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="$CONF_SANITY_EXCEPT"
17 # UPDATE THE COMMENT ABOVE WITH BUG NUMBERS WHEN CHANGING ALWAYS_EXCEPT!
20 PATH=$PWD/$SRCDIR:$SRCDIR:$SRCDIR/../utils:$PATH
22 PTLDEBUG=${PTLDEBUG:--1}
24 LUSTRE=${LUSTRE:-`dirname $0`/..}
25 RLUSTRE=${RLUSTRE:-$LUSTRE}
27 . $LUSTRE/tests/test-framework.sh
30 # use small MDS + OST size to speed formatting time
31 # do not use too small MDSSIZE/OSTSIZE, which affect the default jouranl size
34 . ${CONFIG:=$LUSTRE/tests/cfg/$NAME.sh}
36 # STORED_MDSSIZE is used in test_18
37 if [ -n "$MDSSIZE" ]; then
38 STORED_MDSSIZE=$MDSSIZE
43 require_dsh_mds || exit 0
44 require_dsh_ost || exit 0
46 [ "$SLOW" = "no" ] && EXCEPT_SLOW="0 1 2 3 6 7 15 18 24b 25 30 31 32 33 34a 45"
55 local facet=$SINGLEMDS
56 local dev=${facet}_dev
60 # who knows if/where $TUNEFS is installed? Better reformat if it fails...
61 do_facet ${facet} "$TUNEFS --writeconf ${!dev}" ||
62 echo "tunefs failed, reformatting instead" && reformat_and_config
67 # The MGS must be started before the OSTs for a new fs, so start
68 # and stop to generate the startup logs.
71 wait_osc_import_state mds ost FULL
76 reformat_and_config() {
78 if ! combined_mgs_mds ; then
86 start mgs $MGSDEV $mgs_MOUNT_OPTS
90 local facet=$SINGLEMDS
91 # we can not use MDSDEV1 here because SINGLEMDS could be set not to mds1 only
92 local num=$(echo $facet | tr -d "mds")
93 local dev=$(mdsdevname $num)
94 echo "start mds service on `facet_active_host $facet`"
95 start $facet ${dev} $MDS_MOUNT_OPTS || return 94
99 echo "stop mds service on `facet_active_host $SINGLEMDS`"
100 # These tests all use non-failover stop
101 stop $SINGLEMDS -f || return 97
105 echo "start ost1 service on `facet_active_host ost1`"
106 start ost1 `ostdevname 1` $OST_MOUNT_OPTS || return 95
110 echo "stop ost1 service on `facet_active_host ost1`"
111 # These tests all use non-failover stop
112 stop ost1 -f || return 98
116 echo "start ost2 service on `facet_active_host ost2`"
117 start ost2 `ostdevname 2` $OST_MOUNT_OPTS || return 92
121 echo "stop ost2 service on `facet_active_host ost2`"
122 # These tests all use non-failover stop
123 stop ost2 -f || return 93
128 echo "mount $FSNAME on ${MOUNTPATH}....."
129 zconf_mount `hostname` $MOUNTPATH || return 96
133 local SAVEMOUNTOPT=$MOUNTOPT
134 MOUNTOPT="remount,$1"
136 echo "remount '$1' lustre on ${MOUNTPATH}....."
137 zconf_mount `hostname` $MOUNTPATH || return 96
138 MOUNTOPT=$SAVEMOUNTOPT
143 echo "umount lustre on ${MOUNTPATH}....."
144 zconf_umount `hostname` $MOUNTPATH || return 97
147 manual_umount_client(){
150 echo "manual umount lustre on ${MOUNT}...."
151 do_facet client "umount -d ${FORCE} $MOUNT"
157 start_mds || error "MDT start failed"
158 start_ost || error "OST start failed"
159 mount_client $MOUNT || error "client start failed"
163 if ! combined_mgs_mds ; then
173 stop_mds || return 201
174 stop_ost || return 202
175 unload_modules || return 203
179 umount_client $MOUNT || return 200
180 cleanup_nocli || return $?
184 do_facet client "cp /etc/passwd $DIR/a" || return 71
185 do_facet client "rm $DIR/a" || return 72
186 # make sure lustre is actually mounted (touch will block,
187 # but grep won't, so do it after)
188 do_facet client "grep $MOUNT' ' /proc/mounts > /dev/null" || return 73
189 echo "setup single mount lustre success"
193 do_facet client "touch $DIR/a" || return 71
194 do_facet client "rm $DIR/a" || return 72
195 do_facet client "touch $DIR2/a" || return 73
196 do_facet client "rm $DIR2/a" || return 74
197 echo "setup double mount lustre success"
202 if [ "$ONLY" == "setup" ]; then
207 if [ "$ONLY" == "cleanup" ]; then
214 #create single point mountpoint
220 check_mount || return 41
223 run_test 0 "single mount setup"
226 start_mds || error "MDT start failed"
228 echo "start ost second time..."
229 start_ost && error "2nd OST start should fail"
230 mount_client $MOUNT || error "client start failed"
231 check_mount || return 42
234 run_test 1 "start up ost twice (should return errors)"
238 echo "start mds second time.."
239 start_mds && error "2nd MDT start should fail"
242 check_mount || return 43
245 run_test 2 "start up mds twice (should return err)"
249 #mount.lustre returns an error if already in mtab
250 mount_client $MOUNT && error "2nd client mount should fail"
251 check_mount || return 44
254 run_test 3 "mount client twice (should return err)"
258 touch $DIR/$tfile || return 85
262 # ok for ost to fail shutdown
263 if [ 202 -ne $eno ]; then
268 run_test 4 "force cleanup ost, then cleanup"
270 test_5a() { # was test_5
272 touch $DIR/$tfile || return 1
273 fuser -m -v $MOUNT && echo "$MOUNT is in use by user space process."
275 stop_mds -f || return 2
277 # cleanup may return an error from the failed
278 # disconnects; for now I'll consider this successful
279 # if all the modules have unloaded.
283 echo "killing umount"
284 kill -TERM $UMOUNT_PID
285 echo "waiting for umount to finish"
287 if grep " $MOUNT " /proc/mounts; then
288 echo "test 5: /proc/mounts after failed umount"
292 echo "killing umount"
293 kill -TERM $UMOUNT_PID
294 echo "waiting for umount to finish"
296 grep " $MOUNT " /proc/mounts && echo "test 5: /proc/mounts after second umount" && return 11
300 # stop_mds is a no-op here, and should not fail
301 cleanup_nocli || return $?
302 # df may have lingering entry
304 # mtab may have lingering entry
308 while [ "$WAIT" -ne "$MAX_WAIT" ]; do
310 grep -q $MOUNT" " /etc/mtab || break
311 echo "Waiting /etc/mtab updated ... "
312 WAIT=$(( WAIT + sleep))
314 [ "$WAIT" -eq "$MAX_WAIT" ] && error "/etc/mtab is not updated in $WAIT secs"
315 echo "/etc/mtab updated in $WAIT secs"
317 run_test 5a "force cleanup mds, then cleanup"
321 [ -d $MOUNT ] || mkdir -p $MOUNT
322 grep " $MOUNT " /etc/mtab && echo "test 5b: mtab before mount" && return 10
323 mount_client $MOUNT && return 1
324 grep " $MOUNT " /etc/mtab && echo "test 5b: mtab after failed mount" && return 11
326 # stop_mds is a no-op here, and should not fail
327 cleanup_nocli || return $?
330 run_test 5b "mds down, cleanup after failed mount (bug 2712) (should return errs)"
335 [ -d $MOUNT ] || mkdir -p $MOUNT
336 grep " $MOUNT " /etc/mtab && echo "test 5c: mtab before mount" && return 10
337 local oldfs="${FSNAME}"
338 FSNAME="wrong.${FSNAME}"
339 mount_client $MOUNT || :
341 grep " $MOUNT " /etc/mtab && echo "test 5c: mtab after failed mount" && return 11
343 cleanup_nocli || return $?
345 run_test 5c "cleanup after failed mount (bug 2712) (should return errs)"
351 grep " $MOUNT " /etc/mtab && echo "test 5d: mtab before mount" && return 10
352 mount_client $MOUNT || return 1
354 grep " $MOUNT " /etc/mtab && echo "test 5d: mtab after unmount" && return 11
357 run_test 5d "mount with ost down"
363 #define OBD_FAIL_PTLRPC_DELAY_SEND 0x506
364 do_facet client "lctl set_param fail_loc=0x80000506"
365 grep " $MOUNT " /etc/mtab && echo "test 5e: mtab before mount" && return 10
366 mount_client $MOUNT || echo "mount failed (not fatal)"
368 grep " $MOUNT " /etc/mtab && echo "test 5e: mtab after unmount" && return 11
371 run_test 5e "delayed connect, don't crash (bug 10268)"
376 mount_client ${MOUNT} || return 87
377 touch $DIR/a || return 86
380 run_test 6 "manual umount, then mount again"
385 cleanup_nocli || return $?
387 run_test 7 "manual umount, then cleanup"
392 check_mount2 || return 45
393 umount_client $MOUNT2
396 run_test 8 "double mount setup"
401 do_facet ost1 lctl set_param debug=\'inode trace\' || return 1
402 do_facet ost1 lctl set_param subsystem_debug=\'mds ost\' || return 1
404 CHECK_PTLDEBUG="`do_facet ost1 lctl get_param -n debug`"
405 if [ "$CHECK_PTLDEBUG" ] && { \
406 [ "$CHECK_PTLDEBUG" = "trace inode warning error emerg console" ] ||
407 [ "$CHECK_PTLDEBUG" = "trace inode" ]; }; then
408 echo "lnet.debug success"
410 echo "lnet.debug: want 'trace inode', have '$CHECK_PTLDEBUG'"
413 CHECK_SUBSYS="`do_facet ost1 lctl get_param -n subsystem_debug`"
414 if [ "$CHECK_SUBSYS" ] && [ "$CHECK_SUBSYS" = "mds ost" ]; then
415 echo "lnet.subsystem_debug success"
417 echo "lnet.subsystem_debug: want 'mds ost', have '$CHECK_SUBSYS'"
420 stop_ost || return $?
