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="9 10 11 12 13 13b 14 15"
17 # bug number for skipped test: 2661 13709 10510 12743
18 ALWAYS_EXCEPT=" $CONF_SANITY_EXCEPT $MOUNTCONFSKIP 16 22 23 36"
19 # UPDATE THE COMMENT ABOVE WITH BUG NUMBERS WHEN CHANGING ALWAYS_EXCEPT!
22 PATH=$PWD/$SRCDIR:$SRCDIR:$SRCDIR/../utils:$PATH
24 PTLDEBUG=${PTLDEBUG:--1}
26 LUSTRE=${LUSTRE:-`dirname $0`/..}
27 RLUSTRE=${RLUSTRE:-$LUSTRE}
28 MOUNTLUSTRE=${MOUNTLUSTRE:-/sbin/mount.lustre}
29 MKFSLUSTRE=${MKFSLUSTRE:-/usr/sbin/mkfs.lustre}
32 . $LUSTRE/tests/test-framework.sh
34 # use small MDS + OST size to speed formatting time
37 . ${CONFIG:=$LUSTRE/tests/cfg/local.sh}
48 # who knows if/where $TUNEFS is installed? Better reformat if it fails...
49 do_facet ${facet} "$TUNEFS --writeconf $MDSDEV" || echo "tunefs failed, reformatting instead" && reformat
54 # The MGS must be started before the OSTs for a new fs, so start
55 # and stop to generate the startup logs.
64 echo "start mds service on `facet_active_host mds`"
65 start mds $MDSDEV $MDS_MOUNT_OPTS || return 94
69 echo "stop mds service on `facet_active_host mds`"
70 # These tests all use non-failover stop
71 stop mds -f || return 97
75 echo "start ost1 service on `facet_active_host ost1`"
76 start ost1 `ostdevname 1` $OST_MOUNT_OPTS || return 95
80 echo "stop ost1 service on `facet_active_host ost1`"
81 # These tests all use non-failover stop
82 stop ost1 -f || return 98
86 echo "start ost2 service on `facet_active_host ost2`"
87 start ost2 `ostdevname 2` $OST_MOUNT_OPTS || return 92
91 echo "stop ost2 service on `facet_active_host ost2`"
92 # These tests all use non-failover stop
93 stop ost2 -f || return 93
98 echo "mount $FSNAME on ${MOUNTPATH}....."
99 zconf_mount `hostname` $MOUNTPATH || return 96
103 local SAVEMOUNTOPT=$MOUNTOPT
104 MOUNTOPT="remount,$1"
106 echo "remount '$1' lustre on ${MOUNTPATH}....."
107 zconf_mount `hostname` $MOUNTPATH || return 96
108 MOUNTOPT=$SAVEMOUNTOPT
113 echo "umount lustre on ${MOUNTPATH}....."
114 zconf_umount `hostname` $MOUNTPATH || return 97
117 manual_umount_client(){
118 echo "manual umount lustre on ${MOUNT}...."
119 do_facet client "umount -d $MOUNT"
129 stop_mds || return 201
130 stop_ost || return 202
131 unload_modules || return 203
135 umount_client $MOUNT || return 200
136 cleanup_nocli || return $?
140 do_facet client "cp /etc/passwd $DIR/a" || return 71
141 do_facet client "rm $DIR/a" || return 72
142 # make sure lustre is actually mounted (touch will block,
143 # but grep won't, so do it after)
144 do_facet client "grep $MOUNT' ' /proc/mounts > /dev/null" || return 73
145 echo "setup single mount lustre success"
149 do_facet client "touch $DIR/a" || return 71
150 do_facet client "rm $DIR/a" || return 72
151 do_facet client "touch $DIR2/a" || return 73
152 do_facet client "rm $DIR2/a" || return 74
153 echo "setup double mount lustre success"
158 if [ "$ONLY" == "setup" ]; then
163 if [ "$ONLY" == "cleanup" ]; then
168 #create single point mountpoint
176 check_mount || return 41
179 run_test 0 "single mount setup"
183 echo "start ost second time..."
185 check_mount || return 42
188 run_test 1 "start up ost twice (should return errors)"
193 echo "start mds second time.."
196 check_mount || return 43
199 run_test 2 "start up mds twice (should return err)"
203 #mount.lustre returns an error if already in mtab
204 mount_client $MOUNT && return $?
205 check_mount || return 44
208 run_test 3 "mount client twice (should return err)"
212 touch $DIR/$tfile || return 85
216 # ok for ost to fail shutdown
217 if [ 202 -ne $eno ]; then
222 run_test 4 "force cleanup ost, then cleanup"
226 touch $DIR/$tfile || return 1
227 stop_mds -f || return 2
229 # cleanup may return an error from the failed
230 # disconnects; for now I'll consider this successful
231 # if all the modules have unloaded.
235 echo "killing umount"
236 kill -TERM $UMOUNT_PID
237 echo "waiting for umount to finish"
239 if grep " $MOUNT " /etc/mtab; then
240 echo "test 5: mtab after failed umount"
244 echo "killing umount"
245 kill -TERM $UMOUNT_PID
246 echo "waiting for umount to finish"
248 grep " $MOUNT " /etc/mtab && echo "test 5: mtab after second umount" && return 11
252 # stop_mds is a no-op here, and should not fail
253 cleanup_nocli || return $?
254 # df may have lingering entry
256 # mtab may have lingering entry
257 grep -v $MOUNT" " /etc/mtab > $TMP/mtabtemp
258 mv $TMP/mtabtemp /etc/mtab
260 run_test 5 "force cleanup mds, then cleanup"
264 [ -d $MOUNT ] || mkdir -p $MOUNT
265 grep " $MOUNT " /etc/mtab && echo "test 5b: mtab before mount" && return 10
266 mount_client $MOUNT && return 1
267 grep " $MOUNT " /etc/mtab && echo "test 5b: mtab after failed mount" && return 11
269 # stop_mds is a no-op here, and should not fail
270 cleanup_nocli || return $?
273 run_test 5b "mds down, cleanup after failed mount (bug 2712) (should return errs)"
278 [ -d $MOUNT ] || mkdir -p $MOUNT
279 grep " $MOUNT " /etc/mtab && echo "test 5c: mtab before mount" && return 10
280 mount -t lustre $MGSNID:/wrong.$FSNAME $MOUNT || :
281 grep " $MOUNT " /etc/mtab && echo "test 5c: mtab after failed mount" && return 11
283 cleanup_nocli || return $?
