7 # bug number for skipped test: LU-8972
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=$OSTSIZE
70 if ! combined_mgs_mds; then
71 # bug number for skipped test: 23954
72 ALWAYS_EXCEPT="$ALWAYS_EXCEPT 24b"
75 # pass "-E lazy_itable_init" to mke2fs to speed up the formatting time
76 if [[ "$LDISKFS_MKFS_OPTS" != *lazy_itable_init* ]]; then
77 LDISKFS_MKFS_OPTS=$(csa_add "$LDISKFS_MKFS_OPTS" -E lazy_itable_init)
80 [ $(facet_fstype $SINGLEMDS) = "zfs" ] &&
81 # bug number for skipped test:
82 ALWAYS_EXCEPT="$ALWAYS_EXCEPT"
87 require_dsh_mds || exit 0
88 require_dsh_ost || exit 0
91 [ "$SLOW" = "no" ] && EXCEPT_SLOW="45 69"
96 # The MGS must be started before the OSTs for a new fs, so start
97 # and stop to generate the startup logs.
100 wait_osc_import_state mds ost FULL
105 reformat_and_config() {
107 if ! combined_mgs_mds ; then
113 writeconf_or_reformat() {
114 # There are at most 2 OSTs for write_conf test
115 # who knows if/where $TUNEFS is installed?
116 # Better reformat if it fails...
117 writeconf_all $MDSCOUNT 2 ||
118 { echo "tunefs failed, reformatting instead" &&
119 reformat_and_config && return 0; }
129 start mgs $(mgsdevname) $MGS_MOUNT_OPTS
135 local dev=$(mdsdevname $num)
138 echo "start mds service on `facet_active_host $facet`"
139 start $facet ${dev} $MDS_MOUNT_OPTS $@ || return 94
145 local dev=$(mdsdevname $num)
148 echo "stop mds service on `facet_active_host $facet`"
149 # These tests all use non-failover stop
150 stop $facet -f || return 97
156 for num in $(seq $MDSCOUNT); do
157 start_mdt $num $@ || return 94
162 if ! combined_mgs_mds ; then
170 for num in $(seq $MDSCOUNT); do
171 stop_mdt $num || return 97
176 echo "stop mgs service on `facet_active_host mgs`"
177 # These tests all use non-failover stop
178 stop mgs -f || return 97
182 echo "start ost1 service on `facet_active_host ost1`"
183 start ost1 $(ostdevname 1) $OST_MOUNT_OPTS $@ || return 95
187 echo "stop ost1 service on `facet_active_host ost1`"
188 # These tests all use non-failover stop
189 stop ost1 -f || return 98
193 echo "start ost2 service on `facet_active_host ost2`"
194 start ost2 $(ostdevname 2) $OST_MOUNT_OPTS $@ || return 92
198 echo "stop ost2 service on `facet_active_host ost2`"
199 # These tests all use non-failover stop
200 stop ost2 -f || return 93
205 echo "mount $FSNAME on ${MOUNTPATH}....."
206 zconf_mount $(hostname) $MOUNTPATH || return 96
210 local mountopt="remount,$1"
212 echo "remount '$1' lustre on ${MOUNTPATH}....."
213 zconf_mount $(hostname) $MOUNTPATH "$mountopt" || return 96
219 echo "umount lustre on $mountpath....."
220 zconf_umount $HOSTNAME $mountpath $@ || return 97
223 manual_umount_client(){
226 echo "manual umount lustre on ${MOUNT}...."
227 do_facet client "umount ${FORCE} $MOUNT"
233 start_mds || error "MDT start failed"
234 start_ost || error "Unable to start OST1"
235 mount_client $MOUNT || error "client start failed"
236 client_up || error "client_up failed"
245 unload_modules_conf () {
246 if combined_mgs_mds || ! local_mode; then
247 unload_modules || return 1
252 stop_ost || return 202
253 stop_mds || return 201
254 unload_modules_conf || return 203
258 umount_client $MOUNT || return 200
259 cleanup_nocli || return $?