423 run_test 9 "test ptldebug and subsystem for mkfs"
425 # LOGS/PENDING do not exist anymore since CMD3
427 local TMPMTPT="${TMP}/conf16"
428 local dev=${SINGLEMDS}_dev
430 if [ ! -e "$MDSDEV" ]; then
431 log "no $MDSDEV existing, so mount Lustre to create one"
433 check_mount || return 41
437 [ -f "$MDSDEV" ] && LOOPOPT="-o loop"
439 log "change the mode of $MDSDEV/OBJECTS to 555"
440 do_facet $SINGLEMDS "mkdir -p $TMPMTPT &&
441 mount $LOOPOPT -t $FSTYPE $MDSDEV $TMPMTPT &&
442 chmod 555 $TMPMTPT/OBJECTS &&
443 umount $TMPMTPT" || return $?
445 log "mount Lustre to change the mode of OBJECTS, then umount Lustre"
447 check_mount || return 41
450 log "read the mode of OBJECTS and check if they has been changed properly"
451 EXPECTEDOBJECTSMODE=`do_facet $SINGLEMDS "$DEBUGFS -R 'stat OBJECTS' $MDSDEV 2> /dev/null" | grep 'Mode: ' | sed -e "s/.*Mode: *//" -e "s/ *Flags:.*//"`
453 if [ "$EXPECTEDOBJECTSMODE" = "0777" ]; then
454 log "Success:Lustre change the mode of OBJECTS correctly"
456 error "Lustre does not change mode of OBJECTS properly"
459 run_test 16 "verify that lustre will correct the mode of OBJECTS"
462 local dev=${SINGLEMDS}_dev
465 if [ ! -e "$MDSDEV" ]; then
466 echo "no $MDSDEV existing, so mount Lustre to create one"
468 check_mount || return 41
472 echo "Remove mds config log"
473 do_facet $SINGLEMDS "$DEBUGFS -w -R 'unlink CONFIGS/$FSNAME-MDT0000' $MDSDEV || return \$?" || return $?
476 start_mds && return 42
479 run_test 17 "Verify failed mds_postsetup won't fail assertion (2936) (should return errs)"
482 [ "$FSTYPE" != "ldiskfs" ] && skip "not needed for FSTYPE=$FSTYPE" && return
484 local dev=${SINGLEMDS}_dev
490 # check if current MDSSIZE is large enough
491 [ $MDSSIZE -ge $MIN ] && OK=1 && myMDSSIZE=$MDSSIZE && \
492 log "use MDSSIZE=$MDSSIZE"
494 # check if the global config has a large enough MDSSIZE
495 [ -z "$OK" -a ! -z "$STORED_MDSSIZE" ] && [ $STORED_MDSSIZE -ge $MIN ] && \
496 OK=1 && myMDSSIZE=$STORED_MDSSIZE && \
497 log "use STORED_MDSSIZE=$STORED_MDSSIZE"
499 # check if the block device is large enough
500 [ -z "$OK" -a -b $MDSDEV ] && \
501 [ "$(dd if=$MDSDEV of=/dev/null bs=1k count=1 skip=$MIN 2>&1 |
502 awk '($3 == "in") { print $1 }')" = "1+0" ] && OK=1 && \
503 myMDSSIZE=$MIN && log "use device $MDSDEV with MIN=$MIN"
505 # check if a loopback device has enough space for fs metadata (5%)
506 [ -z "$OK" ] && [ -f $MDSDEV -o ! -e $MDSDEV ] &&
507 SPACE=$(df -P $(dirname $MDSDEV) |
508 awk '($1 != "Filesystem") {print $4}') &&
509 [ $SPACE -gt $((MIN / 20)) ] && OK=1 && myMDSSIZE=$MIN && \
510 log "use file $MDSDEV with MIN=$MIN"
512 [ -z "$OK" ] && skip_env "$MDSDEV too small for ${MIN}kB MDS" && return
515 echo "mount mds with large journal..."
516 local OLD_MDS_MKFS_OPTS=$MDS_MKFS_OPTS
518 MDS_MKFS_OPTS="--mgs --mdt --fsname=$FSNAME --device-size=$myMDSSIZE --param sys.timeout=$TIMEOUT $MDSOPT"
521 echo "mount lustre system..."
523 check_mount || return 41
525 echo "check journal size..."
526 local FOUNDSIZE=`do_facet mds "$DEBUGFS -c -R 'stat <8>' $MDSDEV" | awk '/Size: / { print $NF; exit;}'`
527 if [ $FOUNDSIZE -gt $((32 * 1024 * 1024)) ]; then
528 log "Success: mkfs creates large journals. Size: $((FOUNDSIZE >> 20))M"
530 error "expected journal size > 32M, found $((FOUNDSIZE >> 20))M"
535 MDS_MKFS_OPTS=$OLD_MDS_MKFS_OPTS
538 run_test 18 "check mkfs creates large journals"
541 start_mds || return 1
542 stop_mds -f || return 2
544 run_test 19a "start/stop MDS without OSTs"
547 start_ost || return 1
548 stop_ost -f || return 2
550 run_test 19b "start/stop OSTs without MDS"
553 # first format the ost/mdt
557 check_mount || return 43
559 remount_client ro $MOUNT || return 44
560 touch $DIR/$tfile && echo "$DIR/$tfile created incorrectly" && return 45
561 [ -e $DIR/$tfile ] && echo "$DIR/$tfile exists incorrectly" && return 46
562 remount_client rw $MOUNT || return 47
564 [ ! -f $DIR/$tfile ] && echo "$DIR/$tfile missing" && return 48
565 MCNT=`grep -c $MOUNT /etc/mtab`
566 [ "$MCNT" -ne 1 ] && echo "$MOUNT in /etc/mtab $MCNT times" && return 49
571 run_test 20 "remount ro,rw mounts work and doesn't break /etc/mtab"
576 wait_osc_import_state mds ost FULL
580 run_test 21a "start mds before ost, stop ost first"
585 wait_osc_import_state mds ost FULL
589 run_test 21b "start ost before mds, stop mds first"
595 wait_osc_import_state mds ost2 FULL
599 #writeconf to remove all ost2 traces for subsequent tests
602 run_test 21c "start mds between two osts, stop mds last"
607 echo Client mount with ost in logs, but none running
609 # wait until mds connected to ost and open client connection
610 wait_osc_import_state mds ost FULL
613 # check_mount will block trying to contact ost
614 mcreate $DIR/$tfile || return 40
615 rm -f $DIR/$tfile || return 42
619 echo Client mount with a running ost
622 # if gss enabled, wait full time to let connection from
623 # mds to ost be established, due to the mismatch between
624 # initial connect timeout and gss context negotiation timeout.
625 # This perhaps could be remove after AT landed.
626 echo "sleep $((TIMEOUT + TIMEOUT + TIMEOUT))s"
627 sleep $((TIMEOUT + TIMEOUT + TIMEOUT))
630 check_mount || return 41
635 run_test 22 "start a client before osts (should return errs)"
637 test_23a() { # was test_23
641 # force down client so that recovering mds waits for reconnect
642 local running=$(grep -c $MOUNT /proc/mounts) || true
643 if [ $running -ne 0 ]; then
644 echo "Stopping client $MOUNT (opts: -f)"
648 # enter recovery on mds
650 # try to start a new client
651 mount_client $MOUNT &
653 MOUNT_PID=$(ps -ef | grep "t lustre" | grep -v grep | awk '{print $2}')
654 MOUNT_LUSTRE_PID=`ps -ef | grep mount.lustre | grep -v grep | awk '{print $2}'`
655 echo mount pid is ${MOUNT_PID}, mount.lustre pid is ${MOUNT_LUSTRE_PID}
657 ps --ppid $MOUNT_LUSTRE_PID
658 # FIXME why o why can't I kill these? Manual "ctrl-c" works...
659 kill -TERM $MOUNT_LUSTRE_PID
660 echo "waiting for mount to finish"
662 # we can not wait $MOUNT_PID because it is not a child of this shell
668 while [ "$WAIT" -lt "$MAX_WAIT" ]; do
670 PID1=$(ps -ef | awk '{print $2}' | grep -w $MOUNT_PID)
671 PID2=$(ps -ef | awk '{print $2}' | grep -w $MOUNT_LUSTRE_PID)
674 [ -z "$PID1" -a -z "$PID2" ] && break
675 echo "waiting for mount to finish ... "
676 WAIT=$(( WAIT + sleep))
678 if [ "$WAIT" -eq "$MAX_WAIT" ]; then
679 error "MOUNT_PID $MOUNT_PID and "\
680 "MOUNT_LUSTRE_PID $MOUNT_LUSTRE_PID still not killed in $WAIT secs"
686 run_test 23a "interrupt client during recovery mount delay"
691 test_23b() { # was test_23
694 # Simulate -EINTR during mount OBD_FAIL_LDLM_CLOSE_THREAD
695 lctl set_param fail_loc=0x80000313
699 run_test 23b "Simulate -EINTR during mount"
701 fs2mds_HOST=$mds_HOST
702 fs2ost_HOST=$ost_HOST
706 echo "umount $MOUNT2 ..."
707 umount $MOUNT2 || true
708 echo "stopping fs2mds ..."
709 stop fs2mds -f || true
710 echo "stopping fs2ost ..."
711 stop fs2ost -f || true
718 [ -n "$ost1_HOST" ] && fs2ost_HOST=$ost1_HOST
719 if [ -z "$fs2ost_DEV" -o -z "$fs2mds_DEV" ]; then
720 do_facet $SINGLEMDS [ -b "$MDSDEV" ] && \
721 skip_env "mixed loopback and real device not working" && return
724 local fs2mdsdev=${fs2mds_DEV:-${MDSDEV}_2}
725 local fs2ostdev=${fs2ost_DEV:-$(ostdevname 1)_2}
727 # test 8-char fsname as well
728 local FSNAME2=test1234
729 add fs2mds $MDS_MKFS_OPTS --fsname=${FSNAME2} --nomgs --mgsnode=$MGSNID --reformat $fs2mdsdev || exit 10
731 add fs2ost $OST_MKFS_OPTS --fsname=${FSNAME2} --reformat $fs2ostdev || exit 10
734 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS && trap cleanup_24a EXIT INT
735 start fs2ost $fs2ostdev $OST_MOUNT_OPTS
737 mount -t lustre $MGSNID:/${FSNAME2} $MOUNT2 || return 1
739 check_mount || return 2
740 # files written on 1 should not show up on 2
741 cp /etc/passwd $DIR/$tfile
743 [ -e $MOUNT2/$tfile ] && error "File bleed" && return 7
746 cp /etc/passwd $MOUNT2/b || return 3
747 rm $MOUNT2/b || return 4
748 # 2 is actually mounted
749 grep $MOUNT2' ' /proc/mounts > /dev/null || return 5
751 facet_failover fs2mds
752 facet_failover fs2ost
755 # the MDS must remain up until last MDT
757 MDS=$(do_facet $SINGLEMDS "lctl get_param -n devices" | awk '($3 ~ "mdt" && $4 ~ "MDT") { print $4 }' | head -1)
758 [ -z "$MDS" ] && error "No MDT" && return 8
760 cleanup_nocli || return 6
762 run_test 24a "Multiple MDTs on a single node"
765 if [ -z "$fs2mds_DEV" ]; then
766 do_facet $SINGLEMDS [ -b "$MDSDEV" ] && \
767 skip_env "mixed loopback and real device not working" && return
770 local fs2mdsdev=${fs2mds_DEV:-${MDSDEV}_2}
772 add fs2mds $MDS_MKFS_OPTS --fsname=${FSNAME}2 --mgs --reformat $fs2mdsdev || exit 10
774 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS && return 2
777 run_test 24b "Multiple MGSs on a single node (should return err)"
781 check_mount || return 2
782 local MODULES=$($LCTL modules | awk '{ print $2 }')
783 rmmod $MODULES 2>/dev/null || true
786 run_test 25 "Verify modules are referenced"
790 # we need modules before mount for sysctl, so make sure...
