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: 13709 10510 12743
20 ALWAYS_EXCEPT=" $CONF_SANITY_EXCEPT $MOUNTCONFSKIP $HEAD_EXCEPT 22 23 36"
21 # UPDATE THE COMMENT ABOVE WITH BUG NUMBERS WHEN CHANGING ALWAYS_EXCEPT!
24 [ "$SLOW" = "no" ] && EXCEPT="$EXCEPT 0 1 2 3 6 7 15 18 24b 25 30 31 32 33 34a "
27 PATH=$PWD/$SRCDIR:$SRCDIR:$SRCDIR/../utils:$PATH
29 PTLDEBUG=${PTLDEBUG:--1}
31 LUSTRE=${LUSTRE:-`dirname $0`/..}
32 RLUSTRE=${RLUSTRE:-$LUSTRE}
33 MOUNTLUSTRE=${MOUNTLUSTRE:-/sbin/mount.lustre}
34 MKFSLUSTRE=${MKFSLUSTRE:-/usr/sbin/mkfs.lustre}
37 . $LUSTRE/tests/test-framework.sh
39 # use small MDS + OST size to speed formatting time
42 . ${CONFIG:=$LUSTRE/tests/cfg/local.sh}
53 # who knows if/where $TUNEFS is installed? Better reformat if it fails...
54 do_facet ${facet} "$TUNEFS --writeconf $MDSDEV" || echo "tunefs failed, reformatting instead" && reformat
59 # The MGS must be started before the OSTs for a new fs, so start
60 # and stop to generate the startup logs.
69 echo "start mds service on `facet_active_host mds`"
70 start mds $MDSDEV $MDS_MOUNT_OPTS || return 94
74 echo "stop mds service on `facet_active_host mds`"
75 # These tests all use non-failover stop
76 stop mds -f || return 97
80 echo "start ost1 service on `facet_active_host ost1`"
81 start ost1 `ostdevname 1` $OST_MOUNT_OPTS || return 95
85 echo "stop ost1 service on `facet_active_host ost1`"
86 # These tests all use non-failover stop
87 stop ost1 -f || return 98
91 echo "start ost2 service on `facet_active_host ost2`"
92 start ost2 `ostdevname 2` $OST_MOUNT_OPTS || return 92
96 echo "stop ost2 service on `facet_active_host ost2`"
97 # These tests all use non-failover stop
98 stop ost2 -f || return 93
103 echo "mount $FSNAME on ${MOUNTPATH}....."
104 zconf_mount `hostname` $MOUNTPATH || return 96
108 local SAVEMOUNTOPT=$MOUNTOPT
109 MOUNTOPT="remount,$1"
111 echo "remount '$1' lustre on ${MOUNTPATH}....."
112 zconf_mount `hostname` $MOUNTPATH || return 96
113 MOUNTOPT=$SAVEMOUNTOPT
118 echo "umount lustre on ${MOUNTPATH}....."
119 zconf_umount `hostname` $MOUNTPATH || return 97
122 manual_umount_client(){
123 echo "manual umount lustre on ${MOUNT}...."
124 do_facet client "umount -d $MOUNT"
134 stop_mds || return 201
135 stop_ost || return 202
136 unload_modules || return 203
140 umount_client $MOUNT || return 200
141 cleanup_nocli || return $?
145 do_facet client "cp /etc/passwd $DIR/a" || return 71
146 do_facet client "rm $DIR/a" || return 72
147 # make sure lustre is actually mounted (touch will block,
148 # but grep won't, so do it after)
149 do_facet client "grep $MOUNT' ' /proc/mounts > /dev/null" || return 73
150 echo "setup single mount lustre success"
154 do_facet client "touch $DIR/a" || return 71
155 do_facet client "rm $DIR/a" || return 72
156 do_facet client "touch $DIR2/a" || return 73
157 do_facet client "rm $DIR2/a" || return 74
158 echo "setup double mount lustre success"
163 if [ "$ONLY" == "setup" ]; then
168 if [ "$ONLY" == "cleanup" ]; then
173 #create single point mountpoint
181 check_mount || return 41
184 run_test 0 "single mount setup"
188 echo "start ost second time..."
190 check_mount || return 42
193 run_test 1 "start up ost twice (should return errors)"
198 echo "start mds second time.."
201 check_mount || return 43
204 run_test 2 "start up mds twice (should return err)"
208 #mount.lustre returns an error if already in mtab
209 mount_client $MOUNT && return $?
210 check_mount || return 44
213 run_test 3 "mount client twice (should return err)"
217 touch $DIR/$tfile || return 85
221 # ok for ost to fail shutdown
222 if [ 202 -ne $eno ]; then
227 run_test 4 "force cleanup ost, then cleanup"
231 touch $DIR/$tfile || return 1
232 stop_mds -f || return 2
234 # cleanup may return an error from the failed
235 # disconnects; for now I'll consider this successful
236 # if all the modules have unloaded.
240 echo "killing umount"
241 kill -TERM $UMOUNT_PID
242 echo "waiting for umount to finish"
244 if grep " $MOUNT " /etc/mtab; then
245 echo "test 5: mtab after failed umount"
249 echo "killing umount"
250 kill -TERM $UMOUNT_PID
251 echo "waiting for umount to finish"
253 grep " $MOUNT " /etc/mtab && echo "test 5: mtab after second umount" && return 11
257 # stop_mds is a no-op here, and should not fail
258 cleanup_nocli || return $?
259 # df may have lingering entry
261 # mtab may have lingering entry
262 grep -v $MOUNT" " /etc/mtab > $TMP/mtabtemp
263 mv $TMP/mtabtemp /etc/mtab
265 run_test 5 "force cleanup mds, then cleanup"
269 [ -d $MOUNT ] || mkdir -p $MOUNT
270 grep " $MOUNT " /etc/mtab && echo "test 5b: mtab before mount" && return 10
271 mount_client $MOUNT && return 1
272 grep " $MOUNT " /etc/mtab && echo "test 5b: mtab after failed mount" && return 11
274 # stop_mds is a no-op here, and should not fail
275 cleanup_nocli || return $?
