3 # add uml1 uml2 uml3 in your /etc/hosts
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: 13369
15 ALWAYS_EXCEPT=" $CONF_SANITY_EXCEPT 34a"
16 # UPDATE THE COMMENT ABOVE WITH BUG NUMBERS WHEN CHANGING ALWAYS_EXCEPT!
19 PATH=$PWD/$SRCDIR:$SRCDIR:$SRCDIR/../utils:$PATH
22 LUSTRE=${LUSTRE:-`dirname $0`/..}
23 RLUSTRE=${RLUSTRE:-$LUSTRE}
26 . $LUSTRE/tests/test-framework.sh
28 # STORED_MDSSIZE is used in test_18
29 if [ -n "$MDSSIZE" ]; then
30 STORED_MDSSIZE=$MDSSIZE
32 # use small MDS + OST size to speed formatting time
35 . ${CONFIG:=$LUSTRE/tests/cfg/$NAME.sh}
37 remote_mds_nodsh && skip "remote MDS with nodsh" && exit 0
38 remote_ost_nodsh && skip "remote OST with nodsh" && exit 0
41 [ "$SLOW" = "no" ] && EXCEPT_SLOW="0 1 2 3 6 7 15 18 24b 25 30 31 32 33 34a 45"
54 # who knows if/where $TUNEFS is installed? Better reformat if it fails...
55 do_facet ${facet} "$TUNEFS --writeconf $MDSDEV" || echo "tunefs failed, reformatting instead" && reformat
61 # The MGS must be started before the OSTs for a new fs, so start
62 # and stop to generate the startup logs.
65 wait_osc_import_state mds ost FULL
70 reformat_and_config() {
76 echo "start mds service on `facet_active_host mds`"
77 start mds $MDSDEV $MDS_MOUNT_OPTS || return 94
81 echo "stop mds service on `facet_active_host mds`"
82 # These tests all use non-failover stop
83 stop mds -f || return 97
87 echo "start ost1 service on `facet_active_host ost1`"
88 start ost1 `ostdevname 1` $OST_MOUNT_OPTS || return 95
92 echo "stop ost1 service on `facet_active_host ost1`"
93 # These tests all use non-failover stop
94 stop ost1 -f || return 98
98 echo "start ost2 service on `facet_active_host ost2`"
99 start ost2 `ostdevname 2` $OST_MOUNT_OPTS || return 92
103 echo "stop ost2 service on `facet_active_host ost2`"
104 # These tests all use non-failover stop
105 stop ost2 -f || return 93
109 echo "start client on `facet_active_host client`"
110 start client || return 99
114 echo "stop client on `facet_active_host client`"
115 stop client || return 100
120 echo "mount $FSNAME on ${MOUNTPATH}....."
121 zconf_mount `hostname` $MOUNTPATH || return 96
125 local SAVEMOUNTOPT=$MOUNTOPT
126 MOUNTOPT="remount,$1"
128 echo "remount '$1' lustre on ${MOUNTPATH}....."
129 zconf_mount `hostname` $MOUNTPATH || return 96
130 MOUNTOPT=$SAVEMOUNTOPT
135 echo "umount lustre on ${MOUNTPATH}....."
136 zconf_umount `hostname` $MOUNTPATH || return 97
139 manual_umount_client(){
142 echo "manual umount lustre on ${MOUNT}...."
143 do_facet client "umount -d ${FORCE} $MOUNT"
162 stop_mds || return 201
163 stop_ost || return 202
164 unload_modules || return 203
168 umount_client $MOUNT || return 200
169 cleanup_nocli || return $?
173 do_facet client "cp /etc/passwd $DIR/a" || return 71
174 do_facet client "rm $DIR/a" || return 72
175 # make sure lustre is actually mounted (touch will block,
176 # but grep won't, so do it after)
177 do_facet client "grep $MOUNT' ' /proc/mounts > /dev/null" || return 73
178 echo "setup single mount lustre success"
182 do_facet client "touch $DIR/a" || return 71
183 do_facet client "rm $DIR/a" || return 72
184 do_facet client "touch $DIR2/a" || return 73
185 do_facet client "rm $DIR2/a" || return 74
186 echo "setup double mount lustre success"
191 if [ "$ONLY" == "setup" ]; then
196 if [ "$ONLY" == "cleanup" ]; then
201 #create single point mountpoint
207 check_mount || return 41
210 run_test 0 "single mount setup"
214 echo "start ost second time..."
216 check_mount || return 42
219 run_test 1 "start up ost twice (should return errors)"
224 echo "start mds second time.."
227 check_mount || return 43
230 run_test 2 "start up mds twice (should return err)"
234 #mount.lustre returns an error if already in mtab
235 mount_client $MOUNT && return $?
236 check_mount || return 44
239 run_test 3 "mount client twice (should return err)"
243 touch $DIR/$tfile || return 85
247 # ok for ost to fail shutdown
248 if [ 202 -ne $eno ]; then
253 run_test 4 "force cleanup ost, then cleanup"
255 test_5a() { # was test_5
257 touch $DIR/$tfile || return 1
258 fuser -m -v $MOUNT && echo "$MOUNT is in use by user space process."
260 stop_mds -f || return 2
262 # cleanup may return an error from the failed
263 # disconnects; for now I'll consider this successful
264 # if all the modules have unloaded.
268 echo "killing umount"
269 kill -TERM $UMOUNT_PID
270 echo "waiting for umount to finish"
272 if grep " $MOUNT " /proc/mounts; then
273 echo "test 5: /proc/mounts after failed umount"
277 echo "killing umount"
278 kill -TERM $UMOUNT_PID
279 echo "waiting for umount to finish"
281 grep " $MOUNT " /proc/mounts && echo "test 5: /proc/mounts after second umount" && return 11
285 # stop_mds is a no-op here, and should not fail
286 cleanup_nocli || return $?
287 # df may have lingering entry
289 # mtab may have lingering entry
293 while [ "$WAIT" -ne "$MAX_WAIT" ]; do
295 grep -q $MOUNT" " /etc/mtab || break
296 echo "Waiting /etc/mtab updated ... "
297 WAIT=$(( WAIT + sleep))
299 [ "$WAIT" -eq "$MAX_WAIT" ] && error "/etc/mtab is not updated in $WAIT secs"
300 echo "/etc/mtab updated in $WAIT secs"
302 run_test 5a "force cleanup mds, then cleanup"
306 [ -d $MOUNT ] || mkdir -p $MOUNT
307 grep " $MOUNT " /etc/mtab && echo "test 5b: mtab before mount" && return 10
308 mount_client $MOUNT && return 1
309 grep " $MOUNT " /etc/mtab && echo "test 5b: mtab after failed mount" && return 11
311 # stop_mds is a no-op here, and should not fail
312 cleanup_nocli || return $?
315 run_test 5b "mds down, cleanup after failed mount (bug 2712) (should return errs)"
320 [ -d $MOUNT ] || mkdir -p $MOUNT
321 grep " $MOUNT " /etc/mtab && echo "test 5c: mtab before mount" && return 10
322 local oldfs="${FSNAME}"
323 FSNAME="wrong.${FSNAME}"
324 mount_client $MOUNT || :
326 grep " $MOUNT " /etc/mtab && echo "test 5c: mtab after failed mount" && return 11
328 cleanup_nocli || return $?
330 run_test 5c "cleanup after failed mount (bug 2712) (should return errs)"
336 grep " $MOUNT " /etc/mtab && echo "test 5d: mtab before mount" && return 10
337 mount_client $MOUNT || return 1
339 grep " $MOUNT " /etc/mtab && echo "test 5d: mtab after unmount" && return 11
342 run_test 5d "mount with ost down"
348 #define OBD_FAIL_PTLRPC_DELAY_SEND 0x506
349 do_facet client "lctl set_param fail_loc=0x80000506"
350 grep " $MOUNT " /etc/mtab && echo "test 5e: mtab before mount" && return 10
351 mount_client $MOUNT || echo "mount failed (not fatal)"
353 grep " $MOUNT " /etc/mtab && echo "test 5e: mtab after unmount" && return 11
356 run_test 5e "delayed connect, don't crash (bug 10268)"
361 mount_client ${MOUNT} || return 87
362 touch $DIR/a || return 86
365 run_test 6 "manual umount, then mount again"
370 cleanup_nocli || return $?