791 do_facet $SINGLEMDS "lsmod | grep -q lustre || modprobe lustre"
792 #define OBD_FAIL_MDS_FS_SETUP 0x135
793 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000135"
794 start_mds && echo MDS started && return 1
795 lctl get_param -n devices
796 DEVS=$(lctl get_param -n devices | wc -l)
797 [ $DEVS -gt 0 ] && return 2
798 unload_modules || return 203
800 run_test 26 "MDT startup failure cleans LOV (should return errs)"
806 local ORIG=$(do_facet $myfacet "$TEST")
807 if [ $# -gt 3 ]; then
813 echo "Setting $PARAM from $ORIG to $FINAL"
814 do_facet $SINGLEMDS "$LCTL conf_param $PARAM='$FINAL'" || error conf_param failed
816 wait_update $(facet_host $myfacet) "$TEST" "$FINAL" || error check failed!
820 start_ost || return 1
821 start_mds || return 2
822 echo "Requeue thread should have started: "
823 ps -e | grep ll_cfg_requeue
824 set_and_check ost1 "lctl get_param -n obdfilter.$FSNAME-OST0000.client_cache_seconds" "$FSNAME-OST0000.ost.client_cache_seconds" || return 3
827 run_test 27a "Reacquire MGS lock if OST started first"
832 local device=$(do_facet $SINGLEMDS "lctl get_param -n devices" | awk '($3 ~ "mdt" && $4 ~ "MDT") { print $4 }')
834 facet_failover $SINGLEMDS
835 set_and_check $SINGLEMDS "lctl get_param -n mdt.$device.identity_acquire_expire" "$device.mdt.identity_acquire_expire" || return 3
836 set_and_check client "lctl get_param -n mdc.$device-mdc-*.max_rpcs_in_flight" "$device.mdc.max_rpcs_in_flight" || return 4
840 run_test 27b "Reacquire MGS lock after failover"
844 TEST="lctl get_param -n llite.$FSNAME-*.max_read_ahead_whole_mb"
845 PARAM="$FSNAME.llite.max_read_ahead_whole_mb"
848 set_and_check client "$TEST" "$PARAM" $FINAL || return 3
849 FINAL=$(($FINAL + 1))
850 set_and_check client "$TEST" "$PARAM" $FINAL || return 4
851 umount_client $MOUNT || return 200
854 if [ $RESULT -ne $FINAL ]; then
855 echo "New config not seen: wanted $FINAL got $RESULT"
858 echo "New config success: got $RESULT"
860 set_and_check client "$TEST" "$PARAM" $ORIG || return 5
863 run_test 28 "permanent parameter setting"
866 [ "$OSTCOUNT" -lt "2" ] && skip_env "$OSTCOUNT < 2, skipping" && return
867 setup > /dev/null 2>&1
871 local PARAM="$FSNAME-OST0001.osc.active"
872 local PROC_ACT="osc.$FSNAME-OST0001-osc-[^M]*.active"
873 local PROC_UUID="osc.$FSNAME-OST0001-osc-[^M]*.ost_server_uuid"
875 ACTV=$(lctl get_param -n $PROC_ACT)
877 set_and_check client "lctl get_param -n $PROC_ACT" "$PARAM" $DEAC || return 2
878 # also check ost_server_uuid status
879 RESULT=$(lctl get_param -n $PROC_UUID | grep DEACTIV)
880 if [ -z "$RESULT" ]; then
881 echo "Live client not deactivated: $(lctl get_param -n $PROC_UUID)"
884 echo "Live client success: got $RESULT"
888 local MPROC="osc.$FSNAME-OST0001-osc-[M]*.active"
893 RESULT=`do_facet $SINGLEMDS " lctl get_param -n $MPROC"`
894 [ ${PIPESTATUS[0]} = 0 ] || error "Can't read $MPROC"
895 if [ $RESULT -eq $DEAC ]; then
896 echo "MDT deactivated also after $WAIT sec (got $RESULT)"
900 if [ $WAIT -eq $MAX ]; then
901 echo "MDT not deactivated: wanted $DEAC got $RESULT"
904 echo "Waiting $(($MAX - $WAIT)) secs for MDT deactivated"
907 # test new client starts deactivated
908 umount_client $MOUNT || return 200
910 RESULT=$(lctl get_param -n $PROC_UUID | grep DEACTIV | grep NEW)
911 if [ -z "$RESULT" ]; then
912 echo "New client not deactivated from start: $(lctl get_param -n $PROC_UUID)"
915 echo "New client success: got $RESULT"
918 # make sure it reactivates
919 set_and_check client "lctl get_param -n $PROC_ACT" "$PARAM" $ACTV || return 6
924 #writeconf to remove all ost2 traces for subsequent tests
927 run_test 29 "permanently remove an OST"
933 TEST="lctl get_param -n llite.$FSNAME-*.max_read_ahead_whole_mb"
935 LIST=(1 2 3 4 5 4 3 2 1 2 3 4 5 4 3 2 1 2 3 4 5)
936 for i in ${LIST[@]}; do
937 set_and_check client "$TEST" "$FSNAME.llite.max_read_ahead_whole_mb" $i || return 3
939 # make sure client restart still works
941 mount_client $MOUNT || return 4
942 [ "$($TEST)" -ne "$i" ] && error "Param didn't stick across restart $($TEST) != $i"
945 echo Erase parameter setting
946 do_facet mgs "$LCTL conf_param -d $FSNAME.llite.max_read_ahead_whole_mb" || return 6
948 mount_client $MOUNT || return 6
950 echo "deleted (default) value=$FINAL, orig=$ORIG"
951 # assumes this parameter started at the default value
952 [ "$FINAL" -eq "$ORIG" ] || fail "Deleted value=$FINAL, orig=$ORIG"
956 run_test 30a "Big config llog and conf_param deletion"
962 TEST="lctl get_param -n osc.$FSNAME-OST0000-osc-????????*/import | grep failover_nids | sed -n 's/.*\($NEW\).*/\1/p'"
963 set_and_check client "$TEST" "$FSNAME-OST0000.failover.node" $NEW || error "didn't add failover nid $NEW"
964 do_facet mgs "$LCTL conf_param -d $FSNAME-OST0000.failover.node" || error "conf_param delete failed"
966 mount_client $MOUNT || return 3
968 NIDS=$($LCTL get_param -n osc.$FSNAME-OST0000-osc-????????*/import | grep failover_nids)
969 NIDCOUNT=$(echo "$NIDS" | wc -w)
970 echo "only 1 final nid should remain: $NIDS"
971 [ $NIDCOUNT -eq 2 ] || fail "Failover nids not removed"
976 run_test 30b "Remove failover nids"
978 test_31() { # bug 10734
979 # ipaddr must not exist
980 mount -t lustre 4.3.2.1@tcp:/lustre $MOUNT || true
983 run_test 31 "Connect to non-existent node (shouldn't crash)"
985 # Use these start32/stop32 fn instead of t-f start/stop fn,
986 # for local devices, to skip global facet vars init
990 echo "Stopping local ${MOUNT%/*}/${facet} (opts:$@)"
991 umount -d $@ ${MOUNT%/*}/${facet}
1000 mkdir -p ${MOUNT%/*}/${facet}
1002 echo "Starting local ${facet}: $@ $device ${MOUNT%/*}/${facet}"
1003 mount -t lustre $@ ${device} ${MOUNT%/*}/${facet}
1005 if [ $RC -ne 0 ]; then
1006 echo "mount -t lustre $@ ${device} ${MOUNT%/*}/${facet}"
1007 echo "Start of ${device} of local ${facet} failed ${RC}"
1013 cleanup_nocli32 () {
1021 echo "Cleanup test_32 umount $MOUNT ..."
1022 umount -f $MOUNT || true
1023 echo "Cleanup local mds ost1 ..."
