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 MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
430 if [ ! -e "$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 MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
464 if [ ! -e "$MDSDEV" ]; then
465 echo "no $MDSDEV existing, so mount Lustre to create one"
467 check_mount || return 41
471 echo "Remove mds config log"
472 do_facet $SINGLEMDS "$DEBUGFS -w -R 'unlink CONFIGS/$FSNAME-MDT0000' $MDSDEV || return \$?" || return $?
475 start_mds && return 42
478 run_test 17 "Verify failed mds_postsetup won't fail assertion (2936) (should return errs)"
481 [ "$FSTYPE" != "ldiskfs" ] && skip "not needed for FSTYPE=$FSTYPE" && return
483 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
488 # check if current MDSSIZE is large enough
489 [ $MDSSIZE -ge $MIN ] && OK=1 && myMDSSIZE=$MDSSIZE && \
490 log "use MDSSIZE=$MDSSIZE"
492 # check if the global config has a large enough MDSSIZE
493 [ -z "$OK" -a ! -z "$STORED_MDSSIZE" ] && [ $STORED_MDSSIZE -ge $MIN ] && \
494 OK=1 && myMDSSIZE=$STORED_MDSSIZE && \
495 log "use STORED_MDSSIZE=$STORED_MDSSIZE"
497 # check if the block device is large enough
498 [ -z "$OK" -a -b $MDSDEV ] && \
499 [ "$(dd if=$MDSDEV of=/dev/null bs=1k count=1 skip=$MIN 2>&1 |
500 awk '($3 == "in") { print $1 }')" = "1+0" ] && OK=1 && \
501 myMDSSIZE=$MIN && log "use device $MDSDEV with MIN=$MIN"
503 # check if a loopback device has enough space for fs metadata (5%)
504 [ -z "$OK" ] && [ -f $MDSDEV -o ! -e $MDSDEV ] &&
505 SPACE=$(df -P $(dirname $MDSDEV) |
506 awk '($1 != "Filesystem") {print $4}') &&
507 [ $SPACE -gt $((MIN / 20)) ] && OK=1 && myMDSSIZE=$MIN && \
508 log "use file $MDSDEV with MIN=$MIN"
510 [ -z "$OK" ] && skip_env "$MDSDEV too small for ${MIN}kB MDS" && return
513 echo "mount mds with large journal..."
514 local OLD_MDS_MKFS_OPTS=$MDS_MKFS_OPTS
516 MDS_MKFS_OPTS="--mgs --mdt --fsname=$FSNAME --device-size=$myMDSSIZE --param sys.timeout=$TIMEOUT $MDSOPT"
519 echo "mount lustre system..."
521 check_mount || return 41
523 echo "check journal size..."
524 local FOUNDSIZE=`do_facet mds "$DEBUGFS -c -R 'stat <8>' $MDSDEV" | awk '/Size: / { print $NF; exit;}'`
525 if [ $FOUNDSIZE -gt $((32 * 1024 * 1024)) ]; then
526 log "Success: mkfs creates large journals. Size: $((FOUNDSIZE >> 20))M"
528 error "expected journal size > 32M, found $((FOUNDSIZE >> 20))M"
533 MDS_MKFS_OPTS=$OLD_MDS_MKFS_OPTS
536 run_test 18 "check mkfs creates large journals"
539 start_mds || return 1
540 stop_mds -f || return 2
542 run_test 19a "start/stop MDS without OSTs"
545 start_ost || return 1
546 stop_ost -f || return 2
548 run_test 19b "start/stop OSTs without MDS"
551 # first format the ost/mdt
555 check_mount || return 43
557 remount_client ro $MOUNT || return 44
558 touch $DIR/$tfile && echo "$DIR/$tfile created incorrectly" && return 45
559 [ -e $DIR/$tfile ] && echo "$DIR/$tfile exists incorrectly" && return 46
560 remount_client rw $MOUNT || return 47
562 [ ! -f $DIR/$tfile ] && echo "$DIR/$tfile missing" && return 48
563 MCNT=`grep -c $MOUNT /etc/mtab`
564 [ "$MCNT" -ne 1 ] && echo "$MOUNT in /etc/mtab $MCNT times" && return 49
569 run_test 20 "remount ro,rw mounts work and doesn't break /etc/mtab"
574 wait_osc_import_state mds ost FULL
578 run_test 21a "start mds before ost, stop ost first"
583 wait_osc_import_state mds ost FULL
587 run_test 21b "start ost before mds, stop mds first"
593 wait_osc_import_state mds ost2 FULL
597 #writeconf to remove all ost2 traces for subsequent tests
600 run_test 21c "start mds between two osts, stop mds last"
605 echo Client mount with ost in logs, but none running
607 # wait until mds connected to ost and open client connection
608 wait_osc_import_state mds ost FULL
611 # check_mount will block trying to contact ost
612 mcreate $DIR/$tfile || return 40
613 rm -f $DIR/$tfile || return 42
617 echo Client mount with a running ost
620 # if gss enabled, wait full time to let connection from
621 # mds to ost be established, due to the mismatch between
622 # initial connect timeout and gss context negotiation timeout.
623 # This perhaps could be remove after AT landed.
624 echo "sleep $((TIMEOUT + TIMEOUT + TIMEOUT))s"
625 sleep $((TIMEOUT + TIMEOUT + TIMEOUT))
628 check_mount || return 41
633 run_test 22 "start a client before osts (should return errs)"
635 test_23a() { # was test_23
639 # force down client so that recovering mds waits for reconnect
640 local running=$(grep -c $MOUNT /proc/mounts) || true
641 if [ $running -ne 0 ]; then
642 echo "Stopping client $MOUNT (opts: -f)"
646 # enter recovery on mds
648 # try to start a new client
649 mount_client $MOUNT &
651 MOUNT_PID=$(ps -ef | grep "t lustre" | grep -v grep | awk '{print $2}')
652 MOUNT_LUSTRE_PID=`ps -ef | grep mount.lustre | grep -v grep | awk '{print $2}'`
653 echo mount pid is ${MOUNT_PID}, mount.lustre pid is ${MOUNT_LUSTRE_PID}
655 ps --ppid $MOUNT_LUSTRE_PID
656 # FIXME why o why can't I kill these? Manual "ctrl-c" works...
657 kill -TERM $MOUNT_LUSTRE_PID
658 echo "waiting for mount to finish"
660 # we can not wait $MOUNT_PID because it is not a child of this shell
666 while [ "$WAIT" -lt "$MAX_WAIT" ]; do
668 PID1=$(ps -ef | awk '{print $2}' | grep -w $MOUNT_PID)
669 PID2=$(ps -ef | awk '{print $2}' | grep -w $MOUNT_LUSTRE_PID)
672 [ -z "$PID1" -a -z "$PID2" ] && break
673 echo "waiting for mount to finish ... "
674 WAIT=$(( WAIT + sleep))
676 if [ "$WAIT" -eq "$MAX_WAIT" ]; then
677 error "MOUNT_PID $MOUNT_PID and "\
678 "MOUNT_LUSTRE_PID $MOUNT_LUSTRE_PID still not killed in $WAIT secs"
684 run_test 23a "interrupt client during recovery mount delay"
689 test_23b() { # was test_23
692 # Simulate -EINTR during mount OBD_FAIL_LDLM_CLOSE_THREAD
693 lctl set_param fail_loc=0x80000313
697 run_test 23b "Simulate -EINTR during mount"
699 fs2mds_HOST=$mds_HOST
700 fs2ost_HOST=$ost_HOST
704 echo "umount $MOUNT2 ..."
705 umount $MOUNT2 || true
706 echo "stopping fs2mds ..."
707 stop fs2mds -f || true
708 echo "stopping fs2ost ..."