285 run_test 5c "cleanup after failed mount (bug 2712) (should return errs)"
291 grep " $MOUNT " /etc/mtab && echo "test 5d: mtab before mount" && return 10
292 mount_client $MOUNT || return 1
294 grep " $MOUNT " /etc/mtab && echo "test 5d: mtab after unmount" && return 11
297 run_test 5d "mount with ost down"
303 #define OBD_FAIL_PTLRPC_DELAY_SEND 0x506
304 do_facet client "sysctl -w lustre.fail_loc=0x80000506"
305 grep " $MOUNT " /etc/mtab && echo "test 5e: mtab before mount" && return 10
306 mount_client $MOUNT || echo "mount failed (not fatal)"
308 grep " $MOUNT " /etc/mtab && echo "test 5e: mtab after unmount" && return 11
311 run_test 5e "delayed connect, don't crash (bug 10268)"
316 mount_client ${MOUNT} || return 87
317 touch $DIR/a || return 86
320 run_test 6 "manual umount, then mount again"
325 cleanup_nocli || return $?
327 run_test 7 "manual umount, then cleanup"
332 check_mount2 || return 45
333 umount_client $MOUNT2
336 run_test 8 "double mount setup"
341 do_facet ost1 sysctl lnet.debug=\'inode trace\' || return 1
342 do_facet ost1 sysctl lnet.subsystem_debug=\'mds ost\' || return 1
344 CHECK_PTLDEBUG="`do_facet ost1 sysctl -n lnet.debug`"
345 if [ "$CHECK_PTLDEBUG" ] && [ "$CHECK_PTLDEBUG" = "trace inode" ];then
346 echo "lnet.debug success"
348 echo "lnet.debug: want 'trace inode', have '$CHECK_PTLDEBUG'"
351 CHECK_SUBSYS="`do_facet ost1 sysctl -n lnet.subsystem_debug`"
352 if [ "$CHECK_SUBSYS" ] && [ "$CHECK_SUBSYS" = "mds ost" ]; then
353 echo "lnet.subsystem_debug success"
355 echo "lnet.subsystem_debug: want 'mds ost', have '$CHECK_SUBSYS'"
358 stop_ost || return $?
361 run_test 9 "test ptldebug and subsystem for mkfs"
364 echo "generate configuration with the same name for node and mds"
365 OLDXMLCONFIG=$XMLCONFIG
366 XMLCONFIG="broken.xml"
367 [ -f "$XMLCONFIG" ] && rm -f $XMLCONFIG
371 echo "the name for node and mds is the same"
372 do_lmc --add mds --node ${facet}_facet --mds ${facet}_facet \
373 --dev $MDSDEV --size $MDSSIZE || return $?
374 do_lmc --add lov --mds ${facet}_facet --lov lov1 --stripe_sz \
375 $STRIPE_BYTES --stripe_cnt $STRIPES_PER_OBJ \
376 --stripe_pattern 0 || return $?
377 add_ost ost --lov lov1 --dev $OSTDEV --size $OSTSIZE
379 add_facet $facet --lustre_upcall $UPCALL
380 do_lmc --add mtpt --node ${facet}_facet --mds mds_facet \
381 --lov lov1 --path $MOUNT
387 check_mount || return 41
391 XMLCONFIG=$OLDXMLCONFIG
393 run_test 10 "mount lustre with the same name for node and mds"
396 OLDXMLCONFIG=$XMLCONFIG
397 XMLCONFIG="conf11.xml"
399 [ -f "$XMLCONFIG" ] && rm -f $XMLCONFIG
400 add_mds mds --dev $MDSDEV --size $MDSSIZE
401 add_ost ost --dev $OSTDEV --size $OSTSIZE
402 add_client client mds --path $MOUNT --ost ost_svc || return $?
403 echo "Default lov config success!"
405 [ -f "$XMLCONFIG" ] && rm -f $XMLCONFIG
406 add_mds mds --dev $MDSDEV --size $MDSSIZE
407 add_ost ost --dev $OSTDEV --size $OSTSIZE
408 add_client client mds --path $MOUNT && return $?
409 echo "--add mtpt with neither --lov nor --ost will return error"
413 XMLCONFIG=$OLDXMLCONFIG
415 run_test 11 "use default lov configuration (should return error)"
418 OLDXMLCONFIG=$XMLCONFIG
419 XMLCONFIG="batch.xml"
420 BATCHFILE="batchfile"
423 [ -f "$XMLCONFIG" ] && rm -f $XMLCONFIG
424 [ -f "$BATCHFILE" ] && rm -f $BATCHFILE
425 echo "--add net --node $HOSTNAME --nid $HOSTNAME --nettype tcp" > $BATCHFILE
426 echo "--add mds --node $HOSTNAME --mds mds1 --mkfsoptions \"-I 128\"" >> $BATCHFILE
427 # --mkfsoptions "-I 128"
428 do_lmc -m $XMLCONFIG --batch $BATCHFILE || return $?
429 if [ `sed -n '/>-I 128</p' $XMLCONFIG | wc -l` -eq 1 ]; then
430 echo "matched double quote success"
432 echo "matched double quote fail"
437 echo "--add net --node $HOSTNAME --nid $HOSTNAME --nettype tcp" > $BATCHFILE
438 echo "--add mds --node $HOSTNAME --mds mds1 --mkfsoptions \"-I 128" >> $BATCHFILE
439 # --mkfsoptions "-I 128
440 do_lmc -m $XMLCONFIG --batch $BATCHFILE && return $?
441 echo "unmatched double quote should return error"
445 echo "--add net --node $HOSTNAME --nid $HOSTNAME --nettype tcp" > $BATCHFILE
446 echo "--add mds --node $HOSTNAME --mds mds1 --mkfsoptions '-I 128'" >> $BATCHFILE
447 # --mkfsoptions '-I 128'
448 do_lmc -m $XMLCONFIG --batch $BATCHFILE || return $?