1029 client_only && skip "client only testing" && return 0
1030 [ "$NETTYPE" = "tcp" ] || { skip "NETTYPE != tcp" && return 0; }
1031 [ -z "$TUNEFS" ] && skip_env "No tunefs" && return 0
1033 local DISK1_8=$LUSTRE/tests/disk1_8.tar.bz2
1034 [ ! -r $DISK1_8 ] && skip_env "Cannot find $DISK1_8" && return 0
1035 local tmpdir=$TMP/conf32a
1037 tar xjvf $DISK1_8 -C $tmpdir || \
1038 { skip_env "Cannot untar $DISK1_8" && return 0; }
1041 $LCTL set_param debug=$PTLDEBUG
1043 $TUNEFS $tmpdir/mds || error "tunefs failed"
1045 # nids are wrong, so client wont work, but server should start
1046 start32 mds1 $tmpdir/mds "-o loop,exclude=lustre-OST0000" && \
1047 trap cleanup_32 EXIT INT || return 3
1049 local UUID=$($LCTL get_param -n mdt.lustre-MDT0000.uuid)
1051 [ "$UUID" == "lustre-MDT0000_UUID" ] || error "UUID is wrong: $UUID"
1053 $TUNEFS --mgsnode=$HOSTNAME $tmpdir/ost1 || error "tunefs failed"
1054 start32 ost1 $tmpdir/ost1 "-o loop" || return 5
1055 UUID=$($LCTL get_param -n obdfilter.lustre-OST0000.uuid)
1057 [ "$UUID" == "lustre-OST0000_UUID" ] || error "UUID is wrong: $UUID"
1059 local NID=$($LCTL list_nids | head -1)
1061 echo "OSC changes should succeed:"
1062 $LCTL conf_param lustre-OST0000.osc.max_dirty_mb=15 || return 7
1063 $LCTL conf_param lustre-OST0000.failover.node=$NID || return 8
1066 echo "MDC changes should succeed:"
1067 $LCTL conf_param lustre-MDT0000.mdc.max_rpcs_in_flight=9 || return 9
1068 $LCTL conf_param lustre-MDT0000.failover.node=$NID || return 10
1071 echo "LOV changes should succeed:"
1072 $LCTL pool_new lustre.interop || return 11
1073 $LCTL conf_param lustre-MDT0000.lov.stripesize=4M || return 12
1078 # mount a second time to make sure we didnt leave upgrade flag on
1080 $TUNEFS --dryrun $tmpdir/mds || error "tunefs failed"
1081 start32 mds1 $tmpdir/mds "-o loop,exclude=lustre-OST0000" && \
1082 trap cleanup_32 EXIT INT || return 12
1086 rm -rf $tmpdir || true # true is only for TMP on NFS
1088 run_test 32a "Upgrade from 1.8 (not live)"
1091 client_only && skip "client only testing" && return 0
1092 [ "$NETTYPE" = "tcp" ] || { skip "NETTYPE != tcp" && return 0; }
1093 [ -z "$TUNEFS" ] && skip_env "No tunefs" && return
1095 local DISK1_8=$LUSTRE/tests/disk1_8.tar.bz2
1096 [ ! -r $DISK1_8 ] && skip_env "Cannot find $DISK1_8" && return 0
1097 local tmpdir=$TMP/conf32b
1099 tar xjvf $DISK1_8 -C $tmpdir || \
1100 { skip_env "Cannot untar $DISK1_8" && return ; }
1103 $LCTL set_param debug="config"
1104 local NEWNAME=lustre
1106 # writeconf will cause servers to register with their current nids
1107 $TUNEFS --writeconf --fsname=$NEWNAME $tmpdir/mds || error "tunefs failed"
1108 start32 mds1 $tmpdir/mds "-o loop" && \
1109 trap cleanup_32 EXIT INT || return 3
1111 local UUID=$($LCTL get_param -n mdt.${NEWNAME}-MDT0000.uuid)
1113 [ "$UUID" == "${NEWNAME}-MDT0000_UUID" ] || error "UUID is wrong: $UUID"
1115 $TUNEFS --mgsnode=$HOSTNAME --writeconf --fsname=$NEWNAME $tmpdir/ost1 ||\
1116 error "tunefs failed"
1117 start32 ost1 $tmpdir/ost1 "-o loop" || return 5
1118 UUID=$($LCTL get_param -n obdfilter.${NEWNAME}-OST0000.uuid)
1120 [ "$UUID" == "${NEWNAME}-OST0000_UUID" ] || error "UUID is wrong: $UUID"
1122 local NID=$($LCTL list_nids | head -1)
1124 echo "OSC changes should succeed:"
1125 $LCTL conf_param ${NEWNAME}-OST0000.osc.max_dirty_mb=15 || return 7
1126 $LCTL conf_param ${NEWNAME}-OST0000.failover.node=$NID || return 8
1129 echo "MDC changes should succeed:"
1130 $LCTL conf_param ${NEWNAME}-MDT0000.mdc.max_rpcs_in_flight=9 || return 9
1131 $LCTL conf_param ${NEWNAME}-MDT0000.failover.node=$NID || return 10
1134 echo "LOV changes should succeed:"
1135 $LCTL pool_new ${NEWNAME}.interop || return 11
1136 $LCTL conf_param ${NEWNAME}-MDT0000.lov.stripesize=4M || return 12
1139 # MDT and OST should have registered with new nids, so we should have
1140 # a fully-functioning client
1141 echo "Check client and old fs contents"
1143 local device=`h2$NETTYPE $HOSTNAME`:/$NEWNAME
1144 echo "Starting local client: $HOSTNAME: $device $MOUNT"
1145 mount -t lustre $device $MOUNT || return 1
1147 local old=$($LCTL get_param -n mdc.*.max_rpcs_in_flight)
1148 local new=$((old + 5))
1149 $LCTL conf_param ${NEWNAME}-MDT0000.mdc.max_rpcs_in_flight=$new
1150 wait_update $HOSTNAME "$LCTL get_param -n mdc.*.max_rpcs_in_flight" $new || return 11
1152 [ "$(cksum $MOUNT/passwd | cut -d' ' -f 1,2)" == "94306271 1478" ] || return 12
1157 rm -rf $tmpdir || true # true is only for TMP on NFS
1159 run_test 32b "Upgrade from 1.8 with writeconf"
1161 test_33a() { # bug 12333, was test_33
1163 local FSNAME2=test-123
1164 [ -n "$ost1_HOST" ] && fs2ost_HOST=$ost1_HOST
1166 if [ -z "$fs2ost_DEV" -o -z "$fs2mds_DEV" ]; then
1167 do_facet $SINGLEMDS [ -b "$MDSDEV" ] && \
1168 skip_env "mixed loopback and real device not working" && return
1171 local fs2mdsdev=${fs2mds_DEV:-${MDSDEV}_2}
1172 local fs2ostdev=${fs2ost_DEV:-$(ostdevname 1)_2}
1173 add fs2mds $MDS_MKFS_OPTS --mkfsoptions='\"-J size=8\"' --fsname=${FSNAME2} --reformat $fs2mdsdev || exit 10
1174 add fs2ost $OST_MKFS_OPTS --fsname=${FSNAME2} --index=8191 --mgsnode=$MGSNID --reformat $fs2ostdev || exit 10
1176 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS && trap cleanup_24a EXIT INT
1177 start fs2ost $fs2ostdev $OST_MOUNT_OPTS
1178 do_facet $SINGLEMDS "$LCTL conf_param $FSNAME2.sys.timeout=200" || rc=1
1180 mount -t lustre $MGSNID:/${FSNAME2} $MOUNT2 || rc=2
1183 cp /etc/hosts $MOUNT2/ || rc=3
1184 $LFS getstripe $MOUNT2/hosts
1189 rm -rf $MOUNT2 $fs2mdsdev $fs2ostdev
1190 cleanup_nocli || rc=6
1193 run_test 33a "Mount ost with a large index number"
1195 test_33b() { # was test_34
1198 do_facet client dd if=/dev/zero of=$MOUNT/24 bs=1024k count=1
1199 # Drop lock cancelation reply during umount
1200 #define OBD_FAIL_LDLM_CANCEL 0x304
1201 do_facet client lctl set_param fail_loc=0x80000304
1202 #lctl set_param debug=-1
1203 umount_client $MOUNT
1206 run_test 33b "Drop cancel during umount"
1210 do_facet client "sh runmultiop_bg_pause $DIR/file O_c"
1211 manual_umount_client
1213 do_facet client killall -USR1 multiop
1214 if [ $rc -eq 0 ]; then
1215 error "umount not fail!"
1220 run_test 34a "umount with opened file should be fail"
1225 touch $DIR/$tfile || return 1
1226 stop_mds --force || return 2
1228 manual_umount_client --force
1230 if [ $rc -ne 0 ]; then
1231 error "mtab after failed umount - rc $rc"
1237 run_test 34b "force umount with failed mds should be normal"
1241 touch $DIR/$tfile || return 1
1242 stop_ost --force || return 2
1244 manual_umount_client --force
1246 if [ $rc -ne 0 ]; then
1247 error "mtab after failed umount - rc $rc"
1253 run_test 34c "force umount with failed ost should be normal"
1255 test_35a() { # bug 12459
1258 DBG_SAVE="`lctl get_param -n debug`"
1259 lctl set_param debug="ha"
1261 log "Set up a fake failnode for the MDS"
1263 local device=$(do_facet $SINGLEMDS "lctl get_param -n devices" | awk '($3 ~ "mdt" && $4 ~ "MDT") { print $4 }' | head -1)
1264 do_facet $SINGLEMDS $LCTL conf_param ${device}.failover.node=$FAKENID || return 4
1266 log "Wait for RECONNECT_INTERVAL seconds (10s)"
1269 MSG="conf-sanity.sh test_35a `date +%F%kh%Mm%Ss`"
1272 log "Stopping the MDT:"
1273 stop_mds || return 5
1275 df $MOUNT > /dev/null 2>&1 &
1277 log "Restarting the MDT:"
1278 start_mds || return 6
1279 log "Wait for df ($DFPID) ... "
1282 lctl set_param debug="$DBG_SAVE"
1284 # retrieve from the log the first server that the client tried to
1285 # contact after the connection loss
1286 $LCTL dk $TMP/lustre-log-$TESTNAME.log
1287 NEXTCONN=`awk "/${MSG}/ {start = 1;}
1288 /import_select_connection.*$device-mdc.* using connection/ {
1290 if (\\\$NF ~ /$FAKENID/)
1296 }" $TMP/lustre-log-$TESTNAME.log`
1297 [ "$NEXTCONN" != "0" ] && log "The client didn't try to reconnect to the last active server (tried ${NEXTCONN} instead)" && return 7
1300 run_test 35a "Reconnect to the last active server first"
