7 # bug number for skipped test:
8 ALWAYS_EXCEPT="$CONF_SANITY_EXCEPT "
9 # UPDATE THE COMMENT ABOVE WITH BUG NUMBERS WHEN CHANGING ALWAYS_EXCEPT!
13 if [ -r /etc/SuSE-release ]
15 local vers=$(grep VERSION /etc/SuSE-release | awk '{print $3}')
16 local patchlev=$(grep PATCHLEVEL /etc/SuSE-release |
18 if [ $vers -eq 11 ] && [ $patchlev -eq 2 ]
26 if is_sles11; then # LU-2181
27 ALWAYS_EXCEPT="$ALWAYS_EXCEPT 23a 34b"
30 if [ "$FAILURE_MODE" = "HARD" ]; then
31 CONFIG_EXCEPTIONS="24a " &&
32 echo "Except the tests: $CONFIG_EXCEPTIONS for " \
33 "FAILURE_MODE=$FAILURE_MODE, b=23573" &&
34 ALWAYS_EXCEPT="$ALWAYS_EXCEPT $CONFIG_EXCEPTIONS"
37 # bug number for skipped test:
38 # a tool to create lustre filesystem images
39 ALWAYS_EXCEPT="32newtarball $ALWAYS_EXCEPT"
42 PATH=$PWD/$SRCDIR:$SRCDIR:$SRCDIR/../utils:$PATH
44 PTLDEBUG=${PTLDEBUG:--1}
46 LUSTRE=${LUSTRE:-$(dirname $0)/..}
47 RLUSTRE=${RLUSTRE:-$LUSTRE}
48 export MULTIOP=${MULTIOP:-multiop}
50 . $LUSTRE/tests/test-framework.sh
52 . ${CONFIG:=$LUSTRE/tests/cfg/$NAME.sh}
54 # use small MDS + OST size to speed formatting time
55 # do not use too small MDSSIZE/OSTSIZE, which affect the default journal size
56 # STORED_MDSSIZE is used in test_18
57 STORED_MDSSIZE=$MDSSIZE
58 STORED_OSTSIZE=$OSTSIZE
70 if ! combined_mgs_mds; then
71 # bug number for skipped test: 23954 LU-9860 LU-9860 LU-9860 LU-9860
72 ALWAYS_EXCEPT="$ALWAYS_EXCEPT 24b 33a 43b 53b 54b"
73 # bug number for skipped test: LU-9875 LU-9879 LU-9879 LU-9879 LU-9879
74 ALWAYS_EXCEPT="$ALWAYS_EXCEPT 70e 80 84 87 100"
75 # bug number for skipped test: LU-8110 LU-9400 LU-9879 LU-9879 LU-9879
76 ALWAYS_EXCEPT="$ALWAYS_EXCEPT 102 103 104 105 107"
79 # pass "-E lazy_itable_init" to mke2fs to speed up the formatting time
80 if [[ "$LDISKFS_MKFS_OPTS" != *lazy_itable_init* ]]; then
81 LDISKFS_MKFS_OPTS=$(csa_add "$LDISKFS_MKFS_OPTS" -E lazy_itable_init)
84 [ $(facet_fstype $SINGLEMDS) = "zfs" ] &&
85 # bug number for skipped test:
86 ALWAYS_EXCEPT="$ALWAYS_EXCEPT"
87 # UPDATE THE COMMENT ABOVE WITH BUG NUMBERS WHEN CHANGING ALWAYS_EXCEPT!
92 require_dsh_mds || exit 0
93 require_dsh_ost || exit 0
96 [ "$SLOW" = "no" ] && EXCEPT_SLOW="45 69"
101 # The MGS must be started before the OSTs for a new fs, so start
102 # and stop to generate the startup logs.
105 wait_osc_import_state mds ost FULL
110 reformat_and_config() {
112 if ! combined_mgs_mds ; then
118 writeconf_or_reformat() {
119 # There are at most 2 OSTs for write_conf test
120 # who knows if/where $TUNEFS is installed?
121 # Better reformat if it fails...
122 writeconf_all $MDSCOUNT 2 ||
123 { echo "tunefs failed, reformatting instead" &&
124 reformat_and_config && return 0; }
133 echo "start mgs service on $(facet_active_host mgs)"
134 start mgs $(mgsdevname) $MGS_MOUNT_OPTS $@
140 local dev=$(mdsdevname $num)
143 echo "start mds service on `facet_active_host $facet`"
144 start $facet ${dev} $MDS_MOUNT_OPTS $@ || return 94
150 local dev=$(mdsdevname $num)
153 echo "stop mds service on `facet_active_host $facet`"
154 # These tests all use non-failover stop
155 stop $facet -f || return 97
161 for num in $(seq $MDSCOUNT); do
162 start_mdt $num $@ || return 94
167 if ! combined_mgs_mds ; then
175 for num in $(seq $MDSCOUNT); do
176 stop_mdt $num || return 97
181 echo "stop mgs service on `facet_active_host mgs`"
182 # These tests all use non-failover stop
183 stop mgs -f || return 97
187 echo "start ost1 service on `facet_active_host ost1`"
188 start ost1 $(ostdevname 1) $OST_MOUNT_OPTS $@ || return 95
192 echo "stop ost1 service on `facet_active_host ost1`"
193 # These tests all use non-failover stop
194 stop ost1 -f || return 98
198 echo "start ost2 service on `facet_active_host ost2`"
199 start ost2 $(ostdevname 2) $OST_MOUNT_OPTS $@ || return 92
203 echo "stop ost2 service on `facet_active_host ost2`"
204 # These tests all use non-failover stop
205 stop ost2 -f || return 93
210 echo "mount $FSNAME on ${MOUNTPATH}....."
211 zconf_mount $(hostname) $MOUNTPATH || return 96
215 local mountopt="remount,$1"
217 echo "remount '$1' lustre on ${MOUNTPATH}....."
218 zconf_mount $(hostname) $MOUNTPATH "$mountopt" || return 96
224 echo "umount lustre on $mountpath....."
225 zconf_umount $HOSTNAME $mountpath $@ || return 97
228 manual_umount_client(){
231 echo "manual umount lustre on ${MOUNT}...."
232 do_facet client "umount ${FORCE} $MOUNT"
238 start_mds || error "MDT start failed"
239 start_ost || error "Unable to start OST1"
240 mount_client $MOUNT || error "client start failed"
241 client_up || error "client_up failed"
250 unload_modules_conf () {
251 if combined_mgs_mds || ! local_mode; then
252 unload_modules || return 1
257 stop_ost || return 202
258 stop_mds || return 201
259 unload_modules_conf || return 203
264 [ "x$1" != "x" ] && force='-f'
265 umount_client $MOUNT $force|| return 200
266 cleanup_nocli || return $?
271 echo "umount $MOUNT2 ..."
272 umount $MOUNT2 || true
273 echo "stopping fs2mds ..."
274 stop fs2mds -f || true
275 echo "stopping fs2ost ..."
276 stop fs2ost -f || true
280 do_facet client "cp /etc/passwd $DIR/a" || return 71
281 do_facet client "rm $DIR/a" || return 72
282 # make sure lustre is actually mounted (touch will block,
283 # but grep won't, so do it after)
284 do_facet client "grep $MOUNT' ' /proc/mounts > /dev/null" || return 73
285 echo "setup single mount lustre success"
289 do_facet client "touch $DIR/a" || return 71
290 do_facet client "rm $DIR/a" || return 72
291 do_facet client "touch $DIR2/a" || return 73
292 do_facet client "rm $DIR2/a" || return 74
293 echo "setup double mount lustre success"
298 if [ "$ONLY" == "setup" ]; then
303 if [ "$ONLY" == "cleanup" ]; then
310 #create single point mountpoint
316 check_mount || error "check_mount failed"
317 cleanup || error "cleanup failed with $?"
319 run_test 0 "single mount setup"
322 start_mds || error "MDS start failed"
323 start_ost || error "unable to start OST"
324 echo "start ost second time..."
325 start_ost && error "2nd OST start should fail"
326 mount_client $MOUNT || error "client start failed"
327 check_mount || error "check_mount failed"
328 cleanup || error "cleanup failed with $?"
330 run_test 1 "start up ost twice (should return errors)"
333 start_mds || error "MDT start failed"
334 echo "start mds second time.."
335 start_mds && error "2nd MDT start should fail"
336 start_ost || error "OST start failed"
337 mount_client $MOUNT || error "mount_client failed to start client"
338 check_mount || error "check_mount failed"
339 cleanup || error "cleanup failed with $?"
341 run_test 2 "start up mds twice (should return err)"
345 #mount.lustre returns an error if already in mtab
346 mount_client $MOUNT && error "2nd client mount should fail"
347 check_mount || error "check_mount failed"
348 cleanup || error "cleanup failed with $?"
350 run_test 3 "mount client twice (should return err)"
354 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
355 stop_ost || error "Unable to stop OST1"
356 umount_client $MOUNT -f || error “unmount $MOUNT failed”
359 # ok for ost to fail shutdown
360 if [ 202 -ne $eno ] && [ 0 -ne $eno ]; then
361 error "cleanup failed with $?"
364 run_test 4 "force cleanup ost, then cleanup"
366 test_5a() { # was test_5
368 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
369 fuser -m -v $MOUNT && echo "$MOUNT is in use by user space process."
371 stop_mds || error "Unable to stop MDS"
373 # cleanup may return an error from the failed
374 # disconnects; for now I'll consider this successful
375 # if all the modules have unloaded.
379 echo "killing umount"
380 kill -TERM $UMOUNT_PID
381 echo "waiting for umount to finish"
383 if grep " $MOUNT " /proc/mounts; then
384 echo "test 5: /proc/mounts after failed umount"
388 echo "killing umount"
389 kill -TERM $UMOUNT_PID
390 echo "waiting for umount to finish"
392 grep " $MOUNT " /proc/mounts &&
393 error "/proc/mounts after second umount"
396 # manual_mount_client may fail due to umount succeeding above
398 # stop_mds is a no-op here, and should not fail
399 cleanup_nocli || error "cleanup_nocli failed with $?"
400 # df may have lingering entry
402 # mtab may have lingering entry
406 while [ "$WAIT" -ne "$MAX_WAIT" ]; do
408 grep -q $MOUNT" " /etc/mtab || break
409 echo "Waiting /etc/mtab updated ... "
410 WAIT=$(( WAIT + sleep))
412 [ "$WAIT" -eq "$MAX_WAIT" ] &&
413 error "/etc/mtab is not updated in $WAIT secs"
414 echo "/etc/mtab updated in $WAIT secs"
416 run_test 5a "force cleanup mds, then cleanup"
424 grep " $MOUNT " /etc/mtab &&
425 error false "unexpected entry in mtab before mount" && return 10
427 start_ost || error "OST start failed"
428 if ! combined_mgs_mds ; then
429 trap cleanup_5b EXIT ERR
430 start_mds || error "MDS start failed"
434 mount_client $MOUNT && error "mount_client $MOUNT should fail"
435 grep " $MOUNT " /etc/mtab &&
436 error "$MOUNT entry in mtab after failed mount"
438 # stop_mds is a no-op here, and should not fail
439 cleanup_nocli || error "cleanup_nocli failed with $?"
440 if ! combined_mgs_mds ; then
444 run_test 5b "Try to start a client with no MGS (should return errs)"
447 grep " $MOUNT " /etc/mtab &&
448 error false "unexpected entry in mtab before mount" && return 10
450 start_mds || error "MDS start failed"
451 start_ost || error "OST start failed"
452 local oldfs="${FSNAME}"
453 FSNAME="wrong.${FSNAME}"
454 mount_client $MOUNT || :
456 grep " $MOUNT " /etc/mtab &&
457 error "$MOUNT entry in mtab after failed mount"
459 cleanup_nocli || error "cleanup_nocli failed with $?"
461 run_test 5c "cleanup after failed mount (bug 2712) (should return errs)"
464 grep " $MOUNT " /etc/mtab &&
465 error "unexpected entry in mtab before mount"
467 start_ost || error "OST start failed"
468 start_mds || error "MDS start failed"
469 stop_ost -f || error "Unable to stop OST1"
470 mount_client $MOUNT || error "mount_client $MOUNT failed"
471 umount_client $MOUNT -f || error "umount_client $MOUNT failed"
472 cleanup_nocli || error "cleanup_nocli failed with $?"
473 ! grep " $MOUNT " /etc/mtab ||
474 error "$MOUNT entry in mtab after unmount"
476 run_test 5d "mount with ost down"
479 grep " $MOUNT " /etc/mtab &&
480 error false "unexpected entry in mtab before mount" && return 10
482 start_mds || error "MDS start failed"
483 start_ost || error "OST start failed"
485 #define OBD_FAIL_PTLRPC_DELAY_SEND 0x506
486 do_facet client "$LCTL set_param fail_loc=0x80000506"
487 mount_client $MOUNT || echo "mount failed (not fatal)"
488 cleanup || error "cleanup failed with $?"
489 grep " $MOUNT " /etc/mtab &&
490 error "$MOUNT entry in mtab after unmount"
493 run_test 5e "delayed connect, don't crash (bug 10268)"
496 if combined_mgs_mds ; then
497 skip "needs separate mgs and mds"
501 grep " $MOUNT " /etc/mtab &&
502 error false "unexpected entry in mtab before mount" && return 10
505 start_ost || error "OST start failed"
506 mount_client $MOUNT &
508 echo client_mount pid is $pid
512 if ! ps -f -p $pid >/dev/null; then
515 grep " $MOUNT " /etc/mtab && echo "test 5f: mtab after mount"
516 error "mount returns $rc, expected to hang"
518 cleanup || error "cleanup failed with $?"
523 start_mds || error "start MDS failed"
525 # mount should succeed after start mds
527 grep " $MOUNT " /etc/mtab && echo "test 5f: mtab after mount"
528 cleanup || error "final call to cleanup failed with rc $?"
530 run_test 5f "mds down, cleanup after failed mount (bug 2712)"
534 [ $(lustre_version_code client) -lt $(version_code 2.9.53) ] &&
535 { skip "automount of debugfs missing before 2.9.53" && return 0; }
536 umount /sys/kernel/debug
537 $LCTL get_param -n devices | egrep -v "error" && \
538 error "lctl can't access debugfs data"
539 grep " debugfs " /etc/mtab || error "debugfs failed to remount"
541 run_test 5g "handle missing debugfs"
546 mount_client $MOUNT || error "mount_client $MOUNT failed"
547 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
548 cleanup || error "cleanup failed with rc $?"
550 run_test 6 "manual umount, then mount again"
555 cleanup_nocli || error "cleanup_nocli failed with $?"
557 run_test 7 "manual umount, then cleanup"
561 mount_client $MOUNT2 || error "mount_client $MOUNT2 failed"
562 check_mount2 || error "check_mount2 failed"
563 umount_client $MOUNT2 || error "umount_client $MOUNT2 failed"
564 cleanup || error "cleanup failed with rc $?"
566 run_test 8 "double mount setup"
569 start_ost || error "OST start failed"
571 do_facet ost1 $LCTL set_param debug=\'inode trace\' ||
572 error "do_facet ost1 set_param inode trace failed."
573 do_facet ost1 $LCTL set_param subsystem_debug=\'mds ost\' ||
574 error "do_facet ost1 set_param debug mds ost failed."
576 CHECK_PTLDEBUG="`do_facet ost1 $LCTL get_param -n debug`"
577 if [ "$CHECK_PTLDEBUG" ] && { \
578 [ "$CHECK_PTLDEBUG" = "trace inode warning error emerg console" ] ||
579 [ "$CHECK_PTLDEBUG" = "trace inode" ]; }; then
580 echo "lnet.debug success"
582 error "lnet.debug: want 'trace inode', have '$CHECK_PTLDEBUG'"
584 CHECK_SUBSYS="`do_facet ost1 $LCTL get_param -n subsystem_debug`"
585 if [ "$CHECK_SUBSYS" ] && [ "$CHECK_SUBSYS" = "mds ost" ]; then
586 echo "lnet.subsystem_debug success"
588 error "lnet.subsystem_debug: want 'mds ost' got '$CHECK_SUBSYS'"
590 stop_ost || error "Unable to stop OST1"
592 run_test 9 "test ptldebug and subsystem for mkfs"
600 do_facet $facet "test -b $dev" || rc=1
601 if [[ "$size" ]]; then
602 local in=$(do_facet $facet "dd if=$dev of=/dev/null bs=1k \
603 count=1 skip=$size 2>&1" |
604 awk '($3 == "in") { print $1 }')
605 [[ $in = "1+0" ]] || rc=1
611 # Test 16 was to "verify that lustre will correct the mode of OBJECTS".
612 # But with new MDS stack we don't care about the mode of local objects
613 # anymore, so this test is removed. See bug 22944 for more details.
617 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
618 skip "ldiskfs only test"
623 check_mount || error "check_mount failed"
624 cleanup || error "cleanup failed with rc $?"
626 echo "Remove mds config log"
627 if ! combined_mgs_mds ; then
631 do_facet mgs "$DEBUGFS -w -R 'unlink CONFIGS/$FSNAME-MDT0000' \
632 $(mgsdevname) || return \$?" ||
633 error "do_facet mgs failed with $?"
635 if ! combined_mgs_mds ; then
639 start_ost || error "OST start failed"
640 start_mds && error "MDS start succeeded, but should fail"
643 run_test 17 "Verify failed mds_postsetup won't fail assertion (2936) (should return errs)"
646 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
647 skip "ldiskfs only test"
651 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
656 # check if current MDSSIZE is large enough
657 [ $MDSSIZE -ge $MIN ] && OK=1 && myMDSSIZE=$MDSSIZE &&
658 log "use MDSSIZE=$MDSSIZE"
660 # check if the global config has a large enough MDSSIZE
661 [ -z "$OK" -a ! -z "$STORED_MDSSIZE" ] &&
662 [ $STORED_MDSSIZE -ge $MIN ] &&
663 OK=1 && myMDSSIZE=$STORED_MDSSIZE &&
664 log "use STORED_MDSSIZE=$STORED_MDSSIZE"
666 # check if the block device is large enough
667 is_blkdev $SINGLEMDS $MDSDEV $MIN
668 local large_enough=$?
669 if [ -n "$OK" ]; then
670 [ $large_enough -ne 0 ] && OK=""
672 [ $large_enough -eq 0 ] && OK=1 && myMDSSIZE=$MIN &&
673 log "use device $MDSDEV with MIN=$MIN"
676 # check if a loopback device has enough space for fs metadata (5%)
678 if [ -z "$OK" ]; then
679 local SPACE=$(do_facet $SINGLEMDS "[ -f $MDSDEV -o ! \
680 -e $MDSDEV ] && df -P \\\$(dirname $MDSDEV)" |
681 awk '($1 != "Filesystem") { print $4 }')
682 ! [ -z "$SPACE" ] && [ $SPACE -gt $((MIN / 20)) ] &&
683 OK=1 && myMDSSIZE=$MIN &&
684 log "use file $MDSDEV with MIN=$MIN"
687 [ -z "$OK" ] && skip_env "$MDSDEV too small for ${MIN}kB MDS" && return
689 echo "mount mds with large journal..."
691 local OLD_MDSSIZE=$MDSSIZE
695 echo "mount lustre system..."
697 check_mount || error "check_mount failed"
699 echo "check journal size..."
700 local FOUNDSIZE=$(do_facet $SINGLEMDS "$DEBUGFS -c -R 'stat <8>' $MDSDEV" | awk '/Size: / { print $NF; exit;}')
701 if [ $FOUNDSIZE -gt $((32 * 1024 * 1024)) ]; then
702 log "Success: mkfs creates large journals. Size: $((FOUNDSIZE >> 20))M"
704 error "expected journal size > 32M, found $((FOUNDSIZE >> 20))M"
707 cleanup || error "cleanup failed with rc $?"
712 run_test 18 "check mkfs creates large journals"
715 start_mds || error "MDS start failed"
716 stop_mds || error "Unable to stop MDS"
718 run_test 19a "start/stop MDS without OSTs"
721 start_ost || error "Unable to start OST1"
722 stop_ost -f || error "Unable to stop OST1"
724 run_test 19b "start/stop OSTs without MDS"
727 # first format the ost/mdt
728 start_mds || error "MDS start failed"
729 start_ost || error "Unable to start OST1"
730 mount_client $MOUNT || error "mount_client $MOUNT failed"
731 check_mount || error "check_mount failed"
732 rm -f $DIR/$tfile || error "remove $DIR/$tfile failed."
733 remount_client ro $MOUNT || error "remount_client with ro failed"
734 touch $DIR/$tfile && error "$DIR/$tfile created incorrectly"
735 [ -e $DIR/$tfile ] && error "$DIR/$tfile exists incorrectly"
736 remount_client rw $MOUNT || error "remount_client with rw failed"
737 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
738 MCNT=$(grep -c $MOUNT /etc/mtab)
739 [ "$MCNT" -ne 1 ] && error "$MOUNT in /etc/mtab $MCNT times"
741 stop_mds || error "Unable to stop MDS"
742 stop_ost || error "Unable to stop OST1"
744 run_test 20 "remount ro,rw mounts work and doesn't break /etc/mtab"
747 start_mds || error "MDS start failed"
748 start_ost || error "unable to start OST1"
749 wait_osc_import_state mds ost FULL
750 stop_ost || error "unable to stop OST1"
751 stop_mds || error "unable to stop MDS"
753 run_test 21a "start mds before ost, stop ost first"
756 start_ost || error "unable to start OST1"
757 start_mds || error "MDS start failed"
758 wait_osc_import_state mds ost FULL
759 stop_mds || error "unable to stop MDS"
760 stop_ost || error "unable to stop OST1"
762 run_test 21b "start ost before mds, stop mds first"
765 start_ost || error "Unable to start OST1"
766 start_mds || error "MDS start failed"
767 start_ost2 || error "Unable to start OST2"
768 wait_osc_import_state mds ost2 FULL
769 stop_ost || error "Unable to stop OST1"
770 stop_ost2 || error "Unable to stop OST2"
771 stop_mds || error "Unable to stop MDS"
772 #writeconf to remove all ost2 traces for subsequent tests
773 writeconf_or_reformat
775 run_test 21c "start mds between two osts, stop mds last"
778 if combined_mgs_mds ; then
779 skip "need separate mgs device" && return 0
785 start_mgs || error "unable to start MGS"
786 start_ost || error "unable to start OST1"
787 start_ost2 || error "unable to start OST2"
788 start_mds || error "MDS start failed"
789 wait_osc_import_state mds ost2 FULL
791 stop_ost || error "Unable to stop OST1"
792 stop_ost2 || error "Unable to stop OST2"
793 stop_mds || error "Unable to stop MDS"
795 #writeconf to remove all ost2 traces for subsequent tests
796 writeconf_or_reformat
797 start_mgs || error "unable to start MGS"
799 run_test 21d "start mgs then ost and then mds"
802 MGSNID="$saved_mgsnid"
804 echo "stopping fs2mgs ..."
805 stop $fs2mgs -f || true
808 test_21e() { # LU-5863
809 if [[ -z "$fs3ost_DEV" || -z "$fs2ost_DEV" || -z "$fs2mds_DEV" ]]; then
810 is_blkdev $SINGLEMDS $(mdsdevname ${SINGLEMDS//mds/}) &&
811 skip_env "mixed loopback and real device not working" && return
814 local fs2mdsdev=$(mdsdevname 1_2)
815 local fs2ostdev=$(ostdevname 1_2)
816 local fs3ostdev=$(ostdevname 2_2)
818 local fs2mdsvdev=$(mdsvdevname 1_2)
819 local fs2ostvdev=$(ostvdevname 1_2)
820 local fs3ostvdev=$(ostvdevname 2_2)
822 # temporarily use fs3ost as fs2mgs
824 local fs2mgsdev=$fs3ostdev
825 local fs2mgsvdev=$fs3ostvdev
827 local fsname=test1234
829 add $fs2mgs $(mkfs_opts mgs $fs2mgsdev) --fsname=$fsname \
830 --reformat $fs2mgsdev $fs2mgsvdev || error "add fs2mgs failed"
831 start $fs2mgs $fs2mgsdev $MGS_MOUNT_OPTS && trap cleanup_21e EXIT INT ||
832 error "start fs2mgs failed"
834 local saved_mgsnid="$MGSNID"
835 MGSNID=$(do_facet $fs2mgs $LCTL list_nids | xargs | tr ' ' ,)
837 add fs2mds $(mkfs_opts mds1 $fs2mdsdev $fsname) \
838 --reformat $fs2mdsdev $fs2mdsvdev || error "add fs2mds failed"
839 add fs2ost $(mkfs_opts ost1 $fs2ostdev $fsname) \
840 --reformat $fs2ostdev $fs2ostvdev || error "add fs2ost failed"
842 start fs2ost $fs2ostdev $OST_MOUNT_OPTS || error "start fs2ost failed"
843 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS || error "start fs2mds failed"
845 mkdir -p $MOUNT2 || error "mkdir $MOUNT2 failed"
846 $MOUNT_CMD $MGSNID:/$fsname $MOUNT2 || error "mount $MOUNT2 failed"
847 DIR=$MOUNT2 MOUNT=$MOUNT2 check_mount || error "check $MOUNT2 failed"
851 run_test 21e "separate MGS and MDS"
854 start_mds || error "MDS start failed"
856 echo "Client mount with ost in logs, but none running"
857 start_ost || error "unable to start OST1"
858 # wait until mds connected to ost and open client connection
859 wait_osc_import_state mds ost FULL
860 stop_ost || error "unable to stop OST1"
861 mount_client $MOUNT || error "mount_client $MOUNT failed"
862 # check_mount will block trying to contact ost
863 mcreate $DIR/$tfile || error "mcreate $DIR/$tfile failed"
864 rm -f $DIR/$tfile || error "remove $DIR/$tfile failed"
865 umount_client $MOUNT -f
868 echo "Client mount with a running ost"
869 start_ost || error "unable to start OST1"
871 # if gss enabled, wait full time to let connection from
872 # mds to ost be established, due to the mismatch between
873 # initial connect timeout and gss context negotiation timeout.
874 # This perhaps could be remove after AT landed.
875 echo "sleep $((TIMEOUT + TIMEOUT + TIMEOUT))s"
876 sleep $((TIMEOUT + TIMEOUT + TIMEOUT))
878 mount_client $MOUNT || error "mount_client $MOUNT failed"
879 wait_osc_import_state mds ost FULL
880 wait_osc_import_state client ost FULL
881 check_mount || error "check_mount failed"
884 cleanup || error "cleanup failed with rc $?"
886 run_test 22 "start a client before osts (should return errs)"
888 test_23a() { # was test_23
891 stop $SINGLEMDS || error "failed to stop $SINGLEMDS"
892 # force down client so that recovering mds waits for reconnect
893 local running=$(grep -c $MOUNT /proc/mounts) || true
894 if [ $running -ne 0 ]; then
895 echo "Stopping client $MOUNT (opts: -f)"
899 # enter recovery on failed mds
900 local MDT_DEV=$(mdsdevname ${SINGLEMDS//mds/})
901 start $SINGLEMDS $MDT_DEV $MDS_MOUNT_OPTS || error "MDS start failed"
902 # try to start a new client
903 mount_client $MOUNT &
905 MOUNT_PID=$(ps -ef | grep "t lustre" | grep -v grep | awk '{print $2}')
906 MOUNT_LUSTRE_PID=$(ps -ef | grep mount.lustre |
907 grep -v grep | awk '{print $2}')
908 echo mount pid is ${MOUNT_PID}, mount.lustre pid is ${MOUNT_LUSTRE_PID}
910 ps --ppid $MOUNT_LUSTRE_PID
911 echo "waiting for mount to finish"
913 # "ctrl-c" sends SIGINT but it usually (in script) does not work on child process
914 # SIGTERM works but it does not spread to offspring processses
915 kill -s TERM $MOUNT_PID
916 kill -s TERM $MOUNT_LUSTRE_PID
917 # we can not wait $MOUNT_PID because it is not a child of this shell
923 while [ "$WAIT" -lt "$MAX_WAIT" ]; do
925 PID1=$(ps -ef | awk '{print $2}' | grep -w $MOUNT_PID)
926 PID2=$(ps -ef | awk '{print $2}' | grep -w $MOUNT_LUSTRE_PID)
929 [ -z "$PID1" -a -z "$PID2" ] && break
930 echo "waiting for mount to finish ... "
931 WAIT=$(( WAIT + sleep))
933 if [ "$WAIT" -eq "$MAX_WAIT" ]; then
934 error "MOUNT_PID $MOUNT_PID and "\
935 "MOUNT_LUSTRE_PID $MOUNT_LUSTRE_PID still not killed in $WAIT secs"
938 stop_mds || error "stopping MDSes failed"
939 stop_ost || error "stopping OSSes failed"
941 run_test 23a "interrupt client during recovery mount delay"
946 test_23b() { # was test_23
947 start_mds || error "MDS start failed"
948 start_ost || error "Unable to start OST1"
949 # Simulate -EINTR during mount OBD_FAIL_LDLM_CLOSE_THREAD
950 $LCTL set_param fail_loc=0x80000313
952 cleanup || error "cleanup failed with rc $?"
954 run_test 23b "Simulate -EINTR during mount"
957 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
959 if [ -z "$fs2ost_DEV" -o -z "$fs2mds_DEV" ]; then
960 is_blkdev $SINGLEMDS $MDSDEV &&
961 skip_env "mixed loopback and real device not working" && return
964 [ -n "$ost1_HOST" ] && fs2ost_HOST=$ost1_HOST
966 local fs2mdsdev=$(mdsdevname 1_2)
967 local fs2ostdev=$(ostdevname 1_2)
968 local fs2mdsvdev=$(mdsvdevname 1_2)
969 local fs2ostvdev=$(ostvdevname 1_2)
972 # LU-9733 test fsname started with numbers as well
973 local FSNAME2=969362ae
975 add fs2mds $(mkfs_opts mds1 ${fs2mdsdev} ) --nomgs --mgsnode=$MGSNID \
976 --fsname=${FSNAME2} --reformat $fs2mdsdev $fs2mdsvdev || exit 10
978 add fs2ost $(mkfs_opts ost1 ${fs2ostdev}) --fsname=${FSNAME2} \
979 --reformat $fs2ostdev $fs2ostvdev || exit 10
982 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS && trap cleanup_fs2 EXIT INT
983 start fs2ost $fs2ostdev $OST_MOUNT_OPTS
984 mkdir -p $MOUNT2 || error "mkdir $MOUNT2 failed"
985 $MOUNT_CMD $MGSNID:/${FSNAME2} $MOUNT2 || error "$MOUNT_CMD failed"
987 # LU-9733 test fsname started with numbers
988 cl_user=$(do_facet $SINGLEMDS lctl --device $FSNAME2-MDT0000 \
989 changelog_register -n) ||
990 error "register changelog failed"
992 do_facet $SINGLEMDS lctl --device $FSNAME2-MDT0000 \
993 changelog_deregister $cl_user ||
994 error "deregister changelog failed"
996 check_mount || error "check_mount failed"
997 # files written on 1 should not show up on 2
998 cp /etc/passwd $DIR/$tfile
1000 [ -e $MOUNT2/$tfile ] && error "File bleed"
1003 cp /etc/passwd $MOUNT2/$tfile ||
1004 error "cp /etc/passwd $MOUNT2/$tfile failed"
1005 rm $MOUNT2/$tfile || error "remove $MOUNT2/$tfile failed"
1006 # 2 is actually mounted
1007 grep $MOUNT2' ' /proc/mounts > /dev/null || error "$MOUNT2 not mounted"
1009 facet_failover fs2mds
1010 facet_failover fs2ost
1012 umount_client $MOUNT
1013 # the MDS must remain up until last MDT
1015 MDS=$(do_facet $SINGLEMDS "$LCTL get_param -n devices" |
1016 awk '($3 ~ "mdt" && $4 ~ "MDT") { print $4 }' | head -1)
1017 [ -z "$MDS" ] && error "No MDT"
1019 cleanup_nocli || error "cleanup_nocli failed with rc $?"
1021 run_test 24a "Multiple MDTs on a single node"
1024 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
1026 if [ -z "$fs2mds_DEV" ]; then
1027 local dev=${SINGLEMDS}_dev
1028 local MDSDEV=${!dev}
1029 is_blkdev $SINGLEMDS $MDSDEV &&
1030 skip_env "mixed loopback and real device not working" && return
1033 local fs2mdsdev=$(mdsdevname 1_2)
1034 local fs2mdsvdev=$(mdsvdevname 1_2)
1036 add fs2mds $(mkfs_opts mds1 ${fs2mdsdev} ) --mgs --fsname=${FSNAME}2 \
1037 --reformat $fs2mdsdev $fs2mdsvdev || exit 10
1039 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS &&
1040 error "start MDS should fail"
1042 cleanup || error "cleanup failed with rc $?"
1044 run_test 24b "Multiple MGSs on a single node (should return err)"
1048 check_mount || error "check_mount failed"
1049 local MODULES=$($LCTL modules | awk '{ print $2 }')
1050 rmmod $MODULES 2>/dev/null || true
1051 cleanup || error "cleanup failed with $?"
1053 run_test 25 "Verify modules are referenced"
1057 # we need modules before mount for sysctl, so make sure...
1058 do_facet $SINGLEMDS "lsmod | grep -q lustre || modprobe lustre"
1059 #define OBD_FAIL_MDS_FS_SETUP 0x135
1060 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0x80000135"
1061 start_mds && error "MDS started but should not have started"
1062 $LCTL get_param -n devices
1063 DEVS=$($LCTL get_param -n devices | egrep -v MG | wc -l)
1064 [ $DEVS -gt 0 ] && error "number of devices is $DEVS, should be zero"
1065 # start mds to drop writeconf setting
1066 start_mds || error "Unable to start MDS"
1067 stop_mds || error "Unable to stop MDS"
1068 unload_modules_conf || error "unload_modules_conf failed with $?"
1070 run_test 26 "MDT startup failure cleans LOV (should return errs)"
1073 start_ost || error "Unable to start OST1"
1074 start_mds || error "Unable to start MDS"
1075 echo "Requeue thread should have started: "
1076 ps -e | grep ll_cfg_requeue
1077 set_conf_param_and_check ost1 \
1078 "$LCTL get_param -n obdfilter.$FSNAME-OST0000.client_cache_seconds" \
1079 "$FSNAME-OST0000.ost.client_cache_seconds" ||
1080 error "set_conf_param_and_check ost1 failed"
1081 cleanup_nocli || error "cleanup_nocli failed with rc $?"
1083 run_test 27a "Reacquire MGS lock if OST started first"
1088 local device=$(do_facet $SINGLEMDS "$LCTL get_param -n devices" |
1089 awk '($3 ~ "mdt" && $4 ~ "MDT0000") { print $4 }')
1091 facet_failover $SINGLEMDS
1092 set_conf_param_and_check $SINGLEMDS \
1093 "$LCTL get_param -n mdt.$device.identity_acquire_expire" \
1094 "$device.mdt.identity_acquire_expire" ||
1095 error "set_conf_param_and_check $SINGLEMDS failed"
1096 set_conf_param_and_check client \
1097 "$LCTL get_param -n mdc.$device-mdc-*.max_rpcs_in_flight"\
1098 "$device.mdc.max_rpcs_in_flight" ||
1099 error "set_conf_param_and_check client failed"
1101 cleanup || error "cleanup failed with $?"
