }
2337 do_node $node 'getfattr -d -m "[a-z]*\\." '$@' > '$xattrs
2338 [ $? -eq 0 ] || { error "Unable to read xattrs"; return 3; }
2348 local backup2=${TMP}/backup2
2350 do_node $node mkdir -p $backup2
2351 [ $? -eq 0 ] || { error "Unable to create directory"; return 1; }
2353 do_node $node 'tar cf - '$@' | tar xf - -C '$backup2';
2354 [ \"\${PIPESTATUS[*]}\" = \"0 0\" ] || exit 1'
2355 [ $? -eq 0 ] || { error "Unable to tar files to diff"; return 2; }
2357 do_node $node "diff -rq $backup $backup2"
2358 [ $? -eq 0 ] || { error "contents differ"; return 3; }
2360 local xattrs2=${TMP}/xattrs2
2361 do_node $node 'getfattr -d -m "[a-z]*\\." '$@' > '$xattrs2
2362 [ $? -eq 0 ] || { error "Unable to read xattrs to diff"; return 4; }
2364 do_node $node "diff $xattrs $xattrs2"
2365 [ $? -eq 0 ] || { error "xattrs differ"; return 5; }
2367 do_node $node "rm -rf $backup2 $xattrs2"
2368 [ $? -eq 0 ] || { error "Unable to delete temporary files"; return 6; }
2373 [ $? -eq 0 ] || { error "Unable to start MDS"; return 1; }
2375 [ $? -eq 0 ] || { error "Unable to start OST1"; return 2; }
2377 [ $? -eq 0 ] || { error "Unable to mount client"; return 3; }
2380 local ost1mnt=${MOUNT%/*}/ost1
2381 local ost1node=$(facet_active_host ost1)
2382 local ost1tmp=$TMP/conf52
2385 [ $? -eq 0 ] || { error "Unable to create tdir"; return 4; }
2386 touch $TMP/modified_first
2387 [ $? -eq 0 ] || { error "Unable to create temporary file"; return 5; }
2388 local mtime=$(stat -c %Y $TMP/modified_first)
2389 do_node $ost1node "mkdir -p $ost1tmp && touch -m -d @$mtime $ost1tmp/modified_first"
2391 [ $? -eq 0 ] || { error "Unable to create temporary file"; return 6; }
2394 $LFS setstripe $DIR/$tdir -c -1 -s 1M
2395 [ $? -eq 0 ] || { error "lfs setstripe failed"; return 7; }
2397 for (( i=0; i < nrfiles; i++ )); do
2398 multiop $DIR/$tdir/$tfile-$i Ow1048576w1048576w524288c
2399 [ $? -eq 0 ] || { error "multiop failed"; return 8; }
2405 echo backup files to $TMP/files
2406 local files=$(find $DIR/$tdir -type f -newer $TMP/modified_first)
2407 copy_files_xattrs `hostname` $TMP/files $TMP/file_xattrs $files
2408 [ $? -eq 0 ] || { error "Unable to copy files"; return 9; }
2410 umount_client $MOUNT
2411 [ $? -eq 0 ] || { error "Unable to umount client"; return 10; }
2413 [ $? -eq 0 ] || { error "Unable to stop ost1"; return 11; }
2415 echo mount ost1 as ldiskfs
2416 do_node $ost1node mount -t $FSTYPE $ost1_dev $ost1mnt $OST_MOUNT_OPTS
2417 [ $? -eq 0 ] || { error "Unable to mount ost1 as ldiskfs"; return 12; }
2420 echo backup objects to $ost1tmp/objects
2421 local objects=$(do_node $ost1node 'find '$ost1mnt'/O/0 -type f -size +0'\
2422 '-newer '$ost1tmp'/modified_first -regex ".*\/[0-9]+"')
2423 copy_files_xattrs $ost1node $ost1tmp/objects $ost1tmp/object_xattrs $objects
2424 [ $? -eq 0 ] || { error "Unable to copy objects"; return 13; }
2426 # move objects to lost+found
2427 do_node $ost1node 'mv '$objects' '${ost1mnt}'/lost+found'
2428 [ $? -eq 0 ] || { error "Unable to move objects"; return 14; }
2431 do_node $ost1node "ll_recover_lost_found_objs -d $ost1mnt/lost+found"
2432 [ $? -eq 0 ] || { error "ll_recover_lost_found_objs failed"; return 15; }
2434 # compare restored objects against saved ones
2435 diff_files_xattrs $ost1node $ost1tmp/objects $ost1tmp/object_xattrs $objects
2436 [ $? -eq 0 ] || { error "Unable to diff objects"; return 16; }
2438 do_node $ost1node "umount $ost1_dev"
2439 [ $? -eq 0 ] || { error "Unable to umount ost1 as ldiskfs"; return 17; }
2442 [ $? -eq 0 ] || { error "Unable to start ost1"; return 18; }
2444 [ $? -eq 0 ] || { error "Unable to mount client"; return 19; }
2447 diff_files_xattrs `hostname` $TMP/files $TMP/file_xattrs $files
2448 [ $? -eq 0 ] || { error "Unable to diff files"; return 20; }
2450 rm -rf $TMP/files $TMP/file_xattrs
2451 [ $? -eq 0 ] || { error "Unable to delete temporary files"; return 21; }
2452 do_node $ost1node "rm -rf $ost1tmp"
2453 [ $? -eq 0 ] || { error "Unable to delete temporary files"; return 22; }
2456 run_test 52 "check recovering objects from lost+found"
2458 # Checks threads_min/max/started for some service
2460 # Arguments: service name (OST or MDT), facet (e.g., ost1, $SINGLEMDS), and a
2461 # parameter pattern prefix like 'ost.*.ost'.