1302 test_35b() { # bug 18674
1303 remote_mds || { skip "local MDS" && return 0; }
1307 $LCTL set_param debug="ha"
1309 MSG="conf-sanity.sh test_35b `date +%F%kh%Mm%Ss`"
1312 log "Set up a fake failnode for the MDS"
1314 local device=$(do_facet mds "$LCTL get_param -n devices" | \
1315 awk '($3 ~ "mdt" && $4 ~ "MDT") { print $4 }' | head -1)
1316 do_facet mds "$LCTL conf_param ${device}.failover.node=$FAKENID" || \
1319 local at_max_saved=0
1320 # adaptive timeouts may prevent seeing the issue
1321 if at_is_enabled; then
1322 at_max_saved=$(at_max_get mds)
1323 at_max_set 0 mds client
1326 mkdir -p $MOUNT/testdir
1327 touch $MOUNT/testdir/test
1329 log "Injecting EBUSY on MDS"
1330 # Setting OBD_FAIL_MDS_RESEND=0x136
1331 do_facet mds "$LCTL set_param fail_loc=0x80000136" || return 2
1333 log "Stat on a test file"
1334 stat $MOUNT/testdir/test
1336 log "Stop injecting EBUSY on MDS"
1337 do_facet mds "$LCTL set_param fail_loc=0" || return 3
1338 rm -f $MOUNT/testdir/test
1341 # restore adaptive timeout
1342 [ $at_max_saved -ne 0 ] && at_max_set $at_max_saved mds client
1344 $LCTL dk $TMP/lustre-log-$TESTNAME.log
1346 # retrieve from the log if the client has ever tried to
1347 # contact the fake server after the loss of connection
1348 FAILCONN=`awk "BEGIN {ret = 0;}
1349 /import_select_connection.*${FSNAME}-MDT0000-mdc.* using connection/ {
1351 if (\\\$NF ~ /$FAKENID/) {
1356 END {print ret}" $TMP/lustre-log-$TESTNAME.log`
1358 [ "$FAILCONN" == "0" ] && \
1359 log "ERROR: The client reconnection has not been triggered" && \
1361 [ "$FAILCONN" == "2" ] && \
1362 log "ERROR: The client tried to reconnect to the failover server while the primary was busy" && \
1367 run_test 35b "Continue reconnection retries, if the active server is busy"
1371 local FSNAME2=test1234
1372 local fs3ost_HOST=$ost_HOST
1374 [ -n "$ost1_HOST" ] && fs2ost_HOST=$ost1_HOST && fs3ost_HOST=$ost1_HOST
1377 if [ -z "$fs2ost_DEV" -o -z "$fs2mds_DEV" -o -z "$fs3ost_DEV" ]; then
1378 do_facet $SINGLEMDS [ -b "$MDSDEV" ] && \
1379 skip_env "mixed loopback and real device not working" && return
1381 [ $OSTCOUNT -lt 2 ] && skip_env "skipping test for single OST" && return
1383 [ "$ost_HOST" = "`hostname`" -o "$ost1_HOST" = "`hostname`" ] || \
1384 { skip "remote OST" && return 0; }
1386 local fs2mdsdev=${fs2mds_DEV:-${MDSDEV}_2}
1387 local fs2ostdev=${fs2ost_DEV:-$(ostdevname 1)_2}
1388 local fs3ostdev=${fs3ost_DEV:-$(ostdevname 2)_2}
1389 add fs2mds $MDS_MKFS_OPTS --fsname=${FSNAME2} --reformat $fs2mdsdev || exit 10
1390 # XXX after we support non 4K disk blocksize, change following --mkfsoptions with
1392 add fs2ost $OST_MKFS_OPTS --mkfsoptions='-b4096' --fsname=${FSNAME2} --mgsnode=$MGSNID --reformat $fs2ostdev || exit 10
1393 add fs3ost $OST_MKFS_OPTS --mkfsoptions='-b4096' --fsname=${FSNAME2} --mgsnode=$MGSNID --reformat $fs3ostdev || exit 10
1395 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS
1396 start fs2ost $fs2ostdev $OST_MOUNT_OPTS
1397 start fs3ost $fs3ostdev $OST_MOUNT_OPTS
1399 mount -t lustre $MGSNID:/${FSNAME2} $MOUNT2 || return 1
1401 sleep 5 # until 11778 fixed
1403 dd if=/dev/zero of=$MOUNT2/$tfile bs=1M count=7 || return 2
1405 BKTOTAL=`lctl get_param -n obdfilter.*.kbytestotal | awk 'BEGIN{total=0}; {total+=$1}; END{print total}'`
1406 BKFREE=`lctl get_param -n obdfilter.*.kbytesfree | awk 'BEGIN{free=0}; {free+=$1}; END{print free}'`
1407 BKAVAIL=`lctl get_param -n obdfilter.*.kbytesavail | awk 'BEGIN{avail=0}; {avail+=$1}; END{print avail}'`
1408 STRING=`df -P $MOUNT2 | tail -n 1 | awk '{print $2","$3","$4}'`
1409 DFTOTAL=`echo $STRING | cut -d, -f1`
1410 DFUSED=`echo $STRING | cut -d, -f2`
1411 DFAVAIL=`echo $STRING | cut -d, -f3`
1412 DFFREE=$(($DFTOTAL - $DFUSED))
1414 ALLOWANCE=$((64 * $OSTCOUNT))
1416 if [ $DFTOTAL -lt $(($BKTOTAL - $ALLOWANCE)) ] ||
1417 [ $DFTOTAL -gt $(($BKTOTAL + $ALLOWANCE)) ] ; then
1418 echo "**** FAIL: df total($DFTOTAL) mismatch OST total($BKTOTAL)"
1421 if [ $DFFREE -lt $(($BKFREE - $ALLOWANCE)) ] ||
1422 [ $DFFREE -gt $(($BKFREE + $ALLOWANCE)) ] ; then
1423 echo "**** FAIL: df free($DFFREE) mismatch OST free($BKFREE)"
1426 if [ $DFAVAIL -lt $(($BKAVAIL - $ALLOWANCE)) ] ||
1427 [ $DFAVAIL -gt $(($BKAVAIL + $ALLOWANCE)) ] ; then
1428 echo "**** FAIL: df avail($DFAVAIL) mismatch OST avail($BKAVAIL)"
1433 stop fs3ost -f || return 200
1434 stop fs2ost -f || return 201
1435 stop fs2mds -f || return 202
1436 rm -rf $MOUNT2 $fs2mdsdev $fs2ostdev $fs3ostdev
1437 unload_modules || return 203
1440 run_test 36 "df report consistency on OSTs with different block size"
1443 client_only && skip "client only testing" && return 0
1444 LOCAL_MDSDEV="$TMP/mdt.img"
1445 SYM_MDSDEV="$TMP/sym_mdt.img"
1447 echo "MDS : $LOCAL_MDSDEV"
1448 echo "SYMLINK : $SYM_MDSDEV"
1452 mkfs.lustre --reformat --fsname=lustre --mdt --mgs --device-size=9000 $LOCAL_MDSDEV ||
1453 error "mkfs.lustre $LOCAL_MDSDEV failed"
1454 ln -s $LOCAL_MDSDEV $SYM_MDSDEV
1456 echo "mount symlink device - $SYM_MDSDEV"
1458 mount_op=`mount -v -t lustre -o loop $SYM_MDSDEV ${MOUNT%/*}/mds 2>&1 | grep "unable to set tunable"`
1459 umount -d ${MOUNT%/*}/mds
1460 rm -f $LOCAL_MDSDEV $SYM_MDSDEV
1462 if [ -n "$mount_op" ]; then
1463 error "**** FAIL: set tunables failed for symlink device"
1467 run_test 37 "verify set tunables works for symlink device"
1469 test_38() { # bug 14222
1474 FILES=`find $SRC -type f -mtime +1 | head -n $COUNT`
1475 log "copying $(echo $FILES | wc -w) files to $DIR/$tdir"
1477 tar cf - $FILES | tar xf - -C $DIR/$tdir || \
1478 error "copying $SRC to $DIR/$tdir"
1480 umount_client $MOUNT
1482 log "rename lov_objid file on MDS"
1483 rm -f $TMP/lov_objid.orig
1485 local dev=${SINGLEMDS}_dev
1486 local MDSDEV=${!dev}
1487 do_facet $SINGLEMDS "$DEBUGFS -c -R \\\"dump lov_objid $TMP/lov_objid.orig\\\" $MDSDEV"
1488 do_facet $SINGLEMDS "$DEBUGFS -w -R \\\"rm lov_objid\\\" $MDSDEV"
1490 do_facet $SINGLEMDS "od -Ax -td8 $TMP/lov_objid.orig"
1491 # check create in mds_lov_connect
1495 [ $V ] && log "verifying $DIR/$tdir/$f"
1496 diff -q $f $DIR/$tdir/$f || ERROR=y
1498 do_facet $SINGLEMDS "$DEBUGFS -c -R \\\"dump lov_objid $TMP/lov_objid.new\\\" $MDSDEV"
1499 do_facet $SINGLEMDS "od -Ax -td8 $TMP/lov_objid.new"
1500 [ "$ERROR" = "y" ] && error "old and new files are different after connect" || true
1502 # check it's updates in sync
1503 umount_client $MOUNT
1506 do_facet $SINGLEMDS dd if=/dev/zero of=$TMP/lov_objid.clear bs=4096 count=1
1507 do_facet $SINGLEMDS "$DEBUGFS -w -R \\\"rm lov_objid\\\" $MDSDEV"
1508 do_facet $SINGLEMDS "$DEBUGFS -w -R \\\"write $TMP/lov_objid.clear lov_objid\\\" $MDSDEV "
1513 [ $V ] && log "verifying $DIR/$tdir/$f"
1514 diff -q $f $DIR/$tdir/$f || ERROR=y
1516 do_facet $SINGLEMDS "$DEBUGFS -c -R \\\"dump lov_objid $TMP/lov_objid.new1\\\" $MDSDEV"
1517 do_facet $SINGLEMDS "od -Ax -td8 $TMP/lov_objid.new1"
1518 umount_client $MOUNT
1520 [ "$ERROR" = "y" ] && error "old and new files are different after sync" || true
1522 log "files compared the same"
1525 run_test 38 "MDS recreates missing lov_objid file from OST data"
1531 perl $SRCDIR/leak_finder.pl $TMP/debug 2>&1 | egrep '*** Leak:' &&
1532 error "memory leak detected" || true
1534 run_test 39 "leak_finder recognizes both LUSTRE and LNET malloc messages"
1536 test_40() { # bug 15759
1538 #define OBD_FAIL_TGT_TOOMANY_THREADS 0x706
1539 do_facet $SINGLEMDS "sysctl -w lustre.fail_loc=0x80000706"
1543 run_test 40 "race during service thread startup"
1545 test_41() { #bug 14134
1547 local dev=${SINGLEMDS}_dev
1548 local MDSDEV=${!dev}
1550 start $SINGLEMDS $MDSDEV $MDS_MOUNT_OPTS -o nosvc -n
1551 start ost1 `ostdevname 1` $OST_MOUNT_OPTS
1552 start $SINGLEMDS $MDSDEV $MDS_MOUNT_OPTS -o nomgs,force
1554 mount_client $MOUNT || return 1
1557 echo "blah blah" > $MOUNT/$tfile
1560 umount_client $MOUNT
1561 stop ost1 -f || return 201
1562 stop_mds -f || return 202
1563 stop_mds -f || return 203
1564 unload_modules || return 204