372 run_test 7 "manual umount, then cleanup"
377 check_mount2 || return 45
378 umount_client $MOUNT2
381 run_test 8 "double mount setup"
386 do_facet ost1 lctl set_param debug=\'inode trace\' || return 1
387 do_facet ost1 lctl set_param subsystem_debug=\'mds ost\' || return 1
389 CHECK_PTLDEBUG="`do_facet ost1 lctl get_param -n debug`"
390 if [ "$CHECK_PTLDEBUG" ] && [ "$CHECK_PTLDEBUG" = "trace inode" ];then
391 echo "lnet.debug success"
393 echo "lnet.debug: want 'trace inode', have '$CHECK_PTLDEBUG'"
396 CHECK_SUBSYS="`do_facet ost1 lctl get_param -n subsystem_debug`"
397 if [ "$CHECK_SUBSYS" ] && [ "$CHECK_SUBSYS" = "mds ost" ]; then
398 echo "lnet.subsystem_debug success"
400 echo "lnet.subsystem_debug: want 'mds ost', have '$CHECK_SUBSYS'"
403 stop_ost || return $?
406 run_test 9 "test ptldebug and subsystem for mkfs"
409 local TMPMTPT="${TMP}/conf16"
411 if [ ! -e "$MDSDEV" ]; then
412 log "no $MDSDEV existing, so mount Lustre to create one"
414 check_mount || return 41
418 [ -f "$MDSDEV" ] && LOOPOPT="-o loop"
420 log "change the mode of $MDSDEV/OBJECTS,LOGS,PENDING to 555"
421 do_facet mds "mkdir -p $TMPMTPT &&
422 mount $LOOPOPT -t $FSTYPE $MDSDEV $TMPMTPT &&
423 chmod 555 $TMPMTPT/{OBJECTS,LOGS,PENDING} &&
424 umount $TMPMTPT" || return $?
426 log "mount Lustre to change the mode of OBJECTS/LOGS/PENDING, then umount Lustre"
428 check_mount || return 41
431 log "read the mode of OBJECTS/LOGS/PENDING and check if they has been changed properly"
432 EXPECTEDOBJECTSMODE=`do_facet mds "$DEBUGFS -R 'stat OBJECTS' $MDSDEV 2> /dev/null" | grep 'Mode: ' | sed -e "s/.*Mode: *//" -e "s/ *Flags:.*//"`
433 EXPECTEDLOGSMODE=`do_facet mds "$DEBUGFS -R 'stat LOGS' $MDSDEV 2> /dev/null" | grep 'Mode: ' | sed -e "s/.*Mode: *//" -e "s/ *Flags:.*//"`
434 EXPECTEDPENDINGMODE=`do_facet mds "$DEBUGFS -R 'stat PENDING' $MDSDEV 2> /dev/null" | grep 'Mode: ' | sed -e "s/.*Mode: *//" -e "s/ *Flags:.*//"`
436 if [ "$EXPECTEDOBJECTSMODE" = "0777" ]; then
437 log "Success:Lustre change the mode of OBJECTS correctly"
439 error "Lustre does not change mode of OBJECTS properly"
442 if [ "$EXPECTEDLOGSMODE" = "0777" ]; then
443 log "Success:Lustre change the mode of LOGS correctly"
445 error "Lustre does not change mode of LOGS properly"
448 if [ "$EXPECTEDPENDINGMODE" = "0777" ]; then
449 log "Success:Lustre change the mode of PENDING correctly"
451 error "Lustre does not change mode of PENDING properly"
454 run_test 16 "verify that lustre will correct the mode of OBJECTS/LOGS/PENDING"
457 if [ ! -e "$MDSDEV" ]; then
458 echo "no $MDSDEV existing, so mount Lustre to create one"
460 check_mount || return 41
464 echo "Remove mds config log"
465 do_facet mds "$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
479 # check if current MDSSIZE is large enough
480 [ $MDSSIZE -ge $MIN ] && OK=1 && myMDSSIZE=$MDSSIZE && \
481 log "use MDSSIZE=$MDSSIZE"
483 # check if the global config has a large enough MDSSIZE
484 [ -z "$OK" -a ! -z "$STORED_MDSSIZE" ] && [ $STORED_MDSSIZE -ge $MIN ] && \
485 OK=1 && myMDSSIZE=$STORED_MDSSIZE && \
486 log "use STORED_MDSSIZE=$STORED_MDSSIZE"
488 # check if the block device is large enough
489 [ -z "$OK" -a -b $MDSDEV ] && \
490 [ "$(dd if=$MDSDEV of=/dev/null bs=1k count=1 skip=$MIN 2>&1 |
491 awk '($3 == "in") { print $1 }')" = "1+0" ] && OK=1 && \
492 myMDSSIZE=$MIN && log "use device $MDSDEV with MIN=$MIN"
494 # check if a loopback device has enough space for fs metadata (5%)
495 [ -z "$OK" ] && [ -f $MDSDEV -o ! -e $MDSDEV ] &&
496 SPACE=$(df -P $(dirname $MDSDEV) |
497 awk '($1 != "Filesystem") {print $4}') &&
498 [ $SPACE -gt $((MIN / 20)) ] && OK=1 && myMDSSIZE=$MIN && \
499 log "use file $MDSDEV with MIN=$MIN"
501 [ -z "$OK" ] && skip_env "$MDSDEV too small for ${MIN}kB MDS" && return
504 echo "mount mds with large journal..."
505 local OLD_MDS_MKFS_OPTS=$MDS_MKFS_OPTS
507 MDS_MKFS_OPTS="--mgs --mdt --fsname=$FSNAME --device-size=$myMDSSIZE --param sys.timeout=$TIMEOUT $MDSOPT"
510 echo "mount lustre system..."
512 check_mount || return 41
514 echo "check journal size..."
515 local FOUNDSIZE=`do_facet mds "$DEBUGFS -c -R 'stat <8>' $MDSDEV" | awk '/Size: / { print $NF; exit;}'`
516 if [ $FOUNDSIZE -gt $((32 * 1024 * 1024)) ]; then
517 log "Success: mkfs creates large journals. Size: $((FOUNDSIZE >> 20))M"
519 error "expected journal size > 32M, found $((FOUNDSIZE >> 20))M"
524 MDS_MKFS_OPTS=$OLD_MDS_MKFS_OPTS
527 run_test 18 "check mkfs creates large journals"
530 start_mds || return 1
531 stop_mds -f || return 2
533 run_test 19a "start/stop MDS without OSTs"
536 start_ost || return 1
537 stop_ost -f || return 2
539 run_test 19b "start/stop OSTs without MDS"
542 # first format the ost/mdt
546 check_mount || return 43
548 remount_client ro $MOUNT || return 44
549 touch $DIR/$tfile && echo "$DIR/$tfile created incorrectly" && return 45
550 [ -e $DIR/$tfile ] && echo "$DIR/$tfile exists incorrectly" && return 46
551 remount_client rw $MOUNT || return 47
553 [ ! -f $DIR/$tfile ] && echo "$DIR/$tfile missing" && return 48
554 MCNT=`grep -c $MOUNT /etc/mtab`
555 [ "$MCNT" -ne 1 ] && echo "$MOUNT in /etc/mtab $MCNT times" && return 49
560 run_test 20 "remount ro,rw mounts work and doesn't break /etc/mtab"
565 wait_osc_import_state mds ost FULL
569 run_test 21a "start mds before ost, stop ost first"
574 wait_osc_import_state mds ost FULL
578 run_test 21b "start ost before mds, stop mds first"
584 wait_osc_import_state mds ost2 FULL
589 #writeconf to remove all ost2 traces for subsequent tests
592 run_test 21c "start mds between two osts, stop mds last"
597 echo Client mount with ost in logs, but none running
599 wait_osc_import_state mds ost FULL
602 # check_mount will block trying to contact ost
603 mcreate $DIR/$tfile || return 40
604 rm -f $DIR/$tfile || return 42
608 echo Client mount with a running ost
611 check_mount || return 41
616 run_test 22 "start a client before osts (should return errs)"
618 test_23a() { # was test_23
622 # force down client so that recovering mds waits for reconnect
623 local running=$(grep -c $MOUNT /proc/mounts) || true
624 if [ $running -ne 0 ]; then
625 echo "Stopping client $MOUNT (opts: -f)"
629 # enter recovery on mds
631 # try to start a new client
632 mount_client $MOUNT &
634 MOUNT_PID=$(ps -ef | grep "t lustre" | grep -v grep | awk '{print $2}')
635 MOUNT_LUSTRE_PID=`ps -ef | grep mount.lustre | grep -v grep | awk '{print $2}'`
636 echo mount pid is ${MOUNT_PID}, mount.lustre pid is ${MOUNT_LUSTRE_PID}
638 ps --ppid $MOUNT_LUSTRE_PID
639 # FIXME why o why can't I kill these? Manual "ctrl-c" works...