278 run_test 5b "mds down, cleanup after failed mount (bug 2712) (should return errs)"
283 [ -d $MOUNT ] || mkdir -p $MOUNT
284 grep " $MOUNT " /etc/mtab && echo "test 5c: mtab before mount" && return 10
285 mount -t lustre $MGSNID:/wrong.$FSNAME $MOUNT || :
286 grep " $MOUNT " /etc/mtab && echo "test 5c: mtab after failed mount" && return 11
288 cleanup_nocli || return $?
290 run_test 5c "cleanup after failed mount (bug 2712) (should return errs)"
296 grep " $MOUNT " /etc/mtab && echo "test 5d: mtab before mount" && return 10
297 mount_client $MOUNT || return 1
299 grep " $MOUNT " /etc/mtab && echo "test 5d: mtab after unmount" && return 11
302 run_test 5d "mount with ost down"
308 #define OBD_FAIL_PTLRPC_DELAY_SEND 0x506
309 do_facet client "sysctl -w lustre.fail_loc=0x80000506"
310 grep " $MOUNT " /etc/mtab && echo "test 5e: mtab before mount" && return 10
311 mount_client $MOUNT || echo "mount failed (not fatal)"
313 grep " $MOUNT " /etc/mtab && echo "test 5e: mtab after unmount" && return 11
316 run_test 5e "delayed connect, don't crash (bug 10268)"
321 mount_client ${MOUNT} || return 87
322 touch $DIR/a || return 86
325 run_test 6 "manual umount, then mount again"
330 cleanup_nocli || return $?
332 run_test 7 "manual umount, then cleanup"
337 check_mount2 || return 45
338 umount_client $MOUNT2
341 run_test 8 "double mount setup"
346 do_facet ost1 sysctl lnet.debug=\'inode trace\' || return 1
347 do_facet ost1 sysctl lnet.subsystem_debug=\'mds ost\' || return 1
349 CHECK_PTLDEBUG="`do_facet ost1 sysctl -n lnet.debug`"
350 if [ "$CHECK_PTLDEBUG" ] && [ "$CHECK_PTLDEBUG" = "trace inode" ];then
351 echo "lnet.debug success"
353 echo "lnet.debug: want 'trace inode', have '$CHECK_PTLDEBUG'"
356 CHECK_SUBSYS="`do_facet ost1 sysctl -n lnet.subsystem_debug`"
357 if [ "$CHECK_SUBSYS" ] && [ "$CHECK_SUBSYS" = "mds ost" ]; then
358 echo "lnet.subsystem_debug success"
360 echo "lnet.subsystem_debug: want 'mds ost', have '$CHECK_SUBSYS'"
363 stop_ost || return $?
366 run_test 9 "test ptldebug and subsystem for mkfs"
369 echo "generate configuration with the same name for node and mds"
370 OLDXMLCONFIG=$XMLCONFIG
371 XMLCONFIG="broken.xml"
372 [ -f "$XMLCONFIG" ] && rm -f $XMLCONFIG
376 echo "the name for node and mds is the same"
377 do_lmc --add mds --node ${facet}_facet --mds ${facet}_facet \
378 --dev $MDSDEV --size $MDSSIZE || return $?
379 do_lmc --add lov --mds ${facet}_facet --lov lov1 --stripe_sz \
380 $STRIPE_BYTES --stripe_cnt $STRIPES_PER_OBJ \
381 --stripe_pattern 0 || return $?
382 add_ost ost --lov lov1 --dev $OSTDEV --size $OSTSIZE
384 add_facet $facet --lustre_upcall $UPCALL
385 do_lmc --add mtpt --node ${facet}_facet --mds mds_facet \
386 --lov lov1 --path $MOUNT
392 check_mount || return 41
396 XMLCONFIG=$OLDXMLCONFIG
398 run_test 10 "mount lustre with the same name for node and mds"
401 OLDXMLCONFIG=$XMLCONFIG
402 XMLCONFIG="conf11.xml"
404 [ -f "$XMLCONFIG" ] && rm -f $XMLCONFIG
405 add_mds mds --dev $MDSDEV --size $MDSSIZE
406 add_ost ost --dev $OSTDEV --size $OSTSIZE
407 add_client client mds --path $MOUNT --ost ost_svc || return $?
408 echo "Default lov config success!"
410 [ -f "$XMLCONFIG" ] && rm -f $XMLCONFIG
411 add_mds mds --dev $MDSDEV --size $MDSSIZE
412 add_ost ost --dev $OSTDEV --size $OSTSIZE
413 add_client client mds --path $MOUNT && return $?
414 echo "--add mtpt with neither --lov nor --ost will return error"
418 XMLCONFIG=$OLDXMLCONFIG
420 run_test 11 "use default lov configuration (should return error)"
423 OLDXMLCONFIG=$XMLCONFIG
424 XMLCONFIG="batch.xml"
425 BATCHFILE="batchfile"
428 [ -f "$XMLCONFIG" ] && rm -f $XMLCONFIG
429 [ -f "$BATCHFILE" ] && rm -f $BATCHFILE
430 echo "--add net --node $HOSTNAME --nid $HOSTNAME --nettype tcp" > $BATCHFILE
431 echo "--add mds --node $HOSTNAME --mds mds1 --mkfsoptions \"-I 128\"" >> $BATCHFILE
432 # --mkfsoptions "-I 128"
433 do_lmc -m $XMLCONFIG --batch $BATCHFILE || return $?
434 if [ `sed -n '/>-I 128</p' $XMLCONFIG | wc -l` -eq 1 ]; then
435 echo "matched double quote success"
437 echo "matched double quote fail"
442 echo "--add net --node $HOSTNAME --nid $HOSTNAME --nettype tcp" > $BATCHFILE
443 echo "--add mds --node $HOSTNAME --mds mds1 --mkfsoptions \"-I 128" >> $BATCHFILE
444 # --mkfsoptions "-I 128
445 do_lmc -m $XMLCONFIG --batch $BATCHFILE && return $?