709 stop fs2ost -f || true
717 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
719 [ -n "$ost1_HOST" ] && fs2ost_HOST=$ost1_HOST
720 if [ -z "$fs2ost_DEV" -o -z "$fs2mds_DEV" ]; then
721 do_facet $SINGLEMDS [ -b "$MDSDEV" ] && \
722 skip_env "mixed loopback and real device not working" && return
725 local fs2mdsdev=${fs2mds_DEV:-${MDSDEV}_2}
726 local fs2ostdev=${fs2ost_DEV:-$(ostdevname 1)_2}
728 # test 8-char fsname as well
729 local FSNAME2=test1234
730 add fs2mds $MDS_MKFS_OPTS --fsname=${FSNAME2} --nomgs --mgsnode=$MGSNID --reformat $fs2mdsdev || exit 10
732 add fs2ost $OST_MKFS_OPTS --fsname=${FSNAME2} --reformat $fs2ostdev || exit 10
735 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS && trap cleanup_24a EXIT INT
736 start fs2ost $fs2ostdev $OST_MOUNT_OPTS
738 mount -t lustre $MGSNID:/${FSNAME2} $MOUNT2 || return 1
740 check_mount || return 2
741 # files written on 1 should not show up on 2
742 cp /etc/passwd $DIR/$tfile
744 [ -e $MOUNT2/$tfile ] && error "File bleed" && return 7
747 cp /etc/passwd $MOUNT2/b || return 3
748 rm $MOUNT2/b || return 4
749 # 2 is actually mounted
750 grep $MOUNT2' ' /proc/mounts > /dev/null || return 5
752 facet_failover fs2mds
753 facet_failover fs2ost
756 # the MDS must remain up until last MDT
758 MDS=$(do_facet $SINGLEMDS "lctl get_param -n devices" | awk '($3 ~ "mdt" && $4 ~ "MDT") { print $4 }' | head -1)
759 [ -z "$MDS" ] && error "No MDT" && return 8
761 cleanup_nocli || return 6
763 run_test 24a "Multiple MDTs on a single node"
766 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
768 if [ -z "$fs2mds_DEV" ]; then
769 do_facet $SINGLEMDS [ -b "$MDSDEV" ] && \
770 skip_env "mixed loopback and real device not working" && return
773 local fs2mdsdev=${fs2mds_DEV:-${MDSDEV}_2}
775 add fs2mds $MDS_MKFS_OPTS --fsname=${FSNAME}2 --mgs --reformat $fs2mdsdev || exit 10
777 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS && return 2
780 run_test 24b "Multiple MGSs on a single node (should return err)"
784 check_mount || return 2
785 local MODULES=$($LCTL modules | awk '{ print $2 }')
786 rmmod $MODULES 2>/dev/null || true
789 run_test 25 "Verify modules are referenced"
793 # we need modules before mount for sysctl, so make sure...
794 do_facet $SINGLEMDS "lsmod | grep -q lustre || modprobe lustre"
795 #define OBD_FAIL_MDS_FS_SETUP 0x135
796 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000135"
797 start_mds && echo MDS started && return 1
798 lctl get_param -n devices
799 DEVS=$(lctl get_param -n devices | wc -l)
800 [ $DEVS -gt 0 ] && return 2
801 unload_modules || return 203
803 run_test 26 "MDT startup failure cleans LOV (should return errs)"
809 local ORIG=$(do_facet $myfacet "$TEST")
810 if [ $# -gt 3 ]; then
816 echo "Setting $PARAM from $ORIG to $FINAL"
817 do_facet $SINGLEMDS "$LCTL conf_param $PARAM='$FINAL'" || error conf_param failed
819 wait_update $(facet_host $myfacet) "$TEST" "$FINAL" || error check failed!
823 start_ost || return 1
824 start_mds || return 2
825 echo "Requeue thread should have started: "
826 ps -e | grep ll_cfg_requeue
827 set_and_check ost1 "lctl get_param -n obdfilter.$FSNAME-OST0000.client_cache_seconds" "$FSNAME-OST0000.ost.client_cache_seconds" || return 3
830 run_test 27a "Reacquire MGS lock if OST started first"
835 local device=$(do_facet $SINGLEMDS "lctl get_param -n devices" | awk '($3 ~ "mdt" && $4 ~ "MDT") { print $4 }')
837 facet_failover $SINGLEMDS
838 set_and_check $SINGLEMDS "lctl get_param -n mdt.$device.identity_acquire_expire" "$device.mdt.identity_acquire_expire" || return 3
839 set_and_check client "lctl get_param -n mdc.$device-mdc-*.max_rpcs_in_flight" "$device.mdc.max_rpcs_in_flight" || return 4
843 run_test 27b "Reacquire MGS lock after failover"
847 TEST="lctl get_param -n llite.$FSNAME-*.max_read_ahead_whole_mb"
848 PARAM="$FSNAME.llite.max_read_ahead_whole_mb"
851 set_and_check client "$TEST" "$PARAM" $FINAL || return 3
852 FINAL=$(($FINAL + 1))
853 set_and_check client "$TEST" "$PARAM" $FINAL || return 4
854 umount_client $MOUNT || return 200
857 if [ $RESULT -ne $FINAL ]; then
858 echo "New config not seen: wanted $FINAL got $RESULT"
861 echo "New config success: got $RESULT"
863 set_and_check client "$TEST" "$PARAM" $ORIG || return 5
866 run_test 28 "permanent parameter setting"
869 [ "$OSTCOUNT" -lt "2" ] && skip_env "$OSTCOUNT < 2, skipping" && return
870 setup > /dev/null 2>&1
874 local PARAM="$FSNAME-OST0001.osc.active"
875 local PROC_ACT="osc.$FSNAME-OST0001-osc-[^M]*.active"
876 local PROC_UUID="osc.$FSNAME-OST0001-osc-[^M]*.ost_server_uuid"
878 ACTV=$(lctl get_param -n $PROC_ACT)
880 set_and_check client "lctl get_param -n $PROC_ACT" "$PARAM" $DEAC || return 2
881 # also check ost_server_uuid status
882 RESULT=$(lctl get_param -n $PROC_UUID | grep DEACTIV)
883 if [ -z "$RESULT" ]; then
884 echo "Live client not deactivated: $(lctl get_param -n $PROC_UUID)"
887 echo "Live client success: got $RESULT"
891 local MPROC="osc.$FSNAME-OST0001-osc-[M]*.active"
896 RESULT=`do_facet $SINGLEMDS " lctl get_param -n $MPROC"`
897 [ ${PIPESTATUS[0]} = 0 ] || error "Can't read $MPROC"
898 if [ $RESULT -eq $DEAC ]; then
899 echo "MDT deactivated also after $WAIT sec (got $RESULT)"
903 if [ $WAIT -eq $MAX ]; then
904 echo "MDT not deactivated: wanted $DEAC got $RESULT"
907 echo "Waiting $(($MAX - $WAIT)) secs for MDT deactivated"
910 # test new client starts deactivated
911 umount_client $MOUNT || return 200
913 RESULT=$(lctl get_param -n $PROC_UUID | grep DEACTIV | grep NEW)
914 if [ -z "$RESULT" ]; then
915 echo "New client not deactivated from start: $(lctl get_param -n $PROC_UUID)"
918 echo "New client success: got $RESULT"
921 # make sure it reactivates
922 set_and_check client "lctl get_param -n $PROC_ACT" "$PARAM" $ACTV || return 6
927 #writeconf to remove all ost2 traces for subsequent tests
930 run_test 29 "permanently remove an OST"
936 TEST="lctl get_param -n llite.$FSNAME-*.max_read_ahead_whole_mb"
938 LIST=(1 2 3 4 5 4 3 2 1 2 3 4 5 4 3 2 1 2 3 4 5)
939 for i in ${LIST[@]}; do
940 set_and_check client "$TEST" "$FSNAME.llite.max_read_ahead_whole_mb" $i || return 3
942 # make sure client restart still works
944 mount_client $MOUNT || return 4
945 [ "$($TEST)" -ne "$i" ] && error "Param didn't stick across restart $($TEST) != $i"
948 echo Erase parameter setting
949 do_facet mgs "$LCTL conf_param -d $FSNAME.llite.max_read_ahead_whole_mb" || return 6
951 mount_client $MOUNT || return 6
953 echo "deleted (default) value=$FINAL, orig=$ORIG"
954 # assumes this parameter started at the default value
955 [ "$FINAL" -eq "$ORIG" ] || fail "Deleted value=$FINAL, orig=$ORIG"
959 run_test 30a "Big config llog and conf_param deletion"
964 # Make a fake nid. Use the OST nid, and add 20 to the least significant
965 # numerical part of it. Hopefully that's not already a failover address for
967 OSTNID=$(do_facet ost1 "$LCTL get_param nis" | tail -1 | awk '{print $1}')
968 ORIGVAL=$(echo $OSTNID | egrep -oi "[0-9]*@")
969 NEWVAL=$((($(echo $ORIGVAL | egrep -oi "[0-9]*") + 20) % 256))
970 NEW=$(echo $OSTNID | sed "s/$ORIGVAL/$NEWVAL@/")
971 echo "Using fake nid $NEW"
973 TEST="$LCTL get_param -n osc.$FSNAME-OST0000-osc-[^M]*.import | grep failover_nids | sed -n 's/.*\($NEW\).*/\1/p'"
974 set_and_check client "$TEST" "$FSNAME-OST0000.failover.node" $NEW || error "didn't add failover nid $NEW"
975 NIDS=$($LCTL get_param -n osc.$FSNAME-OST0000-osc-[^M]*.import | grep failover_nids)
977 NIDCOUNT=$(($(echo "$NIDS" | wc -w) - 1))
978 echo "should have 2 failover nids: $NIDCOUNT"
979 [ $NIDCOUNT -eq 2 ] || error "Failover nid not added"
980 do_facet mgs "$LCTL conf_param -d $FSNAME-OST0000.failover.node" || error "conf_param delete failed"
982 mount_client $MOUNT || return 3
984 NIDS=$($LCTL get_param -n osc.$FSNAME-OST0000-osc-[^M]*.import | grep failover_nids)
986 NIDCOUNT=$(($(echo "$NIDS" | wc -w) - 1))
987 echo "only 1 final nid should remain: $NIDCOUNT"
988 [ $NIDCOUNT -eq 1 ] || error "Failover nids not removed"
992 run_test 30b "Remove failover nids"
994 test_31() { # bug 10734
995 # ipaddr must not exist
996 mount -t lustre 4.3.2.1@tcp:/lustre $MOUNT || true
999 run_test 31 "Connect to non-existent node (shouldn't crash)"
1001 # Use these start32/stop32 fn instead of t-f start/stop fn,
1002 # for local devices, to skip global facet vars init
1006 echo "Stopping local ${MOUNT%/*}/${facet} (opts:$@)"
1007 umount -d $@ ${MOUNT%/*}/${facet}
1016 mkdir -p ${MOUNT%/*}/${facet}
1018 echo "Starting local ${facet}: $@ $device ${MOUNT%/*}/${facet}"
1019 mount -t lustre $@ ${device} ${MOUNT%/*}/${facet}
1021 if [ $RC -ne 0 ]; then
1022 echo "mount -t lustre $@ ${device} ${MOUNT%/*}/${facet}"
1023 echo "Start of ${device} of local ${facet} failed ${RC}"
1029 cleanup_nocli32 () {
1037 echo "Cleanup test_32 umount $MOUNT ..."