1567 run_test 41 "mount mds with --nosvc and --nomgs"
1569 test_42() { #bug 14693
1571 check_mount || return 2
1572 do_facet mgs $LCTL conf_param lustre.llite.some_wrong_param=10
1573 umount_client $MOUNT
1574 mount_client $MOUNT || return 1
1578 run_test 42 "invalid config param should not prevent client from mounting"
1581 [ $UID -ne 0 -o $RUNAS_ID -eq 0 ] && skip_env "run as root"
1583 chmod ugo+x $DIR || error "chmod 0 failed"
1585 "lctl get_param -n mdt.$FSNAME-MDT0000.root_squash" \
1586 "$FSNAME.mdt.root_squash" \
1589 "lctl get_param -n mdt.$FSNAME-MDT0000.nosquash_nids" \
1590 "$FSNAME.mdt.nosquash_nids" \
1594 # create set of test files
1596 echo "111" > $DIR/$tfile-userfile || error "write 1 failed"
1597 chmod go-rw $DIR/$tfile-userfile || error "chmod 1 failed"
1598 chown $RUNAS_ID.$RUNAS_ID $DIR/$tfile-userfile || error "chown failed"
1600 echo "222" > $DIR/$tfile-rootfile || error "write 2 failed"
1601 chmod go-rw $DIR/$tfile-rootfile || error "chmod 2 faield"
1603 mkdir $DIR/$tdir-rootdir -p || error "mkdir failed"
1604 chmod go-rwx $DIR/$tdir-rootdir || error "chmod 3 failed"
1605 touch $DIR/$tdir-rootdir/tfile-1 || error "touch failed"
1608 # check root_squash:
1609 # set root squash UID:GID to RUNAS_ID
1610 # root should be able to access only files owned by RUNAS_ID
1613 "lctl get_param -n mdt.$FSNAME-MDT0000.root_squash" \
1614 "$FSNAME.mdt.root_squash" \
1615 "$RUNAS_ID:$RUNAS_ID"
1617 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-userfile)
1618 dd if=$DIR/$tfile-userfile 1>/dev/null 2>/dev/null || \
1619 error "$ST: root read permission is denied"
1620 echo "$ST: root read permission is granted - ok"
1623 dd conv=notrunc if=$DIR/$tfile-userfile 1>/dev/null 2>/dev/null || \
1624 error "$ST: root write permission is denied"
1625 echo "$ST: root write permission is granted - ok"
1627 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-rootfile)
1628 dd if=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null && \
1629 error "$ST: root read permission is granted"
1630 echo "$ST: root read permission is denied - ok"
1633 dd conv=notrunc of=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null && \
1634 error "$ST: root write permission is granted"
1635 echo "$ST: root write permission is denied - ok"
1637 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tdir-rootdir)
1638 rm $DIR/$tdir-rootdir/tfile-1 1>/dev/null 2>/dev/null && \
1639 error "$ST: root unlink permission is granted"
1640 echo "$ST: root unlink permission is denied - ok"
1642 touch $DIR/tdir-rootdir/tfile-2 1>/dev/null 2>/dev/null && \
1643 error "$ST: root create permission is granted"
1644 echo "$ST: root create permission is denied - ok"
1647 # check nosquash_nids:
1648 # put client's NID into nosquash_nids list,
1649 # root should be able to access root file after that
1651 local NIDLIST=$(lctl list_nids all | tr '\n' ' ')
1652 NIDLIST="2@elan $NIDLIST 192.168.0.[2,10]@tcp"
1653 NIDLIST=$(echo $NIDLIST | tr -s ' ' ' ')
1655 "lctl get_param -n mdt.$FSNAME-MDT0000.nosquash_nids" \
1656 "$FSNAME-MDTall.mdt.nosquash_nids" \
1659 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-rootfile)
1660 dd if=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null || \
1661 error "$ST: root read permission is denied"
1662 echo "$ST: root read permission is granted - ok"
1665 dd conv=notrunc of=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null || \
1666 error "$ST: root write permission is denied"
1667 echo "$ST: root write permission is granted - ok"
1669 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tdir-rootdir)
1670 rm $DIR/$tdir-rootdir/tfile-1 || \
1671 error "$ST: root unlink permission is denied"
1672 echo "$ST: root unlink permission is granted - ok"
1673 touch $DIR/$tdir-rootdir/tfile-2 || \
1674 error "$ST: root create permission is denied"
1675 echo "$ST: root create permission is granted - ok"
1679 run_test 43 "check root_squash and nosquash_nids"
1681 umount_client $MOUNT
1686 check_mount || return 2
1687 UUID=$($LCTL get_param llite.${FSNAME}*.uuid | cut -d= -f2)
1689 UUIDS=$(do_facet mds "$LCTL get_param mdt.${FSNAME}*.exports.*.uuid")
1690 for VAL in $UUIDS; do
1691 NID=$(echo $VAL | cut -d= -f1)
1692 CLUUID=$(echo $VAL | cut -d= -f2)
1693 [ "$UUID" = "$CLUUID" ] && STATS_FOUND=yes && break
1695 [ "$STATS_FOUND" = "no" ] && error "stats not found for client"
1699 run_test 44 "mounted client proc entry exists"
1703 check_mount || return 2
1708 #define OBD_FAIL_PTLRPC_LONG_UNLINK 0x50f
1709 do_facet client "lctl set_param fail_loc=0x50f"
1712 manual_umount_client --force || return 3
1713 do_facet client "lctl set_param fail_loc=0x0"
1715 mount_client $MOUNT || return 4
1719 run_test 45 "long unlink handling in ptlrpcd"
1726 umount_client $MOUNT2 || rc=$?
1727 umount_client $MOUNT || rc=$?
1728 while [ $count -gt 0 ]; do
1729 stop ost${count} -f || rc=$?
1733 # writeconf is needed after the test, otherwise,
1734 # we might end up with extra OSTs
1736 cleanup_nocli || rc=$?
1741 echo "Testing with $OSTCOUNT OSTs"
1743 start_mds || return 1
1744 #first client should see only one ost
1745 start_ost || return 2
1746 wait_osc_import_state mds ost FULL
1748 mount_client $MOUNT || return 3
1749 trap "cleanup_46a $OSTCOUNT" EXIT ERR
1752 for (( i=2; i<=$OSTCOUNT; i++ )); do
1753 start ost$i `ostdevname $i` $OST_MOUNT_OPTS || return $((i+2))
1756 # wait until osts in sync
1757 for (( i=2; i<=$OSTCOUNT; i++ )); do
1758 wait_osc_import_state mds ost$i FULL
1762 #second client see all ost's
1764 mount_client $MOUNT2 || return 8
1765 $LFS setstripe $MOUNT2 -c -1 || return 9
1766 $LFS getstripe $MOUNT2 || return 10
1768 echo "ok" > $MOUNT2/widestripe
1769 $LFS getstripe $MOUNT2/widestripe || return 11
1770 # fill acl buffer for avoid expand lsm to them
1771 awk -F : '{if (FNR < 25) { print "u:"$1":rwx" }}' /etc/passwd | while read acl; do
1772 setfacl -m $acl $MOUNT2/widestripe
1776 stat $MOUNT/widestripe || return 12
1778 cleanup_46a $OSTCOUNT || { echo "cleanup_46a failed!" && return 13; }
1781 run_test 46a "handle ost additional - wide striped file"
1786 check_mount || return 2
1787 $LCTL set_param ldlm.namespaces.$FSNAME-*-*-*.lru_size=100
1791 for ns in $($LCTL get_param ldlm.namespaces.$FSNAME-*-*-*.lru_size); do
1792 if echo $ns | grep "MDT[[:digit:]]*"; then
1795 lrs=$(echo $ns | sed 's/.*lru_size=//')
1796 lru_size[count]=$lrs
1801 facet_failover $SINGLEMDS
1802 client_up || return 3
1805 for ns in $($LCTL get_param ldlm.namespaces.$FSNAME-*-*-*.lru_size); do
1806 if echo $ns | grep "MDT[[:digit:]]*"; then
1809 lrs=$(echo $ns | sed 's/.*lru_size=//')
1810 if ! test "$lrs" -eq "${lru_size[count]}"; then
1811 n=$(echo $ns | sed -e 's/ldlm.namespaces.//' -e 's/.lru_size=.*//')
1812 error "$n has lost lru_size: $lrs vs. ${lru_size[count]}"
1820 run_test 47 "server restart does not make client loss lru_resize settings"
1825 # reformat after this test is needed - if test will failed
1826 # we will have unkillable file at FS
1830 test_48() { # bug 17636
1833 check_mount || return 2
1835 $LFS setstripe $MOUNT -c -1 || return 9
1836 $LFS getstripe $MOUNT || return 10
1838 echo "ok" > $MOUNT/widestripe
1839 $LFS getstripe $MOUNT/widestripe || return 11
1841 trap cleanup_48 EXIT ERR
1843 # fill acl buffer for avoid expand lsm to them
1844 getent passwd | awk -F : '{ print "u:"$1":rwx" }' | while read acl; do
1845 setfacl -m $acl $MOUNT/widestripe
1848 stat $MOUNT/widestripe || return 12
1853 run_test 48 "too many acls on file"
1855 # check PARAM_SYS_LDLM_TIMEOUT option of MKFS.LUSTRE
1856 test_49() { # bug 17710
1857 local OLD_MDS_MKFS_OPTS=$MDS_MKFS_OPTS
1858 local OLD_OST_MKFS_OPTS=$OST_MKFS_OPTS
1859 local LOCAL_TIMEOUT=20
1862 OST_MKFS_OPTS="--ost --fsname=$FSNAME --device-size=$OSTSIZE --mgsnode=$MGSNID --param sys.timeout=$LOCAL_TIMEOUT --param sys.ldlm_timeout=$LOCAL_TIMEOUT $MKFSOPT $OSTOPT"
1866 check_mount || return 1
1868 echo "check ldlm_timout..."
