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=$MDSSIZE
44 require_dsh_mds || exit 0
45 require_dsh_ost || exit 0
47 [ "$SLOW" = "no" ] && EXCEPT_SLOW="30a 31 45"
62 # who knows if/where $TUNEFS is installed? Better reformat if it fails...
63 do_facet ${facet} "$TUNEFS --quiet --writeconf $dev" ||
64 { echo "tunefs failed, reformatting instead" && reformat_and_config && return 1; }
71 # if writeconf fails anywhere, we reformat everything
72 writeconf1 mds `mdsdevname 1` || return 0
73 writeconf1 ost1 `ostdevname 1` || return 0
74 writeconf1 ost2 `ostdevname 2` || return 0
78 # The MGS must be started before the OSTs for a new fs, so start
79 # and stop to generate the startup logs.
82 wait_osc_import_state mds ost FULL
87 reformat_and_config() {
89 if ! combined_mgs_mds ; then
97 start mgs $MGSDEV $mgs_MOUNT_OPTS
101 local facet=$SINGLEMDS
102 # we can not use MDSDEV1 here because SINGLEMDS could be set not to mds1 only
103 local num=$(echo $facet | tr -d "mds")
104 local dev=$(mdsdevname $num)
105 echo "start mds service on `facet_active_host $facet`"
106 start $facet ${dev} $MDS_MOUNT_OPTS $@ || return 94
110 if ! combined_mgs_mds ; then
117 echo "stop mds service on `facet_active_host $SINGLEMDS`"
118 # These tests all use non-failover stop
119 stop $SINGLEMDS -f || return 97
123 echo "start ost1 service on `facet_active_host ost1`"
124 start ost1 `ostdevname 1` $OST_MOUNT_OPTS $@ || return 95
128 echo "stop ost1 service on `facet_active_host ost1`"
129 # These tests all use non-failover stop
130 stop ost1 -f || return 98
134 echo "start ost2 service on `facet_active_host ost2`"
135 start ost2 `ostdevname 2` $OST_MOUNT_OPTS $@ || return 92
139 echo "stop ost2 service on `facet_active_host ost2`"
140 # These tests all use non-failover stop
141 stop ost2 -f || return 93
146 echo "mount $FSNAME on ${MOUNTPATH}....."
147 zconf_mount `hostname` $MOUNTPATH || return 96
151 local SAVEMOUNTOPT=$MOUNTOPT
152 MOUNTOPT="remount,$1"
154 echo "remount '$1' lustre on ${MOUNTPATH}....."
155 zconf_mount `hostname` $MOUNTPATH || return 96
156 MOUNTOPT=$SAVEMOUNTOPT
161 echo "umount lustre on ${MOUNTPATH}....."
162 zconf_umount `hostname` $MOUNTPATH || return 97
165 manual_umount_client(){
168 echo "manual umount lustre on ${MOUNT}...."
169 do_facet client "umount -d ${FORCE} $MOUNT"
175 start_mds || error "MDT start failed"
176 start_ost || error "OST start failed"
177 mount_client $MOUNT || error "client start failed"
181 if ! combined_mgs_mds ; then
191 stop_ost || return 202
192 stop_mds || return 201
193 unload_modules || return 203
197 umount_client $MOUNT || return 200
198 cleanup_nocli || return $?
202 do_facet client "cp /etc/passwd $DIR/a" || return 71
203 do_facet client "rm $DIR/a" || return 72
204 # make sure lustre is actually mounted (touch will block,
205 # but grep won't, so do it after)
206 do_facet client "grep $MOUNT' ' /proc/mounts > /dev/null" || return 73
207 echo "setup single mount lustre success"
211 do_facet client "touch $DIR/a" || return 71
212 do_facet client "rm $DIR/a" || return 72
213 do_facet client "touch $DIR2/a" || return 73
214 do_facet client "rm $DIR2/a" || return 74
215 echo "setup double mount lustre success"
220 if [ "$ONLY" == "setup" ]; then
225 if [ "$ONLY" == "cleanup" ]; then
232 #create single point mountpoint
238 check_mount || return 41
241 run_test 0 "single mount setup"
244 start_mds || error "MDT start failed"
246 echo "start ost second time..."
247 start_ost && error "2nd OST start should fail"
248 mount_client $MOUNT || error "client start failed"
249 check_mount || return 42
252 run_test 1 "start up ost twice (should return errors)"
256 echo "start mds second time.."
257 start_mds && error "2nd MDT start should fail"
260 check_mount || return 43
263 run_test 2 "start up mds twice (should return err)"
267 #mount.lustre returns an error if already in mtab
268 mount_client $MOUNT && error "2nd client mount should fail"
269 check_mount || return 44
272 run_test 3 "mount client twice (should return err)"
276 touch $DIR/$tfile || return 85
280 # ok for ost to fail shutdown
281 if [ 202 -ne $eno ]; then
286 run_test 4 "force cleanup ost, then cleanup"
288 test_5a() { # was test_5
290 touch $DIR/$tfile || return 1
291 fuser -m -v $MOUNT && echo "$MOUNT is in use by user space process."
293 stop_mds -f || return 2
295 # cleanup may return an error from the failed
296 # disconnects; for now I'll consider this successful
297 # if all the modules have unloaded.
301 echo "killing umount"
302 kill -TERM $UMOUNT_PID
303 echo "waiting for umount to finish"
305 if grep " $MOUNT " /proc/mounts; then
306 echo "test 5: /proc/mounts after failed umount"
310 echo "killing umount"
311 kill -TERM $UMOUNT_PID
312 echo "waiting for umount to finish"
314 grep " $MOUNT " /proc/mounts && echo "test 5: /proc/mounts after second umount" && return 11
318 # stop_mds is a no-op here, and should not fail
319 cleanup_nocli || return $?
320 # df may have lingering entry
322 # mtab may have lingering entry
326 while [ "$WAIT" -ne "$MAX_WAIT" ]; do
328 grep -q $MOUNT" " /etc/mtab || break
329 echo "Waiting /etc/mtab updated ... "
330 WAIT=$(( WAIT + sleep))
332 [ "$WAIT" -eq "$MAX_WAIT" ] && error "/etc/mtab is not updated in $WAIT secs"
333 echo "/etc/mtab updated in $WAIT secs"
335 run_test 5a "force cleanup mds, then cleanup"
339 [ -d $MOUNT ] || mkdir -p $MOUNT
340 grep " $MOUNT " /etc/mtab && echo "test 5b: mtab before mount" && return 10
341 mount_client $MOUNT && return 1
342 grep " $MOUNT " /etc/mtab && echo "test 5b: mtab after failed mount" && return 11
344 # stop_mds is a no-op here, and should not fail
345 cleanup_nocli || return $?
348 run_test 5b "mds down, cleanup after failed mount (bug 2712) (should return errs)"
353 [ -d $MOUNT ] || mkdir -p $MOUNT
354 grep " $MOUNT " /etc/mtab && echo "test 5c: mtab before mount" && return 10
355 local oldfs="${FSNAME}"
356 FSNAME="wrong.${FSNAME}"
357 mount_client $MOUNT || :
359 grep " $MOUNT " /etc/mtab && echo "test 5c: mtab after failed mount" && return 11
361 cleanup_nocli || return $?
363 run_test 5c "cleanup after failed mount (bug 2712) (should return errs)"
369 grep " $MOUNT " /etc/mtab && echo "test 5d: mtab before mount" && return 10
370 mount_client $MOUNT || return 1
372 grep " $MOUNT " /etc/mtab && echo "test 5d: mtab after unmount" && return 11
375 run_test 5d "mount with ost down"
381 #define OBD_FAIL_PTLRPC_DELAY_SEND 0x506
382 do_facet client "lctl set_param fail_loc=0x80000506"
383 grep " $MOUNT " /etc/mtab && echo "test 5e: mtab before mount" && return 10
384 mount_client $MOUNT || echo "mount failed (not fatal)"
386 grep " $MOUNT " /etc/mtab && echo "test 5e: mtab after unmount" && return 11
389 run_test 5e "delayed connect, don't crash (bug 10268)"
394 mount_client ${MOUNT} || return 87
395 touch $DIR/a || return 86
398 run_test 6 "manual umount, then mount again"
403 cleanup_nocli || return $?
405 run_test 7 "manual umount, then cleanup"
410 check_mount2 || return 45
411 umount_client $MOUNT2
414 run_test 8 "double mount setup"
419 do_facet ost1 lctl set_param debug=\'inode trace\' || return 1
420 do_facet ost1 lctl set_param subsystem_debug=\'mds ost\' || return 1
422 CHECK_PTLDEBUG="`do_facet ost1 lctl get_param -n debug`"
423 if [ "$CHECK_PTLDEBUG" ] && { \
424 [ "$CHECK_PTLDEBUG" = "trace inode warning error emerg console" ] ||
425 [ "$CHECK_PTLDEBUG" = "trace inode" ]; }; then
426 echo "lnet.debug success"
428 echo "lnet.debug: want 'trace inode', have '$CHECK_PTLDEBUG'"
431 CHECK_SUBSYS="`do_facet ost1 lctl get_param -n subsystem_debug`"
432 if [ "$CHECK_SUBSYS" ] && [ "$CHECK_SUBSYS" = "mds ost" ]; then
433 echo "lnet.subsystem_debug success"
435 echo "lnet.subsystem_debug: want 'mds ost', have '$CHECK_SUBSYS'"
438 stop_ost || return $?
440 run_test 9 "test ptldebug and subsystem for mkfs"
443 # Test 16 was to "verify that lustre will correct the mode of OBJECTS".
444 # But with new MDS stack we don't care about the mode of local objects
445 # anymore, so this test is removed. See bug 22944 for more details.
449 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
451 if [ ! -e "$MDSDEV" ]; then
452 echo "no $MDSDEV existing, so mount Lustre to create one"
454 check_mount || return 41
458 echo "Remove mds config log"
459 do_facet $SINGLEMDS "$DEBUGFS -w -R 'unlink CONFIGS/$FSNAME-MDT0000' $MDSDEV || return \$?" || return $?