264 echo "umount $MOUNT2 ..."
265 umount $MOUNT2 || true
266 echo "stopping fs2mds ..."
267 stop fs2mds -f || true
268 echo "stopping fs2ost ..."
269 stop fs2ost -f || true
273 do_facet client "cp /etc/passwd $DIR/a" || return 71
274 do_facet client "rm $DIR/a" || return 72
275 # make sure lustre is actually mounted (touch will block,
276 # but grep won't, so do it after)
277 do_facet client "grep $MOUNT' ' /proc/mounts > /dev/null" || return 73
278 echo "setup single mount lustre success"
282 do_facet client "touch $DIR/a" || return 71
283 do_facet client "rm $DIR/a" || return 72
284 do_facet client "touch $DIR2/a" || return 73
285 do_facet client "rm $DIR2/a" || return 74
286 echo "setup double mount lustre success"
291 if [ "$ONLY" == "setup" ]; then
296 if [ "$ONLY" == "cleanup" ]; then
303 #create single point mountpoint
309 check_mount || error "check_mount failed"
310 cleanup || error "cleanup failed with $?"
312 run_test 0 "single mount setup"
315 start_mds || error "MDS start failed"
316 start_ost || error "unable to start OST"
317 echo "start ost second time..."
318 start_ost && error "2nd OST start should fail"
319 mount_client $MOUNT || error "client start failed"
320 check_mount || error "check_mount failed"
321 cleanup || error "cleanup failed with $?"
323 run_test 1 "start up ost twice (should return errors)"
326 start_mds || error "MDT start failed"
327 echo "start mds second time.."
328 start_mds && error "2nd MDT start should fail"
329 start_ost || error "OST start failed"
330 mount_client $MOUNT || error "mount_client failed to start client"
331 check_mount || error "check_mount failed"
332 cleanup || error "cleanup failed with $?"
334 run_test 2 "start up mds twice (should return err)"
338 #mount.lustre returns an error if already in mtab
339 mount_client $MOUNT && error "2nd client mount should fail"
340 check_mount || error "check_mount failed"
341 cleanup || error "cleanup failed with $?"
343 run_test 3 "mount client twice (should return err)"
347 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
348 stop_ost || error "Unable to stop OST1"
349 umount_client $MOUNT -f || error “unmount $MOUNT failed”
352 # ok for ost to fail shutdown
353 if [ 202 -ne $eno ] && [ 0 -ne $eno ]; then
354 error "cleanup failed with $?"
357 run_test 4 "force cleanup ost, then cleanup"
359 test_5a() { # was test_5
361 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
362 fuser -m -v $MOUNT && echo "$MOUNT is in use by user space process."
364 stop_mds || error "Unable to stop MDS"
366 # cleanup may return an error from the failed
367 # disconnects; for now I'll consider this successful
368 # if all the modules have unloaded.
372 echo "killing umount"
373 kill -TERM $UMOUNT_PID
374 echo "waiting for umount to finish"
376 if grep " $MOUNT " /proc/mounts; then
377 echo "test 5: /proc/mounts after failed umount"
381 echo "killing umount"
382 kill -TERM $UMOUNT_PID
383 echo "waiting for umount to finish"
385 grep " $MOUNT " /proc/mounts &&
386 error "/proc/mounts after second umount"
389 # manual_mount_client may fail due to umount succeeding above
391 # stop_mds is a no-op here, and should not fail
392 cleanup_nocli || error "cleanup_nocli failed with $?"
