3 # add uml1 uml2 uml3 in your /etc/hosts
5 # FIXME - there is no reason to use all of these different
6 # return codes, espcially when most of them are mapped to something
7 # else anyway. The combination of test number and return code
8 # figure out what failed.
14 # These tests don't apply to mountconf
15 MOUNTCONFSKIP="10 11 12 13 13b 14 15"
16 # bug number for skipped test: 13739
17 HEAD_EXCEPT=" 32a 32b "
19 # bug number for skipped test:
20 ALWAYS_EXCEPT=" $CONF_SANITY_EXCEPT $MOUNTCONFSKIP $HEAD_EXCEPT"
21 # UPDATE THE COMMENT ABOVE WITH BUG NUMBERS WHEN CHANGING ALWAYS_EXCEPT!
24 PATH=$PWD/$SRCDIR:$SRCDIR:$SRCDIR/../utils:$PATH
26 PTLDEBUG=${PTLDEBUG:--1}
28 LUSTRE=${LUSTRE:-`dirname $0`/..}
29 RLUSTRE=${RLUSTRE:-$LUSTRE}
30 MOUNTLUSTRE=${MOUNTLUSTRE:-/sbin/mount.lustre}
31 MKFSLUSTRE=${MKFSLUSTRE:-/usr/sbin/mkfs.lustre}
34 . $LUSTRE/tests/test-framework.sh
36 # STORED_MDSSIZE is used in test_18
37 if [ -n "$MDSSIZE" ]; then
38 STORED_MDSSIZE=$MDSSIZE
40 # use small MDS + OST size to speed formatting time
43 . ${CONFIG:=$LUSTRE/tests/cfg/local.sh}
46 [ "$SLOW" = "no" ] && EXCEPT_SLOW="0 1 2 3 6 7 15 18 24b 25 30 31 32 33 34a "
59 # who knows if/where $TUNEFS is installed? Better reformat if it fails...
60 do_facet ${facet} "$TUNEFS --writeconf $MDSDEV" || echo "tunefs failed, reformatting instead" && reformat
65 # The MGS must be started before the OSTs for a new fs, so start
66 # and stop to generate the startup logs.
75 echo "start mds service on `facet_active_host mds`"
76 start mds $MDSDEV $MDS_MOUNT_OPTS || return 94
80 echo "stop mds service on `facet_active_host mds`"
81 # These tests all use non-failover stop
82 stop mds -f || return 97
86 echo "start ost1 service on `facet_active_host ost1`"
87 start ost1 `ostdevname 1` $OST_MOUNT_OPTS || return 95
91 echo "stop ost1 service on `facet_active_host ost1`"
92 # These tests all use non-failover stop
93 stop ost1 -f || return 98
97 echo "start ost2 service on `facet_active_host ost2`"
98 start ost2 `ostdevname 2` $OST_MOUNT_OPTS || return 92
102 echo "stop ost2 service on `facet_active_host ost2`"
103 # These tests all use non-failover stop
104 stop ost2 -f || return 93
109 echo "mount $FSNAME on ${MOUNTPATH}....."
110 zconf_mount `hostname` $MOUNTPATH || return 96
114 local SAVEMOUNTOPT=$MOUNTOPT
115 MOUNTOPT="remount,$1"
117 echo "remount '$1' lustre on ${MOUNTPATH}....."
118 zconf_mount `hostname` $MOUNTPATH || return 96
119 MOUNTOPT=$SAVEMOUNTOPT
124 echo "umount lustre on ${MOUNTPATH}....."
125 zconf_umount `hostname` $MOUNTPATH || return 97
128 manual_umount_client(){
129 echo "manual umount lustre on ${MOUNT}...."
130 do_facet client "umount -d $MOUNT"
140 stop_mds || return 201
141 stop_ost || return 202
142 unload_modules || return 203
146 umount_client $MOUNT || return 200
147 cleanup_nocli || return $?
151 do_facet client "cp /etc/passwd $DIR/a" || return 71
152 do_facet client "rm $DIR/a" || return 72
153 # make sure lustre is actually mounted (touch will block,
154 # but grep won't, so do it after)
155 do_facet client "grep $MOUNT' ' /proc/mounts > /dev/null" || return 73
156 echo "setup single mount lustre success"
160 do_facet client "touch $DIR/a" || return 71
161 do_facet client "rm $DIR/a" || return 72
162 do_facet client "touch $DIR2/a" || return 73
163 do_facet client "rm $DIR2/a" || return 74
164 echo "setup double mount lustre success"
169 if [ "$ONLY" == "setup" ]; then
174 if [ "$ONLY" == "cleanup" ]; then
179 #create single point mountpoint
187 check_mount || return 41
190 run_test 0 "single mount setup"
194 echo "start ost second time..."
196 check_mount || return 42
199 run_test 1 "start up ost twice (should return errors)"
204 echo "start mds second time.."
207 check_mount || return 43
210 run_test 2 "start up mds twice (should return err)"
214 #mount.lustre returns an error if already in mtab
215 mount_client $MOUNT && return $?
216 check_mount || return 44
219 run_test 3 "mount client twice (should return err)"
223 touch $DIR/$tfile || return 85
227 # ok for ost to fail shutdown
228 if [ 202 -ne $eno ]; then
233 run_test 4 "force cleanup ost, then cleanup"
235 test_5a() { # was test_5
237 touch $DIR/$tfile || return 1
238 fuser -m -v $MOUNT && echo "$MOUNT is in use by user space process."
240 stop_mds -f || return 2
242 # cleanup may return an error from the failed
243 # disconnects; for now I'll consider this successful
244 # if all the modules have unloaded.
248 echo "killing umount"
249 kill -TERM $UMOUNT_PID
250 echo "waiting for umount to finish"
252 if grep " $MOUNT " /proc/mounts; then
253 echo "test 5: /proc/mounts after failed umount"
257 echo "killing umount"
258 kill -TERM $UMOUNT_PID
259 echo "waiting for umount to finish"
261 grep " $MOUNT " /proc/mounts && echo "test 5: /proc/mounts after second umount" && return 11
265 # stop_mds is a no-op here, and should not fail
266 cleanup_nocli || return $?
267 # df may have lingering entry
269 # mtab may have lingering entry
273 while [ "$WAIT" -ne "$MAX_WAIT" ]; do
275 grep -q $MOUNT" " /etc/mtab || break
276 echo "Waiting /etc/mtab updated ... "
277 WAIT=$(( WAIT + sleep))
279 [ "$WAIT" -eq "$MAX_WAIT" ] && error "/etc/mtab is not updated in $WAIT secs"
280 echo "/etc/mtab updated in $WAIT secs"
282 run_test 5a "force cleanup mds, then cleanup"
286 [ -d $MOUNT ] || mkdir -p $MOUNT
287 grep " $MOUNT " /etc/mtab && echo "test 5b: mtab before mount" && return 10
288 mount_client $MOUNT && return 1
289 grep " $MOUNT " /etc/mtab && echo "test 5b: mtab after failed mount" && return 11
291 # stop_mds is a no-op here, and should not fail
292 cleanup_nocli || return $?