1038 umount -f $MOUNT || true
1039 echo "Cleanup local mds ost1 ..."
1045 client_only && skip "client only testing" && return 0
1046 [ "$NETTYPE" = "tcp" ] || { skip "NETTYPE != tcp" && return 0; }
1047 [ -z "$TUNEFS" ] && skip_env "No tunefs" && return 0
1049 local DISK1_8=$LUSTRE/tests/disk1_8.tar.bz2
1050 [ ! -r $DISK1_8 ] && skip_env "Cannot find $DISK1_8" && return 0
1051 local tmpdir=$TMP/conf32a
1053 tar xjvf $DISK1_8 -C $tmpdir || \
1054 { skip_env "Cannot untar $DISK1_8" && return 0; }
1057 $LCTL set_param debug=$PTLDEBUG
1059 $TUNEFS $tmpdir/mds || error "tunefs failed"
1061 # nids are wrong, so client wont work, but server should start
1062 start32 mds1 $tmpdir/mds "-o loop,exclude=lustre-OST0000" && \
1063 trap cleanup_32 EXIT INT || return 3
1065 local UUID=$($LCTL get_param -n mdt.lustre-MDT0000.uuid)
1067 [ "$UUID" == "lustre-MDT0000_UUID" ] || error "UUID is wrong: $UUID"
1069 $TUNEFS --mgsnode=$HOSTNAME $tmpdir/ost1 || error "tunefs failed"
1070 start32 ost1 $tmpdir/ost1 "-o loop" || return 5
1071 UUID=$($LCTL get_param -n obdfilter.lustre-OST0000.uuid)
1073 [ "$UUID" == "lustre-OST0000_UUID" ] || error "UUID is wrong: $UUID"
1075 local NID=$($LCTL list_nids | head -1)
1077 echo "OSC changes should succeed:"
1078 $LCTL conf_param lustre-OST0000.osc.max_dirty_mb=15 || return 7
1079 $LCTL conf_param lustre-OST0000.failover.node=$NID || return 8
1082 echo "MDC changes should succeed:"
1083 $LCTL conf_param lustre-MDT0000.mdc.max_rpcs_in_flight=9 || return 9
1084 $LCTL conf_param lustre-MDT0000.failover.node=$NID || return 10
1087 echo "LOV changes should succeed:"
1088 $LCTL pool_new lustre.interop || return 11
1089 $LCTL conf_param lustre-MDT0000.lov.stripesize=4M || return 12
1094 # mount a second time to make sure we didnt leave upgrade flag on
1096 $TUNEFS --dryrun $tmpdir/mds || error "tunefs failed"
1097 start32 mds1 $tmpdir/mds "-o loop,exclude=lustre-OST0000" && \
1098 trap cleanup_32 EXIT INT || return 12
1102 rm -rf $tmpdir || true # true is only for TMP on NFS
1104 run_test 32a "Upgrade from 1.8 (not live)"
1107 client_only && skip "client only testing" && return 0
1108 [ "$NETTYPE" = "tcp" ] || { skip "NETTYPE != tcp" && return 0; }
1109 [ -z "$TUNEFS" ] && skip_env "No tunefs" && return
1111 local DISK1_8=$LUSTRE/tests/disk1_8.tar.bz2
1112 [ ! -r $DISK1_8 ] && skip_env "Cannot find $DISK1_8" && return 0
1113 local tmpdir=$TMP/conf32b
1115 tar xjvf $DISK1_8 -C $tmpdir || \
1116 { skip_env "Cannot untar $DISK1_8" && return ; }
1119 $LCTL set_param debug="config"
1120 local NEWNAME=lustre
1122 # writeconf will cause servers to register with their current nids
1123 $TUNEFS --writeconf --fsname=$NEWNAME $tmpdir/mds || error "tunefs failed"
1124 start32 mds1 $tmpdir/mds "-o loop" && \
1125 trap cleanup_32 EXIT INT || return 3
1127 local UUID=$($LCTL get_param -n mdt.${NEWNAME}-MDT0000.uuid)
1129 [ "$UUID" == "${NEWNAME}-MDT0000_UUID" ] || error "UUID is wrong: $UUID"
1131 $TUNEFS --mgsnode=$HOSTNAME --writeconf --fsname=$NEWNAME $tmpdir/ost1 ||\
1132 error "tunefs failed"
1133 start32 ost1 $tmpdir/ost1 "-o loop" || return 5
1134 UUID=$($LCTL get_param -n obdfilter.${NEWNAME}-OST0000.uuid)
1136 [ "$UUID" == "${NEWNAME}-OST0000_UUID" ] || error "UUID is wrong: $UUID"
1138 local NID=$($LCTL list_nids | head -1)
1140 echo "OSC changes should succeed:"
1141 $LCTL conf_param ${NEWNAME}-OST0000.osc.max_dirty_mb=15 || return 7
1142 $LCTL conf_param ${NEWNAME}-OST0000.failover.node=$NID || return 8
1145 echo "MDC changes should succeed:"
1146 $LCTL conf_param ${NEWNAME}-MDT0000.mdc.max_rpcs_in_flight=9 || return 9
1147 $LCTL conf_param ${NEWNAME}-MDT0000.failover.node=$NID || return 10
1150 echo "LOV changes should succeed:"
1151 $LCTL pool_new ${NEWNAME}.interop || return 11
1152 $LCTL conf_param ${NEWNAME}-MDT0000.lov.stripesize=4M || return 12
1155 # MDT and OST should have registered with new nids, so we should have
1156 # a fully-functioning client
1157 echo "Check client and old fs contents"
1159 local device=`h2$NETTYPE $HOSTNAME`:/$NEWNAME
1160 echo "Starting local client: $HOSTNAME: $device $MOUNT"
1161 mount -t lustre $device $MOUNT || return 1
1163 local old=$($LCTL get_param -n mdc.*.max_rpcs_in_flight)
1164 local new=$((old + 5))
1165 $LCTL conf_param ${NEWNAME}-MDT0000.mdc.max_rpcs_in_flight=$new
1166 wait_update $HOSTNAME "$LCTL get_param -n mdc.*.max_rpcs_in_flight" $new || return 11
1168 [ "$(cksum $MOUNT/passwd | cut -d' ' -f 1,2)" == "94306271 1478" ] || return 12
1173 rm -rf $tmpdir || true # true is only for TMP on NFS
1175 run_test 32b "Upgrade from 1.8 with writeconf"
1177 test_33a() { # bug 12333, was test_33
1179 local FSNAME2=test-123
1180 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
1182 [ -n "$ost1_HOST" ] && fs2ost_HOST=$ost1_HOST
1184 if [ -z "$fs2ost_DEV" -o -z "$fs2mds_DEV" ]; then
1185 do_facet $SINGLEMDS [ -b "$MDSDEV" ] && \
1186 skip_env "mixed loopback and real device not working" && return
1189 local fs2mdsdev=${fs2mds_DEV:-${MDSDEV}_2}
1190 local fs2ostdev=${fs2ost_DEV:-$(ostdevname 1)_2}
1191 add fs2mds $MDS_MKFS_OPTS --mkfsoptions='\"-J size=8\"' --fsname=${FSNAME2} --reformat $fs2mdsdev || exit 10
1192 add fs2ost $OST_MKFS_OPTS --fsname=${FSNAME2} --index=8191 --mgsnode=$MGSNID --reformat $fs2ostdev || exit 10
1194 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS && trap cleanup_24a EXIT INT
1195 start fs2ost $fs2ostdev $OST_MOUNT_OPTS
1196 do_facet $SINGLEMDS "$LCTL conf_param $FSNAME2.sys.timeout=200" || rc=1
1198 mount -t lustre $MGSNID:/${FSNAME2} $MOUNT2 || rc=2
1201 cp /etc/hosts $MOUNT2/ || rc=3
1202 $LFS getstripe $MOUNT2/hosts
1207 rm -rf $MOUNT2 $fs2mdsdev $fs2ostdev
1208 cleanup_nocli || rc=6
1211 run_test 33a "Mount ost with a large index number"
1213 test_33b() { # was test_34
1216 do_facet client dd if=/dev/zero of=$MOUNT/24 bs=1024k count=1
1217 # Drop lock cancelation reply during umount
1218 #define OBD_FAIL_LDLM_CANCEL 0x304
1219 do_facet client lctl set_param fail_loc=0x80000304
1220 #lctl set_param debug=-1
1221 umount_client $MOUNT
1224 run_test 33b "Drop cancel during umount"
1228 do_facet client "sh runmultiop_bg_pause $DIR/file O_c"
1229 manual_umount_client
1231 do_facet client killall -USR1 multiop
1232 if [ $rc -eq 0 ]; then
1233 error "umount not fail!"