393 # df may have lingering entry
395 # mtab may have lingering entry
399 while [ "$WAIT" -ne "$MAX_WAIT" ]; do
401 grep -q $MOUNT" " /etc/mtab || break
402 echo "Waiting /etc/mtab updated ... "
403 WAIT=$(( WAIT + sleep))
405 [ "$WAIT" -eq "$MAX_WAIT" ] &&
406 error "/etc/mtab is not updated in $WAIT secs"
407 echo "/etc/mtab updated in $WAIT secs"
409 run_test 5a "force cleanup mds, then cleanup"
417 grep " $MOUNT " /etc/mtab &&
418 error false "unexpected entry in mtab before mount" && return 10
420 start_ost || error "OST start failed"
421 if ! combined_mgs_mds ; then
422 trap cleanup_5b EXIT ERR
423 start_mds || error "MDS start failed"
427 mount_client $MOUNT && error "mount_client $MOUNT should fail"
428 grep " $MOUNT " /etc/mtab &&
429 error "$MOUNT entry in mtab after failed mount"
431 # stop_mds is a no-op here, and should not fail
432 cleanup_nocli || error "cleanup_nocli failed with $?"
433 if ! combined_mgs_mds ; then
437 run_test 5b "Try to start a client with no MGS (should return errs)"
440 grep " $MOUNT " /etc/mtab &&
441 error false "unexpected entry in mtab before mount" && return 10
443 start_mds || error "MDS start failed"
444 start_ost || error "OST start failed"
445 local oldfs="${FSNAME}"
446 FSNAME="wrong.${FSNAME}"
447 mount_client $MOUNT || :
449 grep " $MOUNT " /etc/mtab &&
450 error "$MOUNT entry in mtab after failed mount"
452 cleanup_nocli || error "cleanup_nocli failed with $?"
454 run_test 5c "cleanup after failed mount (bug 2712) (should return errs)"
457 grep " $MOUNT " /etc/mtab &&
458 error "unexpected entry in mtab before mount"
460 start_ost || error "OST start failed"
461 start_mds || error "MDS start failed"
462 stop_ost -f || error "Unable to stop OST1"
463 mount_client $MOUNT || error "mount_client $MOUNT failed"
464 umount_client $MOUNT -f || error "umount_client $MOUNT failed"
465 cleanup_nocli || error "cleanup_nocli failed with $?"
466 ! grep " $MOUNT " /etc/mtab ||
467 error "$MOUNT entry in mtab after unmount"
469 run_test 5d "mount with ost down"
472 grep " $MOUNT " /etc/mtab &&
473 error false "unexpected entry in mtab before mount" && return 10
475 start_mds || error "MDS start failed"
476 start_ost || error "OST start failed"
478 #define OBD_FAIL_PTLRPC_DELAY_SEND 0x506
479 do_facet client "$LCTL set_param fail_loc=0x80000506"
480 mount_client $MOUNT || echo "mount failed (not fatal)"
481 cleanup || error "cleanup failed with $?"
482 grep " $MOUNT " /etc/mtab &&
483 error "$MOUNT entry in mtab after unmount"
486 run_test 5e "delayed connect, don't crash (bug 10268)"
489 if combined_mgs_mds ; then
490 skip "combined mgs and mds"
494 grep " $MOUNT " /etc/mtab &&
495 error false "unexpected entry in mtab before mount" && return 10
498 start_ost || error "OST start failed"
499 mount_client $MOUNT &
501 echo client_mount pid is $pid
505 if ! ps -f -p $pid >/dev/null; then
508 grep " $MOUNT " /etc/mtab && echo "test 5f: mtab after mount"
509 error "mount returns $rc, expected to hang"
511 cleanup || error "cleanup failed with $?"
516 start_mds || error "start MDS failed"
518 # mount should succeed after start mds
520 grep " $MOUNT " /etc/mtab && echo "test 5f: mtab after mount"
521 cleanup || error "final call to cleanup failed with rc $?"
523 run_test 5f "mds down, cleanup after failed mount (bug 2712)"
527 [ $(lustre_version_code client) -lt $(version_code 2.9.53) ] &&
528 { skip "automount of debugfs missing before 2.9.53" && return 0; }
529 umount /sys/kernel/debug
530 $LCTL get_param -n devices | egrep -v "error" && \
531 error "lctl can't access debugfs data"
532 grep " debugfs " /etc/mtab || error "debugfs failed to remount"
534 run_test 5g "handle missing debugfs"
539 mount_client $MOUNT || error "mount_client $MOUNT failed"
540 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
541 cleanup || error "cleanup failed with rc $?"