462 start_mds && return 42
465 run_test 17 "Verify failed mds_postsetup won't fail assertion (2936) (should return errs)"
468 [ "$FSTYPE" != "ldiskfs" ] && skip "not needed for FSTYPE=$FSTYPE" && return
470 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
475 # check if current MDSSIZE is large enough
476 [ $MDSSIZE -ge $MIN ] && OK=1 && myMDSSIZE=$MDSSIZE && \
477 log "use MDSSIZE=$MDSSIZE"
479 # check if the global config has a large enough MDSSIZE
480 [ -z "$OK" -a ! -z "$STORED_MDSSIZE" ] && [ $STORED_MDSSIZE -ge $MIN ] && \
481 OK=1 && myMDSSIZE=$STORED_MDSSIZE && \
482 log "use STORED_MDSSIZE=$STORED_MDSSIZE"
484 # check if the block device is large enough
485 [ -z "$OK" -a -b $MDSDEV ] && \
486 [ "$(dd if=$MDSDEV of=/dev/null bs=1k count=1 skip=$MIN 2>&1 |
487 awk '($3 == "in") { print $1 }')" = "1+0" ] && OK=1 && \
488 myMDSSIZE=$MIN && log "use device $MDSDEV with MIN=$MIN"
490 # check if a loopback device has enough space for fs metadata (5%)
491 [ -z "$OK" ] && [ -f $MDSDEV -o ! -e $MDSDEV ] &&
492 SPACE=$(df -P $(dirname $MDSDEV) |
493 awk '($1 != "Filesystem") {print $4}') &&
494 [ $SPACE -gt $((MIN / 20)) ] && OK=1 && myMDSSIZE=$MIN && \
495 log "use file $MDSDEV with MIN=$MIN"
497 [ -z "$OK" ] && skip_env "$MDSDEV too small for ${MIN}kB MDS" && return
500 echo "mount mds with large journal..."
501 local OLD_MDS_MKFS_OPTS=$MDS_MKFS_OPTS
503 MDS_MKFS_OPTS="--mgs --mdt --fsname=$FSNAME --device-size=$myMDSSIZE --param sys.timeout=$TIMEOUT $MDSOPT"
506 echo "mount lustre system..."
508 check_mount || return 41
510 echo "check journal size..."
511 local FOUNDSIZE=`do_facet mds "$DEBUGFS -c -R 'stat <8>' $MDSDEV" | awk '/Size: / { print $NF; exit;}'`
512 if [ $FOUNDSIZE -gt $((32 * 1024 * 1024)) ]; then
513 log "Success: mkfs creates large journals. Size: $((FOUNDSIZE >> 20))M"
515 error "expected journal size > 32M, found $((FOUNDSIZE >> 20))M"
520 MDS_MKFS_OPTS=$OLD_MDS_MKFS_OPTS
523 run_test 18 "check mkfs creates large journals"
526 start_mds || return 1
527 stop_mds -f || return 2
529 run_test 19a "start/stop MDS without OSTs"
532 start_ost || return 1
533 stop_ost -f || return 2
535 run_test 19b "start/stop OSTs without MDS"
538 # first format the ost/mdt
542 check_mount || return 43
544 remount_client ro $MOUNT || return 44
545 touch $DIR/$tfile && echo "$DIR/$tfile created incorrectly" && return 45
546 [ -e $DIR/$tfile ] && echo "$DIR/$tfile exists incorrectly" && return 46
547 remount_client rw $MOUNT || return 47
549 [ ! -f $DIR/$tfile ] && echo "$DIR/$tfile missing" && return 48
550 MCNT=`grep -c $MOUNT /etc/mtab`
551 [ "$MCNT" -ne 1 ] && echo "$MOUNT in /etc/mtab $MCNT times" && return 49
556 run_test 20 "remount ro,rw mounts work and doesn't break /etc/mtab"
561 wait_osc_import_state mds ost FULL
565 run_test 21a "start mds before ost, stop ost first"
570 wait_osc_import_state mds ost FULL
574 run_test 21b "start ost before mds, stop mds first"
580 wait_osc_import_state mds ost2 FULL
584 #writeconf to remove all ost2 traces for subsequent tests
587 run_test 21c "start mds between two osts, stop mds last"
592 echo Client mount with ost in logs, but none running
594 # wait until mds connected to ost and open client connection
595 wait_osc_import_state mds ost FULL
598 # check_mount will block trying to contact ost
599 mcreate $DIR/$tfile || return 40
600 rm -f $DIR/$tfile || return 42
604 echo Client mount with a running ost
607 # if gss enabled, wait full time to let connection from
608 # mds to ost be established, due to the mismatch between
609 # initial connect timeout and gss context negotiation timeout.
610 # This perhaps could be remove after AT landed.
611 echo "sleep $((TIMEOUT + TIMEOUT + TIMEOUT))s"
612 sleep $((TIMEOUT + TIMEOUT + TIMEOUT))
615 check_mount || return 41
620 run_test 22 "start a client before osts (should return errs)"
622 test_23a() { # was test_23
626 # force down client so that recovering mds waits for reconnect
627 local running=$(grep -c $MOUNT /proc/mounts) || true
628 if [ $running -ne 0 ]; then
629 echo "Stopping client $MOUNT (opts: -f)"
633 # enter recovery on mds
635 # try to start a new client
636 mount_client $MOUNT &
638 MOUNT_PID=$(ps -ef | grep "t lustre" | grep -v grep | awk '{print $2}')
639 MOUNT_LUSTRE_PID=`ps -ef | grep mount.lustre | grep -v grep | awk '{print $2}'`
640 echo mount pid is ${MOUNT_PID}, mount.lustre pid is ${MOUNT_LUSTRE_PID}
642 ps --ppid $MOUNT_LUSTRE_PID
643 # FIXME why o why can't I kill these? Manual "ctrl-c" works...
644 kill -TERM $MOUNT_LUSTRE_PID
645 echo "waiting for mount to finish"
647 # we can not wait $MOUNT_PID because it is not a child of this shell
653 while [ "$WAIT" -lt "$MAX_WAIT" ]; do
655 PID1=$(ps -ef | awk '{print $2}' | grep -w $MOUNT_PID)
656 PID2=$(ps -ef | awk '{print $2}' | grep -w $MOUNT_LUSTRE_PID)
659 [ -z "$PID1" -a -z "$PID2" ] && break
660 echo "waiting for mount to finish ... "
661 WAIT=$(( WAIT + sleep))
663 if [ "$WAIT" -eq "$MAX_WAIT" ]; then
664 error "MOUNT_PID $MOUNT_PID and "\
665 "MOUNT_LUSTRE_PID $MOUNT_LUSTRE_PID still not killed in $WAIT secs"
671 run_test 23a "interrupt client during recovery mount delay"
676 test_23b() { # was test_23
679 # Simulate -EINTR during mount OBD_FAIL_LDLM_CLOSE_THREAD
680 lctl set_param fail_loc=0x80000313
684 run_test 23b "Simulate -EINTR during mount"
686 fs2mds_HOST=$mds_HOST
687 fs2ost_HOST=$ost_HOST
691 echo "umount $MOUNT2 ..."
692 umount $MOUNT2 || true
693 echo "stopping fs2mds ..."
694 stop fs2mds -f || true
695 echo "stopping fs2ost ..."
696 stop fs2ost -f || true
704 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
706 [ -n "$ost1_HOST" ] && fs2ost_HOST=$ost1_HOST
707 if [ -z "$fs2ost_DEV" -o -z "$fs2mds_DEV" ]; then
708 do_facet $SINGLEMDS [ -b "$MDSDEV" ] && \
709 skip_env "mixed loopback and real device not working" && return
712 local fs2mdsdev=${fs2mds_DEV:-${MDSDEV}_2}
713 local fs2ostdev=${fs2ost_DEV:-$(ostdevname 1)_2}
715 # test 8-char fsname as well
716 local FSNAME2=test1234
717 add fs2mds $MDS_MKFS_OPTS --fsname=${FSNAME2} --nomgs --mgsnode=$MGSNID --reformat $fs2mdsdev || exit 10
719 add fs2ost $OST_MKFS_OPTS --fsname=${FSNAME2} --reformat $fs2ostdev || exit 10
722 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS && trap cleanup_24a EXIT INT
723 start fs2ost $fs2ostdev $OST_MOUNT_OPTS
725 mount -t lustre $MGSNID:/${FSNAME2} $MOUNT2 || return 1
727 check_mount || return 2
728 # files written on 1 should not show up on 2
729 cp /etc/passwd $DIR/$tfile
731 [ -e $MOUNT2/$tfile ] && error "File bleed" && return 7
734 cp /etc/passwd $MOUNT2/b || return 3
735 rm $MOUNT2/b || return 4
736 # 2 is actually mounted
737 grep $MOUNT2' ' /proc/mounts > /dev/null || return 5
739 facet_failover fs2mds
740 facet_failover fs2ost
743 # the MDS must remain up until last MDT
745 MDS=$(do_facet $SINGLEMDS "lctl get_param -n devices" | awk '($3 ~ "mdt" && $4 ~ "MDT") { print $4 }' | head -1)
746 [ -z "$MDS" ] && error "No MDT" && return 8
748 cleanup_nocli || return 6
750 run_test 24a "Multiple MDTs on a single node"
753 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
755 if [ -z "$fs2mds_DEV" ]; then
756 do_facet $SINGLEMDS [ -b "$MDSDEV" ] && \
757 skip_env "mixed loopback and real device not working" && return
760 local fs2mdsdev=${fs2mds_DEV:-${MDSDEV}_2}
762 add fs2mds $MDS_MKFS_OPTS --fsname=${FSNAME}2 --mgs --reformat $fs2mdsdev || exit 10
764 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS && return 2
767 run_test 24b "Multiple MGSs on a single node (should return err)"
771 check_mount || return 2
772 local MODULES=$($LCTL modules | awk '{ print $2 }')
773 rmmod $MODULES 2>/dev/null || true
776 run_test 25 "Verify modules are referenced"
780 # we need modules before mount for sysctl, so make sure...
781 do_facet $SINGLEMDS "lsmod | grep -q lustre || modprobe lustre"
782 #define OBD_FAIL_MDS_FS_SETUP 0x135
783 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000135"
784 start_mds && echo MDS started && return 1
785 lctl get_param -n devices
786 DEVS=$(lctl get_param -n devices | wc -l)
787 [ $DEVS -gt 0 ] && return 2
788 unload_modules || return 203
790 run_test 26 "MDT startup failure cleans LOV (should return errs)"
796 local ORIG=$(do_facet $myfacet "$TEST")
797 if [ $# -gt 3 ]; then
803 echo "Setting $PARAM from $ORIG to $FINAL"
804 do_facet $SINGLEMDS "$LCTL conf_param $PARAM='$FINAL'" || error conf_param failed
806 wait_update $(facet_host $myfacet) "$TEST" "$FINAL" || error check failed!