1238 run_test 34a "umount with opened file should be fail"
1243 touch $DIR/$tfile || return 1
1244 stop_mds --force || return 2
1246 manual_umount_client --force
1248 if [ $rc -ne 0 ]; then
1249 error "mtab after failed umount - rc $rc"
1255 run_test 34b "force umount with failed mds should be normal"
1259 touch $DIR/$tfile || return 1
1260 stop_ost --force || return 2
1262 manual_umount_client --force
1264 if [ $rc -ne 0 ]; then
1265 error "mtab after failed umount - rc $rc"
1271 run_test 34c "force umount with failed ost should be normal"
1273 test_35a() { # bug 12459
1276 DBG_SAVE="`lctl get_param -n debug`"
1277 lctl set_param debug="ha"
1279 log "Set up a fake failnode for the MDS"
1281 local device=$(do_facet $SINGLEMDS "lctl get_param -n devices" | awk '($3 ~ "mdt" && $4 ~ "MDT") { print $4 }' | head -1)
1282 do_facet $SINGLEMDS $LCTL conf_param ${device}.failover.node=$FAKENID || return 4
1284 log "Wait for RECONNECT_INTERVAL seconds (10s)"
1287 MSG="conf-sanity.sh test_35a `date +%F%kh%Mm%Ss`"
1290 log "Stopping the MDT:"
1291 stop_mds || return 5
1293 df $MOUNT > /dev/null 2>&1 &
1295 log "Restarting the MDT:"
1296 start_mds || return 6
1297 log "Wait for df ($DFPID) ... "
1300 lctl set_param debug="$DBG_SAVE"
1302 # retrieve from the log the first server that the client tried to
1303 # contact after the connection loss
1304 $LCTL dk $TMP/lustre-log-$TESTNAME.log
1305 NEXTCONN=`awk "/${MSG}/ {start = 1;}
1306 /import_select_connection.*$device-mdc.* using connection/ {
1308 if (\\\$NF ~ /$FAKENID/)
1314 }" $TMP/lustre-log-$TESTNAME.log`
1315 [ "$NEXTCONN" != "0" ] && log "The client didn't try to reconnect to the last active server (tried ${NEXTCONN} instead)" && return 7
1318 run_test 35a "Reconnect to the last active server first"
1320 test_35b() { # bug 18674
1321 remote_mds || { skip "local MDS" && return 0; }
1325 $LCTL set_param debug="ha"
1327 MSG="conf-sanity.sh test_35b `date +%F%kh%Mm%Ss`"
1330 log "Set up a fake failnode for the MDS"
1332 local device=$(do_facet mds "$LCTL get_param -n devices" | \
1333 awk '($3 ~ "mdt" && $4 ~ "MDT") { print $4 }' | head -1)
1334 do_facet mds "$LCTL conf_param ${device}.failover.node=$FAKENID" || \
1337 local at_max_saved=0
1338 # adaptive timeouts may prevent seeing the issue
1339 if at_is_enabled; then
1340 at_max_saved=$(at_max_get mds)
1341 at_max_set 0 mds client
1344 mkdir -p $MOUNT/testdir
1345 touch $MOUNT/testdir/test
1347 log "Injecting EBUSY on MDS"
1348 # Setting OBD_FAIL_MDS_RESEND=0x136
1349 do_facet mds "$LCTL set_param fail_loc=0x80000136" || return 2
1351 log "Stat on a test file"
1352 stat $MOUNT/testdir/test
1354 log "Stop injecting EBUSY on MDS"
1355 do_facet mds "$LCTL set_param fail_loc=0" || return 3
1356 rm -f $MOUNT/testdir/test
1359 # restore adaptive timeout
1360 [ $at_max_saved -ne 0 ] && at_max_set $at_max_saved mds client
1362 $LCTL dk $TMP/lustre-log-$TESTNAME.log
1364 # retrieve from the log if the client has ever tried to
1365 # contact the fake server after the loss of connection
1366 FAILCONN=`awk "BEGIN {ret = 0;}
1367 /import_select_connection.*${FSNAME}-MDT0000-mdc.* using connection/ {
1369 if (\\\$NF ~ /$FAKENID/) {
1374 END {print ret}" $TMP/lustre-log-$TESTNAME.log`
1376 [ "$FAILCONN" == "0" ] && \
1377 log "ERROR: The client reconnection has not been triggered" && \
1379 [ "$FAILCONN" == "2" ] && \
1380 log "ERROR: The client tried to reconnect to the failover server while the primary was busy" && \
1385 run_test 35b "Continue reconnection retries, if the active server is busy"
1389 local FSNAME2=test1234
1390 local fs3ost_HOST=$ost_HOST
1391 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
1393 [ -n "$ost1_HOST" ] && fs2ost_HOST=$ost1_HOST && fs3ost_HOST=$ost1_HOST
1396 if [ -z "$fs2ost_DEV" -o -z "$fs2mds_DEV" -o -z "$fs3ost_DEV" ]; then
1397 do_facet $SINGLEMDS [ -b "$MDSDEV" ] && \
1398 skip_env "mixed loopback and real device not working" && return
1400 [ $OSTCOUNT -lt 2 ] && skip_env "skipping test for single OST" && return
1402 [ "$ost_HOST" = "`hostname`" -o "$ost1_HOST" = "`hostname`" ] || \
1403 { skip "remote OST" && return 0; }
1405 local fs2mdsdev=${fs2mds_DEV:-${MDSDEV}_2}
1406 local fs2ostdev=${fs2ost_DEV:-$(ostdevname 1)_2}
1407 local fs3ostdev=${fs3ost_DEV:-$(ostdevname 2)_2}
1408 add fs2mds $MDS_MKFS_OPTS --fsname=${FSNAME2} --reformat $fs2mdsdev || exit 10
1409 # XXX after we support non 4K disk blocksize, change following --mkfsoptions with
1411 add fs2ost $OST_MKFS_OPTS --mkfsoptions='-b4096' --fsname=${FSNAME2} --mgsnode=$MGSNID --reformat $fs2ostdev || exit 10
1412 add fs3ost $OST_MKFS_OPTS --mkfsoptions='-b4096' --fsname=${FSNAME2} --mgsnode=$MGSNID --reformat $fs3ostdev || exit 10
1414 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS
1415 start fs2ost $fs2ostdev $OST_MOUNT_OPTS
1416 start fs3ost $fs3ostdev $OST_MOUNT_OPTS
1418 mount -t lustre $MGSNID:/${FSNAME2} $MOUNT2 || return 1
1420 sleep 5 # until 11778 fixed
1422 dd if=/dev/zero of=$MOUNT2/$tfile bs=1M count=7 || return 2
1424 BKTOTAL=`lctl get_param -n obdfilter.*.kbytestotal | awk 'BEGIN{total=0}; {total+=$1}; END{print total}'`
1425 BKFREE=`lctl get_param -n obdfilter.*.kbytesfree | awk 'BEGIN{free=0}; {free+=$1}; END{print free}'`
1426 BKAVAIL=`lctl get_param -n obdfilter.*.kbytesavail | awk 'BEGIN{avail=0}; {avail+=$1}; END{print avail}'`
1427 STRING=`df -P $MOUNT2 | tail -n 1 | awk '{print $2","$3","$4}'`
1428 DFTOTAL=`echo $STRING | cut -d, -f1`
1429 DFUSED=`echo $STRING | cut -d, -f2`
1430 DFAVAIL=`echo $STRING | cut -d, -f3`
1431 DFFREE=$(($DFTOTAL - $DFUSED))
1433 ALLOWANCE=$((64 * $OSTCOUNT))
1435 if [ $DFTOTAL -lt $(($BKTOTAL - $ALLOWANCE)) ] ||
1436 [ $DFTOTAL -gt $(($BKTOTAL + $ALLOWANCE)) ] ; then
1437 echo "**** FAIL: df total($DFTOTAL) mismatch OST total($BKTOTAL)"
1440 if [ $DFFREE -lt $(($BKFREE - $ALLOWANCE)) ] ||
1441 [ $DFFREE -gt $(($BKFREE + $ALLOWANCE)) ] ; then
1442 echo "**** FAIL: df free($DFFREE) mismatch OST free($BKFREE)"
1445 if [ $DFAVAIL -lt $(($BKAVAIL - $ALLOWANCE)) ] ||
1446 [ $DFAVAIL -gt $(($BKAVAIL + $ALLOWANCE)) ] ; then
1447 echo "**** FAIL: df avail($DFAVAIL) mismatch OST avail($BKAVAIL)"
1452 stop fs3ost -f || return 200
1453 stop fs2ost -f || return 201
1454 stop fs2mds -f || return 202
1455 rm -rf $MOUNT2 $fs2mdsdev $fs2ostdev $fs3ostdev
1456 unload_modules || return 203
1459 run_test 36 "df report consistency on OSTs with different block size"
1462 client_only && skip "client only testing" && return 0
1463 LOCAL_MDSDEV="$TMP/mdt.img"
1464 SYM_MDSDEV="$TMP/sym_mdt.img"
1466 echo "MDS : $LOCAL_MDSDEV"
1467 echo "SYMLINK : $SYM_MDSDEV"
1471 mkfs.lustre --reformat --fsname=lustre --mdt --mgs --device-size=9000 $LOCAL_MDSDEV ||
1472 error "mkfs.lustre $LOCAL_MDSDEV failed"
1473 ln -s $LOCAL_MDSDEV $SYM_MDSDEV
1475 echo "mount symlink device - $SYM_MDSDEV"
1477 mount_op=`mount -v -t lustre -o loop $SYM_MDSDEV ${MOUNT%/*}/mds 2>&1 | grep "unable to set tunable"`
1478 umount -d ${MOUNT%/*}/mds
1479 rm -f $LOCAL_MDSDEV $SYM_MDSDEV
1481 if [ -n "$mount_op" ]; then
1482 error "**** FAIL: set tunables failed for symlink device"
1486 run_test 37 "verify set tunables works for symlink device"
1488 test_38() { # bug 14222
1493 FILES=`find $SRC -type f -mtime +1 | head -n $COUNT`