1103 run_test 27b "Reacquire MGS lock after failover"
1107 TEST="$LCTL get_param -n llite.$FSNAME-*.max_read_ahead_whole_mb"
1108 PARAM="$FSNAME.llite.max_read_ahead_whole_mb"
1110 FINAL=$(($ORIG + 1))
1111 set_conf_param_and_check client "$TEST" "$PARAM" $FINAL ||
1112 error "first set_conf_param_and_check client failed"
1113 FINAL=$(($FINAL + 1))
1114 set_conf_param_and_check client "$TEST" "$PARAM" $FINAL ||
1115 error "second set_conf_param_and_check client failed"
1116 umount_client $MOUNT || error "umount_client $MOUNT failed"
1117 mount_client $MOUNT || error "mount_client $MOUNT failed"
1119 if [ $RESULT -ne $FINAL ]; then
1120 error "New config not seen: wanted $FINAL got $RESULT"
1122 echo "New config success: got $RESULT"
1124 set_conf_param_and_check client "$TEST" "$PARAM" $ORIG ||
1125 error "third set_conf_param_and_check client failed"
1126 cleanup || error "cleanup failed with rc $?"
1128 run_test 28 "permanent parameter setting"
1130 test_28a() { # LU-4221
1131 [[ $(lustre_version_code ost1) -ge $(version_code 2.5.52) ]] ||
1132 { skip "Need OST version at least 2.5.52" && return 0; }
1133 [ "$(facet_fstype ost1)" = "zfs" ] &&
1134 skip "LU-4221: no such proc params for ZFS OSTs" && return
1141 local device="$FSNAME-OST0000"
1145 # In this test we will set three kinds of proc parameters with
1147 # 1. the ones moved from the OFD to the OSD, and only their
1148 # symlinks kept in obdfilter
1149 # 2. non-symlink ones in the OFD
1150 # 3. non-symlink ones in the OSD
1153 # prepare a symlink parameter in the OFD
1154 name="writethrough_cache_enable"
1155 param="$device.ost.$name"
1156 cmd="$LCTL get_param -n obdfilter.$device.$name"
1158 # conf_param the symlink parameter in the OFD
1159 old=$(do_facet ost1 $cmd)
1160 new=$(((old + 1) % 2))
1161 set_conf_param_and_check ost1 "$cmd" "$param" $new ||
1162 error "lctl conf_param $device.ost.$param=$new failed"
1164 # conf_param the target parameter in the OSD
1165 param="$device.osd.$name"
1166 cmd="$LCTL get_param -n osd-*.$device.$name"
1167 set_conf_param_and_check ost1 "$cmd" "$param" $old ||
1168 error "lctl conf_param $device.osd.$param=$old failed"
1171 # prepare a non-symlink parameter in the OFD
1172 name="client_cache_seconds"
1173 param="$device.ost.$name"
1174 cmd="$LCTL get_param -n obdfilter.$device.$name"
1176 # conf_param the parameter in the OFD
1177 old=$(do_facet ost1 $cmd)
1179 set_conf_param_and_check ost1 "$cmd" "$param" $new ||
1180 error "lctl conf_param $device.ost.$param=$new failed"
1181 set_conf_param_and_check ost1 "$cmd" "$param" $old ||
1182 error "lctl conf_param $device.ost.$param=$old failed"
1185 # prepare a non-symlink parameter in the OSD
1187 param="$device.osd.$name"
1188 cmd="$LCTL get_param -n osd-*.$device.$name"
1190 # conf_param the parameter in the OSD
1191 old=$(do_facet ost1 $cmd)
1192 new=$(((old + 1) % 2))
1193 set_conf_param_and_check ost1 "$cmd" "$param" $new ||
1194 error "lctl conf_param $device.osd.$param=$new failed"
1195 set_conf_param_and_check ost1 "$cmd" "$param" $old ||
1196 error "lctl conf_param $device.osd.$param=$old failed"
1198 cleanup || error "cleanup failed with $?"
1200 run_test 28a "set symlink parameters permanently with conf_param"
1203 [ "$OSTCOUNT" -lt "2" ] && skip_env "needs >= 2 OSTs" && return
1204 setup > /dev/null 2>&1
1205 start_ost2 || error "Unable to start OST2"
1208 local PARAM="$FSNAME-OST0001.osc.active"
1209 local PROC_ACT="osc.$FSNAME-OST0001-osc-[^M]*.active"
1210 local PROC_UUID="osc.$FSNAME-OST0001-osc-[^M]*.ost_server_uuid"
1212 ACTV=$($LCTL get_param -n $PROC_ACT)
1214 set_conf_param_and_check client \
1215 "$LCTL get_param -n $PROC_ACT" "$PARAM" $DEAC ||
1216 error "set_conf_param_and_check client failed"
1217 # also check ost_server_uuid status
1218 RESULT=$($LCTL get_param -n $PROC_UUID | grep DEACTIV)
1219 if [ -z "$RESULT" ]; then
1220 error "Client not deactivated: $($LCTL get_param \
1223 echo "Live client success: got $RESULT"
1227 wait_osp_active ost ${FSNAME}-OST0001 1 0
1229 # test new client starts deactivated
1230 umount_client $MOUNT || error "umount_client $MOUNT failed"
1231 mount_client $MOUNT || error "mount_client $MOUNT failed"
1232 RESULT=$($LCTL get_param -n $PROC_UUID | grep DEACTIV | grep NEW)
1233 if [ -z "$RESULT" ]; then
1234 error "New client start active: $(lctl get_param -n $PROC_UUID)"
1236 echo "New client success: got $RESULT"
1239 # make sure it reactivates
1240 set_conf_param_and_check client \
1241 "$LCTL get_param -n $PROC_ACT" "$PARAM" $ACTV ||
1242 error "lctl get_param $PROC_ACT $PARAM $ACTV failed"
1244 umount_client $MOUNT
1245 stop_ost2 || error "Unable to stop OST2"
1246 cleanup_nocli || error "cleanup_nocli failed with $?"
1247 #writeconf to remove all ost2 traces for subsequent tests
1248 writeconf_or_reformat
1250 run_test 29 "permanently remove an OST"
1255 echo Big config llog
1256 TEST="$LCTL get_param -n llite.$FSNAME-*.max_read_ahead_whole_mb"
1258 LIST=(1 2 3 4 5 4 3 2 1 2 3 4 5 4 3 2 1 2 3 4 5)
1259 for i in ${LIST[@]}; do
1260 set_conf_param_and_check client "$TEST" \
1261 "$FSNAME.llite.max_read_ahead_whole_mb" $i ||
1262 error "Set $FSNAME.llite.max_read_ahead_whole_mb failed"
1264 # make sure client restart still works
1265 umount_client $MOUNT
1266 mount_client $MOUNT || error "mount_client $MOUNT failed"
1267 [ "$($TEST)" -ne "$i" ] &&
1268 error "Param didn't stick across restart $($TEST) != $i"
1271 echo Erase parameter setting
1272 do_facet mgs "$LCTL conf_param \
1273 -d $FSNAME.llite.max_read_ahead_whole_mb" ||
1274 error "Erase param $FSNAME.llite.max_read_ahead_whole_mb failed"
1275 umount_client $MOUNT
1276 mount_client $MOUNT || error "mount_client $MOUNT failed"
1278 echo "deleted (default) value=$FINAL, orig=$ORIG"
1279 # assumes this parameter started at the default value
1280 [ "$FINAL" -eq "$ORIG" ] || fail "Deleted value=$FINAL, orig=$ORIG"
1282 cleanup || error "cleanup failed with rc $?"
1284 run_test 30a "Big config llog and conf_param deletion"
1289 local orignids=$($LCTL get_param -n \
1290 osc.$FSNAME-OST0000-osc-[^M]*.import | grep failover_nids)
1292 local orignidcount=$(echo "$orignids" | wc -w)
1294 # Make a fake nid. Use the OST nid, and add 20 to the least significant
1295 # numerical part of it. Hopefully that's not already a failover address
1297 local OSTNID=$(do_facet ost1 "$LCTL get_param nis" | tail -1 | \
1299 local ORIGVAL=$(echo $OSTNID | egrep -oi "[0-9]*@")
1300 local NEWVAL=$((($(echo $ORIGVAL | egrep -oi "[0-9]*") + 20) % 256))
1301 local NEW=$(echo $OSTNID | sed "s/$ORIGVAL/$NEWVAL@/")
1302 echo "Using fake nid $NEW"
1304 local TEST="$LCTL get_param -n osc.$FSNAME-OST0000-osc-[^M]*.import |
1305 grep failover_nids | sed -n 's/.*\($NEW\).*/\1/p'"
1306 set_conf_param_and_check client "$TEST" \
1307 "$FSNAME-OST0000.failover.node" $NEW ||
1308 error "didn't add failover nid $NEW"
1309 local NIDS=$($LCTL get_param -n osc.$FSNAME-OST0000-osc-[^M]*.import |
1312 local NIDCOUNT=$(echo "$NIDS" | wc -w)
1313 echo "should have $((orignidcount + 1)) entries \
1314 in failover nids string, have $NIDCOUNT"
1315 [ $NIDCOUNT -eq $((orignidcount + 1)) ] ||
1316 error "Failover nid not added"
1318 do_facet mgs "$LCTL conf_param -d $FSNAME-OST0000.failover.node" ||
1319 error "conf_param delete failed"
1320 umount_client $MOUNT
1321 mount_client $MOUNT || error "mount_client $MOUNT failed"
1323 NIDS=$($LCTL get_param -n osc.$FSNAME-OST0000-osc-[^M]*.import |
1326 NIDCOUNT=$(echo "$NIDS" | wc -w)
1327 echo "only $orignidcount final entries should remain \
1328 in failover nids string, have $NIDCOUNT"
1329 [ $NIDCOUNT -eq $orignidcount ] || error "Failover nids not removed"
1331 cleanup || error "cleanup failed with rc $?"
1333 run_test 30b "Remove failover nids"
1335 test_31() { # bug 10734
1336 # ipaddr must not exist
1337 $MOUNT_CMD 4.3.2.1@tcp:/lustre $MOUNT || true
1338 cleanup || error "cleanup failed with rc $?"
1340 run_test 31 "Connect to non-existent node (shouldn't crash)"
1344 T32_BLIMIT=40960 # Kbytes
1348 # This is not really a test but a tool to create new disk
1349 # image tarballs for the upgrade tests.
1351 # Disk image tarballs should be created on single-node
1352 # clusters by running this test with default configurations
1353 # plus a few mandatory environment settings that are verified
1354 # at the beginning of the test.
1356 test_32newtarball() {
1360 local tmp=$TMP/t32_image_create
1361 local server_version=$(lustre_version_code $SINGLEMDS)
1366 if [ $FSNAME != t32fs -o \( -z "$MDSDEV" -a -z "$MDSDEV1" \) -o \
1367 $OSTCOUNT -ne 1 -o -z "$OSTDEV1" ]; then
1368 error "Needs FSNAME=t32fs MDSCOUNT=2 " \
1369 "MDSDEV1=<nonexistent_file>" \
1370 "MDSDEV2=<nonexistent_file>" \
1371 "(or MDSDEV, in the case of b1_8)" \
1372 "OSTCOUNT=1 OSTDEV1=<nonexistent_file>"
1376 echo "Found stale $tmp"
1380 mkdir $tmp/src || return 1
1381 tar cf - -C $src . | tar xf - -C $tmp/src
1382 dd if=/dev/zero of=$tmp/src/t32_qf_old bs=1M \
1383 count=$(($T32_BLIMIT / 1024 / 4))
1384 chown $T32_QID.$T32_QID $tmp/src/t32_qf_old
1386 # format ost with comma-separated NIDs to verify LU-4460
1387 local failnid="$(h2nettype 1.2.3.4),$(h2nettype 4.3.2.1)"
1388 MGSNID="$MGSNID,$MGSNID" OSTOPT="--failnode=$failnid" formatall
1392 [[ $server_version -ge $(version_code 2.3.50) ]] ||
1393 $LFS quotacheck -ug /mnt/$FSNAME
1394 $LFS setquota -u $T32_QID -b 0 -B $T32_BLIMIT -i 0 -I $T32_ILIMIT \
1397 tar cf - -C $tmp/src . | tar xf - -C /mnt/$FSNAME
1399 if [[ $MDSCOUNT -ge 2 ]]; then
1400 remote_dir=/mnt/$FSNAME/remote_dir
1401 $LFS mkdir -i 1 $remote_dir
1402 tar cf - -C $tmp/src . | tar xf - -C $remote_dir
1404 if [[ $server_version -ge $(version_code 2.7.0) ]]; then
1405 striped_dir=/mnt/$FSNAME/striped_dir_old
1406 $LFS mkdir -i 1 -c 2 $striped_dir
1407 tar cf - -C $tmp/src . | tar xf - -C $striped_dir
1413 mkdir $tmp/img || return 1
1417 pushd_dir=/mnt/$FSNAME
1418 if [[ $MDSCOUNT -ge 2 ]]; then
1419 pushd_dir=$remote_dir
1420 if [[ $server_version -ge $(version_code 2.7.0) ]]; then
1422 ls -Rni --time-style=+%s >$tmp/img/list2
1428 ls -Rni --time-style=+%s >$tmp/img/list
1429 find ! -name .lustre -type f -exec sha1sum {} \; |
1430 sort -k 2 >$tmp/img/sha1sums
1432 $LCTL get_param -n version | head -n 1 |
1433 sed -e 's/^lustre: *//' >$tmp/img/commit
1435 [[ $server_version -ge $(version_code 2.3.50) ]] ||
1436 $LFS quotaon -ug /mnt/$FSNAME
1437 $LFS quota -u $T32_QID -v /mnt/$FSNAME
1438 $LFS quota -v -u $T32_QID /mnt/$FSNAME |
1439 awk 'BEGIN { num='1' } { if ($1 == "'/mnt/$FSNAME'") \
1440 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1441 | tr -d "*" > $tmp/img/bspace
1442 $LFS quota -v -u $T32_QID /mnt/$FSNAME |
1443 awk 'BEGIN { num='5' } { if ($1 == "'/mnt/$FSNAME'") \
1444 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1445 | tr -d "*" > $tmp/img/ispace
1446 echo $T32_BLIMIT > $tmp/img/blimit
1447 echo $T32_ILIMIT > $tmp/img/ilimit
1452 find -type f -exec sha1sum {} \; | sort -k 2 >$tmp/sha1sums.src
1455 if ! diff -u $tmp/sha1sums.src $tmp/img/sha1sums; then
1456 echo "Data verification failed"
1459 uname -r >$tmp/img/kernel
1460 uname -m >$tmp/img/arch
1462 mv ${MDSDEV1:-$MDSDEV} $tmp/img
1463 for num in $(seq 2 $MDSCOUNT); do
1464 local devname=$(mdsdevname $num)
1466 [[ $(facet_fstype $facet) != zfs ]] ||
1467 devname=$(mdsvdevname $num)
1468 mv $devname $tmp/img
1470 mv $OSTDEV1 $tmp/img
1472 version=$(sed -e 's/\(^[0-9]\+\.[0-9]\+\)\(.*$\)/\1/' $tmp/img/commit |
1473 sed -e 's/\./_/g') # E.g., "1.8.7" -> "1_8"
1476 tar cjvf $dst/disk$version-$(facet_fstype $SINGLEMDS).tar.bz2 -S *
1481 #run_test 32newtarball "Create a new test_32 disk image tarball for this version"
1484 # The list of applicable tarballs is returned via the caller's
1485 # variable "tarballs".
1488 local node=$(facet_active_host $SINGLEMDS)
1489 local r="do_node $node"
1491 if [ "$CLIENTONLY" ]; then
1492 skip "Client-only testing"
1496 if ! $r which $TUNEFS; then
1497 skip_env "tunefs.lustre required on $node"
1501 local IMGTYPE=$(facet_fstype $SINGLEMDS)
1503 tarballs=$($r find $RLUSTRE/tests -maxdepth 1 \
1504 -name \'disk*-$IMGTYPE.tar.bz2\')
1506 if [ -z "$tarballs" ]; then
1507 skip "No applicable tarballs found"
1512 t32_test_cleanup() {
1514 local facet=$SINGLEMDS
1515 local fstype=$(facet_fstype $facet)
1518 if $shall_cleanup_lustre; then
1519 umount $tmp/mnt/lustre || rc=$?
1521 if $shall_cleanup_mdt; then
1522 $r $UMOUNT $tmp/mnt/mdt || rc=$?
1524 if $shall_cleanup_mdt1; then
1525 $r $UMOUNT $tmp/mnt/mdt1 || rc=$?
1527 if $shall_cleanup_ost; then
1528 $r $UMOUNT $tmp/mnt/ost || rc=$?
1533 if [[ $fstype == zfs ]]; then
1535 local poolname_list="t32fs-mdt1 t32fs-ost1"
1537 ! $mdt2_is_available || poolname_list+=" t32fs-mdt2"
1539 for poolname in $poolname_list; do
1540 destroy_zpool $facet $poolname
1543 combined_mgs_mds || start_mgs || rc=$?
1547 t32_bits_per_long() {
1549 # Yes, this is not meant to be perfect.
1559 t32_reload_modules() {
1561 local all_removed=false
1563 local fstype=$(facet_fstype $SINGLEMDS)
1565 [ $fstype == "zfs" ] && do_rpc_nodes $node "service zed stop"
1567 while ((i < 20)); do
1568 echo "Unloading modules on $node: Attempt $i"
1569 do_rpc_nodes $node $LUSTRE_RMMOD $fstype &&
1571 do_rpc_nodes $node check_mem_leak || return 1
1572 if $all_removed; then
1573 do_rpc_nodes $node load_modules
1576 if [ $fstype == "zfs" ]; then
1577 do_rpc_nodes $node "$ZPOOL status -v"
1582 echo "Unloading modules on $node: Given up"
1586 t32_wait_til_devices_gone() {
1592 echo wait for devices to go
1593 while ((i < 20)); do
1594 devices=$(do_rpc_nodes $node $LCTL device_list | wc -l)
1595 loops=$(do_rpc_nodes $node losetup -a | grep -c t32)
1596 ((devices == 0 && loops == 0)) && return 0
1600 echo "waiting for dev on $node: dev $devices loop $loops given up"
1601 do_rpc_nodes $node "losetup -a"
1602 do_rpc_nodes $node "$LCTL device_list"
1606 t32_verify_quota() {
1610 local fstype=$(facet_fstype $SINGLEMDS)
1614 # LU-2435: if the underlying zfs doesn't support userobj_accounting,
1615 # lustre will estimate the object count usage. This fails quota
1616 # verification in 32b. The object quota usage should be accurate after
1617 # zfs-0.7.0 is released.
1618 [ $fstype == "zfs" ] && {
1619 local zfs_version=$(do_node $node cat /sys/module/zfs/version)
1621 [ $(version_code $zfs_version) -lt $(version_code 0.7.0) ] && {
1622 echo "Skip quota verify for zfs: $zfs_version"
1627 $LFS quota -u $T32_QID -v $mnt
1629 qval=$($LFS quota -v -u $T32_QID $mnt |
1630 awk 'BEGIN { num='1' } { if ($1 == "'$mnt'") \
1631 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1633 [ $qval -eq $img_bspace ] || {
1634 echo "bspace, act:$qval, exp:$img_bspace"
1638 qval=$($LFS quota -v -u $T32_QID $mnt |
1639 awk 'BEGIN { num='5' } { if ($1 == "'$mnt'") \
1640 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1642 [ $qval -eq $img_ispace ] || {
1643 echo "ispace, act:$qval, exp:$img_ispace"
1647 qval=$($LFS quota -v -u $T32_QID $mnt |
1648 awk 'BEGIN { num='3' } { if ($1 == "'$mnt'") \
1649 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1651 [ $qval -eq $img_blimit ] || {
1652 echo "blimit, act:$qval, exp:$img_blimit"
1656 qval=$($LFS quota -v -u $T32_QID $mnt |
1657 awk 'BEGIN { num='7' } { if ($1 == "'$mnt'") \
1658 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1660 [ $qval -eq $img_ilimit ] || {
1661 echo "ilimit, act:$qval, exp:$img_ilimit"
1665 do_node $node $LCTL conf_param $fsname.quota.mdt=ug
1666 cmd="$LCTL get_param -n osd-$fstype.$fsname-MDT0000"
1667 cmd=$cmd.quota_slave.enabled
1668 wait_update $node "$cmd" "ug" || {
1669 echo "Enable mdt quota failed"
1673 do_node $node $LCTL conf_param $fsname.quota.ost=ug
1674 cmd="$LCTL get_param -n osd-$fstype.$fsname-OST0000"
1675 cmd=$cmd.quota_slave.enabled
1676 wait_update $node "$cmd" "ug" || {
1677 echo "Enable ost quota failed"
1682 runas -u $T32_QID -g $T32_QID dd if=/dev/zero of=$mnt/t32_qf_new \
1683 bs=1M count=$((img_blimit / 1024)) oflag=sync && {
1684 echo "Write succeed, but expect -EDQUOT"
1687 rm -f $mnt/t32_qf_new
1689 runas -u $T32_QID -g $T32_QID createmany -m $mnt/t32_qf_ \
1691 echo "Create succeed, but expect -EDQUOT"
1694 unlinkmany $mnt/t32_qf_ $img_ilimit
1702 local dne_upgrade=${dne_upgrade:-"no"}
1703 local ff_convert=${ff_convert:-"no"}
1704 local shall_cleanup_mdt=false
1705 local shall_cleanup_mdt1=false
1706 local shall_cleanup_ost=false
1707 local shall_cleanup_lustre=false
1708 local mdt2_is_available=false
1709 local node=$(facet_active_host $SINGLEMDS)
1710 local r="do_node $node"
1726 local fstype=$(facet_fstype $SINGLEMDS)
1727 local mdt_dev=$tmp/mdt
1728 local mdt2_dev=$tmp/mdt2
1729 local ost_dev=$tmp/ost
1734 combined_mgs_mds || stop_mgs || error "Unable to stop MGS"
1735 trap 'trap - RETURN; t32_test_cleanup' RETURN
1738 nid=$($r $LCTL list_nids | head -1)
1740 mkdir -p $tmp/mnt/lustre || error "mkdir $tmp/mnt/lustre failed"
1741 $r mkdir -p $tmp/mnt/{mdt,mdt1,ost}
1742 $r tar xjvf $tarball -S -C $tmp || {
1743 error_noexit "Unpacking the disk image tarball"
1746 img_commit=$($r cat $tmp/commit)
1747 img_kernel=$($r cat $tmp/kernel)
1748 img_arch=$($r cat $tmp/arch)
1749 img_bspace=$($r cat $tmp/bspace)
1750 img_ispace=$($r cat $tmp/ispace)
1752 # older images did not have "blimit" and "ilimit" files
1753 # use old values for T32_BLIMIT and T32_ILIMIT
1754 $r test -f $tmp/blimit && img_blimit=$($r cat $tmp/blimit) ||
1756 $r test -f $tmp/ilimit && img_ilimit=$($r cat $tmp/ilimit) ||
1759 echo "Upgrading from $(basename $tarball), created with:"
1760 echo " Commit: $img_commit"
1761 echo " Kernel: $img_kernel"
1762 echo " Arch: $img_arch"
1763 echo "OST version: $(lustre_build_version ost1)"
1765 # The conversion can be made only when both of the following
1766 # conditions are satisfied:
1767 # - ost device img version < 2.3.64
1768 # - ost server version >= 2.5
1769 [ $(version_code $img_commit) -ge $(version_code 2.3.64) -o \
1770 $(lustre_version_code ost1) -lt $(version_code 2.5.0) ] &&
1773 ! $r test -f $mdt2_dev || mdt2_is_available=true
1775 if [[ $fstype == zfs ]]; then
1778 local poolname_list="t32fs-mdt1 t32fs-ost1"
1780 ! $mdt2_is_available || poolname_list+=" t32fs-mdt2"
1782 for poolname in $poolname_list; do
1784 $ZPOOL list -H $poolname >/dev/null 2>&1 ||
1785 $ZPOOL import -f -d $tmp $poolname"
1788 # upgrade zpool to latest supported features, including
1789 # dnode quota accounting in 0.7.0
1790 $r "$ZPOOL upgrade -a"
1792 mdt_dev=t32fs-mdt1/mdt1
1793 ost_dev=t32fs-ost1/ost1
1794 ! $mdt2_is_available || mdt2_dev=t32fs-mdt2/mdt2
1795 wait_update_facet $SINGLEMDS "$ZPOOL list |
1796 awk '/^t32fs-mdt1/ { print \\\$1 }'" "t32fs-mdt1" || {
1797 error_noexit "import zfs pool failed"
1802 $r $LCTL set_param debug="$PTLDEBUG"
1804 $r $TUNEFS --dryrun $mdt_dev || {
1806 error_noexit "tunefs.lustre before mounting the MDT"
1810 if $mdt2_is_available; then
1811 $r $TUNEFS --dryrun $mdt2_dev || {
1813 error_noexit "tunefs.lustre before mounting the MDT"
1818 if [ "$writeconf" ]; then
1820 if [ $fstype == "ldiskfs" ]; then
1822 $r $TUNEFS --quota $mdt_dev || {
1824 error_noexit "Enable mdt quota feature"
1827 if $mdt2_is_available; then
1828 $r $TUNEFS --quota $mdt2_dev || {
1830 error_noexit "Enable mdt quota feature"
1836 if [ -n "$($LCTL list_nids | grep -v '\(tcp\|lo\)[[:digit:]]*$')" ]; then
1837 [[ $(lustre_version_code mgs) -ge $(version_code 2.3.59) ]] ||
1838 { skip "LU-2200: Cannot run over Inifiniband w/o lctl replace_nids "
1839 "(Need MGS version at least 2.3.59)"; return 0; }
1841 local osthost=$(facet_active_host ost1)
1842 local ostnid=$(do_node $osthost $LCTL list_nids | head -1)
1845 if [ $fstype == "ldiskfs" ]; then
1848 $r $MOUNT_CMD -o $mopts $mdt_dev $tmp/mnt/mdt
1849 $r $LCTL replace_nids $fsname-OST0000 $ostnid
1850 $r $LCTL replace_nids $fsname-MDT0000 $nid
1851 $r $UMOUNT $tmp/mnt/mdt
1854 mopts=exclude=$fsname-OST0000
1855 if [ $fstype == "ldiskfs" ]; then
1860 t32_wait_til_devices_gone $node
1862 $r $MOUNT_CMD -o $mopts $mdt_dev $tmp/mnt/mdt || {
1864 error_noexit "Mounting the MDT"
1867 shall_cleanup_mdt=true
1869 if $mdt2_is_available; then
1870 mopts=mgsnode=$nid,$mopts
1871 $r $MOUNT_CMD -o $mopts $mdt2_dev $tmp/mnt/mdt1 || {
1873 error_noexit "Mounting the MDT"
1877 echo "mount new MDT....$mdt2_dev"
1878 $r $LCTL set_param -n mdt.${fsname}*.enable_remote_dir=1 ||
1879 error_noexit "enable remote dir create failed"
1881 shall_cleanup_mdt1=true
1882 elif [ "$dne_upgrade" != "no" ]; then
1883 local fs2mdsdev=$(mdsdevname 1_2)
1884 local fs2mdsvdev=$(mdsvdevname 1_2)
1886 echo "mkfs new MDT on ${fs2mdsdev}...."
1887 if [ $(facet_fstype mds1) == ldiskfs ]; then
1888 mkfsoptions="--mkfsoptions=\\\"-J size=8\\\""
1891 add $SINGLEMDS $(mkfs_opts mds2 $fs2mdsdev $fsname) --reformat \
1892 $mkfsoptions $fs2mdsdev $fs2mdsvdev > /dev/null || {
1893 error_noexit "Mkfs new MDT failed"
1897 [[ $(facet_fstype mds1) != zfs ]] || import_zpool mds1
1899 $r $TUNEFS --dryrun $fs2mdsdev || {
1900 error_noexit "tunefs.lustre before mounting the MDT"
1904 echo "mount new MDT....$fs2mdsdev"
1905 $r $MOUNT_CMD -o $mopts $fs2mdsdev $tmp/mnt/mdt1 || {
1906 error_noexit "mount mdt1 failed"
1910 $r $LCTL set_param -n mdt.${fsname}*.enable_remote_dir=1 ||
1911 error_noexit "enable remote dir create failed"
1913 shall_cleanup_mdt1=true
1916 uuid=$($r $LCTL get_param -n mdt.$fsname-MDT0000.uuid) || {
1917 error_noexit "Getting MDT UUID"
1920 if [ "$uuid" != $fsname-MDT0000_UUID ]; then
1921 error_noexit "Unexpected MDT UUID: \"$uuid\""
1925 $r $TUNEFS --dryrun $ost_dev || {
1926 error_noexit "tunefs.lustre before mounting the OST"
1929 if [ "$writeconf" ]; then
1930 mopts=mgsnode=$nid,$writeconf
1931 if [ $fstype == "ldiskfs" ]; then
1933 $r $TUNEFS --quota $ost_dev || {
1935 error_noexit "Enable ost quota feature"
1941 if [ $fstype == "ldiskfs" ]; then
1946 $r $MOUNT_CMD -onomgs -o$mopts $ost_dev $tmp/mnt/ost || {
1947 error_noexit "Mounting the OST"
1950 shall_cleanup_ost=true
1952 uuid=$($r $LCTL get_param -n obdfilter.$fsname-OST0000.uuid) || {
1953 error_noexit "Getting OST UUID"
1956 if [ "$uuid" != $fsname-OST0000_UUID ]; then
1957 error_noexit "Unexpected OST UUID: \"$uuid\""
1961 $r $LCTL conf_param $fsname-OST0000.osc.max_dirty_mb=15 || {
1962 error_noexit "Setting \"max_dirty_mb\""
1965 $r $LCTL conf_param $fsname-OST0000.failover.node=$nid || {
1966 error_noexit "Setting OST \"failover.node\""
1969 $r $LCTL conf_param $fsname-MDT0000.mdc.max_rpcs_in_flight=9 || {
1970 error_noexit "Setting \"max_rpcs_in_flight\""
1973 $r $LCTL conf_param $fsname-MDT0000.failover.node=$nid || {
1974 error_noexit "Setting MDT \"failover.node\""
1977 $r $LCTL pool_new $fsname.interop || {
1978 error_noexit "Setting \"interop\""
1981 $r $LCTL conf_param $fsname-MDT0000.lov.stripesize=4M || {
1982 error_noexit "Setting \"lov.stripesize\""
1985 $r $LCTL conf_param $fsname-MDT0000.mdd.atime_diff=70 || {
1986 error_noexit "Setting \"mdd.atime_diff\""
1990 if [ "$ff_convert" != "no" -a $(facet_fstype ost1) == "ldiskfs" ]; then
1991 $r $LCTL lfsck_start -M $fsname-OST0000 || {
1992 error_noexit "Start OI scrub on OST0"
1996 # The oi_scrub should be on ost1, but for test_32(),
1997 # all on the SINGLEMDS.
1998 wait_update_facet $SINGLEMDS "$LCTL get_param -n \
1999 osd-ldiskfs.$fsname-OST0000.oi_scrub |
2000 awk '/^status/ { print \\\$2 }'" "completed" 30 || {
2001 error_noexit "Failed to get the expected 'completed'"
2005 local UPDATED=$($r $LCTL get_param -n \
2006 osd-ldiskfs.$fsname-OST0000.oi_scrub |
2007 awk '/^updated/ { print $2 }')
2008 [ $UPDATED -ge 1 ] || {
2009 error_noexit "Only $UPDATED objects have been converted"
2014 if [ "$dne_upgrade" != "no" ]; then
2015 $r $LCTL conf_param \
2016 $fsname-MDT0001.mdc.max_rpcs_in_flight=9 || {
2017 error_noexit "Setting MDT1 \"max_rpcs_in_flight\""
2020 $r $LCTL conf_param $fsname-MDT0001.failover.node=$nid || {
2021 error_noexit "Setting MDT1 \"failover.node\""
2024 $r $LCTL conf_param $fsname-MDT0001.lov.stripesize=4M || {
2025 error_noexit "Setting MDT1 \"lov.stripesize\""
2030 if [ "$writeconf" ]; then
2031 $MOUNT_CMD $nid:/$fsname $tmp/mnt/lustre || {
2032 error_noexit "Mounting the client"
2036 shall_cleanup_lustre=true
2037 $r $LCTL set_param debug="$PTLDEBUG"
2039 t32_verify_quota $node $fsname $tmp/mnt/lustre || {
2040 error_noexit "verify quota failed"
2044 if $r test -f $tmp/list; then
2046 # There is not a Test Framework API to copy files to or
2047 # from a remote node.
2049 # LU-2393 - do both sorts on same node to ensure locale
2051 local list_file=$tmp/list
2053 if $mdt2_is_available; then
2054 if [[ -d $tmp/mnt/lustre/striped_dir_old ]] &&
2055 $r test -f $tmp/list2; then
2056 list_file=$tmp/list2
2057 pushd $tmp/mnt/lustre/striped_dir_old
2059 pushd $tmp/mnt/lustre/remote_dir
2062 pushd $tmp/mnt/lustre
2064 $r cat $list_file | sort -k 6 >$tmp/list.orig
2065 ls -Rni --time-style=+%s | sort -k 6 |
2066 sed 's/\. / /' >$tmp/list || {
2072 # 32-bit and 64-bit clients use different algorithms to
2073 # convert FIDs into inode numbers. Hence, remove the
2074 # inode numbers from the lists, if the original list was
2075 # created on an architecture with different number of
2078 if [ $(t32_bits_per_long $(uname -m)) != \
2079 $(t32_bits_per_long $img_arch) ]; then
2080 echo "Different number of bits per \"long\"" \
2081 "from the disk image"
2082 for list in list.orig list; do
2083 sed -i -e 's/^[0-9]\+[ \t]\+//' \
2087 if ! diff -ub $tmp/list.orig $tmp/list; then
2088 error_noexit "list verification failed"
2092 echo "list verification skipped"
2095 if [ "$dne_upgrade" != "no" ]; then
2096 $LFS mkdir -i 1 -c2 $tmp/mnt/lustre/striped_dir || {
2097 error_noexit "set striped dir failed"
2101 $LFS setdirstripe -D -c2 $tmp/mnt/lustre/striped_dir
2103 pushd $tmp/mnt/lustre
2104 tar -cf - . --exclude=./striped_dir \
2105 --exclude=./striped_dir_old \
2106 --exclude=./remote_dir |
2107 tar -xvf - -C striped_dir 1>/dev/null || {
2108 error_noexit "cp to striped dir failed"
2114 # If it is upgrade from DNE (2.5), then rename the remote dir,
2115 # which is created in 2.5 to striped dir.