1869 LDLM_MDS="`do_facet mds lctl get_param -n ldlm_timeout`"
1870 LDLM_OST1="`do_facet ost1 lctl get_param -n ldlm_timeout`"
1871 LDLM_CLIENT="`do_facet client lctl get_param -n ldlm_timeout`"
1873 if [ $LDLM_MDS -ne $LDLM_OST1 ] || [ $LDLM_MDS -ne $LDLM_CLIENT ]; then
1874 error "Different LDLM_TIMEOUT:$LDLM_MDS $LDLM_OST1 $LDLM_CLIENT"
1877 if [ $LDLM_MDS -ne $((LOCAL_TIMEOUT / 3)) ]; then
1878 error "LDLM_TIMEOUT($LDLM_MDS) is not correct"
1881 umount_client $MOUNT
1882 stop_ost || return 2
1883 stop_mds || return 3
1885 OST_MKFS_OPTS="--ost --fsname=$FSNAME --device-size=$OSTSIZE --mgsnode=$MGSNID --param sys.timeout=$LOCAL_TIMEOUT --param sys.ldlm_timeout=$((LOCAL_TIMEOUT - 1)) $MKFSOPT $OSTOPT"
1889 check_mount || return 7
1891 LDLM_MDS="`do_facet mds lctl get_param -n ldlm_timeout`"
1892 LDLM_OST1="`do_facet ost1 lctl get_param -n ldlm_timeout`"
1893 LDLM_CLIENT="`do_facet client lctl get_param -n ldlm_timeout`"
1895 if [ $LDLM_MDS -ne $LDLM_OST1 ] || [ $LDLM_MDS -ne $LDLM_CLIENT ]; then
1896 error "Different LDLM_TIMEOUT:$LDLM_MDS $LDLM_OST1 $LDLM_CLIENT"
1899 if [ $LDLM_MDS -ne $((LOCAL_TIMEOUT - 1)) ]; then
1900 error "LDLM_TIMEOUT($LDLM_MDS) is not correct"
1903 cleanup || return $?
1905 MDS_MKFS_OPTS=$OLD_MDS_MKFS_OPTS
1906 OST_MKFS_OPTS=$OLD_OST_MKFS_OPTS
1908 run_test 49 "check PARAM_SYS_LDLM_TIMEOUT option of MKFS.LUSTRE"
1911 # Test both statfs and lfs df and fail if either one fails
1912 multiop_bg_pause $1 f_
1915 killall -USR1 multiop
1916 [ $RC1 -ne 0 ] && log "lazystatfs multiop failed"
1917 wait $PID || { RC1=$?; log "multiop return error "; }
1924 if [ $RC2 -eq 0 ]; then
1926 log "lazystatfs df failed"
1930 [[ $RC1 -ne 0 || $RC2 -eq 0 ]] && RC=1
1936 lctl set_param llite.$FSNAME-*.lazystatfs=1
1939 lazystatfs $MOUNT || error "lazystatfs failed but no down servers"
1941 cleanup || return $?
1943 run_test 50a "lazystatfs all servers available =========================="
1947 lctl set_param llite.$FSNAME-*.lazystatfs=1
1950 # Wait for client to detect down OST
1951 stop_ost || error "Unable to stop OST1"
1952 wait_osc_import_state mds ost DISCONN
1954 lazystatfs $MOUNT || error "lazystatfs should don't have returned EIO"
1956 umount_client $MOUNT || error "Unable to unmount client"
1957 stop_mds || error "Unable to stop MDS"
1959 run_test 50b "lazystatfs all servers down =========================="
1962 start_mds || error "Unable to start MDS"
1963 start_ost || error "Unable to start OST1"
1964 start_ost2 || error "Unable to start OST2"
1965 mount_client $MOUNT || error "Unable to mount client"
1966 lctl set_param llite.$FSNAME-*.lazystatfs=1
1969 # Wait for client to detect down OST
1970 stop_ost || error "Unable to stop OST1"
1971 wait_osc_import_state mds ost DISCONN
1972 lazystatfs $MOUNT || error "lazystatfs failed with one down server"
1974 umount_client $MOUNT || error "Unable to unmount client"
1975 stop_ost2 || error "Unable to stop OST2"
1976 stop_mds || error "Unable to stop MDS"
1978 run_test 50c "lazystatfs one server down =========================="
1981 start_mds || error "Unable to start MDS"
1982 start_ost || error "Unable to start OST1"
1983 start_ost2 || error "Unable to start OST2"
1984 mount_client $MOUNT || error "Unable to mount client"
1985 lctl set_param llite.$FSNAME-*.lazystatfs=1
1988 # Issue the statfs during the window where the client still
1989 # belives the OST to be available but it is in fact down.
1990 # No failure just a statfs which hangs for a timeout interval.
1991 stop_ost || error "Unable to stop OST1"
1992 lazystatfs $MOUNT || error "lazystatfs failed with one down server"
1994 umount_client $MOUNT || error "Unable to unmount client"
1995 stop_ost2 || error "Unable to stop OST2"
1996 stop_mds || error "Unable to stop MDS"
1998 run_test 50d "lazystatfs client/server conn race =========================="
2005 start_mds || return 1
2006 #first client should see only one ost
2007 start_ost || return 2
2008 wait_osc_import_state mds ost FULL
2010 # Wait for client to detect down OST
2011 stop_ost || error "Unable to stop OST1"
2012 wait_osc_import_state mds ost DISCONN
2014 mount_client $MOUNT || error "Unable to mount client"
2015 lctl set_param llite.$FSNAME-*.lazystatfs=0
2017 multiop_bg_pause $MOUNT _f
2021 if [ $RC1 -ne 0 ]; then
2022 log "multiop failed $RC1"
2025 sleep $(( $TIMEOUT+1 ))
2027 [ $? -ne 0 ] && error "process isn't sleep"
2028 start_ost || error "Unable to start OST1"
2029 wait $pid || error "statfs failed"
2032 umount_client $MOUNT || error "Unable to unmount client"
2033 stop_ost || error "Unable to stop OST1"
2034 stop_mds || error "Unable to stop MDS"
2036 run_test 50e "normal statfs all servers down =========================="
2041 CONN_PROC="osc.$FSNAME-OST0001-osc-[M]*.ost_server_uuid"
2043 start_mds || error "Unable to start mds"
2044 #first client should see only one ost
2045 start_ost || error "Unable to start OST1"
2046 wait_osc_import_state mds ost FULL
2048 start_ost2 || error "Unable to start OST2"
2049 wait_osc_import_state mds ost2 FULL
2051 # Wait for client to detect down OST
2052 stop_ost2 || error "Unable to stop OST2"
2054 wait_osc_import_state mds ost2 DISCONN
2055 mount_client $MOUNT || error "Unable to mount client"
2056 lctl set_param llite.$FSNAME-*.lazystatfs=0
2058 multiop_bg_pause $MOUNT _f
2062 if [ $RC1 -ne 0 ]; then
2063 log "lazystatfs multiop failed $RC1"
2066 sleep $(( $TIMEOUT+1 ))
2068 [ $? -ne 0 ] && error "process isn't sleep"
2069 start_ost2 || error "Unable to start OST2"
2070 wait $pid || error "statfs failed"
2071 stop_ost2 || error "Unable to stop OST2"
2074 umount_client $MOUNT || error "Unable to unmount client"
2075 stop_ost || error "Unable to stop OST1"
2076 stop_mds || error "Unable to stop MDS"
2079 run_test 50f "normal statfs one server in down =========================="
2082 [ "$OSTCOUNT" -lt "2" ] && skip_env "$OSTCOUNT < 2, skipping" && return
2084 start_ost2 || error "Unable to start OST2"
2086 local PARAM="${FSNAME}-OST0001.osc.active"
2088 $LFS setstripe -c -1 $DIR/$tfile || error "Unable to lfs setstripe"
2089 do_facet mgs $LCTL conf_param $PARAM=0 || error "Unable to deactivate OST"
2091 umount_client $MOUNT || error "Unable to unmount client"
2092 mount_client $MOUNT || error "Unable to mount client"
2093 # This df should not cause a panic
2096 do_facet mgs $LCTL conf_param $PARAM=1 || error "Unable to activate OST"
2098 umount_client $MOUNT || error "Unable to unmount client"
2099 stop_ost2 || error "Unable to stop OST2"
2100 stop_ost || error "Unable to stop OST1"
2101 stop_mds || error "Unable to stop MDS"
2104 run_test 50g "deactivated OST should not cause panic====================="
2107 local LOCAL_TIMEOUT=20
2111 check_mount || return 1
2114 $LFS setstripe -c -1 $MOUNT/d1
2115 #define OBD_FAIL_MDS_REINT_DELAY 0x142
2116 do_facet $SINGLEMDS "lctl set_param fail_loc=0x142"
2117 touch $MOUNT/d1/f1 &
2120 start_ost2 || return 2
2122 stop_ost2 || return 3
2125 run_test 51 "Verify that mdt_reint handles RMF_MDT_MD correctly when an OST is added"
2134 do_node $node mkdir -p $dest
2135 [ $? -eq 0 ] || { error "Unable to create directory"; return 1; }
2137 do_node $node 'tar cf - '$@' | tar xf - -C '$dest';
2138 [ \"\${PIPESTATUS[*]}\" = \"0 0\" ] || exit 1'
2139 [ $? -eq 0 ] || { error "Unable to tar files"; return 2; }
2141 do_node $node 'getfattr -d -m "[a-z]*\\." '$@' > '$xattrs
2142 [ $? -eq 0 ] || { error "Unable to read xattrs"; return 3; }
2152 local backup2=${TMP}/backup2
2154 do_node $node mkdir -p $backup2
2155 [ $? -eq 0 ] || { error "Unable to create directory"; return 1; }
2157 do_node $node 'tar cf - '$@' | tar xf - -C '$backup2';
2158 [ \"\${PIPESTATUS[*]}\" = \"0 0\" ] || exit 1'
2159 [ $? -eq 0 ] || { error "Unable to tar files to diff"; return 2; }
2161 do_node $node "diff -rq $backup $backup2"
2162 [ $? -eq 0 ] || { error "contents differ"; return 3; }
2164 local xattrs2=${TMP}/xattrs2
2165 do_node $node 'getfattr -d -m "[a-z]*\\." '$@' > '$xattrs2
2166 [ $? -eq 0 ] || { error "Unable to read xattrs to diff"; return 4; }
2168 do_node $node "diff $xattrs $xattrs2"
2169 [ $? -eq 0 ] || { error "xattrs differ"; return 5; }