295 run_test 5b "mds down, cleanup after failed mount (bug 2712) (should return errs)"
300 [ -d $MOUNT ] || mkdir -p $MOUNT
301 grep " $MOUNT " /etc/mtab && echo "test 5c: mtab before mount" && return 10
302 local oldfs="${FSNAME}"
303 FSNAME="wrong.${FSNAME}"
304 mount_client $MOUNT || :
306 grep " $MOUNT " /etc/mtab && echo "test 5c: mtab after failed mount" && return 11
308 cleanup_nocli || return $?
310 run_test 5c "cleanup after failed mount (bug 2712) (should return errs)"
316 grep " $MOUNT " /etc/mtab && echo "test 5d: mtab before mount" && return 10
317 mount_client $MOUNT || return 1
319 grep " $MOUNT " /etc/mtab && echo "test 5d: mtab after unmount" && return 11
322 run_test 5d "mount with ost down"
328 #define OBD_FAIL_PTLRPC_DELAY_SEND 0x506
329 do_facet client "lctl set_param fail_loc=0x80000506"
330 grep " $MOUNT " /etc/mtab && echo "test 5e: mtab before mount" && return 10
331 mount_client $MOUNT || echo "mount failed (not fatal)"
333 grep " $MOUNT " /etc/mtab && echo "test 5e: mtab after unmount" && return 11
336 run_test 5e "delayed connect, don't crash (bug 10268)"
341 mount_client ${MOUNT} || return 87
342 touch $DIR/a || return 86
345 run_test 6 "manual umount, then mount again"
350 cleanup_nocli || return $?
352 run_test 7 "manual umount, then cleanup"
357 check_mount2 || return 45
358 umount_client $MOUNT2
361 run_test 8 "double mount setup"
366 do_facet ost1 lctl set_param debug=\'inode trace\' || return 1
367 do_facet ost1 lctl set_param subsystem_debug=\'mds ost\' || return 1
369 CHECK_PTLDEBUG="`do_facet ost1 lctl get_param -n debug`"
370 if [ "$CHECK_PTLDEBUG" ] && [ "$CHECK_PTLDEBUG" = "trace inode" ];then
371 echo "lnet.debug success"
373 echo "lnet.debug: want 'trace inode', have '$CHECK_PTLDEBUG'"
376 CHECK_SUBSYS="`do_facet ost1 lctl get_param -n subsystem_debug`"
377 if [ "$CHECK_SUBSYS" ] && [ "$CHECK_SUBSYS" = "mds ost" ]; then
378 echo "lnet.subsystem_debug success"
380 echo "lnet.subsystem_debug: want 'mds ost', have '$CHECK_SUBSYS'"
383 stop_ost || return $?
386 run_test 9 "test ptldebug and subsystem for mkfs"
389 echo "generate configuration with the same name for node and mds"
390 OLDXMLCONFIG=$XMLCONFIG
391 XMLCONFIG="broken.xml"
392 [ -f "$XMLCONFIG" ] && rm -f $XMLCONFIG
396 echo "the name for node and mds is the same"
397 do_lmc --add mds --node ${facet}_facet --mds ${facet}_facet \
398 --dev $MDSDEV --size $MDSSIZE || return $?
399 do_lmc --add lov --mds ${facet}_facet --lov lov1 --stripe_sz \
400 $STRIPE_BYTES --stripe_cnt $STRIPES_PER_OBJ \
401 --stripe_pattern 0 || return $?
402 add_ost ost --lov lov1 --dev $OSTDEV --size $OSTSIZE
404 add_facet $facet --lustre_upcall $UPCALL
405 do_lmc --add mtpt --node ${facet}_facet --mds mds_facet \
406 --lov lov1 --path $MOUNT
412 check_mount || return 41
416 XMLCONFIG=$OLDXMLCONFIG
418 run_test 10 "mount lustre with the same name for node and mds"
421 OLDXMLCONFIG=$XMLCONFIG
422 XMLCONFIG="conf11.xml"
424 [ -f "$XMLCONFIG" ] && rm -f $XMLCONFIG
425 add_mds mds --dev $MDSDEV --size $MDSSIZE
426 add_ost ost --dev $OSTDEV --size $OSTSIZE
427 add_client client mds --path $MOUNT --ost ost_svc || return $?
428 echo "Default lov config success!"
430 [ -f "$XMLCONFIG" ] && rm -f $XMLCONFIG
431 add_mds mds --dev $MDSDEV --size $MDSSIZE
432 add_ost ost --dev $OSTDEV --size $OSTSIZE
433 add_client client mds --path $MOUNT && return $?
434 echo "--add mtpt with neither --lov nor --ost will return error"
438 XMLCONFIG=$OLDXMLCONFIG
440 run_test 11 "use default lov configuration (should return error)"
443 OLDXMLCONFIG=$XMLCONFIG
444 XMLCONFIG="batch.xml"
445 BATCHFILE="batchfile"
448 [ -f "$XMLCONFIG" ] && rm -f $XMLCONFIG
449 [ -f "$BATCHFILE" ] && rm -f $BATCHFILE
450 echo "--add net --node $HOSTNAME --nid $HOSTNAME --nettype tcp" > $BATCHFILE
451 echo "--add mds --node $HOSTNAME --mds mds1 --mkfsoptions \"-I 128\"" >> $BATCHFILE
452 # --mkfsoptions "-I 128"
453 do_lmc -m $XMLCONFIG --batch $BATCHFILE || return $?
454 if [ `sed -n '/>-I 128</p' $XMLCONFIG | wc -l` -eq 1 ]; then
455 echo "matched double quote success"
457 echo "matched double quote fail"
462 echo "--add net --node $HOSTNAME --nid $HOSTNAME --nettype tcp" > $BATCHFILE
463 echo "--add mds --node $HOSTNAME --mds mds1 --mkfsoptions \"-I 128" >> $BATCHFILE
464 # --mkfsoptions "-I 128
465 do_lmc -m $XMLCONFIG --batch $BATCHFILE && return $?
466 echo "unmatched double quote should return error"
470 echo "--add net --node $HOSTNAME --nid $HOSTNAME --nettype tcp" > $BATCHFILE
471 echo "--add mds --node $HOSTNAME --mds mds1 --mkfsoptions '-I 128'" >> $BATCHFILE
472 # --mkfsoptions '-I 128'
473 do_lmc -m $XMLCONFIG --batch $BATCHFILE || return $?
474 if [ `sed -n '/>-I 128</p' $XMLCONFIG | wc -l` -eq 1 ]; then
475 echo "matched single quote success"
477 echo "matched single quote fail"
482 echo "--add net --node $HOSTNAME --nid $HOSTNAME --nettype tcp" > $BATCHFILE
483 echo "--add mds --node $HOSTNAME --mds mds1 --mkfsoptions '-I 128" >> $BATCHFILE
484 # --mkfsoptions '-I 128
485 do_lmc -m $XMLCONFIG --batch $BATCHFILE && return $?
486 echo "unmatched single quote should return error"
490 echo "--add net --node $HOSTNAME --nid $HOSTNAME --nettype tcp" > $BATCHFILE
491 echo "--add mds --node $HOSTNAME --mds mds1 --mkfsoptions \-\I\ \128" >> $BATCHFILE
492 # --mkfsoptions \-\I\ \128
493 do_lmc -m $XMLCONFIG --batch $BATCHFILE || return $?