2116 if $mdt2_is_available && [[ "$dne_upgrade" != "no" ]]; then
2117 stripe_index=$($LFS getdirstripe -i \
2118 $tmp/mnt/lustre/remote_dir)
2120 [[ $stripe_index -eq 1 ]] || {
2121 error_noexit "get index \"$stripe_index\"" \
2122 "from remote dir failed"
2125 mv $tmp/mnt/lustre/remote_dir \
2126 $tmp/mnt/lustre/striped_dir/ || {
2127 error_noexit "mv remote dir failed"
2132 # If it is upgraded from DNE (2.7), then move the striped dir
2133 # which was created in 2.7 to the new striped dir.
2134 if $mdt2_is_available && [[ "$dne_upgrade" != "no" ]] &&
2135 [[ -d $tmp/mnt/lustre/striped_dir_old ]]; then
2136 stripe_count=$($LFS getdirstripe -c \
2137 $tmp/mnt/lustre/striped_dir_old)
2138 [[ $stripe_count -eq 2 ]] || {
2139 error_noexit "get count $stripe_count" \
2140 "from striped dir failed"
2143 mv $tmp/mnt/lustre/striped_dir_old \
2144 $tmp/mnt/lustre/striped_dir/ || {
2145 error_noexit "mv striped dir failed"
2151 $r $LCTL set_param -n osd*.*.force_sync=1
2152 dd if=/dev/zero of=$tmp/mnt/lustre/tmp_file bs=10k count=10 || {
2153 error_noexit "dd failed"
2156 rm -rf $tmp/mnt/lustre/tmp_file || {
2157 error_noexit "rm failed"
2161 if $r test -f $tmp/sha1sums; then
2162 # LU-2393 - do both sorts on same node to ensure locale
2164 $r cat $tmp/sha1sums | sort -k 2 >$tmp/sha1sums.orig
2165 if [ "$dne_upgrade" != "no" ]; then
2166 pushd $tmp/mnt/lustre/striped_dir
2168 pushd $tmp/mnt/lustre
2171 find ! -path "*remote_dir*" ! -path "*striped_dir*" \
2172 ! -name .lustre -type f -exec sha1sum {} \; |
2173 sort -k 2 >$tmp/sha1sums || {
2175 error_noexit "sha1sum"
2179 if ! diff -ub $tmp/sha1sums.orig $tmp/sha1sums; then
2180 error_noexit "sha1sum verification failed"
2184 # if upgrade from DNE(2.5), then check remote directory
2185 # if upgrade from DNE(2.7), then check striped directory
2186 if $mdt2_is_available &&
2187 [[ "$dne_upgrade" != "no" ]]; then
2188 local new_dir="$tmp/mnt/lustre/striped_dir"
2189 local striped_dir_old="$new_dir/striped_dir_old"
2191 local dir_list="$new_dir/remote_dir"
2192 [[ ! -d $triped_dir_old ]] ||
2193 dir_list+=" $striped_dir_old"
2195 for dir in $dir_list; do
2197 find ! -name .lustre -type f \
2198 -exec sha1sum {} \; |
2199 sort -k 2 >$tmp/sha1sums || {
2201 error_noexit "sha1sum"
2205 if ! diff -ub $tmp/sha1sums.orig \
2207 error_noexit "sha1sum $dir" \
2214 echo "sha1sum verification skipped"
2217 if [ "$dne_upgrade" != "no" ]; then
2218 rm -rf $tmp/mnt/lustre/striped_dir || {
2219 error_noexit "remove remote dir failed"
2224 # migrate files/dirs to remote MDT, then move them back
2225 if [ $(lustre_version_code mds1) -ge $(version_code 2.7.50) -a \
2226 $dne_upgrade != "no" ]; then
2227 $r $LCTL set_param -n \
2228 mdt.${fsname}*.enable_remote_dir=1 2>/dev/null
2230 echo "test migration"
2231 pushd $tmp/mnt/lustre
2232 for dir in $(find ! -name .lustre ! -name . -type d); do
2233 mdt_index=$($LFS getdirstripe -i $dir)
2234 stripe_cnt=$($LFS getdirstripe -c $dir)
2235 if [ $mdt_index = 0 -a $stripe_cnt -le 1 ]; then
2236 $LFS mv -M 1 $dir || {
2238 error_noexit "migrate MDT1 failed"
2244 for dir in $(find ! -name . ! -name .lustre -type d); do
2245 mdt_index=$($LFS getdirstripe -i $dir)
2246 stripe_cnt=$($LFS getdirstripe -c $dir)
2247 if [ $mdt_index = 1 -a $stripe_cnt -le 1 ]; then
2248 $LFS mv -M 0 $dir || {
2250 error_noexit "migrate MDT0 failed"
2259 # When adding new data verification tests, please check for
2260 # the presence of the required reference files first, like
2261 # the "sha1sums" and "list" tests above, to avoid the need to
2262 # regenerate every image for each test addition.
2265 nrpcs_orig=$($LCTL get_param \
2266 -n mdc.*MDT0000*.max_rpcs_in_flight) || {
2267 error_noexit "Getting \"max_rpcs_in_flight\""
2270 nrpcs=$((nrpcs_orig + 5))
2271 $r $LCTL conf_param $fsname-MDT0000.mdc.max_rpcs_in_flight=$nrpcs || {
2272 error_noexit "Changing \"max_rpcs_in_flight\""
2275 wait_update $HOSTNAME "$LCTL get_param \
2276 -n mdc.*MDT0000*.max_rpcs_in_flight" $nrpcs || {
2277 error_noexit "Verifying \"max_rpcs_in_flight\""
2281 umount $tmp/mnt/lustre || {
2282 error_noexit "Unmounting the client"
2285 shall_cleanup_lustre=false
2287 if [[ "$dne_upgrade" != "no" ]] || $mdt2_is_available; then
2288 $r $UMOUNT $tmp/mnt/mdt1 || {
2289 error_noexit "Unmounting the MDT2"
2292 if [[ $fstype == zfs ]]; then
2293 $r "$ZPOOL export t32fs-mdt2"
2295 shall_cleanup_mdt1=false
2298 $r $UMOUNT $tmp/mnt/mdt || {
2299 error_noexit "Unmounting the MDT"
2302 if [[ $fstype == zfs ]]; then
2303 $r "$ZPOOL export t32fs-mdt1"
2305 shall_cleanup_mdt=false
2307 $r $UMOUNT $tmp/mnt/ost || {
2308 error_noexit "Unmounting the OST"
2311 if [[ $fstype == zfs ]]; then
2312 $r "$ZPOOL export t32fs-ost1"
2314 shall_cleanup_ost=false
2316 t32_reload_modules $node || {
2317 error_noexit "Reloading modules"
2321 if [[ $fstype == zfs ]]; then
2322 local poolname=t32fs-mdt1
2324 $ZPOOL list -H $poolname >/dev/null 2>&1 ||
2325 $ZPOOL import -f -d $tmp $poolname"
2327 # upgrade zpool to latest supported features,
2328 # including dnode quota accounting in 0.7.0
2329 $r "$ZPOOL upgrade $poolname"
2332 # mount a second time to make sure we didnt leave upgrade flag on
2333 $r $TUNEFS --dryrun $mdt_dev || {
2335 error_noexit "tunefs.lustre before remounting the MDT"
2339 mopts=exclude=$fsname-OST0000
2340 if [ $fstype == "ldiskfs" ]; then
2343 $r $MOUNT_CMD -o $mopts $mdt_dev $tmp/mnt/mdt || {
2344 error_noexit "Remounting the MDT"
2347 shall_cleanup_mdt=true
2357 for tarball in $tarballs; do
2358 t32_test $tarball || let "rc += $?"
2362 run_test 32a "Upgrade (not live)"
2370 for tarball in $tarballs; do
2371 t32_test $tarball writeconf || let "rc += $?"
2375 run_test 32b "Upgrade with writeconf"
2382 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
2384 for tarball in $tarballs; do
2385 # Do not support 1_8 and 2_1 direct upgrade to DNE2 anymore */
2386 echo $tarball | grep "1_8" && continue
2387 echo $tarball | grep "2_1" && continue
2389 dne_upgrade=yes t32_test $tarball writeconf || rc=$?
2393 run_test 32c "dne upgrade test"
2401 for tarball in $tarballs; do
2402 ff_convert=yes t32_test $tarball || rc=$?
2406 run_test 32d "convert ff test"
2408 test_33a() { # bug 12333, was test_33
2409 local FSNAME2=test-123
2410 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
2413 [ -n "$ost1_HOST" ] && fs2ost_HOST=$ost1_HOST
2415 if [ -z "$fs2ost_DEV" -o -z "$fs2mds_DEV" ]; then
2416 local dev=${SINGLEMDS}_dev
2417 local MDSDEV=${!dev}
2418 is_blkdev $SINGLEMDS $MDSDEV &&
2419 skip_env "mixed loopback and real device not working" &&
2423 local fs2mdsdev=$(mdsdevname 1_2)
2424 local fs2ostdev=$(ostdevname 1_2)
2425 local fs2mdsvdev=$(mdsvdevname 1_2)
2426 local fs2ostvdev=$(ostvdevname 1_2)
2428 if [ $(facet_fstype mds1) == ldiskfs ]; then
2429 mkfsoptions="--mkfsoptions=\\\"-J size=8\\\"" # See bug 17931.
2432 add fs2mds $(mkfs_opts mds1 ${fs2mdsdev}) --mgs --fsname=${FSNAME2} \
2433 --reformat $mkfsoptions $fs2mdsdev $fs2mdsvdev || exit 10
2434 add fs2ost $(mkfs_opts ost1 ${fs2ostdev}) --mgsnode=$MGSNID \
2435 --fsname=${FSNAME2} --index=8191 --reformat $fs2ostdev \
2436 $fs2ostvdev || exit 10
2438 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS && trap cleanup_fs2 EXIT INT
2439 start fs2ost $fs2ostdev $OST_MOUNT_OPTS
2440 do_facet $SINGLEMDS "$LCTL conf_param $FSNAME2.sys.timeout=200" ||
2441 error "$LCTL conf_param $FSNAME2.sys.timeout=200 failed"
2442 mkdir -p $MOUNT2 || error "mkdir $MOUNT2 failed"
2443 $MOUNT_CMD $MGSNID:/${FSNAME2} $MOUNT2 || error "$MOUNT_CMD failed"
2446 cp /etc/hosts $MOUNT2/ || error "copy /etc/hosts $MOUNT2/ failed"
2447 $GETSTRIPE $MOUNT2/hosts || error "$GETSTRIPE $MOUNT2/hosts failed"
2452 cleanup_nocli || error "cleanup_nocli failed with $?"
2454 run_test 33a "Mount ost with a large index number"
2456 test_33b() { # was test_34
2459 do_facet client dd if=/dev/zero of=$MOUNT/24 bs=1024k count=1
2460 # Drop lock cancelation reply during umount
2461 #define OBD_FAIL_LDLM_CANCEL_NET 0x304
2462 do_facet client $LCTL set_param fail_loc=0x80000304
2463 #lctl set_param debug=-1
2464 umount_client $MOUNT
2465 cleanup || error "cleanup failed with $?"
2467 run_test 33b "Drop cancel during umount"
2471 do_facet client "sh runmultiop_bg_pause $DIR/file O_c"
2472 manual_umount_client
2474 do_facet client killall -USR1 multiop
2475 if [ $rc -eq 0 ]; then
2476 error "umount not fail!"
2479 cleanup || error "cleanup failed with rc $?"
2481 run_test 34a "umount with opened file should be fail"
2485 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
2486 stop_mds || error "Unable to stop MDS"
2488 manual_umount_client --force || error "mtab after failed umount with $?"
2490 cleanup || error "cleanup failed with $?"
2492 run_test 34b "force umount with failed mds should be normal"
2496 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
2497 stop_ost || error "Unable to stop OST1"
2499 manual_umount_client --force || error "mtab after failed umount with $?"
2501 cleanup || error "cleanup failed with $?"
2503 run_test 34c "force umount with failed ost should be normal"
2505 test_35a() { # bug 12459
2508 DBG_SAVE="`$LCTL get_param -n debug`"
2509 $LCTL set_param debug="ha"
2511 log "Set up a fake failnode for the MDS"
2513 local device=$(do_facet $SINGLEMDS "$LCTL get_param -n devices" |
2514 awk '($3 ~ "mdt" && $4 ~ "MDT") { print $4 }' | head -1)
2515 do_facet mgs "$LCTL conf_param \
2516 ${device}.failover.node=$(h2nettype $FAKENID)" ||
2517 error "Setting ${device}.failover.node=\
2518 $(h2nettype $FAKENID) failed."
2520 log "Wait for RECONNECT_INTERVAL seconds (10s)"
2523 MSG="conf-sanity.sh test_35a `date +%F%kh%Mm%Ss`"
2526 log "Stopping the MDT: $device"
2527 stop_mdt 1 || error "MDT0 stop fail"
2529 df $MOUNT > /dev/null 2>&1 &
2531 log "Restarting the MDT: $device"
2532 start_mdt 1 || error "MDT0 start fail"
2533 log "Wait for df ($DFPID) ... "
2536 $LCTL set_param debug="$DBG_SAVE"
2538 # retrieve from the log the first server that the client tried to
2539 # contact after the connection loss
2540 $LCTL dk $TMP/lustre-log-$TESTNAME.log
2541 NEXTCONN=`awk "/${MSG}/ {start = 1;}
2542 /import_select_connection.*$device-mdc.* using connection/ {
2544 if (\\\$NF ~ /$FAKENID/)
2550 }" $TMP/lustre-log-$TESTNAME.log`
2551 [ "$NEXTCONN" != "0" ] &&
2552 error "Tried to connect to ${NEXTCONN} not last active server"
2553 cleanup || error "cleanup failed with $?"
2554 # remove nid settings
2555 writeconf_or_reformat
2557 run_test 35a "Reconnect to the last active server first"
2559 test_35b() { # bug 18674
2560 remote_mds || { skip "local MDS" && return 0; }
2564 $LCTL set_param debug="ha"
2566 MSG="conf-sanity.sh test_35b `date +%F%kh%Mm%Ss`"
2569 log "Set up a fake failnode for the MDS"
2571 local device=$(do_facet $SINGLEMDS "$LCTL get_param -n devices" |
2572 awk '($3 ~ "mdt" && $4 ~ "MDT") { print $4 }' | head -1)
2573 do_facet mgs "$LCTL conf_param \
2574 ${device}.failover.node=$(h2nettype $FAKENID)" ||
2575 error "Set ${device}.failover.node=\
2576 $(h2nettype $FAKENID) failed"
2578 local at_max_saved=0
2579 # adaptive timeouts may prevent seeing the issue
2580 if at_is_enabled; then
2581 at_max_saved=$(at_max_get mds)
2582 at_max_set 0 mds client
2585 mkdir $MOUNT/$tdir || error "mkdir $MOUNT/$tdir failed"
2587 log "Injecting EBUSY on MDS"
2588 # Setting OBD_FAIL_MDS_RESEND=0x136
2589 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0x80000136" ||
2590 error "unable to set param fail_loc=0x80000136"
2592 $LCTL set_param mdc.${FSNAME}*.stats=clear
2594 log "Creating a test file and stat it"
2595 touch $MOUNT/$tdir/$tfile || error "touch $MOUNT/$tdir/$tfile failed"
2596 stat $MOUNT/$tdir/$tfile
2598 log "Stop injecting EBUSY on MDS"
2599 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0" ||
2600 error "unable to set param fail_loc=0"
2601 rm -f $MOUNT/$tdir/$tfile || error "remove $MOUNT/$tdir/$tfile failed"
2604 # restore adaptive timeout
2605 [ $at_max_saved -ne 0 ] && at_max_set $at_max_saved mds client
2607 $LCTL dk $TMP/lustre-log-$TESTNAME.log
2609 CONNCNT=$($LCTL get_param mdc.${FSNAME}*.stats |
2610 awk '/mds_connect/{print $2}')
2612 # retrieve from the log if the client has ever tried to
2613 # contact the fake server after the loss of connection
2614 FAILCONN=`awk "BEGIN {ret = 0;}
2615 /import_select_connection.*${FSNAME}-MDT0000-mdc.* using connection/ {
2617 if (\\\$NF ~ /$FAKENID/) {
2622 END {print ret}" $TMP/lustre-log-$TESTNAME.log`
2624 [ "$FAILCONN" == "0" ] &&
2625 error "The client reconnection has not been triggered"
2626 [ "$FAILCONN" == "2" ] &&
2627 error "Primary server busy, client reconnect to failover failed"
2630 # When OBD_FAIL_MDS_RESEND is hit, we sleep for 2 * obd_timeout
2631 # Reconnects are supposed to be rate limited to one every 5s
2632 [ $CONNCNT -gt $((2 * $TIMEOUT / 5 + 1)) ] &&
2633 error "Too many reconnects $CONNCNT"
2635 cleanup || error "cleanup failed with $?"
2636 # remove nid settings
2637 writeconf_or_reformat
2639 run_test 35b "Continue reconnection retries, if the active server is busy"
2642 [ $OSTCOUNT -lt 2 ] && skip_env "needs >= 2 OSTs" && return
2644 [ "$ost_HOST" = "`hostname`" -o "$ost1_HOST" = "`hostname`" ] ||
2645 { skip "remote OST" && return 0; }
2648 local FSNAME2=test1234
2649 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
2651 [ -n "$ost1_HOST" ] && fs2ost_HOST=$ost1_HOST && fs3ost_HOST=$ost1_HOST
2653 if [ -z "$fs2ost_DEV" -o -z "$fs2mds_DEV" -o -z "$fs3ost_DEV" ]; then
2654 is_blkdev $SINGLEMDS $MDSDEV &&
2655 skip_env "mixed loopback and real device not working" && return
2658 local fs2mdsdev=$(mdsdevname 1_2)
2659 local fs2ostdev=$(ostdevname 1_2)
2660 local fs3ostdev=$(ostdevname 2_2)
2661 local fs2mdsvdev=$(mdsvdevname 1_2)
2662 local fs2ostvdev=$(ostvdevname 1_2)
2663 local fs3ostvdev=$(ostvdevname 2_2)
2665 add fs2mds $(mkfs_opts mds1 ${fs2mdsdev}) --mgs --fsname=${FSNAME2} \
2666 --reformat $fs2mdsdev $fs2mdsvdev || exit 10
2667 # XXX after we support non 4K disk blocksize in ldiskfs, specify a
2668 # different one than the default value here.
2669 add fs2ost $(mkfs_opts ost1 ${fs2ostdev}) --mgsnode=$MGSNID \
2670 --fsname=${FSNAME2} --reformat $fs2ostdev $fs2ostvdev || exit 10
2671 add fs3ost $(mkfs_opts ost2 ${fs3ostdev}) --mgsnode=$MGSNID \
2672 --fsname=${FSNAME2} --reformat $fs3ostdev $fs3ostvdev || exit 10
2674 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS
2675 start fs2ost $fs2ostdev $OST_MOUNT_OPTS
2676 start fs3ost $fs3ostdev $OST_MOUNT_OPTS
2677 mkdir -p $MOUNT2 || error "mkdir $MOUNT2 failed"
2678 $MOUNT_CMD $MGSNID:/${FSNAME2} $MOUNT2 || error "$MOUNT_CMD failed"
2680 sleep 5 # until 11778 fixed
2682 dd if=/dev/zero of=$MOUNT2/$tfile bs=1M count=7 || error "dd failed"
2684 BKTOTAL=$($LCTL get_param -n obdfilter.*.kbytestotal |
2685 awk 'BEGIN{total=0}; {total+=$1}; END{print total}')
2686 BKFREE=$($LCTL get_param -n obdfilter.*.kbytesfree |
2687 awk 'BEGIN{free=0}; {free+=$1}; END{print free}')
2688 BKAVAIL=$($LCTL get_param -n obdfilter.*.kbytesavail |
2689 awk 'BEGIN{avail=0}; {avail+=$1}; END{print avail}')
2690 STRING=$(df -P $MOUNT2 | tail -n 1 | awk '{print $2","$3","$4}')
2691 DFTOTAL=$(echo $STRING | cut -d, -f1)
2692 DFUSED=$(echo $STRING | cut -d, -f2)
2693 DFAVAIL=$(echo $STRING | cut -d, -f3)
2694 DFFREE=$(($DFTOTAL - $DFUSED))
2696 ALLOWANCE=$((64 * $OSTCOUNT))
2698 if [ $DFTOTAL -lt $(($BKTOTAL - $ALLOWANCE)) ] ||
2699 [ $DFTOTAL -gt $(($BKTOTAL + $ALLOWANCE)) ] ; then
2700 echo "**** FAIL: df total($DFTOTAL) mismatch OST total($BKTOTAL)"
2703 if [ $DFFREE -lt $(($BKFREE - $ALLOWANCE)) ] ||
2704 [ $DFFREE -gt $(($BKFREE + $ALLOWANCE)) ] ; then
2705 echo "**** FAIL: df free($DFFREE) mismatch OST free($BKFREE)"
2708 if [ $DFAVAIL -lt $(($BKAVAIL - $ALLOWANCE)) ] ||
2709 [ $DFAVAIL -gt $(($BKAVAIL + $ALLOWANCE)) ] ; then
2710 echo "**** FAIL: df avail($DFAVAIL) mismatch OST avail($BKAVAIL)"
2715 stop fs3ost -f || error "unable to stop OST3"
2716 stop fs2ost -f || error "unable to stop OST2"
2717 stop fs2mds -f || error "unable to stop second MDS"
2718 unload_modules_conf || error "unable unload modules"
2721 run_test 36 "df report consistency on OSTs with different block size"
2724 local mntpt=$(facet_mntpt $SINGLEMDS)
2725 local mdsdev=$(mdsdevname ${SINGLEMDS//mds/})
2726 local mdsdev_sym="$TMP/sym_mdt.img"
2727 local opts=$MDS_MOUNT_OPTS
2730 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
2731 skip "ldiskfs only test"
2735 echo "MDS : $mdsdev"
2736 echo "SYMLINK : $mdsdev_sym"
2737 do_facet $SINGLEMDS rm -f $mdsdev_sym
2739 do_facet $SINGLEMDS ln -s $mdsdev $mdsdev_sym
2741 echo "mount symlink device - $mdsdev_sym"
2743 if ! do_facet $SINGLEMDS test -b $mdsdev; then
2744 opts=$(csa_add "$opts" -o loop)
2746 mount_op=$(do_facet $SINGLEMDS mount -v -t lustre $opts \
2747 $mdsdev_sym $mntpt 2>&1)
2750 echo mount_op=$mount_op
2752 do_facet $SINGLEMDS "$UMOUNT $mntpt && rm -f $mdsdev_sym"
2754 if $(echo $mount_op | grep -q "unable to set tunable"); then
2755 error "set tunables failed for symlink device"
2758 [ $rc -eq 0 ] || error "mount symlink $mdsdev_sym failed! rc=$rc"
2760 run_test 37 "verify set tunables works for symlink device"
2762 test_38() { # bug 14222
2763 local fstype=$(facet_fstype $SINGLEMDS)
2764 local mntpt=$(facet_mntpt $SINGLEMDS)
2769 local SRC="/etc /bin"
2770 local FILES=$(find $SRC -type f -mtime +1 | head -n $COUNT)
2771 log "copying $(echo $FILES | wc -w) files to $DIR/$tdir"
2772 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
2773 tar cf - $FILES | tar xf - -C $DIR/$tdir ||
2774 error "copying $SRC to $DIR/$tdir"
2776 umount_client $MOUNT || error "umount_client $MOUNT failed"
2777 do_facet $SINGLEMDS "$LCTL get_param osp.*.prealloc_next_id"
2778 stop_mds || error "Unable to stop MDS"
2779 log "delete lov_objid file on MDS"
2781 mount_fstype $SINGLEMDS || error "mount MDS failed (1)"
2783 do_facet $SINGLEMDS "od -Ax -td8 $mntpt/lov_objid; rm $mntpt/lov_objid"
2785 unmount_fstype $SINGLEMDS || error "umount failed (1)"
2787 # check create in mds_lov_connect
2788 start_mds || error "unable to start MDS"
2789 mount_client $MOUNT || error "mount_client $MOUNT failed"
2791 [ $V ] && log "verifying $DIR/$tdir/$f"
2792 diff -q $f $DIR/$tdir/$f || ERROR=y
2794 do_facet $SINGLEMDS "$LCTL get_param osp.*.prealloc_next_id"
2795 if [ "$ERROR" = "y" ]; then
2796 # check it's updates in sync
2797 umount_client $MOUNT
2799 mount_fstype $SIGNLEMDS
2800 do_facet $SINGLEMDS "od -Ax -td8 $mntpt/lov_objid"
2801 unmount_fstype $SINGLEMDS
2802 error "old and new files are different after connect" || true
2804 touch $DIR/$tdir/f2 || error "f2 file create failed"
2806 # check it's updates in sync
2807 umount_client $MOUNT || error "second umount_client $MOUNT failed"
2810 mount_fstype $SINGLEMDS || error "mount MDS failed (3)"
2812 do_facet $SINGLEMDS "od -Ax -td8 $mntpt/lov_objid"
2813 do_facet $SINGLEMDS dd if=/dev/zero of=$mntpt/lov_objid.clear count=8
2815 unmount_fstype $SINGLEMDS || error "umount failed (3)"
2817 start_mds || error "unable to start MDS"
2818 mount_client $MOUNT || error "mount_client $MOUNT failed"
2820 [ $V ] && log "verifying $DIR/$tdir/$f"
2821 diff -q $f $DIR/$tdir/$f || ERROR=y
2823 touch $DIR/$tdir/f3 || error "f3 file create failed"
2824 do_facet $SINGLEMDS "$LCTL get_param osp.*.prealloc_next_id"
2825 umount_client $MOUNT || error "third umount_client $MOUNT failed"
2827 mount_fstype $SINGLEMDS || error "mount MDS failed (4)"
2828 do_facet $SINGLEMDS "od -Ax -td8 $mntpt/lov_objid"
2829 unmount_fstype $SINGLEMDS || error "umount failed (4)"
2831 [ "$ERROR" = "y" ] &&
2832 error "old and new files are different after sync" || true
2834 log "files compared the same"
2835 cleanup || error "cleanup failed with $?"
2837 run_test 38 "MDS recreates missing lov_objid file from OST data"
2842 cleanup || error "cleanup failed with $?"
2843 perl $SRCDIR/leak_finder.pl $TMP/debug 2>&1 | egrep '*** Leak:' &&
2844 error "memory leak detected" || true
2846 run_test 39 "leak_finder recognizes both LUSTRE and LNET malloc messages"
2848 test_40() { # bug 15759
2849 start_ost || error "Unable to start OST1"
2850 #define OBD_FAIL_TGT_TOOMANY_THREADS 0x706
2851 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0x80000706"
2853 cleanup || error "cleanup failed with rc $?"
2855 run_test 40 "race during service thread startup"
2857 test_41a() { #bug 14134
2858 if [ $(facet_fstype $SINGLEMDS) == ldiskfs ] &&
2859 ! do_facet $SINGLEMDS test -b $(mdsdevname 1); then
2860 skip "Loop devices does not work with nosvc option"
2865 { skip "needs combined MGT and MDT device" && return 0; }
2867 start_mdt 1 -o nosvc -n
2868 if [ $MDSCOUNT -ge 2 ]; then
2869 for num in $(seq 2 $MDSCOUNT); do
2870 start_mdt $num || return
2873 start ost1 $(ostdevname 1) $OST_MOUNT_OPTS
2874 start_mdt 1 -o nomgs,force
2875 mount_client $MOUNT || error "mount_client $MOUNT failed"
2878 echo "blah blah" > $MOUNT/$tfile
2881 umount_client $MOUNT || error "umount_client $MOUNT failed"
2882 stop ost1 -f || error "unable to stop OST1"
2883 stop_mds || error "Unable to stop MDS"
2884 stop_mds || error "Unable to stop MDS on second try"
2886 run_test 41a "mount mds with --nosvc and --nomgs"
2889 if [ $(facet_fstype $SINGLEMDS) == ldiskfs ] &&
2890 ! do_facet $SINGLEMDS test -b $(mdsdevname 1); then
2891 skip "Loop devices does not work with nosvc option"
2895 ! combined_mgs_mds && skip "needs combined mgs device" && return 0
2899 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
2901 start_mdt 1 -o nosvc -n
2902 if [ $MDSCOUNT -ge 2 ]; then
2903 for num in $(seq 2 $MDSCOUNT); do
2904 start_mdt $num || return
2907 start_ost || error "Unable to start OST1"
2908 start_mdt 1 -o nomgs,force
2909 mount_client $MOUNT || error "mount_client $MOUNT failed"
2912 echo "blah blah" > $MOUNT/$tfile
2913 cat $MOUNT/$tfile || error "cat $MOUNT/$tfile failed"
2915 umount_client $MOUNT -f || error "umount_client $MOUNT failed"
2916 stop_ost || error "Unable to stop OST1"
2917 stop_mds || error "Unable to stop MDS"
2918 stop_mds || error "Unable to stop MDS on second try"
2920 run_test 41b "mount mds with --nosvc and --nomgs on first mount"
2923 local server_version=$(lustre_version_code $SINGLEMDS)
2924 local oss_list=$(comma_list $(osts_nodes))
2926 [[ $server_version -ge $(version_code 2.6.52) ]] ||
2927 [[ $server_version -ge $(version_code 2.5.26) &&
2928 $server_version -lt $(version_code 2.5.50) ]] ||
2929 [[ $server_version -ge $(version_code 2.5.4) &&
2930 $server_version -lt $(version_code 2.5.11) ]] ||
2931 { skip "Need MDS version 2.5.4+ or 2.5.26+ or 2.6.52+"; return; }
2933 # ensure mds1 ost1 have been created even if running sub-test standalone
2936 cleanup || error "cleanup failed"
2938 # using directly mount command instead of start() function to avoid
2939 # any side effect of // with others/externals tools/features
2940 # ("zpool import", ...)
2942 # MDT concurrent start
2944 LOAD_MODULES_REMOTE=true load_modules
2945 do_facet $SINGLEMDS "lsmod | grep -q libcfs" ||
2946 error "MDT concurrent start: libcfs module not loaded"
2948 local mds1dev=$(mdsdevname 1)
2949 local mds1mnt=$(facet_mntpt mds1)
2950 local mds1fstype=$(facet_fstype mds1)
2951 local mds1opts=$MDS_MOUNT_OPTS
2953 if [ $mds1fstype == ldiskfs ] &&
2954 ! do_facet mds1 test -b $mds1dev; then
2955 mds1opts=$(csa_add "$mds1opts" -o loop)
2957 if [[ $mds1fstype == zfs ]]; then
2958 import_zpool mds1 || return ${PIPESTATUS[0]}
2961 #define OBD_FAIL_TGT_MOUNT_RACE 0x716
2962 do_facet mds1 "$LCTL set_param fail_loc=0x80000716"
2964 do_facet mds1 mount -t lustre $mds1dev $mds1mnt $mds1opts &
2967 do_facet mds1 mount -t lustre $mds1dev $mds1mnt $mds1opts
2971 do_facet mds1 "$LCTL set_param fail_loc=0x0"
2972 if [ $rc -eq 0 ] && [ $rc2 -ne 0 ]; then
2973 echo "1st MDT start succeed"
2974 echo "2nd MDT start failed with $rc2"
2975 elif [ $rc2 -eq 0 ] && [ $rc -ne 0 ]; then
2976 echo "1st MDT start failed with $rc"
2977 echo "2nd MDT start succeed"
2980 error "unexpected concurrent MDT mounts result, rc=$rc rc2=$rc2"
2983 if [ $MDSCOUNT -ge 2 ]; then
2984 for num in $(seq 2 $MDSCOUNT); do
2985 start_mdt $num || return
2989 # OST concurrent start
2991 do_rpc_nodes $oss_list "lsmod | grep -q libcfs" ||
2992 error "OST concurrent start: libcfs module not loaded"
2994 local ost1dev=$(ostdevname 1)
2995 local ost1mnt=$(facet_mntpt ost1)
2996 local ost1fstype=$(facet_fstype ost1)
2997 local ost1opts=$OST_MOUNT_OPTS
2999 if [ $ost1fstype == ldiskfs ] &&
3000 ! do_facet ost1 test -b $ost1dev; then
3001 ost1opts=$(csa_add "$ost1opts" -o loop)
3003 if [[ $ost1fstype == zfs ]]; then
3004 import_zpool ost1 || return ${PIPESTATUS[0]}
3007 #define OBD_FAIL_TGT_MOUNT_RACE 0x716
3008 do_facet ost1 "$LCTL set_param fail_loc=0x80000716"
3010 do_facet ost1 mount -t lustre $ost1dev $ost1mnt $ost1opts &
3013 do_facet ost1 mount -t lustre $ost1dev $ost1mnt $ost1opts
3017 do_facet ost1 "$LCTL set_param fail_loc=0x0"
3018 if [ $rc -eq 0 ] && [ $rc2 -ne 0 ]; then
3019 echo "1st OST start succeed"
3020 echo "2nd OST start failed with $rc2"
3021 elif [ $rc2 -eq 0 ] && [ $rc -ne 0 ]; then
3022 echo "1st OST start failed with $rc"
3023 echo "2nd OST start succeed"
3027 error "unexpected concurrent OST mounts result, rc=$rc rc2=$rc2"
3033 # verify everything ok
3038 error "MDT(s) start failed"
3046 error "OST(s) start failed"
3054 error "client start failed"
3061 error "client mount failed"
3065 run_test 41c "concurrent mounts of MDT/OST should all fail but one"
3067 test_42() { #bug 14693
3069 check_mount || error "client was not mounted"
3071 do_facet mgs $LCTL conf_param $FSNAME.llite.some_wrong_param=10
3072 umount_client $MOUNT ||
3073 error "unmounting client failed with invalid llite param"
3074 mount_client $MOUNT ||
3075 error "mounting client failed with invalid llite param"
3077 do_facet mgs $LCTL conf_param $FSNAME.sys.some_wrong_param=20
3078 cleanup || error "stopping $FSNAME failed with invalid sys param"
3080 check_mount || error "client was not mounted with invalid sys param"
3081 cleanup || error "stopping $FSNAME failed with invalid sys param"
3083 run_test 42 "allow client/server mount/unmount with invalid config param"
3086 [[ $(lustre_version_code mgs) -ge $(version_code 2.5.58) ]] ||
3087 { skip "Need MDS version at least 2.5.58" && return 0; }
3088 [ $UID -ne 0 -o $RUNAS_ID -eq 0 ] && skip_env "run as root"
3091 USER1=$(getent passwd | grep :$ID1:$ID1: | cut -d: -f1)
3092 [ -z "$USER1" ] && skip_env "missing user with uid=$ID1 gid=$ID1" &&
3096 chmod ugo+x $DIR || error "chmod 0 failed"
3097 set_conf_param_and_check mds1 \
3098 "$LCTL get_param -n mdt.$FSNAME-MDT0000.root_squash" \
3099 "$FSNAME.mdt.root_squash" \
3101 wait_update $HOSTNAME \
3102 "$LCTL get_param -n llite.${FSNAME}*.root_squash" \
3104 error "check llite root_squash failed!"