810 start_ost || return 1
811 start_mds || return 2
812 echo "Requeue thread should have started: "
813 ps -e | grep ll_cfg_requeue
814 set_and_check ost1 "lctl get_param -n obdfilter.$FSNAME-OST0000.client_cache_seconds" "$FSNAME-OST0000.ost.client_cache_seconds" || return 3
817 run_test 27a "Reacquire MGS lock if OST started first"
822 local device=$(do_facet $SINGLEMDS "lctl get_param -n devices" | awk '($3 ~ "mdt" && $4 ~ "MDT") { print $4 }')
824 facet_failover $SINGLEMDS
825 set_and_check $SINGLEMDS "lctl get_param -n mdt.$device.identity_acquire_expire" "$device.mdt.identity_acquire_expire" || return 3
826 set_and_check client "lctl get_param -n mdc.$device-mdc-*.max_rpcs_in_flight" "$device.mdc.max_rpcs_in_flight" || return 4
830 run_test 27b "Reacquire MGS lock after failover"
834 TEST="lctl get_param -n llite.$FSNAME-*.max_read_ahead_whole_mb"
835 PARAM="llite.$FSNAME.max_read_ahead_whole_mb"
838 set_and_check client "$TEST" "$PARAM" $FINAL || return 3
839 FINAL=$(($FINAL + 1))
840 set_and_check client "$TEST" "$PARAM" $FINAL || return 4
841 umount_client $MOUNT || return 200
844 if [ $RESULT -ne $FINAL ]; then
845 echo "New config not seen: wanted $FINAL got $RESULT"
848 echo "New config success: got $RESULT"
850 set_and_check client "$TEST" "$PARAM" $ORIG || return 5
853 run_test 28a "permanent parameter setting"
858 set_and_check $NODE "$LCTL get_param -n $1*.$2 | head -1" "$1.$2" "$3" || \
859 error "conf_param $1.$2 failed"
865 do_facet mgs "$LCTL conf_param foo=1 2>/dev/null" && \
866 error "Bad format should fail"
867 do_facet mgs "$LCTL conf_param osc.notanfs-OST0000.active=0 2>/dev/null" && \
868 error "Setting on unknown fs should fail"
869 do_facet mgs "$LCTL conf_param osc.$FSNAME-OST00000.active=0 2>/dev/null" && \
870 error "Bad target name should fail"
872 check_28b mds mdt.$FSNAME-MDT0000 capa_timeout 1500
873 check_28b mds mdt.$FSNAME-MDT* identity_expire 150
874 check_28b mds mdd.$FSNAME-MDT0000 atime_diff 15
875 check_28b mds mdd.$FSNAME-MDT* sync_permission 0
876 check_28b ost1 obdfilter.$FSNAME-OST0000 client_cache_seconds 15
877 check_28b ost1 obdfilter.$FSNAME-OST* client_cache_count 15
878 check_28b mds lov.$FSNAME-MDT0000 qos_maxage "15 Sec"
879 check_28b mds lov.$FSNAME-MDT0000 qos_prio_free "15%"
880 check_28b client mdc.$FSNAME-MDT0000 max_rpcs_in_flight 15
881 check_28b client osc.$FSNAME-OST0000 active 0
882 check_28b client osc.$FSNAME-OST0000 active 1
883 check_28b client osc.$FSNAME-OST0000 max_dirty_mb 15
884 check_28b client llite.$FSNAME max_read_ahead_mb 15
885 set_and_check client "$LCTL get_param -n at_max" "sys.$FSNAME.at_max" 1500 || \
886 error "conf_param sys.fsname.at_max failed"
889 run_test 28b "permanent parameter setting, set_param syntax"
892 [ "$OSTCOUNT" -lt "2" ] && skip_env "$OSTCOUNT < 2, skipping" && return
893 setup > /dev/null 2>&1
897 local PARAM="osc.$FSNAME-OST0001.active"
898 local PROC_ACT="osc.$FSNAME-OST0001-osc-[^M]*.active"
899 local PROC_UUID="osc.$FSNAME-OST0001-osc-[^M]*.ost_server_uuid"
901 ACTV=$(lctl get_param -n $PROC_ACT)
903 set_and_check client "lctl get_param -n $PROC_ACT" "$PARAM" $DEAC || return 2
904 # also check ost_server_uuid status
905 RESULT=$(lctl get_param -n $PROC_UUID | grep DEACTIV)
906 if [ -z "$RESULT" ]; then
907 echo "Live client not deactivated: $(lctl get_param -n $PROC_UUID)"
910 echo "Live client success: got $RESULT"
914 local mdtosc=$(get_mdtosc_proc_path $SINGLEMDS $FSNAME-OST0001)
915 mdtosc=${mdtosc/-MDT*/-MDT\*}
916 local MPROC="osc.$mdtosc.active"
921 RESULT=`do_facet $SINGLEMDS " lctl get_param -n $MPROC"`
922 [ ${PIPESTATUS[0]} = 0 ] || error "Can't read $MPROC"
923 if [ $RESULT -eq $DEAC ]; then
924 echo "MDT deactivated also after $WAIT sec (got $RESULT)"
928 if [ $WAIT -eq $MAX ]; then
929 echo "MDT not deactivated: wanted $DEAC got $RESULT"
932 echo "Waiting $(($MAX - $WAIT)) secs for MDT deactivated"
935 # test new client starts deactivated
936 umount_client $MOUNT || return 200
938 RESULT=$(lctl get_param -n $PROC_UUID | grep DEACTIV | grep NEW)
939 if [ -z "$RESULT" ]; then
940 echo "New client not deactivated from start: $(lctl get_param -n $PROC_UUID)"
943 echo "New client success: got $RESULT"
946 # make sure it reactivates
947 set_and_check client "lctl get_param -n $PROC_ACT" "$PARAM" $ACTV || return 6
952 #writeconf to remove all ost2 traces for subsequent tests
955 run_test 29 "permanently remove an OST"
961 TEST="lctl get_param -n llite.$FSNAME-*.max_read_ahead_whole_mb"
963 LIST=(1 2 3 4 5 4 3 2 1 2 3 4 5 4 3 2 1 2 3 4 5)
964 for i in ${LIST[@]}; do
965 set_and_check client "$TEST" "llite.$FSNAME.max_read_ahead_whole_mb" $i || return 3
967 # make sure client restart still works
969 mount_client $MOUNT || return 4
970 [ "$($TEST)" -ne "$i" ] && error "Param didn't stick across restart $($TEST) != $i"
973 echo Erase parameter setting
974 do_facet mgs "$LCTL conf_param -d llite.$FSNAME.max_read_ahead_whole_mb" || return 6
976 mount_client $MOUNT || return 6
978 echo "deleted (default) value=$FINAL, orig=$ORIG"
979 # assumes this parameter started at the default value
980 [ "$FINAL" -eq "$ORIG" ] || fail "Deleted value=$FINAL, orig=$ORIG"
984 run_test 30a "Big config llog and conf_param deletion"
989 # Make a fake nid. Use the OST nid, and add 20 to the least significant
990 # numerical part of it. Hopefully that's not already a failover address for
992 OSTNID=$(do_facet ost1 "$LCTL get_param nis" | tail -1 | awk '{print $1}')
993 ORIGVAL=$(echo $OSTNID | egrep -oi "[0-9]*@")
994 NEWVAL=$((($(echo $ORIGVAL | egrep -oi "[0-9]*") + 20) % 256))
995 NEW=$(echo $OSTNID | sed "s/$ORIGVAL/$NEWVAL@/")
996 echo "Using fake nid $NEW"
998 TEST="$LCTL get_param -n osc.$FSNAME-OST0000-osc-[^M]*.import | grep failover_nids | sed -n 's/.*\($NEW\).*/\1/p'"
999 set_and_check client "$TEST" "osc.$FSNAME-OST0000.failover.node" $NEW || error "didn't add failover nid $NEW"
1000 NIDS=$($LCTL get_param -n osc.$FSNAME-OST0000-osc-[^M]*.import | grep failover_nids)
1002 NIDCOUNT=$(($(echo "$NIDS" | wc -w) - 1))
1003 echo "should have 2 failover nids: $NIDCOUNT"
1004 [ $NIDCOUNT -eq 2 ] || error "Failover nid not added"
1005 do_facet mgs "$LCTL conf_param -d osc.$FSNAME-OST0000.failover.node" || error "conf_param delete failed"
1006 umount_client $MOUNT
1007 mount_client $MOUNT || return 3
1009 NIDS=$($LCTL get_param -n osc.$FSNAME-OST0000-osc-[^M]*.import | grep failover_nids)
1011 NIDCOUNT=$(($(echo "$NIDS" | wc -w) - 1))
1012 echo "only 1 final nid should remain: $NIDCOUNT"
1013 [ $NIDCOUNT -eq 1 ] || error "Failover nids not removed"
1017 run_test 30b "Remove failover nids"
1019 test_31() { # bug 10734
1020 # ipaddr must not exist
1021 mount -t lustre 4.3.2.1@tcp:/lustre $MOUNT || true
1024 run_test 31 "Connect to non-existent node (shouldn't crash)"
1026 # Use these start32/stop32 fn instead of t-f start/stop fn,
1027 # for local devices, to skip global facet vars init
1031 echo "Stopping local ${MOUNT%/*}/${facet} (opts:$@)"
1032 umount -d $@ ${MOUNT%/*}/${facet}
1041 mkdir -p ${MOUNT%/*}/${facet}
1043 echo "Starting local ${facet}: $@ $device ${MOUNT%/*}/${facet}"
1044 mount -t lustre $@ ${device} ${MOUNT%/*}/${facet}
1046 if [ $RC -ne 0 ]; then
1047 echo "mount -t lustre $@ ${device} ${MOUNT%/*}/${facet}"
1048 echo "Start of ${device} of local ${facet} failed ${RC}"
1054 cleanup_nocli32 () {
1062 echo "Cleanup test_32 umount $MOUNT ..."
1063 umount -f $MOUNT || true
1064 echo "Cleanup local mds ost1 ..."