494 if [ `sed -n '/>-I 128</p' $XMLCONFIG | wc -l` -eq 1 ]; then
495 echo "backslash followed by a whitespace/letter success"
497 echo "backslash followed by a whitespace/letter fail"
502 echo "--add net --node $HOSTNAME --nid $HOSTNAME --nettype tcp" > $BATCHFILE
503 echo "--add mds --node $HOSTNAME --mds mds1 --mkfsoptions -I\ 128\\" >> $BATCHFILE
504 # --mkfsoptions -I\ 128\
505 do_lmc -m $XMLCONFIG --batch $BATCHFILE && return $?
506 echo "backslash followed by nothing should return error"
509 XMLCONFIG=$OLDXMLCONFIG
511 run_test 12 "lmc --batch, with single/double quote, backslash in batchfile"
513 test_13a() { # was test_13
514 OLDXMLCONFIG=$XMLCONFIG
515 XMLCONFIG="conf13-1.xml"
517 # check long uuid will be truncated properly and uniquely
518 echo "To generate XML configuration file(with long ost name): $XMLCONFIG"
519 [ -f "$XMLCONFIG" ] && rm -f $XMLCONFIG
520 do_lmc --add net --node $HOSTNAME --nid $HOSTNAME --nettype tcp
521 do_lmc --add mds --node $HOSTNAME --mds mds1_name_longer_than_31characters
522 do_lmc --add mds --node $HOSTNAME --mds mds2_name_longer_than_31characters
523 if [ ! -f "$XMLCONFIG" ]; then
524 echo "Error:no file $XMLCONFIG created!"
527 EXPECTEDMDS1UUID="e_longer_than_31characters_UUID"
528 EXPECTEDMDS2UUID="longer_than_31characters_UUID_2"
529 FOUNDMDS1UUID=`awk -F"'" '/<mds .*uuid=/' $XMLCONFIG | sed -n '1p' \
530 | sed "s/ /\n\r/g" | awk -F"'" '/uuid=/{print $2}'`
531 FOUNDMDS2UUID=`awk -F"'" '/<mds .*uuid=/' $XMLCONFIG | sed -n '2p' \
532 | sed "s/ /\n\r/g" | awk -F"'" '/uuid=/{print $2}'`
533 [ -z "$FOUNDMDS1UUID" ] && echo "MDS1 UUID empty" && return 1
534 [ -z "$FOUNDMDS2UUID" ] && echo "MDS2 UUID empty" && return 1
535 if ([ $EXPECTEDMDS1UUID = $FOUNDMDS1UUID ] && [ $EXPECTEDMDS2UUID = $FOUNDMDS2UUID ]) || \
536 ([ $EXPECTEDMDS1UUID = $FOUNDMDS2UUID ] && [ $EXPECTEDMDS2UUID = $FOUNDMDS1UUID ]); then
537 echo "Success:long uuid truncated successfully and being unique."
539 echo "Error:expected uuid for mds1 and mds2: $EXPECTEDMDS1UUID; $EXPECTEDMDS2UUID"
540 echo "but: found uuid for mds1 and mds2: $FOUNDMDS1UUID; $FOUNDMDS2UUID"
544 XMLCONFIG=$OLDXMLCONFIG
546 run_test 13a "check new_uuid of lmc operating correctly"
549 OLDXMLCONFIG=$XMLCONFIG
550 XMLCONFIG="conf13-1.xml"
551 SECONDXMLCONFIG="conf13-2.xml"
552 # check multiple invocations for lmc generate same XML configuration file
554 echo "Generate the first XML configuration file"
556 echo "mv $XMLCONFIG to $SECONDXMLCONFIG"
557 sed -e "s/mtime[^ ]*//" $XMLCONFIG > $SECONDXMLCONFIG || return $?
558 echo "Generate the second XML configuration file"
560 # don't compare .xml mtime, it will always be different
561 if [ `sed -e "s/mtime[^ ]*//" $XMLCONFIG | diff - $SECONDXMLCONFIG | wc -l` -eq 0 ]; then
562 echo "Success:multiple invocations for lmc generate same XML file"
564 echo "Error: multiple invocations for lmc generate different XML file"
568 rm -f $XMLCONFIG $SECONDXMLCONFIG
569 XMLCONFIG=$OLDXMLCONFIG
571 run_test 13b "check lmc generates consistent .xml file"
576 # create xml file with --mkfsoptions for ost
577 echo "create xml file with --mkfsoptions for ost"
578 add_mds mds --dev $MDSDEV --size $MDSSIZE
579 add_lov lov1 mds --stripe_sz $STRIPE_BYTES\
580 --stripe_cnt $STRIPES_PER_OBJ --stripe_pattern 0
581 add_ost ost --lov lov1 --dev $OSTDEV --size $OSTSIZE \
582 --mkfsoptions "-Llabel_conf_14"
583 add_client client mds --lov lov1 --path $MOUNT
585 FOUNDSTRING=`awk -F"<" '/<mkfsoptions>/{print $2}' $XMLCONFIG`
586 EXPECTEDSTRING="mkfsoptions>-Llabel_conf_14"
587 if [ "$EXPECTEDSTRING" != "$FOUNDSTRING" ]; then
588 echo "Error: expected: $EXPECTEDSTRING; found: $FOUNDSTRING"
591 echo "Success:mkfsoptions for ost written to xml file correctly."
593 # mount lustre to test lconf mkfsoptions-parsing
597 mount_client $MOUNT || return $?
598 if [ -z "`do_facet ost1 dumpe2fs -h $OSTDEV | grep label_conf_14`" ]; then
599 echo "Error: the mkoptions not applied to mke2fs of ost."
603 echo "lconf mkfsoptions for ost success"
607 run_test 14 "test mkfsoptions of ost for lmc and lconf"
611 [ -f $MOUNTLUSTRE ] && echo "remove $MOUNTLUSTRE" && rm -f $MOUNTLUSTRE
612 if [ -f $MOUNTLUSTRE.sav ]; then
613 echo "return original $MOUNTLUSTRE.sav to $MOUNTLUSTRE"
614 mv $MOUNTLUSTRE.sav $MOUNTLUSTRE
618 # this only tests the kernel mount command, not anything about lustre.
620 MOUNTLUSTRE=${MOUNTLUSTRE:-/sbin/mount.lustre}
624 echo "mount lustre on ${MOUNT} without $MOUNTLUSTRE....."
625 if [ -f "$MOUNTLUSTRE" ]; then
626 echo "save $MOUNTLUSTRE to $MOUNTLUSTRE.sav"
627 mv $MOUNTLUSTRE $MOUNTLUSTRE.sav && trap cleanup_15 EXIT INT
628 if [ -f $MOUNTLUSTRE ]; then
629 skip "$MOUNTLUSTRE cannot be moved, skipping test"
634 mount_client $MOUNT && error "mount succeeded" && return 1
635 echo "mount lustre on $MOUNT without $MOUNTLUSTRE failed as expected"
639 run_test 15 "zconf-mount without /sbin/mount.lustre (should return error)"
641 # LOGS/PENDING do not exist anymore since CMD3
643 TMPMTPT="${TMP}/conf16"
645 if [ ! -e "$MDSDEV" ]; then
646 log "no $MDSDEV existing, so mount Lustre to create one"
648 check_mount || return 41
652 [ -f "$MDSDEV" ] && LOOPOPT="-o loop"
654 log "change the mode of $MDSDEV/OBJECTS to 555"
655 do_facet mds "mkdir -p $TMPMTPT &&
656 mount $LOOPOPT -t $FSTYPE $MDSDEV $TMPMTPT &&
657 chmod 555 $TMPMTPT/OBJECTS &&
658 umount $TMPMTPT" || return $?