446 echo "unmatched double quote should return error"
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 single quote success"
457 echo "matched single 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 single 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 "backslash followed by a whitespace/letter success"
477 echo "backslash followed by a whitespace/letter 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 $?
486 echo "backslash followed by nothing should return error"
489 XMLCONFIG=$OLDXMLCONFIG
491 run_test 12 "lmc --batch, with single/double quote, backslash in batchfile"
494 OLDXMLCONFIG=$XMLCONFIG
495 XMLCONFIG="conf13-1.xml"
497 # check long uuid will be truncated properly and uniquely
498 echo "To generate XML configuration file(with long ost name): $XMLCONFIG"
499 [ -f "$XMLCONFIG" ] && rm -f $XMLCONFIG
500 do_lmc --add net --node $HOSTNAME --nid $HOSTNAME --nettype tcp
501 do_lmc --add mds --node $HOSTNAME --mds mds1_name_longer_than_31characters
502 do_lmc --add mds --node $HOSTNAME --mds mds2_name_longer_than_31characters
503 if [ ! -f "$XMLCONFIG" ]; then
504 echo "Error:no file $XMLCONFIG created!"
507 EXPECTEDMDS1UUID="e_longer_than_31characters_UUID"
508 EXPECTEDMDS2UUID="longer_than_31characters_UUID_2"
509 FOUNDMDS1UUID=`awk -F"'" '/<mds .*uuid=/' $XMLCONFIG | sed -n '1p' \
510 | sed "s/ /\n\r/g" | awk -F"'" '/uuid=/{print $2}'`
511 FOUNDMDS2UUID=`awk -F"'" '/<mds .*uuid=/' $XMLCONFIG | sed -n '2p' \
512 | sed "s/ /\n\r/g" | awk -F"'" '/uuid=/{print $2}'`
513 [ -z "$FOUNDMDS1UUID" ] && echo "MDS1 UUID empty" && return 1
514 [ -z "$FOUNDMDS2UUID" ] && echo "MDS2 UUID empty" && return 1
515 if ([ $EXPECTEDMDS1UUID = $FOUNDMDS1UUID ] && [ $EXPECTEDMDS2UUID = $FOUNDMDS2UUID ]) || \
516 ([ $EXPECTEDMDS1UUID = $FOUNDMDS2UUID ] && [ $EXPECTEDMDS2UUID = $FOUNDMDS1UUID ]); then
517 echo "Success:long uuid truncated successfully and being unique."
519 echo "Error:expected uuid for mds1 and mds2: $EXPECTEDMDS1UUID; $EXPECTEDMDS2UUID"
520 echo "but: found uuid for mds1 and mds2: $FOUNDMDS1UUID; $FOUNDMDS2UUID"
524 XMLCONFIG=$OLDXMLCONFIG
526 run_test 13 "check new_uuid of lmc operating correctly"
529 OLDXMLCONFIG=$XMLCONFIG
530 XMLCONFIG="conf13-1.xml"
531 SECONDXMLCONFIG="conf13-2.xml"
532 # check multiple invocations for lmc generate same XML configuration file
534 echo "Generate the first XML configuration file"
536 echo "mv $XMLCONFIG to $SECONDXMLCONFIG"
537 sed -e "s/mtime[^ ]*//" $XMLCONFIG > $SECONDXMLCONFIG || return $?
538 echo "Generate the second XML configuration file"
540 # don't compare .xml mtime, it will always be different
541 if [ `sed -e "s/mtime[^ ]*//" $XMLCONFIG | diff - $SECONDXMLCONFIG | wc -l` -eq 0 ]; then
542 echo "Success:multiple invocations for lmc generate same XML file"
544 echo "Error: multiple invocations for lmc generate different XML file"
548 rm -f $XMLCONFIG $SECONDXMLCONFIG
549 XMLCONFIG=$OLDXMLCONFIG
551 run_test 13b "check lmc generates consistent .xml file"
556 # create xml file with --mkfsoptions for ost
557 echo "create xml file with --mkfsoptions for ost"
558 add_mds mds --dev $MDSDEV --size $MDSSIZE
559 add_lov lov1 mds --stripe_sz $STRIPE_BYTES\
560 --stripe_cnt $STRIPES_PER_OBJ --stripe_pattern 0
561 add_ost ost --lov lov1 --dev $OSTDEV --size $OSTSIZE \
562 --mkfsoptions "-Llabel_conf_14"
563 add_client client mds --lov lov1 --path $MOUNT
565 FOUNDSTRING=`awk -F"<" '/<mkfsoptions>/{print $2}' $XMLCONFIG`
566 EXPECTEDSTRING="mkfsoptions>-Llabel_conf_14"
567 if [ "$EXPECTEDSTRING" != "$FOUNDSTRING" ]; then
568 echo "Error: expected: $EXPECTEDSTRING; found: $FOUNDSTRING"
571 echo "Success:mkfsoptions for ost written to xml file correctly."
573 # mount lustre to test lconf mkfsoptions-parsing
577 mount_client $MOUNT || return $?
578 if [ -z "`do_facet ost1 dumpe2fs -h $OSTDEV | grep label_conf_14`" ]; then
579 echo "Error: the mkoptions not applied to mke2fs of ost."
583 echo "lconf mkfsoptions for ost success"
587 run_test 14 "test mkfsoptions of ost for lmc and lconf"
591 [ -f $MOUNTLUSTRE ] && echo "remove $MOUNTLUSTRE" && rm -f $MOUNTLUSTRE
592 if [ -f $MOUNTLUSTRE.sav ]; then
593 echo "return original $MOUNTLUSTRE.sav to $MOUNTLUSTRE"
594 mv $MOUNTLUSTRE.sav $MOUNTLUSTRE
598 # this only tests the kernel mount command, not anything about lustre.
600 MOUNTLUSTRE=${MOUNTLUSTRE:-/sbin/mount.lustre}
604 echo "mount lustre on ${MOUNT} without $MOUNTLUSTRE....."