449 if [ `sed -n '/>-I 128</p' $XMLCONFIG | wc -l` -eq 1 ]; then
450 echo "matched single quote success"
452 echo "matched single quote fail"
457 echo "--add net --node $HOSTNAME --nid $HOSTNAME --nettype tcp" > $BATCHFILE
458 echo "--add mds --node $HOSTNAME --mds mds1 --mkfsoptions '-I 128" >> $BATCHFILE
459 # --mkfsoptions '-I 128
460 do_lmc -m $XMLCONFIG --batch $BATCHFILE && return $?
461 echo "unmatched single quote should return error"
465 echo "--add net --node $HOSTNAME --nid $HOSTNAME --nettype tcp" > $BATCHFILE
466 echo "--add mds --node $HOSTNAME --mds mds1 --mkfsoptions \-\I\ \128" >> $BATCHFILE
467 # --mkfsoptions \-\I\ \128
468 do_lmc -m $XMLCONFIG --batch $BATCHFILE || return $?
469 if [ `sed -n '/>-I 128</p' $XMLCONFIG | wc -l` -eq 1 ]; then
470 echo "backslash followed by a whitespace/letter success"
472 echo "backslash followed by a whitespace/letter fail"
477 echo "--add net --node $HOSTNAME --nid $HOSTNAME --nettype tcp" > $BATCHFILE
478 echo "--add mds --node $HOSTNAME --mds mds1 --mkfsoptions -I\ 128\\" >> $BATCHFILE
479 # --mkfsoptions -I\ 128\
480 do_lmc -m $XMLCONFIG --batch $BATCHFILE && return $?
481 echo "backslash followed by nothing should return error"
484 XMLCONFIG=$OLDXMLCONFIG
486 run_test 12 "lmc --batch, with single/double quote, backslash in batchfile"
489 OLDXMLCONFIG=$XMLCONFIG
490 XMLCONFIG="conf13-1.xml"
492 # check long uuid will be truncated properly and uniquely
493 echo "To generate XML configuration file(with long ost name): $XMLCONFIG"
494 [ -f "$XMLCONFIG" ] && rm -f $XMLCONFIG
495 do_lmc --add net --node $HOSTNAME --nid $HOSTNAME --nettype tcp
496 do_lmc --add mds --node $HOSTNAME --mds mds1_name_longer_than_31characters
497 do_lmc --add mds --node $HOSTNAME --mds mds2_name_longer_than_31characters
498 if [ ! -f "$XMLCONFIG" ]; then
499 echo "Error:no file $XMLCONFIG created!"
502 EXPECTEDMDS1UUID="e_longer_than_31characters_UUID"
503 EXPECTEDMDS2UUID="longer_than_31characters_UUID_2"
504 FOUNDMDS1UUID=`awk -F"'" '/<mds .*uuid=/' $XMLCONFIG | sed -n '1p' \
505 | sed "s/ /\n\r/g" | awk -F"'" '/uuid=/{print $2}'`
506 FOUNDMDS2UUID=`awk -F"'" '/<mds .*uuid=/' $XMLCONFIG | sed -n '2p' \
507 | sed "s/ /\n\r/g" | awk -F"'" '/uuid=/{print $2}'`
508 [ -z "$FOUNDMDS1UUID" ] && echo "MDS1 UUID empty" && return 1
509 [ -z "$FOUNDMDS2UUID" ] && echo "MDS2 UUID empty" && return 1
510 if ([ $EXPECTEDMDS1UUID = $FOUNDMDS1UUID ] && [ $EXPECTEDMDS2UUID = $FOUNDMDS2UUID ]) || \
511 ([ $EXPECTEDMDS1UUID = $FOUNDMDS2UUID ] && [ $EXPECTEDMDS2UUID = $FOUNDMDS1UUID ]); then
512 echo "Success:long uuid truncated successfully and being unique."
514 echo "Error:expected uuid for mds1 and mds2: $EXPECTEDMDS1UUID; $EXPECTEDMDS2UUID"
515 echo "but: found uuid for mds1 and mds2: $FOUNDMDS1UUID; $FOUNDMDS2UUID"
519 XMLCONFIG=$OLDXMLCONFIG
521 run_test 13 "check new_uuid of lmc operating correctly"
524 OLDXMLCONFIG=$XMLCONFIG
525 XMLCONFIG="conf13-1.xml"
526 SECONDXMLCONFIG="conf13-2.xml"
527 # check multiple invocations for lmc generate same XML configuration file
529 echo "Generate the first XML configuration file"
531 echo "mv $XMLCONFIG to $SECONDXMLCONFIG"
532 sed -e "s/mtime[^ ]*//" $XMLCONFIG > $SECONDXMLCONFIG || return $?
533 echo "Generate the second XML configuration file"
535 # don't compare .xml mtime, it will always be different
536 if [ `sed -e "s/mtime[^ ]*//" $XMLCONFIG | diff - $SECONDXMLCONFIG | wc -l` -eq 0 ]; then
537 echo "Success:multiple invocations for lmc generate same XML file"
539 echo "Error: multiple invocations for lmc generate different XML file"
543 rm -f $XMLCONFIG $SECONDXMLCONFIG
544 XMLCONFIG=$OLDXMLCONFIG
546 run_test 13b "check lmc generates consistent .xml file"
551 # create xml file with --mkfsoptions for ost
552 echo "create xml file with --mkfsoptions for ost"
553 add_mds mds --dev $MDSDEV --size $MDSSIZE
554 add_lov lov1 mds --stripe_sz $STRIPE_BYTES\
555 --stripe_cnt $STRIPES_PER_OBJ --stripe_pattern 0
556 add_ost ost --lov lov1 --dev $OSTDEV --size $OSTSIZE \
557 --mkfsoptions "-Llabel_conf_14"
558 add_client client mds --lov lov1 --path $MOUNT
560 FOUNDSTRING=`awk -F"<" '/<mkfsoptions>/{print $2}' $XMLCONFIG`
561 EXPECTEDSTRING="mkfsoptions>-Llabel_conf_14"
562 if [ "$EXPECTEDSTRING" != "$FOUNDSTRING" ]; then
563 echo "Error: expected: $EXPECTEDSTRING; found: $FOUNDSTRING"
566 echo "Success:mkfsoptions for ost written to xml file correctly."
568 # mount lustre to test lconf mkfsoptions-parsing
572 mount_client $MOUNT || return $?