543 run_test 6 "manual umount, then mount again"
548 cleanup_nocli || error "cleanup_nocli failed with $?"
550 run_test 7 "manual umount, then cleanup"
554 mount_client $MOUNT2 || error "mount_client $MOUNT2 failed"
555 check_mount2 || error "check_mount2 failed"
556 umount_client $MOUNT2 || error "umount_client $MOUNT2 failed"
557 cleanup || error "cleanup failed with rc $?"
559 run_test 8 "double mount setup"
562 start_ost || error "OST start failed"
564 do_facet ost1 $LCTL set_param debug=\'inode trace\' ||
565 error "do_facet ost1 set_param inode trace failed."
566 do_facet ost1 $LCTL set_param subsystem_debug=\'mds ost\' ||
567 error "do_facet ost1 set_param debug mds ost failed."
569 CHECK_PTLDEBUG="`do_facet ost1 $LCTL get_param -n debug`"
570 if [ "$CHECK_PTLDEBUG" ] && { \
571 [ "$CHECK_PTLDEBUG" = "trace inode warning error emerg console" ] ||
572 [ "$CHECK_PTLDEBUG" = "trace inode" ]; }; then
573 echo "lnet.debug success"
575 error "lnet.debug: want 'trace inode', have '$CHECK_PTLDEBUG'"
577 CHECK_SUBSYS="`do_facet ost1 $LCTL get_param -n subsystem_debug`"
578 if [ "$CHECK_SUBSYS" ] && [ "$CHECK_SUBSYS" = "mds ost" ]; then
579 echo "lnet.subsystem_debug success"
581 error "lnet.subsystem_debug: want 'mds ost' got '$CHECK_SUBSYS'"
583 stop_ost || error "Unable to stop OST1"
585 run_test 9 "test ptldebug and subsystem for mkfs"
593 do_facet $facet "test -b $dev" || rc=1
594 if [[ "$size" ]]; then
595 local in=$(do_facet $facet "dd if=$dev of=/dev/null bs=1k \
596 count=1 skip=$size 2>&1" |
597 awk '($3 == "in") { print $1 }')
598 [[ $in = "1+0" ]] || rc=1
604 # Test 16 was to "verify that lustre will correct the mode of OBJECTS".
605 # But with new MDS stack we don't care about the mode of local objects
606 # anymore, so this test is removed. See bug 22944 for more details.
610 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
611 skip "Only applicable to ldiskfs-based MDTs"
616 check_mount || error "check_mount failed"
617 cleanup || error "cleanup failed with rc $?"
619 echo "Remove mds config log"
620 if ! combined_mgs_mds ; then
624 do_facet mgs "$DEBUGFS -w -R 'unlink CONFIGS/$FSNAME-MDT0000' \
625 $(mgsdevname) || return \$?" ||
626 error "do_facet mgs failed with $?"
628 if ! combined_mgs_mds ; then
632 start_ost || error "OST start failed"
633 start_mds && error "MDS start succeeded, but should fail"
636 run_test 17 "Verify failed mds_postsetup won't fail assertion (2936) (should return errs)"
639 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
640 skip "Only applicable to ldiskfs-based MDTs"
644 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
649 # check if current MDSSIZE is large enough
650 [ $MDSSIZE -ge $MIN ] && OK=1 && myMDSSIZE=$MDSSIZE &&
651 log "use MDSSIZE=$MDSSIZE"
653 # check if the global config has a large enough MDSSIZE
654 [ -z "$OK" -a ! -z "$STORED_MDSSIZE" ] &&
655 [ $STORED_MDSSIZE -ge $MIN ] &&
656 OK=1 && myMDSSIZE=$STORED_MDSSIZE &&
657 log "use STORED_MDSSIZE=$STORED_MDSSIZE"
659 # check if the block device is large enough
660 is_blkdev $SINGLEMDS $MDSDEV $MIN
661 local large_enough=$?