605 if [ -f "$MOUNTLUSTRE" ]; then
606 echo "save $MOUNTLUSTRE to $MOUNTLUSTRE.sav"
607 mv $MOUNTLUSTRE $MOUNTLUSTRE.sav && trap cleanup_15 EXIT INT
608 if [ -f $MOUNTLUSTRE ]; then
609 skip "$MOUNTLUSTRE cannot be moved, skipping test"
614 mount_client $MOUNT && error "mount succeeded" && return 1
615 echo "mount lustre on $MOUNT without $MOUNTLUSTRE failed as expected"
619 run_test 15 "zconf-mount without /sbin/mount.lustre (should return error)"
621 # LOGS/PENDING do not exist anymore since CMD3
623 TMPMTPT="${MOUNT%/*}/conf16"
625 if [ ! -e "$MDSDEV" ]; then
626 log "no $MDSDEV existing, so mount Lustre to create one"
628 check_mount || return 41
632 [ -f "$MDSDEV" ] && LOOPOPT="-o loop"
634 log "change the mode of $MDSDEV/OBJECTS to 555"
635 do_facet mds "mkdir -p $TMPMTPT &&
636 mount $LOOPOPT -t $FSTYPE $MDSDEV $TMPMTPT &&
637 chmod 555 $TMPMTPT/OBJECTS &&
638 umount $TMPMTPT" || return $?
640 log "mount Lustre to change the mode of OBJECTS, then umount Lustre"
642 check_mount || return 41
645 log "read the mode of OBJECTS and check if they has been changed properly"
646 EXPECTEDOBJECTSMODE=`do_facet mds "debugfs -R 'stat OBJECTS' $MDSDEV 2> /dev/null" | grep 'Mode: ' | sed -e "s/.*Mode: *//" -e "s/ *Flags:.*//"`
648 if [ "$EXPECTEDOBJECTSMODE" = "0777" ]; then
649 log "Success:Lustre change the mode of OBJECTS correctly"
651 error "Lustre does not change mode of OBJECTS properly"
654 run_test 16 "verify that lustre will correct the mode of OBJECTS"
657 if [ ! -e "$MDSDEV" ]; then
658 echo "no $MDSDEV existing, so mount Lustre to create one"
660 check_mount || return 41
664 echo "Remove mds config log"
665 do_facet mds "debugfs -w -R 'unlink CONFIGS/$FSNAME-MDT0000' $MDSDEV || return \$?" || return $?
668 start_mds && return 42
671 run_test 17 "Verify failed mds_postsetup won't fail assertion (2936) (should return errs)"
674 [ -f $MDSDEV ] && echo "remove $MDSDEV" && rm -f $MDSDEV
675 echo "mount mds with large journal..."
676 local myMDSSIZE=2000000
677 OLD_MDS_MKFS_OPTS=$MDS_MKFS_OPTS
679 MDS_MKFS_OPTS="--mgs --mdt --fsname=$FSNAME --device-size=$myMDSSIZE --param sys.timeout=$TIMEOUT $MDSOPT"
682 echo "mount lustre system..."
684 check_mount || return 41
686 echo "check journal size..."
687 FOUNDSIZE=`do_facet mds "debugfs -c -R 'stat <8>' $MDSDEV" | awk '/Size: / { print $NF; exit;}'`
688 if [ $FOUNDSIZE -gt $((32 * 1024 * 1024)) ]; then
689 log "Success: mkfs creates large journals. Size: $((FOUNDSIZE >> 20))M"
691 error "expected journal size > 32M, found $((FOUNDSIZE >> 20))M"
696 MDS_MKFS_OPTS=$OLD_MDS_MKFS_OPTS
699 run_test 18 "check mkfs creates large journals"
702 start_mds || return 1
703 stop_mds -f || return 2
705 run_test 19a "start/stop MDS without OSTs"
708 start_ost || return 1
709 stop_ost -f || return 2
711 run_test 19b "start/stop OSTs without MDS"
714 # first format the ost/mdt
718 check_mount || return 43
720 remount_client ro $MOUNT || return 44
721 touch $DIR/$tfile && echo "$DIR/$tfile created incorrectly" && return 45
722 [ -e $DIR/$tfile ] && echo "$DIR/$tfile exists incorrectly" && return 46
723 remount_client rw $MOUNT || return 47
725 [ ! -f $DIR/$tfile ] && echo "$DIR/$tfile missing" && return 48
726 MCNT=`grep -c $MOUNT /etc/mtab`
727 [ "$MCNT" -ne 1 ] && echo "$MOUNT in /etc/mtab $MCNT times" && return 49
732 run_test 20 "remount ro,rw mounts work and doesn't break /etc/mtab"
740 run_test 21a "start mds before ost, stop ost first"
748 run_test 21b "start ost before mds, stop mds first"
758 run_test 21c "start mds between two osts, stop mds last"
761 #reformat to remove all logs
764 echo Client mount before any osts are in the logs
766 check_mount && return 41
769 echo Client mount with ost in logs, but none running
773 # check_mount will block trying to contact ost
777 echo Client mount with a running ost
780 check_mount || return 41
785 run_test 22 "start a client before osts (should return errs)"
791 # force down client so that recovering mds waits for reconnect
792 zconf_umount `hostname` $MOUNT -f
793 # enter recovery on mds
795 # try to start a new client
796 mount_client $MOUNT &
799 MOUNT_LUSTRE_PID=`ps -ef | grep mount.lustre | grep -v grep | awk '{print $2}'`
800 echo mount pid is ${MOUNT_PID}, mount.lustre pid is ${MOUNT_LUSTRE_PID}
802 ps --ppid $MOUNT_LUSTRE_PID
803 # FIXME why o why can't I kill these? Manual "ctrl-c" works...