573 if [ -z "`do_facet ost1 dumpe2fs -h $OSTDEV | grep label_conf_14`" ]; then
574 echo "Error: the mkoptions not applied to mke2fs of ost."
578 echo "lconf mkfsoptions for ost success"
582 run_test 14 "test mkfsoptions of ost for lmc and lconf"
586 [ -f $MOUNTLUSTRE ] && echo "remove $MOUNTLUSTRE" && rm -f $MOUNTLUSTRE
587 if [ -f $MOUNTLUSTRE.sav ]; then
588 echo "return original $MOUNTLUSTRE.sav to $MOUNTLUSTRE"
589 mv $MOUNTLUSTRE.sav $MOUNTLUSTRE
593 # this only tests the kernel mount command, not anything about lustre.
595 MOUNTLUSTRE=${MOUNTLUSTRE:-/sbin/mount.lustre}
599 echo "mount lustre on ${MOUNT} without $MOUNTLUSTRE....."
600 if [ -f "$MOUNTLUSTRE" ]; then
601 echo "save $MOUNTLUSTRE to $MOUNTLUSTRE.sav"
602 mv $MOUNTLUSTRE $MOUNTLUSTRE.sav && trap cleanup_15 EXIT INT
603 if [ -f $MOUNTLUSTRE ]; then
604 skip "$MOUNTLUSTRE cannot be moved, skipping test"
609 mount_client $MOUNT && error "mount succeeded" && return 1
610 echo "mount lustre on $MOUNT without $MOUNTLUSTRE failed as expected"
614 run_test 15 "zconf-mount without /sbin/mount.lustre (should return error)"
617 TMPMTPT="${MOUNT%/*}/conf16"
619 if [ ! -e "$MDSDEV" ]; then
620 log "no $MDSDEV existing, so mount Lustre to create one"
622 check_mount || return 41
626 [ -f "$MDSDEV" ] && LOOPOPT="-o loop"
628 log "change the mode of $MDSDEV/OBJECTS,LOGS,PENDING to 555"
629 do_facet mds "mkdir -p $TMPMTPT &&
630 mount $LOOPOPT -t $FSTYPE $MDSDEV $TMPMTPT &&
631 chmod 555 $TMPMTPT/{OBJECTS,LOGS,PENDING} &&
632 umount $TMPMTPT" || return $?
634 log "mount Lustre to change the mode of OBJECTS/LOGS/PENDING, then umount Lustre"
636 check_mount || return 41
639 log "read the mode of OBJECTS/LOGS/PENDING and check if they has been changed properly"
640 EXPECTEDOBJECTSMODE=`do_facet mds "debugfs -R 'stat OBJECTS' $MDSDEV 2> /dev/null" | grep 'Mode: ' | sed -e "s/.*Mode: *//" -e "s/ *Flags:.*//"`
641 EXPECTEDLOGSMODE=`do_facet mds "debugfs -R 'stat LOGS' $MDSDEV 2> /dev/null" | grep 'Mode: ' | sed -e "s/.*Mode: *//" -e "s/ *Flags:.*//"`
642 EXPECTEDPENDINGMODE=`do_facet mds "debugfs -R 'stat PENDING' $MDSDEV 2> /dev/null" | grep 'Mode: ' | sed -e "s/.*Mode: *//" -e "s/ *Flags:.*//"`
644 if [ "$EXPECTEDOBJECTSMODE" = "0777" ]; then
645 log "Success:Lustre change the mode of OBJECTS correctly"
647 error "Lustre does not change mode of OBJECTS properly"
650 if [ "$EXPECTEDLOGSMODE" = "0777" ]; then
651 log "Success:Lustre change the mode of LOGS correctly"
653 error "Lustre does not change mode of LOGS properly"
656 if [ "$EXPECTEDPENDINGMODE" = "0777" ]; then
657 log "Success:Lustre change the mode of PENDING correctly"
659 error "Lustre does not change mode of PENDING properly"
662 run_test 16 "verify that lustre will correct the mode of OBJECTS/LOGS/PENDING"
665 if [ ! -e "$MDSDEV" ]; then
666 echo "no $MDSDEV existing, so mount Lustre to create one"
668 check_mount || return 41
672 echo "Remove mds config log"
673 do_facet mds "debugfs -w -R 'unlink CONFIGS/$FSNAME-MDT0000' $MDSDEV || return \$?" || return $?
676 start_mds && return 42
679 run_test 17 "Verify failed mds_postsetup won't fail assertion (2936) (should return errs)"
682 [ -f $MDSDEV ] && echo "remove $MDSDEV" && rm -f $MDSDEV
683 echo "mount mds with large journal..."
684 local myMDSSIZE=2000000
685 OLD_MDS_MKFS_OPTS=$MDS_MKFS_OPTS
687 MDS_MKFS_OPTS="--mgs --mdt --fsname=$FSNAME --device-size=$myMDSSIZE --param sys.timeout=$TIMEOUT $MDSOPT"
690 echo "mount lustre system..."
692 check_mount || return 41
694 echo "check journal size..."