1070 client_only && skip "client only testing" && return 0
1071 [ "$NETTYPE" = "tcp" ] || { skip "NETTYPE != tcp" && return 0; }
1072 [ -z "$TUNEFS" ] && skip_env "No tunefs" && return 0
1074 local DISK1_8=$LUSTRE/tests/disk1_8.tar.bz2
1075 [ ! -r $DISK1_8 ] && skip_env "Cannot find $DISK1_8" && return 0
1076 local tmpdir=$TMP/conf32a
1078 tar xjvf $DISK1_8 -C $tmpdir || \
1079 { skip_env "Cannot untar $DISK1_8" && return 0; }
1082 $LCTL set_param debug=$PTLDEBUG
1084 $TUNEFS $tmpdir/mds || error "tunefs failed"
1086 # nids are wrong, so client wont work, but server should start
1087 start32 mds1 $tmpdir/mds "-o loop,exclude=lustre-OST0000" && \
1088 trap cleanup_32 EXIT INT || return 3
1090 local UUID=$($LCTL get_param -n mdt.lustre-MDT0000.uuid)
1092 [ "$UUID" == "lustre-MDT0000_UUID" ] || error "UUID is wrong: $UUID"
1094 $TUNEFS --mgsnode=$HOSTNAME $tmpdir/ost1 || error "tunefs failed"
1095 start32 ost1 $tmpdir/ost1 "-o loop" || return 5
1096 UUID=$($LCTL get_param -n obdfilter.lustre-OST0000.uuid)
1098 [ "$UUID" == "lustre-OST0000_UUID" ] || error "UUID is wrong: $UUID"
1100 local NID=$($LCTL list_nids | head -1)
1102 echo "OSC changes should succeed:"
1103 $LCTL conf_param lustre-OST0000.osc.max_dirty_mb=15 || return 7
1104 $LCTL conf_param lustre-OST0000.failover.node=$NID || return 8
1107 echo "MDC changes should succeed:"
1108 $LCTL conf_param lustre-MDT0000.mdc.max_rpcs_in_flight=9 || return 9
1109 $LCTL conf_param lustre-MDT0000.failover.node=$NID || return 10
1112 echo "LOV changes should succeed:"
1113 $LCTL pool_new lustre.interop || return 11
1114 $LCTL conf_param lustre-MDT0000.lov.stripesize=4M || return 12
1119 # mount a second time to make sure we didnt leave upgrade flag on
1121 $TUNEFS --dryrun $tmpdir/mds || error "tunefs failed"
1122 start32 mds1 $tmpdir/mds "-o loop,exclude=lustre-OST0000" && \
1123 trap cleanup_32 EXIT INT || return 12
1127 rm -rf $tmpdir || true # true is only for TMP on NFS
1129 run_test 32a "Upgrade from 1.8 (not live)"
1132 client_only && skip "client only testing" && return 0
1133 [ "$NETTYPE" = "tcp" ] || { skip "NETTYPE != tcp" && return 0; }
1134 [ -z "$TUNEFS" ] && skip_env "No tunefs" && return
1136 local DISK1_8=$LUSTRE/tests/disk1_8.tar.bz2
1137 [ ! -r $DISK1_8 ] && skip_env "Cannot find $DISK1_8" && return 0
1138 local tmpdir=$TMP/conf32b
1140 tar xjvf $DISK1_8 -C $tmpdir || \
1141 { skip_env "Cannot untar $DISK1_8" && return ; }
1144 $LCTL set_param debug="config"
1145 local NEWNAME=lustre
1147 # writeconf will cause servers to register with their current nids
1148 $TUNEFS --writeconf --fsname=$NEWNAME $tmpdir/mds || error "tunefs failed"
1149 start32 mds1 $tmpdir/mds "-o loop" && \
1150 trap cleanup_32 EXIT INT || return 3
1152 local UUID=$($LCTL get_param -n mdt.${NEWNAME}-MDT0000.uuid)
1154 [ "$UUID" == "${NEWNAME}-MDT0000_UUID" ] || error "UUID is wrong: $UUID"
1156 $TUNEFS --mgsnode=$HOSTNAME --writeconf --fsname=$NEWNAME $tmpdir/ost1 ||\
1157 error "tunefs failed"
1158 start32 ost1 $tmpdir/ost1 "-o loop" || return 5
1159 UUID=$($LCTL get_param -n obdfilter.${NEWNAME}-OST0000.uuid)
1161 [ "$UUID" == "${NEWNAME}-OST0000_UUID" ] || error "UUID is wrong: $UUID"
1163 local NID=$($LCTL list_nids | head -1)
1165 echo "OSC changes should succeed:"
1166 $LCTL conf_param ${NEWNAME}-OST0000.osc.max_dirty_mb=15 || return 7
1167 $LCTL conf_param ${NEWNAME}-OST0000.failover.node=$NID || return 8
1170 echo "MDC changes should succeed:"
1171 $LCTL conf_param ${NEWNAME}-MDT0000.mdc.max_rpcs_in_flight=9 || return 9
1172 $LCTL conf_param ${NEWNAME}-MDT0000.failover.node=$NID || return 10
1175 echo "LOV changes should succeed:"
1176 $LCTL pool_new ${NEWNAME}.interop || return 11
1177 $LCTL conf_param ${NEWNAME}-MDT0000.lov.stripesize=4M || return 12
1180 # MDT and OST should have registered with new nids, so we should have
1181 # a fully-functioning client
1182 echo "Check client and old fs contents"
1184 local device=`h2$NETTYPE $HOSTNAME`:/$NEWNAME
1185 echo "Starting local client: $HOSTNAME: $device $MOUNT"
1186 mount -t lustre $device $MOUNT || return 1
1188 local old=$($LCTL get_param -n mdc.*.max_rpcs_in_flight)
1189 local new=$((old + 5))
1190 $LCTL conf_param ${NEWNAME}-MDT0000.mdc.max_rpcs_in_flight=$new
1191 wait_update $HOSTNAME "$LCTL get_param -n mdc.*.max_rpcs_in_flight" $new || return 11
1193 [ "$(cksum $MOUNT/passwd | cut -d' ' -f 1,2)" == "94306271 1478" ] || return 12
1198 rm -rf $tmpdir || true # true is only for TMP on NFS
1200 run_test 32b "Upgrade from 1.8 with writeconf"
1202 test_33a() { # bug 12333, was test_33
1204 local FSNAME2=test-123
1205 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
1207 [ -n "$ost1_HOST" ] && fs2ost_HOST=$ost1_HOST
1209 if [ -z "$fs2ost_DEV" -o -z "$fs2mds_DEV" ]; then
1210 do_facet $SINGLEMDS [ -b "$MDSDEV" ] && \
1211 skip_env "mixed loopback and real device not working" && return
1214 local fs2mdsdev=${fs2mds_DEV:-${MDSDEV}_2}
1215 local fs2ostdev=${fs2ost_DEV:-$(ostdevname 1)_2}
1216 add fs2mds $MDS_MKFS_OPTS --mkfsoptions='\"-J size=8\"' --fsname=${FSNAME2} --reformat $fs2mdsdev || exit 10
1217 add fs2ost $OST_MKFS_OPTS --fsname=${FSNAME2} --index=8191 --mgsnode=$MGSNID --reformat $fs2ostdev || exit 10
1219 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS && trap cleanup_24a EXIT INT
1220 start fs2ost $fs2ostdev $OST_MOUNT_OPTS
1221 do_facet mgs "$LCTL conf_param sys.$FSNAME2.timeout=200" || rc=1
1223 mount -t lustre $MGSNID:/${FSNAME2} $MOUNT2 || rc=2
1226 cp /etc/hosts $MOUNT2/ || rc=3
1227 $LFS getstripe $MOUNT2/hosts
1232 rm -rf $MOUNT2 $fs2mdsdev $fs2ostdev
1233 cleanup_nocli || rc=6
1236 run_test 33a "Mount ost with a large index number"
1238 test_33b() { # was test_34
1241 do_facet client dd if=/dev/zero of=$MOUNT/24 bs=1024k count=1
1242 # Drop lock cancelation reply during umount
1243 #define OBD_FAIL_LDLM_CANCEL 0x304
1244 do_facet client lctl set_param fail_loc=0x80000304
1245 #lctl set_param debug=-1
1246 umount_client $MOUNT
1249 run_test 33b "Drop cancel during umount"
1253 do_facet client "sh runmultiop_bg_pause $DIR/file O_c"
1254 manual_umount_client
1256 do_facet client killall -USR1 multiop
1257 if [ $rc -eq 0 ]; then
1258 error "umount not fail!"
1263 run_test 34a "umount with opened file should fail"
1268 touch $DIR/$tfile || return 1
1269 stop_mds --force || return 2
1271 manual_umount_client --force
1273 if [ $rc -ne 0 ]; then
1274 error "mtab after failed umount - rc $rc"
1280 run_test 34b "force umount with failed mds should be normal"
1284 touch $DIR/$tfile || return 1
1285 stop_ost --force || return 2
1287 manual_umount_client --force
1289 if [ $rc -ne 0 ]; then
1290 error "mtab after failed umount - rc $rc"
1296 run_test 34c "force umount with failed ost should be normal"
1298 test_35a() { # bug 12459
1301 DBG_SAVE="`lctl get_param -n debug`"
1302 lctl set_param debug="ha"
1304 log "Set up a fake failnode for the MDS"
1306 local device=$(do_facet $SINGLEMDS "lctl get_param -n devices" | awk '($3 ~ "mdt" && $4 ~ "MDT") { print $4 }' | head -1)
1307 do_facet $SINGLEMDS $LCTL conf_param mdc.${device}.failover.node=$FAKENID || return 4
1309 log "Wait for RECONNECT_INTERVAL seconds (10s)"
1312 MSG="conf-sanity.sh test_35a `date +%F%kh%Mm%Ss`"
1315 log "Stopping the MDT:"
1316 stop_mds || return 5
1318 df $MOUNT > /dev/null 2>&1 &
1320 log "Restarting the MDT:"
1321 start_mds || return 6
1322 log "Wait for df ($DFPID) ... "
1325 lctl set_param debug="$DBG_SAVE"
1327 # retrieve from the log the first server that the client tried to
1328 # contact after the connection loss
1329 $LCTL dk $TMP/lustre-log-$TESTNAME.log
1330 NEXTCONN=`awk "/${MSG}/ {start = 1;}
1331 /import_select_connection.*$device-mdc.* using connection/ {
1333 if (\\\$NF ~ /$FAKENID/)
1339 }" $TMP/lustre-log-$TESTNAME.log`
1340 [ "$NEXTCONN" != "0" ] && log "The client didn't try to reconnect to the last active server (tried ${NEXTCONN} instead)" && return 7
1342 # remove nid settings
1345 run_test 35a "Reconnect to the last active server first"
1347 test_35b() { # bug 18674
1348 remote_mds || { skip "local MDS" && return 0; }
1352 $LCTL set_param debug="ha"
1354 MSG="conf-sanity.sh test_35b `date +%F%kh%Mm%Ss`"
1357 log "Set up a fake failnode for the MDS"
1359 local device=$(do_facet mds "$LCTL get_param -n devices" | \
1360 awk '($3 ~ "mdt" && $4 ~ "MDT") { print $4 }' | head -1)
1361 do_facet mds "$LCTL conf_param mdc.${device}.failover.node=$FAKENID" || \
1364 local at_max_saved=0
1365 # adaptive timeouts may prevent seeing the issue