804 kill -TERM $MOUNT_PID
805 echo "waiting for mount to finish"
812 #this test isn't working yet
813 #run_test 23 "interrupt client during recovery mount delay"
819 local fs2mds_HOST=$mds_HOST
820 local fs2ost_HOST=$ost_HOST
821 [ -n "$ost1_HOST" ] && fs2ost_HOST=$ost1_HOST
822 if [ -z "$fs2ost_DEV" -o -z "$fs2mds_DEV" ]; then
823 do_facet $SINGLEMDS [ -b "$MDSDEV" ] && \
824 skip "mixed loopback and real device not working" && return
827 local fs2mdsdev=${fs2mds_DEV:-${MDSDEV}_2}
828 local fs2ostdev=${fs2ost_DEV:-$(ostdevname 1)_2}
830 # test 8-char fsname as well
831 local FSNAME2=test1234
832 add fs2mds $MDS_MKFS_OPTS --fsname=${FSNAME2} --nomgs --mgsnode=$MGSNID --reformat $fs2mdsdev || exit 10
834 add fs2ost $OST_MKFS_OPTS --fsname=${FSNAME2} --reformat $fs2ostdev || exit 10
837 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS
838 start fs2ost $fs2ostdev $OST_MOUNT_OPTS
840 mount -t lustre $MGSNID:/${FSNAME2} $MOUNT2 || return 1
842 check_mount || return 2
843 # files written on 1 should not show up on 2
844 cp /etc/passwd $DIR/$tfile
846 [ -e $MOUNT2/$tfile ] && error "File bleed" && return 7
849 cp /etc/passwd $MOUNT2/b || return 3
850 rm $MOUNT2/b || return 4
851 # 2 is actually mounted
852 grep $MOUNT2' ' /proc/mounts > /dev/null || return 5
854 facet_failover fs2mds
855 facet_failover fs2ost
858 # the MDS must remain up until last MDT
860 MDS=$(do_facet $SINGLEMDS "cat $LPROC/devices" | awk '($3 ~ "mdt" && $4 ~ "MDT") { print $4 }')
861 [ -z "$MDS" ] && error "No MDT" && return 8
865 cleanup_nocli || return 6
867 run_test 24a "Multiple MDTs on a single node"
870 local fs2mds_HOST=$mds_HOST
871 if [ -z "$fs2mds_DEV" ]; then
872 do_facet $SINGLEMDS [ -b "$MDSDEV" ] && \
873 skip "mixed loopback and real device not working" && return
876 local fs2mdsdev=${fs2mds_DEV:-${MDSDEV}_2}
878 add fs2mds $MDS_MKFS_OPTS --fsname=${FSNAME}2 --mgs --reformat $fs2mdsdev || exit 10
880 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS && return 2
883 run_test 24b "Multiple MGSs on a single node (should return err)"
887 check_mount || return 2
888 local MODULES=$($LCTL modules | awk '{ print $2 }')
889 rmmod $MODULES 2>/dev/null || true
892 run_test 25 "Verify modules are referenced"
896 # we need modules before mount for sysctl, so make sure...
897 do_facet mds "lsmod | grep -q lustre || modprobe lustre"
898 #define OBD_FAIL_MDS_FS_SETUP 0x135
899 do_facet mds "sysctl -w lustre.fail_loc=0x80000135"
900 start_mds && echo MDS started && return 1
902 DEVS=$(cat $LPROC/devices | wc -l)
903 [ $DEVS -gt 0 ] && return 2
904 unload_modules || return 203
906 run_test 26 "MDT startup failure cleans LOV (should return errs)"
912 local ORIG=$(do_facet $myfacet "$TEST")
913 if [ $# -gt 3 ]; then
919 echo "Setting $PARAM from $ORIG to $FINAL"
920 do_facet mds "$LCTL conf_param $PARAM=$FINAL" || error conf_param failed
926 RESULT=$(do_facet $myfacet "$TEST")
927 if [ $RESULT -eq $FINAL ]; then
928 echo "Updated config after $WAIT sec (got $RESULT)"
932 if [ $WAIT -eq $MAX ]; then
933 echo "Config update not seen: wanted $FINAL got $RESULT"
936 echo "Waiting $(($MAX - $WAIT)) secs for config update"
941 start_ost || return 1
942 start_mds || return 2
943 echo "Requeue thread should have started: "
944 ps -e | grep ll_cfg_requeue
945 set_and_check ost1 "cat $LPROC/obdfilter/$FSNAME-OST0000/client_cache_seconds" "$FSNAME-OST0000.ost.client_cache_seconds" || return 3
948 run_test 27a "Reacquire MGS lock if OST started first"
953 set_and_check mds "cat $LPROC/mdt/$FSNAME-MDT0000/identity_acquire_expire" "$FSNAME-MDT0000.mdt.identity_acquire_expire" || return 3
954 set_and_check client "cat $LPROC/mdc/$FSNAME-MDT0000-mdc-*/max_rpcs_in_flight" "$FSNAME-MDT0000.mdc.max_rpcs_in_flight" || return 4
957 run_test 27b "Reacquire MGS lock after failover"
961 TEST="cat $LPROC/llite/$FSNAME-*/max_read_ahead_whole_mb"
964 FINAL=$(($ORIG + 10))
965 set_and_check client "$TEST" "$FSNAME.llite.max_read_ahead_whole_mb" || return 3
966 set_and_check client "$TEST" "$FSNAME.llite.max_read_ahead_whole_mb" || return 3
967 umount_client $MOUNT || return 200