660 log "mount Lustre to change the mode of OBJECTS, then umount Lustre"
662 check_mount || return 41
665 log "read the mode of OBJECTS and check if they has been changed properly"
666 EXPECTEDOBJECTSMODE=`do_facet mds "debugfs -R 'stat OBJECTS' $MDSDEV 2> /dev/null" | grep 'Mode: ' | sed -e "s/.*Mode: *//" -e "s/ *Flags:.*//"`
668 if [ "$EXPECTEDOBJECTSMODE" = "0777" ]; then
669 log "Success:Lustre change the mode of OBJECTS correctly"
671 error "Lustre does not change mode of OBJECTS properly"
674 run_test 16 "verify that lustre will correct the mode of OBJECTS"
677 if [ ! -e "$MDSDEV" ]; then
678 echo "no $MDSDEV existing, so mount Lustre to create one"
680 check_mount || return 41
684 echo "Remove mds config log"
685 do_facet mds "debugfs -w -R 'unlink CONFIGS/$FSNAME-MDT0000' $MDSDEV || return \$?" || return $?
688 start_mds && return 42
691 run_test 17 "Verify failed mds_postsetup won't fail assertion (2936) (should return errs)"
694 [ "$FSTYPE" != "ldiskfs" ] && skip "not needed for FSTYPE=$FSTYPE" && return
699 # check if current MDSSIZE is large enough
700 [ $MDSSIZE -ge $MIN ] && OK=1 && myMDSSIZE=$MDSSIZE && \
701 log "use MDSSIZE=$MDSSIZE"
703 # check if the global config has a large enough MDSSIZE
704 [ -z "$OK" -a ! -z "$STORED_MDSSIZE" ] && [ $STORED_MDSSIZE -ge $MIN ] && \
705 OK=1 && myMDSSIZE=$STORED_MDSSIZE && \
706 log "use STORED_MDSSIZE=$STORED_MDSSIZE"
708 # check if the block device is large enough
709 [ -z "$OK" -a -b $MDSDEV ] && \
710 [ "$(dd if=$MDSDEV of=/dev/null bs=1k count=1 skip=$MIN 2>&1 |
711 awk '($3 == "in") { print $1 }')" = "1+0" ] && OK=1 && \
712 myMDSSIZE=$MIN && log "use device $MDSDEV with MIN=$MIN"
714 # check if a loopback device has enough space for fs metadata (5%)
715 [ -z "$OK" ] && [ -f $MDSDEV -o ! -e $MDSDEV ] &&
716 SPACE=$(df -P $(dirname $MDSDEV) |
717 awk '($1 != "Filesystem") {print $4}') &&
718 [ $SPACE -gt $((MIN / 20)) ] && OK=1 && myMDSSIZE=$MIN && \
719 log "use file $MDSDEV with MIN=$MIN"
721 [ -z "$OK" ] && skip "$MDSDEV too small for ${MIN}kB MDS" && return
724 echo "mount mds with large journal..."
725 local OLD_MDS_MKFS_OPTS=$MDS_MKFS_OPTS
727 MDS_MKFS_OPTS="--mgs --mdt --fsname=$FSNAME --device-size=$myMDSSIZE --param sys.timeout=$TIMEOUT $MDSOPT"
730 echo "mount lustre system..."
732 check_mount || return 41
734 echo "check journal size..."
735 local FOUNDSIZE=`do_facet mds "debugfs -c -R 'stat <8>' $MDSDEV" | awk '/Size: / { print $NF; exit;}'`
736 if [ $FOUNDSIZE -gt $((32 * 1024 * 1024)) ]; then
737 log "Success: mkfs creates large journals. Size: $((FOUNDSIZE >> 20))M"
739 error "expected journal size > 32M, found $((FOUNDSIZE >> 20))M"
744 MDS_MKFS_OPTS=$OLD_MDS_MKFS_OPTS
747 run_test 18 "check mkfs creates large journals"
750 start_mds || return 1
751 stop_mds -f || return 2
753 run_test 19a "start/stop MDS without OSTs"
756 start_ost || return 1
757 stop_ost -f || return 2
759 run_test 19b "start/stop OSTs without MDS"
762 # first format the ost/mdt
766 check_mount || return 43
768 remount_client ro $MOUNT || return 44
769 touch $DIR/$tfile && echo "$DIR/$tfile created incorrectly" && return 45
770 [ -e $DIR/$tfile ] && echo "$DIR/$tfile exists incorrectly" && return 46
771 remount_client rw $MOUNT || return 47
773 [ ! -f $DIR/$tfile ] && echo "$DIR/$tfile missing" && return 48
774 MCNT=`grep -c $MOUNT /etc/mtab`
775 [ "$MCNT" -ne 1 ] && echo "$MOUNT in /etc/mtab $MCNT times" && return 49
780 run_test 20 "remount ro,rw mounts work and doesn't break /etc/mtab"
788 run_test 21a "start mds before ost, stop ost first"
796 run_test 21b "start ost before mds, stop mds first"
806 run_test 21c "start mds between two osts, stop mds last"
809 #reformat to remove all logs
812 echo Client mount before any osts are in the logs
814 check_mount && return 41
817 echo Client mount with ost in logs, but none running
821 # check_mount will block trying to contact ost
825 echo Client mount with a running ost
828 # if gss enabled, wait full time to let connection from
829 # mds to ost be established, due to the mismatch between
830 # initial connect timeout and gss context negotiation timeout.
831 # This perhaps could be remove after AT landed.
832 echo "sleep $((TIMEOUT + TIMEOUT + TIMEOUT))s"
833 sleep $((TIMEOUT + TIMEOUT + TIMEOUT))
836 check_mount || return 41
841 run_test 22 "start a client before osts (should return errs)"
843 test_23a() { # was test_23
847 # force down client so that recovering mds waits for reconnect
848 local running=$(grep -c $MOUNT /proc/mounts) || true
849 if [ $running -ne 0 ]; then
850 echo "Stopping client $MOUNT (opts: -f)"
854 # enter recovery on mds
856 # try to start a new client
857 mount_client $MOUNT &
859 MOUNT_PID=$(ps -ef | grep "t lustre" | grep -v grep | awk '{print $2}')
860 MOUNT_LUSTRE_PID=`ps -ef | grep mount.lustre | grep -v grep | awk '{print $2}'`
861 echo mount pid is ${MOUNT_PID}, mount.lustre pid is ${MOUNT_LUSTRE_PID}
863 ps --ppid $MOUNT_LUSTRE_PID
864 # FIXME why o why can't I kill these? Manual "ctrl-c" works...