1494 log "copying $(echo $FILES | wc -w) files to $DIR/$tdir"
1496 tar cf - $FILES | tar xf - -C $DIR/$tdir || \
1497 error "copying $SRC to $DIR/$tdir"
1499 umount_client $MOUNT
1501 log "rename lov_objid file on MDS"
1502 rm -f $TMP/lov_objid.orig
1504 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
1505 do_facet $SINGLEMDS "$DEBUGFS -c -R \\\"dump lov_objid $TMP/lov_objid.orig\\\" $MDSDEV"
1506 do_facet $SINGLEMDS "$DEBUGFS -w -R \\\"rm lov_objid\\\" $MDSDEV"
1508 do_facet $SINGLEMDS "od -Ax -td8 $TMP/lov_objid.orig"
1509 # check create in mds_lov_connect
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.new\\\" $MDSDEV"
1517 do_facet $SINGLEMDS "od -Ax -td8 $TMP/lov_objid.new"
1518 [ "$ERROR" = "y" ] && error "old and new files are different after connect" || true
1520 # check it's updates in sync
1521 umount_client $MOUNT
1524 do_facet $SINGLEMDS dd if=/dev/zero of=$TMP/lov_objid.clear bs=4096 count=1
1525 do_facet $SINGLEMDS "$DEBUGFS -w -R \\\"rm lov_objid\\\" $MDSDEV"
1526 do_facet $SINGLEMDS "$DEBUGFS -w -R \\\"write $TMP/lov_objid.clear lov_objid\\\" $MDSDEV "
1531 [ $V ] && log "verifying $DIR/$tdir/$f"
1532 diff -q $f $DIR/$tdir/$f || ERROR=y
1534 do_facet $SINGLEMDS "$DEBUGFS -c -R \\\"dump lov_objid $TMP/lov_objid.new1\\\" $MDSDEV"
1535 do_facet $SINGLEMDS "od -Ax -td8 $TMP/lov_objid.new1"
1536 umount_client $MOUNT
1538 [ "$ERROR" = "y" ] && error "old and new files are different after sync" || true
1540 log "files compared the same"
1543 run_test 38 "MDS recreates missing lov_objid file from OST data"
1549 perl $SRCDIR/leak_finder.pl $TMP/debug 2>&1 | egrep '*** Leak:' &&
1550 error "memory leak detected" || true
1552 run_test 39 "leak_finder recognizes both LUSTRE and LNET malloc messages"
1554 test_40() { # bug 15759
1556 #define OBD_FAIL_TGT_TOOMANY_THREADS 0x706
1557 do_facet $SINGLEMDS "sysctl -w lustre.fail_loc=0x80000706"
1561 run_test 40 "race during service thread startup"
1563 test_41() { #bug 14134
1565 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
1567 start $SINGLEMDS $MDSDEV $MDS_MOUNT_OPTS -o nosvc -n
1568 start ost1 `ostdevname 1` $OST_MOUNT_OPTS
1569 start $SINGLEMDS $MDSDEV $MDS_MOUNT_OPTS -o nomgs,force
1571 mount_client $MOUNT || return 1
1574 echo "blah blah" > $MOUNT/$tfile
1577 umount_client $MOUNT
1578 stop ost1 -f || return 201
1579 stop_mds -f || return 202
1580 stop_mds -f || return 203
1581 unload_modules || return 204
1584 run_test 41 "mount mds with --nosvc and --nomgs"
1586 test_42() { #bug 14693
1588 check_mount || return 2
1589 do_facet mgs $LCTL conf_param lustre.llite.some_wrong_param=10
1590 umount_client $MOUNT
1591 mount_client $MOUNT || return 1
1595 run_test 42 "invalid config param should not prevent client from mounting"
1598 [ $UID -ne 0 -o $RUNAS_ID -eq 0 ] && skip_env "run as root"
1600 chmod ugo+x $DIR || error "chmod 0 failed"
1602 "lctl get_param -n mdt.$FSNAME-MDT0000.root_squash" \
1603 "$FSNAME.mdt.root_squash" \
1606 "lctl get_param -n mdt.$FSNAME-MDT0000.nosquash_nids" \
1607 "$FSNAME.mdt.nosquash_nids" \
1611 # create set of test files
1613 echo "111" > $DIR/$tfile-userfile || error "write 1 failed"
1614 chmod go-rw $DIR/$tfile-userfile || error "chmod 1 failed"
1615 chown $RUNAS_ID.$RUNAS_ID $DIR/$tfile-userfile || error "chown failed"
1617 echo "222" > $DIR/$tfile-rootfile || error "write 2 failed"
1618 chmod go-rw $DIR/$tfile-rootfile || error "chmod 2 faield"
1620 mkdir $DIR/$tdir-rootdir -p || error "mkdir failed"
1621 chmod go-rwx $DIR/$tdir-rootdir || error "chmod 3 failed"
1622 touch $DIR/$tdir-rootdir/tfile-1 || error "touch failed"
1625 # check root_squash:
1626 # set root squash UID:GID to RUNAS_ID
1627 # root should be able to access only files owned by RUNAS_ID
1630 "lctl get_param -n mdt.$FSNAME-MDT0000.root_squash" \
1631 "$FSNAME.mdt.root_squash" \
1632 "$RUNAS_ID:$RUNAS_ID"
1634 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-userfile)
1635 dd if=$DIR/$tfile-userfile 1>/dev/null 2>/dev/null || \
1636 error "$ST: root read permission is denied"
1637 echo "$ST: root read permission is granted - ok"
1640 dd conv=notrunc if=$DIR/$tfile-userfile 1>/dev/null 2>/dev/null || \
1641 error "$ST: root write permission is denied"
1642 echo "$ST: root write permission is granted - ok"
1644 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-rootfile)
1645 dd if=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null && \
1646 error "$ST: root read permission is granted"
1647 echo "$ST: root read permission is denied - ok"
1650 dd conv=notrunc of=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null && \
1651 error "$ST: root write permission is granted"
1652 echo "$ST: root write permission is denied - ok"
1654 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tdir-rootdir)
1655 rm $DIR/$tdir-rootdir/tfile-1 1>/dev/null 2>/dev/null && \
1656 error "$ST: root unlink permission is granted"
1657 echo "$ST: root unlink permission is denied - ok"
1659 touch $DIR/tdir-rootdir/tfile-2 1>/dev/null 2>/dev/null && \
1660 error "$ST: root create permission is granted"
1661 echo "$ST: root create permission is denied - ok"
1664 # check nosquash_nids:
1665 # put client's NID into nosquash_nids list,
1666 # root should be able to access root file after that
1668 local NIDLIST=$(lctl list_nids all | tr '\n' ' ')
1669 NIDLIST="2@elan $NIDLIST 192.168.0.[2,10]@tcp"
1670 NIDLIST=$(echo $NIDLIST | tr -s ' ' ' ')
1672 "lctl get_param -n mdt.$FSNAME-MDT0000.nosquash_nids" \
1673 "$FSNAME-MDTall.mdt.nosquash_nids" \
1676 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-rootfile)
1677 dd if=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null || \
1678 error "$ST: root read permission is denied"
1679 echo "$ST: root read permission is granted - ok"
1682 dd conv=notrunc of=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null || \
1683 error "$ST: root write permission is denied"
1684 echo "$ST: root write permission is granted - ok"
1686 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tdir-rootdir)
1687 rm $DIR/$tdir-rootdir/tfile-1 || \
1688 error "$ST: root unlink permission is denied"
1689 echo "$ST: root unlink permission is granted - ok"
1690 touch $DIR/$tdir-rootdir/tfile-2 || \
1691 error "$ST: root create permission is denied"
1692 echo "$ST: root create permission is granted - ok"
1696 run_test 43 "check root_squash and nosquash_nids"
1698 umount_client $MOUNT
1703 check_mount || return 2
1704 UUID=$($LCTL get_param llite.${FSNAME}*.uuid | cut -d= -f2)
1706 UUIDS=$(do_facet mds "$LCTL get_param mdt.${FSNAME}*.exports.*.uuid")
1707 for VAL in $UUIDS; do
1708 NID=$(echo $VAL | cut -d= -f1)
1709 CLUUID=$(echo $VAL | cut -d= -f2)
1710 [ "$UUID" = "$CLUUID" ] && STATS_FOUND=yes && break
1712 [ "$STATS_FOUND" = "no" ] && error "stats not found for client"
1716 run_test 44 "mounted client proc entry exists"
1720 check_mount || return 2
1725 #define OBD_FAIL_PTLRPC_LONG_UNLINK 0x50f
1726 do_facet client "lctl set_param fail_loc=0x50f"
1729 manual_umount_client --force || return 3
1730 do_facet client "lctl set_param fail_loc=0x0"
1732 mount_client $MOUNT || return 4
1736 run_test 45 "long unlink handling in ptlrpcd"
1743 umount_client $MOUNT2 || rc=$?