970 if [ $RESULT -ne $FINAL ]; then
971 echo "New config not seen: wanted $FINAL got $RESULT"
974 echo "New config success: got $RESULT"
978 run_test 28 "permanent parameter setting"
981 [ "$OSTCOUNT" -lt "2" ] && skip "$OSTCOUNT < 2, skipping" && return
982 setup > /dev/null 2>&1
986 local PARAM="$FSNAME-OST0001.osc.active"
987 local PROC_ACT="$LPROC/osc/$FSNAME-OST0001-osc-[^M]*/active"
988 local PROC_UUID="$LPROC/osc/$FSNAME-OST0001-osc-[^M]*/ost_server_uuid"
989 if [ ! -r $PROC_ACT ]; then
990 echo "Can't read $PROC_ACT"
991 ls $LPROC/osc/$FSNAME-*
994 ACTV=$(cat $PROC_ACT)
996 set_and_check client "cat $PROC_ACT" "$PARAM" $DEAC || return 2
997 # also check ost_server_uuid status
998 RESULT=$(grep DEACTIV $PROC_UUID)
999 if [ -z "$RESULT" ]; then
1000 echo "Live client not deactivated: $(cat $PROC_UUID)"
1003 echo "Live client success: got $RESULT"
1007 local MPROC="$LPROC/osc/$FSNAME-OST0001-osc-[M]*/active"
1012 RESULT=`do_facet mds " [ -r $MPROC ] && cat $MPROC"`
1013 [ ${PIPESTATUS[0]} = 0 ] || error "Can't read $MPROC"
1014 if [ $RESULT -eq $DEAC ]; then
1015 echo "MDT deactivated also after $WAIT sec (got $RESULT)"
1019 if [ $WAIT -eq $MAX ]; then
1020 echo "MDT not deactivated: wanted $DEAC got $RESULT"
1023 echo "Waiting $(($MAX - $WAIT)) secs for MDT deactivated"
1026 # test new client starts deactivated
1027 umount_client $MOUNT || return 200
1029 RESULT=$(grep DEACTIV $PROC_UUID | grep NEW)
1030 if [ -z "$RESULT" ]; then
1031 echo "New client not deactivated from start: $(cat $PROC_UUID)"
1034 echo "New client success: got $RESULT"
1037 # make sure it reactivates
1038 set_and_check client "cat $PROC_ACT" "$PARAM" $ACTV || return 6
1040 umount_client $MOUNT
1043 #writeconf to remove all ost2 traces for subsequent tests
1046 run_test 29 "permanently remove an OST"
1049 # start mds first after writeconf
1053 TEST="cat $LPROC/llite/$FSNAME-*/max_read_ahead_whole_mb"
1055 for i in $(seq 1 20); do
1056 set_and_check client "$TEST" "$FSNAME.llite.max_read_ahead_whole_mb" $i || return 3
1058 # make sure client restart still works
1059 umount_client $MOUNT
1060 mount_client $MOUNT || return 4
1061 [ "$($TEST)" -ne "$i" ] && return 5
1062 set_and_check client "$TEST" "$FSNAME.llite.max_read_ahead_whole_mb" $ORIG || return 6
1065 run_test 30 "Big config llog"
1067 test_31() { # bug 10734
1068 # ipaddr must not exist
1069 mount -t lustre 4.3.2.1@tcp:/lustre $MOUNT || true
1072 run_test 31 "Connect to non-existent node (shouldn't crash)"
1075 # XXX - make this run on client-only systems with real hardware on
1077 # there appears to be a lot of assumption here about loopback
1079 # or maybe this test is just totally useless on a client-only system
1080 [ "$mds_HOST" = "`hostname`" ] || { skip "remote MDS" && return 0; }
1081 [ "$ost_HOST" = "`hostname`" -o "$ost1_HOST" = "`hostname`" ] || \
1082 { skip "remote OST" && return 0; }
1084 [ -z "$TUNEFS" ] && skip "No tunefs" && return
1085 local DISK1_4=$LUSTRE/tests/disk1_4.zip
1086 [ ! -r $DISK1_4 ] && skip "Cant find $DISK1_4, skipping" && return
1088 unzip -o -j -d $TMP/$tdir $DISK1_4 || { skip "Cant unzip $DISK1_4, skipping" && return ; }
1090 sysctl lnet.debug=$PTLDEBUG
1092 $TUNEFS $TMP/$tdir/mds || error "tunefs failed"
1093 # nids are wrong, so client wont work, but server should start
1094 start mds $TMP/$tdir/mds "-o loop,exclude=lustre-OST0000" || return 3
1095 local UUID=$(cat $LPROC/mds/lustre-MDT0000/uuid)
1097 [ "$UUID" == "mdsA_UUID" ] || error "UUID is wrong: $UUID"
1099 $TUNEFS --mgsnode=`hostname` $TMP/$tdir/ost1 || error "tunefs failed"
1100 start ost1 $TMP/$tdir/ost1 "-o loop" || return 5
1101 UUID=$(cat $LPROC/obdfilter/lustre-OST0000/uuid)
1103 [ "$UUID" == "ost1_UUID" ] || error "UUID is wrong: $UUID"
1105 local NID=$($LCTL list_nids | head -1)
1107 echo "OSC changes should return err:"
1108 $LCTL conf_param lustre-OST0000.osc.max_dirty_mb=15 && return 7
1109 $LCTL conf_param lustre-OST0000.failover.node=$NID && return 8
1111 echo "MDC changes should succeed:"
1112 $LCTL conf_param lustre-MDT0000.mdc.max_rpcs_in_flight=9 || return 9
1113 $LCTL conf_param lustre-MDT0000.failover.node=$NID || return 10