865 kill -TERM $MOUNT_LUSTRE_PID
866 echo "waiting for mount to finish"
868 # we can not wait $MOUNT_PID because it is not a child of this shell
874 while [ "$WAIT" -lt "$MAX_WAIT" ]; do
876 PID1=$(ps -ef | awk '{print $2}' | grep -w $MOUNT_PID)
877 PID2=$(ps -ef | awk '{print $2}' | grep -w $MOUNT_LUSTRE_PID)
880 [ -z "$PID1" -a -z "$PID2" ] && break
881 echo "waiting for mount to finish ... "
882 WAIT=$(( WAIT + sleep))
884 [ "$WAIT" -eq "$MAX_WAIT" ] && error "MOUNT_PID $MOUNT_PID and \
885 MOUNT__LUSTRE_PID $MOUNT__LUSTRE_PID still not killed in $WAIT secs"
890 run_test 23a "interrupt client during recovery mount delay"
895 test_23b() { # was test_23
898 # Simulate -EINTR during mount OBD_FAIL_LDLM_CLOSE_THREAD
899 lctl set_param fail_loc=0x80000313
903 run_test 23b "Simulate -EINTR during mount"
905 fs2mds_HOST=$mds_HOST
906 fs2ost_HOST=$ost_HOST
910 echo "umount $MOUNT2 ..."
911 umount $MOUNT2 || true
912 echo "stopping fs2mds ..."
913 stop fs2mds -f || true
914 echo "stopping fs2ost ..."
915 stop fs2ost -f || true
922 [ -n "$ost1_HOST" ] && fs2ost_HOST=$ost1_HOST
923 if [ -z "$fs2ost_DEV" -o -z "$fs2mds_DEV" ]; then
924 do_facet $SINGLEMDS [ -b "$MDSDEV" ] && \
925 skip "mixed loopback and real device not working" && return
928 local fs2mdsdev=${fs2mds_DEV:-${MDSDEV}_2}
929 local fs2ostdev=${fs2ost_DEV:-$(ostdevname 1)_2}
931 # test 8-char fsname as well
932 local FSNAME2=test1234
933 add fs2mds $MDS_MKFS_OPTS --fsname=${FSNAME2} --nomgs --mgsnode=$MGSNID --reformat $fs2mdsdev || exit 10
935 add fs2ost $OST_MKFS_OPTS --fsname=${FSNAME2} --reformat $fs2ostdev || exit 10
938 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS && trap cleanup_24a EXIT INT
939 start fs2ost $fs2ostdev $OST_MOUNT_OPTS
941 mount -t lustre $MGSNID:/${FSNAME2} $MOUNT2 || return 1
943 check_mount || return 2
944 # files written on 1 should not show up on 2
945 cp /etc/passwd $DIR/$tfile
947 [ -e $MOUNT2/$tfile ] && error "File bleed" && return 7
950 cp /etc/passwd $MOUNT2/b || return 3
951 rm $MOUNT2/b || return 4
952 # 2 is actually mounted
953 grep $MOUNT2' ' /proc/mounts > /dev/null || return 5
955 facet_failover fs2mds
956 facet_failover fs2ost
959 # the MDS must remain up until last MDT
961 MDS=$(do_facet $SINGLEMDS "lctl get_param -n devices" | awk '($3 ~ "mdt" && $4 ~ "MDT") { print $4 }')
962 [ -z "$MDS" ] && error "No MDT" && return 8
964 cleanup_nocli || return 6
966 run_test 24a "Multiple MDTs on a single node"
969 if [ -z "$fs2mds_DEV" ]; then
970 do_facet $SINGLEMDS [ -b "$MDSDEV" ] && \
971 skip "mixed loopback and real device not working" && return
974 local fs2mdsdev=${fs2mds_DEV:-${MDSDEV}_2}
976 add fs2mds $MDS_MKFS_OPTS --fsname=${FSNAME}2 --mgs --reformat $fs2mdsdev || exit 10
978 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS && return 2
981 run_test 24b "Multiple MGSs on a single node (should return err)"
985 check_mount || return 2
986 local MODULES=$($LCTL modules | awk '{ print $2 }')
987 rmmod $MODULES 2>/dev/null || true
990 run_test 25 "Verify modules are referenced"
994 # we need modules before mount for sysctl, so make sure...
995 do_facet mds "lsmod | grep -q lustre || modprobe lustre"
996 #define OBD_FAIL_MDS_FS_SETUP 0x135
997 do_facet mds "lctl set_param fail_loc=0x80000135"
998 start_mds && echo MDS started && return 1
999 lctl get_param -n devices
1000 DEVS=$(lctl get_param -n devices | wc -l)
1001 [ $DEVS -gt 0 ] && return 2
1002 unload_modules || return 203
1004 run_test 26 "MDT startup failure cleans LOV (should return errs)"
1010 local ORIG=$(do_facet $myfacet "$TEST")
1011 if [ $# -gt 3 ]; then
1015 FINAL=$(($ORIG + 5))
1017 echo "Setting $PARAM from $ORIG to $FINAL"
1018 do_facet mds "$LCTL conf_param $PARAM=$FINAL" || error conf_param failed
1024 RESULT=$(do_facet $myfacet "$TEST")
1025 if [ $RESULT -eq $FINAL ]; then
1026 echo "Updated config after $WAIT sec (got $RESULT)"
1030 if [ $WAIT -eq $MAX ]; then
1031 echo "Config update not seen: wanted $FINAL got $RESULT"
1034 echo "Waiting $(($MAX - $WAIT)) secs for config update"
1039 start_ost || return 1
1040 start_mds || return 2
1041 echo "Requeue thread should have started: "
1042 ps -e | grep ll_cfg_requeue
1043 set_and_check ost1 "lctl get_param -n obdfilter.$FSNAME-OST0000.client_cache_seconds" "$FSNAME-OST0000.ost.client_cache_seconds" || return 3
1046 run_test 27a "Reacquire MGS lock if OST started first"
1051 set_and_check mds "lctl get_param -n mdt.$FSNAME-MDT0000.identity_acquire_expire" "$FSNAME-MDT0000.mdt.identity_acquire_expire" || return 3
1052 set_and_check client "lctl get_param -n mdc.$FSNAME-MDT0000-mdc-*.max_rpcs_in_flight" "$FSNAME-MDT0000.mdc.max_rpcs_in_flight" || return 4
1055 run_test 27b "Reacquire MGS lock after failover"
1059 TEST="lctl get_param -n llite.$FSNAME-*.max_read_ahead_whole_mb"
1062 FINAL=$(($ORIG + 10))
1063 set_and_check client "$TEST" "$FSNAME.llite.max_read_ahead_whole_mb" || return 3
1064 set_and_check client "$TEST" "$FSNAME.llite.max_read_ahead_whole_mb" || return 3
1065 umount_client $MOUNT || return 200
1068 if [ $RESULT -ne $FINAL ]; then
1069 echo "New config not seen: wanted $FINAL got $RESULT"
1072 echo "New config success: got $RESULT"
1076 run_test 28 "permanent parameter setting"
1079 [ "$OSTCOUNT" -lt "2" ] && skip "$OSTCOUNT < 2, skipping" && return
1080 setup > /dev/null 2>&1
1084 local PARAM="$FSNAME-OST0001.osc.active"
1085 local PROC_ACT="osc.$FSNAME-OST0001-osc-[^M]*.active"