1744 umount_client $MOUNT || rc=$?
1745 while [ $count -gt 0 ]; do
1746 stop ost${count} -f || rc=$?
1750 # writeconf is needed after the test, otherwise,
1751 # we might end up with extra OSTs
1753 cleanup_nocli || rc=$?
1758 echo "Testing with $OSTCOUNT OSTs"
1760 start_mds || return 1
1761 #first client should see only one ost
1762 start_ost || return 2
1763 wait_osc_import_state mds ost FULL
1765 mount_client $MOUNT || return 3
1766 trap "cleanup_46a $OSTCOUNT" EXIT ERR
1769 for (( i=2; i<=$OSTCOUNT; i++ )); do
1770 start ost$i `ostdevname $i` $OST_MOUNT_OPTS || return $((i+2))
1773 # wait until osts in sync
1774 for (( i=2; i<=$OSTCOUNT; i++ )); do
1775 wait_osc_import_state mds ost$i FULL
1779 #second client see all ost's
1781 mount_client $MOUNT2 || return 8
1782 $LFS setstripe $MOUNT2 -c -1 || return 9
1783 $LFS getstripe $MOUNT2 || return 10
1785 echo "ok" > $MOUNT2/widestripe
1786 $LFS getstripe $MOUNT2/widestripe || return 11
1787 # fill acl buffer for avoid expand lsm to them
1788 awk -F : '{if (FNR < 25) { print "u:"$1":rwx" }}' /etc/passwd | while read acl; do
1789 setfacl -m $acl $MOUNT2/widestripe
1793 stat $MOUNT/widestripe || return 12
1795 cleanup_46a $OSTCOUNT || { echo "cleanup_46a failed!" && return 13; }
1798 run_test 46a "handle ost additional - wide striped file"
1803 check_mount || return 2
1804 $LCTL set_param ldlm.namespaces.$FSNAME-*-*-*.lru_size=100
1808 for ns in $($LCTL get_param ldlm.namespaces.$FSNAME-*-*-*.lru_size); do
1809 if echo $ns | grep "MDT[[:digit:]]*"; then
1812 lrs=$(echo $ns | sed 's/.*lru_size=//')
1813 lru_size[count]=$lrs
1818 facet_failover $SINGLEMDS
1819 client_up || return 3
1822 for ns in $($LCTL get_param ldlm.namespaces.$FSNAME-*-*-*.lru_size); do
1823 if echo $ns | grep "MDT[[:digit:]]*"; then
1826 lrs=$(echo $ns | sed 's/.*lru_size=//')
1827 if ! test "$lrs" -eq "${lru_size[count]}"; then
1828 n=$(echo $ns | sed -e 's/ldlm.namespaces.//' -e 's/.lru_size=.*//')
1829 error "$n has lost lru_size: $lrs vs. ${lru_size[count]}"
1837 run_test 47 "server restart does not make client loss lru_resize settings"
1842 # reformat after this test is needed - if test will failed
1843 # we will have unkillable file at FS
1847 test_48() { # bug 17636
1850 check_mount || return 2
1852 $LFS setstripe $MOUNT -c -1 || return 9
1853 $LFS getstripe $MOUNT || return 10
1855 echo "ok" > $MOUNT/widestripe
1856 $LFS getstripe $MOUNT/widestripe || return 11
1858 trap cleanup_48 EXIT ERR
1860 # fill acl buffer for avoid expand lsm to them
1861 getent passwd | awk -F : '{ print "u:"$1":rwx" }' | while read acl; do
1862 setfacl -m $acl $MOUNT/widestripe
1865 stat $MOUNT/widestripe || return 12
1870 run_test 48 "too many acls on file"
1872 # check PARAM_SYS_LDLM_TIMEOUT option of MKFS.LUSTRE
1873 test_49() { # bug 17710
1874 local OLD_MDS_MKFS_OPTS=$MDS_MKFS_OPTS
1875 local OLD_OST_MKFS_OPTS=$OST_MKFS_OPTS
1876 local LOCAL_TIMEOUT=20
1879 OST_MKFS_OPTS="--ost --fsname=$FSNAME --device-size=$OSTSIZE --mgsnode=$MGSNID --param sys.timeout=$LOCAL_TIMEOUT --param sys.ldlm_timeout=$LOCAL_TIMEOUT $MKFSOPT $OSTOPT"
1883 check_mount || return 1
1885 echo "check ldlm_timout..."
1886 LDLM_MDS="`do_facet mds lctl get_param -n ldlm_timeout`"
1887 LDLM_OST1="`do_facet ost1 lctl get_param -n ldlm_timeout`"
1888 LDLM_CLIENT="`do_facet client lctl get_param -n ldlm_timeout`"
1890 if [ $LDLM_MDS -ne $LDLM_OST1 ] || [ $LDLM_MDS -ne $LDLM_CLIENT ]; then
1891 error "Different LDLM_TIMEOUT:$LDLM_MDS $LDLM_OST1 $LDLM_CLIENT"
1894 if [ $LDLM_MDS -ne $((LOCAL_TIMEOUT / 3)) ]; then
1895 error "LDLM_TIMEOUT($LDLM_MDS) is not correct"
1898 umount_client $MOUNT
1899 stop_ost || return 2
1900 stop_mds || return 3
1902 OST_MKFS_OPTS="--ost --fsname=$FSNAME --device-size=$OSTSIZE --mgsnode=$MGSNID --param sys.timeout=$LOCAL_TIMEOUT --param sys.ldlm_timeout=$((LOCAL_TIMEOUT - 1)) $MKFSOPT $OSTOPT"
1906 check_mount || return 7
1908 LDLM_MDS="`do_facet mds lctl get_param -n ldlm_timeout`"
1909 LDLM_OST1="`do_facet ost1 lctl get_param -n ldlm_timeout`"
1910 LDLM_CLIENT="`do_facet client lctl get_param -n ldlm_timeout`"
1912 if [ $LDLM_MDS -ne $LDLM_OST1 ] || [ $LDLM_MDS -ne $LDLM_CLIENT ]; then
1913 error "Different LDLM_TIMEOUT:$LDLM_MDS $LDLM_OST1 $LDLM_CLIENT"
1916 if [ $LDLM_MDS -ne $((LOCAL_TIMEOUT - 1)) ]; then
1917 error "LDLM_TIMEOUT($LDLM_MDS) is not correct"
1920 cleanup || return $?
1922 MDS_MKFS_OPTS=$OLD_MDS_MKFS_OPTS
1923 OST_MKFS_OPTS=$OLD_OST_MKFS_OPTS
1925 run_test 49 "check PARAM_SYS_LDLM_TIMEOUT option of MKFS.LUSTRE"
1928 # Test both statfs and lfs df and fail if either one fails
1929 multiop_bg_pause $1 f_
1932 killall -USR1 multiop
1933 [ $RC1 -ne 0 ] && log "lazystatfs multiop failed"
1934 wait $PID || { RC1=$?; log "multiop return error "; }
1941 if [ $RC2 -eq 0 ]; then
1943 log "lazystatfs df failed"
1947 [[ $RC1 -ne 0 || $RC2 -eq 0 ]] && RC=1
1953 lctl set_param llite.$FSNAME-*.lazystatfs=1
1956 lazystatfs $MOUNT || error "lazystatfs failed but no down servers"
1958 cleanup || return $?