1116 # With a new good MDT failover nid, we should be able to mount a client
1117 # (but it cant talk to OST)
1118 local OLDMOUNTOPT=$MOUNTOPT
1119 MOUNTOPT="exclude=lustre-OST0000"
1121 MOUNTOPT=$OLDMOUNTOPT
1122 set_and_check client "cat $LPROC/mdc/*/max_rpcs_in_flight" "lustre-MDT0000.mdc.max_rpcs_in_flight" || return 11
1124 zconf_umount `hostname` $MOUNT -f
1127 # mount a second time to make sure we didnt leave upgrade flag on
1128 $TUNEFS --dryrun $TMP/$tdir/mds || error "tunefs failed"
1129 start mds $TMP/$tdir/mds "-o loop,exclude=lustre-OST0000" || return 12
1132 [ -d $TMP/$tdir ] && rm -rf $TMP/$tdir
1134 run_test 32a "Upgrade from 1.4 (not live)"
1137 # XXX - make this run on client-only systems with real hardware on
1139 # there appears to be a lot of assumption here about loopback
1141 # or maybe this test is just totally useless on a client-only system
1142 [ "$mds_HOST" = "`hostname`" ] || { skip "remote MDS" && return 0; }
1143 [ "$ost_HOST" = "`hostname`" -o "$ost1_HOST" = "`hostname`" ] || \
1144 { skip "remote OST" && return 0; }
1146 [ -z "$TUNEFS" ] && skip "No tunefs" && return
1147 local DISK1_4=$LUSTRE/tests/disk1_4.zip
1148 [ ! -r $DISK1_4 ] && skip "Cant find $DISK1_4, skipping" && return
1150 unzip -o -j -d $TMP/$tdir $DISK1_4 || { skip "Cant unzip $DISK1_4, skipping" && return ; }
1152 sysctl lnet.debug=$PTLDEBUG
1154 # writeconf will cause servers to register with their current nids
1155 $TUNEFS --writeconf $TMP/$tdir/mds || error "tunefs failed"
1156 start mds $TMP/$tdir/mds "-o loop" || return 3
1157 local UUID=$(cat $LPROC/mds/lustre-MDT0000/uuid)
1159 [ "$UUID" == "mdsA_UUID" ] || error "UUID is wrong: $UUID"
1161 $TUNEFS --mgsnode=`hostname` $TMP/$tdir/ost1 || error "tunefs failed"
1162 start ost1 $TMP/$tdir/ost1 "-o loop" || return 5
1163 UUID=$(cat $LPROC/obdfilter/lustre-OST0000/uuid)
1165 [ "$UUID" == "ost1_UUID" ] || error "UUID is wrong: $UUID"
1167 echo "OSC changes should succeed:"
1168 $LCTL conf_param lustre-OST0000.osc.max_dirty_mb=15 || return 7
1169 $LCTL conf_param lustre-OST0000.failover.node=$NID || return 8
1171 echo "MDC changes should succeed:"
1172 $LCTL conf_param lustre-MDT0000.mdc.max_rpcs_in_flight=9 || return 9
1175 # MDT and OST should have registered with new nids, so we should have
1176 # a fully-functioning client
1177 echo "Check client and old fs contents"
1179 set_and_check client "cat $LPROC/mdc/*/max_rpcs_in_flight" "${NEWNAME}-MDT0000.mdc.max_rpcs_in_flight" || return 11
1180 [ "$(cksum $MOUNT/passwd | cut -d' ' -f 1,2)" == "2479747619 779" ] || return 12
1184 [ -d $TMP/$tdir ] && rm -rf $TMP/$tdir
1186 run_test 32b "Upgrade from 1.4 with writeconf"
1188 test_33() { # bug 12333
1189 local FSNAME2=test1234
1190 local fs2mds_HOST=$mds_HOST
1191 local fs2ost_HOST=$ost_HOST
1192 [ -n "$ost1_HOST" ] && fs2ost_HOST=$ost1_HOST
1194 if [ -z "$fs2ost_DEV" -o -z "$fs2mds_DEV" ]; then
1195 do_facet $SINGLEMDS [ -b "$MDSDEV" ] && \
1196 skip "mixed loopback and real device not working" && return
1199 local fs2mdsdev=${fs2mds_DEV:-${MDSDEV}_2}
1200 local fs2ostdev=${fs2ost_DEV:-$(ostdevname 1)_2}
1201 add fs2mds $MDS_MKFS_OPTS --fsname=${FSNAME2} --reformat $fs2mdsdev || exit 10
1202 add fs2ost $OST_MKFS_OPTS --fsname=${FSNAME2} --index=8191 --mgsnode=$MGSNID --reformat $fs2ostdev || exit 10
1204 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS
1205 start fs2ost $fs2ostdev $OST_MOUNT_OPTS
1207 mount -t lustre $MGSNID:/${FSNAME2} $MOUNT2 || return 1
1213 rm -rf $MOUNT2 $fs2mdsdev $fs2ostdev
1214 cleanup_nocli || return 6
1216 run_test 33 "Mount ost with a large index number"
1221 do_facet client dd if=/dev/zero of=$MOUNT/24 bs=1024k count=1
1222 # Drop lock cancelation reply during umount
1223 #define OBD_FAIL_LDLM_CANCEL 0x304
1224 do_facet client sysctl -w lustre.fail_loc=0x80000304
1225 #sysctl -w lnet.debug=-1
1226 umount_client $MOUNT
1229 run_test 34 "Drop cancel during umount"
1233 do_facet client multiop $DIR/file O_c &
1235 manual_umount_client
1237 do_facet client killall -USR1 multiop
1238 if [ $rc -eq 0 ]; then
1239 error "umount not fail!"