662 if [ -n "$OK" ]; then
663 [ $large_enough -ne 0 ] && OK=""
665 [ $large_enough -eq 0 ] && OK=1 && myMDSSIZE=$MIN &&
666 log "use device $MDSDEV with MIN=$MIN"
669 # check if a loopback device has enough space for fs metadata (5%)
671 if [ -z "$OK" ]; then
672 local SPACE=$(do_facet $SINGLEMDS "[ -f $MDSDEV -o ! \
673 -e $MDSDEV ] && df -P \\\$(dirname $MDSDEV)" |
674 awk '($1 != "Filesystem") { print $4 }')
675 ! [ -z "$SPACE" ] && [ $SPACE -gt $((MIN / 20)) ] &&
676 OK=1 && myMDSSIZE=$MIN &&
677 log "use file $MDSDEV with MIN=$MIN"
680 [ -z "$OK" ] && skip_env "$MDSDEV too small for ${MIN}kB MDS" && return
682 echo "mount mds with large journal..."
684 local OLD_MDSSIZE=$MDSSIZE
688 echo "mount lustre system..."
690 check_mount || error "check_mount failed"
692 echo "check journal size..."
693 local FOUNDSIZE=$(do_facet $SINGLEMDS "$DEBUGFS -c -R 'stat <8>' $MDSDEV" | awk '/Size: / { print $NF; exit;}')
694 if [ $FOUNDSIZE -gt $((32 * 1024 * 1024)) ]; then
695 log "Success: mkfs creates large journals. Size: $((FOUNDSIZE >> 20))M"
697 error "expected journal size > 32M, found $((FOUNDSIZE >> 20))M"
700 cleanup || error "cleanup failed with rc $?"
705 run_test 18 "check mkfs creates large journals"
708 start_mds || error "MDS start failed"
709 stop_mds || error "Unable to stop MDS"
711 run_test 19a "start/stop MDS without OSTs"
714 start_ost || error "Unable to start OST1"
715 stop_ost -f || error "Unable to stop OST1"
717 run_test 19b "start/stop OSTs without MDS"
720 # first format the ost/mdt
721 start_mds || error "MDS start failed"
722 start_ost || error "Unable to start OST1"
723 mount_client $MOUNT || error "mount_client $MOUNT failed"
724 check_mount || error "check_mount failed"
725 rm -f $DIR/$tfile || error "remove $DIR/$tfile failed."
726 remount_client ro $MOUNT || error "remount_client with ro failed"
727 touch $DIR/$tfile && error "$DIR/$tfile created incorrectly"
728 [ -e $DIR/$tfile ] && error "$DIR/$tfile exists incorrectly"
729 remount_client rw $MOUNT || error "remount_client with rw failed"
730 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
731 MCNT=$(grep -c $MOUNT /etc/mtab)
732 [ "$MCNT" -ne 1 ] && error "$MOUNT in /etc/mtab $MCNT times"
734 stop_mds || error "Unable to stop MDS"
735 stop_ost || error "Unable to stop OST1"
737 run_test 20 "remount ro,rw mounts work and doesn't break /etc/mtab"
740 start_mds || error "MDS start failed"
741 start_ost || error "unable to start OST1"
742 wait_osc_import_state mds ost FULL
743 stop_ost || error "unable to stop OST1"
744 stop_mds || error "unable to stop MDS"
746 run_test 21a "start mds before ost, stop ost first"
749 start_ost || error "unable to start OST1"
750 start_mds || error "MDS start failed"
751 wait_osc_import_state mds ost FULL
752 stop_mds || error "unable to stop MDS"
753 stop_ost || error "unable to stop OST1"
755 run_test 21b "start ost before mds, stop mds first"
758 start_ost || error "Unable to start OST1"
759 start_mds || error "MDS start failed"
760 start_ost2 || error "Unable to start OST2"
761 wait_osc_import_state mds ost2 FULL
762 stop_ost || error "Unable to stop OST1"
763 stop_ost2 || error "Unable to stop OST2"
764 stop_mds || error "Unable to stop MDS"
765 #writeconf to remove all ost2 traces for subsequent tests
766 writeconf_or_reformat
768 run_test 21c "start mds between two osts, stop mds last"
771 if combined_mgs_mds ; then
772 skip "need separate mgs device" && return 0
778 start_mgs || error "unable to start MGS"
779 start_ost || error "unable to start OST1"
780 start_ost2 || error "unable to start OST2"
781 start_mds || error "MDS start failed"
782 wait_osc_import_state mds ost2 FULL
784 stop_ost || error "Unable to stop OST1"
785 stop_ost2 || error "Unable to stop OST2"
786 stop_mds || error "Unable to stop MDS"
788 #writeconf to remove all ost2 traces for subsequent tests
789 writeconf_or_reformat
790 start_mgs || error "unable to start MGS"
792 run_test 21d "start mgs then ost and then mds"
795 MGSNID="$saved_mgsnid"
797 echo "stopping fs2mgs ..."