695 FOUNDSIZE=`do_facet mds "debugfs -c -R 'stat <8>' $MDSDEV" | awk '/Size: / { print $NF; exit;}'`
696 if [ $FOUNDSIZE -gt $((32 * 1024 * 1024)) ]; then
697 log "Success: mkfs creates large journals. Size: $((FOUNDSIZE >> 20))M"
699 error "expected journal size > 32M, found $((FOUNDSIZE >> 20))M"
704 MDS_MKFS_OPTS=$OLD_MDS_MKFS_OPTS
707 run_test 18 "check mkfs creates large journals"
710 start_mds || return 1
711 stop_mds -f || return 2
713 run_test 19a "start/stop MDS without OSTs"
716 start_ost || return 1
717 stop_ost -f || return 2
719 run_test 19b "start/stop OSTs without MDS"
722 # first format the ost/mdt
726 check_mount || return 43
728 remount_client ro $MOUNT || return 44
729 touch $DIR/$tfile && echo "$DIR/$tfile created incorrectly" && return 45
730 [ -e $DIR/$tfile ] && echo "$DIR/$tfile exists incorrectly" && return 46
731 remount_client rw $MOUNT || return 47
733 [ ! -f $DIR/$tfile ] && echo "$DIR/$tfile missing" && return 48
734 MCNT=`grep -c $MOUNT /etc/mtab`
735 [ "$MCNT" -ne 1 ] && echo "$MOUNT in /etc/mtab $MCNT times" && return 49
740 run_test 20 "remount ro,rw mounts work and doesn't break /etc/mtab"
748 run_test 21a "start mds before ost, stop ost first"
756 run_test 21b "start ost before mds, stop mds first"
766 run_test 21c "start mds between two osts, stop mds last"
769 #reformat to remove all logs
772 echo Client mount before any osts are in the logs
774 check_mount && return 41
777 echo Client mount with ost in logs, but none running
781 # check_mount will block trying to contact ost
785 echo Client mount with a running ost
788 check_mount || return 41
793 run_test 22 "start a client before osts (should return errs)"
799 # force down client so that recovering mds waits for reconnect
800 zconf_umount `hostname` $MOUNT -f
801 # enter recovery on mds
803 # try to start a new client
804 mount_client $MOUNT &
807 MOUNT_LUSTRE_PID=`ps -ef | grep mount.lustre | grep -v grep | awk '{print $2}'`
808 echo mount pid is ${MOUNT_PID}, mount.lustre pid is ${MOUNT_LUSTRE_PID}
810 ps --ppid $MOUNT_LUSTRE_PID
811 # FIXME why o why can't I kill these? Manual "ctrl-c" works...
812 kill -TERM $MOUNT_PID
813 echo "waiting for mount to finish"
820 #this test isn't working yet
821 #run_test 23 "interrupt client during recovery mount delay"
824 local fs2mds_HOST=$mds_HOST
825 local fs2ost_HOST=$ost_HOST
826 [ -n "$ost1_HOST" ] && fs2ost_HOST=$ost1_HOST
827 if [ -z "$fs2ost_DEV" -o -z "$fs2mds_DEV" ]; then
828 do_facet $SINGLEMDS [ -b "$MDSDEV" ] && \
829 skip "mixed loopback and real device not working" && return
832 local fs2mdsdev=${fs2mds_DEV:-${MDSDEV}_2}
833 local fs2ostdev=${fs2ost_DEV:-$(ostdevname 1)_2}
835 # test 8-char fsname as well
836 local FSNAME2=test1234
837 add fs2mds $MDS_MKFS_OPTS --fsname=${FSNAME2} --nomgs --mgsnode=$MGSNID --reformat $fs2mdsdev || exit 10
839 add fs2ost $OST_MKFS_OPTS --fsname=${FSNAME2} --reformat $fs2ostdev || exit 10
842 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS
843 start fs2ost $fs2ostdev $OST_MOUNT_OPTS
845 mount -t lustre $MGSNID:/${FSNAME2} $MOUNT2 || return 1
847 check_mount || return 2
848 # files written on 1 should not show up on 2
849 cp /etc/passwd $DIR/$tfile
851 [ -e $MOUNT2/$tfile ] && error "File bleed" && return 7
854 cp /etc/passwd $MOUNT2/b || return 3
855 rm $MOUNT2/b || return 4
856 # 2 is actually mounted
857 grep $MOUNT2' ' /proc/mounts > /dev/null || return 5
859 facet_failover fs2mds
860 facet_failover fs2ost
863 # the MDS must remain up until last MDT
865 MDS=$(do_facet $SINGLEMDS "cat $LPROC/devices" | awk '($3 ~ "mdt" && $4 ~ "MDT") { print $4 }')
866 [ -z "$MDS" ] && error "No MDT" && return 8
870 cleanup_nocli || return 6
872 run_test 24a "Multiple MDTs on a single node"
875 local fs2mds_HOST=$mds_HOST
876 if [ -z "$fs2mds_DEV" ]; then
877 do_facet $SINGLEMDS [ -b "$MDSDEV" ] && \
878 skip "mixed loopback and real device not working" && return
881 local fs2mdsdev=${fs2mds_DEV:-${MDSDEV}_2}
883 add fs2mds $MDS_MKFS_OPTS --fsname=${FSNAME}2 --mgs --reformat $fs2mdsdev || exit 10
885 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS && return 2
888 run_test 24b "Multiple MGSs on a single node (should return err)"
892 check_mount || return 2
893 local MODULES=$($LCTL modules | awk '{ print $2 }')
894 rmmod $MODULES 2>/dev/null || true
897 run_test 25 "Verify modules are referenced"
901 # we need modules before mount for sysctl, so make sure...