640 kill -TERM $MOUNT_LUSTRE_PID
641 echo "waiting for mount to finish"
643 # we can not wait $MOUNT_PID because it is not a child of this shell
649 while [ "$WAIT" -lt "$MAX_WAIT" ]; do
651 PID1=$(ps -ef | awk '{print $2}' | grep -w $MOUNT_PID)
652 PID2=$(ps -ef | awk '{print $2}' | grep -w $MOUNT_LUSTRE_PID)
655 [ -z "$PID1" -a -z "$PID2" ] && break
656 echo "waiting for mount to finish ... "
657 WAIT=$(( WAIT + sleep))
659 [ "$WAIT" -eq "$MAX_WAIT" ] && error "MOUNT_PID $MOUNT_PID and \
660 MOUNT__LUSTRE_PID $MOUNT__LUSTRE_PID still not killed in $WAIT secs"
665 run_test 23a "interrupt client during recovery mount delay"
670 test_23b() { # was test_23
673 # Simulate -EINTR during mount OBD_FAIL_LDLM_CLOSE_THREAD
674 lctl set_param fail_loc=0x80000313
678 run_test 23b "Simulate -EINTR during mount"
680 fs2mds_HOST=$mds_HOST
681 fs2ost_HOST=$ost_HOST
685 echo "umount $MOUNT2 ..."
686 umount $MOUNT2 || true
687 echo "stopping fs2mds ..."
688 stop fs2mds -f || true
689 echo "stopping fs2ost ..."
690 stop fs2ost -f || true
694 [ -n "$ost1_HOST" ] && fs2ost_HOST=$ost1_HOST
695 if [ -z "$fs2ost_DEV" -o -z "$fs2mds_DEV" ]; then
696 do_facet mds [ -b "$MDSDEV" ] && \
697 skip_env "mixed loopback and real device not working" && return
700 local fs2mdsdev=${fs2mds_DEV:-${MDSDEV}_2}
701 local fs2ostdev=${fs2ost_DEV:-$(ostdevname 1)_2}
703 # test 8-char fsname as well
704 local FSNAME2=test1234
705 add fs2mds $MDS_MKFS_OPTS --fsname=${FSNAME2} --nomgs --mgsnode=$MGSNID --reformat $fs2mdsdev || exit 10
707 add fs2ost $OST_MKFS_OPTS --fsname=${FSNAME2} --reformat $fs2ostdev || exit 10
710 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS && trap cleanup_24a EXIT INT
711 start fs2ost $fs2ostdev $OST_MOUNT_OPTS
713 mount -t lustre $MGSNID:/${FSNAME2} $MOUNT2 || return 1
715 check_mount || return 2
716 # files written on 1 should not show up on 2
717 cp /etc/passwd $DIR/$tfile
719 [ -e $MOUNT2/$tfile ] && error "File bleed" && return 7
721 cp /etc/passwd $MOUNT2/b || return 3
722 rm $MOUNT2/b || return 4
723 # 2 is actually mounted
724 grep $MOUNT2' ' /proc/mounts > /dev/null || return 5
726 facet_failover fs2mds
727 facet_failover fs2ost
730 # the MDS must remain up until last MDT
732 MDS=$(do_facet mds "lctl get_param -n devices" | awk '($3 ~ "mdt" && $4 ~ "MDS") { print $4 }')
733 [ -z "$MDS" ] && error "No MDS" && return 8
735 cleanup_nocli || return 6
737 run_test 24a "Multiple MDTs on a single node"
740 if [ -z "$fs2mds_DEV" ]; then
741 do_facet mds [ -b "$MDSDEV" ] && \
742 skip_env "mixed loopback and real device not working" && return
745 local fs2mdsdev=${fs2mds_DEV:-${MDSDEV}_2}
747 add fs2mds $MDS_MKFS_OPTS --fsname=${FSNAME}2 --mgs --reformat $fs2mdsdev || exit 10
749 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS && return 2
752 run_test 24b "Multiple MGSs on a single node (should return err)"
756 check_mount || return 2
757 local MODULES=$($LCTL modules | awk '{ print $2 }')
758 rmmod $MODULES 2>/dev/null || true
761 run_test 25 "Verify modules are referenced"
765 # we need modules before mount for sysctl, so make sure...
766 do_facet mds "lsmod | grep -q lustre || modprobe lustre"
767 #define OBD_FAIL_MDS_FS_SETUP 0x135
768 do_facet mds "lctl set_param fail_loc=0x80000135"
769 start_mds && echo MDS started && return 1
770 lctl get_param -n devices
771 DEVS=$(lctl get_param -n devices | wc -l)
772 [ $DEVS -gt 0 ] && return 2
773 unload_modules || return 203
775 run_test 26 "MDT startup failure cleans LOV (should return errs)"
781 local ORIG=$(do_facet $myfacet "$TEST")
782 if [ $# -gt 3 ]; then
788 echo "Setting $PARAM from $ORIG to $FINAL"
789 do_facet mds "$LCTL conf_param $PARAM=$FINAL" || error conf_param failed
791 wait_update $(facet_host $myfacet) "$TEST" $FINAL || error check failed!