3105 set_conf_param_and_check mds1 \
3106 "$LCTL get_param -n mdt.$FSNAME-MDT0000.nosquash_nids" \
3107 "$FSNAME.mdt.nosquash_nids" \
3109 wait_update $HOSTNAME \
3110 "$LCTL get_param -n llite.${FSNAME}*.nosquash_nids" \
3112 error "check llite nosquash_nids failed!"
3115 # create set of test files
3117 echo "111" > $DIR/$tfile-userfile || error "write 1 failed"
3118 chmod go-rw $DIR/$tfile-userfile || error "chmod 1 failed"
3119 chown $RUNAS_ID.$RUNAS_ID $DIR/$tfile-userfile || error "chown failed"
3121 echo "222" > $DIR/$tfile-rootfile || error "write 2 failed"
3122 chmod go-rw $DIR/$tfile-rootfile || error "chmod 2 faield"
3124 mkdir $DIR/$tdir-rootdir || error "mkdir failed"
3125 chmod go-rwx $DIR/$tdir-rootdir || error "chmod 3 failed"
3126 touch $DIR/$tdir-rootdir/tfile-1 || error "touch failed"
3128 echo "777" > $DIR/$tfile-user1file || error "write 7 failed"
3129 chmod go-rw $DIR/$tfile-user1file || error "chmod 7 failed"
3130 chown $ID1.$ID1 $DIR/$tfile-user1file || error "chown failed"
3133 # check root_squash:
3134 # set root squash UID:GID to RUNAS_ID
3135 # root should be able to access only files owned by RUNAS_ID
3137 set_conf_param_and_check mds1 \
3138 "$LCTL get_param -n mdt.$FSNAME-MDT0000.root_squash" \
3139 "$FSNAME.mdt.root_squash" \
3140 "$RUNAS_ID:$RUNAS_ID"
3141 wait_update $HOSTNAME \
3142 "$LCTL get_param -n llite.${FSNAME}*.root_squash" \
3143 "$RUNAS_ID:$RUNAS_ID" ||
3144 error "check llite root_squash failed!"
3146 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-userfile)
3147 dd if=$DIR/$tfile-userfile 1>/dev/null 2>/dev/null ||
3148 error "$ST: root read permission is denied"
3149 echo "$ST: root read permission is granted - ok"
3152 dd conv=notrunc of=$DIR/$tfile-userfile 1>/dev/null 2>/dev/null ||
3153 error "$ST: root write permission is denied"
3154 echo "$ST: root write permission is granted - ok"
3156 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-rootfile)
3157 dd if=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null &&
3158 error "$ST: root read permission is granted"
3159 echo "$ST: root read permission is denied - ok"
3162 dd conv=notrunc of=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null &&
3163 error "$ST: root write permission is granted"
3164 echo "$ST: root write permission is denied - ok"
3166 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tdir-rootdir)
3167 rm $DIR/$tdir-rootdir/tfile-1 1>/dev/null 2>/dev/null &&
3168 error "$ST: root unlink permission is granted"
3169 echo "$ST: root unlink permission is denied - ok"
3171 touch $DIR/tdir-rootdir/tfile-2 1>/dev/null 2>/dev/null &&
3172 error "$ST: root create permission is granted"
3173 echo "$ST: root create permission is denied - ok"
3176 # check root_squash is enforced independently
3177 # of client cache content
3179 # access file by USER1, keep access open
3180 # root should be denied access to user file
3182 runas -u $ID1 tail -f $DIR/$tfile-user1file 1>/dev/null 2>&1 &
3186 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-user1file)
3187 dd if=$DIR/$tfile-user1file 1>/dev/null 2>&1 &&
3188 { kill $pid; error "$ST: root read permission is granted"; }
3189 echo "$ST: root read permission is denied - ok"
3192 dd conv=notrunc of=$DIR/$tfile-user1file 1>/dev/null 2>&1 &&
3193 { kill $pid; error "$ST: root write permission is granted"; }
3194 echo "$ST: root write permission is denied - ok"
3200 # check nosquash_nids:
3201 # put client's NID into nosquash_nids list,
3202 # root should be able to access root file after that
3204 local NIDLIST=$($LCTL list_nids all | tr '\n' ' ')
3205 NIDLIST="2@gni $NIDLIST 192.168.0.[2,10]@tcp"
3206 NIDLIST=$(echo $NIDLIST | tr -s ' ' ' ')
3207 set_conf_param_and_check mds1 \
3208 "$LCTL get_param -n mdt.$FSNAME-MDT0000.nosquash_nids" \
3209 "$FSNAME-MDTall.mdt.nosquash_nids" \
3211 wait_update $HOSTNAME \
3212 "$LCTL get_param -n llite.${FSNAME}*.nosquash_nids" \
3214 error "check llite nosquash_nids failed!"
3216 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-rootfile)
3217 dd if=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null ||
3218 error "$ST: root read permission is denied"
3219 echo "$ST: root read permission is granted - ok"
3222 dd conv=notrunc of=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null ||
3223 error "$ST: root write permission is denied"
3224 echo "$ST: root write permission is granted - ok"
3226 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tdir-rootdir)
3227 rm $DIR/$tdir-rootdir/tfile-1 ||
3228 error "$ST: root unlink permission is denied"
3229 echo "$ST: root unlink permission is granted - ok"
3230 touch $DIR/$tdir-rootdir/tfile-2 ||
3231 error "$ST: root create permission is denied"
3232 echo "$ST: root create permission is granted - ok"
3233 cleanup || error "cleanup failed with $?"
3235 run_test 43a "check root_squash and nosquash_nids"
3237 test_43b() { # LU-5690
3238 [[ $(lustre_version_code mgs) -ge $(version_code 2.7.62) ]] ||
3239 { skip "Need MGS version 2.7.62+"; return; }
3241 if [[ -z "$fs2mds_DEV" ]]; then
3242 is_blkdev $SINGLEMDS $(mdsdevname ${SINGLEMDS//mds/}) &&
3243 skip_env "mixed loopback and real device not working" && return
3246 local fs2mdsdev=$(mdsdevname 1_2)
3247 local fs2mdsvdev=$(mdsvdevname 1_2)
3249 # temporarily use fs2mds as fs2mgs
3251 local fs2mgsdev=$fs2mdsdev
3252 local fs2mgsvdev=$fs2mdsvdev
3254 local fsname=test1234
3256 load_module llite/lustre
3257 local client_ip=$(host_nids_address $HOSTNAME $NETTYPE)
3258 local host=${client_ip//*./}
3259 local net=${client_ip/%$host/}
3260 local nosquash_nids=$(h2nettype $net[$host,$host,$host])
3262 add $fs2mgs $(mkfs_opts mgs $fs2mgsdev) --fsname=$fsname \
3263 --param mdt.root_squash=$RUNAS_ID:$RUNAS_ID \
3264 --param mdt.nosquash_nids=$nosquash_nids \
3265 --reformat $fs2mgsdev $fs2mgsvdev || error "add fs2mgs failed"
3266 start $fs2mgs $fs2mgsdev $MGS_MOUNT_OPTS || error "start fs2mgs failed"
3267 stop $fs2mgs -f || error "stop fs2mgs failed"
3269 run_test 43b "parse nosquash_nids with commas in expr_list"
3271 umount_client $MOUNT
3276 check_mount || error "check_mount"
3277 UUID=$($LCTL get_param llite.${FSNAME}*.uuid | cut -d= -f2)
3279 UUIDS=$(do_facet $SINGLEMDS "$LCTL get_param mdt.${FSNAME}*.exports.*.uuid")
3280 for VAL in $UUIDS; do
3281 NID=$(echo $VAL | cut -d= -f1)
3282 CLUUID=$(echo $VAL | cut -d= -f2)
3283 [ "$UUID" = "$CLUUID" ] && STATS_FOUND=yes && break
3285 [ "$STATS_FOUND" = "no" ] && error "stats not found for client"
3286 cleanup || error "cleanup failed with $?"
3288 run_test 44 "mounted client proc entry exists"
3292 check_mount || error "check_mount"
3293 stop_mds || error "Unable to stop MDS"
3297 #define OBD_FAIL_PTLRPC_LONG_REPL_UNLINK 0x50f
3298 do_facet client "$LCTL set_param fail_loc=0x8000050f"
3301 manual_umount_client --force || error "manual_umount_client failed"
3302 do_facet client "$LCTL set_param fail_loc=0x0"
3303 start_mds || error "unable to start MDS"
3304 mount_client $MOUNT || error "mount_client $MOUNT failed"
3305 cleanup || error "cleanup failed with $?"
3307 run_test 45 "long unlink handling in ptlrpcd"
3314 umount_client $MOUNT2 || rc=$?
3315 umount_client $MOUNT || rc=$?
3316 while [ $count -gt 0 ]; do
3317 stop ost${count} -f || rc=$?
3321 cleanup_nocli || rc=$?
3322 #writeconf to remove all ost2 traces for subsequent tests
3323 writeconf_or_reformat
3328 echo "Testing with $OSTCOUNT OSTs"
3330 start_mds || error "unable to start MDS"
3331 #first client should see only one ost
3332 start_ost || error "Unable to start OST1"
3333 wait_osc_import_state mds ost FULL
3335 mount_client $MOUNT || error "mount_client $MOUNT failed"
3336 trap "cleanup_46a $OSTCOUNT" EXIT ERR
3339 for (( i=2; i<=$OSTCOUNT; i++ )); do
3340 start ost$i $(ostdevname $i) $OST_MOUNT_OPTS ||
3341 error "start_ost$i $(ostdevname $i) failed"
3344 # wait until osts in sync
3345 for (( i=2; i<=$OSTCOUNT; i++ )); do
3346 wait_osc_import_state mds ost$i FULL
3347 wait_osc_import_state client ost$i FULL
3350 #second client see all ost's
3352 mount_client $MOUNT2 || error "mount_client failed"
3353 $SETSTRIPE -c -1 $MOUNT2 || error "$SETSTRIPE -c -1 $MOUNT2 failed"
3354 $GETSTRIPE $MOUNT2 || error "$GETSTRIPE $MOUNT2 failed"
3356 echo "ok" > $MOUNT2/widestripe
3357 $GETSTRIPE $MOUNT2/widestripe ||
3358 error "$GETSTRIPE $MOUNT2/widestripe failed"
3359 # fill acl buffer for avoid expand lsm to them
3360 awk -F : '{if (FNR < 25) { print "u:"$1":rwx" }}' /etc/passwd |
3362 setfacl -m $acl $MOUNT2/widestripe
3366 stat $MOUNT/widestripe || error "stat $MOUNT/widestripe failed"
3368 cleanup_46a $OSTCOUNT || error "cleanup_46a failed"
3370 run_test 46a "handle ost additional - wide striped file"
3375 check_mount || error "check_mount failed"
3376 $LCTL set_param ldlm.namespaces.$FSNAME-*-*-*.lru_size=100
3380 for ns in $($LCTL get_param ldlm.namespaces.$FSNAME-*-*-*.lru_size); do
3381 if echo $ns | grep "MDT[[:digit:]]*"; then
3384 lrs=$(echo $ns | sed 's/.*lru_size=//')
3385 lru_size[count]=$lrs
3390 facet_failover $SINGLEMDS
3391 client_up || error "client_up failed"
3394 for ns in $($LCTL get_param ldlm.namespaces.$FSNAME-*-*-*.lru_size); do
3395 if echo $ns | grep "MDT[[:digit:]]*"; then
3398 lrs=$(echo $ns | sed 's/.*lru_size=//')
3399 if ! test "$lrs" -eq "${lru_size[count]}"; then
3400 n=$(echo $ns | sed -e 's/ldlm.namespaces.//' -e 's/.lru_size=.*//')
3401 error "$n has lost lru_size: $lrs vs. ${lru_size[count]}"
3406 cleanup || error "cleanup failed with $?"
3408 run_test 47 "server restart does not make client loss lru_resize settings"
3413 # reformat after this test is needed - if the test fails,
3414 # we will have unkillable file at FS
3418 test_48() { # bz-17636 LU-7473
3422 check_mount || error "check_mount failed"
3424 $SETSTRIPE -c -1 $MOUNT || error "$SETSTRIPE -c -1 $MOUNT failed"
3425 $GETSTRIPE $MOUNT || error "$GETSTRIPE $MOUNT failed"
3427 echo "ok" > $MOUNT/widestripe
3428 $GETSTRIPE $MOUNT/widestripe ||
3429 error "$GETSTRIPE $MOUNT/widestripe failed"
3431 # In the future, we may introduce more EAs, such as selinux, enlarged
3432 # LOV EA, and so on. These EA will use some EA space that is shared by
3433 # ACL entries. So here we only check some reasonable ACL entries count,
3434 # instead of the max number that is calculated from the max_ea_size.
3435 if [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.8.57) ];
3437 count=28 # hard coded of RPC protocol
3438 elif [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
3439 count=4000 # max_num 4091 max_ea_size = 32768
3440 elif ! large_xattr_enabled; then
3441 count=450 # max_num 497 max_ea_size = 4012
3443 count=4500 # max_num 8187 max_ea_size = 1048492
3444 # not create too much (>5000) to save test time
3447 echo "It is expected to hold at least $count ACL entries"
3448 trap cleanup_48 EXIT ERR
3449 for ((i = 0; i < $count; i++)) do
3450 setfacl -m u:$((i + 100)):rw $MOUNT/widestripe ||
3451 error "Fail to setfacl for $MOUNT/widestripe at $i"
3454 cancel_lru_locks mdc
3455 stat $MOUNT/widestripe || error "stat $MOUNT/widestripe failed"
3456 local r_count=$(getfacl $MOUNT/widestripe | grep "user:" | wc -l)
3457 count=$((count + 1)) # for the entry "user::rw-"
3459 [ $count -eq $r_count ] ||
3460 error "Expected ACL entries $count, but got $r_count"
3464 run_test 48 "too many acls on file"
3466 # check PARAM_SYS_LDLM_TIMEOUT option of MKFS.LUSTRE
3467 test_49a() { # bug 17710
3468 local timeout_orig=$TIMEOUT
3469 local ldlm_timeout_orig=$LDLM_TIMEOUT
3470 local LOCAL_TIMEOUT=20
3472 LDLM_TIMEOUT=$LOCAL_TIMEOUT
3473 TIMEOUT=$LOCAL_TIMEOUT
3477 check_mount || error "client mount failed"
3479 echo "check ldlm_timout..."
3480 local LDLM_MDS="$(do_facet $SINGLEMDS $LCTL get_param -n ldlm_timeout)"
3481 local LDLM_OST1="$(do_facet ost1 $LCTL get_param -n ldlm_timeout)"
3482 local LDLM_CLIENT="$(do_facet client $LCTL get_param -n ldlm_timeout)"
3484 if [ $LDLM_MDS -ne $LDLM_OST1 -o $LDLM_MDS -ne $LDLM_CLIENT ]; then
3485 error "Different LDLM_TIMEOUT:$LDLM_MDS $LDLM_OST1 $LDLM_CLIENT"
3488 if [ $LDLM_MDS -ne $((LOCAL_TIMEOUT / 3)) ]; then
3489 error "LDLM_TIMEOUT($LDLM_MDS) is not $((LOCAL_TIMEOUT / 3))"
3492 umount_client $MOUNT || error "umount_client $MOUNT failed"
3493 stop_ost || error "problem stopping OSS"
3494 stop_mds || error "problem stopping MDS"
3496 LDLM_TIMEOUT=$ldlm_timeout_orig
3497 TIMEOUT=$timeout_orig
3499 run_test 49a "check PARAM_SYS_LDLM_TIMEOUT option of mkfs.lustre"
3501 test_49b() { # bug 17710
3502 local timeout_orig=$TIMEOUT
3503 local ldlm_timeout_orig=$LDLM_TIMEOUT
3504 local LOCAL_TIMEOUT=20
3506 LDLM_TIMEOUT=$((LOCAL_TIMEOUT - 1))
3507 TIMEOUT=$LOCAL_TIMEOUT
3511 check_mount || error "client mount failed"
3513 local LDLM_MDS="$(do_facet $SINGLEMDS $LCTL get_param -n ldlm_timeout)"
3514 local LDLM_OST1="$(do_facet ost1 $LCTL get_param -n ldlm_timeout)"
3515 local LDLM_CLIENT="$(do_facet client $LCTL get_param -n ldlm_timeout)"
3517 if [ $LDLM_MDS -ne $LDLM_OST1 -o $LDLM_MDS -ne $LDLM_CLIENT ]; then
3518 error "Different LDLM_TIMEOUT:$LDLM_MDS $LDLM_OST1 $LDLM_CLIENT"
3521 if [ $LDLM_MDS -ne $((LOCAL_TIMEOUT - 1)) ]; then
3522 error "LDLM_TIMEOUT($LDLM_MDS) is not $((LOCAL_TIMEOUT - 1))"
3525 cleanup || error "cleanup failed"
3527 LDLM_TIMEOUT=$ldlm_timeout_orig
3528 TIMEOUT=$timeout_orig
3530 run_test 49b "check PARAM_SYS_LDLM_TIMEOUT option of mkfs.lustre"
3533 # Test both statfs and lfs df and fail if either one fails
3534 multiop_bg_pause $1 f_
3537 killall -USR1 multiop
3538 [ $RC1 -ne 0 ] && log "lazystatfs multiop failed"
3539 wait $PID || { RC1=$?; log "multiop return error "; }
3546 if [ $RC2 -eq 0 ]; then
3548 log "lazystatfs df failed"
3552 [[ $RC1 -ne 0 || $RC2 -eq 0 ]] && RC=1
3558 $LCTL set_param llite.$FSNAME-*.lazystatfs=1
3559 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
3561 lazystatfs $MOUNT || error "lazystatfs failed but no down servers"
3563 cleanup || error "cleanup failed with rc $?"
3565 run_test 50a "lazystatfs all servers available"
3569 $LCTL set_param llite.$FSNAME-*.lazystatfs=1
3570 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
3572 # Wait for client to detect down OST
3573 stop_ost || error "Unable to stop OST1"
3574 wait_osc_import_state mds ost DISCONN
3576 lazystatfs $MOUNT || error "lazystatfs should not return EIO"
3578 umount_client $MOUNT || error "Unable to unmount client"
3579 stop_mds || error "Unable to stop MDS"
3581 run_test 50b "lazystatfs all servers down"
3584 start_mds || error "Unable to start MDS"
3585 start_ost || error "Unable to start OST1"
3586 start_ost2 || error "Unable to start OST2"
3587 mount_client $MOUNT || error "Unable to mount client"
3588 $LCTL set_param llite.$FSNAME-*.lazystatfs=1
3589 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
3591 # Wait for client to detect down OST
3592 stop_ost || error "Unable to stop OST1"
3593 wait_osc_import_state mds ost DISCONN
3594 lazystatfs $MOUNT || error "lazystatfs failed with one down server"
3596 umount_client $MOUNT || error "Unable to unmount client"
3597 stop_ost2 || error "Unable to stop OST2"
3598 stop_mds || error "Unable to stop MDS"
3599 #writeconf to remove all ost2 traces for subsequent tests
3600 writeconf_or_reformat
3602 run_test 50c "lazystatfs one server down"
3605 start_mds || error "Unable to start MDS"
3606 start_ost || error "Unable to start OST1"
3607 start_ost2 || error "Unable to start OST2"
3608 mount_client $MOUNT || error "Unable to mount client"
3609 $LCTL set_param llite.$FSNAME-*.lazystatfs=1
3610 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
3612 # Issue the statfs during the window where the client still
3613 # belives the OST to be available but it is in fact down.
3614 # No failure just a statfs which hangs for a timeout interval.
3615 stop_ost || error "Unable to stop OST1"
3616 lazystatfs $MOUNT || error "lazystatfs failed with one down server"
3618 umount_client $MOUNT || error "Unable to unmount client"
3619 stop_ost2 || error "Unable to stop OST2"
3620 stop_mds || error "Unable to stop MDS"
3621 #writeconf to remove all ost2 traces for subsequent tests
3622 writeconf_or_reformat
3624 run_test 50d "lazystatfs client/server conn race"
3631 start_mds || error "Unable to start MDS"
3632 #first client should see only one ost
3633 start_ost || error "Unable to start OST1"
3634 wait_osc_import_state mds ost FULL
3636 # Wait for client to detect down OST
3637 stop_ost || error "Unable to stop OST1"
3638 wait_osc_import_state mds ost DISCONN
3640 mount_client $MOUNT || error "Unable to mount client"
3641 $LCTL set_param llite.$FSNAME-*.lazystatfs=0
3643 multiop_bg_pause $MOUNT _f
3647 if [ $RC1 -ne 0 ]; then
3648 log "multiop failed $RC1"
3651 sleep $(( $TIMEOUT+1 ))
3653 [ $? -ne 0 ] && error "process isn't sleep"
3654 start_ost || error "Unable to start OST1"
3655 wait $pid || error "statfs failed"
3658 umount_client $MOUNT || error "Unable to unmount client"
3659 stop_ost || error "Unable to stop OST1"
3660 stop_mds || error "Unable to stop MDS"
3662 run_test 50e "normal statfs all servers down"
3667 CONN_PROC="osc.$FSNAME-OST0001-osc-[M]*.ost_server_uuid"
3669 start_mds || error "Unable to start mds"
3670 #first client should see only one ost
3671 start_ost || error "Unable to start OST1"
3672 wait_osc_import_state mds ost FULL
3674 start_ost2 || error "Unable to start OST2"
3675 wait_osc_import_state mds ost2 FULL
3677 # Wait for client to detect down OST
3678 stop_ost2 || error "Unable to stop OST2"
3680 wait_osc_import_state mds ost2 DISCONN
3681 mount_client $MOUNT || error "Unable to mount client"
3682 $LCTL set_param llite.$FSNAME-*.lazystatfs=0
3684 multiop_bg_pause $MOUNT _f
3688 if [ $RC1 -ne 0 ]; then
3689 log "lazystatfs multiop failed $RC1"
3692 sleep $(( $TIMEOUT+1 ))
3694 [ $? -ne 0 ] && error "process isn't sleep"
3695 start_ost2 || error "Unable to start OST2"
3696 wait $pid || error "statfs failed"
3697 stop_ost2 || error "Unable to stop OST2"
3700 umount_client $MOUNT -f || error "Unable to unmount client"
3701 stop_ost || error "Unable to stop OST1"
3702 stop_mds || error "Unable to stop MDS"
3703 #writeconf to remove all ost2 traces for subsequent tests
3704 writeconf_or_reformat
3706 run_test 50f "normal statfs one server in down"
3709 [ "$OSTCOUNT" -lt "2" ] && skip_env "needs >=2 OSTs" && return
3711 start_ost2 || error "Unable to start OST2"
3712 wait_osc_import_state mds ost2 FULL
3713 wait_osc_import_state client ost2 FULL
3715 local PARAM="${FSNAME}-OST0001.osc.active"
3717 $SETSTRIPE -c -1 $DIR/$tfile || error "$SETSTRIPE failed"
3718 do_facet mgs $LCTL conf_param $PARAM=0 ||
3719 error "Unable to deactivate OST"
3721 umount_client $MOUNT || error "Unable to unmount client"
3722 mount_client $MOUNT || error "Unable to mount client"
3723 # This df should not cause a panic
3726 do_facet mgs $LCTL conf_param $PARAM=1 || error "Unable to activate OST"
3727 rm -f $DIR/$tfile || error "unable to remove file $DIR/$tfile"
3728 umount_client $MOUNT || error "Unable to unmount client"
3729 stop_ost2 || error "Unable to stop OST2"
3730 stop_ost || error "Unable to stop OST1"
3731 stop_mds || error "Unable to stop MDS"
3732 #writeconf to remove all ost2 traces for subsequent tests
3733 writeconf_or_reformat
3735 run_test 50g "deactivated OST should not cause panic"
3739 # prepare MDT/OST, make OSC inactive for OST1
3740 [ "$OSTCOUNT" -lt "2" ] && skip_env "needs >=2 OSTs" && return
3742 [ $(facet_fstype ost1) == zfs ] && import_zpool ost1
3743 do_facet ost1 "$TUNEFS --param osc.active=0 `ostdevname 1`" ||
3744 error "tunefs OST1 failed"
3745 start_mds || error "Unable to start MDT"
3746 start_ost || error "Unable to start OST1"
3747 start_ost2 || error "Unable to start OST2"
3748 mount_client $MOUNT || error "client start failed"
3750 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
3752 # activatate OSC for OST1
3753 local TEST="$LCTL get_param -n osc.${FSNAME}-OST0000-osc-[!M]*.active"
3754 set_conf_param_and_check client \
3755 "$TEST" "${FSNAME}-OST0000.osc.active" 1 ||
3756 error "Unable to activate OST1"
3758 mkdir $DIR/$tdir/2 || error "mkdir $DIR/$tdir/2 failed"
3759 $SETSTRIPE -c -1 -i 0 $DIR/$tdir/2 ||
3760 error "$SETSTRIPE $DIR/$tdir/2 failed"
3761 sleep 1 && echo "create a file after OST1 is activated"
3762 # doing some io, shouldn't crash
3763 dd if=/dev/zero of=$DIR/$tdir/2/$tfile-io bs=1M count=10
3765 # check OSC import is working
3766 stat $DIR/$tdir/2/* >/dev/null 2>&1 ||
3767 error "some OSC imports are still not connected"
3771 umount_client $MOUNT || error "Unable to umount client"
3772 stop_ost2 || error "Unable to stop OST2"
3773 cleanup_nocli || error "cleanup_nocli failed with $?"
3775 run_test 50h "LU-642: activate deactivated OST"
3778 # prepare MDT/OST, make OSC inactive for OST1
3779 [ "$MDSCOUNT" -lt "2" ] && skip_env "needs >= 2 MDTs" && return
3782 [ $(facet_fstype mds2) == zfs ] && import_zpool mds2
3783 do_facet mds2 "$TUNEFS --param mdc.active=0 $(mdsdevname 2)" ||
3784 error "tunefs MDT2 failed"
3785 start_mds || error "Unable to start MDT"
3786 start_ost || error "Unable to start OST1"
3787 start_ost2 || error "Unable to start OST2"
3788 mount_client $MOUNT || error "client start failed"
3790 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
3792 $LCTL conf_param ${FSNAME}-MDT0000.mdc.active=0 &&
3793 error "deactive MDC0 succeeds"
3794 # activate MDC for MDT2
3795 local TEST="$LCTL get_param -n mdc.${FSNAME}-MDT0001-mdc-[!M]*.active"
3796 set_conf_param_and_check client \
3797 "$TEST" "${FSNAME}-MDT0001.mdc.active" 1 ||
3798 error "Unable to activate MDT2"
3800 wait_clients_import_state ${CLIENTS:-$HOSTNAME} mds2 FULL
3801 if [ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.7.60) ]
3803 wait_dne_interconnect
3805 $LFS mkdir -i1 $DIR/$tdir/2 || error "mkdir $DIR/$tdir/2 failed"
3807 createmany -o $DIR/$tdir/2/$tfile-%d 1 || error "create files failed"
3809 rm -rf $DIR/$tdir/2 || error "unlink dir failed"
3811 # deactivate MDC for MDT2
3812 local TEST="$LCTL get_param -n mdc.${FSNAME}-MDT0001-mdc-[!M]*.active"
3813 set_conf_param_and_check client \
3814 "$TEST" "${FSNAME}-MDT0001.mdc.active" 0 ||
3815 error "Unable to deactivate MDT2"
3817 wait_osp_active mds ${FSNAME}-MDT0001 1 0
3819 $LFS mkdir -i1 $DIR/$tdir/2 &&
3820 error "mkdir $DIR/$tdir/2 succeeds after deactive MDT"
3822 $LFS mkdir -i0 -c$MDSCOUNT $DIR/$tdir/striped_dir ||
3823 error "mkdir $DIR/$tdir/striped_dir fails after deactive MDT2"
3825 local stripe_count=$($LFS getdirstripe -c $DIR/$tdir/striped_dir)
3826 [ $stripe_count -eq $((MDSCOUNT - 1)) ] ||
3827 error "wrong $stripe_count != $((MDSCOUNT -1)) for striped_dir"
3830 umount_client $MOUNT || error "Unable to umount client"
3835 run_test 50i "activate deactivated MDT"
3838 local LOCAL_TIMEOUT=20
3842 check_mount || error "check_mount failed"
3844 mkdir $MOUNT/$tdir || error "mkdir $MOUNT/$tdir failed"
3845 $SETSTRIPE -c -1 $MOUNT/$tdir ||
3846 error "$SETSTRIPE -c -1 $MOUNT/$tdir failed"
3847 #define OBD_FAIL_MDS_REINT_DELAY 0x142
3848 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0x142"
3849 touch $MOUNT/$tdir/$tfile &
3852 start_ost2 || error "Unable to start OST1"
3854 stop_ost2 || error "Unable to stop OST1"
3855 umount_client $MOUNT -f || error “unmount $MOUNT failed”
3856 cleanup_nocli || error “stop server failed”
3857 #writeconf to remove all ost2 traces for subsequent tests
3858 writeconf_or_reformat
3860 run_test 51 "Verify that mdt_reint handles RMF_MDT_MD correctly when an OST is added"
3869 do_node $node mkdir -p $dest
3870 [ $? -eq 0 ] || { error "Unable to create directory"; return 1; }
3872 do_node $node 'tar cf - '$@' | tar xf - -C '$dest';
3873 [ \"\${PIPESTATUS[*]}\" = \"0 0\" ] || exit 1'
3874 [ $? -eq 0 ] || { error "Unable to tar files"; return 2; }
3876 do_node $node 'getfattr -d -m "[a-z]*\\." '$@' > '$xattrs
3877 [ $? -eq 0 ] || { error "Unable to read xattrs"; return 3; }
3887 local backup2=${TMP}/backup2
3889 do_node $node mkdir -p $backup2
3890 [ $? -eq 0 ] || { error "Unable to create directory"; return 1; }
3892 do_node $node 'tar cf - '$@' | tar xf - -C '$backup2';
3893 [ \"\${PIPESTATUS[*]}\" = \"0 0\" ] || exit 1'
3894 [ $? -eq 0 ] || { error "Unable to tar files to diff"; return 2; }
3896 do_node $node "diff -rq $backup $backup2"
3897 [ $? -eq 0 ] || { error "contents differ"; return 3; }
3899 local xattrs2=${TMP}/xattrs2
3900 do_node $node 'getfattr -d -m "[a-z]*\\." '$@' > '$xattrs2
3901 [ $? -eq 0 ] || { error "Unable to read xattrs to diff"; return 4; }
3903 do_node $node "diff $xattrs $xattrs2"
3904 [ $? -eq 0 ] || { error "xattrs differ"; return 5; }
3906 do_node $node "rm -rf $backup2 $xattrs2"
3907 [ $? -eq 0 ] || { error "Unable to delete temporary files"; return 6; }
3911 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
3912 skip "ldiskfs only test"
3916 start_mds || error "Unable to start MDS"
3917 start_ost || error "Unable to start OST1"
3918 mount_client $MOUNT || error "Unable to mount client"
3921 local ost1mnt=$(facet_mntpt ost1)
3922 local ost1node=$(facet_active_host ost1)
3923 local ost1tmp=$TMP/conf52
3926 mkdir $DIR/$tdir || error "Unable to create $DIR/$tdir"
3927 touch $TMP/modified_first || error "Unable to create temporary file"
3928 local mtime=$(stat -c %Y $TMP/modified_first)
3929 do_node $ost1node "mkdir -p $ost1tmp &&
3930 touch -m -d @$mtime $ost1tmp/modified_first" ||
3931 error "Unable to create temporary file"
3934 $SETSTRIPE -c -1 -S 1M $DIR/$tdir || error "$SETSTRIPE failed"
3936 for (( i=0; i < nrfiles; i++ )); do
3937 multiop $DIR/$tdir/$tfile-$i Ow1048576w1048576w524288c ||
3938 error "multiop failed"
3944 echo backup files to $TMP/$tdir
3945 local files=$(find $DIR/$tdir -type f -newer $TMP/modified_first)
3946 copy_files_xattrs $(hostname) $TMP/$tdir $TMP/file_xattrs $files ||
3947 error "Unable to copy files"
3949 umount_client $MOUNT || error "Unable to umount client"
3950 stop_ost || error "Unable to stop ost1"
3952 echo mount ost1 as ldiskfs
3953 do_node $ost1node mkdir -p $ost1mnt || error "Unable to create $ost1mnt"
3954 if ! do_node $ost1node test -b $ost1_dev; then
3957 do_node $ost1node mount -t $(facet_fstype ost1) $loop $ost1_dev \
3959 error "Unable to mount ost1 as ldiskfs"
3962 echo backup objects to $ost1tmp/objects
3963 local objects=$(do_node $ost1node 'find '$ost1mnt'/O/[0-9]* -type f'\
3964 '-size +0 -newer '$ost1tmp'/modified_first -regex ".*\/[0-9]+"')
3965 copy_files_xattrs $ost1node $ost1tmp/objects $ost1tmp/object_xattrs \
3967 error "Unable to copy objects"
3969 # move objects to lost+found
3970 do_node $ost1node 'mv '$objects' '${ost1mnt}'/lost+found'
3971 [ $? -eq 0 ] || { error "Unable to move objects"; return 14; }
3973 do_node $ost1node "umount $ost1mnt" ||
3974 error "Unable to umount ost1 as ldiskfs"
3976 start_ost || error "Unable to start OST1"
3977 mount_client $MOUNT || error "Unable to mount client"
3979 local REPAIRED=$(do_node $ost1node "$LCTL get_param \
3980 -n osd-ldiskfs.$FSNAME-OST0000.oi_scrub" |
3981 awk '/^lf_repa[ri]*ed/ { print $2 }')
3982 [ $REPAIRED -gt 0 ] ||
3983 error "Some entry under /lost+found should be repaired"
3986 diff_files_xattrs $(hostname) $TMP/$tdir $TMP/file_xattrs $files ||
3987 error "Unable to diff files"
3989 rm -rf $TMP/$tdir $TMP/file_xattrs ||
3990 error "Unable to delete temporary files"
3991 do_node $ost1node "rm -rf $ost1tmp" ||
3992 error "Unable to delete temporary files"
3993 cleanup || error "cleanup failed with $?"
3995 run_test 52 "check recovering objects from lost+found"
3997 # Checks threads_min/max/started for some service
3999 # Arguments: service name (OST or MDT), facet (e.g., ost1, $SINGLEMDS), and a
4000 # parameter pattern prefix like 'ost.*.ost'.