2473 local msg="Insane $modname thread counts"
2477 check_mount || return 41
2479 # We need to expand $parampat, but it may match multiple parameters, so
2480 # we'll pick the first one
2481 if ! paramp=$(do_facet $facet "lctl get_param -N ${parampat}.threads_min"|head -1); then
2482 error "Couldn't expand ${parampat}.threads_min parameter name"
2486 # Remove the .threads_min part
2487 paramp=${paramp%.threads_min}
2489 # Check for sanity in defaults
2490 tmin=$(do_facet $facet "lctl get_param -n ${paramp}.threads_min" || echo 0)
2491 tmax=$(do_facet $facet "lctl get_param -n ${paramp}.threads_max" || echo 0)
2492 tstarted=$(do_facet $facet "lctl get_param -n ${paramp}.threads_started" || echo 0)
2493 lassert 23 "$msg (PDSH problems?)" '(($tstarted && $tmin && $tmax))' || return $?
2494 lassert 24 "$msg" '(($tstarted >= $tmin && $tstarted <= $tmax ))' || return $?
2496 # Check that we can lower min/max
2497 do_facet $facet "lctl set_param ${paramp}.threads_min=$((tmin - 1))"
2498 do_facet $facet "lctl set_param ${paramp}.threads_max=$((tmax - 1))"
2499 tmin2=$(do_facet $facet "lctl get_param -n ${paramp}.threads_min" || echo 0)
2500 tmax2=$(do_facet $facet "lctl get_param -n ${paramp}.threads_max" || echo 0)
2501 lassert 25 "$msg" '(($tmin2 == ($tmin - 1) && $tmax2 == ($tmax -1)))' || return $?
2503 # Check that we can set min/max to the same value
2504 do_facet $facet "lctl set_param ${paramp}.threads_max=$((tmin - 1))"
2505 tmin2=$(do_facet $facet "lctl get_param -n ${paramp}.threads_min" || echo 0)
2506 tmax2=$(do_facet $facet "lctl get_param -n ${paramp}.threads_max" || echo 0)
2507 lassert 26 "$msg" '(($tmin2 == ($tmin - 1) && $tmax2 == ($tmin - 1)))' || return $?
2509 # Check that we can't set max < min
2510 do_facet $facet "lctl set_param ${paramp}.threads_max=$((tmin - 2))"
2511 tmin2=$(do_facet $facet "lctl get_param -n ${paramp}.threads_min" || echo 0)
2512 tmax2=$(do_facet $facet "lctl get_param -n ${paramp}.threads_max" || echo 0)
2513 lassert 27 "$msg" '(($tmin2 <= $tmax2))' || return $?
2515 # We need to ensure that we get the module options desired; to do this
2516 # we set LOAD_MODULES_REMOTE=true and we call setmodopts below.
2517 LOAD_MODULES_REMOTE=true
2520 setmodopts -a $modname "$opts" oldvalue
2524 check_mount || return 41
2526 # Restore previous setting of MODOPTS_*
2527 setmodopts $modname "$oldvalue"
2529 # Check that $opts took
2530 tmin=$(do_facet $facet "lctl get_param -n ${paramp}.threads_min")
2531 tmax=$(do_facet $facet "lctl get_param -n ${paramp}.threads_max")
2532 tstarted=$(do_facet $facet "lctl get_param -n ${paramp}.threads_started")
2533 lassert 28 "$msg" '(($tstarted == $tmin && $tstarted == $tmax ))' || return $?
2536 # Workaround a YALA bug where YALA expects that modules will remain
2537 # loaded on the servers
2538 LOAD_MODULES_REMOTE=false
2545 thread_sanity OST ost1 'ost.*.ost' 'oss_num_threads=64'
2547 run_test 53a "check OSS thread count params"
2550 thread_sanity MDT $SINGLEMDS 'mdt.*.*.' 'mdt_num_threads=64'
2552 run_test 53b "check MDT thread count params"
2557 local partial_arg=""
2558 local size=$(df -B G $dir | tail -1 | awk '{print $2}' | sed 's/G//') # Gb
2560 # Run in partial (fast) mode if the size
2561 # of a partition > 10 GB
2562 [ $size -gt 10 ] && partial_arg="-p"
2564 llverfs $partial_arg $dir
2568 do_rpc_nodes $(facet_host ost1) run_llverdev $(ostdevname 1)
2569 [ $? -eq 0 ] || error "llverdev failed!"