1960 run_test 50a "lazystatfs all servers available =========================="
1964 lctl set_param llite.$FSNAME-*.lazystatfs=1
1967 # Wait for client to detect down OST
1968 stop_ost || error "Unable to stop OST1"
1969 wait_osc_import_state mds ost DISCONN
1971 lazystatfs $MOUNT || error "lazystatfs should don't have returned EIO"
1973 umount_client $MOUNT || error "Unable to unmount client"
1974 stop_mds || error "Unable to stop MDS"
1976 run_test 50b "lazystatfs all servers down =========================="
1979 start_mds || error "Unable to start MDS"
1980 start_ost || error "Unable to start OST1"
1981 start_ost2 || error "Unable to start OST2"
1982 mount_client $MOUNT || error "Unable to mount client"
1983 lctl set_param llite.$FSNAME-*.lazystatfs=1
1986 # Wait for client to detect down OST
1987 stop_ost || error "Unable to stop OST1"
1988 wait_osc_import_state mds ost DISCONN
1989 lazystatfs $MOUNT || error "lazystatfs failed with one down server"
1991 umount_client $MOUNT || error "Unable to unmount client"
1992 stop_ost2 || error "Unable to stop OST2"
1993 stop_mds || error "Unable to stop MDS"
1995 run_test 50c "lazystatfs one server down =========================="
1998 start_mds || error "Unable to start MDS"
1999 start_ost || error "Unable to start OST1"
2000 start_ost2 || error "Unable to start OST2"
2001 mount_client $MOUNT || error "Unable to mount client"
2002 lctl set_param llite.$FSNAME-*.lazystatfs=1
2005 # Issue the statfs during the window where the client still
2006 # belives the OST to be available but it is in fact down.
2007 # No failure just a statfs which hangs for a timeout interval.
2008 stop_ost || error "Unable to stop OST1"
2009 lazystatfs $MOUNT || error "lazystatfs failed with one down server"
2011 umount_client $MOUNT || error "Unable to unmount client"
2012 stop_ost2 || error "Unable to stop OST2"
2013 stop_mds || error "Unable to stop MDS"
2015 run_test 50d "lazystatfs client/server conn race =========================="
2022 start_mds || return 1
2023 #first client should see only one ost
2024 start_ost || return 2
2025 wait_osc_import_state mds ost FULL
2027 # Wait for client to detect down OST
2028 stop_ost || error "Unable to stop OST1"
2029 wait_osc_import_state mds ost DISCONN
2031 mount_client $MOUNT || error "Unable to mount client"
2032 lctl set_param llite.$FSNAME-*.lazystatfs=0
2034 multiop_bg_pause $MOUNT _f
2038 if [ $RC1 -ne 0 ]; then
2039 log "multiop failed $RC1"
2042 sleep $(( $TIMEOUT+1 ))
2044 [ $? -ne 0 ] && error "process isn't sleep"
2045 start_ost || error "Unable to start OST1"
2046 wait $pid || error "statfs failed"
2049 umount_client $MOUNT || error "Unable to unmount client"
2050 stop_ost || error "Unable to stop OST1"
2051 stop_mds || error "Unable to stop MDS"
2053 run_test 50e "normal statfs all servers down =========================="
2058 CONN_PROC="osc.$FSNAME-OST0001-osc-[M]*.ost_server_uuid"
2060 start_mds || error "Unable to start mds"
2061 #first client should see only one ost
2062 start_ost || error "Unable to start OST1"
2063 wait_osc_import_state mds ost FULL
2065 start_ost2 || error "Unable to start OST2"
2066 wait_osc_import_state mds ost2 FULL
2068 # Wait for client to detect down OST
2069 stop_ost2 || error "Unable to stop OST2"
2071 wait_osc_import_state mds ost2 DISCONN
2072 mount_client $MOUNT || error "Unable to mount client"
2073 lctl set_param llite.$FSNAME-*.lazystatfs=0
2075 multiop_bg_pause $MOUNT _f
2079 if [ $RC1 -ne 0 ]; then
2080 log "lazystatfs multiop failed $RC1"
2083 sleep $(( $TIMEOUT+1 ))
2085 [ $? -ne 0 ] && error "process isn't sleep"
2086 start_ost2 || error "Unable to start OST2"
2087 wait $pid || error "statfs failed"
2088 stop_ost2 || error "Unable to stop OST2"
2091 umount_client $MOUNT || error "Unable to unmount client"
2092 stop_ost || error "Unable to stop OST1"
2093 stop_mds || error "Unable to stop MDS"
2096 run_test 50f "normal statfs one server in down =========================="
2099 [ "$OSTCOUNT" -lt "2" ] && skip_env "$OSTCOUNT < 2, skipping" && return
2101 start_ost2 || error "Unable to start OST2"
2103 local PARAM="${FSNAME}-OST0001.osc.active"
2105 $LFS setstripe -c -1 $DIR/$tfile || error "Unable to lfs setstripe"
2106 do_facet mgs $LCTL conf_param $PARAM=0 || error "Unable to deactivate OST"
2108 umount_client $MOUNT || error "Unable to unmount client"
2109 mount_client $MOUNT || error "Unable to mount client"
2110 # This df should not cause a panic
2113 do_facet mgs $LCTL conf_param $PARAM=1 || error "Unable to activate OST"
2115 umount_client $MOUNT || error "Unable to unmount client"
2116 stop_ost2 || error "Unable to stop OST2"
2117 stop_ost || error "Unable to stop OST1"
2118 stop_mds || error "Unable to stop MDS"
2121 run_test 50g "deactivated OST should not cause panic====================="
2124 local LOCAL_TIMEOUT=20
2128 check_mount || return 1
2131 $LFS setstripe -c -1 $MOUNT/d1
2132 #define OBD_FAIL_MDS_REINT_DELAY 0x142
2133 do_facet $SINGLEMDS "lctl set_param fail_loc=0x142"
2134 touch $MOUNT/d1/f1 &
2137 start_ost2 || return 2
2139 stop_ost2 || return 3
2142 run_test 51 "Verify that mdt_reint handles RMF_MDT_MD correctly when an OST is added"
2151 do_node $node mkdir -p $dest
2152 [ $? -eq 0 ] || { error "Unable to create directory"; return 1; }
2154 do_node $node 'tar cf - '$@' | tar xf - -C '$dest';
2155 [ \"\${PIPESTATUS[*]}\" = \"0 0\" ] || exit 1'
2156 [ $? -eq 0 ] || { error "Unable to tar files"; return 2; }
2158 do_node $node 'getfattr -d -m "[a-z]*\\." '$@' > '$xattrs
2159 [ $? -eq 0 ] || { error "Unable to read xattrs"; return 3; }
2169 local backup2=${TMP}/backup2
2171 do_node $node mkdir -p $backup2
2172 [ $? -eq 0 ] || { error "Unable to create directory"; return 1; }
2174 do_node $node 'tar cf - '$@' | tar xf - -C '$backup2';
2175 [ \"\${PIPESTATUS[*]}\" = \"0 0\" ] || exit 1'
2176 [ $? -eq 0 ] || { error "Unable to tar files to diff"; return 2; }
2178 do_node $node "diff -rq $backup $backup2"
2179 [ $? -eq 0 ] || { error "contents differ"; return 3; }
2181 local xattrs2=${TMP}/xattrs2
2182 do_node $node 'getfattr -d -m "[a-z]*\\." '$@' > '$xattrs2
2183 [ $? -eq 0 ] || { error "Unable to read xattrs to diff"; return 4; }
2185 do_node $node "diff $xattrs $xattrs2"
2186 [ $? -eq 0 ] || { error "xattrs differ"; return 5; }
2188 do_node $node "rm -rf $backup2 $xattrs2"
2189 [ $? -eq 0 ] || { error "Unable to delete temporary files"; return 6; }
2194 [ $? -eq 0 ] || { error "Unable to start MDS"; return 1; }
2196 [ $? -eq 0 ] || { error "Unable to start OST1"; return 2; }
2198 [ $? -eq 0 ] || { error "Unable to mount client"; return 3; }
2201 local ost1mnt=${MOUNT%/*}/ost1
2202 local ost1node=$(facet_active_host ost1)
2203 local ost1tmp=$TMP/conf52
2206 [ $? -eq 0 ] || { error "Unable to create tdir"; return 4; }
2207 touch $TMP/modified_first
2208 [ $? -eq 0 ] || { error "Unable to create temporary file"; return 5; }
2209 do_node $ost1node "mkdir -p $ost1tmp && touch $ost1tmp/modified_first"
2210 [ $? -eq 0 ] || { error "Unable to create temporary file"; return 6; }
2213 $LFS setstripe $DIR/$tdir -c -1 -s 1M
2214 [ $? -eq 0 ] || { error "lfs setstripe failed"; return 7; }
2216 for (( i=0; i < nrfiles; i++ )); do