4013 local msg="Insane $modname thread counts"
4014 local ncpts=$(check_cpt_number $facet)
4018 check_mount || return 41
4020 # We need to expand $parampat, but it may match multiple parameters, so
4021 # we'll pick the first one
4022 if ! paramp=$(do_facet $facet "lctl get_param -N ${parampat}.threads_min"|head -1); then
4023 error "Couldn't expand ${parampat}.threads_min parameter name"
4027 # Remove the .threads_min part
4028 paramp=${paramp%.threads_min}
4030 # Check for sanity in defaults
4031 tmin=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min" ||
4033 tmax=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_max" ||
4035 tstarted=$(do_facet $facet "$LCTL get_param \
4036 -n ${paramp}.threads_started" || echo 0)
4037 lassert 23 "$msg (PDSH problems?)" '(($tstarted && $tmin && $tmax))' ||
4039 lassert 24 "$msg" '(($tstarted >= $tmin && $tstarted <= $tmax ))' ||
4041 nthrs=$(expr $tmax - $tmin)
4042 if [ $nthrs -lt $ncpts ]; then
4048 [ $tmin -eq $tmax -a $tmin -eq $tstarted ] &&
4049 skip_env "module parameter forced $facet thread count" &&
4050 tmin=3 && tmax=$((3 * tmax))
4052 # Check that we can change min/max
4053 do_facet $facet "$LCTL set_param \
4054 ${paramp}.threads_min=$((tmin + nthrs))"
4055 do_facet $facet "$LCTL set_param \
4056 ${paramp}.threads_max=$((tmax - nthrs))"
4057 tmin2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min" ||
4059 tmax2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_max" ||
4061 lassert 25 "$msg" '(($tmin2 == ($tmin + $nthrs) &&
4062 $tmax2 == ($tmax - $nthrs)))' || return $?
4064 # Check that we can set min/max to the same value
4065 tmin=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min" ||
4067 do_facet $facet "$LCTL set_param ${paramp}.threads_max=$tmin"
4068 tmin2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min" ||
4070 tmax2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_max" ||
4072 lassert 26 "$msg" '(($tmin2 == $tmin && $tmax2 == $tmin))' || return $?
4074 # Check that we can't set max < min
4075 do_facet $facet "$LCTL set_param ${paramp}.threads_max=$((tmin - 1))"
4076 tmin2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min" ||
4078 tmax2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_max" ||
4080 lassert 27 "$msg" '(($tmin2 <= $tmax2))' || return $?
4082 # We need to ensure that we get the module options desired; to do this
4083 # we set LOAD_MODULES_REMOTE=true and we call setmodopts below.
4084 LOAD_MODULES_REMOTE=true
4087 local newvalue="${opts}=$(expr $basethr \* $ncpts)"
4088 setmodopts -a $modname "$newvalue" oldvalue
4091 check_mount || return 41
4093 # Restore previous setting of MODOPTS_*
4094 setmodopts $modname "$oldvalue"
4096 # Check that $opts took
4097 tmin=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min")
4098 tmax=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_max")
4099 tstarted=$(do_facet $facet \
4100 "$LCTL get_param -n ${paramp}.threads_started")
4101 lassert 28 "$msg" '(($tstarted >= $tmin && $tstarted <= $tmax ))' ||
4110 thread_sanity OST ost1 'ost.*.ost' 'oss_num_threads' '16'
4111 cleanup || error "cleanup failed with rc $?"
4113 run_test 53a "check OSS thread count params"
4117 local mds=$(do_facet $SINGLEMDS "$LCTL get_param \
4118 -N mds.*.*.threads_max 2>/dev/null")
4119 if [ -z "$mds" ]; then
4120 #running this on an old MDT
4121 thread_sanity MDT $SINGLEMDS 'mdt.*.*.' 'mdt_num_threads' 16
4123 thread_sanity MDT $SINGLEMDS 'mds.*.*.' 'mds_num_threads' 16
4125 cleanup || error "cleanup failed with $?"
4127 run_test 53b "check MDS thread count params"
4130 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
4131 skip "ldiskfs only test"
4135 do_rpc_nodes $(facet_host ost1) run_llverdev $(ostdevname 1) -p ||
4136 error "llverdev failed with rc=$?"
4139 run_test 54a "test llverdev and partial verify of device"
4142 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
4143 skip "ldiskfs only test"
4148 run_llverfs $MOUNT -p || error "llverfs failed with rc=$?"
4149 cleanup || error "cleanup failed with rc=$?"
4151 run_test 54b "test llverfs and partial verify of filesystem"
4155 local max_ost_index=$1
4156 echo -n $(((max_ost_index + 1) * 8))
4160 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
4161 skip "ldiskfs only test"
4165 local mdsdev=$(mdsdevname 1)
4166 local mdsvdev=$(mdsvdevname 1)
4170 if ! combined_mgs_mds; then
4171 stop_mgs || error "stopping MGS service failed"
4172 format_mgs || error "formatting MGT failed"
4174 add mds1 $(mkfs_opts mds1 ${mdsdev}) --reformat $mdsdev \
4176 add ost1 $(mkfs_opts ost1 $(ostdevname 1)) --index=$i \
4177 --reformat $(ostdevname 1) $(ostvdevname 1)
4183 echo checking size of lov_objid for ost index $i
4184 LOV_OBJID_SIZE=$(do_facet mds1 "$DEBUGFS -R 'stat lov_objid' $mdsdev 2>/dev/null" |
4185 grep ^User | awk -F 'Size: ' '{print $2}')
4186 if [ "$LOV_OBJID_SIZE" != $(lov_objid_size $i) ]; then
4187 error "lov_objid size has to be $(lov_objid_size $i), not $LOV_OBJID_SIZE"
4189 echo ok, lov_objid size is correct: $LOV_OBJID_SIZE
4196 run_test 55 "check lov_objid size"
4199 local server_version=$(lustre_version_code $SINGLEMDS)
4200 local mds_journal_size_orig=$MDSJOURNALSIZE
4206 add ost1 $(mkfs_opts ost1 $(ostdevname 1)) --index=10000 --reformat \
4207 $(ostdevname 1) $(ostvdevname 1)
4208 add ost2 $(mkfs_opts ost2 $(ostdevname 2)) --index=1000 --reformat \
4209 $(ostdevname 2) $(ostvdevname 2)
4212 start_ost || error "Unable to start first ost (idx 10000)"
4213 start_ost2 || error "Unable to start second ost (idx 1000)"
4214 mount_client $MOUNT || error "Unable to mount client"
4218 if [[ $server_version -ge $(version_code 2.6.54) ]] ||
4219 [[ $server_version -ge $(version_code 2.5.4) &&
4220 $server_version -lt $(version_code 2.5.11) ]]; then
4221 wait_osc_import_state mds ost1 FULL
4222 wait_osc_import_state mds ost2 FULL
4223 $SETSTRIPE --stripe-count=-1 $DIR/$tfile ||
4224 error "Unable to setstripe $DIR/$tfile"
4225 n=$($LFS getstripe --stripe-count $DIR/$tfile)
4226 [ "$n" -eq 2 ] || error "Stripe count not two: $n"
4231 MDSJOURNALSIZE=$mds_journal_size_orig
4234 run_test 56a "check big OST indexes and out-of-index-order start"
4239 umount_client $MOUNT -f || error "unmount client failed"
4248 [ $MDSCOUNT -lt 3 ] && skip "needs >= 3 MDTs" && return
4250 trap cleanup_56b EXIT RETURN ERR
4253 if ! combined_mgs_mds ; then
4258 add mds1 $(mkfs_opts mds1 $(mdsdevname 1)) --index=0 --reformat \
4259 $(mdsdevname 1) $(mdsvdevname 1)
4260 add mds2 $(mkfs_opts mds2 $(mdsdevname 2)) --index=1 --reformat \
4261 $(mdsdevname 2) $(mdsvdevname 2)
4262 add mds3 $(mkfs_opts mds3 $(mdsdevname 3)) --index=1000 --reformat \
4263 $(mdsdevname 3) $(mdsvdevname 3)
4267 start_mdt 1 || error "MDT 1 (idx 0) start failed"
4268 start_mdt 2 || error "MDT 2 (idx 1) start failed"
4269 start_mdt 3 || error "MDT 3 (idx 1000) start failed"
4270 start_ost || error "Unable to start first ost"
4271 start_ost2 || error "Unable to start second ost"
4273 do_nodes $(comma_list $(mdts_nodes)) \
4274 "$LCTL set_param mdt.*.enable_remote_dir=1 \
4275 mdt.*.enable_remote_dir_gid=-1"
4277 mount_client $MOUNT || error "Unable to mount client"
4279 $LFS mkdir -c3 $MOUNT/$tdir || error "failed to make testdir"
4281 echo "This is test file 1!" > $MOUNT/$tdir/$tfile.1 ||
4282 error "failed to make test file 1"
4283 echo "This is test file 2!" > $MOUNT/$tdir/$tfile.2 ||
4284 error "failed to make test file 2"
4285 echo "This is test file 1000!" > $MOUNT/$tdir/$tfile.1000 ||
4286 error "failed to make test file 1000"
4288 rm -rf $MOUNT/$tdir || error "failed to remove testdir"
4290 $LFS mkdir -i1000 $MOUNT/$tdir.1000 ||
4291 error "create remote dir at idx 1000 failed"
4294 echo "=== START lfs df OUTPUT ==="
4296 echo "==== END lfs df OUTPUT ===="
4298 mdtcnt=$(echo -e "$output" | grep $FSNAME-MDT | wc -l)
4299 ostcnt=$(echo -e "$output" | grep $FSNAME-OST | wc -l)
4301 echo "lfs df returned mdt count $mdtcnt and ost count $ostcnt"
4302 [ $mdtcnt -eq 3 ] || error "lfs df returned wrong mdt count"
4303 [ $ostcnt -eq 2 ] || error "lfs df returned wrong ost count"
4305 echo "This is test file 1!" > $MOUNT/$tdir.1000/$tfile.1 ||
4306 error "failed to make test file 1"
4307 echo "This is test file 2!" > $MOUNT/$tdir.1000/$tfile.2 ||
4308 error "failed to make test file 2"
4309 echo "This is test file 1000!" > $MOUNT/$tdir.1000/$tfile.1000 ||
4310 error "failed to make test file 1000"
4311 rm -rf $MOUNT/$tdir.1000 || error "failed to remove remote_dir"
4314 echo "=== START lfs mdts OUTPUT ==="
4316 echo "==== END lfs mdts OUTPUT ===="
4318 echo -e "$output" | grep -v "MDTS:" | awk '{print $1}' |
4319 sed 's/://g' > $TMP/mdts-actual.txt
4320 sort $TMP/mdts-actual.txt -o $TMP/mdts-actual.txt
4322 echo -e "0\n1\n1000" > $TMP/mdts-expected.txt
4324 diff $TMP/mdts-expected.txt $TMP/mdts-actual.txt
4327 rm $TMP/mdts-expected.txt $TMP/mdts-actual.txt
4329 [ $result -eq 0 ] || error "target_obd proc file is incorrect!"
4331 run_test 56b "test target_obd correctness with nonconsecutive MDTs"
4333 test_57a() { # bug 22656
4334 do_rpc_nodes $(facet_active_host ost1) load_modules_local
4335 local NID=$(do_facet ost1 "$LCTL get_param nis" |
4336 tail -1 | awk '{print $1}')
4337 writeconf_or_reformat
4338 [ $(facet_fstype ost1) == zfs ] && import_zpool ost1
4339 do_facet ost1 "$TUNEFS --failnode=$NID `ostdevname 1`" ||
4340 error "tunefs failed"
4342 start_ost && error "OST registration from failnode should fail"
4345 run_test 57a "initial registration from failnode should fail (should return errs)"
4348 do_rpc_nodes $(facet_active_host ost1) load_modules_local
4349 local NID=$(do_facet ost1 "$LCTL get_param nis" |
4350 tail -1 | awk '{print $1}')
4351 writeconf_or_reformat
4352 [ $(facet_fstype ost1) == zfs ] && import_zpool ost1
4353 do_facet ost1 "$TUNEFS --servicenode=$NID `ostdevname 1`" ||
4354 error "tunefs failed"
4356 start_ost || error "OST registration from servicenode should not fail"
4359 run_test 57b "initial registration from servicenode should not fail"
4362 do_facet mgs $LCTL get_param mgs.MGS.live.$FSNAME | grep OST | wc -l
4365 test_58() { # bug 22658
4366 combined_mgs_mds || stop_mgs || error "stopping MGS service failed"
4368 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
4369 createmany -o $DIR/$tdir/$tfile-%d 100
4370 # make sure that OSTs do not cancel llog cookies before we unmount the MDS
4371 #define OBD_FAIL_OBD_LOG_CANCEL_NET 0x601
4372 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0x601"
4373 unlinkmany $DIR/$tdir/$tfile-%d 100
4374 stop_mds || error "Unable to stop MDS"
4376 local MNTDIR=$(facet_mntpt $SINGLEMDS)
4377 local devname=$(mdsdevname ${SINGLEMDS//mds/})
4379 # remove all files from the OBJECTS dir
4380 mount_fstype $SINGLEMDS
4382 do_facet $SINGLEMDS "find $MNTDIR/O/1/d* -type f -delete"
4384 unmount_fstype $SINGLEMDS
4385 # restart MDS with missing llog files
4386 start_mds || error "unable to start MDS"
4387 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0"
4390 run_test 58 "missing llog files must not prevent MDT from mounting"
4393 start_mgsmds >> /dev/null
4394 local C1=$(count_osts)
4395 if [ $C1 -eq 0 ]; then
4396 start_ost >> /dev/null
4400 echo "original ost count: $C1 (expect > 0)"
4401 [ $C1 -gt 0 ] || error "No OSTs in $FSNAME log"
4402 start_mgsmds -o writeconf >> /dev/null || error "MDT start failed"
4403 local C2=$(count_osts)
4404 echo "after mdt writeconf count: $C2 (expect 0)"
4405 [ $C2 -gt 0 ] && error "MDT writeconf should erase OST logs"
4406 echo "OST start without writeconf should fail:"
4407 start_ost >> /dev/null &&
4408 error "OST start without writeconf didn't fail"
4409 echo "OST start with writeconf should succeed:"
4410 start_ost -o writeconf >> /dev/null || error "OST1 start failed"
4411 local C3=$(count_osts)
4412 echo "after ost writeconf count: $C3 (expect 1)"
4413 [ $C3 -eq 1 ] || error "new OST writeconf should add:"
4414 start_ost2 -o writeconf >> /dev/null || error "OST2 start failed"
4415 local C4=$(count_osts)
4416 echo "after ost2 writeconf count: $C4 (expect 2)"
4417 [ $C4 -eq 2 ] || error "OST2 writeconf should add log"
4418 stop_ost2 >> /dev/null
4419 cleanup_nocli >> /dev/null
4420 #writeconf to remove all ost2 traces for subsequent tests
4421 writeconf_or_reformat
4423 run_test 59 "writeconf mount option"
4425 test_60() { # LU-471
4428 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
4429 skip "ldiskfs only test"
4433 for num in $(seq $MDSCOUNT); do
4434 add mds${num} $(mkfs_opts mds${num} $(mdsdevname $num)) \
4435 --mkfsoptions='\" -E stride=64 -O ^uninit_bg\"' \
4436 --reformat $(mdsdevname $num) $(mdsvdevname $num) ||
4440 dump=$(do_facet $SINGLEMDS dumpe2fs $(mdsdevname 1))
4441 [ ${PIPESTATUS[0]} -eq 0 ] || error "dumpe2fs $(mdsdevname 1) failed"
4443 # MDT default has dirdata feature
4444 echo $dump | grep dirdata > /dev/null || error "dirdata is not set"
4445 # we disable uninit_bg feature
4446 echo $dump | grep uninit_bg > /dev/null && error "uninit_bg is set"
4447 # we set stride extended options
4448 echo $dump | grep stride > /dev/null || error "stride is not set"
4451 run_test 60 "check mkfs.lustre --mkfsoptions -E -O options setting"
4456 [ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.1.53) ] ||
4457 { skip "Need MDS version at least 2.1.53"; return 0; }
4459 if [ $(facet_fstype $SINGLEMDS) == ldiskfs ] &&
4460 ! large_xattr_enabled; then
4463 for num in $(seq $MDSCOUNT); do
4464 do_facet mds${num} $TUNE2FS -O ea_inode \
4465 $(mdsdevname $num) ||
4466 error "tune2fs on mds $num failed"
4470 combined_mgs_mds || stop_mgs || error "stopping MGS service failed"
4471 setup_noconfig || error "setting up the filesystem failed"
4472 client_up || error "starting client failed"
4474 local file=$DIR/$tfile
4475 touch $file || error "touch $file failed"
4477 local large_value="$(generate_string $(max_xattr_size))"
4478 local small_value="bar"
4480 local name="trusted.big"
4481 log "save large xattr $name on $file"
4482 setfattr -n $name -v $large_value $file ||
4483 error "saving $name on $file failed"
4485 local new_value=$(get_xattr_value $name $file)
4486 [[ "$new_value" != "$large_value" ]] &&
4487 error "$name different after saving"
4489 log "shrink value of $name on $file"
4490 setfattr -n $name -v $small_value $file ||
4491 error "shrinking value of $name on $file failed"
4493 new_value=$(get_xattr_value $name $file)
4494 [[ "$new_value" != "$small_value" ]] &&
4495 error "$name different after shrinking"
4497 log "grow value of $name on $file"
4498 setfattr -n $name -v $large_value $file ||
4499 error "growing value of $name on $file failed"
4501 new_value=$(get_xattr_value $name $file)
4502 [[ "$new_value" != "$large_value" ]] &&
4503 error "$name different after growing"
4505 log "check value of $name on $file after remounting MDS"
4507 new_value=$(get_xattr_value $name $file)
4508 [[ "$new_value" != "$large_value" ]] &&
4509 error "$name different after remounting MDS"
4511 log "remove large xattr $name from $file"
4512 setfattr -x $name $file || error "removing $name from $file failed"
4515 stopall || error "stopping for e2fsck run"
4516 for num in $(seq $MDSCOUNT); do
4517 run_e2fsck $(facet_active_host mds$num) \
4518 $(mdsdevname $num) "-y" ||
4519 error "e2fsck MDT$num failed"
4521 setup_noconfig || error "remounting the filesystem failed"
4524 # need to delete this file to avoid problems in other tests
4526 stopall || error "stopping systems failed"
4528 run_test 61 "large xattr"
4531 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
4532 skip "ldiskfs only test"
4537 local mdsdev=$(mdsdevname 1)
4538 local ostdev=$(ostdevname 1)
4540 [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.2.51) ]] ||
4541 { skip "Need MDS version at least 2.2.51"; return 0; }
4543 echo "disable journal for mds"
4544 do_facet mds1 $TUNE2FS -O ^has_journal $mdsdev || error "tune2fs failed"
4545 start_mds && error "MDT start should fail"
4546 echo "disable journal for ost"
4547 do_facet ost1 $TUNE2FS -O ^has_journal $ostdev || error "tune2fs failed"
4548 start_ost && error "OST start should fail"
4549 cleanup || error "cleanup failed with rc $?"
4552 run_test 62 "start with disabled journal"
4555 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
4556 skip "ldiskfs only test"
4560 do_rpc_nodes $(facet_active_host $SINGLEMDS) load_module ldiskfs
4561 local inode_slab=$(do_facet $SINGLEMDS "cat /proc/slabinfo" |
4562 awk '/ldiskfs_inode_cache/ { print $5 / $6 }')
4563 if [ -z "$inode_slab" ]; then
4564 skip "ldiskfs module has not been loaded"
4568 echo "$inode_slab ldiskfs inodes per page"
4569 [ "${inode_slab%.*}" -ge "3" ] && return 0
4571 # If kmalloc-128 is also 1 per page - this is a debug kernel
4572 # and so this is not an error.
4573 local kmalloc128=$(do_facet $SINGLEMDS "cat /proc/slabinfo" |
4574 awk '/^(kmalloc|size)-128 / { print $5 / $6 }')
4575 # 32 128-byte chunks in 4k
4576 [ "${kmalloc128%.*}" -lt "32" ] ||
4577 error "ldiskfs inode too big, only $inode_slab objs/page, " \
4578 "kmalloc128 = $kmalloc128 objs/page"
4580 run_test 63 "Verify each page can at least hold 3 ldiskfs inodes"
4583 start_mds || error "unable to start MDS"
4584 start_ost || error "Unable to start OST1"
4585 start_ost2 || error "Unable to start second ost"
4586 mount_client $MOUNT || error "Unable to mount client"
4587 stop_ost2 || error "Unable to stop second ost"
4590 umount_client $MOUNT -f || error “unmount $MOUNT failed”
4591 cleanup_nocli || error "cleanup_nocli failed with $?"
4592 #writeconf to remove all ost2 traces for subsequent tests
4593 writeconf_or_reformat
4595 run_test 64 "check lfs df --lazy "
4597 test_65() { # LU-2237
4598 # Currently, the test is only valid for ldiskfs backend
4599 [ "$(facet_fstype $SINGLEMDS)" != "ldiskfs" ] &&
4600 skip "ldiskfs only test" && return
4602 local devname=$(mdsdevname ${SINGLEMDS//mds/})
4603 local brpt=$(facet_mntpt brpt)
4606 if ! do_facet $SINGLEMDS "test -b $devname"; then
4610 stop_mds || error "Unable to stop MDS"
4611 local obj=$(do_facet $SINGLEMDS \
4612 "$DEBUGFS -c -R \\\"stat last_rcvd\\\" $devname" |
4614 if [ -z "$obj" ]; then
4615 # The MDT may be just re-formatted, mount the MDT for the
4616 # first time to guarantee the "last_rcvd" file is there.
4617 start_mds || error "fail to mount the MDS for the first time"
4618 stop_mds || error "Unable to stop MDS"
4621 # remove the "last_rcvd" file
4622 do_facet $SINGLEMDS "mkdir -p $brpt"
4623 do_facet $SINGLEMDS \
4624 "mount -t $(facet_fstype $SINGLEMDS) $opts $devname $brpt"
4625 do_facet $SINGLEMDS "rm -f ${brpt}/last_rcvd"
4626 do_facet $SINGLEMDS "$UMOUNT $brpt"
4628 # restart MDS, the "last_rcvd" file should be recreated.
4629 start_mds || error "fail to restart the MDS"
4630 stop_mds || error "Unable to stop MDS"
4631 obj=$(do_facet $SINGLEMDS \
4632 "$DEBUGFS -c -R \\\"stat last_rcvd\\\" $devname" | grep Inode)
4633 [ -n "$obj" ] || error "fail to re-create the last_rcvd"
4635 run_test 65 "re-create the lost last_rcvd file when server mount"
4638 [[ $(lustre_version_code mgs) -ge $(version_code 2.3.59) ]] ||
4639 { skip "Need MGS version at least 2.3.59"; return 0; }
4642 local OST1_NID=$(do_facet ost1 $LCTL list_nids | head -1)
4643 local MDS_NID=$(do_facet $SINGLEMDS $LCTL list_nids | head -1)
4645 # add EXCLUDE records to config log, they are not to be
4646 # removed by lctl replace_nids
4647 set_conf_param_and_check mds \
4648 "$LCTL get_param -n osc.$FSNAME-OST0000-osc-MDT0000.active" \
4649 "$FSNAME-OST0000.osc.active" \
4652 echo "replace_nids should fail if MDS, OSTs and clients are UP"
4653 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 $OST1_NID &&
4654 error "replace_nids fail"
4656 umount_client $MOUNT || error "unmounting client failed"
4657 echo "replace_nids should fail if MDS and OSTs are UP"
4658 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 $OST1_NID &&
4659 error "replace_nids fail"
4661 stop_ost || error "Unable to stop OST1"
4662 echo "replace_nids should fail if MDS is UP"
4663 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 $OST1_NID &&
4664 error "replace_nids fail"
4666 stop_mds || error "stopping mds failed"
4668 if combined_mgs_mds; then
4669 start_mdt 1 "-o nosvc" ||
4670 error "starting mds with nosvc option failed"
4673 echo "command should accept two parameters"
4674 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 &&
4675 error "command should accept two params"
4677 echo "correct device name should be passed"
4678 do_facet mgs $LCTL replace_nids $FSNAME-WRONG0000 $OST1_NID &&
4679 error "wrong devname"
4681 echo "wrong nids list should not destroy the system"
4682 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 "wrong nids list" &&
4685 echo "replace OST nid"
4686 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 $OST1_NID ||
4687 error "replace nids failed"
4689 echo "command should accept two parameters"
4690 do_facet mgs $LCTL replace_nids $FSNAME-MDT0000 &&
4691 error "command should accept two params"
4693 echo "wrong nids list should not destroy the system"
4694 do_facet mgs $LCTL replace_nids $FSNAME-MDT0000 "wrong nids list" &&
4697 echo "replace MDS nid"
4698 do_facet mgs $LCTL replace_nids $FSNAME-MDT0000 $MDS_NID ||
4699 error "replace nids failed"
4701 if ! combined_mgs_mds ; then
4704 stop_mds || error "Unable to stop MDS"
4707 start_mgsmds || error "start mgsmds failed"
4708 set_conf_param_and_check mds \
4709 "$LCTL get_param -n osc.$FSNAME-OST0000-osc-MDT0000.active" \
4710 "$FSNAME-OST0000.osc.active" \
4712 start_ost || error "unable to start OST"
4713 mount_client $MOUNT || error "mount client failed"
4715 check_mount || error "error after nid replace"
4716 cleanup || error "cleanup failed"
4719 run_test 66 "replace nids"
4721 test_67() { #LU-2950
4722 local legacy="$TMP/legacy_lnet_config"
4723 local new="$TMP/new_routes_test"
4724 local out="$TMP/config_out_file"
4725 local verify="$TMP/conv_verify"
4726 local verify_conf="$TMP/conf_verify"
4728 # Create the legacy file that will be run through the
4729 # lustre_routes_conversion script
4730 cat <<- LEGACY_LNET_CONFIG > $legacy
4731 tcp1 23 192.168.213.1@tcp:1; tcp5 34 193.30.4.3@tcp:4;
4732 tcp2 54 10.1.3.2@tcp;
4733 tcp3 10.3.4.3@tcp:3;
4737 # Create the verification file to verify the output of
4738 # lustre_routes_conversion script against.
4739 cat <<- VERIFY_LNET_CONFIG > $verify
4740 tcp1: { gateway: 192.168.213.1@tcp, hop: 23, priority: 1 }
4741 tcp5: { gateway: 193.30.4.3@tcp, hop: 34, priority: 4 }
4742 tcp2: { gateway: 10.1.3.2@tcp, hop: 54 }
4743 tcp3: { gateway: 10.3.4.3@tcp, priority: 3 }
4744 tcp4: { gateway: 10.3.3.4@tcp }
4747 # Create the verification file to verify the output of
4748 # lustre_routes_config script against
4749 cat <<- VERIFY_LNET_CONFIG > $verify_conf
4750 lctl --net tcp1 add_route 192.168.213.1@tcp 23 1
4751 lctl --net tcp5 add_route 193.30.4.3@tcp 34 4
4752 lctl --net tcp2 add_route 10.1.3.2@tcp 54 4
4753 lctl --net tcp3 add_route 10.3.4.3@tcp 1 3
4754 lctl --net tcp4 add_route 10.3.3.4@tcp 1 3
4757 lustre_routes_conversion $legacy $new > /dev/null
4758 if [ -f $new ]; then
4759 # verify the conversion output
4760 cmp -s $new $verify > /dev/null
4761 if [ $? -eq 1 ]; then
4762 error "routes conversion failed"
4765 lustre_routes_config --dry-run --verbose $new > $out
4766 # check that the script succeeded
4767 cmp -s $out $verify_conf > /dev/null
4768 if [ $? -eq 1 ]; then
4769 error "routes config failed"
4772 error "routes conversion test failed"
4774 # remove generated files
4775 rm -f $new $legacy $verify $verify_conf $out
4777 run_test 67 "test routes conversion and configuration"
4785 [ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.4.53) ] ||
4786 { skip "Need MDS version at least 2.4.53"; return 0; }
4788 umount_client $MOUNT || error "umount client failed"
4790 if ! combined_mgs_mds; then
4791 start_mgs || error "start mgs failed"
4794 start_mdt 1 || error "MDT start failed"
4795 start_ost || error "Unable to start OST1"
4797 # START-END - the sequences we'll be reserving
4798 START=$(do_facet $SINGLEMDS \
4799 $LCTL get_param -n seq.ctl*.space | awk -F'[[ ]' '{print $2}')
4800 END=$((START + (1 << 30)))
4801 do_facet $SINGLEMDS \
4802 $LCTL set_param seq.ctl*.fldb="[$START-$END\):0:mdt"
4804 # reset the sequences MDT0000 has already assigned
4805 do_facet $SINGLEMDS \
4806 $LCTL set_param seq.srv*MDT0000.space=clear
4808 # remount to let the client allocate new sequence
4809 mount_client $MOUNT || error "mount client failed"
4811 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
4812 do_facet $SINGLEMDS \
4813 $LCTL get_param seq.srv*MDT0000.space
4814 $LFS path2fid $DIR/$tfile
4816 local old_ifs="$IFS"
4818 fid=($($LFS path2fid $DIR/$tfile))
4822 if [[ $seq < $END ]]; then
4823 error "used reserved sequence $seq?"
4825 cleanup || error "cleanup failed with $?"
4827 run_test 68 "be able to reserve specific sequences in FLDB"
4829 # Test 69: is about the total number of objects ever created on an OST.
4830 # so that when it is reformatted the normal MDS->OST orphan recovery won't
4831 # just "precreate" the missing objects. In the past it might try to recreate
4832 # millions of objects after an OST was reformatted
4834 local server_version=$(lustre_version_code $SINGLEMDS)
4836 [[ $server_version -lt $(version_code 2.4.2) ]] &&
4837 skip "Need MDS version at least 2.4.2" && return
4839 [[ $server_version -ge $(version_code 2.4.50) ]] &&
4840 [[ $server_version -lt $(version_code 2.5.0) ]] &&
4841 skip "Need MDS version at least 2.5.0" && return
4844 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
4846 # use OST0000 since it probably has the most creations
4847 local OSTNAME=$(ostname_from_index 0)
4848 local mdtosc_proc1=$(get_mdtosc_proc_path mds1 $OSTNAME)
4849 local last_id=$(do_facet mds1 $LCTL get_param -n \
4850 osc.$mdtosc_proc1.prealloc_last_id)
4852 # Want to have OST LAST_ID over 5 * OST_MAX_PRECREATE to
4853 # verify that the LAST_ID recovery is working properly. If
4854 # not, then the OST will refuse to allow the MDS connect
4855 # because the LAST_ID value is too different from the MDS
4856 #define OST_MAX_PRECREATE=20000
4857 local ost_max_pre=20000
4858 local num_create=$(( ost_max_pre * 5 + 1 - last_id))
4860 # If the LAST_ID is already over 5 * OST_MAX_PRECREATE, we don't
4861 # need to create any files. So, skip this section.
4862 if [ $num_create -gt 0 ]; then
4863 # Check the number of inodes available on OST0
4865 local ifree=$($LFS df -i $MOUNT | awk '/OST0000/ { print $4 }')
4866 log "On OST0, $ifree inodes available. Want $num_create."
4868 $SETSTRIPE -i 0 $DIR/$tdir ||
4869 error "$SETSTRIPE -i 0 $DIR/$tdir failed"
4870 if [ $ifree -lt 10000 ]; then
4871 files=$(( ifree - 50 ))
4876 local j=$((num_create / files + 1))
4877 for i in $(seq 1 $j); do
4878 createmany -o $DIR/$tdir/$tfile-$i- $files ||
4879 error "createmany fail create $files files: $?"
4880 unlinkmany $DIR/$tdir/$tfile-$i- $files ||
4881 error "unlinkmany failed unlink $files files"
4885 # delete all of the files with objects on OST0 so the
4886 # filesystem is not inconsistent later on
4887 $LFS find $MOUNT --ost 0 -print0 | xargs -0 rm
4889 umount_client $MOUNT || error "umount client failed"
4890 stop_ost || error "OST0 stop failure"
4891 add ost1 $(mkfs_opts ost1 $(ostdevname 1)) --reformat --replace \
4892 $(ostdevname 1) $(ostvdevname 1) ||
4893 error "reformat and replace $ostdev failed"
4894 start_ost || error "OST0 restart failure"
4895 wait_osc_import_state mds ost FULL
4897 mount_client $MOUNT || error "mount client failed"
4898 touch $DIR/$tdir/$tfile-last || error "create file after reformat"
4899 local idx=$($GETSTRIPE -i $DIR/$tdir/$tfile-last)
4900 [ $idx -ne 0 ] && error "$DIR/$tdir/$tfile-last on $idx not 0" || true
4902 local iused=$($LFS df -i $MOUNT | awk '/OST0000/ { print $3 }')
4903 log "On OST0, $iused used inodes"
4904 [ $iused -ge $((ost_max_pre/2 + 1000)) ] &&
4905 error "OST replacement created too many inodes; $iused"
4906 cleanup || error "cleanup failed with $?"
4908 run_test 69 "replace an OST with the same index"
4911 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
4914 cleanup || error "cleanup failed with $?"
4916 start_mdt 1 || error "MDT0 start fail"
4918 start_ost || error "OST0 start fail"
4919 for num in $(seq 2 $MDSCOUNT); do
4920 start_mdt $num || return
4923 mount_client $MOUNT || error "mount client fails"
4925 mkdir $DIR/$tdir || error "create $DIR/$tdir failed"
4927 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
4928 error "create remote dir fail"
4930 rm -rf $DIR/$tdir || error "delete dir fail"
4931 cleanup || error "cleanup failed with $?"
4933 run_test 70a "start MDT0, then OST, then MDT1"
4936 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
4939 start_ost || error "OST0 start fail"
4941 start_mds || error "MDS start fail"
4943 mount_client $MOUNT || error "mount client fails"
4945 mkdir $DIR/$tdir || error "create $DIR/$tdir failed"
4947 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
4948 error "create remote dir fail"
4950 rm -rf $DIR/$tdir || error "delete dir fail"
4952 cleanup || error "cleanup failed with $?"
4954 run_test 70b "start OST, MDT1, MDT0"
4957 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
4960 start_mds || error "MDS start fail"
4961 start_ost || error "OST0 start fail"
4963 mount_client $MOUNT || error "mount client fails"
4964 stop_mdt 1 || error "MDT1 start fail"
4966 local mdc_for_mdt1=$($LCTL dl | grep MDT0000-mdc | awk '{print $4}')
4967 echo "deactivate $mdc_for_mdt1"
4968 $LCTL --device $mdc_for_mdt1 deactivate ||
4969 error "set $mdc_for_mdt1 deactivate failed"
4971 mkdir $DIR/$tdir && error "mkdir succeed"
4973 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir &&
4974 error "create remote dir succeed"
4976 cleanup || error "cleanup failed with $?"
4978 run_test 70c "stop MDT0, mkdir fail, create remote dir fail"
4981 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
4984 start_mds || error "MDS start fail"
4985 start_ost || error "OST0 start fail"
4987 mount_client $MOUNT || error "mount client fails"
4989 stop_mdt 2 || error "MDT1 start fail"
4991 local mdc_for_mdt2=$($LCTL dl | grep MDT0001-mdc |
4993 echo "deactivate $mdc_for_mdt2"
4994 $LCTL --device $mdc_for_mdt2 deactivate ||
4995 error "set $mdc_for_mdt2 deactivate failed"
4997 mkdir $DIR/$tdir || error "mkdir fail"
4998 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir &&
4999 error "create remote dir succeed"
5001 rm -rf $DIR/$tdir || error "delete dir fail"
5003 cleanup || error "cleanup failed with $?"