1366 if at_is_enabled; then
1367 at_max_saved=$(at_max_get mds)
1368 at_max_set 0 mds client
1371 mkdir -p $MOUNT/$tdir
1373 log "Injecting EBUSY on MDS"
1374 # Setting OBD_FAIL_MDS_RESEND=0x136
1375 do_facet mds "$LCTL set_param fail_loc=0x80000136" || return 2
1377 $LCTL set_param mdc.${FSNAME}*.stats=clear
1379 log "Creating a test file and stat it"
1380 touch $MOUNT/$tdir/$tfile
1381 stat $MOUNT/$tdir/$tfile
1383 log "Stop injecting EBUSY on MDS"
1384 do_facet mds "$LCTL set_param fail_loc=0" || return 3
1385 rm -f $MOUNT/$tdir/$tfile
1388 # restore adaptive timeout
1389 [ $at_max_saved -ne 0 ] && at_max_set $at_max_saved mds client
1391 $LCTL dk $TMP/lustre-log-$TESTNAME.log
1393 CONNCNT=`$LCTL get_param mdc.${FSNAME}*.stats | awk '/mds_connect/{print $2}'`
1395 # retrieve from the log if the client has ever tried to
1396 # contact the fake server after the loss of connection
1397 FAILCONN=`awk "BEGIN {ret = 0;}
1398 /import_select_connection.*${FSNAME}-MDT0000-mdc.* using connection/ {
1400 if (\\\$NF ~ /$FAKENID/) {
1405 END {print ret}" $TMP/lustre-log-$TESTNAME.log`
1407 [ "$FAILCONN" == "0" ] && \
1408 log "ERROR: The client reconnection has not been triggered" && \
1410 [ "$FAILCONN" == "2" ] && \
1411 log "ERROR: The client tried to reconnect to the failover server while the primary was busy" && \
1414 # When OBD_FAIL_MDS_RESEND is hit, we sleep for 2 * obd_timeout
1415 # Reconnects are supposed to be rate limited to one every 5s
1416 [ $CONNCNT -gt $((2 * $TIMEOUT / 5 + 1)) ] && \
1417 log "ERROR: Too many reconnects $CONNCNT" && \
1421 # remove nid settings
1424 run_test 35b "Continue reconnection retries, if the active server is busy"
1428 local FSNAME2=test1234
1429 local fs3ost_HOST=$ost_HOST
1430 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
1432 [ -n "$ost1_HOST" ] && fs2ost_HOST=$ost1_HOST && fs3ost_HOST=$ost1_HOST
1435 if [ -z "$fs2ost_DEV" -o -z "$fs2mds_DEV" -o -z "$fs3ost_DEV" ]; then
1436 do_facet $SINGLEMDS [ -b "$MDSDEV" ] && \
1437 skip_env "mixed loopback and real device not working" && return
1439 [ $OSTCOUNT -lt 2 ] && skip_env "skipping test for single OST" && return
1441 [ "$ost_HOST" = "`hostname`" -o "$ost1_HOST" = "`hostname`" ] || \
1442 { skip "remote OST" && return 0; }
1444 local fs2mdsdev=${fs2mds_DEV:-${MDSDEV}_2}
1445 local fs2ostdev=${fs2ost_DEV:-$(ostdevname 1)_2}
1446 local fs3ostdev=${fs3ost_DEV:-$(ostdevname 2)_2}
1447 add fs2mds $MDS_MKFS_OPTS --fsname=${FSNAME2} --reformat $fs2mdsdev || exit 10
1448 # XXX after we support non 4K disk blocksize, change following --mkfsoptions with
1450 add fs2ost $OST_MKFS_OPTS --mkfsoptions='-b4096' --fsname=${FSNAME2} --mgsnode=$MGSNID --reformat $fs2ostdev || exit 10
1451 add fs3ost $OST_MKFS_OPTS --mkfsoptions='-b4096' --fsname=${FSNAME2} --mgsnode=$MGSNID --reformat $fs3ostdev || exit 10
1453 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS
1454 start fs2ost $fs2ostdev $OST_MOUNT_OPTS
1455 start fs3ost $fs3ostdev $OST_MOUNT_OPTS
1457 mount -t lustre $MGSNID:/${FSNAME2} $MOUNT2 || return 1
1459 sleep 5 # until 11778 fixed
1461 dd if=/dev/zero of=$MOUNT2/$tfile bs=1M count=7 || return 2
1463 BKTOTAL=`lctl get_param -n obdfilter.*.kbytestotal | awk 'BEGIN{total=0}; {total+=$1}; END{print total}'`
1464 BKFREE=`lctl get_param -n obdfilter.*.kbytesfree | awk 'BEGIN{free=0}; {free+=$1}; END{print free}'`
1465 BKAVAIL=`lctl get_param -n obdfilter.*.kbytesavail | awk 'BEGIN{avail=0}; {avail+=$1}; END{print avail}'`
1466 STRING=`df -P $MOUNT2 | tail -n 1 | awk '{print $2","$3","$4}'`
1467 DFTOTAL=`echo $STRING | cut -d, -f1`
1468 DFUSED=`echo $STRING | cut -d, -f2`
1469 DFAVAIL=`echo $STRING | cut -d, -f3`
1470 DFFREE=$(($DFTOTAL - $DFUSED))
1472 ALLOWANCE=$((64 * $OSTCOUNT))
1474 if [ $DFTOTAL -lt $(($BKTOTAL - $ALLOWANCE)) ] ||
1475 [ $DFTOTAL -gt $(($BKTOTAL + $ALLOWANCE)) ] ; then
1476 echo "**** FAIL: df total($DFTOTAL) mismatch OST total($BKTOTAL)"
1479 if [ $DFFREE -lt $(($BKFREE - $ALLOWANCE)) ] ||
1480 [ $DFFREE -gt $(($BKFREE + $ALLOWANCE)) ] ; then
1481 echo "**** FAIL: df free($DFFREE) mismatch OST free($BKFREE)"
1484 if [ $DFAVAIL -lt $(($BKAVAIL - $ALLOWANCE)) ] ||
1485 [ $DFAVAIL -gt $(($BKAVAIL + $ALLOWANCE)) ] ; then
1486 echo "**** FAIL: df avail($DFAVAIL) mismatch OST avail($BKAVAIL)"
1491 stop fs3ost -f || return 200
1492 stop fs2ost -f || return 201
1493 stop fs2mds -f || return 202
1494 rm -rf $MOUNT2 $fs2mdsdev $fs2ostdev $fs3ostdev
1495 unload_modules || return 203
1498 run_test 36 "df report consistency on OSTs with different block size"
1501 client_only && skip "client only testing" && return 0
1502 LOCAL_MDSDEV="$TMP/mdt.img"
1503 SYM_MDSDEV="$TMP/sym_mdt.img"
1505 echo "MDS : $LOCAL_MDSDEV"
1506 echo "SYMLINK : $SYM_MDSDEV"
1510 mkfs.lustre --reformat --fsname=lustre --mdt --mgs --device-size=9000 $LOCAL_MDSDEV ||
1511 error "mkfs.lustre $LOCAL_MDSDEV failed"
1512 ln -s $LOCAL_MDSDEV $SYM_MDSDEV
1514 echo "mount symlink device - $SYM_MDSDEV"
1516 mount_op=`mount -v -t lustre -o loop $SYM_MDSDEV ${MOUNT%/*}/mds 2>&1 | grep "unable to set tunable"`
1517 umount -d ${MOUNT%/*}/mds
1518 rm -f $LOCAL_MDSDEV $SYM_MDSDEV
1520 if [ -n "$mount_op" ]; then
1521 error "**** FAIL: set tunables failed for symlink device"
1525 run_test 37 "verify set tunables works for symlink device"
1527 test_38() { # bug 14222
1532 FILES=`find $SRC -type f -mtime +1 | head -n $COUNT`
1533 log "copying $(echo $FILES | wc -w) files to $DIR/$tdir"
1535 tar cf - $FILES | tar xf - -C $DIR/$tdir || \
1536 error "copying $SRC to $DIR/$tdir"
1538 umount_client $MOUNT
1540 log "rename lov_objid file on MDS"
1541 rm -f $TMP/lov_objid.orig
1543 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
1544 do_facet $SINGLEMDS "$DEBUGFS -c -R \\\"dump lov_objid $TMP/lov_objid.orig\\\" $MDSDEV"
1545 do_facet $SINGLEMDS "$DEBUGFS -w -R \\\"rm lov_objid\\\" $MDSDEV"
1547 do_facet $SINGLEMDS "od -Ax -td8 $TMP/lov_objid.orig"
1548 # check create in mds_lov_connect
1552 [ $V ] && log "verifying $DIR/$tdir/$f"
1553 diff -q $f $DIR/$tdir/$f || ERROR=y
1555 do_facet $SINGLEMDS "$DEBUGFS -c -R \\\"dump lov_objid $TMP/lov_objid.new\\\" $MDSDEV"
1556 do_facet $SINGLEMDS "od -Ax -td8 $TMP/lov_objid.new"
1557 [ "$ERROR" = "y" ] && error "old and new files are different after connect" || true
1559 # check it's updates in sync
1560 umount_client $MOUNT
1563 do_facet $SINGLEMDS dd if=/dev/zero of=$TMP/lov_objid.clear bs=4096 count=1
1564 do_facet $SINGLEMDS "$DEBUGFS -w -R \\\"rm lov_objid\\\" $MDSDEV"
1565 do_facet $SINGLEMDS "$DEBUGFS -w -R \\\"write $TMP/lov_objid.clear lov_objid\\\" $MDSDEV "
1570 [ $V ] && log "verifying $DIR/$tdir/$f"
1571 diff -q $f $DIR/$tdir/$f || ERROR=y
1573 do_facet $SINGLEMDS "$DEBUGFS -c -R \\\"dump lov_objid $TMP/lov_objid.new1\\\" $MDSDEV"
1574 do_facet $SINGLEMDS "od -Ax -td8 $TMP/lov_objid.new1"
1575 umount_client $MOUNT
1577 [ "$ERROR" = "y" ] && error "old and new files are different after sync" || true
1579 log "files compared the same"
1582 run_test 38 "MDS recreates missing lov_objid file from OST data"
1588 perl $SRCDIR/leak_finder.pl $TMP/debug 2>&1 | egrep '*** Leak:' &&
1589 error "memory leak detected" || true
1591 run_test 39 "leak_finder recognizes both LUSTRE and LNET malloc messages"
1593 test_40() { # bug 15759
1595 #define OBD_FAIL_TGT_TOOMANY_THREADS 0x706
1596 do_facet $SINGLEMDS "sysctl -w lustre.fail_loc=0x80000706"
1600 run_test 40 "race during service thread startup"
1602 test_41() { #bug 14134
1604 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
1606 start $SINGLEMDS $MDSDEV $MDS_MOUNT_OPTS -o nosvc -n
1607 start ost1 `ostdevname 1` $OST_MOUNT_OPTS
1608 start $SINGLEMDS $MDSDEV $MDS_MOUNT_OPTS -o nomgs,force
1610 mount_client $MOUNT || return 1
1613 echo "blah blah" > $MOUNT/$tfile
1616 umount_client $MOUNT
1617 stop ost1 -f || return 201
1618 stop_mds -f || return 202
1619 stop_mds -f || return 203
1620 unload_modules || return 204
1623 run_test 41 "mount mds with --nosvc and --nomgs"
1625 test_42() { #bug 14693
1627 check_mount || return 2
1628 do_facet mgs $LCTL conf_param llite.$FSNAME.some_wrong_param=10
1629 umount_client $MOUNT
1630 mount_client $MOUNT || return 1
1634 run_test 42 "invalid config param should not prevent client from mounting"
1637 [ $UID -ne 0 -o $RUNAS_ID -eq 0 ] && skip_env "run as root"
1639 chmod ugo+x $DIR || error "chmod 0 failed"
1641 "lctl get_param -n mdt.$FSNAME-MDT0000.root_squash" \
1642 "mdt.$FSNAME-MDT*.root_squash" \
1645 "lctl get_param -n mdt.$FSNAME-MDT0000.nosquash_nids" \