795 start_ost || return 1
796 start_mds || return 2
797 echo "Requeue thread should have started: "
798 ps -e | grep ll_cfg_requeue
799 set_and_check ost1 "lctl get_param -n obdfilter.$FSNAME-OST0000.client_cache_seconds" "$FSNAME-OST0000.ost.client_cache_seconds" || return 3
802 run_test 27a "Reacquire MGS lock if OST started first"
807 set_and_check mds "lctl get_param -n mds.$FSNAME-MDT0000.group_acquire_expire" "$FSNAME-MDT0000.mdt.group_acquire_expire" || return 3
808 set_and_check client "lctl get_param -n mdc.$FSNAME-MDT0000-mdc-*.max_rpcs_in_flight" "$FSNAME-MDT0000.mdc.max_rpcs_in_flight" || return 4
812 run_test 27b "Reacquire MGS lock after failover"
816 TEST="lctl get_param -n llite.$FSNAME-*.max_read_ahead_whole_mb"
817 PARAM="$FSNAME.llite.max_read_ahead_whole_mb"
820 set_and_check client "$TEST" "$PARAM" $FINAL || return 3
821 FINAL=$(($FINAL + 1))
822 set_and_check client "$TEST" "$PARAM" $FINAL || return 4
823 umount_client $MOUNT || return 200
826 if [ $RESULT -ne $FINAL ]; then
827 echo "New config not seen: wanted $FINAL got $RESULT"
830 echo "New config success: got $RESULT"
832 set_and_check client "$TEST" "$PARAM" $ORIG || return 5
835 run_test 28 "permanent parameter setting"
838 [ "$OSTCOUNT" -lt "2" ] && skip_env "$OSTCOUNT < 2, skipping" && return
839 setup > /dev/null 2>&1
843 local PARAM="$FSNAME-OST0001.osc.active"
844 local PROC_ACT="osc.$FSNAME-OST0001-osc-*.active"
845 local PROC_UUID="osc.$FSNAME-OST0001-osc-*.ost_server_uuid"
847 ACTV=$(lctl get_param -n $PROC_ACT)
849 set_and_check client "lctl get_param -n $PROC_ACT" "$PARAM" $DEAC || return 2
850 # also check ost_server_uuid status
851 RESULT=$(lctl get_param -n $PROC_UUID | grep DEACTIV)
852 if [ -z "$RESULT" ]; then
853 echo "Live client not deactivated: $(lctl get_param -n $PROC_UUID)"
856 echo "Live client success: got $RESULT"
860 local MPROC="osc.$FSNAME-OST0001-osc.active"
865 RESULT=`do_facet mds " lctl get_param -n $MPROC"`
866 [ ${PIPESTATUS[0]} = 0 ] || error "Can't read $MPROC"
867 if [ $RESULT -eq $DEAC ]; then
868 echo "MDT deactivated also after $WAIT sec (got $RESULT)"
872 if [ $WAIT -eq $MAX ]; then
873 echo "MDT not deactivated: wanted $DEAC got $RESULT"
876 echo "Waiting $(($MAX - $WAIT)) secs for MDT deactivated"
879 # test new client starts deactivated
880 umount_client $MOUNT || return 200
882 RESULT=$(lctl get_param -n $PROC_UUID | grep DEACTIV | grep NEW)
883 if [ -z "$RESULT" ]; then
884 echo "New client not deactivated from start: $(lctl get_param -n $PROC_UUID)"
887 echo "New client success: got $RESULT"
890 # make sure it reactivates
891 set_and_check client "lctl get_param -n $PROC_ACT" "$PARAM" $ACTV || return 6
896 #writeconf to remove all ost2 traces for subsequent tests
899 run_test 29 "permanently remove an OST"
904 TEST="lctl get_param -n llite.$FSNAME-*.max_read_ahead_whole_mb"
906 LIST=(1 2 3 4 5 4 3 2 1 2 3 4 5 4 3 2 1 2 3 4 5)
907 for i in ${LIST[@]}; do
908 set_and_check client "$TEST" "$FSNAME.llite.max_read_ahead_whole_mb" $i || return 3
910 # make sure client restart still works
912 mount_client $MOUNT || return 4
913 [ "$($TEST)" -ne "$i" ] && return 5
914 set_and_check client "$TEST" "$FSNAME.llite.max_read_ahead_whole_mb" $ORIG || return 6
917 run_test 30 "Big config llog"
919 test_31() { # bug 10734
920 # ipaddr must not exist
921 mount -t lustre 4.3.2.1@tcp:/lustre $MOUNT || true
924 run_test 31 "Connect to non-existent node (returns errors, should not crash)"
926 # Use these start32/stop32 fn instead of t-f start/stop fn,
927 # for local devices, to skip global facet vars init
931 echo "Stopping local ${MOUNT%/*}/${facet} (opts:$@)"
932 umount -d $@ ${MOUNT%/*}/${facet}
941 mkdir -p ${MOUNT%/*}/${facet}
943 echo "Starting local ${facet}: $@ $device ${MOUNT%/*}/${facet}"
944 mount -t lustre $@ ${device} ${MOUNT%/*}/${facet}
946 if [ $RC -ne 0 ]; then
947 echo "mount -t lustre $@ ${device} ${MOUNT%/*}/${facet}"
948 echo "Start of ${device} of local ${facet} failed ${RC}"
962 echo "Cleanup test_32 umount $MOUNT ..."
963 umount -f $MOUNT || true
964 echo "Cleanup local mds ost1 ..."
970 # this test is totally useless on a client-only system
971 client_only && skip "client only testing" && return 0
972 [ "$NETTYPE" = "tcp" ] || { skip "NETTYPE != tcp" && return 0; }
973 [ -z "$TUNEFS" ] && skip_env "No tunefs" && return 0
975 local DISK1_6=$LUSTRE/tests/disk1_6.tar.bz2
976 [ ! -r $DISK1_6 ] && skip_env "Cant find $DISK1_6, skipping" && return
978 local tmpdir=$TMP/conf32a
981 tar xjvf $DISK1_6 -C $tmpdir ||
982 { skip_env "Cant untar $DISK1_6, skipping" && return ; }
984 lctl set_param debug=$PTLDEBUG
986 $TUNEFS $tmpdir/mds || error "tunefs failed"
988 # nids are wrong, so client wont work, but server should start
989 start32 mds $tmpdir/mds "-o loop,exclude=lustre-OST0000" && \
990 trap cleanup_32 EXIT INT || return 3
992 local UUID=$(lctl get_param -n mds.lustre-MDT0000.uuid)
994 [ "$UUID" == "lustre-MDT0000_UUID" ] || error "UUID is wrong: $UUID"
996 $TUNEFS --mgsnode=`hostname` $tmpdir/ost1 || error "tunefs failed"
997 start32 ost1 $tmpdir/ost1 "-o loop" || return 5
998 UUID=$(lctl get_param -n obdfilter.lustre-OST0000.uuid)
1000 [ "$UUID" == "lustre-OST0000_UUID" ] || error "UUID is wrong: $UUID"
1002 local NID=$($LCTL list_nids | head -1)
1004 echo "OSC changes should succeed:"
1005 $LCTL conf_param lustre-OST0000.osc.max_dirty_mb=15 || return 7
1006 $LCTL conf_param lustre-OST0000.failover.node=$NID || return 8
1008 echo "MDC changes should succeed:"
1009 $LCTL conf_param lustre-MDT0000.mdc.max_rpcs_in_flight=9 || return 9
1010 $LCTL conf_param lustre-MDT0000.failover.node=$NID || return 10
1015 # mount a second time to make sure we didnt leave upgrade flag on
1017 $TUNEFS --dryrun $tmpdir/mds || error "tunefs failed"
1018 start32 mds $tmpdir/mds "-o loop,exclude=lustre-OST0000" && \
1019 trap cleanup_32 EXIT INT || return 12
1023 rm -rf $tmpdir || true # true is only for TMP on NFS
1025 run_test 32a "Upgrade from 1.6 (not live)"
1028 # this test is totally useless on a client-only system
1029 client_only && skip "client only testing" && return 0
1030 [ "$NETTYPE" = "tcp" ] || { skip "NETTYPE != tcp" && return 0; }