5005 run_test 70d "stop MDT1, mkdir succeed, create remote dir fail"
5008 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
5010 [ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.7.62) ] ||
5011 { skip "Need MDS version at least 2.7.62"; return 0; }
5013 cleanup || error "cleanup failed with $?"
5015 local mdsdev=$(mdsdevname 1)
5016 local ostdev=$(ostdevname 1)
5017 local mdsvdev=$(mdsvdevname 1)
5018 local ostvdev=$(ostvdevname 1)
5019 local opts_mds="$(mkfs_opts mds1 $mdsdev) --reformat $mdsdev $mdsvdev"
5020 local opts_ost="$(mkfs_opts ost1 $ostdev) --reformat $ostdev $ostvdev"
5022 add mds1 $opts_mds || error "add mds1 failed"
5023 start_mdt 1 || error "start mdt1 failed"
5024 add ost1 $opts_ost || error "add ost1 failed"
5025 start_ost || error "start ost failed"
5026 mount_client $MOUNT > /dev/null || error "mount client $MOUNT failed"
5028 local soc=$(do_facet mds1 "$LCTL get_param -n \
5029 mdt.*MDT0000.sync_lock_cancel")
5030 [ $soc == "never" ] || error "SoC enabled on single MDS"
5032 for i in $(seq 2 $MDSCOUNT); do
5033 mdsdev=$(mdsdevname $i)
5034 mdsvdev=$(mdsvdevname $i)
5035 opts_mds="$(mkfs_opts mds$i $mdsdev) --reformat $mdsdev \
5037 add mds$i $opts_mds || error "add mds$i failed"
5038 start_mdt $i || error "start mdt$i fail"
5041 wait_dne_interconnect
5043 for i in $(seq $MDSCOUNT); do
5044 soc=$(do_facet mds$i "$LCTL get_param -n \
5045 mdt.*MDT000$((i - 1)).sync_lock_cancel")
5046 [ $soc == "blocking" ] || error "SoC not enabled on DNE"
5049 for i in $(seq 2 $MDSCOUNT); do
5050 stop_mdt $i || error "stop mdt$i fail"
5052 soc=$(do_facet mds1 "$LCTL get_param -n \
5053 mdt.*MDT0000.sync_lock_cancel")
5054 [ $soc == "never" ] || error "SoC enabled on single MDS"
5055 umount_client $MOUNT -f > /dev/null
5057 cleanup || error "cleanup failed with $?"
5059 run_test 70e "Sync-on-Cancel will be enabled by default on DNE"
5062 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
5063 if combined_mgs_mds; then
5064 skip "needs separate MGS/MDT" && return
5068 start_mdt 1 || error "MDT0 start fail"
5069 start_ost || error "OST0 start fail"
5070 for num in $(seq 2 $MDSCOUNT); do
5071 start_mdt $num || return
5074 start_ost2 || error "OST1 start fail"
5076 mount_client $MOUNT || error "mount client fails"
5078 mkdir $DIR/$tdir || error "mkdir fail"
5079 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
5080 error "create remote dir succeed"
5082 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
5083 rm -rf $DIR/$tdir || error "delete dir fail"
5085 umount_client $MOUNT || error "umount_client failed"
5086 stop_mds || error "MDS stop fail"
5087 stop_ost || error "OST0 stop fail"
5088 stop_ost2 || error "OST1 stop fail"
5090 run_test 71a "start MDT0 OST0, MDT1, OST1"
5093 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
5094 if combined_mgs_mds; then
5095 skip "needs separate MGS/MDT" && return
5099 for num in $(seq 2 $MDSCOUNT); do
5100 start_mdt $num || return
5102 start_ost || error "OST0 start fail"
5103 start_mdt 1 || error "MDT0 start fail"
5104 start_ost2 || error "OST1 start fail"
5106 mount_client $MOUNT || error "mount client fails"
5108 mkdir $DIR/$tdir || error "mkdir fail"
5109 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
5110 error "create remote dir succeed"
5112 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
5113 rm -rf $DIR/$tdir || error "delete dir fail"
5115 umount_client $MOUNT || error "umount_client failed"
5116 stop_mds || error "MDT0 stop fail"
5117 stop_ost || error "OST0 stop fail"
5118 stop_ost2 || error "OST1 stop fail"
5120 run_test 71b "start MDT1, OST0, MDT0, OST1"
5123 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
5124 if combined_mgs_mds; then
5125 skip "needs separate MGS/MDT" && return
5129 start_ost || error "OST0 start fail"
5130 start_ost2 || error "OST1 start fail"
5131 for num in $(seq 2 $MDSCOUNT); do
5132 start_mdt $num || return
5134 start_mdt 1 || error "MDT0 start fail"
5136 mount_client $MOUNT || error "mount client fails"
5138 mkdir $DIR/$tdir || error "mkdir fail"
5139 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
5140 error "create remote dir succeed"
5142 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
5143 rm -rf $DIR/$tdir || error "delete dir fail"
5145 umount_client $MOUNT || error "umount_client failed"
5146 stop_mds || error "MDS stop fail"
5147 stop_ost || error "OST0 stop fail"
5148 stop_ost2 || error "OST1 stop fail"
5151 run_test 71c "start OST0, OST1, MDT1, MDT0"
5154 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
5155 if combined_mgs_mds; then
5156 skip "needs separate MGS/MDT" && return
5160 start_ost || error "OST0 start fail"
5161 for num in $(seq 2 $MDSCOUNT); do
5162 start_mdt $num || return
5164 start_mdt 1 || error "MDT0 start fail"
5165 start_ost2 || error "OST1 start fail"
5167 mount_client $MOUNT || error "mount client fails"
5169 mkdir $DIR/$tdir || error "mkdir fail"
5170 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
5171 error "create remote dir succeed"
5173 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
5174 rm -rf $DIR/$tdir || error "delete dir fail"
5176 umount_client $MOUNT || error "umount_client failed"
5177 stop_mds || error "MDS stop fail"
5178 stop_ost || error "OST0 stop fail"
5179 stop_ost2 || error "OST1 stop fail"
5182 run_test 71d "start OST0, MDT1, MDT0, OST1"
5185 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
5186 if combined_mgs_mds; then
5187 skip "needs separate MGS/MDT" && return
5191 start_ost || error "OST0 start fail"
5192 for num in $(seq 2 $MDSCOUNT); do
5193 start_mdt $num || return
5195 start_ost2 || error "OST1 start fail"
5196 start_mdt 1 || error "MDT0 start fail"
5198 mount_client $MOUNT || error "mount client fails"
5200 mkdir $DIR/$tdir || error "mkdir fail"
5201 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
5202 error "create remote dir succeed"
5204 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
5205 rm -rf $DIR/$tdir || error "delete dir fail"
5207 umount_client $MOUNT || error "umount_client failed"
5208 stop_mds || error "MDS stop fail"
5209 stop_ost || error "OST0 stop fail"
5210 stop_ost2 || error "OST1 stop fail"
5213 run_test 71e "start OST0, MDT1, OST1, MDT0"
5215 test_72() { #LU-2634
5216 local mdsdev=$(mdsdevname 1)
5217 local ostdev=$(ostdevname 1)
5218 local cmd="$E2FSCK -fnvd $mdsdev"
5222 [ "$(facet_fstype $SINGLEMDS)" != "ldiskfs" ] &&
5223 skip "ldiskfs only test" && return
5225 if combined_mgs_mds; then
5226 add_options='--reformat'
5228 add_options='--reformat --replace'
5231 #tune MDT with "-O extents"
5233 for num in $(seq $MDSCOUNT); do
5234 add mds${num} $(mkfs_opts mds$num $(mdsdevname $num)) \
5235 $add_options $(mdsdevname $num) $(mdsvdevname $num) ||
5236 error "add mds $num failed"
5237 do_facet mds${num} "$TUNE2FS -O extents $(mdsdevname $num)" ||
5238 error "$TUNE2FS failed on mds${num}"
5241 add ost1 $(mkfs_opts ost1 $ostdev) $add_options $ostdev ||
5242 error "add $ostdev failed"
5243 start_mds || error "start mds failed"
5244 start_ost || error "start ost failed"
5245 mount_client $MOUNT || error "mount client failed"
5247 #create some short symlinks
5248 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
5249 createmany -o $DIR/$tdir/$tfile-%d $fn
5250 echo "create $fn short symlinks"
5251 for i in $(seq -w 1 $fn); do
5252 ln -s $DIR/$tdir/$tfile-$i $MOUNT/$tfile-$i
5257 umount_client $MOUNT || error "umount client failed"
5258 stop_mds || error "stop mds failed"
5259 stop_ost || error "stop ost failed"
5262 run_e2fsck $(facet_active_host $SINGLEMDS) $mdsdev "-n"
5264 run_test 72 "test fast symlink with extents flag enabled"
5266 test_73() { #LU-3006
5267 [ $(facet_fstype ost1) == zfs ] && import_zpool ost1
5268 do_facet ost1 "$TUNEFS --failnode=1.2.3.4@$NETTYPE $(ostdevname 1)" ||
5269 error "1st tunefs failed"
5270 start_mgsmds || error "start mds failed"
5271 start_ost || error "start ost failed"
5272 mount_client $MOUNT || error "mount client failed"
5273 $LCTL get_param -n osc.*OST0000-osc-[^M]*.import | grep failover_nids |
5274 grep 1.2.3.4@$NETTYPE || error "failover nids haven't changed"
5275 umount_client $MOUNT || error "umount client failed"
5279 run_test 73 "failnode to update from mountdata properly"
5281 test_75() { # LU-2374
5282 [[ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.4.1) ]] &&
5283 skip "Need MDS version at least 2.4.1" && return
5286 local opts_mds="$(mkfs_opts mds1 $(mdsdevname 1)) \
5287 --reformat $(mdsdevname 1) $(mdsvdevname 1)"
5288 local opts_ost="$(mkfs_opts ost1 $(ostdevname 1)) \
5289 --reformat $(ostdevname 1) $(ostvdevname 1)"
5291 #check with default parameters
5292 add mds1 $opts_mds || error "add mds1 failed for default params"
5293 add ost1 $opts_ost || error "add ost1 failed for default params"
5295 opts_mds=$(echo $opts_mds | sed -e "s/--mdt//")
5296 opts_mds=$(echo $opts_mds |
5297 sed -e "s/--index=$index/--index=$index --mdt/")
5298 opts_ost=$(echo $opts_ost | sed -e "s/--ost//")
5299 opts_ost=$(echo $opts_ost |
5300 sed -e "s/--index=$index/--index=$index --ost/")
5302 add mds1 $opts_mds || error "add mds1 failed for new params"
5303 add ost1 $opts_ost || error "add ost1 failed for new params"
5304 if ! combined_mgs_mds; then
5305 stop_mgs || error "stop mgs failed"
5310 run_test 75 "The order of --index should be irrelevant"
5313 [[ $(lustre_version_code mgs) -ge $(version_code 2.4.52) ]] ||
5314 { skip "Need MDS version at least 2.4.52" && return 0; }
5316 if ! combined_mgs_mds; then
5317 start_mgs || error "start mgs failed"
5320 local MDMB_PARAM="osc.*.max_dirty_mb"
5321 echo "Change MGS params"
5322 local MAX_DIRTY_MB=$($LCTL get_param -n $MDMB_PARAM |
5324 echo "max_dirty_mb: $MAX_DIRTY_MB"
5325 local NEW_MAX_DIRTY_MB=$((MAX_DIRTY_MB + MAX_DIRTY_MB))
5326 echo "new_max_dirty_mb: $NEW_MAX_DIRTY_MB"
5327 do_facet mgs $LCTL set_param -P $MDMB_PARAM=$NEW_MAX_DIRTY_MB
5328 wait_update $HOSTNAME "$LCTL get_param -n $MDMB_PARAM |
5329 head -1" $NEW_MAX_DIRTY_MB
5330 MAX_DIRTY_MB=$($LCTL get_param -n $MDMB_PARAM | head -1)
5331 echo "$MAX_DIRTY_MB"
5332 [ $MAX_DIRTY_MB = $NEW_MAX_DIRTY_MB ] ||
5333 error "error while apply max_dirty_mb"
5335 echo "Check the value is stored after remount"
5338 wait_update $HOSTNAME "$LCTL get_param -n $MDMB_PARAM |
5339 head -1" $NEW_MAX_DIRTY_MB
5340 MAX_DIRTY_MB=$($LCTL get_param -n $MDMB_PARAM | head -1)
5341 [ $MAX_DIRTY_MB = $NEW_MAX_DIRTY_MB ] ||
5342 error "max_dirty_mb is not saved after remount"
5344 echo "Change OST params"
5345 CLIENT_PARAM="obdfilter.*.client_cache_count"
5346 local CLIENT_CACHE_COUNT
5347 CLIENT_CACHE_COUNT=$(do_facet ost1 $LCTL get_param -n $CLIENT_PARAM |
5349 echo "client_cache_count: $CLIENT_CACHE_COUNT"
5350 NEW_CLIENT_CACHE_COUNT=$((CLIENT_CACHE_COUNT+CLIENT_CACHE_COUNT))
5351 echo "new_client_cache_count: $NEW_CLIENT_CACHE_COUNT"
5352 do_facet mgs $LCTL set_param -P $CLIENT_PARAM=$NEW_CLIENT_CACHE_COUNT
5353 wait_update $(facet_host ost1) "$LCTL get_param -n $CLIENT_PARAM |
5354 head -1" $NEW_CLIENT_CACHE_COUNT
5355 CLIENT_CACHE_COUNT=$(do_facet ost1 $LCTL get_param -n $CLIENT_PARAM |
5357 echo "$CLIENT_CACHE_COUNT"
5358 [ $CLIENT_CACHE_COUNT = $NEW_CLIENT_CACHE_COUNT ] ||
5359 error "error while apply client_cache_count"
5361 echo "Check the value is stored after remount"
5364 wait_update $(facet_host ost1) "$LCTL get_param -n $CLIENT_PARAM |
5365 head -1" $NEW_CLIENT_CACHE_COUNT
5366 CLIENT_CACHE_COUNT=$(do_facet ost1 $LCTL get_param -n $CLIENT_PARAM |
5368 echo "$CLIENT_CACHE_COUNT"
5369 [ $CLIENT_CACHE_COUNT = $NEW_CLIENT_CACHE_COUNT ] ||
5370 error "client_cache_count is not saved after remount"
5373 run_test 76a "set permanent params set_param -P"
5375 test_76b() { # LU-4783
5376 [[ $(lustre_version_code mgs) -ge $(version_code 2.5.57) ]] ||
5377 { skip "Need MGS version at least 2.5.57" && return 0; }
5380 do_facet mgs $LCTL get_param mgs.MGS.live.params ||
5381 error "start params log failed"
5384 run_test 76b "verify params log setup correctly"
5387 [[ $(lustre_version_code mgs) -ge $(version_code 2.8.54) ]] ||
5388 { skip "Need MDS version at least 2.4.52" && return 0; }
5390 local MASK_PARAM="mdd.*.changelog_mask"
5391 echo "Change changelog_mask"
5392 do_facet mgs $LCTL set_param -P $MASK_PARAM=-CLOSE ||
5393 error "Can't change changlog_mask"
5394 wait_update $(facet_host mds) "$LCTL get_param -n $MASK_PARAM |
5397 echo "Check the value is stored after mds remount"
5398 stop_mds || error "Failed to stop MDS"
5399 start_mds || error "Failed to start MDS"
5400 local CHANGELOG_MASK=$(do_facet mgs $LCTL get_param -n $MASK_PARAM)
5401 echo $CHANGELOG_MASK | grep CLOSE > /dev/null &&
5402 error "changelog_mask is not changed"
5406 run_test 76c "verify changelog_mask is applied with set_param -P"
5408 test_76d() { #LU-9399
5411 local xattr_cache="llite.*.xattr_cache"
5412 local cmd="$LCTL get_param -n $xattr_cache | head -1"
5413 local new=$((($(eval $cmd) + 1) % 2))
5415 echo "lctl set_param -P llite.*.xattr_cache=$new"
5416 do_facet mgs $LCTL set_param -P $xattr_cache=$new ||
5417 error "Can't change xattr_cache"
5418 wait_update $HOSTNAME "$cmd" "$new"
5420 echo "Check $xattr_cache on client $MOUNT"
5421 umount_client $MOUNT || error "umount $MOUNT failed"
5422 mount_client $MOUNT || error "mount $MOUNT failed"
5423 [ $(eval $cmd) -eq $new ] ||
5424 error "$xattr_cache != $new on client $MOUNT"
5426 echo "Check $xattr_cache on the new client $MOUNT2"
5427 mount_client $MOUNT2 || error "mount $MOUNT2 failed"
5428 [ $(eval $cmd) -eq $new ] ||
5429 error "$xattr_cache != $new on client $MOUNT2"
5430 umount_client $MOUNT2 || error "umount $MOUNT2 failed"
5434 run_test 76d "verify llite.*.xattr_cache can be set by 'set_param -P' correctly"
5436 test_77() { # LU-3445
5437 local server_version=$(lustre_version_code $SINGLEMDS)
5438 [[ $server_version -ge $(version_code 2.8.55) ]] ||
5439 { skip "Need MDS version 2.8.55+ "; return; }
5441 if [[ -z "$fs2ost_DEV" || -z "$fs2mds_DEV" ]]; then
5442 is_blkdev $SINGLEMDS $(mdsdevname ${SINGLEMDS//mds/}) &&
5443 skip_env "mixed loopback and real device not working" && return
5446 local fs2mdsdev=$(mdsdevname 1_2)
5447 local fs2ostdev=$(ostdevname 1_2)
5448 local fs2mdsvdev=$(mdsvdevname 1_2)
5449 local fs2ostvdev=$(ostvdevname 1_2)
5450 local fsname=test1234
5452 local failnid="$(h2nettype 1.2.3.4),$(h2nettype 4.3.2.1)"
5454 combined_mgs_mds || stop_mgs || error "stopping MGS service failed"
5456 add fs2mds $(mkfs_opts mds1 $fs2mdsdev) --mgs --fsname=$fsname \
5457 --reformat $fs2mdsdev $fs2mdsvdev || error "add fs2mds failed"
5458 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS && trap cleanup_fs2 EXIT INT ||
5459 error "start fs2mds failed"
5461 mgsnid=$(do_facet fs2mds $LCTL list_nids | xargs | tr ' ' ,)
5462 mgsnid="0.0.0.0@tcp,$mgsnid,$mgsnid:$mgsnid"
5464 add fs2ost --mgsnode=$mgsnid $(mkfs_opts ost1 $fs2ostdev) \
5465 --failnode=$failnid --fsname=$fsname \
5466 --reformat $fs2ostdev $fs2ostvdev ||
5467 error "add fs2ost failed"
5468 start fs2ost $fs2ostdev $OST_MOUNT_OPTS || error "start fs2ost failed"
5470 mkdir -p $MOUNT2 || error "mkdir $MOUNT2 failed"
5471 $MOUNT_CMD $mgsnid:/$fsname $MOUNT2 || error "mount $MOUNT2 failed"
5472 DIR=$MOUNT2 MOUNT=$MOUNT2 check_mount || error "check $MOUNT2 failed"
5475 run_test 77 "comma-separated MGS NIDs and failover node NIDs"
5478 [[ $(facet_fstype $SINGLEMDS) != ldiskfs ||
5479 $(facet_fstype ost1) != ldiskfs ]] &&
5480 skip "ldiskfs only test" && return
5482 # reformat the Lustre filesystem with a smaller size
5483 local saved_MDSCOUNT=$MDSCOUNT
5484 local saved_MDSSIZE=$MDSSIZE
5485 local saved_OSTCOUNT=$OSTCOUNT
5486 local saved_OSTSIZE=$OSTSIZE
5489 MDSSIZE=$((MDSSIZE - 20000))
5490 OSTSIZE=$((OSTSIZE - 20000))
5491 reformat || error "(1) reformat Lustre filesystem failed"
5492 MDSSIZE=$saved_MDSSIZE
5493 OSTSIZE=$saved_OSTSIZE
5495 # mount the Lustre filesystem
5496 setup_noconfig || error "(2) setup Lustre filesystem failed"
5499 log "create test files"
5504 mkdir $MOUNT/$tdir || error "(3) mkdir $MOUNT/$tdir failed"
5506 for i in $(seq $num_files); do
5507 file=$MOUNT/$tdir/$tfile-$i
5508 dd if=/dev/urandom of=$file count=1 bs=1M || {
5509 $LCTL get_param osc.*.cur*grant*
5510 $LFS df; $LFS df -i;
5511 # stop creating files if there is no more space
5512 if [ ! -e $file ]; then
5513 num_files=$((i - 1))
5517 $LFS getstripe -v $file
5518 local ost_idx=$(LFS getstripe -i $file)
5519 do_facet ost$((ost_idx + 1)) \
5520 $LCTL get_param obdfilter.*.*grant*
5521 error "(4) create $file failed"
5525 # unmount the Lustre filesystem
5526 cleanup || error "(5) cleanup Lustre filesystem failed"
5528 # run e2fsck on the MDT and OST devices
5529 local mds_host=$(facet_active_host $SINGLEMDS)
5530 local ost_host=$(facet_active_host ost1)
5531 local mds_dev=$(mdsdevname ${SINGLEMDS//mds/})
5532 local ost_dev=$(ostdevname 1)
5534 run_e2fsck $mds_host $mds_dev "-y"
5535 run_e2fsck $ost_host $ost_dev "-y"
5537 # get the original block count of the MDT and OST filesystems
5538 local mds_orig_blks=$(get_block_count $SINGLEMDS $mds_dev)
5539 local ost_orig_blks=$(get_block_count ost1 $ost_dev)
5541 # expand the MDT and OST filesystems to the device size
5542 run_resize2fs $SINGLEMDS $mds_dev "" || error "expand $SINGLEMDS failed"
5543 run_resize2fs ost1 $ost_dev "" || error "expand ost1 failed"
5545 # run e2fsck on the MDT and OST devices again
5546 run_e2fsck $mds_host $mds_dev "-y"
5547 run_e2fsck $ost_host $ost_dev "-y"
5549 # mount the Lustre filesystem
5553 log "check files after expanding the MDT and OST filesystems"
5554 for i in $(seq $num_files); do
5555 file=$MOUNT/$tdir/$tfile-$i
5556 $CHECKSTAT -t file -s 1048576 $file ||
5557 error "(6) checkstat $file failed"
5561 log "create more files after expanding the MDT and OST filesystems"
5562 for i in $(seq $((num_files + 1)) $((num_files + 10))); do
5563 file=$MOUNT/$tdir/$tfile-$i
5564 dd if=/dev/urandom of=$file count=1 bs=1M ||
5565 error "(7) create $file failed"
5568 # unmount the Lustre filesystem
5569 cleanup || error "(8) cleanup Lustre filesystem failed"
5571 # run e2fsck on the MDT and OST devices
5572 run_e2fsck $mds_host $mds_dev "-y"
5573 run_e2fsck $ost_host $ost_dev "-y"
5575 # get the maximum block count of the MDT and OST filesystems
5576 local mds_max_blks=$(get_block_count $SINGLEMDS $mds_dev)
5577 local ost_max_blks=$(get_block_count ost1 $ost_dev)
5579 # get the minimum block count of the MDT and OST filesystems
5580 local mds_min_blks=$(run_resize2fs $SINGLEMDS $mds_dev "" "-P" 2>&1 |
5581 grep minimum | sed -e 's/^.*filesystem: //g')
5582 local ost_min_blks=$(run_resize2fs ost1 $ost_dev "" "-P" 2>&1 |
5583 grep minimum | sed -e 's/^.*filesystem: //g')
5585 # shrink the MDT and OST filesystems to a smaller size
5589 if [[ $mds_max_blks -gt $mds_min_blks &&
5590 $mds_max_blks -gt $mds_orig_blks ]]; then
5591 [[ $mds_orig_blks -gt $mds_min_blks ]] &&
5592 base_blks=$mds_orig_blks || base_blks=$mds_min_blks
5593 new_blks=$(( (mds_max_blks - base_blks) / 2 + base_blks ))
5594 run_resize2fs $SINGLEMDS $mds_dev $new_blks ||
5595 error "shrink $SINGLEMDS to $new_blks failed"
5599 if [[ $ost_max_blks -gt $ost_min_blks &&
5600 $ost_max_blks -gt $ost_orig_blks ]]; then
5601 [[ $ost_orig_blks -gt $ost_min_blks ]] &&
5602 base_blks=$ost_orig_blks || base_blks=$ost_min_blks
5603 new_blks=$(( (ost_max_blks - base_blks) / 2 + base_blks ))
5604 run_resize2fs ost1 $ost_dev $new_blks ||
5605 error "shrink ost1 to $new_blks failed"
5609 # check whether the MDT or OST filesystem was shrunk or not
5611 combined_mgs_mds || stop_mgs || error "(9) stop mgs failed"
5612 reformat || error "(10) reformat Lustre filesystem failed"
5616 # run e2fsck on the MDT and OST devices again
5617 run_e2fsck $mds_host $mds_dev "-y"
5618 run_e2fsck $ost_host $ost_dev "-y"
5620 # mount the Lustre filesystem again
5624 log "check files after shrinking the MDT and OST filesystems"
5625 for i in $(seq $((num_files + 10))); do
5626 file=$MOUNT/$tdir/$tfile-$i
5627 $CHECKSTAT -t file -s 1048576 $file ||
5628 error "(11) checkstat $file failed"
5631 # unmount and reformat the Lustre filesystem
5632 cleanup || error "(12) cleanup Lustre filesystem failed"
5633 combined_mgs_mds || stop_mgs || error "(13) stop mgs failed"
5635 MDSCOUNT=$saved_MDSCOUNT
5636 OSTCOUNT=$saved_OSTCOUNT
5637 reformat || error "(14) reformat Lustre filesystem failed"
5639 run_test 78 "run resize2fs on MDT and OST filesystems"
5641 test_79() { # LU-4227
5642 [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.5.59) ]] ||
5643 { skip "Need MDS version at least 2.5.59"; return 0; }
5645 local mdsdev1=$(mdsdevname 1)
5646 local mdsvdev1=$(mdsvdevname 1)
5647 local mdsdev2=$(mdsdevname 2)
5648 local mdsvdev2=$(mdsvdevname 2)
5649 local ostdev1=$(ostdevname 1)
5650 local ostvdev1=$(ostvdevname 1)
5651 local opts_mds1="$(mkfs_opts mds1 $mdsdev1) --reformat"
5652 local opts_mds2="$(mkfs_opts mds2 $mdsdev2) --reformat"
5653 local opts_ost1="$(mkfs_opts ost1 $ostdev1) --reformat"
5656 # remove --mgs/--mgsnode from mkfs.lustre options
5657 opts_mds1=$(echo $opts_mds1 | sed -e "s/--mgs//")
5659 mgsnode_opt=$(echo $opts_mds2 |
5660 awk '{ for ( i = 1; i < NF; i++ )
5661 if ( $i ~ "--mgsnode" ) { print $i; break } }')
5662 [ -n $mgsnode_opt ] &&
5663 opts_mds2=$(echo $opts_mds2 | sed -e "s/$mgsnode_opt//")
5665 mgsnode_opt=$(echo $opts_ost1 |
5666 awk '{ for ( i = 1; i < NF; i++ )
5667 if ( $i ~ "--mgsnode" ) { print $i; break } }')
5668 [ -n $mgsnode_opt ] &&
5669 opts_ost1=$(echo $opts_ost1 | sed -e "s/$mgsnode_opt//")
5671 # -MGS, format a mdt without --mgs option
5672 add mds1 $opts_mds1 $mdsdev1 $mdsvdev1 &&
5673 error "Must specify --mgs when formatting mdt combined with mgs"
5675 # +MGS, format a mdt/ost without --mgsnode option
5676 add mds1 $(mkfs_opts mds1 $mdsdev1) --reformat $mdsdev1 $mdsvdev1 \
5677 > /dev/null || error "start mds1 failed"
5678 add mds2 $opts_mds2 $mdsdev2 $mdsvdev2 &&
5679 error "Must specify --mgsnode when formatting a mdt"
5680 add ost1 $opts_ost1 $ostdev1 $ostvdev2 &&
5681 error "Must specify --mgsnode when formatting an ost"
5685 run_test 79 "format MDT/OST without mgs option (should return errors)"
5688 start_mds || error "Failed to start MDT"
5689 start_ost || error "Failed to start OST1"
5690 uuid=$(do_facet ost1 $LCTL get_param -n mgc.*.uuid)
5691 #define OBD_FAIL_MGS_PAUSE_TARGET_CON 0x906
5692 do_facet ost1 "$LCTL set_param fail_val=10 fail_loc=0x906"
5693 do_facet mgs "$LCTL set_param fail_val=10 fail_loc=0x906"
5694 do_facet mgs "$LCTL set_param -n mgs/MGS/evict_client $uuid"
5696 start_ost2 || error "Failed to start OST2"
5698 do_facet ost1 "$LCTL set_param fail_loc=0"
5701 run_test 80 "mgc import reconnect race"
5703 #Save the original values of $OSTCOUNT and $OSTINDEX$i.
5705 local new_ostcount=$1
5706 saved_ostcount=$OSTCOUNT
5707 OSTCOUNT=$new_ostcount
5711 for i in $(seq $OSTCOUNT); do
5713 eval saved_ostindex$i=${!index}
5718 # Restore the original values of $OSTCOUNT and $OSTINDEX$i.
5719 restore_ostindex() {
5724 for i in $(seq $OSTCOUNT); do
5725 index=saved_ostindex$i
5726 eval OSTINDEX$i=${!index}
5728 OSTCOUNT=$saved_ostcount
5733 # The main purpose of this test is to ensure the OST_INDEX_LIST functions as
5734 # expected. This test uses OST_INDEX_LIST to format OSTs with a randomly
5735 # assigned index and ensures we can mount such a formatted file system
5736 test_81() { # LU-4665
5737 [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.6.54) ]] ||
5738 { skip "Need MDS version at least 2.6.54" && return; }
5739 [[ $OSTCOUNT -ge 3 ]] || { skip_env "needs >= 3 OSTs" && return; }
5743 # Each time RANDOM is referenced, a random integer between 0 and 32767
5746 local saved_ostindex1=$OSTINDEX1
5747 for i in 65535 $((RANDOM + 65536)); do
5748 echo -e "\nFormat ost1 with --index=$i, should fail"
5750 if add ost1 $(mkfs_opts ost1 $(ostdevname 1)) --reformat \
5751 $(ostdevname 1) $(ostvdevname 1); then
5752 OSTINDEX1=$saved_ostindex1
5753 error "format ost1 with --index=$i should fail"
5756 OSTINDEX1=$saved_ostindex1
5760 # Format OSTs with random sparse indices.
5761 trap "restore_ostindex" EXIT
5762 echo -e "\nFormat $OSTCOUNT OSTs with sparse indices"
5763 OST_INDEX_LIST=[0,$((RANDOM * 2 % 65533 + 1)),65534] formatall
5765 # Setup and check Lustre filesystem.
5766 start_mgsmds || error "start_mgsmds failed"
5767 for i in $(seq $OSTCOUNT); do
5768 start ost$i $(ostdevname $i) $OST_MOUNT_OPTS ||
5769 error "start ost$i failed"
5772 mount_client $MOUNT || error "mount client $MOUNT failed"
5773 check_mount || error "check client $MOUNT failed"
5776 local max_easize=$($LCTL get_param -n llite.*.max_easize)
5777 [[ $max_easize -eq 128 ]] ||
5778 error "max_easize is $max_easize, should be 128 bytes"
5782 run_test 81 "sparse OST indexing"
5784 # Here we exercise the stripe placement functionality on a file system that
5785 # has formatted the OST with a random index. With the file system the following
5786 # functionality is tested:
5788 # 1. Creating a new file with a specific stripe layout.
5790 # 2. Modifiy a existing empty file with a specific stripe layout.
5792 # 3. Ensure we fail to set the stripe layout of a file that already has one.
5794 # 4. If ost-index is defined we need to ensure it is the first entry in the
5795 # ost index list returned by lfs getstripe.
5797 # 5. Lastly ensure this functionality fails with directories.
5798 test_82a() { # LU-4665
5799 [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.6.54) ]] ||
5800 { skip "Need MDS version at least 2.6.54" && return; }
5801 [[ $OSTCOUNT -ge 3 ]] || { skip_env "needs >= 3 OSTs" && return; }
5807 # Format OSTs with random sparse indices.
5811 local LOV_V1_INSANE_STRIPE_COUNT=65532
5812 for i in $(seq $OSTCOUNT); do
5813 index=$(((RANDOM * 2) % LOV_V1_INSANE_STRIPE_COUNT))
5814 ost_indices+=" $index"
5816 ost_indices=$(comma_list $ost_indices)
5818 stack_trap "restore_ostindex" EXIT
5819 echo -e "\nFormat $OSTCOUNT OSTs with sparse indices $ost_indices"
5820 OST_INDEX_LIST=[$ost_indices] formatall
5822 # Setup Lustre filesystem.
5823 start_mgsmds || error "start_mgsmds failed"
5824 for i in $(seq $OSTCOUNT); do
5825 start ost$i $(ostdevname $i) $OST_MOUNT_OPTS ||
5826 error "start ost$i failed"
5829 # Collect debug information - start of test
5830 do_nodes $(comma_list $(mdts_nodes)) \
5831 $LCTL get_param osc.*.prealloc_*_id
5833 mount_client $MOUNT || error "mount client $MOUNT failed"
5836 $LFS df $MOUNT || error "$LFS df $MOUNT failed"
5837 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
5839 stack_trap "do_nodes $(comma_list $(mdts_nodes)) \
5840 $LCTL get_param osc.*.prealloc_*_id" EXIT
5842 # 1. If the file does not exist, new file will be created
5843 # with specified OSTs.
5844 local file=$DIR/$tdir/$tfile-1
5845 local cmd="$SETSTRIPE -o $ost_indices $file"
5847 eval $cmd || error "$cmd failed"
5848 check_stripe_count $file $OSTCOUNT
5849 check_obdidx $file $ost_indices
5850 dd if=/dev/urandom of=$file count=1 bs=1M > /dev/null 2>&1 ||
5851 error "write $file failed"
5853 # 2. If the file already exists and is an empty file, the file
5854 # will be attached with specified layout.
5855 file=$DIR/$tdir/$tfile-2
5856 mcreate $file || error "mcreate $file failed"
5857 cmd="$SETSTRIPE -o $ost_indices $file"
5859 eval $cmd || error "$cmd failed"
5860 dd if=/dev/urandom of=$file count=1 bs=1M > /dev/null 2>&1 ||
5861 error "write $file failed"
5862 check_stripe_count $file $OSTCOUNT
5863 check_obdidx $file $ost_indices
5865 # 3. If the file already has a valid layout attached, the command
5866 # should fail with EBUSY.