2171 do_node $node "rm -rf $backup2 $xattrs2"
2172 [ $? -eq 0 ] || { error "Unable to delete temporary files"; return 6; }
2177 [ $? -eq 0 ] || { error "Unable to start MDS"; return 1; }
2179 [ $? -eq 0 ] || { error "Unable to start OST1"; return 2; }
2181 [ $? -eq 0 ] || { error "Unable to mount client"; return 3; }
2184 local ost1mnt=${MOUNT%/*}/ost1
2185 local ost1node=$(facet_active_host ost1)
2186 local ost1tmp=$TMP/conf52
2189 [ $? -eq 0 ] || { error "Unable to create tdir"; return 4; }
2190 touch $TMP/modified_first
2191 [ $? -eq 0 ] || { error "Unable to create temporary file"; return 5; }
2192 do_node $ost1node "mkdir -p $ost1tmp && touch $ost1tmp/modified_first"
2193 [ $? -eq 0 ] || { error "Unable to create temporary file"; return 6; }
2196 $LFS setstripe $DIR/$tdir -c -1 -s 1M
2197 [ $? -eq 0 ] || { error "lfs setstripe failed"; return 7; }
2199 for (( i=0; i < nrfiles; i++ )); do
2200 multiop $DIR/$tdir/$tfile-$i Ow1048576w1048576w524288c
2201 [ $? -eq 0 ] || { error "multiop failed"; return 8; }
2207 echo backup files to $TMP/files
2208 local files=$(find $DIR/$tdir -type f -newer $TMP/modified_first)
2209 copy_files_xattrs `hostname` $TMP/files $TMP/file_xattrs $files
2210 [ $? -eq 0 ] || { error "Unable to copy files"; return 9; }
2212 umount_client $MOUNT
2213 [ $? -eq 0 ] || { error "Unable to umount client"; return 10; }
2215 [ $? -eq 0 ] || { error "Unable to stop ost1"; return 11; }
2217 echo mount ost1 as ldiskfs
2218 do_node $ost1node mount -t $FSTYPE $ost1_dev $ost1mnt $OST_MOUNT_OPTS
2219 [ $? -eq 0 ] || { error "Unable to mount ost1 as ldiskfs"; return 12; }
2222 echo backup objects to $ost1tmp/objects
2223 local objects=$(do_node $ost1node 'find '$ost1mnt'/O/0 -type f -size +0'\
2224 '-newer '$ost1tmp'/modified_first -regex ".*\/[0-9]+"')
2225 copy_files_xattrs $ost1node $ost1tmp/objects $ost1tmp/object_xattrs $objects
2226 [ $? -eq 0 ] || { error "Unable to copy objects"; return 13; }
2228 # move objects to lost+found
2229 do_node $ost1node 'mv '$objects' '${ost1mnt}'/lost+found'
2230 [ $? -eq 0 ] || { error "Unable to move objects"; return 14; }
2233 do_node $ost1node "ll_recover_lost_found_objs -d $ost1mnt/lost+found"
2234 [ $? -eq 0 ] || { error "ll_recover_lost_found_objs failed"; return 15; }
2236 # compare restored objects against saved ones
2237 diff_files_xattrs $ost1node $ost1tmp/objects $ost1tmp/object_xattrs $objects
2238 [ $? -eq 0 ] || { error "Unable to diff objects"; return 16; }
2240 do_node $ost1node "umount $ost1_dev"
2241 [ $? -eq 0 ] || { error "Unable to umount ost1 as ldiskfs"; return 17; }
2244 [ $? -eq 0 ] || { error "Unable to start ost1"; return 18; }
2246 [ $? -eq 0 ] || { error "Unable to mount client"; return 19; }
2249 diff_files_xattrs `hostname` $TMP/files $TMP/file_xattrs $files
2250 [ $? -eq 0 ] || { error "Unable to diff files"; return 20; }
2252 rm -rf $TMP/files $TMP/file_xattrs
2253 [ $? -eq 0 ] || { error "Unable to delete temporary files"; return 21; }
2254 do_node $ost1node "rm -rf $ost1tmp"
2255 [ $? -eq 0 ] || { error "Unable to delete temporary files"; return 22; }
2258 run_test 52 "check recovering objects from lost+found"
2260 # Checks threads_min/max/started for some service
2262 # Arguments: service name (OST or MDT), facet (e.g., ost1, $SINGLEMDS), and a
2263 # parameter pattern prefix like 'ost.*.ost'.
2275 local msg="Insane $modname thread counts"
2279 check_mount || return 41
2281 # We need to expand $parampat, but it may match multiple parameters, so
2282 # we'll pick the first one
2283 if ! paramp=$(do_facet $facet "lctl get_param -N ${parampat}.threads_min"|head -1); then
2284 error "Couldn't expand ${parampat}.threads_min parameter name"
2288 # Remove the .threads_min part
2289 paramp=${paramp%.threads_min}
2291 # Check for sanity in defaults
2292 tmin=$(do_facet $facet "lctl get_param -n ${paramp}.threads_min" || echo 0)
2293 tmax=$(do_facet $facet "lctl get_param -n ${paramp}.threads_max" || echo 0)
2294 tstarted=$(do_facet $facet "lctl get_param -n ${paramp}.threads_started" || echo 0)
2295 lassert 23 "$msg (PDSH problems?)" '(($tstarted && $tmin && $tmax))' || return $?
2296 lassert 24 "$msg" '(($tstarted >= $tmin && $tstarted <= tmax ))' || return $?
2298 # Check that we can lower min/max
2299 do_facet $facet "lctl set_param ${paramp}.threads_min=$((tmin - 1))"
2300 do_facet $facet "lctl set_param ${paramp}.threads_max=$((tmax - 10))"
2301 tmin2=$(do_facet $facet "lctl get_param -n ${paramp}.threads_min" || echo 0)
2302 tmax2=$(do_facet $facet "lctl get_param -n ${paramp}.threads_max" || echo 0)
2303 lassert 25 "$msg" '(($tmin2 == ($tmin - 1) && $tmax2 == ($tmax -10)))' || return $?
2305 # Check that we can set min/max to the same value
2306 do_facet $facet "lctl set_param ${paramp}.threads_min=$tmin"
2307 do_facet $facet "lctl set_param ${paramp}.threads_max=$tmin"
2308 tmin2=$(do_facet $facet "lctl get_param -n ${paramp}.threads_min" || echo 0)
2309 tmax2=$(do_facet $facet "lctl get_param -n ${paramp}.threads_max" || echo 0)
2310 lassert 26 "$msg" '(($tmin2 == $tmin && $tmax2 == $tmin))' || return $?
2312 # Check that we can't set max < min
2313 do_facet $facet "lctl set_param ${paramp}.threads_max=$((tmin - 1))"
2314 tmin2=$(do_facet $facet "lctl get_param -n ${paramp}.threads_min" || echo 0)
2315 tmax2=$(do_facet $facet "lctl get_param -n ${paramp}.threads_max" || echo 0)
2316 lassert 27 "$msg" '(($tmin <= $tmax2))' || return $?
2318 # We need to ensure that we get the module options desired; to do this
2319 # we set LOAD_MODULES_REMOTE=true and we call setmodopts below.
2320 LOAD_MODULES_REMOTE=true
2323 setmodopts -a $modname "$opts" oldvalue
2327 check_mount || return 41
2329 # Restore previous setting of MODOPTS_*
2330 setmodopts $modname "$oldvalue"
2332 # Check that $opts took
2333 tmin=$(do_facet $facet "lctl get_param -n ${paramp}.threads_min")
2334 tmax=$(do_facet $facet "lctl get_param -n ${paramp}.threads_max")
2335 tstarted=$(do_facet $facet "lctl get_param -n ${paramp}.threads_started")
2336 lassert 28 "$msg" '(($tstarted == $tmin && $tstarted == $tmax ))' || return $?
2339 # Workaround a YALA bug where YALA expects that modules will remain
2340 # loaded on the servers
2341 LOAD_MODULES_REMOTE=false
2348 thread_sanity OST ost1 'ost.*.ost' 'oss_num_threads=64'
2350 run_test 53a "check OSS thread count params"
2353 thread_sanity MDT $SINGLEMDS 'mdt.*.*.' 'mdt_num_threads=64'
2355 run_test 53b "check MDT thread count params"
2357 if ! combined_mgs_mds ; then
2364 local partial_arg=""
2365 local size=$(df -B G $dir | tail -1 | awk '{print $2}' | sed 's/G//') # Gb
2367 # Run in partial (fast) mode if the size
2368 # of a partition > 10 GB
2369 [ $size -gt 10 ] && partial_arg="-p"
2371 llverfs $partial_arg $dir
2375 do_rpc_nodes $(facet_host ost1) run_llverdev $(ostdevname 1)
2376 [ $? -eq 0 ] || error "llverdev failed!"
2379 run_test 54a "llverdev"
2384 [ $? -eq 0 ] || error "llverfs failed!"
2387 run_test 54b "llverfs"
2391 local max_ost_index=$1
2392 echo -n $(((max_ost_index + 1) * 8))
2396 local mdsdev=$(mdsdevname 1)
2397 local ostdev=$(ostdevname 1)
2398 local saved_opts=$OST_MKFS_OPTS
2400 for i in 0 1023 2048
2402 OST_MKFS_OPTS="$saved_opts --index $i"
2410 echo checking size of lov_objid for ost index $i
2411 LOV_OBJID_SIZE=$(do_facet mds1 "$DEBUGFS -R 'stat lov_objid' $mdsdev 2>/dev/null" | grep ^User | awk '{print $6}')
2412 if [ "$LOV_OBJID_SIZE" != $(lov_objid_size $i) ]; then
2413 error "lov_objid size has to be $(lov_objid_size $i), not $LOV_OBJID_SIZE"
2415 echo ok, lov_objid size is correct: $LOV_OBJID_SIZE
2420 OST_MKFS_OPTS=$saved_opts
2423 run_test 55 "check lov_objid size"
2426 add mds1 $MDS_MKFS_OPTS --mkfsoptions='\"-J size=16\"' --reformat $(mdsdevname 1)
2427 add ost1 $OST_MKFS_OPTS --index=1000 --reformat $(ostdevname 1)
2428 add ost2 $OST_MKFS_OPTS --index=10000 --reformat $(ostdevname 2)
2432 start_ost2 || error "Unable to start second ost"
2433 mount_client $MOUNT || error "Unable to mount client"
2439 run_test 56 "check big indexes"
2442 equals_msg `basename $0`: test complete
2443 [ -f "$TESTSUITELOG" ] && cat $TESTSUITELOG && grep -q FAIL $TESTSUITELOG && exit 1 || true