1031 [ -z "$TUNEFS" ] && skip_env "No tunefs" && return
1033 local DISK1_6=$LUSTRE/tests/disk1_6.tar.bz2
1034 [ ! -r $DISK1_6 ] && skip_env "Cant find $DISK1_6, skipping" && return
1036 local tmpdir=$TMP/conf32b
1039 tar xjvf $DISK1_6 -C $tmpdir ||
1040 { skip_env "Cant untar $DISK1_6, skipping" && return ; }
1043 lctl set_param debug=$PTLDEBUG
1045 # Ugrade process in according to comment 2 bug 20246
1046 # writeconf will cause servers to register with their current nids
1047 $TUNEFS --mdt --writeconf --erase-param \
1048 --param="mdt.group_upcall=/usr/sbin/l_getgroups" $tmpdir/mds || \
1049 error "tunefs mds failed"
1051 start32 mds $tmpdir/mds "-o loop,abort_recov" && \
1052 trap cleanup_32 EXIT INT || return 3
1055 start32 mds "-o loop" $tmpdir/mds || return 4
1056 local UUID=$(lctl get_param -n mds.lustre-MDT0000.uuid)
1058 [ "$UUID" == "lustre-MDT0000_UUID" ] || error "UUID is wrong: $UUID"
1060 $TUNEFS --ost --writeconf --erase-param --mgsnode=`hostname`@$NETTYPE $tmpdir/ost1 || \
1061 error "tunefs ost failed"
1062 start32 ost1 $tmpdir/ost1 "-o loop,abort_recov" || return 5
1064 UUID=$(lctl get_param -n obdfilter.lustre-OST0000.uuid)
1066 [ "$UUID" == "lustre-OST0000_UUID" ] || error "UUID is wrong: $UUID"
1068 echo "OSC changes should succeed:"
1069 $LCTL conf_param lustre-OST0000.osc.max_dirty_mb=15 || return 7
1070 $LCTL conf_param lustre-OST0000.failover.node=$NID || return 8
1072 echo "MDC changes should succeed:"
1073 $LCTL conf_param lustre-MDT0000.mdc.max_rpcs_in_flight=9 || return 9
1076 # MDT and OST should have registered with new nids, so we should have
1077 # a fully-functioning client
1078 echo "Check client and old fs contents"
1080 local device=`h2$NETTYPE $HOSTNAME`:/lustre
1081 echo "Starting local client: $HOSTNAME: $device $MOUNT"
1082 mount -t lustre $device $MOUNT || return 1
1084 local old=$(lctl get_param -n mdc.*.max_rpcs_in_flight)
1085 local new=$((old + 5))
1086 lctl conf_param lustre-MDT0000.mdc.max_rpcs_in_flight=$new
1087 wait_update $HOSTNAME "lctl get_param -n mdc.*.max_rpcs_in_flight" $new || return 11
1089 [ "$(cksum $MOUNT/passwd | cut -d' ' -f 1,2)" == "2940530074 2837" ] || return 12
1094 rm -rf $tmpdir || true # true is only for TMP on NFS
1096 run_test 32b "Upgrade from 1.6 with writeconf"
1098 test_33a() { # bug 12333, was test_33
1100 local FSNAME2=test-123
1101 [ -n "$ost1_HOST" ] && fs2ost_HOST=$ost1_HOST
1103 if [ -z "$fs2ost_DEV" -o -z "$fs2mds_DEV" ]; then
1104 do_facet mds [ -b "$MDSDEV" ] && \
1105 skip_env "mixed loopback and real device not working" && return
1108 local fs2mdsdev=${fs2mds_DEV:-${MDSDEV}_2}
1109 local fs2ostdev=${fs2ost_DEV:-$(ostdevname 1)_2}
1110 add fs2mds $MDS_MKFS_OPTS --fsname=${FSNAME2} --reformat $fs2mdsdev || exit 10
1111 add fs2ost $OST_MKFS_OPTS --fsname=${FSNAME2} --index=8191 --mgsnode=$MGSNID --reformat $fs2ostdev || exit 10
1113 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS && trap cleanup_24a EXIT INT
1114 start fs2ost $fs2ostdev $OST_MOUNT_OPTS
1115 do_facet mds "$LCTL conf_param $FSNAME2.sys.timeout=200" || rc=1
1117 mount -t lustre $MGSNID:/${FSNAME2} $MOUNT2 || rc=2
1118 cp /etc/hosts $MOUNT2/. || rc=3
1121 cp /etc/hosts $MOUNT2/ || rc=3
1122 $LFS getstripe $MOUNT2/hosts
1127 rm -rf $MOUNT2 $fs2mdsdev $fs2ostdev
1128 cleanup_nocli || rc=6
1131 run_test 33a "Mount ost with a large index number"
1133 test_33b() { # was test_33a
1136 do_facet client dd if=/dev/zero of=$MOUNT/24 bs=1024k count=1
1137 # Drop lock cancelation reply during umount
1138 #define OBD_FAIL_LDLM_CANCEL 0x304
1139 do_facet client lctl set_param fail_loc=0x80000304
1140 #lctl set_param debug=-1
1141 umount_client $MOUNT
1144 run_test 33b "Drop cancel during umount"
1148 do_facet client "sh runmultiop_bg_pause $DIR/file O_c"
1149 manual_umount_client
1151 do_facet client killall -USR1 multiop
1152 if [ $rc -eq 0 ]; then
1153 error "umount not fail!"
1158 run_test 34a "umount with opened file should be fail"
1163 touch $DIR/$tfile || return 1
1164 stop_mds --force || return 2
1166 manual_umount_client --force
1168 if [ $rc -ne 0 ]; then
1169 error "mtab after failed umount - rc $rc"
1175 run_test 34b "force umount with failed mds should be normal"
1179 touch $DIR/$tfile || return 1
1180 stop_ost --force || return 2
1182 manual_umount_client --force
1184 if [ $rc -ne 0 ]; then
1185 error "mtab after failed umount - rc $rc"
1191 run_test 34c "force umount with failed ost should be normal"
1193 test_35a() { # bug 12459
1197 lctl set_param debug="ha"
1199 log "Set up a fake failnode for the MDS"
1201 do_facet mds $LCTL conf_param ${FSNAME}-MDT0000.failover.node=$FAKENID || return 4
1203 log "Wait for RECONNECT_INTERVAL seconds (10s)"
1206 MSG="conf-sanity.sh test_35a `date +%F%kh%Mm%Ss`"
1209 log "Stopping the MDT:"
1210 stop_mds || return 5
1212 df $MOUNT > /dev/null 2>&1 &
1214 log "Restarting the MDT:"
1215 start_mds || return 6
1216 log "Wait for df ($DFPID) ... "
1221 # retrieve from the log the first server that the client tried to
1222 # contact after the connection loss
1223 $LCTL dk $TMP/lustre-log-$TESTNAME.log
1224 NEXTCONN=`awk "/${MSG}/ {start = 1;}
1225 /import_select_connection.*${FSNAME}-MDT0000-mdc.* using connection/ {
1227 if (\\\$NF ~ /$FAKENID/)
1233 }" $TMP/lustre-log-$TESTNAME.log`
1234 [ "$NEXTCONN" != "0" ] && log "The client didn't try to reconnect to the last active server (tried ${NEXTCONN} instead)" && return 7
1237 run_test 35a "Reconnect to the last active server first"
1239 test_35b() { # bug 18674
1240 remote_mds || { skip "local MDS" && return 0; }
1244 $LCTL set_param debug="ha"
1246 MSG="conf-sanity.sh test_35b `date +%F%kh%Mm%Ss`"
1249 log "Set up a fake failnode for the MDS"
1251 do_facet mds $LCTL conf_param ${FSNAME}-MDT0000.failover.node=$FAKENID || \
1254 local at_max_saved=0
1255 # adaptive timeouts may prevent seeing the issue
1256 if at_is_enabled; then
1257 at_max_saved=$(at_max_get mds)
1258 at_max_set 0 mds client
1261 mkdir -p $MOUNT/testdir
1262 touch $MOUNT/testdir/test
1264 log "Injecting EBUSY on MDS"
1265 # Setting OBD_FAIL_MDS_RESEND=0x136