1646 "mdt.$FSNAME-MDT*.nosquash_nids" \
1650 # create set of test files
1652 echo "111" > $DIR/$tfile-userfile || error "write 1 failed"
1653 chmod go-rw $DIR/$tfile-userfile || error "chmod 1 failed"
1654 chown $RUNAS_ID.$RUNAS_ID $DIR/$tfile-userfile || error "chown failed"
1656 echo "222" > $DIR/$tfile-rootfile || error "write 2 failed"
1657 chmod go-rw $DIR/$tfile-rootfile || error "chmod 2 faield"
1659 mkdir $DIR/$tdir-rootdir -p || error "mkdir failed"
1660 chmod go-rwx $DIR/$tdir-rootdir || error "chmod 3 failed"
1661 touch $DIR/$tdir-rootdir/tfile-1 || error "touch failed"
1664 # check root_squash:
1665 # set root squash UID:GID to RUNAS_ID
1666 # root should be able to access only files owned by RUNAS_ID
1669 "lctl get_param -n mdt.$FSNAME-MDT0000.root_squash" \
1670 "mdt.$FSNAME-MDT*.root_squash" \
1671 "$RUNAS_ID:$RUNAS_ID"
1673 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-userfile)
1674 dd if=$DIR/$tfile-userfile 1>/dev/null 2>/dev/null || \
1675 error "$ST: root read permission is denied"
1676 echo "$ST: root read permission is granted - ok"
1679 dd conv=notrunc if=$DIR/$tfile-userfile 1>/dev/null 2>/dev/null || \
1680 error "$ST: root write permission is denied"
1681 echo "$ST: root write permission is granted - ok"
1683 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-rootfile)
1684 dd if=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null && \
1685 error "$ST: root read permission is granted"
1686 echo "$ST: root read permission is denied - ok"
1689 dd conv=notrunc of=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null && \
1690 error "$ST: root write permission is granted"
1691 echo "$ST: root write permission is denied - ok"
1693 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tdir-rootdir)
1694 rm $DIR/$tdir-rootdir/tfile-1 1>/dev/null 2>/dev/null && \
1695 error "$ST: root unlink permission is granted"
1696 echo "$ST: root unlink permission is denied - ok"
1698 touch $DIR/tdir-rootdir/tfile-2 1>/dev/null 2>/dev/null && \
1699 error "$ST: root create permission is granted"
1700 echo "$ST: root create permission is denied - ok"
1703 # check nosquash_nids:
1704 # put client's NID into nosquash_nids list,
1705 # root should be able to access root file after that
1707 local NIDLIST=$(lctl list_nids all | tr '\n' ' ')
1708 NIDLIST="2@elan $NIDLIST 192.168.0.[2,10]@tcp"
1709 NIDLIST=$(echo $NIDLIST | tr -s ' ' ' ')
1711 "lctl get_param -n mdt.$FSNAME-MDT0000.nosquash_nids" \
1712 "mdt.$FSNAME-MDT*.nosquash_nids" \
1715 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-rootfile)
1716 dd if=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null || \
1717 error "$ST: root read permission is denied"
1718 echo "$ST: root read permission is granted - ok"
1721 dd conv=notrunc of=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null || \
1722 error "$ST: root write permission is denied"
1723 echo "$ST: root write permission is granted - ok"
1725 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tdir-rootdir)
1726 rm $DIR/$tdir-rootdir/tfile-1 || \
1727 error "$ST: root unlink permission is denied"
1728 echo "$ST: root unlink permission is granted - ok"
1729 touch $DIR/$tdir-rootdir/tfile-2 || \
1730 error "$ST: root create permission is denied"
1731 echo "$ST: root create permission is granted - ok"
1735 run_test 43 "check root_squash and nosquash_nids"
1737 umount_client $MOUNT
1742 check_mount || return 2
1743 UUID=$($LCTL get_param llite.${FSNAME}*.uuid | cut -d= -f2)
1745 UUIDS=$(do_facet mds "$LCTL get_param mdt.${FSNAME}*.exports.*.uuid")
1746 for VAL in $UUIDS; do
1747 NID=$(echo $VAL | cut -d= -f1)
1748 CLUUID=$(echo $VAL | cut -d= -f2)
1749 [ "$UUID" = "$CLUUID" ] && STATS_FOUND=yes && break
1751 [ "$STATS_FOUND" = "no" ] && error "stats not found for client"
1755 run_test 44 "mounted client proc entry exists"
1759 check_mount || return 2
1764 #define OBD_FAIL_PTLRPC_LONG_UNLINK 0x50f
1765 do_facet client "lctl set_param fail_loc=0x50f"
1768 manual_umount_client --force || return 3
1769 do_facet client "lctl set_param fail_loc=0x0"
1771 mount_client $MOUNT || return 4
1775 run_test 45 "long unlink handling in ptlrpcd"
1782 umount_client $MOUNT2 || rc=$?
1783 umount_client $MOUNT || rc=$?
1784 while [ $count -gt 0 ]; do
1785 stop ost${count} -f || rc=$?
1789 cleanup_nocli || rc=$?
1790 #writeconf to remove all ost2 traces for subsequent tests
1796 echo "Testing with $OSTCOUNT OSTs"
1798 start_mds || return 1
1799 #first client should see only one ost
1800 start_ost || return 2
1801 wait_osc_import_state mds ost FULL
1803 mount_client $MOUNT || return 3
1804 trap "cleanup_46a $OSTCOUNT" EXIT ERR
1807 for (( i=2; i<=$OSTCOUNT; i++ )); do
1808 start ost$i `ostdevname $i` $OST_MOUNT_OPTS || return $((i+2))
1811 # wait until osts in sync
1812 for (( i=2; i<=$OSTCOUNT; i++ )); do
1813 wait_osc_import_state mds ost$i FULL
1817 #second client see all ost's
1819 mount_client $MOUNT2 || return 8
1820 $LFS setstripe $MOUNT2 -c -1 || return 9
1821 $LFS getstripe $MOUNT2 || return 10
1823 echo "ok" > $MOUNT2/widestripe
1824 $LFS getstripe $MOUNT2/widestripe || return 11
1825 # fill acl buffer for avoid expand lsm to them
1826 awk -F : '{if (FNR < 25) { print "u:"$1":rwx" }}' /etc/passwd | while read acl; do
1827 setfacl -m $acl $MOUNT2/widestripe
1831 stat $MOUNT/widestripe || return 12
1833 cleanup_46a $OSTCOUNT || { echo "cleanup_46a failed!" && return 13; }
1836 run_test 46a "handle ost additional - wide striped file"
1841 check_mount || return 2
1842 $LCTL set_param ldlm.namespaces.$FSNAME-*-*-*.lru_size=100
1846 for ns in $($LCTL get_param ldlm.namespaces.$FSNAME-*-*-*.lru_size); do
1847 if echo $ns | grep "MDT[[:digit:]]*"; then
1850 lrs=$(echo $ns | sed 's/.*lru_size=//')
1851 lru_size[count]=$lrs
1856 facet_failover $SINGLEMDS
1857 client_up || return 3
1860 for ns in $($LCTL get_param ldlm.namespaces.$FSNAME-*-*-*.lru_size); do
1861 if echo $ns | grep "MDT[[:digit:]]*"; then
1864 lrs=$(echo $ns | sed 's/.*lru_size=//')
1865 if ! test "$lrs" -eq "${lru_size[count]}"; then
1866 n=$(echo $ns | sed -e 's/ldlm.namespaces.//' -e 's/.lru_size=.*//')
1867 error "$n has lost lru_size: $lrs vs. ${lru_size[count]}"
1875 run_test 47 "server restart does not make client loss lru_resize settings"
1880 # reformat after this test is needed - if test will failed
1881 # we will have unkillable file at FS
1885 test_48() { # bug 17636
1888 check_mount || return 2
1890 $LFS setstripe $MOUNT -c -1 || return 9
1891 $LFS getstripe $MOUNT || return 10
1893 echo "ok" > $MOUNT/widestripe
1894 $LFS getstripe $MOUNT/widestripe || return 11
1896 trap cleanup_48 EXIT ERR
1898 # fill acl buffer for avoid expand lsm to them
1899 getent passwd | awk -F : '{ print "u:"$1":rwx" }' | while read acl; do
1900 setfacl -m $acl $MOUNT/widestripe
1903 stat $MOUNT/widestripe || return 12
1908 run_test 48 "too many acls on file"
1910 # check PARAM_SYS_LDLM_TIMEOUT option of MKFS.LUSTRE
1911 test_49() { # bug 17710
1912 local OLD_MDS_MKFS_OPTS=$MDS_MKFS_OPTS
1913 local OLD_OST_MKFS_OPTS=$OST_MKFS_OPTS
1914 local LOCAL_TIMEOUT=20
1917 OST_MKFS_OPTS="--ost --fsname=$FSNAME --device-size=$OSTSIZE --mgsnode=$MGSNID --param sys.timeout=$LOCAL_TIMEOUT --param sys.ldlm_timeout=$LOCAL_TIMEOUT $MKFSOPT $OSTOPT"
1921 check_mount || return 1
1923 echo "check ldlm_timout..."
1924 LDLM_MDS="`do_facet mds lctl get_param -n ldlm_timeout`"
1925 LDLM_OST1="`do_facet ost1 lctl get_param -n ldlm_timeout`"
1926 LDLM_CLIENT="`do_facet client lctl get_param -n ldlm_timeout`"
1928 if [ $LDLM_MDS -ne $LDLM_OST1 ] || [ $LDLM_MDS -ne $LDLM_CLIENT ]; then
1929 error "Different LDLM_TIMEOUT:$LDLM_MDS $LDLM_OST1 $LDLM_CLIENT"
1932 if [ $LDLM_MDS -ne $((LOCAL_TIMEOUT / 3)) ]; then
1933 error "LDLM_TIMEOUT($LDLM_MDS) is not correct"
1936 umount_client $MOUNT
1937 stop_ost || return 2
1938 stop_mds || return 3
1940 OST_MKFS_OPTS="--ost --fsname=$FSNAME --device-size=$OSTSIZE --mgsnode=$MGSNID --param sys.timeout=$LOCAL_TIMEOUT --param sys.ldlm_timeout=$((LOCAL_TIMEOUT - 1)) $MKFSOPT $OSTOPT"
1944 check_mount || return 7
1946 LDLM_MDS="`do_facet mds lctl get_param -n ldlm_timeout`"
1947 LDLM_OST1="`do_facet ost1 lctl get_param -n ldlm_timeout`"
1948 LDLM_CLIENT="`do_facet client lctl get_param -n ldlm_timeout`"
1950 if [ $LDLM_MDS -ne $LDLM_OST1 ] || [ $LDLM_MDS -ne $LDLM_CLIENT ]; then
1951 error "Different LDLM_TIMEOUT:$LDLM_MDS $LDLM_OST1 $LDLM_CLIENT"
1954 if [ $LDLM_MDS -ne $((LOCAL_TIMEOUT - 1)) ]; then
1955 error "LDLM_TIMEOUT($LDLM_MDS) is not correct"
1958 cleanup || return $?