902 do_facet mds "lsmod | grep -q lustre || modprobe lustre"
903 #define OBD_FAIL_MDS_FS_SETUP 0x135
904 do_facet mds "sysctl -w lustre.fail_loc=0x80000135"
905 start_mds && echo MDS started && return 1
907 DEVS=$(cat $LPROC/devices | wc -l)
908 [ $DEVS -gt 0 ] && return 2
909 unload_modules || return 203
911 run_test 26 "MDT startup failure cleans LOV (should return errs)"
917 local ORIG=$(do_facet $myfacet "$TEST")
918 if [ $# -gt 3 ]; then
924 echo "Setting $PARAM from $ORIG to $FINAL"
925 do_facet mds "$LCTL conf_param $PARAM=$FINAL" || error conf_param failed
931 RESULT=$(do_facet $myfacet "$TEST")
932 if [ $RESULT -eq $FINAL ]; then
933 echo "Updated config after $WAIT sec (got $RESULT)"
937 if [ $WAIT -eq $MAX ]; then
938 echo "Config update not seen: wanted $FINAL got $RESULT"
941 echo "Waiting $(($MAX - $WAIT)) secs for config update"
946 start_ost || return 1
947 start_mds || return 2
948 echo "Requeue thread should have started: "
949 ps -e | grep ll_cfg_requeue
950 set_and_check ost1 "cat $LPROC/obdfilter/$FSNAME-OST0000/client_cache_seconds" "$FSNAME-OST0000.ost.client_cache_seconds" || return 3
953 run_test 27a "Reacquire MGS lock if OST started first"
958 set_and_check mds "cat $LPROC/mdt/$FSNAME-MDT0000/identity_acquire_expire" "$FSNAME-MDT0000.mdt.identity_acquire_expire" || return 3
959 set_and_check client "cat $LPROC/mdc/$FSNAME-MDT0000-mdc-*/max_rpcs_in_flight" "$FSNAME-MDT0000.mdc.max_rpcs_in_flight" || return 4
962 run_test 27b "Reacquire MGS lock after failover"
966 TEST="cat $LPROC/llite/$FSNAME-*/max_read_ahead_whole_mb"
969 FINAL=$(($ORIG + 10))
970 set_and_check client "$TEST" "$FSNAME.llite.max_read_ahead_whole_mb" || return 3
971 set_and_check client "$TEST" "$FSNAME.llite.max_read_ahead_whole_mb" || return 3
972 umount_client $MOUNT || return 200
975 if [ $RESULT -ne $FINAL ]; then
976 echo "New config not seen: wanted $FINAL got $RESULT"
979 echo "New config success: got $RESULT"
983 run_test 28 "permanent parameter setting"
986 [ "$OSTCOUNT" -lt "2" ] && skip "$OSTCOUNT < 2, skipping" && return
987 setup > /dev/null 2>&1
991 local PARAM="$FSNAME-OST0001.osc.active"
992 local PROC_ACT="$LPROC/osc/$FSNAME-OST0001-osc-[^M]*/active"
993 local PROC_UUID="$LPROC/osc/$FSNAME-OST0001-osc-[^M]*/ost_server_uuid"
994 if [ ! -r $PROC_ACT ]; then
995 echo "Can't read $PROC_ACT"
996 ls $LPROC/osc/$FSNAME-*
999 ACTV=$(cat $PROC_ACT)
1001 set_and_check client "cat $PROC_ACT" "$PARAM" $DEAC || return 2
1002 # also check ost_server_uuid status
1003 RESULT=$(grep DEACTIV $PROC_UUID)
1004 if [ -z "$RESULT" ]; then
1005 echo "Live client not deactivated: $(cat $PROC_UUID)"
1008 echo "Live client success: got $RESULT"
1012 local MPROC="$LPROC/osc/$FSNAME-OST0001-osc-[M]*/active"
1017 RESULT=`do_facet mds " [ -r $MPROC ] && cat $MPROC"`
1018 [ ${PIPESTATUS[0]} = 0 ] || error "Can't read $MPROC"
1019 if [ $RESULT -eq $DEAC ]; then
1020 echo "MDT deactivated also after $WAIT sec (got $RESULT)"
1024 if [ $WAIT -eq $MAX ]; then
1025 echo "MDT not deactivated: wanted $DEAC got $RESULT"
1028 echo "Waiting $(($MAX - $WAIT)) secs for MDT deactivated"
1031 # test new client starts deactivated
1032 umount_client $MOUNT || return 200
1034 RESULT=$(grep DEACTIV $PROC_UUID | grep NEW)
1035 if [ -z "$RESULT" ]; then
1036 echo "New client not deactivated from start: $(cat $PROC_UUID)"
1039 echo "New client success: got $RESULT"
1042 # make sure it reactivates
1043 set_and_check client "cat $PROC_ACT" "$PARAM" $ACTV || return 6
1045 umount_client $MOUNT
1048 #writeconf to remove all ost2 traces for subsequent tests
1051 run_test 29 "permanently remove an OST"
1054 # start mds first after writeconf
1058 TEST="cat $LPROC/llite/$FSNAME-*/max_read_ahead_whole_mb"
1060 for i in $(seq 1 20); do
1061 set_and_check client "$TEST" "$FSNAME.llite.max_read_ahead_whole_mb" $i || return 3
1063 # make sure client restart still works
1064 umount_client $MOUNT
1065 mount_client $MOUNT || return 4
1066 [ "$($TEST)" -ne "$i" ] && return 5
1067 set_and_check client "$TEST" "$FSNAME.llite.max_read_ahead_whole_mb" $ORIG || return 6
1070 run_test 30 "Big config llog"
1072 test_31() { # bug 10734
1073 # ipaddr must not exist
1074 mount -t lustre 4.3.2.1@tcp:/lustre $MOUNT || true
1077 run_test 31 "Connect to non-existent node (shouldn't crash)"
1080 # XXX - make this run on client-only systems with real hardware on
1082 # there appears to be a lot of assumption here about loopback
1084 # or maybe this test is just totally useless on a client-only system
1085 [ "$mds_HOST" = "`hostname`" ] || { skip "remote MDS" && return 0; }
1086 [ "$ost_HOST" = "`hostname`" -o "$ost1_HOST" = "`hostname`" ] || \
1087 { skip "remote OST" && return 0; }
1089 [ -z "$TUNEFS" ] && skip "No tunefs" && return
1090 local DISK1_4=$LUSTRE/tests/disk1_4.zip
1091 [ ! -r $DISK1_4 ] && skip "Cant find $DISK1_4, skipping" && return
1092 unzip -o -j -d $TMP/$tdir $DISK1_4 || { skip "Cant unzip $DISK1_4, skipping" && return ; }
1094 sysctl lnet.debug=$PTLDEBUG
1096 $TUNEFS $TMP/$tdir/mds || error "tunefs failed"