1086 local PROC_UUID="osc.$FSNAME-OST0001-osc-[^M]*.ost_server_uuid"
1088 ACTV=$(lctl get_param -n $PROC_ACT)
1090 set_and_check client "lctl get_param -n $PROC_ACT" "$PARAM" $DEAC || return 2
1091 # also check ost_server_uuid status
1092 RESULT=$(lctl get_param -n $PROC_UUID | grep DEACTIV)
1093 if [ -z "$RESULT" ]; then
1094 echo "Live client not deactivated: $(lctl get_param -n $PROC_UUID)"
1097 echo "Live client success: got $RESULT"
1101 local MPROC="osc.$FSNAME-OST0001-osc-[M]*.active"
1106 RESULT=`do_facet mds " lctl get_param -n $MPROC"`
1107 [ ${PIPESTATUS[0]} = 0 ] || error "Can't read $MPROC"
1108 if [ $RESULT -eq $DEAC ]; then
1109 echo "MDT deactivated also after $WAIT sec (got $RESULT)"
1113 if [ $WAIT -eq $MAX ]; then
1114 echo "MDT not deactivated: wanted $DEAC got $RESULT"
1117 echo "Waiting $(($MAX - $WAIT)) secs for MDT deactivated"
1120 # test new client starts deactivated
1121 umount_client $MOUNT || return 200
1123 RESULT=$(lctl get_param -n $PROC_UUID | grep DEACTIV | grep NEW)
1124 if [ -z "$RESULT" ]; then
1125 echo "New client not deactivated from start: $(lctl get_param -n $PROC_UUID)"
1128 echo "New client success: got $RESULT"
1131 # make sure it reactivates
1132 set_and_check client "lctl get_param -n $PROC_ACT" "$PARAM" $ACTV || return 6
1134 umount_client $MOUNT
1137 #writeconf to remove all ost2 traces for subsequent tests
1143 run_test 29 "permanently remove an OST"
1148 TEST="lctl get_param -n llite.$FSNAME-*.max_read_ahead_whole_mb"
1150 for i in $(seq 1 20); do
1151 set_and_check client "$TEST" "$FSNAME.llite.max_read_ahead_whole_mb" $i || return 3
1153 # make sure client restart still works
1154 umount_client $MOUNT
1155 mount_client $MOUNT || return 4
1156 [ "$($TEST)" -ne "$i" ] && return 5
1157 set_and_check client "$TEST" "$FSNAME.llite.max_read_ahead_whole_mb" $ORIG || return 6
1160 run_test 30 "Big config llog"
1162 test_31() { # bug 10734
1163 # ipaddr must not exist
1164 mount -t lustre 4.3.2.1@tcp:/lustre $MOUNT || true
1167 run_test 31 "Connect to non-existent node (shouldn't crash)"
1170 # XXX - make this run on client-only systems with real hardware on
1172 # there appears to be a lot of assumption here about loopback
1174 # or maybe this test is just totally useless on a client-only system
1175 [ "$NETTYPE" = "tcp" ] || { skip "NETTYPE != tcp" && return 0; }
1176 [ "$mds_HOST" = "`hostname`" ] || { skip "remote MDS" && return 0; }
1177 [ "$ost_HOST" = "`hostname`" -o "$ost1_HOST" = "`hostname`" ] || \
1178 { skip "remote OST" && return 0; }
1180 [ -z "$TUNEFS" ] && skip "No tunefs" && return
1181 local DISK1_4=$LUSTRE/tests/disk1_4.zip
1182 [ ! -r $DISK1_4 ] && skip "Cant find $DISK1_4, skipping" && return
1184 local tmpdir=$TMP/conf32a
1185 unzip -o -j -d $tmpdir $DISK1_4 || { skip "Cant unzip $DISK1_4, skipping" && return ; }
1187 lctl set_param debug=$PTLDEBUG
1189 $TUNEFS $tmpdir/mds || error "tunefs failed"
1190 # nids are wrong, so client wont work, but server should start
1191 start mds $tmpdir/mds "-o loop,exclude=lustre-OST0000" || return 3
1192 local UUID=$(lctl get_param -n mdt.lustre-MDT0000.uuid)
1194 [ "$UUID" == "mdsA_UUID" ] || error "UUID is wrong: $UUID"
1196 $TUNEFS --mgsnode=`hostname` $tmpdir/ost1 || error "tunefs failed"
1197 start ost1 $tmpdir/ost1 "-o loop" || return 5
1198 UUID=$(lctl get_param -n obdfilter.lustre-OST0000.uuid)
1200 [ "$UUID" == "ost1_UUID" ] || error "UUID is wrong: $UUID"
1202 local NID=$($LCTL list_nids | head -1)
1204 echo "OSC changes should return err:"
1205 $LCTL conf_param lustre-OST0000.osc.max_dirty_mb=15 && return 7
1206 $LCTL conf_param lustre-OST0000.failover.node=$NID && return 8
1208 echo "MDC changes should succeed:"
1209 $LCTL conf_param lustre-MDT0000.mdc.max_rpcs_in_flight=9 || return 9
1210 $LCTL conf_param lustre-MDT0000.failover.node=$NID || return 10
1213 # With a new good MDT failover nid, we should be able to mount a client
1214 # (but it cant talk to OST)
1215 local OLDMOUNTOPT=$MOUNTOPT
1216 MOUNTOPT="exclude=lustre-OST0000"
1218 MOUNTOPT=$OLDMOUNTOPT
1219 set_and_check client "lctl get_param -n mdc.*.max_rpcs_in_flight" "lustre-MDT0000.mdc.max_rpcs_in_flight" ||
1222 zconf_umount `hostname` $MOUNT -f
1226 # mount a second time to make sure we didnt leave upgrade flag on
1228 $TUNEFS --dryrun $tmpdir/mds || error "tunefs failed"
1230 start mds $tmpdir/mds "-o loop,exclude=lustre-OST0000" || return 12
1233 rm -rf $tmpdir || true # true is only for TMP on NFS
1235 run_test 32a "Upgrade from 1.4 (not live)"
1238 # XXX - make this run on client-only systems with real hardware on
1240 # there appears to be a lot of assumption here about loopback
1242 # or maybe this test is just totally useless on a client-only system
1243 [ "$NETTYPE" = "tcp" ] || { skip "NETTYPE != tcp" && return 0; }
1244 [ "$mds_HOST" = "`hostname`" ] || { skip "remote MDS" && return 0; }
1245 [ "$ost_HOST" = "`hostname`" -o "$ost1_HOST" = "`hostname`" ] || \
1246 { skip "remote OST" && return 0; }
1248 [ -z "$TUNEFS" ] && skip "No tunefs" && return
1249 local DISK1_4=$LUSTRE/tests/disk1_4.zip
1250 [ ! -r $DISK1_4 ] && skip "Cant find $DISK1_4, skipping" && return
1252 local tmpdir=$TMP/conf32b
1253 unzip -o -j -d $tmpdir $DISK1_4 || { skip "Cant unzip $DISK1_4, skipping" && return ; }
1255 lctl set_param debug=$PTLDEBUG
1258 # writeconf will cause servers to register with their current nids
1259 $TUNEFS --writeconf --fsname=$NEWNAME $tmpdir/mds || error "tunefs failed"