2217 multiop $DIR/$tdir/$tfile-$i Ow1048576w1048576w524288c
2218 [ $? -eq 0 ] || { error "multiop failed"; return 8; }
2224 echo backup files to $TMP/files
2225 local files=$(find $DIR/$tdir -type f -newer $TMP/modified_first)
2226 copy_files_xattrs `hostname` $TMP/files $TMP/file_xattrs $files
2227 [ $? -eq 0 ] || { error "Unable to copy files"; return 9; }
2229 umount_client $MOUNT
2230 [ $? -eq 0 ] || { error "Unable to umount client"; return 10; }
2232 [ $? -eq 0 ] || { error "Unable to stop ost1"; return 11; }
2234 echo mount ost1 as ldiskfs
2235 do_node $ost1node mount -t $FSTYPE $ost1_dev $ost1mnt $OST_MOUNT_OPTS
2236 [ $? -eq 0 ] || { error "Unable to mount ost1 as ldiskfs"; return 12; }
2239 echo backup objects to $ost1tmp/objects
2240 local objects=$(do_node $ost1node 'find '$ost1mnt'/O/0 -type f -size +0'\
2241 '-newer '$ost1tmp'/modified_first -regex ".*\/[0-9]+"')
2242 copy_files_xattrs $ost1node $ost1tmp/objects $ost1tmp/object_xattrs $objects
2243 [ $? -eq 0 ] || { error "Unable to copy objects"; return 13; }
2245 # move objects to lost+found
2246 do_node $ost1node 'mv '$objects' '${ost1mnt}'/lost+found'
2247 [ $? -eq 0 ] || { error "Unable to move objects"; return 14; }
2250 do_node $ost1node "ll_recover_lost_found_objs -d $ost1mnt/lost+found"
2251 [ $? -eq 0 ] || { error "ll_recover_lost_found_objs failed"; return 15; }
2253 # compare restored objects against saved ones
2254 diff_files_xattrs $ost1node $ost1tmp/objects $ost1tmp/object_xattrs $objects
2255 [ $? -eq 0 ] || { error "Unable to diff objects"; return 16; }
2257 do_node $ost1node "umount $ost1_dev"
2258 [ $? -eq 0 ] || { error "Unable to umount ost1 as ldiskfs"; return 17; }
2261 [ $? -eq 0 ] || { error "Unable to start ost1"; return 18; }
2263 [ $? -eq 0 ] || { error "Unable to mount client"; return 19; }
2266 diff_files_xattrs `hostname` $TMP/files $TMP/file_xattrs $files
2267 [ $? -eq 0 ] || { error "Unable to diff files"; return 20; }
2269 rm -rf $TMP/files $TMP/file_xattrs
2270 [ $? -eq 0 ] || { error "Unable to delete temporary files"; return 21; }
2271 do_node $ost1node "rm -rf $ost1tmp"
2272 [ $? -eq 0 ] || { error "Unable to delete temporary files"; return 22; }
2275 run_test 52 "check recovering objects from lost+found"
2277 # Checks threads_min/max/started for some service
2279 # Arguments: service name (OST or MDT), facet (e.g., ost1, $SINGLEMDS), and a
2280 # parameter pattern prefix like 'ost.*.ost'.
2292 local msg="Insane $modname thread counts"
2296 check_mount || return 41
2298 # We need to expand $parampat, but it may match multiple parameters, so
2299 # we'll pick the first one
2300 if ! paramp=$(do_facet $facet "lctl get_param -N ${parampat}.threads_min"|head -1); then
2301 error "Couldn't expand ${parampat}.threads_min parameter name"
2305 # Remove the .threads_min part
2306 paramp=${paramp%.threads_min}
2308 # Check for sanity in defaults
2309 tmin=$(do_facet $facet "lctl get_param -n ${paramp}.threads_min" || echo 0)
2310 tmax=$(do_facet $facet "lctl get_param -n ${paramp}.threads_max" || echo 0)
2311 tstarted=$(do_facet $facet "lctl get_param -n ${paramp}.threads_started" || echo 0)
2312 lassert 23 "$msg (PDSH problems?)" '(($tstarted && $tmin && $tmax))' || return $?
2313 lassert 24 "$msg" '(($tstarted >= $tmin && $tstarted <= tmax ))' || return $?
2315 # Check that we can lower min/max
2316 do_facet $facet "lctl set_param ${paramp}.threads_min=$((tmin - 1))"
2317 do_facet $facet "lctl set_param ${paramp}.threads_max=$((tmax - 10))"
2318 tmin2=$(do_facet $facet "lctl get_param -n ${paramp}.threads_min" || echo 0)
2319 tmax2=$(do_facet $facet "lctl get_param -n ${paramp}.threads_max" || echo 0)
2320 lassert 25 "$msg" '(($tmin2 == ($tmin - 1) && $tmax2 == ($tmax -10)))' || return $?
2322 # Check that we can set min/max to the same value
2323 do_facet $facet "lctl set_param ${paramp}.threads_min=$tmin"
2324 do_facet $facet "lctl set_param ${paramp}.threads_max=$tmin"
2325 tmin2=$(do_facet $facet "lctl get_param -n ${paramp}.threads_min" || echo 0)
2326 tmax2=$(do_facet $facet "lctl get_param -n ${paramp}.threads_max" || echo 0)
2327 lassert 26 "$msg" '(($tmin2 == $tmin && $tmax2 == $tmin))' || return $?
2329 # Check that we can't set max < min
2330 do_facet $facet "lctl set_param ${paramp}.threads_max=$((tmin - 1))"
2331 tmin2=$(do_facet $facet "lctl get_param -n ${paramp}.threads_min" || echo 0)
2332 tmax2=$(do_facet $facet "lctl get_param -n ${paramp}.threads_max" || echo 0)
2333 lassert 27 "$msg" '(($tmin <= $tmax2))' || return $?
2335 # We need to ensure that we get the module options desired; to do this
2336 # we set LOAD_MODULES_REMOTE=true and we call setmodopts below.
2337 LOAD_MODULES_REMOTE=true
2340 setmodopts -a $modname "$opts" oldvalue
2344 check_mount || return 41
2346 # Restore previous setting of MODOPTS_*
2347 setmodopts $modname "$oldvalue"
2349 # Check that $opts took
2350 tmin=$(do_facet $facet "lctl get_param -n ${paramp}.threads_min")
2351 tmax=$(do_facet $facet "lctl get_param -n ${paramp}.threads_max")
2352 tstarted=$(do_facet $facet "lctl get_param -n ${paramp}.threads_started")
2353 lassert 28 "$msg" '(($tstarted == $tmin && $tstarted == $tmax ))' || return $?
2356 # Workaround a YALA bug where YALA expects that modules will remain
2357 # loaded on the servers
2358 LOAD_MODULES_REMOTE=false
2365 thread_sanity OST ost1 'ost.*.ost' 'oss_num_threads=64'
2367 run_test 53a "check OSS thread count params"
2370 thread_sanity MDT $SINGLEMDS 'mdt.*.*.' 'mdt_num_threads=64'
2372 run_test 53b "check MDT thread count params"
2374 if ! combined_mgs_mds ; then
2381 local partial_arg=""
2382 local size=$(df -B G $dir | tail -1 | awk '{print $2}' | sed 's/G//') # Gb
2384 # Run in partial (fast) mode if the size
2385 # of a partition > 10 GB
2386 [ $size -gt 10 ] && partial_arg="-p"
2388 llverfs $partial_arg $dir
2392 do_rpc_nodes $(facet_host ost1) run_llverdev $(ostdevname 1)
2393 [ $? -eq 0 ] || error "llverdev failed!"
2396 run_test 54a "llverdev"
2401 [ $? -eq 0 ] || error "llverfs failed!"
2404 run_test 54b "llverfs"
2408 local max_ost_index=$1
2409 echo -n $(((max_ost_index + 1) * 8))
2413 local mdsdev=$(mdsdevname 1)
2414 local ostdev=$(ostdevname 1)
2415 local saved_opts=$OST_MKFS_OPTS
2417 for i in 0 1023 2048
2419 OST_MKFS_OPTS="$saved_opts --index $i"
2427 echo checking size of lov_objid for ost index $i
2428 LOV_OBJID_SIZE=$(do_facet mds1 "$DEBUGFS -R 'stat lov_objid' $mdsdev 2>/dev/null" | grep ^User | awk '{print $6}')
2429 if [ "$LOV_OBJID_SIZE" != $(lov_objid_size $i) ]; then
2430 error "lov_objid size has to be $(lov_objid_size $i), not $LOV_OBJID_SIZE"
2432 echo ok, lov_objid size is correct: $LOV_OBJID_SIZE
2437 OST_MKFS_OPTS=$saved_opts
2440 run_test 55 "check lov_objid size"
2443 add mds1 $MDS_MKFS_OPTS --mkfsoptions='\"-J size=16\"' --reformat $(mdsdevname 1)
2444 add ost1 $OST_MKFS_OPTS --index=1000 --reformat $(ostdevname 1)
2445 add ost2 $OST_MKFS_OPTS --index=10000 --reformat $(ostdevname 2)
2449 start_ost2 || error "Unable to start second ost"
2450 mount_client $MOUNT || error "Unable to mount client"
2456 run_test 56 "check big indexes"
2459 equals_msg `basename $0`: test complete
2460 [ -f "$TESTSUITELOG" ] && cat $TESTSUITELOG && grep -q FAIL $TESTSUITELOG && exit 1 || true