2572 run_test 54a "llverdev"
2577 [ $? -eq 0 ] || error "llverfs failed!"
2580 run_test 54b "llverfs"
2584 local max_ost_index=$1
2585 echo -n $(((max_ost_index + 1) * 8))
2589 local mdsdev=$(mdsdevname 1)
2590 local ostdev=$(ostdevname 1)
2591 local saved_opts=$OST_MKFS_OPTS
2593 for i in 0 1023 2048
2595 OST_MKFS_OPTS="$saved_opts --index $i"
2603 echo checking size of lov_objid for ost index $i
2604 LOV_OBJID_SIZE=$(do_facet mds1 "$DEBUGFS -R 'stat lov_objid' $mdsdev 2>/dev/null" | grep ^User | awk '{print $6}')
2605 if [ "$LOV_OBJID_SIZE" != $(lov_objid_size $i) ]; then
2606 error "lov_objid size has to be $(lov_objid_size $i), not $LOV_OBJID_SIZE"
2608 echo ok, lov_objid size is correct: $LOV_OBJID_SIZE
2613 OST_MKFS_OPTS=$saved_opts
2616 run_test 55 "check lov_objid size"
2619 add mds1 $MDS_MKFS_OPTS --mkfsoptions='\"-J size=16\"' --reformat $(mdsdevname 1)
2620 add ost1 $OST_MKFS_OPTS --index=1000 --reformat $(ostdevname 1)
2621 add ost2 $OST_MKFS_OPTS --index=10000 --reformat $(ostdevname 2)
2625 start_ost2 || error "Unable to start second ost"
2626 mount_client $MOUNT || error "Unable to mount client"
2629 [ -n "$ENABLE_QUOTA" ] && { $LFS quotacheck -ug $MOUNT || error "quotacheck has failed" ; }
2633 run_test 56 "check big indexes"
2635 test_57() { # bug 22656
2636 local NID=$(do_facet ost1 "$LCTL get_param nis" | tail -1 | awk '{print $1}')
2638 do_facet ost1 "$TUNEFS --failnode=$NID `ostdevname 1`" || error "tunefs failed"
2640 start_ost && error "OST registration from failnode should fail"
2644 run_test 57 "initial registration from failnode should fail (should return errs)"
2647 do_facet mgs $LCTL get_param mgs.MGS.live.$FSNAME | grep OST | wc -l
2651 start_mgsmds >> /dev/null
2652 local C1=$(count_osts)
2653 if [ $C1 -eq 0 ]; then
2654 start_ost >> /dev/null
2658 echo "original ost count: $C1 (expect > 0)"
2659 [ $C1 -gt 0 ] || error "No OSTs in $FSNAME log"
2660 start_mgsmds -o writeconf >> /dev/null || error "MDT start failed"
2661 local C2=$(count_osts)
2662 echo "after mdt writeconf count: $C2 (expect 0)"
2663 [ $C2 -gt 0 ] && error "MDT writeconf should erase OST logs"
2664 echo "OST start without writeconf should fail:"
2665 start_ost >> /dev/null && error "OST start without writeconf didn't fail"
2666 echo "OST start with writeconf should succeed:"
2667 start_ost -o writeconf >> /dev/null || error "OST1 start failed"
2668 local C3=$(count_osts)
2669 echo "after ost writeconf count: $C3 (expect 1)"
2670 [ $C3 -eq 1 ] || error "new OST writeconf should add:"
2671 start_ost2 -o writeconf >> /dev/null || error "OST2 start failed"
2672 local C4=$(count_osts)
2673 echo "after ost2 writeconf count: $C4 (expect 2)"
2674 [ $C4 -eq 2 ] || error "OST2 writeconf should add log"
2675 stop_ost2 >> /dev/null
2676 cleanup_nocli >> /dev/null
2678 run_test 59 "writeconf mount option"
2681 test_58() { # bug 22658
2682 [ "$FSTYPE" != "ldiskfs" ] && skip "not supported for $FSTYPE" && return
2685 createmany -o $DIR/$tdir/$tfile-%d 100
2686 # make sure that OSTs do not cancel llog cookies before we unmount the MDS
2687 #define OBD_FAIL_OBD_LOG_CANCEL_NET 0x601
2688 do_facet mds "lctl set_param fail_loc=0x601"
2689 unlinkmany $DIR/$tdir/$tfile-%d 100
2691 local MNTDIR=$(facet_mntpt mds)
2692 # remove all files from the OBJECTS dir
2693 do_facet mds "mount -t ldiskfs $MDSDEV $MNTDIR"
2694 do_facet mds "find $MNTDIR/OBJECTS -type f -delete"
2695 do_facet mds "umount $MNTDIR"
2696 # restart MDS with missing llog files
2698 do_facet mds "lctl set_param fail_loc=0"
2701 run_test 58 "missing llog files must not prevent MDT from mounting"
2703 if ! combined_mgs_mds ; then
2707 equals_msg `basename $0`: test complete
2708 [ -f "$TESTSUITELOG" ] && cat $TESTSUITELOG && grep -q FAIL $TESTSUITELOG && exit 1 || true