1266 do_facet mds "$LCTL set_param fail_loc=0x80000136" || return 2
1268 log "Stat on a test file"
1269 stat $MOUNT/testdir/test
1271 log "Stop injecting EBUSY on MDS"
1272 do_facet mds "$LCTL set_param fail_loc=0" || return 3
1273 rm -f $MOUNT/testdir/test
1276 # restore adaptive timeout
1277 [ $at_max_saved -ne 0 ] && at_max_set $at_max_saved mds client
1279 $LCTL dk $TMP/lustre-log-$TESTNAME.log
1281 # retrieve from the log if the client has ever tried to
1282 # contact the fake server after the loss of connection
1283 FAILCONN=`awk "BEGIN {ret = 0;}
1284 /import_select_connection.*${FSNAME}-MDT0000-mdc.* using connection/ {
1286 if (\\\$NF ~ /$FAKENID/) {
1291 END {print ret}" $TMP/lustre-log-$TESTNAME.log`
1293 [ "$FAILCONN" == "0" ] && \
1294 log "ERROR: The client reconnection has not been triggered" && \
1296 [ "$FAILCONN" == "2" ] && \
1297 log "ERROR: The client tried to reconnect to the failover server while the primary was busy" && \
1302 run_test 35b "Continue reconnection retries, if the active server is busy"
1306 local FSNAME2=test1234
1307 local fs3ost_HOST=$ost_HOST
1309 [ -n "$ost1_HOST" ] && fs2ost_HOST=$ost1_HOST && fs3ost_HOST=$ost1_HOST
1312 if [ -z "$fs2ost_DEV" -o -z "$fs2mds_DEV" -o -z "$fs3ost_DEV" ]; then
1313 do_facet mds [ -b "$MDSDEV" ] && \
1314 skip_env "mixed loopback and real device not working" && return
1316 [ $OSTCOUNT -lt 2 ] && skip_env "skipping test for single OST" && return
1318 [ "$ost_HOST" = "`hostname`" -o "$ost1_HOST" = "`hostname`" ] || \
1319 { skip "remote OST" && return 0; }
1321 local fs2mdsdev=${fs2mds_DEV:-${MDSDEV}_2}
1322 local fs2ostdev=${fs2ost_DEV:-$(ostdevname 1)_2}
1323 local fs3ostdev=${fs3ost_DEV:-$(ostdevname 2)_2}
1324 add fs2mds $MDS_MKFS_OPTS --fsname=${FSNAME2} --reformat $fs2mdsdev || exit 10
1325 # XXX after we support non 4K disk blocksize, change following --mkfsoptions with
1327 add fs2ost $OST_MKFS_OPTS --mkfsoptions='-b4096' --fsname=${FSNAME2} --mgsnode=$MGSNID --reformat $fs2ostdev || exit 10
1328 add fs3ost $OST_MKFS_OPTS --mkfsoptions='-b4096' --fsname=${FSNAME2} --mgsnode=$MGSNID --reformat $fs3ostdev || exit 10
1330 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS
1331 start fs2ost $fs2ostdev $OST_MOUNT_OPTS
1332 start fs3ost $fs3ostdev $OST_MOUNT_OPTS
1334 mount -t lustre $MGSNID:/${FSNAME2} $MOUNT2 || return 1
1336 sleep 5 # until 11778 fixed
1338 dd if=/dev/zero of=$MOUNT2/$tfile bs=1M count=7 || return 2
1340 BKTOTAL=`lctl get_param -n obdfilter.*.kbytestotal | awk 'BEGIN{total=0}; {total+=$1}; END{print total}'`
1341 BKFREE=`lctl get_param -n obdfilter.*.kbytesfree | awk 'BEGIN{free=0}; {free+=$1}; END{print free}'`
1342 BKAVAIL=`lctl get_param -n obdfilter.*.kbytesavail | awk 'BEGIN{avail=0}; {avail+=$1}; END{print avail}'`
1343 STRING=`df -P $MOUNT2 | tail -n 1 | awk '{print $2","$3","$4}'`
1344 DFTOTAL=`echo $STRING | cut -d, -f1`
1345 DFUSED=`echo $STRING | cut -d, -f2`
1346 DFAVAIL=`echo $STRING | cut -d, -f3`
1347 DFFREE=$(($DFTOTAL - $DFUSED))
1349 ALLOWANCE=$((64 * $OSTCOUNT))
1351 if [ $DFTOTAL -lt $(($BKTOTAL - $ALLOWANCE)) ] ||
1352 [ $DFTOTAL -gt $(($BKTOTAL + $ALLOWANCE)) ] ; then
1353 echo "**** FAIL: df total($DFTOTAL) mismatch OST total($BKTOTAL)"
1356 if [ $DFFREE -lt $(($BKFREE - $ALLOWANCE)) ] ||
1357 [ $DFFREE -gt $(($BKFREE + $ALLOWANCE)) ] ; then
1358 echo "**** FAIL: df free($DFFREE) mismatch OST free($BKFREE)"
1361 if [ $DFAVAIL -lt $(($BKAVAIL - $ALLOWANCE)) ] ||
1362 [ $DFAVAIL -gt $(($BKAVAIL + $ALLOWANCE)) ] ; then
1363 echo "**** FAIL: df avail($DFAVAIL) mismatch OST avail($BKAVAIL)"
1368 stop fs3ost -f || return 200
1369 stop fs2ost -f || return 201
1370 stop fs2mds -f || return 202
1371 rm -rf $MOUNT2 $fs2mdsdev $fs2ostdev $fs3ostdev
1372 unload_modules || return 203
1375 run_test 36 "df report consistency on OSTs with different block size"
1378 client_only && skip "client only testing" && return 0
1379 LOCAL_MDSDEV="$TMP/mdt.img"
1380 SYM_MDSDEV="$TMP/sym_mdt.img"
1382 echo "MDS : $LOCAL_MDSDEV"
1383 echo "SYMLINK : $SYM_MDSDEV"
1387 mkfs.lustre --reformat --fsname=lustre --mdt --mgs --device-size=9000 $LOCAL_MDSDEV ||
1388 error "mkfs.lustre $LOCAL_MDSDEV failed"
1389 ln -s $LOCAL_MDSDEV $SYM_MDSDEV
1391 echo "mount symlink device - $SYM_MDSDEV"
1393 mount_op=`mount -v -t lustre -o loop $SYM_MDSDEV ${MOUNT%/*}/mds 2>&1 | grep "unable to set tunable"`
1394 umount -d ${MOUNT%/*}/mds
1395 rm -f $LOCAL_MDSDEV $SYM_MDSDEV
1397 if [ -n "$mount_op" ]; then
1398 error "**** FAIL: set tunables failed for symlink device"
1402 run_test 37 "verify set tunables works for symlink device"
1404 test_38() { # bug 14222
1409 FILES=`find $SRC -type f -mtime +1 | head -n $COUNT`
1410 log "copying $(echo $FILES | wc -w) files to $DIR/$tdir"
1412 tar cf - $FILES | tar xf - -C $DIR/$tdir || \
1413 error "copying $SRC to $DIR/$tdir"
1415 umount_client $MOUNT
1417 log "rename lov_objid file on MDS"
1418 rm -f $TMP/lov_objid.orig
1419 do_facet mds "$DEBUGFS -c -R \\\"dump lov_objid $TMP/lov_objid.orig\\\" $MDSDEV"
1420 do_facet mds "$DEBUGFS -w -R \\\"rm lov_objid\\\" $MDSDEV"
1422 do_facet mds "od -Ax -td8 $TMP/lov_objid.orig"
1423 # check create in mds_lov_connect
1427 [ $V ] && log "verifying $DIR/$tdir/$f"
1428 diff -q $f $DIR/$tdir/$f || ERROR=y
1430 do_facet mds "$DEBUGFS -c -R \\\"dump lov_objid $TMP/lov_objid.new\\\" $MDSDEV"
1431 do_facet mds "od -Ax -td8 $TMP/lov_objid.new"
1432 [ "$ERROR" = "y" ] && error "old and new files are different after connect" || true
1435 # check it's updates in sync
1436 umount_client $MOUNT
1439 do_facet mds dd if=/dev/zero of=$TMP/lov_objid.clear bs=4096 count=1
1440 do_facet mds "$DEBUGFS -w -R \\\"rm lov_objid\\\" $MDSDEV"
1441 do_facet mds "$DEBUGFS -w -R \\\"write $TMP/lov_objid.clear lov_objid\\\" $MDSDEV "
1446 [ $V ] && log "verifying $DIR/$tdir/$f"
1447 diff -q $f $DIR/$tdir/$f || ERROR=y
1449 do_facet mds "$DEBUGFS -c -R \\\"dump lov_objid $TMP/lov_objid.new1\\\" $MDSDEV"
1450 do_facet mds "od -Ax -td8 $TMP/lov_objid.new1"
1451 umount_client $MOUNT
1453 [ "$ERROR" = "y" ] && error "old and new files are different after sync" || true
1455 log "files compared the same"
1458 run_test 38 "MDS recreates missing lov_objid file from OST data"