5868 eval $cmd && error "stripe is already set on $file, $cmd should fail"
5870 # 4. If [--stripe-index|-i <start_ost_idx>] is used, the index must
5871 # be in the OST indices list.
5872 local start_ost_idx=${ost_indices##*,}
5873 file=$DIR/$tdir/$tfile-3
5874 cmd="$SETSTRIPE -o $ost_indices -i $start_ost_idx $file"
5876 eval $cmd || error "$cmd failed"
5877 check_stripe_count $file $OSTCOUNT
5878 check_obdidx $file $ost_indices
5879 check_start_ost_idx $file $start_ost_idx
5881 file=$DIR/$tdir/$tfile-4
5883 cmd+=" -o $(exclude_items_from_list $ost_indices $start_ost_idx)"
5884 cmd+=" -i $start_ost_idx $file"
5886 eval $cmd && error "index $start_ost_idx should be in $ost_indices"
5888 # 5. Specifying OST indices for directory should fail with ENOSUPP.
5889 local dir=$DIR/$tdir/$tdir
5890 mkdir $dir || error "mkdir $dir failed"
5891 cmd="$SETSTRIPE -o $ost_indices $dir"
5893 eval $cmd && error "$cmd should fail, specifying OST indices" \
5894 "for directory is not supported"
5898 run_test 82a "specify OSTs for file (succeed) or directory (fail)"
5903 # Remove OSTs from a pool and destroy the pool.
5904 destroy_pool $ost_pool || true
5906 if ! combined_mgs_mds ; then
5912 # Test 82b is run to ensure that if the user supplies a pool with a specific
5913 # stripe layout that it behaves proprerly. It should fail in the case that
5914 # the supplied OST index list points to OSTs not contained in the user
5916 test_82b() { # LU-4665
5917 [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.6.54) ]] ||
5918 { skip "Need MDS version at least 2.6.54" && return; }
5919 [[ $OSTCOUNT -ge 4 ]] || { skip_env "needs >= 4 OSTs" && return; }
5925 # Format OSTs with random sparse indices.
5929 local LOV_V1_INSANE_STRIPE_COUNT=65532
5930 for i in $(seq $OSTCOUNT); do
5931 index=$(((RANDOM * 2) % LOV_V1_INSANE_STRIPE_COUNT))
5932 ost_indices+=" $index"
5934 ost_indices=$(comma_list $ost_indices)
5936 trap "restore_ostindex" EXIT
5937 echo -e "\nFormat $OSTCOUNT OSTs with sparse indices $ost_indices"
5938 OST_INDEX_LIST=[$ost_indices] formatall
5940 # Setup Lustre filesystem.
5941 start_mgsmds || error "start_mgsmds failed"
5942 for i in $(seq $OSTCOUNT); do
5943 start ost$i $(ostdevname $i) $OST_MOUNT_OPTS ||
5944 error "start ost$i failed"
5947 mount_client $MOUNT || error "mount client $MOUNT failed"
5948 if ! combined_mgs_mds ; then
5953 $LFS df $MOUNT || error "$LFS df $MOUNT failed"
5954 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
5956 # Create a new pool and add OSTs into it.
5957 local ost_pool=$FSNAME.$TESTNAME
5958 create_pool $ost_pool || error "create OST pool $ost_pool failed"
5961 trap "cleanup_82b" EXIT
5963 local ost_idx_in_list=${ost_indices##*,}
5964 local ost_idx_in_pool=$(exclude_items_from_list $ost_indices \
5967 local ost_targets="$FSNAME-OST["
5968 for i in ${ost_idx_in_pool//,/ }; do
5969 ost_targets=$ost_targets$(printf "%04x," $i)
5971 ost_targets="${ost_targets%,}]"
5973 local ost_targets_uuid=$(for i in ${ost_idx_in_pool//,/ }; \
5974 do printf "$FSNAME-OST%04x_UUID\n" $i; done |
5975 sort -u | tr '\n' ' ')
5977 local cmd="$LCTL pool_add $ost_pool $ost_targets"
5978 do_facet mgs $cmd || error "$cmd failed"
5979 wait_update $HOSTNAME "$LCTL get_param -n lov.$FSNAME-*.pools.$TESTNAME|
5980 sort -u | tr '\n' ' ' " "$ost_targets_uuid" ||
5981 error "wait_update $ost_pool failed"
5982 wait_update_facet $SINGLEMDS "$LCTL pool_list $ost_pool | wc -l" 4 ||
5983 error "wait_update pool_list $ost_pool failed"
5985 # If [--pool|-p <pool_name>] is set with [--ost-list|-o <ost_indices>],
5986 # then the OSTs must be the members of the pool.
5987 local file=$DIR/$tdir/$tfile
5988 cmd="$SETSTRIPE -p $ost_pool -o $ost_idx_in_list $file"
5990 eval $cmd && error "OST with index $ost_idx_in_list should be" \
5991 "in OST pool $ost_pool"
5993 # Only select OST $ost_idx_in_list from $ost_pool for file.
5994 ost_idx_in_list=${ost_idx_in_pool#*,}
5995 cmd="$SETSTRIPE -p $ost_pool -o $ost_idx_in_list $file"
5997 eval $cmd || error "$cmd failed"
5998 cmd="$GETSTRIPE $file"
6000 eval $cmd || error "$cmd failed"
6001 check_stripe_count $file 2
6002 check_obdidx $file $ost_idx_in_list
6003 dd if=/dev/urandom of=$file count=1 bs=1M > /dev/null 2>&1 ||
6004 error "write $file failed"
6008 run_test 82b "specify OSTs for file with --pool and --ost-list options"
6011 [[ $(lustre_version_code ost1) -ge $(version_code 2.6.91) ]] ||
6012 { skip "Need OST version at least 2.6.91" && return 0; }
6013 if [ $(facet_fstype ost1) != ldiskfs ]; then
6014 skip "ldiskfs only test"
6024 ostmnt=$(facet_mntpt ost1)
6025 fstype=$(facet_fstype ost1)
6027 # Mount the OST as an ldiskfs filesystem.
6028 log "mount the OST $dev as a $fstype filesystem"
6029 add ost1 $(mkfs_opts ost1 $dev) $FSTYPE_OPT \
6030 --reformat $dev > /dev/null ||
6031 error "format ost1 error"
6033 if ! test -b $dev; then
6034 mnt_opts=$(csa_add "$OST_MOUNT_OPTS" -o loop)
6036 echo "mnt_opts $mnt_opts"
6037 do_facet ost1 mount -t $fstype $dev \
6039 # Run llverfs on the mounted ldiskfs filesystem.
6040 # It is needed to get ENOSPACE.
6041 log "run llverfs in partial mode on the OST $fstype $ostmnt"
6042 do_rpc_nodes $(facet_host ost1) run_llverfs $ostmnt -vpl \
6043 "no" || error "run_llverfs error on $fstype"
6046 log "unmount the OST $dev"
6049 # Delete file IO_scrub. Later osd_scrub_setup will try to
6050 # create "IO_scrub" but will get ENOSPACE.
6052 echo "start ost1 service on `facet_active_host ost1`"
6053 start ost1 `ostdevname 1` $OST_MOUNT_OPTS
6056 err=$(do_facet ost1 dmesg | grep "VFS: Busy inodes after unmount of")
6057 echo "string err $err"
6058 [ -z "$err" ] || error $err
6061 run_test 83 "ENOSPACE on OST doesn't cause message VFS: \
6062 Busy inodes after unmount ..."
6065 local facet=$SINGLEMDS
6066 local num=$(echo $facet | tr -d "mds")
6067 local dev=$(mdsdevname $num)
6068 local time_min=$(recovery_time_min)
6069 local recovery_duration
6070 local completed_clients
6071 local correct_clients
6074 echo "start mds service on $(facet_active_host $facet)"
6076 "-o recovery_time_hard=$time_min,recovery_time_soft=$time_min" $@ ||
6077 error "start MDS failed"
6079 start_ost || error "start OST0000 failed"
6080 start_ost2 || error "start OST0001 failed"
6082 echo "recovery_time=$time_min, timeout=$TIMEOUT, wrap_up=$wrap_up"
6084 mount_client $MOUNT1 || error "mount $MOUNT1 failed"
6085 mount_client $MOUNT2 || error "mount $MOUNT2 failed"
6086 # make sure new superblock labels are sync'd before disabling writes
6090 replay_barrier $SINGLEMDS
6091 createmany -o $DIR1/$tfile-%d 1000
6093 # We need to catch the end of recovery window to extend it.
6094 # Skip 5 requests and add delay to request handling.
6095 #define OBD_FAIL_TGT_REPLAY_DELAY 0x709 | FAIL_SKIP
6096 do_facet $SINGLEMDS "lctl set_param fail_loc=0x20000709 fail_val=5"
6098 facet_failover --fsck $SINGLEMDS || error "failover: $?"
6101 echo "recovery status"
6102 do_facet $SINGLEMDS \
6103 "$LCTL get_param -n mdt.$FSNAME-MDT0000.recovery_status"
6105 recovery_duration=$(do_facet $SINGLEMDS \
6106 "$LCTL get_param -n mdt.$FSNAME-MDT0000.recovery_status" |
6107 awk '/recovery_duration/ { print $2 }')
6108 (( $recovery_duration > $time_min + $wrap_up )) &&
6109 error "recovery_duration > recovery_time_hard + wrap up"
6110 completed_clients=$(do_facet $SINGLEMDS \
6111 "$LCTL get_param -n mdt.$FSNAME-MDT0000.recovery_status" |
6112 awk '/completed_clients/ { print $2 }')
6114 correct_clients="$MDSCOUNT/$((MDSCOUNT+1))"
6115 [ "$completed_clients" = "${correct_clients}" ] ||
6116 error "$completed_clients != $correct_clients"
6118 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
6119 umount_client $MOUNT1
6120 umount_client $MOUNT2
6126 run_test 84 "check recovery_hard_time"
6129 [[ $(lustre_version_code ost1) -ge $(version_code 2.7.55) ]] ||
6130 { skip "Need OST version at least 2.7.55" && return 0; }
6131 ##define OBD_FAIL_OSD_OST_EA_FID_SET 0x197
6132 do_facet ost1 "lctl set_param fail_loc=0x197"
6136 run_test 85 "osd_ost init: fail ea_fid_set"
6141 # ost1 has already registered to the MGS before the reformat.
6142 # So after reformatting it with option "-G", it could not be
6143 # mounted to the MGS. Cleanup the system for subsequent tests.
6148 local server_version=$(lustre_version_code $SINGLEMDS)
6149 [ "$(facet_fstype ost1)" = "zfs" ] &&
6150 skip "LU-6442: no such mkfs params for ZFS OSTs" && return
6151 [[ $server_version -ge $(version_code 2.7.56) ]] ||
6152 { skip "Need server version newer than 2.7.55"; return 0; }
6154 local OST_OPTS="$(mkfs_opts ost1 $(ostdevname 1)) \
6155 --reformat $(ostdevname 1) $(ostvdevname 1)"
6158 local OLDSIZE=$(do_facet ost1 "$DEBUGFS -c -R stats $(ostdevname 1)" |
6159 awk '/Flex block group size: / { print $NF; exit; }')
6162 if [[ ${!opts} != *mkfsoptions* ]]; then
6163 eval opts=\"${!opts} \
6164 --mkfsoptions='\\\"-O flex_bg -G $NEWSIZE\\\"'\"
6166 val=${!opts//--mkfsoptions=\\\"/ \
6167 --mkfsoptions=\\\"-O flex_bg -G $NEWSIZE }
6171 echo "params: $opts"
6173 trap cleanup_86 EXIT ERR
6176 add ost1 $opts || error "add ost1 failed with new params"
6178 local FOUNDSIZE=$(do_facet ost1 "$DEBUGFS -c -R stats $(ostdevname 1)" |
6179 awk '/Flex block group size: / { print $NF; exit; }')
6181 [[ $FOUNDSIZE == $NEWSIZE ]] ||
6182 error "Flex block group size: $FOUNDSIZE, expected: $NEWSIZE"
6186 run_test 86 "Replacing mkfs.lustre -G option"
6188 test_87() { #LU-6544
6189 [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.9.51) ]] ||
6190 { skip "Need MDS version at least 2.9.51" && return; }
6191 [[ $(facet_fstype $SINGLEMDS) != ldiskfs ]] &&
6192 { skip "ldiskfs only test" && return; }
6193 [[ $OSTCOUNT -gt 59 ]] &&
6194 { skip "Ignore wide striping situation" && return; }
6196 local mdsdev=$(mdsdevname 1)
6197 local mdsvdev=$(mdsvdevname 1)
6198 local file=$DIR/$tfile
6199 local mntpt=$(facet_mntpt $SINGLEMDS)
6200 local used_xattr_blk=0
6201 local inode_size=${1:-1024}
6203 local xtest="trusted.test"
6207 local stripe_cnt=$(($OSTCOUNT + 2))
6209 #Please see ldiskfs_make_lustre() for MDT inode size calculation
6210 if [ $stripe_cnt -gt 16 ]; then
6213 left_size=$(expr $inode_size - \
6216 32 - 40 \* 3 - 32 \* 3 - $stripe_cnt \* 24 - 16 - 3 - \
6218 24 - 18 - $(expr length $tfile) - 16 - 4)
6219 if [ $left_size -le 0 ]; then
6220 echo "No space($left_size) is expected in inode."
6221 echo "Try 1-byte xattr instead to verify this."
6224 echo "Estimate: at most $left_size-byte space left in inode."
6230 add mds1 $(mkfs_opts mds1 ${mdsdev}) --stripe-count-hint=$stripe_cnt \
6231 --reformat $mdsdev $mdsvdev || error "add mds1 failed"
6232 start_mdt 1 > /dev/null || error "start mdt1 failed"
6233 for i in $(seq $OSTCOUNT); do
6234 start ost$i $(ostdevname $i) $OST_MOUNT_OPTS > /dev/null ||
6235 error "start ost$i failed"
6237 mount_client $MOUNT > /dev/null || error "mount client $MOUNT failed"
6238 check_mount || error "check client $MOUNT failed"
6241 $SETSTRIPE -E 1M -c 1 -E 64M -c 1 -E -1 -c -1 $file ||
6242 error "Create file with 3 components failed"
6243 $TRUNCATE $file $((1024*1024*64+1)) || error "truncate file failed"
6244 i=$($GETSTRIPE -I3 -c $file) || error "get 3rd stripe count failed"
6245 if [ $i -ne $OSTCOUNT ]; then
6246 left_size=$(expr $left_size + $(expr $OSTCOUNT - $i) \* 24)
6247 echo -n "Since only $i out $OSTCOUNT OSTs are used, "
6248 echo -n "the expected left space is changed to "
6249 echo "$left_size bytes at most."
6251 value=$(generate_string $left_size)
6252 setfattr -n $xtest -v $value $file
6253 orig=$(get_xattr_value $xtest $file)
6254 [[ "$orig" != "$value" ]] && error "$xtest changed"
6256 #Verify if inode has some expected space left
6257 umount $MOUNT > /dev/null || error "umount $MOUNT failed"
6258 stop_mdt 1 > /dev/null || error "stop mdt1 failed"
6259 mount_ldiskfs $SINGLEMDS || error "mount -t ldiskfs $SINGLEMDS failed"
6261 do_facet $SINGLEMDS ls -sal $mntpt/ROOT/$tfile
6262 used_xattr_blk=$(do_facet $SINGLEMDS ls -s $mntpt/ROOT/$tfile |
6264 [[ $used_xattr_blk -eq 0 ]] &&
6265 error "Please check MDS inode size calculation: \
6266 more than $left_size-byte space left in inode."
6267 echo "Verified: at most $left_size-byte space left in inode."
6271 run_test 87 "check if MDT inode can hold EAs with N stripes properly"
6274 [ "$(facet_fstype mds1)" == "zfs" ] &&
6275 skip "LU-6662: no implementation for ZFS" && return
6279 add mds1 $(mkfs_opts mds1 $(mdsdevname 1)) \
6280 --reformat $(mdsdevname 1) || error "add mds1 failed"
6282 do_facet mds1 "$TUNEFS $(mdsdevname 1) |
6283 grep -e \".*opts:.*errors=remount-ro.*\"" ||
6284 error "default mount options is missing"
6286 add mds1 $(mkfs_opts mds1 $(mdsdevname 1)) \
6287 --mountfsoptions="user_xattr,errors=panic" \
6288 --reformat $(mdsdevname 1) || error "add mds1 failed"
6290 do_facet mds1 "$TUNEFS $(mdsdevname 1) |
6291 grep -e \".*opts:.*errors=panic.*\"" ||
6292 error "user can't override default mount options"
6294 run_test 88 "check the default mount options can be overridden"
6296 test_89() { # LU-7131
6297 [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.9.54) ]] ||
6298 { skip "Need MDT version at least 2.9.54" && return 0; }
6300 local key=failover.node
6301 local val1=192.0.2.254@tcp0 # Reserved IPs, see RFC 5735
6302 local val2=192.0.2.255@tcp0
6303 local mdsdev=$(mdsdevname 1)
6308 [ $(facet_fstype mds1) == zfs ] && import_zpool mds1
6309 # Check that parameters are added correctly
6310 echo "tunefs --param $key=$val1"
6311 do_facet mds "$TUNEFS --param $key=$val1 $mdsdev >/dev/null" ||
6312 error "tunefs --param $key=$val1 failed"
6313 params=$(do_facet mds $TUNEFS --dryrun $mdsdev) ||
6314 error "tunefs --dryrun failed"
6315 params=${params##*Parameters:}
6316 params=${params%%exiting*}
6317 [ $(echo $params | tr ' ' '\n' | grep -c $key=$val1) = "1" ] ||
6318 error "on-disk parameter not added correctly via tunefs"
6320 # Check that parameters replace existing instances when added
6321 echo "tunefs --param $key=$val2"
6322 do_facet mds "$TUNEFS --param $key=$val2 $mdsdev >/dev/null" ||
6323 error "tunefs --param $key=$val2 failed"
6324 params=$(do_facet mds $TUNEFS --dryrun $mdsdev) ||
6325 error "tunefs --dryrun failed"
6326 params=${params##*Parameters:}
6327 params=${params%%exiting*}
6328 [ $(echo $params | tr ' ' '\n' | grep -c $key=) = "1" ] ||
6329 error "on-disk parameter not replaced via tunefs"
6330 [ $(echo $params | tr ' ' '\n' | grep -c $key=$val2) = "1" ] ||
6331 error "on-disk parameter not replaced correctly via tunefs"
6333 # Check that a parameter is erased properly
6334 echo "tunefs --erase-param $key"
6335 do_facet mds "$TUNEFS --erase-param $key $mdsdev >/dev/null" ||
6336 error "tunefs --erase-param $key failed"
6337 params=$(do_facet mds $TUNEFS --dryrun $mdsdev) ||
6338 error "tunefs --dryrun failed"
6339 params=${params##*Parameters:}
6340 params=${params%%exiting*}
6341 [ $(echo $params | tr ' ' '\n' | grep -c $key=) = "0" ] ||
6342 error "on-disk parameter not erased correctly via tunefs"
6344 # Check that all the parameters are erased
6345 echo "tunefs --erase-params"
6346 do_facet mds "$TUNEFS --erase-params $mdsdev >/dev/null" ||
6347 error "tunefs --erase-params failed"
6348 params=$(do_facet mds $TUNEFS --dryrun $mdsdev) ||
6349 error "tunefs --dryrun failed"
6350 params=${params##*Parameters:}
6351 params=${params%%exiting*}
6353 error "all on-disk parameters not erased correctly via tunefs"
6355 # Check the order of options --erase-params and --param
6356 echo "tunefs --param $key=$val1 --erase-params"
6358 "$TUNEFS --param $key=$val1 --erase-params $mdsdev >/dev/null"||
6359 error "tunefs --param $key=$val1 --erase-params failed"
6360 params=$(do_facet mds $TUNEFS --dryrun $mdsdev) ||
6361 error "tunefs --dryrun failed"
6362 params=${params##*Parameters:}
6363 params=${params%%exiting*}
6364 [ $(echo $params | tr ' ' '\n') == "$key=$val1" ] ||
6365 error "on-disk param not added correctly with --erase-params"
6369 run_test 89 "check tunefs --param and --erase-param{s} options"
6372 # $2 (optional) value of max_mod_rpcs_in_flight to set
6373 check_max_mod_rpcs_in_flight() {
6381 idx=$(printf "%04x" $($LFS getdirstripe -i $dir))
6382 facet="mds$((0x$idx + 1))"
6384 if [ -z "$mmr" ]; then
6385 # get value of max_mod_rcps_in_flight
6386 mmr=$($LCTL get_param -n \
6387 mdc.$FSNAME-MDT$idx-mdc-*.max_mod_rpcs_in_flight) ||
6388 error "Unable to get max_mod_rpcs_in_flight"
6389 echo "max_mod_rcps_in_flight is $mmr"
6391 # set value of max_mod_rpcs_in_flight
6393 mdc.$FSNAME-MDT$idx-mdc-*.max_mod_rpcs_in_flight=$mmr ||
6394 error "Unable to set max_mod_rpcs_in_flight to $mmr"
6395 echo "max_mod_rpcs_in_flight set to $mmr"
6398 # create mmr+1 files
6399 echo "creating $((mmr + 1)) files ..."
6401 for i in $(seq $((mmr + 1))); do
6407 # consumes mmr-1 modify RPC slots
6408 #define OBD_FAIL_MDS_REINT_MULTI_NET 0x159
6409 # drop requests on MDT so that RPC slots are consumed
6410 # during all the request resend interval
6411 do_facet $facet "$LCTL set_param fail_loc=0x159"
6412 echo "launch $((mmr - 1)) chmod in parallel ..."
6413 for i in $(seq $((mmr - 1))); do
6414 chmod 0600 $dir/file-$i &
6418 # send one additional modify RPC
6419 do_facet $facet "$LCTL set_param fail_loc=0"
6420 echo "launch 1 additional chmod in parallel ..."
6421 chmod 0600 $dir/file-$mmr &
6424 # check this additional modify RPC get a modify RPC slot
6425 # and succeed its operation
6426 checkstat -vp 0600 $dir/file-$mmr ||
6427 error "Unable to send $mmr modify RPCs in parallel"
6432 # consumes mmr modify RPC slots
6433 #define OBD_FAIL_MDS_REINT_MULTI_NET 0x159
6434 # drop requests on MDT so that RPC slots are consumed
6435 # during all the request resend interval
6436 do_facet $facet "$LCTL set_param fail_loc=0x159"
6437 echo "launch $mmr chmod in parallel ..."
6438 for i in $(seq $mmr); do
6439 chmod 0666 $dir/file-$i &
6443 # send one additional modify RPC
6444 do_facet $facet "$LCTL set_param fail_loc=0"
6445 echo "launch 1 additional chmod in parallel ..."
6446 chmod 0666 $dir/file-$((mmr + 1)) &
6449 # check this additional modify RPC blocked getting a modify RPC slot
6450 checkstat -vp 0644 $dir/file-$((mmr + 1)) ||
6451 error "Unexpectedly send $(($mmr + 1)) modify RPCs in parallel"
6457 if ! combined_mgs_mds ; then
6462 [[ $($LCTL get_param mdc.*.import |
6463 grep "connect_flags:.*multi_mod_rpc") ]] ||
6464 { skip "Need MDC with 'multi_mod_rpcs' feature"; return 0; }
6466 # check default value
6467 $LFS mkdir -c1 $DIR/$tdir || error "mkdir $DIR/$tdir failed"
6468 check_max_mod_rpcs_in_flight $DIR/$tdir
6472 run_test 90a "check max_mod_rpcs_in_flight is enforced"
6482 [[ $($LCTL get_param mdc.*.import |
6483 grep "connect_flags:.*multi_mod_rpc") ]] ||
6484 { skip "Need MDC with 'multi_mod_rpcs' feature"; return 0; }
6487 # update max_mod_rpcs_in_flight
6488 $LFS mkdir -c1 $DIR/${tdir}1 || error "mkdir $DIR/${tdir}1 failed"
6489 check_max_mod_rpcs_in_flight $DIR/${tdir}1 1
6492 # check client is able to send multiple modify RPCs in paralell
6493 tmp=$($LCTL get_param -n mdc.$FSNAME-MDT*-mdc-*.import |
6494 grep -c "multi_mod_rpcs")
6495 if [ "$tmp" -ne $MDSCOUNT ]; then
6496 echo "Client not able to send multiple modify RPCs in parallel"
6501 # update max_mod_rpcs_in_flight
6502 $LFS mkdir -c1 $DIR/${tdir}2 || error "mkdir $DIR/${tdir}2 failed"
6503 check_max_mod_rpcs_in_flight $DIR/${tdir}2 5
6506 $LFS mkdir -c1 $DIR/${tdir}3 || error "mkdir $DIR/${tdir}3 failed"
6507 idx=$(printf "%04x" $($LFS getdirstripe -i $DIR/${tdir}3))
6508 facet="mds$((0x$idx + 1))"
6510 # save MDT max_mod_rpcs_per_client
6511 mmrpc=$(do_facet $facet \
6512 cat /sys/module/mdt/parameters/max_mod_rpcs_per_client)
6514 # update max_mod_rpcs_in_flight
6515 umount_client $MOUNT
6517 "echo 16 > /sys/module/mdt/parameters/max_mod_rpcs_per_client"
6519 $LCTL set_param mdc.$FSNAME-MDT$idx-mdc-*.max_rpcs_in_flight=17
6520 check_max_mod_rpcs_in_flight $DIR/${tdir}3 16
6522 # restore MDT max_mod_rpcs_per_client initial value
6524 "echo $mmrpc > /sys/module/mdt/parameters/max_mod_rpcs_per_client"
6526 rm -rf $DIR/${tdir}?
6529 run_test 90b "check max_mod_rpcs_in_flight is enforced after update"
6538 [[ $($LCTL get_param mdc.*.import |
6539 grep "connect_flags:.*multi_mod_rpc") ]] ||
6540 { skip "Need MDC with 'multi_mod_rpcs' feature"; return 0; }
6542 # check client is able to send multiple modify RPCs in paralell
6543 tmp=$($LCTL get_param -n mdc.$FSNAME-MDT*-mdc-*.import |
6544 grep -c "multi_mod_rpcs")
6545 if [ "$tmp" -ne $MDSCOUNT ]; then
6546 skip "Client not able to send multiple modify RPCs in parallel"
6551 # get max_rpcs_in_flight value
6552 mrif=$($LCTL get_param -n mdc.$FSNAME-MDT0000-mdc-*.max_rpcs_in_flight)
6553 echo "max_rpcs_in_flight is $mrif"
6555 # get MDT max_mod_rpcs_per_client
6556 mmrpc=$(do_facet mds1 \
6557 cat /sys/module/mdt/parameters/max_mod_rpcs_per_client)
6558 echo "max_mod_rpcs_per_client is $mmrpc"
6561 # attempt to set max_mod_rpcs_in_flight to max_rpcs_in_flight value
6562 # prerequisite: set max_mod_rpcs_per_client to max_rpcs_in_flight value
6563 umount_client $MOUNT
6565 "echo $mrif > /sys/module/mdt/parameters/max_mod_rpcs_per_client"
6569 mdc.$FSNAME-MDT0000-mdc-*.max_mod_rpcs_in_flight=$mrif &&
6570 error "set max_mod_rpcs_in_flight to $mrif should fail"
6572 umount_client $MOUNT
6574 "echo $mmrpc > /sys/module/mdt/parameters/max_mod_rpcs_per_client"
6578 # attempt to set max_mod_rpcs_in_flight to max_mod_rpcs_per_client+1
6579 # prerequisite: set max_rpcs_in_flight to max_mod_rpcs_per_client+2
6581 mdc.$FSNAME-MDT0000-mdc-*.max_rpcs_in_flight=$((mmrpc + 2))
6584 mdc.$FSNAME-MDT0000-mdc-*.max_mod_rpcs_in_flight=$((mmrpc + 1)) &&
6585 error "set max_mod_rpcs_in_flight to $((mmrpc + 1)) should fail"
6589 run_test 90c "check max_mod_rpcs_in_flight update limits"
6600 [[ $($LCTL get_param mdc.*.import |
6601 grep "connect_flags:.*multi_mod_rpc") ]] ||
6602 { skip "Need MDC with 'multi_mod_rpcs' feature"; return 0; }
6604 $LFS mkdir -c1 $DIR/$tdir || error "mkdir $DIR/$tdir failed"
6605 idx=$(printf "%04x" $($LFS getdirstripe -i $DIR/$tdir))
6606 facet="mds$((0x$idx + 1))"
6608 # check client version supports multislots
6609 tmp=$($LCTL get_param -N \
6610 mdc.$FSNAME-MDT$idx-mdc-*.max_mod_rpcs_in_flight)
6611 if [ -z "$tmp" ]; then
6612 skip "Client does not support multiple modify RPCs in flight"
6617 # get current value of max_mod_rcps_in_flight
6618 mmr=$($LCTL get_param -n \
6619 mdc.$FSNAME-MDT$idx-mdc-*.max_mod_rpcs_in_flight)
6620 echo "max_mod_rcps_in_flight is $mmr"
6623 echo "creating $mmr files ..."
6625 for i in $(seq $mmr); do
6626 touch $DIR/$tdir/file-$i
6629 # prepare for close RPC
6630 multiop_bg_pause $DIR/$tdir/file-close O_c
6633 # consumes mmr modify RPC slots
6634 #define OBD_FAIL_MDS_REINT_MULTI_NET 0x159
6635 # drop requests on MDT so that RPC slots are consumed
6636 # during all the request resend interval
6637 do_facet $facet "$LCTL set_param fail_loc=0x159"
6638 echo "launch $mmr chmod in parallel ..."
6639 for i in $(seq $mmr); do
6640 chmod 0600 $DIR/$tdir/file-$i &
6643 # send one additional close RPC
6644 do_facet $facet "$LCTL set_param fail_loc=0"
6645 echo "launch 1 additional close in parallel ..."
6647 cancel_lru_locks mdc
6650 # check this additional close RPC get a modify RPC slot
6651 # and multiop process completed
6652 [ -d /proc/$pid ] &&
6653 error "Unable to send the additional close RPC in parallel"
6658 run_test 90d "check one close RPC is allowed above max_mod_rpcs_in_flight"
6660 check_uuid_on_ost() {
6662 do_facet ost1 "$LCTL get_param obdfilter.${FSNAME}*.exports.'$nid'.uuid"
6665 check_uuid_on_mdt() {
6667 do_facet $SINGLEMDS "$LCTL get_param mdt.${FSNAME}*.exports.'$nid'.uuid"
6675 [[ $(lustre_version_code ost1) -ge $(version_code 2.7.63) ]] ||
6676 { skip "Need OST version at least 2.7.63" && return 0; }
6677 [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.7.63) ]] ||
6678 { skip "Need MDT version at least 2.7.63" && return 0; }
6680 start_mds || error "MDS start failed"
6681 start_ost || error "unable to start OST"
6682 mount_client $MOUNT || error "client start failed"
6683 check_mount || error "check_mount failed"
6686 nid=$($LCTL list_nids | head -1 | sed "s/\./\\\./g")
6690 uuid=$(get_client_uuid $MOUNT)
6692 echo "list nids on mdt:"
6693 do_facet $SINGLEMDS "$LCTL list_param mdt.${FSNAME}*.exports.*"
6694 echo "uuid from $nid:"
6695 do_facet $SINGLEMDS "$LCTL get_param mdt.${FSNAME}*.exports.'$nid'.uuid"
6697 found=$(check_uuid_on_mdt $nid | grep $uuid)
6698 [ -z "$found" ] && error "can't find $uuid $nid on MDT"
6699 found=$(check_uuid_on_ost $nid | grep $uuid)
6700 [ -z "$found" ] && error "can't find $uuid $nid on OST"
6702 # umount the client so it won't reconnect
6703 manual_umount_client --force || error "failed to umount $?"
6704 # shouldn't disappear on MDS after forced umount
6705 found=$(check_uuid_on_mdt $nid | grep $uuid)
6706 [ -z "$found" ] && error "can't find $uuid $nid"
6709 do_facet $SINGLEMDS \
6710 "$LCTL set_param -n mdt.${mds1_svc}.evict_client nid:$nid"
6712 found=$(check_uuid_on_mdt $nid | grep $uuid)
6713 [ -n "$found" ] && error "found $uuid $nid on MDT"
6714 found=$(check_uuid_on_ost $nid | grep $uuid)
6715 [ -n "$found" ] && error "found $uuid $nid on OST"
6717 # check it didn't reconnect (being umounted)
6718 sleep $((TIMEOUT+1))
6719 found=$(check_uuid_on_mdt $nid | grep $uuid)
6720 [ -n "$found" ] && error "found $uuid $nid on MDT"
6721 found=$(check_uuid_on_ost $nid | grep $uuid)
6722 [ -n "$found" ] && error "found $uuid $nid on OST"
6726 run_test 91 "evict-by-nid support"
6728 generate_ldev_conf() {
6729 # generate an ldev.conf file
6730 local ldevconfpath=$1
6732 local fsldevformat=""
6735 fstype=$(facet_fstype mgs)
6736 if [ "$fstype" == "zfs" ]; then
6737 fsldevformat="$fstype:"
6742 printf "%s\t-\t%s-MGS0000\t%s%s\n" \
6746 $(mgsdevname) > $ldevconfpath
6748 local mdsfo_host=$mdsfailover_HOST;
6749 if [ -z "$mdsfo_host" ]; then
6753 for num in $(seq $MDSCOUNT); do
6754 fstype=$(facet_fstype mds$num)
6755 if [ "$fstype" == "zfs" ]; then
6756 fsldevformat="$fstype:"
6761 printf "%s\t%s\t%s-MDT%04d\t%s%s\n" \
6767 $(mdsdevname $num) >> $ldevconfpath
6770 local ostfo_host=$ostfailover_HOST;
6771 if [ -z "$ostfo_host" ]; then
6775 for num in $(seq $OSTCOUNT); do
6776 fstype=$(facet_fstype ost$num)
6777 if [ "$fstype" == "zfs" ]; then
6778 fsldevformat="$fstype:"
6783 printf "%s\t%s\t%s-OST%04d\t%s%s\n" \
6789 $(ostdevname $num) >> $ldevconfpath
6792 echo "----- $ldevconfpath -----"
6794 echo "--- END $ldevconfpath ---"
6799 # generate a nids file (mapping between hostname to nid)
6800 # looks like we only have the MGS nid available to us
6801 # so just echo that to a file
6803 echo -e "${mgs_HOST}\t${MGSNID}" > $nidspath
6805 echo "----- $nidspath -----"
6807 echo "--- END $nidspath ---"
6810 compare_ldev_output() {
6814 sort $expected_output -o $expected_output
6815 sort $ldev_output -o $ldev_output
6817 echo "-- START OF LDEV OUTPUT --"
6819 echo "--- END OF LDEV OUTPUT ---"
6821 echo "-- START OF EXPECTED OUTPUT --"
6822 cat $expected_output
6823 echo "--- END OF EXPECTED OUTPUT ---"
6825 diff $expected_output $ldev_output
6830 if [ -z "$LDEV" ]; then
6831 error "ldev is missing!"