1260 start mds $tmpdir/mds "-o loop" || return 3
1261 local UUID=$(lctl get_param -n mdt.${NEWNAME}-MDT0000.uuid)
1263 [ "$UUID" == "mdsA_UUID" ] || error "UUID is wrong: $UUID"
1265 $TUNEFS --mgsnode=`hostname` --fsname=$NEWNAME --writeconf $tmpdir/ost1 || error "tunefs failed"
1266 start ost1 $tmpdir/ost1 "-o loop" || return 5
1267 UUID=$(lctl get_param -n obdfilter.${NEWNAME}-OST0000.uuid)
1269 [ "$UUID" == "ost1_UUID" ] || error "UUID is wrong: $UUID"
1271 echo "OSC changes should succeed:"
1272 $LCTL conf_param ${NEWNAME}-OST0000.osc.max_dirty_mb=15 || return 7
1273 $LCTL conf_param ${NEWNAME}-OST0000.failover.node=$NID || return 8
1275 echo "MDC changes should succeed:"
1276 $LCTL conf_param ${NEWNAME}-MDT0000.mdc.max_rpcs_in_flight=9 || return 9
1279 # MDT and OST should have registered with new nids, so we should have
1280 # a fully-functioning client
1281 echo "Check client and old fs contents"
1286 set_and_check client "lctl get_param -n mdc.*.max_rpcs_in_flight" "${NEWNAME}-MDT0000.mdc.max_rpcs_in_flight" || return 11
1287 [ "$(cksum $MOUNT/passwd | cut -d' ' -f 1,2)" == "2479747619 779" ] || return 12
1291 rm -rf $tmpdir || true # true is only for TMP on NFS
1293 run_test 32b "Upgrade from 1.4 with writeconf"
1295 test_33a() { # bug 12333, was test_33
1297 local FSNAME2=test-123
1298 [ -n "$ost1_HOST" ] && fs2ost_HOST=$ost1_HOST
1300 if [ -z "$fs2ost_DEV" -o -z "$fs2mds_DEV" ]; then
1301 do_facet $SINGLEMDS [ -b "$MDSDEV" ] && \
1302 skip "mixed loopback and real device not working" && return
1305 local fs2mdsdev=${fs2mds_DEV:-${MDSDEV}_2}
1306 local fs2ostdev=${fs2ost_DEV:-$(ostdevname 1)_2}
1307 add fs2mds $MDS_MKFS_OPTS --fsname=${FSNAME2} --reformat $fs2mdsdev || exit 10
1308 add fs2ost $OST_MKFS_OPTS --fsname=${FSNAME2} --index=8191 --mgsnode=$MGSNID --reformat $fs2ostdev || exit 10
1310 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS && trap cleanup_24a EXIT INT
1311 start fs2ost $fs2ostdev $OST_MOUNT_OPTS
1312 do_facet mds "$LCTL conf_param $FSNAME2.sys.timeout=200" || rc=1
1314 mount -t lustre $MGSNID:/${FSNAME2} $MOUNT2 || rc=2
1320 rm -rf $MOUNT2 $fs2mdsdev $fs2ostdev
1321 cleanup_nocli || rc=6
1324 run_test 33a "Mount ost with a large index number"
1326 test_33b() { # was test_34
1329 do_facet client dd if=/dev/zero of=$MOUNT/24 bs=1024k count=1
1330 # Drop lock cancelation reply during umount
1331 #define OBD_FAIL_LDLM_CANCEL 0x304
1332 do_facet client lctl set_param fail_loc=0x80000304
1333 #lctl set_param debug=-1
1334 umount_client $MOUNT
1337 run_test 33b "Drop cancel during umount"
1341 do_facet client "sh runmultiop_bg_pause $DIR/file O_c"
1342 manual_umount_client
1344 do_facet client killall -USR1 multiop
1345 if [ $rc -eq 0 ]; then
1346 error "umount not fail!"
1351 run_test 34a "umount with opened file should be fail"
1356 touch $DIR/$tfile || return 1
1357 stop_mds --force || return 2
1359 manual_umount_client --force
1361 if [ $rc -ne 0 ]; then
1362 error "mtab after failed umount - rc $rc"
1368 run_test 34b "force umount with failed mds should be normal"
1372 touch $DIR/$tfile || return 1
1373 stop_ost --force || return 2
1375 manual_umount_client --force
1377 if [ $rc -ne 0 ]; then
1378 error "mtab after failed umount - rc $rc"
1384 run_test 34c "force umount with failed ost should be normal"
1386 test_35() { # bug 12459
1389 DBG_SAVE="`lctl get_param -n debug`"
1390 lctl set_param debug="ha"
1392 log "Set up a fake failnode for the MDS"
1394 do_facet mds $LCTL conf_param ${FSNAME}-MDT0000.failover.node=$FAKENID || return 4
1396 log "Wait for RECONNECT_INTERVAL seconds (10s)"
1399 MSG="conf-sanity.sh test_35 `date +%F%kh%Mm%Ss`"
1402 log "Stopping the MDT:"
1403 stop_mds || return 5
1405 df $MOUNT > /dev/null 2>&1 &
1407 log "Restarting the MDT:"
1408 start_mds || return 6
1409 log "Wait for df ($DFPID) ... "
1412 lctl set_param debug="$DBG_SAVE"
1414 # retrieve from the log the first server that the client tried to
1415 # contact after the connection loss
1416 $LCTL dk $TMP/lustre-log-$TESTNAME.log
1417 NEXTCONN=`awk "/${MSG}/ {start = 1;}
1418 /import_select_connection.*${FSNAME}-MDT0000-mdc.* using connection/ {
1420 if (\\\$NF ~ /$FAKENID/)
1426 }" $TMP/lustre-log-$TESTNAME.log`
1427 [ "$NEXTCONN" != "0" ] && log "The client didn't try to reconnect to the last active server (tried ${NEXTCONN} instead)" && return 7
1430 run_test 35 "Reconnect to the last active server first"
1434 local FSNAME2=test1234
1435 local fs3ost_HOST=$ost_HOST
1437 [ -n "$ost1_HOST" ] && fs2ost_HOST=$ost1_HOST && fs3ost_HOST=$ost1_HOST
1440 if [ -z "$fs2ost_DEV" -o -z "$fs2mds_DEV" -o -z "$fs3ost_DEV" ]; then
1441 do_facet $SINGLEMDS [ -b "$MDSDEV" ] && \
1442 skip "mixed loopback and real device not working" && return
1444 [ $OSTCOUNT -lt 2 ] && skip "skipping test for single OST" && return
1446 [ "$ost_HOST" = "`hostname`" -o "$ost1_HOST" = "`hostname`" ] || \
1447 { skip "remote OST" && return 0; }
1449 local fs2mdsdev=${fs2mds_DEV:-${MDSDEV}_2}
1450 local fs2ostdev=${fs2ost_DEV:-$(ostdevname 1)_2}
1451 local fs3ostdev=${fs3ost_DEV:-$(ostdevname 2)_2}
1452 add fs2mds $MDS_MKFS_OPTS --fsname=${FSNAME2} --reformat $fs2mdsdev || exit 10
1453 # XXX after we support non 4K disk blocksize, change following --mkfsoptions with
1455 add fs2ost $OST_MKFS_OPTS --mkfsoptions='-b4096' --fsname=${FSNAME2} --mgsnode=$MGSNID --reformat $fs2ostdev || exit 10