1460 test_39() { #bug 14413
1464 perl $SRCDIR/leak_finder.pl $TMP/debug 2>&1 | egrep '*** Leak:' &&
1465 error "memory leak detected" || true
1467 run_test 39 "leak_finder recognizes both LUSTRE and LNET malloc messages"
1469 test_40() { # bug 15759
1471 #define OBD_FAIL_TGT_TOOMANY_THREADS 0x706
1472 do_facet mds "lctl set_param fail_loc=0x80000706"
1476 run_test 40 "race during service thread startup"
1478 test_41() { #bug 14134
1480 start mds $MDSDEV $MDS_MOUNT_OPTS -o nosvc -n
1481 start ost1 `ostdevname 1` $OST_MOUNT_OPTS
1482 start mds $MDSDEV $MDS_MOUNT_OPTS -o nomgs,force
1484 mount_client $MOUNT || return 1
1487 echo "blah blah" > $MOUNT/$tfile
1490 umount_client $MOUNT
1491 stop ost1 -f || return 201
1492 stop mds -f || return 202
1493 stop mds -f || return 203
1494 unload_modules || return 204
1497 run_test 41 "mount mds with --nosvc and --nomgs"
1499 test_42() { #bug 14693
1501 check_mount || return 2
1502 do_facet client lctl conf_param lustre.llite.some_wrong_param=10
1503 umount_client $MOUNT
1504 mount_client $MOUNT || return 1
1508 run_test 42 "invalid config param should not prevent client from mounting"
1510 test_43() { #bug 15993
1512 VERSION_1_8=$(do_facet mds $LCTL get_param version | grep ^lustre.*1\.[78])
1513 if [ -z "$VERSION_1_8" ]; then
1514 skip "skipping test for non 1.8 MDS"
1519 check_mount || return 2
1520 testfile=$DIR/$tfile
1521 lma="this-should-be-removed-after-remount-and-accessed"
1523 echo "set/get trusted.lma"
1524 #define OBD_FAIL_MDS_ALLOW_COMMON_EA_SETTING 0x13f
1525 do_facet mds "lctl set_param fail_loc=0x13f"
1526 lctl set_param fail_loc=0x13f
1527 setfattr -n trusted.lma -v $lma $testfile || error "create common EA"
1528 do_facet mds "lctl set_param fail_loc=0"
1529 lctl set_param fail_loc=0
1530 ATTR=$(getfattr -n trusted.lma $testfile 2> /dev/null | grep trusted.lma)
1531 [ "$ATTR" = "trusted.lma=\"$lma\"" ] || error "check common EA"
1532 umount_client $MOUNT
1537 check_mount || return 3
1538 #define OBD_FAIL_MDS_REMOVE_COMMON_EA 0x13e
1539 do_facet mds "lctl set_param fail_loc=0x13e"
1541 do_facet mds "lctl set_param fail_loc=0"
1542 getfattr -d -m trusted $testfile 2> /dev/null | \
1543 grep "trusted.lma" && error "common EA not removed" || true
1547 run_test 43 "remove common EA if it exists"
1551 check_mount || return 2
1552 UUID=$($LCTL get_param llite.${FSNAME}*.uuid | cut -d= -f2)
1554 UUIDS=$(do_facet mds "$LCTL get_param mds.${FSNAME}*.exports.*.uuid")
1555 for VAL in $UUIDS; do
1556 NID=$(echo $VAL | cut -d= -f1)
1557 CLUUID=$(echo $VAL | cut -d= -f2)
1558 [ "$UUID" = "$CLUUID" ] && STATS_FOUND=yes && break
1560 [ "$STATS_FOUND" = "no" ] && error "stats not found for client"
1564 run_test 44 "mounted client proc entry exists"
1568 check_mount || return 2
1573 #define OBD_FAIL_PTLRPC_LONG_UNLINK 0x50f
1574 do_facet client "lctl set_param fail_loc=0x50f"
1577 manual_umount_client --force || return 3
1578 do_facet client "lctl set_param fail_loc=0x0"
1580 mount_client $MOUNT || return 4
1584 run_test 45 "long unlink handling in ptlrpcd"
1591 umount_client $MOUNT2 || rc=$?
1592 umount_client $MOUNT || rc=$?
1593 while [ $count -gt 0 ]; do
1594 stop ost${count} -f || rc=$?
1598 # writeconf is needed after the test, otherwise,
1599 # we might end up with extra OSTs
1601 cleanup_nocli || rc=$?
1606 echo "Testing with $OSTCOUNT OSTs"
1608 start_mds || return 1
1609 #first client should see only one ost
1610 start_ost || return 2
1611 wait_osc_import_state mds ost FULL
1613 mount_client $MOUNT || return 3
1614 trap "cleanup_46a $OSTCOUNT" EXIT ERR
1617 for (( i=2; i<=$OSTCOUNT; i++ )); do
1618 start ost$i `ostdevname $i` $OST_MOUNT_OPTS || return $((i+2))
1621 # wait until osts in sync
1622 for (( i=2; i<=$OSTCOUNT; i++ )); do
1623 wait_osc_import_state mds ost$i FULL
1626 #second client see all ost's
1628 mount_client $MOUNT2 || return 8
1629 $LFS setstripe $MOUNT2 -c -1 || return 9
1630 $LFS getstripe $MOUNT2 || return 10
1632 echo "ok" > $MOUNT2/widestripe
1633 $LFS getstripe $MOUNT2/widestripe || return 11
1634 # fill acl buffer for avoid expand lsm to them
1635 awk -F : '{if (FNR < 25) { print "u:"$1":rwx" }}' /etc/passwd | while read acl; do
1636 setfacl -m $acl $MOUNT2/widestripe
1640 stat $MOUNT/widestripe || return 12
1642 cleanup_46a $OSTCOUNT || { echo "cleanup_46a failed!" && return 13; }
1645 run_test 46a "handle ost additional - wide striped file"
1650 check_mount || return 2
1651 $LCTL set_param ldlm.namespaces.$FSNAME-*-*-*.lru_size=100
1655 for ns in $($LCTL get_param ldlm.namespaces.$FSNAME-*-*-*.lru_size); do
1656 lrs=$(echo $ns | sed 's/.*lru_size=//')
1657 lru_size[count]=$lrs
1663 client_up || return 3
1666 for ns in $($LCTL get_param ldlm.namespaces.$FSNAME-*-*-*.lru_size); do
1667 lrs=$(echo $ns | sed 's/.*lru_size=//')
1668 if ! test "$lrs" -eq "${lru_size[count]}"; then
1669 n=$(echo $ns | sed -e 's/ldlm.namespaces.//' -e 's/.lru_size=.*//')
1670 error "$n has lost lru_size: $lrs vs. ${lru_size[count]}"
1678 run_test 47 "server restart does not make client loss lru_resize settings"
1683 # reformat after this test is needed - if test will failed
1684 # we will have unkillable file at FS
1688 test_48() { # bug 17636
1691 check_mount || return 2
1693 $LFS setstripe $MOUNT -c -1 || return 9
1694 $LFS getstripe $MOUNT || return 10
1696 echo "ok" > $MOUNT/widestripe
1697 $LFS getstripe $MOUNT/widestripe || return 11
1699 trap cleanup_48 EXIT ERR
1701 # fill acl buffer for avoid expand lsm to them
1702 getent passwd | awk -F : '{ print "u:"$1":rwx" }' | while read acl; do
1703 setfacl -m $acl $MOUNT/widestripe
1706 stat $MOUNT/widestripe || return 12
1711 run_test 48 "too many acls on file"
1713 # check PARAM_SYS_LDLM_TIMEOUT option of MKFS.LUSTRE
1714 test_49() { # bug 17710
1715 local OLD_MDS_MKFS_OPTS=$MDS_MKFS_OPTS
1716 local OLD_OST_MKFS_OPTS=$OST_MKFS_OPTS
1717 local LOCAL_TIMEOUT=20
1719 OST_MKFS_OPTS="--ost --fsname=$FSNAME --device-size=$OSTSIZE --mgsnode=$MGSNID --param sys.timeout=$LOCAL_TIMEOUT --param sys.ldlm_timeout=$LOCAL_TIMEOUT $MKFSOPT $OSTOPT"
1725 check_mount || return 1
1727 echo "check ldlm_timout..."