6834 local LDEVCONFPATH=$TMP/ldev.conf
6835 local NIDSPATH=$TMP/nids
6837 echo "Host is $(hostname)"
6839 generate_ldev_conf $LDEVCONFPATH
6840 generate_nids $NIDSPATH
6842 # echo the mgs nid and compare it to environment variable MGSNID
6843 # also, ldev.conf and nids is a server side thing, use the OSS
6846 output=$($LDEV -c $LDEVCONFPATH -H $ost_HOST -n $NIDSPATH echo %m)
6848 echo "-- START OF LDEV OUTPUT --"
6850 echo "--- END OF LDEV OUTPUT ---"
6852 # ldev failed, error
6853 if [ $? -ne 0 ]; then
6854 rm $LDEVCONFPATH $NIDSPATH
6855 error "ldev failed to execute!"
6858 # need to process multiple lines because of combined MGS and MDS
6859 echo -e $output | awk '{ print $2 }' | while read -r line ; do
6860 if [ "$line" != "$MGSNID" ]; then
6861 rm $LDEVCONFPATH $NIDSPATH
6862 error "ldev failed mgs nid '$line', expected '$MGSNID'"
6866 rm $LDEVCONFPATH $NIDSPATH
6868 run_test 92 "ldev returns MGS NID correctly in command substitution"
6871 [ $MDSCOUNT -lt 3 ] && skip "needs >= 3 MDTs" && return
6874 #start mgs or mgs/mdt0
6875 if ! combined_mgs_mds ; then
6882 start_ost || error "OST0 start fail"
6884 #define OBD_FAIL_MGS_WRITE_TARGET_DELAY 0x90e
6885 do_facet mgs "$LCTL set_param fail_val = 10 fail_loc=0x8000090e"
6886 for num in $(seq 2 $MDSCOUNT); do
6890 mount_client $MOUNT || error "mount client fails"
6891 wait_osc_import_state mds ost FULL
6892 wait_osc_import_state client ost FULL
6893 check_mount || error "check_mount failed"
6895 cleanup || error "cleanup failed with $?"
6897 run_test 93 "register mulitple MDT at the same time"
6900 if [ -z "$LDEV" ]; then
6901 error "ldev is missing!"
6904 local LDEVCONFPATH=$TMP/ldev.conf
6905 local NIDSPATH=$TMP/nids
6907 generate_ldev_conf $LDEVCONFPATH
6908 generate_nids $NIDSPATH
6910 local LDEV_OUTPUT=$TMP/ldev-output.txt
6911 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -F $FSNAME > $LDEV_OUTPUT
6913 # ldev failed, error
6914 if [ $? -ne 0 ]; then
6915 rm $LDEVCONFPATH $NIDSPATH $LDEV_OUTPUT
6916 error "ldev failed to execute!"
6920 local EXPECTED_OUTPUT=$TMP/ldev-expected.txt
6922 printf "%s-MGS0000\n" $FSNAME > $EXPECTED_OUTPUT
6924 for num in $(seq $MDSCOUNT); do
6925 printf "%s-MDT%04d\n" $FSNAME $num >> $EXPECTED_OUTPUT
6928 for num in $(seq $OSTCOUNT); do
6929 printf "%s-OST%04d\n" $FSNAME $num >> $EXPECTED_OUTPUT
6932 compare_ldev_output $LDEV_OUTPUT $EXPECTED_OUTPUT
6934 if [ $? -ne 0 ]; then
6935 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
6936 error "ldev failed to produce the correct hostlist!"
6939 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
6941 run_test 94 "ldev outputs correct labels for file system name query"
6944 if [ -z "$LDEV" ]; then
6945 error "ldev is missing!"
6948 local LDEVCONFPATH=$TMP/ldev.conf
6949 local NIDSPATH=$TMP/nids
6951 generate_ldev_conf $LDEVCONFPATH
6952 generate_nids $NIDSPATH
6956 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -F $FSNAME &>/dev/null
6957 if [ $? -ne 0 ]; then
6958 rm $LDEVCONFPATH $NIDSPATH
6959 error "ldev label filtering w/ -F failed!"
6963 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -l &>/dev/null
6964 if [ $? -ne 0 ]; then
6965 rm $LDEVCONFPATH $NIDSPATH
6966 error "ldev label filtering w/ -l failed!"
6970 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -f &>/dev/null
6971 if [ $? -ne 0 ]; then
6972 rm $LDEVCONFPATH $NIDSPATH
6973 error "ldev label filtering w/ -f failed!"
6977 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -a &>/dev/null
6978 if [ $? -ne 0 ]; then
6979 rm $LDEVCONFPATH $NIDSPATH
6980 error "ldev label filtering w/ -a failed!"
6985 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -a -F $FSNAME &>/dev/null
6986 if [ $? -eq 0 ]; then
6987 rm $LDEVCONFPATH $NIDSPATH
6988 error "ldev label filtering w/ -a and -F incorrectly succeeded"
6992 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -a -f &>/dev/null
6993 if [ $? -eq 0 ]; then
6994 rm $LDEVCONFPATH $NIDSPATH
6995 error "ldev label filtering w/ -a and -f incorrectly succeeded"
6999 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -a -l &>/dev/null
7000 if [ $? -eq 0 ]; then
7001 rm $LDEVCONFPATH $NIDSPATH
7002 error "ldev label filtering w/ -a and -l incorrectly succeeded"
7006 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -f -l &>/dev/null
7007 if [ $? -eq 0 ]; then
7008 rm $LDEVCONFPATH $NIDSPATH
7009 error "ldev label filtering w/ -f and -l incorrectly succeeded"
7013 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -F $FSNAME -l &>/dev/null
7014 if [ $? -eq 0 ]; then
7015 rm $LDEVCONFPATH $NIDSPATH
7016 error "ldev label filtering w/ -F and -l incorrectly succeeded"
7019 # file sys & foreign
7020 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -F $FSNAME -f &>/dev/null
7021 if [ $? -eq 0 ]; then
7022 rm $LDEVCONFPATH $NIDSPATH
7023 error "ldev label filtering w/ -F and -f incorrectly succeeded"
7026 rm $LDEVCONFPATH $NIDSPATH
7028 run_test 95 "ldev should only allow one label filter"
7031 if [ -z "$LDEV" ]; then
7032 error "ldev is missing!"
7035 local LDEVCONFPATH=$TMP/ldev.conf
7036 local NIDSPATH=$TMP/nids
7038 generate_ldev_conf $LDEVCONFPATH
7039 generate_nids $NIDSPATH
7041 local LDEV_OUTPUT=$TMP/ldev-output.txt
7042 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -H $mgs_HOST \
7044 awk '{print $2}' > $LDEV_OUTPUT
7046 # ldev failed, error
7047 if [ $? -ne 0 ]; then
7048 rm $LDEVCONFPATH $NIDSPATH $LDEV_OUTPUT
7049 error "ldev failed to execute!"
7053 local EXPECTED_OUTPUT=$TMP/ldev-expected-output.txt
7055 echo "$mgs_HOST-$(facet_fstype mgs)" > $EXPECTED_OUTPUT
7057 if [ "$mgs_HOST" == "$mds_HOST" ]; then
7058 for num in $(seq $MDSCOUNT); do
7059 echo "$mds_HOST-$(facet_fstype mds$num)" \
7064 if [ "$mgs_HOST" == "$ost_HOST" ]; then
7065 for num in $(seq $OSTCOUNT); do
7066 echo "$ost_HOST-$(facet_fstype ost$num)" \
7071 compare_ldev_output $LDEV_OUTPUT $EXPECTED_OUTPUT
7073 if [ $? -ne 0 ]; then
7074 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
7075 error "ldev failed to produce the correct output!"
7078 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
7080 run_test 96 "ldev returns hostname and backend fs correctly in command sub"
7083 if [ -z "$LDEV" ]; then
7084 error "ldev is missing!"
7087 local LDEVCONFPATH=$TMP/ldev.conf
7088 local NIDSPATH=$TMP/nids
7090 generate_ldev_conf $LDEVCONFPATH
7091 generate_nids $NIDSPATH
7093 local LDEV_OUTPUT=$TMP/ldev-output.txt
7094 local EXPECTED_OUTPUT=$TMP/ldev-expected-output.txt
7096 echo -e "\nMDT role"
7097 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -F $FSNAME -R mdt > $LDEV_OUTPUT
7099 if [ $? -ne 0 ]; then
7100 rm $LDEVCONFPATH $NIDSPATH $LDEV_OUTPUT
7101 error "ldev failed to execute for mdt role!"
7104 for num in $(seq $MDSCOUNT); do
7105 printf "%s-MDT%04d\n" $FSNAME $num >> $EXPECTED_OUTPUT
7108 compare_ldev_output $LDEV_OUTPUT $EXPECTED_OUTPUT
7110 if [ $? -ne 0 ]; then
7111 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
7112 error "ldev failed to produce the correct output for mdt role!"
7115 echo -e "\nOST role"
7116 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -F $FSNAME -R ost > $LDEV_OUTPUT
7118 if [ $? -ne 0 ]; then
7119 rm $LDEVCONFPATH $NIDSPATH $LDEV_OUTPUT $EXPECTED_OUTPUT
7120 error "ldev failed to execute for ost role!"
7124 for num in $(seq $OSTCOUNT); do
7125 printf "%s-OST%04d\n" $FSNAME $num >> $EXPECTED_OUTPUT
7128 compare_ldev_output $LDEV_OUTPUT $EXPECTED_OUTPUT
7130 if [ $? -ne 0 ]; then
7131 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
7132 error "ldev failed to produce the correct output for ost role!"
7135 echo -e "\nMGS role"
7136 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -F $FSNAME -R mgs > $LDEV_OUTPUT
7138 if [ $? -ne 0 ]; then
7139 rm $LDEVCONFPATH $NIDSPATH $LDEV_OUTPUT $EXPECTED_OUTPUT
7140 error "ldev failed to execute for mgs role!"
7143 printf "%s-MGS0000\n" $FSNAME > $EXPECTED_OUTPUT
7145 compare_ldev_output $LDEV_OUTPUT $EXPECTED_OUTPUT
7147 if [ $? -ne 0 ]; then
7148 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
7149 error "ldev failed to produce the correct output for mgs role!"
7152 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
7154 run_test 97 "ldev returns correct ouput when querying based on role"
7159 local temp=$MDS_MOUNT_OPTS
7162 check_mount || error "mount failed"
7163 mountopt="user_xattr"
7164 for ((x = 1; x <= 400; x++)); do
7165 mountopt="$mountopt,user_xattr"
7167 remount_client $mountopt $MOUNT 2>&1 | grep "too long" ||
7168 error "Buffer overflow check failed"
7169 cleanup || error "cleanup failed"
7171 run_test 98 "Buffer-overflow check while parsing mount_opts"
7175 [[ $(facet_fstype ost1) != ldiskfs ]] &&
7176 { skip "ldiskfs only test" && return; }
7177 [[ $(lustre_version_code ost1) -ge $(version_code 2.8.57) ]] ||
7178 { skip "Need OST version at least 2.8.57" && return 0; }
7180 local ost_opts="$(mkfs_opts ost1 $(ostdevname 1)) \
7181 --reformat $(ostdevname 1) $(ostvdevname 1)"
7182 do_facet ost1 $DEBUGFS -c -R stats `ostdevname 1` | grep "meta_bg" &&
7183 skip "meta_bg already set" && return
7186 if [[ ${!opts} != *mkfsoptions* ]]; then
7187 eval opts=\"${!opts} \
7188 --mkfsoptions='\\\"-O ^resize_inode,meta_bg\\\"'\"
7190 local val=${!opts//--mkfsoptions=\\\"/ \
7191 --mkfsoptions=\\\"-O ^resize_inode,meta_bg }
7195 echo "params: $opts"
7197 add ost1 $opts || error "add ost1 failed with new params"
7199 do_facet ost1 $DEBUGFS -c -R stats `ostdevname 1` | grep "meta_bg" ||
7200 error "meta_bg is not set"
7204 run_test 99 "Adding meta_bg option"
7208 start_mds || error "MDS start failed"
7209 start_ost || error "unable to start OST"
7210 mount_client $MOUNT || error "client start failed"
7211 check_mount || error "check_mount failed"
7220 do_facet mgs 'lshowmount -v' | awk 'BEGIN {NR == 0; rc=1} /MGS:/ {rc=0}
7221 END {exit rc}' || error "lshowmount have no output MGS"
7223 do_facet mds1 'lshowmount -v' | awk 'BEGIN {NR == 2; rc=1} /-MDT0000:/
7224 {rc=0} END {exit rc}' || error "lshowmount have no output MDT0"
7226 do_facet ost1 'lshowmount -v' | awk 'BEGIN {NR == 4; rc=1} /-OST0000:/
7227 {rc=0} END {exit rc}' || error "lshowmount have no output OST0"
7229 cleanup || error "cleanup failed with $?"
7231 run_test 100 "check lshowmount lists MGS, MDT, OST and 0@lo"
7234 local createmany_pid
7235 local dev=$FSNAME-OST0000-osc-MDT0000
7239 createmany -o $DIR1/$tdir/$tfile-%d 50000 &
7241 # MDT->OST reconnection causes MDT<->OST last_id synchornisation
7242 # via osp_precreate_cleanup_orphans.
7243 for ((i = 0; i < 100; i++)); do
7244 for ((k = 0; k < 10; k++)); do
7245 do_facet $SINGLEMDS "$LCTL --device $dev deactivate;" \
7246 "$LCTL --device $dev activate"
7249 ls -asl $MOUNT | grep '???' &&
7250 { kill -9 $createmany_pid &>/dev/null;
7251 error "File has no object on OST"; }
7253 kill -s 0 $createmany_pid || break
7255 wait $createmany_pid
7257 unlinkmany $DIR1/$tdir/$tfile-%d 50000
7260 run_test 101 "Race MDT->OST reconnection with create"
7263 [[ $(lustre_version_code $SINGLEMDS) -gt $(version_code 2.9.53) ]] ||
7264 { skip "Need server version greater than 2.9.53" && return 0; }
7265 cleanup || error "cleanup failed with $?"
7267 local mds1dev=$(mdsdevname 1)
7268 local mds1mnt=$(facet_mntpt mds1)
7269 local mds1fstype=$(facet_fstype mds1)
7270 local mds1opts=$MDS_MOUNT_OPTS
7272 if [ $mds1fstype == ldiskfs ] &&
7273 ! do_facet mds1 test -b $mds1dev; then
7274 mds1opts=$(csa_add "$mds1opts" -o loop)
7276 if [[ $mds1fstype == zfs ]]; then
7277 import_zpool mds1 || return ${PIPESTATUS[0]}
7280 # unload all and only load libcfs to allow fail_loc setting
7281 do_facet mds1 lustre_rmmod || error "unable to unload modules"
7282 do_facet mds1 modprobe libcfs || error "libcfs not loaded"
7283 do_facet mds1 lsmod \| grep libcfs || error "libcfs not loaded"
7285 #define OBD_FAIL_OBDCLASS_MODULE_LOAD 0x60a
7286 do_facet mds1 "$LCTL set_param fail_loc=0x8000060a"
7288 do_facet mds1 $MOUNT_CMD $mds1dev $mds1mnt $mds1opts &&
7289 error "mdt start must fail"
7290 do_facet mds1 lsmod \| grep obdclass && error "obdclass must not load"
7292 do_facet mds1 "$LCTL set_param fail_loc=0x0"
7294 do_facet mds1 $MOUNT_CMD $mds1dev $mds1mnt $mds1opts ||
7295 error "mdt start must not fail"
7297 cleanup || error "cleanup failed with $?"
7299 run_test 102 "obdclass module cleanup upon error"
7304 echo "rename $FSNAME to $newname"
7306 if ! combined_mgs_mds ; then
7307 local facet=$(mgsdevname)
7310 "$TUNEFS --fsname=$newname --rename=$FSNAME -v $facet"||
7311 error "(7) Fail to rename MGS"
7312 if [ "$(facet_fstype $facet)" = "zfs" ]; then
7313 reimport_zpool mgs $newname-mgs
7317 for num in $(seq $MDSCOUNT); do
7318 local facet=$(mdsdevname $num)
7320 do_facet mds${num} \
7321 "$TUNEFS --fsname=$newname --rename=$FSNAME -v $facet"||
7322 error "(8) Fail to rename MDT $num"
7323 if [ "$(facet_fstype $facet)" = "zfs" ]; then
7324 reimport_zpool mds${num} $newname-mdt${num}
7328 for num in $(seq $OSTCOUNT); do
7329 local facet=$(ostdevname $num)
7331 do_facet ost${num} \
7332 "$TUNEFS --fsname=$newname --rename=$FSNAME -v $facet"||
7333 error "(9) Fail to rename OST $num"
7334 if [ "$(facet_fstype $facet)" = "zfs" ]; then
7335 reimport_zpool ost${num} $newname-ost${num}
7340 test_103_set_pool() {
7344 do_facet mgs $LCTL pool_add $FSNAME.$pname ${FSNAME}-$ost_x ||
7345 error "Fail to add $ost_x to $FSNAME.$pname"
7346 wait_update $HOSTNAME \
7347 "lctl get_param -n lov.$FSNAME-clilov-*.pools.$pname |
7348 grep $ost_x" "$FSNAME-${ost_x}_UUID" ||
7349 error "$ost_x is NOT in pool $FSNAME.$pname"
7352 test_103_check_pool() {
7353 local save_fsname=$1
7356 stat $DIR/$tdir/test-framework.sh ||
7357 error "($errno) Fail to stat"
7358 do_facet mgs $LCTL pool_list $FSNAME.pool1 ||
7359 error "($errno) Fail to list $FSNAME.pool1"
7360 do_facet mgs $LCTL pool_list $FSNAME.$save_fsname ||
7361 error "($errno) Fail to list $FSNAME.$save_fsname"
7362 do_facet mgs $LCTL pool_list $FSNAME.$save_fsname |
7363 grep ${FSNAME}-OST0000 ||
7364 error "($errno) List $FSNAME.$save_fsname is invalid"
7366 local pname=$($LFS getstripe --pool $DIR/$tdir/d0)
7367 [ "$pname" = "$save_fsname" ] ||
7368 error "($errno) Unexpected pool name $pname"
7372 check_mount_and_prep
7374 mkdir $DIR/$tdir || error "(1) Fail to mkdir $DIR/$tdir"
7375 cp $LUSTRE/tests/test-framework.sh $DIR/$tdir ||
7376 error "(2) Fail to copy test-framework.sh"
7378 if ! combined_mgs_mds ; then
7381 do_facet mgs $LCTL pool_new $FSNAME.pool1 ||
7382 error "(3) Fail to create $FSNAME.pool1"
7383 # name the pool name as the fsname
7384 do_facet mgs $LCTL pool_new $FSNAME.$FSNAME ||
7385 error "(4) Fail to create $FSNAME.$FSNAME"
7387 test_103_set_pool $FSNAME OST0000
7389 $SETSTRIPE -p $FSNAME $DIR/$tdir/d0 ||
7390 error "(6) Fail to setstripe on $DIR/$tdir/d0"
7392 if ! combined_mgs_mds ; then
7398 test_renamefs mylustre
7400 local save_fsname=$FSNAME
7404 if ! combined_mgs_mds ; then
7407 test_103_check_pool $save_fsname 7
7409 if [ $OSTCOUNT -ge 2 ]; then
7410 test_103_set_pool $save_fsname OST0001
7413 $SETSTRIPE -p $save_fsname $DIR/$tdir/f0 ||
7414 error "(16) Fail to setstripe on $DIR/$tdir/f0"
7415 if ! combined_mgs_mds ; then
7426 if ! combined_mgs_mds ; then
7429 test_103_check_pool $save_fsname 17
7431 if ! combined_mgs_mds ; then
7436 test_renamefs $save_fsname
7442 run_test 103 "rename filesystem name"
7444 test_104() { # LU-6952
7445 local mds_mountopts=$MDS_MOUNT_OPTS
7446 local ost_mountopts=$OST_MOUNT_OPTS
7447 local mds_mountfsopts=$MDS_MOUNT_FS_OPTS
7448 local lctl_ver=$(do_facet $SINGLEMDS $LCTL --version |
7451 [[ $(version_code $lctl_ver) -lt $(version_code 2.9.55) ]] &&
7452 { skip "this test needs utils above 2.9.55" && return 0; }
7454 # specify "acl" in mount options used by mkfs.lustre
7455 if [ -z "$MDS_MOUNT_FS_OPTS" ]; then
7456 MDS_MOUNT_FS_OPTS="acl,user_xattr"
7459 MDS_MOUNT_FS_OPTS="${MDS_MOUNT_FS_OPTS},acl,user_xattr"
7462 echo "mountfsopt: $MDS_MOUNT_FS_OPTS"
7464 #reformat/remount the MDT to apply the MDT_MOUNT_FS_OPT options
7466 if [ -z "$MDS_MOUNT_OPTS" ]; then
7467 MDS_MOUNT_OPTS="-o noacl"
7469 MDS_MOUNT_OPTS="${MDS_MOUNT_OPTS},noacl"
7472 for num in $(seq $MDSCOUNT); do
7473 start mds$num $(mdsdevname $num) $MDS_MOUNT_OPTS ||
7474 error "Failed to start MDS"
7477 for num in $(seq $OSTCOUNT); do
7478 start ost$num $(ostdevname $num) $OST_MOUNT_OPTS ||
7479 error "Failed to start OST"
7483 setfacl -m "d:$RUNAS_ID:rwx" $MOUNT &&
7484 error "ACL is applied when FS is mounted with noacl."
7486 MDS_MOUNT_OPTS=$mds_mountopts
7487 OST_MOUNT_OPTS=$ost_mountopts
7488 MDS_MOUNT_FS_OPTS=$mds_mountfsopts
7493 run_test 104 "Make sure user defined options are reflected in mount"
7495 error_and_umount() {
7506 mount --bind $DIR $TMP/$tdir || error "mount bind mnt pt failed"
7507 rm -f $TMP/$tdir/$tfile
7508 rm -f $TMP/$tdir/${tfile}1
7510 # Files should not be created in ro bind mount point
7511 # remounting from rw to ro
7512 mount -o remount,ro $TMP/$tdir ||
7513 error_and_umount "readonly remount of bind mnt pt failed"
7514 touch $TMP/$tdir/$tfile &&
7515 error_and_umount "touch succeeds on ro bind mnt pt"
7516 [ -e $TMP/$tdir/$tfile ] &&
7517 error_and_umount "file created on ro bind mnt pt"
7519 # Files should be created in rw bind mount point
7520 # remounting from ro to rw
7521 mount -o remount,rw $TMP/$tdir ||
7522 error_and_umount "read-write remount of bind mnt pt failed"
7523 touch $TMP/$tdir/${tfile}1 ||
7524 error_and_umount "touch fails on rw bind mnt pt"
7525 [ -e $TMP/$tdir/${tfile}1 ] ||
7526 error_and_umount "file not created on rw bind mnt pt"
7527 umount $TMP/$tdir || error "umount of bind mnt pt failed"
7529 cleanup || error "cleanup failed with $?"
7531 run_test 105 "check file creation for ro and rw bind mnt pt"
7538 mkdir -p $DIR/$tdir || error "create $tdir failed"
7539 lfs setstripe -c 1 -i 0 $DIR/$tdir
7540 #define OBD_FAIL_CAT_RECORDS 0x1312
7541 do_facet mds1 $LCTL set_param fail_loc=0x1312 fail_val=$repeat
7543 for ((i = 1; i <= $repeat; i++)); do
7545 #one full plain llog
7546 createmany -o $DIR/$tdir/f- 64768
7548 createmany -u $DIR/$tdir/f- 64768
7550 wait_delete_completed $((TIMEOUT * 7))
7551 #ASSERTION osp_sync_thread() ( thread->t_flags != SVC_RUNNING ) failed
7552 #shows that osp code is buggy
7553 do_facet mds1 $LCTL set_param fail_loc=0 fail_val=0
7557 run_test 106 "check osp llog processing when catalog is wrapped"
7560 [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.10.50) ]] ||
7561 { skip "Need MDS version > 2.10.50"; return; }
7563 start_mgsmds || error "start_mgsmds failed"
7564 start_ost || error "unable to start OST"
7566 # add unknown configuration parameter.
7567 local PARAM="$FSNAME-OST0000.ost.unknown_param=50"
7568 do_facet mgs "$LCTL conf_param $PARAM"
7569 cleanup_nocli || error "cleanup_nocli failed with $?"
7572 # unknown param should be ignored while mounting.
7573 start_ost || error "unable to start OST after unknown param set"
7575 cleanup || error "cleanup failed with $?"
7577 run_test 107 "Unknown config param should not fail target mounting"
7580 # set number of permanent parameters
7582 test_109_set_params() {
7585 set_conf_param_and_check mds \
7586 "$LCTL get_param -n mdd.$fsname-MDT0000.atime_diff" \
7587 "$fsname-MDT0000.mdd.atime_diff" \
7589 set_conf_param_and_check mds \
7590 "$LCTL get_param -n mdd.$fsname-MDT0000.atime_diff" \
7591 "$fsname-MDT0000.mdd.atime_diff" \
7593 set_conf_param_and_check client \
7594 "$LCTL get_param -n llite.$fsname*.max_read_ahead_mb" \
7595 "$fsname.llite.max_read_ahead_mb" \
7597 set_conf_param_and_check client \
7598 "$LCTL get_param -n llite.$fsname*.max_read_ahead_mb" \
7599 "$fsname.llite.max_read_ahead_mb" \
7601 create_pool $fsname.pool1 || error "create pool failed"
7602 do_facet mgs $LCTL pool_add $fsname.pool1 OST0000 ||
7603 error "pool_add failed"
7604 do_facet mgs $LCTL pool_remove $fsname.pool1 OST0000 ||
7605 error "pool_remove failed"
7606 do_facet mgs $LCTL pool_add $fsname.pool1 OST0000 ||
7607 error "pool_add failed"
7611 # check permanent parameters
7613 test_109_test_params() {
7616 local atime_diff=$(do_facet mds $LCTL \
7617 get_param -n mdd.$fsname-MDT0000.atime_diff)
7618 [ $atime_diff == 63 ] || error "wrong mdd parameter after clear_conf"
7619 local max_read_ahead_mb=$(do_facet client $LCTL \
7620 get_param -n llite.$fsname*.max_read_ahead_mb)
7621 [ $max_read_ahead_mb == 64 ] ||
7622 error "wrong llite parameter after clear_conf"
7623 local ost_in_pool=$(do_facet mds $LCTL pool_list $fsname.pool1 |
7624 grep -v "^Pool:" | sed 's/_UUID//')
7625 [ $ost_in_pool = "$fsname-OST0000" ] ||
7626 error "wrong pool after clear_conf"
7630 # run lctl clear_conf, store CONFIGS before and after that
7632 test_109_clear_conf()
7634 local clear_conf_arg=$1
7637 if ! combined_mgs_mds ; then
7639 stop_mgs || error "stop_mgs failed"
7640 start_mgs "-o nosvc" || error "start_mgs nosvc failed"
7642 mgsdev=$(mdsdevname 1)
7643 start_mdt 1 "-o nosvc" || error "start_mdt 1 nosvc failed"
7646 do_facet mgs "rm -rf $TMP/${tdir}/conf1; mkdir -p $TMP/${tdir}/conf1;" \
7647 "$DEBUGFS -c -R \\\"rdump CONFIGS $TMP/${tdir}/conf1\\\" \
7651 # the command being tested
7653 do_facet mgs $LCTL clear_conf $clear_conf_arg ||
7654 error "clear_conf failed"
7655 if ! combined_mgs_mds ; then
7656 stop_mgs || error "stop_mgs failed"
7658 stop_mdt 1 || error "stop_mdt 1 failed"
7661 do_facet mgs "rm -rf $TMP/${tdir}/conf2; mkdir -p $TMP/${tdir}/conf2;" \
7662 "$DEBUGFS -c -R \\\"rdump CONFIGS $TMP/${tdir}/conf2\\\" \
7666 test_109_file_shortened() {
7668 local sizes=($(do_facet mgs "stat -c %s " \
7669 "$TMP/${tdir}/conf1/CONFIGS/$file" \
7670 "$TMP/${tdir}/conf2/CONFIGS/$file"))
7671 [ ${sizes[1]} -lt ${sizes[0]} ] && return 0
7677 [ "$(facet_fstype mgs)" == "zfs" ] &&
7678 skip "LU-8727: no implementation for ZFS" && return
7682 client_up || error "client_up failed"
7685 # set number of permanent parameters
7687 test_109_set_params $FSNAME
7689 umount_client $MOUNT || error "umount_client failed"
7690 stop_ost || error "stop_ost failed"
7691 stop_mds || error "stop_mds failed"
7693 test_109_clear_conf $FSNAME
7695 # make sure that all configs are cleared
7697 test_109_file_shortened $FSNAME-MDT0000 ||
7698 error "failed to clear MDT0000 config"
7699 test_109_file_shortened $FSNAME-client ||
7700 error "failed to clear client config"
7705 # check that configurations are intact
7707 test_109_test_params $FSNAME
7712 destroy_test_pools || error "destroy test pools failed"
7716 run_test 109a "test lctl clear_conf fsname"
7720 [ "$(facet_fstype mgs)" == "zfs" ] &&
7721 skip "LU-8727: no implementation for ZFS" && return
7725 client_up || error "client_up failed"
7728 # set number of permanent parameters
7730 test_109_set_params $FSNAME
7732 umount_client $MOUNT || error "umount_client failed"
7733 stop_ost || error "stop_ost failed"
7734 stop_mds || error "stop_mds failed"
7736 test_109_clear_conf $FSNAME-MDT0000
7738 # make sure that only one config is cleared
7740 test_109_file_shortened $FSNAME-MDT0000 ||
7741 error "failed to clear MDT0000 config"
7742 test_109_file_shortened $FSNAME-client &&
7743 error "failed to clear client config"
7748 # check that configurations are intact
7750 test_109_test_params $FSNAME
7755 destroy_test_pools || error "destroy test pools failed"
7759 run_test 109b "test lctl clear_conf one config"
7762 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
7763 [[ $(lustre_version_code ost1) -ge $(version_code 2.10.6) ]] ||
7764 { skip "Need OST version at least 2.10.6" && return 0; }
7768 LOAD_MODULES_REMOTE=true load_modules
7769 #define OBD_FAIL_OFD_SET_OID 0x1e0
7770 do_facet ost1 $LCTL set_param fail_loc=0x00001e0
7773 $LFS mkdir -i1 -c1 $DIR/$tdir
7774 $LFS setstripe -i0 -c1 $DIR/$tdir
7775 do_facet ost1 $LCTL set_param fail_loc=0
7776 createmany -o $DIR/$tdir/file_ 1000 ||
7777 error "Fail to create a new sequence"
7781 run_test 122 "Check OST sequence update"
7784 remote_mgs_nodsh && skip "remote MGS with nodsh"
7785 [ -d $MOUNT/.lustre ] || setupall
7787 # test old logid format until removal from llog_ioctl.c::str2logid()
7788 if [ $(lustre_version_code MGS) -lt $(version_code 3.1.53) ]; then
7789 do_facet mgs $LCTL dl | grep MGS
7790 do_facet mgs "$LCTL --device %MGS llog_print \
7791 \\\\\\\$$FSNAME-client 1 10" ||
7792 error "old llog_print failed"
7795 # test new logid format
7796 if [ $(lustre_version_code MGS) -ge $(version_code 2.9.53) ]; then
7797 do_facet mgs "$LCTL --device MGS llog_print $FSNAME-client" ||
7798 error "new llog_print failed"
7801 run_test 123aa "llog_print works with FIDs and simple names"
7804 remote_mgs_nodsh && skip "remote MGS with nodsh"
7805 [[ $(lustre_version_code MGS) -gt $(version_code 2.11.51) ]] ||
7806 skip "Need server with working llog_print support"
7808 [ -d $MOUNT/.lustre ] || setupall
7813 orig_val=$(do_facet mgs $LCTL get_param jobid_name)
7814 do_facet mgs $LCTL set_param -P jobid_name="testname"
7816 yaml=$(do_facet mgs $LCTL --device MGS llog_print params |
7817 grep jobid_name | tail -n 1)
7819 local param=$(awk '{ print $10 }' <<< "$yaml")
7820 local val=$(awk '{ print $12 }' <<< "$yaml")
7821 #return to the default
7822 do_facet mgs $LCTL set_param -P jobid_name=$orig_val
7823 [ $val = "testname" ] || error "bad value: $val"
7824 [ $param = "jobid_name," ] || error "Bad param: $param"
7826 run_test 123ab "llog_print params output values from set_param -P"
7828 test_123ac() { # LU-11566
7829 remote_mgs_nodsh && skip "remote MGS with nodsh"
7830 do_facet mgs "$LCTL help llog_print" 2>&1 | grep -q -- --start ||
7831 skip "Need 'lctl llog_print --start' on MGS"
7836 [ -d $MOUNT/.lustre ] || setupall
7838 # - { index: 10, event: add_uuid, nid: 192.168.20.1@tcp(0x20000c0a81401,
7839 # node: 192.168.20.1@tcp }
7840 do_facet mgs $LCTL --device MGS \
7841 llog_print --start $start --end $end $FSNAME-client | tr -d , |
7842 while read DASH BRACE INDEX idx EVENT BLAH BLAH BLAH; do
7843 (( idx >= start )) || error "llog_print index $idx < $start"
7844 (( idx <= end )) || error "llog_print index $idx > $end"
7847 run_test 123ac "llog_print with --start and --end"
7849 test_123ad() { # LU-11566
7850 remote_mgs_nodsh && skip "remote MGS with nodsh"
7851 # older versions of lctl may not print all records properly
7852 do_facet mgs "$LCTL help llog_print" 2>&1 | grep -q -- --start ||
7853 skip "Need 'lctl llog_print --start' on MGS"
7855 [ -d $MOUNT/.lustre ] || setupall
7857 # append a new record, to avoid issues if last record was cancelled
7858 local old=$($LCTL get_param -n osc.*-OST0000-*.max_dirty_mb | head -1)
7859 do_facet mgs $LCTL conf_param $FSNAME-OST0000.osc.max_dirty_mb=$old
7861 # logid: [0x3:0xa:0x0]:0
7865 local num=$(do_facet mgs $LCTL --device MGS llog_info $FSNAME-client |
7866 awk '/last_index:/ { print $2 - 1 }')
7868 # - { index: 71, event: set_timeout, num: 0x14, param: sys.timeout=20 }
7869 local last=$(do_facet mgs $LCTL --device MGS llog_print $FSNAME-client |
7870 tail -1 | awk '{ print $4 }' | tr -d , )
7871 (( last == num )) || error "llog_print only showed $last/$num records"
7873 run_test 123ad "llog_print shows all records"
7875 if ! combined_mgs_mds ; then
7881 # restore the values of MDSSIZE and OSTSIZE
7882 MDSSIZE=$STORED_MDSSIZE
7883 OSTSIZE=$STORED_OSTSIZE