1097 # nids are wrong, so client wont work, but server should start
1098 start mds $TMP/$tdir/mds "-o loop,exclude=lustre-OST0000" || return 3
1099 local UUID=$(cat $LPROC/mds/lustre-MDT0000/uuid)
1101 [ "$UUID" == "mdsA_UUID" ] || error "UUID is wrong: $UUID"
1103 $TUNEFS --mgsnode=`hostname` $TMP/$tdir/ost1 || error "tunefs failed"
1104 start ost1 $TMP/$tdir/ost1 "-o loop" || return 5
1105 UUID=$(cat $LPROC/obdfilter/lustre-OST0000/uuid)
1107 [ "$UUID" == "ost1_UUID" ] || error "UUID is wrong: $UUID"
1109 local NID=$($LCTL list_nids | head -1)
1111 echo "OSC changes should return err:"
1112 $LCTL conf_param lustre-OST0000.osc.max_dirty_mb=15 && return 7
1113 $LCTL conf_param lustre-OST0000.failover.node=$NID && return 8
1115 echo "MDC changes should succeed:"
1116 $LCTL conf_param lustre-MDT0000.mdc.max_rpcs_in_flight=9 || return 9
1117 $LCTL conf_param lustre-MDT0000.failover.node=$NID || return 10
1120 # With a new good MDT failover nid, we should be able to mount a client
1121 # (but it cant talk to OST)
1122 local OLDMOUNTOPT=$MOUNTOPT
1123 MOUNTOPT="exclude=lustre-OST0000"
1125 MOUNTOPT=$OLDMOUNTOPT
1126 set_and_check client "cat $LPROC/mdc/*/max_rpcs_in_flight" "lustre-MDT0000.mdc.max_rpcs_in_flight" || return 11
1128 zconf_umount `hostname` $MOUNT -f
1131 # mount a second time to make sure we didnt leave upgrade flag on
1132 $TUNEFS --dryrun $TMP/$tdir/mds || error "tunefs failed"
1133 start mds $TMP/$tdir/mds "-o loop,exclude=lustre-OST0000" || return 12
1136 [ -d $TMP/$tdir ] && rm -rf $TMP/$tdir
1138 run_test 32a "Upgrade from 1.4 (not live)"
1141 # XXX - make this run on client-only systems with real hardware on
1143 # there appears to be a lot of assumption here about loopback
1145 # or maybe this test is just totally useless on a client-only system
1146 [ "$mds_HOST" = "`hostname`" ] || { skip "remote MDS" && return 0; }
1147 [ "$ost_HOST" = "`hostname`" -o "$ost1_HOST" = "`hostname`" ] || \
1148 { skip "remote OST" && return 0; }
1150 [ -z "$TUNEFS" ] && skip "No tunefs" && return
1151 local DISK1_4=$LUSTRE/tests/disk1_4.zip
1152 [ ! -r $DISK1_4 ] && skip "Cant find $DISK1_4, skipping" && return
1153 unzip -o -j -d $TMP/$tdir $DISK1_4 || { skip "Cant unzip $DISK1_4, skipping" && return ; }
1155 sysctl lnet.debug=$PTLDEBUG
1157 # writeconf will cause servers to register with their current nids
1158 $TUNEFS --writeconf $TMP/$tdir/mds || error "tunefs failed"
1159 start mds $TMP/$tdir/mds "-o loop" || return 3
1160 local UUID=$(cat $LPROC/mds/lustre-MDT0000/uuid)
1162 [ "$UUID" == "mdsA_UUID" ] || error "UUID is wrong: $UUID"
1164 $TUNEFS --mgsnode=`hostname` $TMP/$tdir/ost1 || error "tunefs failed"
1165 start ost1 $TMP/$tdir/ost1 "-o loop" || return 5
1166 UUID=$(cat $LPROC/obdfilter/lustre-OST0000/uuid)
1168 [ "$UUID" == "ost1_UUID" ] || error "UUID is wrong: $UUID"
1170 echo "OSC changes should succeed:"
1171 $LCTL conf_param lustre-OST0000.osc.max_dirty_mb=15 || return 7
1172 $LCTL conf_param lustre-OST0000.failover.node=$NID || return 8
1174 echo "MDC changes should succeed:"
1175 $LCTL conf_param lustre-MDT0000.mdc.max_rpcs_in_flight=9 || return 9
1178 # MDT and OST should have registered with new nids, so we should have
1179 # a fully-functioning client
1180 echo "Check client and old fs contents"
1182 set_and_check client "cat $LPROC/mdc/*/max_rpcs_in_flight" "${NEWNAME}-MDT0000.mdc.max_rpcs_in_flight" || return 11
1183 [ "$(cksum $MOUNT/passwd | cut -d' ' -f 1,2)" == "2479747619 779" ] || return 12
1187 [ -d $TMP/$tdir ] && rm -rf $TMP/$tdir
1189 run_test 32b "Upgrade from 1.4 with writeconf"
1191 test_33() { # bug 12333
1192 local FSNAME2=test1234
1193 local fs2mds_HOST=$mds_HOST
1194 local fs2ost_HOST=$ost_HOST
1195 [ -n "$ost1_HOST" ] && fs2ost_HOST=$ost1_HOST
1197 if [ -z "$fs2ost_DEV" -o -z "$fs2mds_DEV" ]; then
1198 do_facet $SINGLEMDS [ -b "$MDSDEV" ] && \
1199 skip "mixed loopback and real device not working" && return
1202 local fs2mdsdev=${fs2mds_DEV:-${MDSDEV}_2}
1203 local fs2ostdev=${fs2ost_DEV:-$(ostdevname 1)_2}
1204 add fs2mds $MDS_MKFS_OPTS --fsname=${FSNAME2} --reformat $fs2mdsdev || exit 10
1205 add fs2ost $OST_MKFS_OPTS --fsname=${FSNAME2} --index=8191 --mgsnode=$MGSNID --reformat $fs2ostdev || exit 10
1207 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS
1208 start fs2ost $fs2ostdev $OST_MOUNT_OPTS
1210 mount -t lustre $MGSNID:/${FSNAME2} $MOUNT2 || return 1
1216 rm -rf $MOUNT2 $fs2mdsdev $fs2ostdev
1217 cleanup_nocli || return 6
1219 run_test 33 "Mount ost with a large index number"
1224 do_facet client dd if=/dev/zero of=$MOUNT/24 bs=1024k count=1
1225 # Drop lock cancelation reply during umount
1226 #define OBD_FAIL_LDLM_CANCEL 0x304
1227 do_facet client sysctl -w lustre.fail_loc=0x80000304
1228 #sysctl -w lnet.debug=-1
1229 umount_client $MOUNT
1232 run_test 34 "Drop cancel during umount"
1236 do_facet client multiop $DIR/file O_c &
1238 manual_umount_client
1240 do_facet client killall -USR1 multiop
1241 if [ $rc -eq 0 ]; then
1242 error "umount not fail!"