1728 LDLM_MDS="`do_facet mds lctl get_param -n ldlm_timeout`"
1729 LDLM_OST1="`do_facet ost1 lctl get_param -n ldlm_timeout`"
1730 LDLM_CLIENT="`do_facet client lctl get_param -n ldlm_timeout`"
1732 if [ $LDLM_MDS -ne $LDLM_OST1 ] || [ $LDLM_MDS -ne $LDLM_CLIENT ]; then
1733 error "Different LDLM_TIMEOUT:$LDLM_MDS $LDLM_OST1 $LDLM_CLIENT"
1736 if [ $LDLM_MDS -ne $((LOCAL_TIMEOUT / 3)) ]; then
1737 error "LDLM_TIMEOUT($LDLM_MDS) is not correct"
1740 umount_client $MOUNT
1741 stop_ost || return 2
1742 stop_mds || return 3
1744 OST_MKFS_OPTS="--ost --fsname=$FSNAME --device-size=$OSTSIZE --mgsnode=$MGSNID --param sys.timeout=$LOCAL_TIMEOUT --param sys.ldlm_timeout=$((LOCAL_TIMEOUT - 1)) $MKFSOPT $OSTOPT"
1747 start_mds || return 4
1748 start_ost || return 5
1749 mount_client $MOUNT || return 6
1750 check_mount || return 7
1752 LDLM_MDS="`do_facet mds lctl get_param -n ldlm_timeout`"
1753 LDLM_OST1="`do_facet ost1 lctl get_param -n ldlm_timeout`"
1754 LDLM_CLIENT="`do_facet client lctl get_param -n ldlm_timeout`"
1756 if [ $LDLM_MDS -ne $LDLM_OST1 ] || [ $LDLM_MDS -ne $LDLM_CLIENT ]; then
1757 error "Different LDLM_TIMEOUT:$LDLM_MDS $LDLM_OST1 $LDLM_CLIENT"
1760 if [ $LDLM_MDS -ne $((LOCAL_TIMEOUT - 1)) ]; then
1761 error "LDLM_TIMEOUT($LDLM_MDS) is not correct"
1764 cleanup || return $?
1766 MDS_MKFS_OPTS=$OLD_MDS_MKFS_OPTS
1767 OST_MKFS_OPTS=$OLD_OST_MKFS_OPTS
1769 run_test 49 "check PARAM_SYS_LDLM_TIMEOUT option of MKFS.LUSTRE"
1772 # Test both statfs and lfs df and fail if either one fails
1773 multiop_bg_pause $1 f_
1776 killall -USR1 multiop
1777 [ $RC1 -ne 0 ] && log "lazystatfs multiop failed"
1778 wait $PID || { RC1=$?; log "multiop return error "; }
1785 if [ $RC2 -eq 0 ]; then
1787 log "lazystatfs df failed"
1791 [[ $RC1 -ne 0 || $RC2 -eq 0 ]] && RC=1
1797 lctl set_param llite.$FSNAME-*.lazystatfs=1
1800 lazystatfs $MOUNT || error "lazystatfs failed but no down servers"
1802 cleanup || return $?
1804 run_test 50a "lazystatfs all servers available =========================="
1808 lctl set_param llite.$FSNAME-*.lazystatfs=1
1811 # Wait for client to detect down OST
1812 stop_ost || error "Unable to stop OST1"
1813 wait_osc_import_state mds ost DISCONN
1815 lazystatfs $MOUNT || error "lazystatfs should don't have returned EIO"
1817 umount_client $MOUNT || error "Unable to unmount client"
1818 stop_mds || error "Unable to stop MDS"
1820 run_test 50b "lazystatfs all servers down =========================="
1823 start_mds || error "Unable to start MDS"
1824 start_ost || error "Unable to start OST1"
1825 start_ost2 || error "Unable to start OST2"
1826 mount_client $MOUNT || error "Unable to mount client"
1827 lctl set_param llite.$FSNAME-*.lazystatfs=1
1830 # Wait for client to detect down OST
1831 stop_ost || error "Unable to stop OST1"
1832 wait_osc_import_state mds ost DISCONN
1833 lazystatfs $MOUNT || error "lazystatfs failed with one down server"
1835 umount_client $MOUNT || error "Unable to unmount client"
1836 stop_ost2 || error "Unable to stop OST2"
1837 stop_mds || error "Unable to stop MDS"
1839 run_test 50c "lazystatfs one server down =========================="
1842 start_mds || error "Unable to start MDS"
1843 start_ost || error "Unable to start OST1"
1844 start_ost2 || error "Unable to start OST2"
1845 mount_client $MOUNT || error "Unable to mount client"
1846 lctl set_param llite.$FSNAME-*.lazystatfs=1
1849 # Issue the statfs during the window where the client still
1850 # belives the OST to be available but it is in fact down.
1851 # No failure just a statfs which hangs for a timeout interval.
1852 stop_ost || error "Unable to stop OST1"
1853 lazystatfs $MOUNT || error "lazystatfs failed with one down server"
1855 umount_client $MOUNT || error "Unable to unmount client"
1856 stop_ost2 || error "Unable to stop OST2"
1857 stop_mds || error "Unable to stop MDS"
1859 run_test 50d "lazystatfs client/server conn race =========================="
1866 start_mds || return 1
1867 #first client should see only one ost
1868 start_ost || return 2
1869 wait_osc_import_state mds ost FULL
1871 # Wait for client to detect down OST
1872 stop_ost || error "Unable to stop OST1"
1873 wait_osc_import_state mds ost DISCONN
1875 mount_client $MOUNT || error "Unable to mount client"
1876 lctl set_param llite.$FSNAME-*.lazystatfs=0
1878 multiop_bg_pause $MOUNT _f
1882 if [ $RC1 -ne 0 ]; then
1883 log "lazystatfs multiop failed $RC1"
1886 sleep $(( $TIMEOUT+1 ))
1888 [ $? -ne 0 ] && error "process isn't sleep"
1889 start_ost || error "Unable to start OST1"
1890 wait $pid || error "statfs failed"
1893 umount_client $MOUNT || error "Unable to unmount client"
1894 stop_ost || error "Unable to stop OST1"
1895 stop_mds || error "Unable to stop MDS"
1897 run_test 50e "normal statfs all servers down =========================="
1902 CONN_PROC="osc.$FSNAME-OST0001-osc.ost_server_uuid"
1904 start_mds || error "Unable to start mds"
1905 #first client should see only one ost
1906 start_ost || error "Unable to start OST1"
1907 wait_osc_import_state mds ost FULL
1909 start_ost2 || error "Unable to start OST2"
1910 wait_osc_import_state mds ost2 FULL
1912 # Wait for client to detect down OST
1913 stop_ost2 || error "Unable to stop OST2"
1914 wait_osc_import_state mds ost2 DISCONN
1916 mount_client $MOUNT || error "Unable to mount client"
1917 lctl set_param llite.$FSNAME-*.lazystatfs=0
1919 multiop_bg_pause $MOUNT _f
1923 if [ $RC1 -ne 0 ]; then
1924 log "lazystatfs multiop failed $RC1"
1927 sleep $(( $TIMEOUT+1 ))
1929 [ $? -ne 0 ] && error "process isn't sleep"
1930 start_ost2 || error "Unable to start OST2"
1931 wait $pid || error "statfs failed"
1932 stop_ost2 || error "Unable to stop OST2"
1935 umount_client $MOUNT || error "Unable to unmount client"
1936 stop_ost || error "Unable to stop OST1"
1937 stop_mds || error "Unable to stop MDS"
1940 run_test 50f "normal statfs one server in down =========================="
1943 [ "$OSTCOUNT" -lt "2" ] && skip_env "$OSTCOUNT < 2, skipping" && return
1945 start_ost2 || error "Unable to start OST2"
1947 local PARAM="${FSNAME}-OST0001.osc.active"
1949 $LFS setstripe -c -1 $DIR/$tfile || error "Unable to lfs setstripe"
1950 do_facet mgs $LCTL conf_param $PARAM=0 || error "Unable to deactivate OST"
1952 umount_client $MOUNT || error "Unable to unmount client"
1953 mount_client $MOUNT || error "Unable to mount client"
1954 # This df should not cause a panic
1957 do_facet mgs $LCTL conf_param $PARAM=1 || error "Unable to activate OST"
1959 umount_client $MOUNT || error "Unable to unmount client"
1960 stop_ost2 || error "Unable to stop OST2"
1961 stop_ost || error "Unable to stop OST1"
1962 stop_mds || error "Unable to stop MDS"
1965 run_test 50g "deactivated OST should not cause panic====================="
1967 equals_msg `basename $0`: test complete
1968 [ -f "$TESTSUITELOG" ] && cat $TESTSUITELOG && grep -q FAIL $TESTSUITELOG && exit 1 || true