1244 run_test 34a "umount with opened file should be fail"
1249 touch $DIR/$tfile || return 1
1250 stop_mds --force || return 2
1252 manual_umount_client --force
1254 if [ $rc -ne 0 ]; then
1255 error "mtab after failed umount - rc $rc"
1261 run_test 34b "force umount with failed mds should be normal"
1265 touch $DIR/$tfile || return 1
1266 stop_ost --force || return 2
1268 manual_umount_client --force
1270 if [ $rc -ne 0 ]; then
1271 error "mtab after failed umount - rc $rc"
1277 run_test 34c "force umount with failed ost should be normal"
1279 test_35() { # bug 12459
1282 DBG_SAVE="`sysctl -n lnet.debug`"
1283 sysctl -w lnet.debug="ha"
1285 log "Set up a fake failnode for the MDS"
1287 do_facet mds $LCTL conf_param ${FSNAME}-MDT0000.failover.node=$FAKENID || return 4
1289 log "Wait for RECONNECT_INTERVAL seconds (10s)"
1292 MSG="conf-sanity.sh test_35 `date +%F%kh%Mm%Ss`"
1295 log "Stopping the MDT:"
1296 stop_mds || return 5
1298 df $MOUNT > /dev/null 2>&1 &
1300 log "Restarting the MDT:"
1301 start_mds || return 6
1302 log "Wait for df ($DFPID) ... "
1305 sysctl -w lnet.debug="$DBG_SAVE"
1307 # retrieve from the log the first server that the client tried to
1308 # contact after the connection loss
1309 $LCTL dk $TMP/lustre-log-$TESTNAME.log
1310 NEXTCONN=`awk "/${MSG}/ {start = 1;}
1311 /import_select_connection.*${FSNAME}-MDT0000-mdc.* using connection/ {
1313 if (\\\$NF ~ /$FAKENID/)
1319 }" $TMP/lustre-log-$TESTNAME.log`
1320 [ "$NEXTCONN" != "0" ] && log "The client didn't try to reconnect to the last active server (tried ${NEXTCONN} instead)" && return 7
1323 run_test 35 "Reconnect to the last active server first"
1327 local FSNAME2=test1234
1328 local fs2mds_HOST=$mds_HOST
1329 local fs2ost_HOST=$ost_HOST
1330 local fs3ost_HOST=$ost_HOST
1332 [ -n "$ost1_HOST" ] && fs2ost_HOST=$ost1_HOST && fs3ost_HOST=$ost1_HOST
1335 if [ -z "$fs2ost_DEV" -o -z "$fs2mds_DEV" -o -z "$fs3ost_DEV" ]; then
1336 do_facet $SINGLEMDS [ -b "$MDSDEV" ] && \
1337 skip "mixed loopback and real device not working" && return
1339 [ $OSTCOUNT -lt 2 ] && skip "skipping test for single OST" && return
1341 [ $(grep -c obdfilter $LPROC/devices) -eq 0 ] &&
1342 skip "skipping test for remote OST" && return
1344 local fs2mdsdev=${fs2mds_DEV:-${MDSDEV}_2}
1345 local fs2ostdev=${fs2ost_DEV:-$(ostdevname 1)_2}
1346 local fs3ostdev=${fs3ost_DEV:-$(ostdevname 2)_2}
1347 add fs2mds $MDS_MKFS_OPTS --fsname=${FSNAME2} --reformat $fs2mdsdev || exit 10
1348 add fs2ost $OST_MKFS_OPTS --mkfsoptions='-b1024' --fsname=${FSNAME2} --mgsnode=$MGSNID --reformat $fs2ostdev || exit 10
1349 add fs3ost $OST_MKFS_OPTS --mkfsoptions='-b4096' --fsname=${FSNAME2} --mgsnode=$MGSNID --reformat $fs3ostdev || exit 10
1351 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS
1352 start fs2ost $fs2ostdev $OST_MOUNT_OPTS
1353 start fs3ost $fs3ostdev $OST_MOUNT_OPTS
1355 mount -t lustre $MGSNID:/${FSNAME2} $MOUNT2 || return 1
1357 sleep 5 # until 11778 fixed
1359 dd if=/dev/zero of=$MOUNT2/$tfile bs=1M count=7 || return 2
1361 BKTOTAL=`awk 'BEGIN{total=0}; {total+=$1}; END{print total}' \
1362 $LPROC/obdfilter/*/kbytestotal`
1363 BKFREE=`awk 'BEGIN{free=0}; {free+=$1}; END{print free}' \
1364 $LPROC/obdfilter/*/kbytesfree`
1365 BKAVAIL=`awk 'BEGIN{avail=0}; {avail+=$1}; END{print avail}' \
1366 $LPROC/obdfilter/*/kbytesavail`
1367 STRING=`df -P $MOUNT2 | tail -n 1 | awk '{print $2","$3","$4}'`
1368 DFTOTAL=`echo $STRING | cut -d, -f1`
1369 DFUSED=`echo $STRING | cut -d, -f2`
1370 DFAVAIL=`echo $STRING | cut -d, -f3`
1371 DFFREE=$(($DFTOTAL - $DFUSED))
1373 ALLOWANCE=$((64 * $OSTCOUNT))
1375 if [ $DFTOTAL -lt $(($BKTOTAL - $ALLOWANCE)) ] ||
1376 [ $DFTOTAL -gt $(($BKTOTAL + $ALLOWANCE)) ] ; then
1377 echo "**** FAIL: df total($DFTOTAL) mismatch OST total($BKTOTAL)"
1380 if [ $DFFREE -lt $(($BKFREE - $ALLOWANCE)) ] ||
1381 [ $DFFREE -gt $(($BKFREE + $ALLOWANCE)) ] ; then
1382 echo "**** FAIL: df free($DFFREE) mismatch OST free($BKFREE)"
1385 if [ $DFAVAIL -lt $(($BKAVAIL - $ALLOWANCE)) ] ||
1386 [ $DFAVAIL -gt $(($BKAVAIL + $ALLOWANCE)) ] ; then
1387 echo "**** FAIL: df avail($DFAVAIL) mismatch OST avail($BKAVAIL)"
1392 stop fs3ost -f || return 200
1393 stop fs2ost -f || return 201
1394 stop fs2mds -f || return 202
1395 rm -rf $MOUNT2 $fs2mdsdev $fs2ostdev $fs3ostdev
1396 unload_modules || return 203
1399 run_test 36 "df report consistency on OSTs with different block size"
1402 LOCAL_MDSDEV="$TMP/mdt.img"
1403 SYM_MDSDEV="$TMP/sym_mdt.img"
1405 echo "MDS : $LOCAL_MDSDEV"
1406 echo "SYMLINK : $SYM_MDSDEV"
1410 mkfs.lustre --reformat --fsname=lustre --mdt --mgs --device-size=9000 $LOCAL_MDSDEV ||
1411 error "mkfs.lustre $LOCAL_MDSDEV failed"
1412 ln -s $LOCAL_MDSDEV $SYM_MDSDEV
1414 echo "mount symlink device - $SYM_MDSDEV"
1416 mount_op=`mount -v -t lustre -o loop $SYM_MDSDEV ${MOUNT%/*}/mds 2>&1 | grep "unable to set tunable"`
1417 umount -d ${MOUNT%/*}/mds
1418 rm -f $LOCAL_MDSDEV $SYM_MDSDEV
1420 if [ -n "$mount_op" ]; then
1421 error "**** FAIL: set tunables failed for symlink device"
1425 run_test 37 "verify set tunables works for symlink device"
1427 umount_client $MOUNT
1431 equals_msg `basename $0`: test complete
1432 [ -f "$TESTSUITELOG" ] && cat $TESTSUITELOG || true