1960 MDS_MKFS_OPTS=$OLD_MDS_MKFS_OPTS
1961 OST_MKFS_OPTS=$OLD_OST_MKFS_OPTS
1963 run_test 49 "check PARAM_SYS_LDLM_TIMEOUT option of MKFS.LUSTRE"
1966 # Test both statfs and lfs df and fail if either one fails
1967 multiop_bg_pause $1 f_
1970 killall -USR1 multiop
1971 [ $RC1 -ne 0 ] && log "lazystatfs multiop failed"
1972 wait $PID || { RC1=$?; log "multiop return error "; }
1979 if [ $RC2 -eq 0 ]; then
1981 log "lazystatfs df failed"
1985 [[ $RC1 -ne 0 || $RC2 -eq 0 ]] && RC=1
1991 lctl set_param llite.$FSNAME-*.lazystatfs=1
1994 lazystatfs $MOUNT || error "lazystatfs failed but no down servers"
1996 cleanup || return $?
1998 run_test 50a "lazystatfs all servers available =========================="
2002 lctl set_param llite.$FSNAME-*.lazystatfs=1
2005 # Wait for client to detect down OST
2006 stop_ost || error "Unable to stop OST1"
2007 wait_osc_import_state mds ost DISCONN
2009 lazystatfs $MOUNT || error "lazystatfs should don't have returned EIO"
2011 umount_client $MOUNT || error "Unable to unmount client"
2012 stop_mds || error "Unable to stop MDS"
2014 run_test 50b "lazystatfs all servers down =========================="
2017 start_mds || error "Unable to start MDS"
2018 start_ost || error "Unable to start OST1"
2019 start_ost2 || error "Unable to start OST2"
2020 mount_client $MOUNT || error "Unable to mount client"
2021 lctl set_param llite.$FSNAME-*.lazystatfs=1
2024 # Wait for client to detect down OST
2025 stop_ost || error "Unable to stop OST1"
2026 wait_osc_import_state mds ost DISCONN
2027 lazystatfs $MOUNT || error "lazystatfs failed with one down server"
2029 umount_client $MOUNT || error "Unable to unmount client"
2030 stop_ost2 || error "Unable to stop OST2"
2031 stop_mds || error "Unable to stop MDS"
2033 run_test 50c "lazystatfs one server down =========================="
2036 start_mds || error "Unable to start MDS"
2037 start_ost || error "Unable to start OST1"
2038 start_ost2 || error "Unable to start OST2"
2039 mount_client $MOUNT || error "Unable to mount client"
2040 lctl set_param llite.$FSNAME-*.lazystatfs=1
2043 # Issue the statfs during the window where the client still
2044 # belives the OST to be available but it is in fact down.
2045 # No failure just a statfs which hangs for a timeout interval.
2046 stop_ost || error "Unable to stop OST1"
2047 lazystatfs $MOUNT || error "lazystatfs failed with one down server"
2049 umount_client $MOUNT || error "Unable to unmount client"
2050 stop_ost2 || error "Unable to stop OST2"
2051 stop_mds || error "Unable to stop MDS"
2053 run_test 50d "lazystatfs client/server conn race =========================="
2060 start_mds || return 1
2061 #first client should see only one ost
2062 start_ost || return 2
2063 wait_osc_import_state mds ost FULL
2065 # Wait for client to detect down OST
2066 stop_ost || error "Unable to stop OST1"
2067 wait_osc_import_state mds ost DISCONN
2069 mount_client $MOUNT || error "Unable to mount client"
2070 lctl set_param llite.$FSNAME-*.lazystatfs=0
2072 multiop_bg_pause $MOUNT _f
2076 if [ $RC1 -ne 0 ]; then
2077 log "multiop failed $RC1"
2080 sleep $(( $TIMEOUT+1 ))
2082 [ $? -ne 0 ] && error "process isn't sleep"
2083 start_ost || error "Unable to start OST1"
2084 wait $pid || error "statfs failed"
2087 umount_client $MOUNT || error "Unable to unmount client"
2088 stop_ost || error "Unable to stop OST1"
2089 stop_mds || error "Unable to stop MDS"
2091 run_test 50e "normal statfs all servers down =========================="
2096 CONN_PROC="osc.$FSNAME-OST0001-osc-[M]*.ost_server_uuid"
2098 start_mds || error "Unable to start mds"
2099 #first client should see only one ost
2100 start_ost || error "Unable to start OST1"
2101 wait_osc_import_state mds ost FULL
2103 start_ost2 || error "Unable to start OST2"
2104 wait_osc_import_state mds ost2 FULL
2106 # Wait for client to detect down OST
2107 stop_ost2 || error "Unable to stop OST2"
2109 wait_osc_import_state mds ost2 DISCONN
2110 mount_client $MOUNT || error "Unable to mount client"
2111 lctl set_param llite.$FSNAME-*.lazystatfs=0
2113 multiop_bg_pause $MOUNT _f
2117 if [ $RC1 -ne 0 ]; then
2118 log "lazystatfs multiop failed $RC1"
2121 sleep $(( $TIMEOUT+1 ))
2123 [ $? -ne 0 ] && error "process isn't sleep"
2124 start_ost2 || error "Unable to start OST2"
2125 wait $pid || error "statfs failed"
2126 stop_ost2 || error "Unable to stop OST2"
2129 umount_client $MOUNT || error "Unable to unmount client"
2130 stop_ost || error "Unable to stop OST1"
2131 stop_mds || error "Unable to stop MDS"
2132 #writeconf to remove all ost2 traces for subsequent tests
2135 run_test 50f "normal statfs one server in down =========================="
2138 [ "$OSTCOUNT" -lt "2" ] && skip_env "$OSTCOUNT < 2, skipping" && return
2140 start_ost2 || error "Unable to start OST2"
2142 local PARAM="${FSNAME}-OST0001.osc.active"
2144 $LFS setstripe -c -1 $DIR/$tfile || error "Unable to lfs setstripe"
2145 do_facet mgs $LCTL conf_param $PARAM=0 || error "Unable to deactivate OST"
2147 umount_client $MOUNT || error "Unable to unmount client"
2148 mount_client $MOUNT || error "Unable to mount client"
2149 # This df should not cause a panic
2152 do_facet mgs $LCTL conf_param $PARAM=1 || error "Unable to activate OST"
2154 umount_client $MOUNT || error "Unable to unmount client"
2155 stop_ost2 || error "Unable to stop OST2"
2156 stop_ost || error "Unable to stop OST1"
2157 stop_mds || error "Unable to stop MDS"
2158 #writeconf to remove all ost2 traces for subsequent tests
2161 run_test 50g "deactivated OST should not cause panic====================="
2164 local LOCAL_TIMEOUT=20
2168 check_mount || return 1
2171 $LFS setstripe -c -1 $MOUNT/d1
2172 #define OBD_FAIL_MDS_REINT_DELAY 0x142
2173 do_facet $SINGLEMDS "lctl set_param fail_loc=0x142"
2174 touch $MOUNT/d1/f1 &
2177 start_ost2 || return 2
2179 stop_ost2 || return 3
2182 run_test 51 "Verify that mdt_reint handles RMF_MDT_MD correctly when an OST is added"
2191 do_node $node mkdir -p $dest
2192 [ $? -eq 0 ] || { error "Unable to create directory"; return 1; }
2194 do_node $node 'tar cf - '$@' | tar xf - -C '$dest';
2195 [ \"\${PIPESTATUS[*]}\" = \"0 0\" ] || exit 1'
2196 [ $? -eq 0 ] || { error "Unable to tar files"; return 2; }
2198 do_node $node 'getfattr -d -m "[a-z]*\\." '$@' > '$xattrs
2199 [ $? -eq 0 ] || { error "Unable to read xattrs"; return 3; }
2209 local backup2=${TMP}/backup2
2211 do_node $node mkdir -p $backup2
2212 [ $? -eq 0 ] || { error "Unable to create directory"; return 1; }
2214 do_node $node 'tar cf - '$@' | tar xf - -C '$backup2';
2215 [ \"\${PIPESTATUS[*]}\" = \"0 0\" ] || exit 1'
2216 [ $? -eq 0 ] || { error "Unable to tar files to diff"; return 2; }
2218 do_node $node "diff -rq $backup $backup2"
2219 [ $? -eq 0 ] || { error "contents differ"; return 3; }
2221 local xattrs2=${TMP}/xattrs2
2222 do_node $node 'getfattr -d -m "[a-z]*\\." '$@' > '$xattrs2
2223 [ $? -eq 0 ] || { error "Unable to read xattrs to diff"; return 4; }
2225 do_node $node "diff $xattrs $xattrs2"
2226 [ $? -eq 0 ] || { error "xattrs differ"; return 5; }
2228 do_node $node "rm -rf $backup2 $xattrs2"
2229 [ $? -eq 0 ] || { error "Unable to delete temporary files"; return 6; }
2234 [ $? -eq 0 ] || { error "Unable to start MDS"; return 1; }
2236 [ $? -eq 0 ] || { error "Unable to start OST1"; return 2; }
2238 [ $? -eq 0 ] || { error "Unable to mount client"; return 3; }
2241 local ost1mnt=${MOUNT%/*}/ost1
2242 local ost1node=$(facet_active_host ost1)
2243 local ost1tmp=$TMP/conf52
2246 [ $? -eq 0 ] || { error "Unable to create tdir"; return 4; }
2247 touch $TMP/modified_first
2248 [ $? -eq 0 ] || { error "Unable to create temporary file"; return 5; }
2249 do_node $ost1node "mkdir -p $ost1tmp && touch $ost1tmp/modified_first"
2250 [ $? -eq 0 ] || { error "Unable to create temporary file"; return 6; }
2253 $LFS setstripe $DIR/$tdir -c -1 -s 1M
2254 [ $? -eq 0 ] || { error "lfs setstripe failed"; return 7; }
2256 for (( i=0; i < nrfiles; i++ )); do
2257 multiop $DIR/$tdir/$tfile-$i Ow1048576w1048576w524288c
2258 [ $? -eq 0 ] || { error "multiop failed"; return 8; }
2264 echo backup files to $TMP/files
2265 local files=$(find $DIR/$tdir -type f -newer $TMP/modified_first)
2266 copy_files_xattrs `hostname` $TMP/files $TMP/file_xattrs $files
2267 [ $? -eq 0 ] || { error "Unable to copy files"; return 9; }
2269 umount_client $MOUNT
2270 [ $? -eq 0 ] || { error "Unable to umount client"; return 10; }
2272 [ $? -eq 0 ] || { error "Unable to stop ost1"; return 11; }
2274 echo mount ost1 as ldiskfs
2275 do_node $ost1node mount -t $FSTYPE $ost1_dev $ost1mnt $OST_MOUNT_OPTS
2276 [ $? -eq 0 ] || { error "Unable to mount ost1 as ldiskfs"; return 12; }
2279 echo backup objects to $ost1tmp/objects
2280 local objects=$(do_node $ost1node 'find '$ost1mnt'/O/0 -type f -size +0'\
2281 '-newer '$ost1tmp'/modified_first -regex ".*\/[0-9]+"')
2282 copy_files_xattrs $ost1node $ost1tmp/objects $ost1tmp/object_xattrs $objects
2283 [ $? -eq 0 ] || { error "Unable to copy objects"; return 13; }
2285 # move objects to lost+found
2286 do_node $ost1node 'mv '$objects' '${ost1mnt}'/lost+found'
2287 [ $? -eq 0 ] || { error "Unable to move objects"; return 14; }
2290 do_node $ost1node "ll_recover_lost_found_objs -d $ost1mnt/lost+found"
2291 [ $? -eq 0 ] || { error "ll_recover_lost_found_objs failed"; return 15; }
2293 # compare restored objects against saved ones
2294 diff_files_xattrs $ost1node $ost1tmp/objects $ost1tmp/object_xattrs $objects
2295 [ $? -eq 0 ] || { error "Unable to diff objects"; return 16; }
2297 do_node $ost1node "umount $ost1_dev"
2298 [ $? -eq 0 ] || { error "Unable to umount ost1 as ldiskfs"; return 17; }
2301 [ $? -eq 0 ] || { error "Unable to start ost1"; return 18; }
2303 [ $? -eq 0 ] || { error "Unable to mount client"; return 19; }
2306 diff_files_xattrs `hostname` $TMP/files $TMP/file_xattrs $files
2307 [ $? -eq 0 ] || { error "Unable to diff files"; return 20; }
2309 rm -rf $TMP/files $TMP/file_xattrs
2310 [ $? -eq 0 ] || { error "Unable to delete temporary files"; return 21; }
2311 do_node $ost1node "rm -rf $ost1tmp"
2312 [ $? -eq 0 ] || { error "Unable to delete temporary files"; return 22; }
2315 run_test 52 "check recovering objects from lost+found"
2317 # Checks threads_min/max/started for some service
2319 # Arguments: service name (OST or MDT), facet (e.g., ost1, $SINGLEMDS), and a
2320 # parameter pattern prefix like 'ost.*.ost'.