1247 run_test 34a "umount with opened file should be fail"
1252 touch $DIR/$tfile || return 1
1253 stop_mds --force || return 2
1255 manual_umount_client --force
1257 if [ $rc -ne 0 ]; then
1258 error "mtab after failed umount - rc $rc"
1264 run_test 34b "force umount with failed mds should be normal"
1268 touch $DIR/$tfile || return 1
1269 stop_ost --force || return 2
1271 manual_umount_client --force
1273 if [ $rc -ne 0 ]; then
1274 error "mtab after failed umount - rc $rc"
1280 run_test 34c "force umount with failed mds should be normal"
1282 test_35() { # bug 12459
1285 DBG_SAVE="`sysctl -n lnet.debug`"
1286 sysctl -w lnet.debug="ha"
1288 log "Set up a fake failnode for the MDS"
1290 do_facet mds $LCTL conf_param ${FSNAME}-MDT0000.failover.node=$FAKENID || return 4
1292 log "Wait for RECONNECT_INTERVAL seconds (10s)"
1295 MSG="conf-sanity.sh test_33 `date +%F%kh%Mm%Ss`"
1298 log "Stopping the MDT:"
1299 stop_mds || return 5
1301 df $MOUNT > /dev/null 2>&1 &
1303 log "Restarting the MDT:"
1304 start_mds || return 6
1305 log "Wait for df ($DFPID) ... "
1308 sysctl -w lnet.debug="$DBG_SAVE"
1310 # retrieve from the log the first server that the client tried to
1311 # contact after the connection loss
1312 $LCTL dk $TMP/lustre-log-$TESTNAME.log
1313 NEXTCONN=`awk "/${MSG}/ {start = 1;}
1314 /import_select_connection.*${FSNAME}-MDT0000-mdc.* using connection/ {
1316 if (\\\$NF ~ /$FAKENID/)
1322 }" $TMP/lustre-log-$TESTNAME.log`
1323 [ "$NEXTCONN" != "0" ] && log "The client didn't try to reconnect to the last active server (tried ${NEXTCONN} instead)" && return 7
1326 run_test 35 "Reconnect to the last active server first"
1330 local FSNAME2=test1234
1331 local fs2mds_HOST=$mds_HOST
1332 local fs2ost_HOST=$ost_HOST
1333 local fs3ost_HOST=$ost_HOST
1335 [ -n "$ost1_HOST" ] && fs2ost_HOST=$ost1_HOST && fs3ost_HOST=$ost1_HOST
1338 if [ -z "$fs2ost_DEV" -o -z "$fs2mds_DEV" -o -z "$fs3ost_DEV" ]; then
1339 do_facet $SINGLEMDS [ -b "$MDSDEV" ] && \
1340 skip "mixed loopback and real device not working" && return
1342 [ $OSTCOUNT -lt 2 ] && skip "skipping test for single OST" && return
1344 [ $(grep -c obdfilter $LPROC/devices) -eq 0 ] &&
1345 skip "skipping test for remote OST" && return
1347 local fs2mdsdev=${fs2mds_DEV:-${MDSDEV}_2}
1348 local fs2ostdev=${fs2ost_DEV:-$(ostdevname 1)_2}
1349 local fs3ostdev=${fs3ost_DEV:-$(ostdevname 2)_2}
1350 add fs2mds $MDS_MKFS_OPTS --fsname=${FSNAME2} --reformat $fs2mdsdev || exit 10
1351 add fs2ost $OST_MKFS_OPTS --mkfsoptions='-b1024' --fsname=${FSNAME2} --mgsnode=$MGSNID --reformat $fs2ostdev || exit 10
1352 add fs3ost $OST_MKFS_OPTS --mkfsoptions='-b4096' --fsname=${FSNAME2} --mgsnode=$MGSNID --reformat $fs3ostdev || exit 10
1354 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS
1355 start fs2ost $fs2ostdev $OST_MOUNT_OPTS
1356 start fs3ost $fs3ostdev $OST_MOUNT_OPTS
1358 mount -t lustre $MGSNID:/${FSNAME2} $MOUNT2 || return 1
1360 sleep 5 # until 11778 fixed
1362 dd if=/dev/zero of=$MOUNT2/$tfile bs=1M count=7 || return 2
1364 BKTOTAL=`awk 'BEGIN{total=0}; {total+=$1}; END{print total}' \
1365 $LPROC/obdfilter/*/kbytestotal`
1366 BKFREE=`awk 'BEGIN{free=0}; {free+=$1}; END{print free}' \
1367 $LPROC/obdfilter/*/kbytesfree`
1368 BKAVAIL=`awk 'BEGIN{avail=0}; {avail+=$1}; END{print avail}' \
1369 $LPROC/obdfilter/*/kbytesavail`
1370 STRING=`df -P $MOUNT2 | tail -n 1 | awk '{print $2","$3","$4}'`
1371 DFTOTAL=`echo $STRING | cut -d, -f1`
1372 DFUSED=`echo $STRING | cut -d, -f2`
1373 DFAVAIL=`echo $STRING | cut -d, -f3`
1374 DFFREE=$(($DFTOTAL - $DFUSED))
1376 ALLOWANCE=$((64 * $OSTCOUNT))
1378 if [ $DFTOTAL -lt $(($BKTOTAL - $ALLOWANCE)) ] ||
1379 [ $DFTOTAL -gt $(($BKTOTAL + $ALLOWANCE)) ] ; then
1380 echo "**** FAIL: df total($DFTOTAL) mismatch OST total($BKTOTAL)"
1383 if [ $DFFREE -lt $(($BKFREE - $ALLOWANCE)) ] ||
1384 [ $DFFREE -gt $(($BKFREE + $ALLOWANCE)) ] ; then
1385 echo "**** FAIL: df free($DFFREE) mismatch OST free($BKFREE)"
1388 if [ $DFAVAIL -lt $(($BKAVAIL - $ALLOWANCE)) ] ||
1389 [ $DFAVAIL -gt $(($BKAVAIL + $ALLOWANCE)) ] ; then
1390 echo "**** FAIL: df avail($DFAVAIL) mismatch OST avail($BKAVAIL)"
1395 stop fs3ost -f || return 200
1396 stop fs2ost -f || return 201
1397 stop fs2mds -f || return 202
1398 rm -rf $MOUNT2 $fs2mdsdev $fs2ostdev $fs3ostdev
1399 unload_modules || return 203
1402 run_test 36 "df report consistency on OSTs with different block size"
1404 umount_client $MOUNT
1408 equals_msg `basename $0`: test complete
1409 [ -f "$TESTSUITELOG" ] && cat $TESTSUITELOG || true