1456 add fs3ost $OST_MKFS_OPTS --mkfsoptions='-b4096' --fsname=${FSNAME2} --mgsnode=$MGSNID --reformat $fs3ostdev || exit 10
1458 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS
1459 start fs2ost $fs2ostdev $OST_MOUNT_OPTS
1460 start fs3ost $fs3ostdev $OST_MOUNT_OPTS
1462 mount -t lustre $MGSNID:/${FSNAME2} $MOUNT2 || return 1
1464 sleep 5 # until 11778 fixed
1466 dd if=/dev/zero of=$MOUNT2/$tfile bs=1M count=7 || return 2
1468 BKTOTAL=`lctl get_param -n obdfilter.*.kbytestotal | awk 'BEGIN{total=0}; {total+=$1}; END{print total}'`
1469 BKFREE=`lctl get_param -n obdfilter.*.kbytesfree | awk 'BEGIN{free=0}; {free+=$1}; END{print free}'`
1470 BKAVAIL=`lctl get_param -n obdfilter.*.kbytesavail | awk 'BEGIN{avail=0}; {avail+=$1}; END{print avail}'`
1471 STRING=`df -P $MOUNT2 | tail -n 1 | awk '{print $2","$3","$4}'`
1472 DFTOTAL=`echo $STRING | cut -d, -f1`
1473 DFUSED=`echo $STRING | cut -d, -f2`
1474 DFAVAIL=`echo $STRING | cut -d, -f3`
1475 DFFREE=$(($DFTOTAL - $DFUSED))
1477 ALLOWANCE=$((64 * $OSTCOUNT))
1479 if [ $DFTOTAL -lt $(($BKTOTAL - $ALLOWANCE)) ] ||
1480 [ $DFTOTAL -gt $(($BKTOTAL + $ALLOWANCE)) ] ; then
1481 echo "**** FAIL: df total($DFTOTAL) mismatch OST total($BKTOTAL)"
1484 if [ $DFFREE -lt $(($BKFREE - $ALLOWANCE)) ] ||
1485 [ $DFFREE -gt $(($BKFREE + $ALLOWANCE)) ] ; then
1486 echo "**** FAIL: df free($DFFREE) mismatch OST free($BKFREE)"
1489 if [ $DFAVAIL -lt $(($BKAVAIL - $ALLOWANCE)) ] ||
1490 [ $DFAVAIL -gt $(($BKAVAIL + $ALLOWANCE)) ] ; then
1491 echo "**** FAIL: df avail($DFAVAIL) mismatch OST avail($BKAVAIL)"
1496 stop fs3ost -f || return 200
1497 stop fs2ost -f || return 201
1498 stop fs2mds -f || return 202
1499 rm -rf $MOUNT2 $fs2mdsdev $fs2ostdev $fs3ostdev
1500 unload_modules || return 203
1503 run_test 36 "df report consistency on OSTs with different block size"
1506 [ -n "$CLIENTONLY" -o -n "$CLIENTMODSONLY" ] && skip "client only testing" && return 0
1507 LOCAL_MDSDEV="$TMP/mdt.img"
1508 SYM_MDSDEV="$TMP/sym_mdt.img"
1510 echo "MDS : $LOCAL_MDSDEV"
1511 echo "SYMLINK : $SYM_MDSDEV"
1515 mkfs.lustre --reformat --fsname=lustre --mdt --mgs --device-size=9000 $LOCAL_MDSDEV ||
1516 error "mkfs.lustre $LOCAL_MDSDEV failed"
1517 ln -s $LOCAL_MDSDEV $SYM_MDSDEV
1519 echo "mount symlink device - $SYM_MDSDEV"
1521 mount_op=`mount -v -t lustre -o loop $SYM_MDSDEV ${MOUNT%/*}/mds 2>&1 | grep "unable to set tunable"`
1522 umount -d ${MOUNT%/*}/mds
1523 rm -f $LOCAL_MDSDEV $SYM_MDSDEV
1525 if [ -n "$mount_op" ]; then
1526 error "**** FAIL: set tunables failed for symlink device"
1530 run_test 37 "verify set tunables works for symlink device"
1532 test_38() { # bug 14222
1537 FILES=`find $SRC -type f -mtime +1 | head -n $COUNT`
1538 log "copying $(echo $FILES | wc -w) files to $DIR/$tdir"
1540 tar cf - $FILES | tar xf - -C $DIR/$tdir || \
1541 error "copying $SRC to $DIR/$tdir"
1543 umount_client $MOUNT
1545 log "rename lov_objid file on MDS"
1546 rm -f $TMP/lov_objid.orig
1547 do_facet mds "debugfs -c -R \\\"dump lov_objid $TMP/lov_objid.orig\\\" $MDSDEV"
1548 do_facet mds "debugfs -w -R \\\"rm lov_objid\\\" $MDSDEV"
1550 do_facet mds "od -Ax -td8 $TMP/lov_objid.orig"
1551 # check create in mds_lov_connect
1555 [ $V ] && log "verifying $DIR/$tdir/$f"
1556 diff -q $f $DIR/$tdir/$f || ERROR=y
1558 do_facet mds "debugfs -c -R \\\"dump lov_objid $TMP/lov_objid.new\\\" $MDSDEV"
1559 do_facet mds "od -Ax -td8 $TMP/lov_objid.new"
1560 [ "$ERROR" = "y" ] && error "old and new files are different after connect" || true
1562 # check it's updates in sync
1563 umount_client $MOUNT
1566 do_facet mds dd if=/dev/zero of=$TMP/lov_objid.clear bs=4096 count=1
1567 do_facet mds "debugfs -w -R \\\"rm lov_objid\\\" $MDSDEV"
1568 do_facet mds "debugfs -w -R \\\"write $TMP/lov_objid.clear lov_objid\\\" $MDSDEV "
1573 [ $V ] && log "verifying $DIR/$tdir/$f"
1574 diff -q $f $DIR/$tdir/$f || ERROR=y
1576 do_facet mds "debugfs -c -R \\\"dump lov_objid $TMP/lov_objid.new1\\\" $MDSDEV"
1577 do_facet mds "od -Ax -td8 $TMP/lov_objid.new1"
1578 umount_client $MOUNT
1580 [ "$ERROR" = "y" ] && error "old and new files are different after sync" || true
1582 log "files compared the same"
1585 run_test 38 "MDS recreates missing lov_objid file from OST data"
1591 perl $SRCDIR/leak_finder.pl $TMP/debug 2>&1 | egrep '*** Leak:' &&
1592 error "memory leak detected" || true
1594 run_test 39 "leak_finder recognizes both LUSTRE and LNET malloc messages"
1596 test_40() { # bug 15759
1598 #define OBD_FAIL_TGT_TOOMANY_THREADS 0x706
1599 do_facet mds "sysctl -w lustre.fail_loc=0x80000706"
1603 run_test 40 "race during service thread startup"
1605 test_41() { #bug 14134
1607 start mds $MDSDEV $MDS_MOUNT_OPTS -o nosvc -n
1608 start ost1 `ostdevname 1` $OST_MOUNT_OPTS
1609 start mds $MDSDEV $MDS_MOUNT_OPTS -o nomgs
1611 mount_client $MOUNT || return 1
1614 echo "blah blah" > $MOUNT/$tfile
1617 umount_client $MOUNT
1618 stop ost1 -f || return 201
1619 stop mds -f || return 202
1620 stop mds -f || return 203
1621 unload_modules || return 204
1624 run_test 41 "mount mds with --nosvc and --nomgs"
1626 test_42() { #bug 14693
1628 check_mount || return 2
1629 do_facet client lctl conf_param lustre.llite.some_wrong_param=10
1630 umount_client $MOUNT
1631 mount_client $MOUNT || return 1
1635 run_test 42 "invalid config param should not prevent client from mounting"
1637 umount_client $MOUNT
1641 equals_msg `basename $0`: test complete
1642 [ -f "$TESTSUITELOG" ] && cat $TESTSUITELOG || true