798 stop $fs2mgs -f || true
801 test_21e() { # LU-5863
802 if [[ -z "$fs3ost_DEV" || -z "$fs2ost_DEV" || -z "$fs2mds_DEV" ]]; then
803 is_blkdev $SINGLEMDS $(mdsdevname ${SINGLEMDS//mds/}) &&
804 skip_env "mixed loopback and real device not working" && return
807 local fs2mdsdev=$(mdsdevname 1_2)
808 local fs2ostdev=$(ostdevname 1_2)
809 local fs3ostdev=$(ostdevname 2_2)
811 local fs2mdsvdev=$(mdsvdevname 1_2)
812 local fs2ostvdev=$(ostvdevname 1_2)
813 local fs3ostvdev=$(ostvdevname 2_2)
815 # temporarily use fs3ost as fs2mgs
817 local fs2mgsdev=$fs3ostdev
818 local fs2mgsvdev=$fs3ostvdev
820 local fsname=test1234
822 add $fs2mgs $(mkfs_opts mgs $fs2mgsdev) --fsname=$fsname \
823 --reformat $fs2mgsdev $fs2mgsvdev || error "add fs2mgs failed"
824 start $fs2mgs $fs2mgsdev $MGS_MOUNT_OPTS && trap cleanup_21e EXIT INT ||
825 error "start fs2mgs failed"
827 local saved_mgsnid="$MGSNID"
828 MGSNID=$(do_facet $fs2mgs $LCTL list_nids | xargs | tr ' ' ,)
830 add fs2mds $(mkfs_opts mds1 $fs2mdsdev $fsname) \
831 --reformat $fs2mdsdev $fs2mdsvdev || error "add fs2mds failed"
832 add fs2ost $(mkfs_opts ost1 $fs2ostdev $fsname) \
833 --reformat $fs2ostdev $fs2ostvdev || error "add fs2ost failed"
835 start fs2ost $fs2ostdev $OST_MOUNT_OPTS || error "start fs2ost failed"
836 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS || error "start fs2mds failed"
838 mkdir -p $MOUNT2 || error "mkdir $MOUNT2 failed"
839 $MOUNT_CMD $MGSNID:/$fsname $MOUNT2 || error "mount $MOUNT2 failed"
840 DIR=$MOUNT2 MOUNT=$MOUNT2 check_mount || error "check $MOUNT2 failed"
844 run_test 21e "separate MGS and MDS"
847 start_mds || error "MDS start failed"
849 echo "Client mount with ost in logs, but none running"
850 start_ost || error "unable to start OST1"
851 # wait until mds connected to ost and open client connection
852 wait_osc_import_state mds ost FULL
853 stop_ost || error "unable to stop OST1"
854 mount_client $MOUNT || error "mount_client $MOUNT failed"
855 # check_mount will block trying to contact ost
856 mcreate $DIR/$tfile || error "mcreate $DIR/$tfile failed"
857 rm -f $DIR/$tfile || error "remove $DIR/$tfile failed"
858 umount_client $MOUNT -f
861 echo "Client mount with a running ost"
862 start_ost || error "unable to start OST1"
864 # if gss enabled, wait full time to let connection from
865 # mds to ost be established, due to the mismatch between
866 # initial connect timeout and gss context negotiation timeout.