2332 local msg="Insane $modname thread counts"
2336 check_mount || return 41
2338 # We need to expand $parampat, but it may match multiple parameters, so
2339 # we'll pick the first one
2340 paramp=$(do_facet $facet "lctl get_param -N ${parampat}.threads_min"|head -1)
2341 if [ -z "$paramp" ]; then
2342 error "Couldn't expand ${parampat}.threads_min parameter name"
2346 # Remove the .threads_min part
2347 paramp=${paramp%.threads_min}
2349 # Check for sanity in defaults
2350 tmin=$(do_facet $facet "lctl get_param -n ${paramp}.threads_min" || echo 0)
2351 tmax=$(do_facet $facet "lctl get_param -n ${paramp}.threads_max" || echo 0)
2352 tstarted=$(do_facet $facet "lctl get_param -n ${paramp}.threads_started" || echo 0)
2353 lassert 23 "$msg (PDSH problems?)" '(($tstarted && $tmin && $tmax))' || return $?
2354 lassert 24 "$msg" '(($tstarted >= $tmin && $tstarted <= tmax ))' || return $?
2356 # Check that we can lower min/max
2357 do_facet $facet "lctl set_param ${paramp}.threads_min=$((tmin - 1))"
2358 do_facet $facet "lctl set_param ${paramp}.threads_max=$((tmax - 10))"
2359 tmin2=$(do_facet $facet "lctl get_param -n ${paramp}.threads_min" || echo 0)
2360 tmax2=$(do_facet $facet "lctl get_param -n ${paramp}.threads_max" || echo 0)
2361 lassert 25 "$msg" '(($tmin2 == ($tmin - 1) && $tmax2 == ($tmax -10)))' || return $?
2363 # Check that we can set min/max to the same value
2364 do_facet $facet "lctl set_param ${paramp}.threads_min=$tmin"
2365 do_facet $facet "lctl set_param ${paramp}.threads_max=$tmin"
2366 tmin2=$(do_facet $facet "lctl get_param -n ${paramp}.threads_min" || echo 0)
2367 tmax2=$(do_facet $facet "lctl get_param -n ${paramp}.threads_max" || echo 0)
2368 lassert 26 "$msg" '(($tmin2 == $tmin && $tmax2 == $tmin))' || return $?
2370 # Check that we can't set max < min
2371 do_facet $facet "lctl set_param ${paramp}.threads_max=$((tmin - 1))"
2372 tmin2=$(do_facet $facet "lctl get_param -n ${paramp}.threads_min" || echo 0)
2373 tmax2=$(do_facet $facet "lctl get_param -n ${paramp}.threads_max" || echo 0)
2374 lassert 27 "$msg" '(($tmin <= $tmax2))' || return $?
2376 # We need to ensure that we get the module options desired; to do this
2377 # we set LOAD_MODULES_REMOTE=true and we call setmodopts below.
2378 LOAD_MODULES_REMOTE=true
2381 setmodopts -a $modname "$opts" oldvalue
2385 check_mount || return 41
2387 # Restore previous setting of MODOPTS_*
2388 setmodopts $modname "$oldvalue"
2390 # Check that $opts took
2391 tmin=$(do_facet $facet "lctl get_param -n ${paramp}.threads_min")
2392 tmax=$(do_facet $facet "lctl get_param -n ${paramp}.threads_max")
2393 tstarted=$(do_facet $facet "lctl get_param -n ${paramp}.threads_started")
2394 lassert 28 "$msg" '(($tstarted == $tmin && $tstarted == $tmax ))' || return $?
2397 # Workaround a YALA bug where YALA expects that modules will remain
2398 # loaded on the servers
2399 LOAD_MODULES_REMOTE=false
2406 thread_sanity OST ost1 'ost.*.ost' 'oss_num_threads=64'
2408 run_test 53a "check OSS thread count params"
2411 thread_sanity MDT $SINGLEMDS 'mdt.*.*.' 'mdt_num_threads=64'
2413 run_test 53b "check MDT thread count params"
2418 local partial_arg=""
2419 local size=$(df -B G $dir | tail -1 | awk '{print $2}' | sed 's/G//') # Gb
2421 # Run in partial (fast) mode if the size
2422 # of a partition > 10 GB
2423 [ $size -gt 10 ] && partial_arg="-p"
2425 llverfs $partial_arg $dir
2429 do_rpc_nodes $(facet_host ost1) run_llverdev $(ostdevname 1)
2430 [ $? -eq 0 ] || error "llverdev failed!"
2433 run_test 54a "llverdev"
2438 [ $? -eq 0 ] || error "llverfs failed!"
2441 run_test 54b "llverfs"
2445 local max_ost_index=$1
2446 echo -n $(((max_ost_index + 1) * 8))
2450 local mdsdev=$(mdsdevname 1)
2451 local ostdev=$(ostdevname 1)
2452 local saved_opts=$OST_MKFS_OPTS
2454 for i in 0 1023 2048
2456 OST_MKFS_OPTS="$saved_opts --index $i"
2464 echo checking size of lov_objid for ost index $i
2465 LOV_OBJID_SIZE=$(do_facet mds1 "$DEBUGFS -R 'stat lov_objid' $mdsdev 2>/dev/null" | grep ^User | awk '{print $6}')
2466 if [ "$LOV_OBJID_SIZE" != $(lov_objid_size $i) ]; then
2467 error "lov_objid size has to be $(lov_objid_size $i), not $LOV_OBJID_SIZE"
2469 echo ok, lov_objid size is correct: $LOV_OBJID_SIZE
2474 OST_MKFS_OPTS=$saved_opts
2477 run_test 55 "check lov_objid size"
2480 add mds1 $MDS_MKFS_OPTS --mkfsoptions='\"-J size=16\"' --reformat $(mdsdevname 1)
2481 add ost1 $OST_MKFS_OPTS --index=1000 --reformat $(ostdevname 1)
2482 add ost2 $OST_MKFS_OPTS --index=10000 --reformat $(ostdevname 2)
2486 start_ost2 || error "Unable to start second ost"
2487 mount_client $MOUNT || error "Unable to mount client"
2493 run_test 56 "check big indexes"
2495 test_57() { # bug 22656
2496 local NID=$(do_facet ost1 "$LCTL get_param nis" | tail -1 | awk '{print $1}')
2498 do_facet ost1 "$TUNEFS --failnode=$NID `ostdevname 1`" || error "tunefs failed"
2500 start_ost && error "OST registration from failnode should fail"
2504 run_test 57 "initial registration from failnode should fail (should return errs)"
2507 do_facet mgs $LCTL get_param mgs.MGS.live.$FSNAME | grep OST | wc -l
2511 start_mgsmds >> /dev/null
2512 local C1=$(count_osts)
2513 if [ $C1 -eq 0 ]; then
2514 start_ost >> /dev/null
2518 echo "original ost count: $C1 (expect > 0)"
2519 [ $C1 -gt 0 ] || error "No OSTs in $FSNAME log"
2520 start_mgsmds -o writeconf >> /dev/null || error "MDT start failed"
2521 local C2=$(count_osts)
2522 echo "after mdt writeconf count: $C2 (expect 0)"
2523 [ $C2 -gt 0 ] && error "MDT writeconf should erase OST logs"
2524 echo "OST start without writeconf should fail:"
2525 start_ost >> /dev/null && error "OST start without writeconf didn't fail"
2526 echo "OST start with writeconf should succeed:"
2527 start_ost -o writeconf >> /dev/null || error "OST1 start failed"
2528 local C3=$(count_osts)
2529 echo "after ost writeconf count: $C3 (expect 1)"
2530 [ $C3 -eq 1 ] || error "new OST writeconf should add:"
2531 start_ost2 -o writeconf >> /dev/null || error "OST2 start failed"
2532 local C4=$(count_osts)
2533 echo "after ost2 writeconf count: $C4 (expect 2)"
2534 [ $C4 -eq 2 ] || error "OST2 writeconf should add log"
2535 stop_ost2 >> /dev/null
2536 cleanup_nocli >> /dev/null
2538 run_test 59 "writeconf mount option"
2541 if ! combined_mgs_mds ; then
2545 equals_msg `basename $0`: test complete
2546 [ -f "$TESTSUITELOG" ] && cat $TESTSUITELOG && grep -q FAIL $TESTSUITELOG && exit 1 || true