867 # This perhaps could be remove after AT landed.
868 echo "sleep $((TIMEOUT + TIMEOUT + TIMEOUT))s"
869 sleep $((TIMEOUT + TIMEOUT + TIMEOUT))
871 mount_client $MOUNT || error "mount_client $MOUNT failed"
872 wait_osc_import_state mds ost FULL
873 wait_osc_import_state client ost FULL
874 check_mount || error "check_mount failed"
877 cleanup || error "cleanup failed with rc $?"
879 run_test 22 "start a client before osts (should return errs)"
881 test_23a() { # was test_23
884 stop $SINGLEMDS || error "failed to stop $SINGLEMDS"
885 # force down client so that recovering mds waits for reconnect
886 local running=$(grep -c $MOUNT /proc/mounts) || true
887 if [ $running -ne 0 ]; then
888 echo "Stopping client $MOUNT (opts: -f)"
892 # enter recovery on failed mds
893 local MDT_DEV=$(mdsdevname ${SINGLEMDS//mds/})
894 start $SINGLEMDS $MDT_DEV $MDS_MOUNT_OPTS || error "MDS start failed"
895 # try to start a new client
896 mount_client $MOUNT &
898 MOUNT_PID=$(ps -ef | grep "t lustre" | grep -v grep | awk '{print $2}')
899 MOUNT_LUSTRE_PID=$(ps -ef | grep mount.lustre |
900 grep -v grep | awk '{print $2}')
901 echo mount pid is ${MOUNT_PID}, mount.lustre pid is ${MOUNT_LUSTRE_PID}
903 ps --ppid $MOUNT_LUSTRE_PID
904 echo "waiting for mount to finish"
906 # "ctrl-c" sends SIGINT but it usually (in script) does not work on child process
907 # SIGTERM works but it does not spread to offspring processses
908 kill -s TERM $MOUNT_PID
909 kill -s TERM $MOUNT_LUSTRE_PID
910 # we can not wait $MOUNT_PID because it is not a child of this shell
916 while [ "$WAIT" -lt "$MAX_WAIT" ]; do
918 PID1=$(ps -ef | awk '{print $2}' | grep -w $MOUNT_PID)
919 PID2=$(ps -ef | awk '{print $2}' | grep -w $MOUNT_LUSTRE_PID)
922 [ -z "$PID1" -a -z "$PID2" ] && break
923 echo "waiting for mount to finish ... "
924 WAIT=$(( WAIT + sleep))
926 if [ "$WAIT" -eq "$MAX_WAIT" ]; then
927 error "MOUNT_PID $MOUNT_PID and "\
928 "MOUNT_LUSTRE_PID $MOUNT_LUSTRE_PID still not killed in $WAIT secs"
931 stop_mds || error "stopping MDSes failed"
932 stop_ost || error "stopping OSSes failed"
934 run_test 23a "interrupt client during recovery mount delay"
939 test_23b() { # was test_23
940 start_mds || error "MDS start failed"
941 start_ost || error "Unable to start OST1"
942 # Simulate -EINTR during mount OBD_FAIL_LDLM_CLOSE_THREAD
943 $LCTL set_param fail_loc=0x80000313
945 cleanup || error "cleanup failed with rc $?"
947 run_test 23b "Simulate -EINTR during mount"
950 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
952 if [ -z "$fs2ost_DEV" -o -z "$fs2mds_DEV" ]; then
953 is_blkdev $SINGLEMDS $MDSDEV &&
954 skip_env "mixed loopback and real device not working" && return
957 [ -n "$ost1_HOST" ] && fs2ost_HOST=$ost1_HOST
959 local fs2mdsdev=$(mdsdevname 1_2)
960 local fs2ostdev=$(ostdevname 1_2)
961 local fs2mdsvdev=$(mdsvdevname 1_2)
962 local fs2ostvdev=$(ostvdevname 1_2)
964 # test 8-char fsname as well
965 local FSNAME2=test1234
967 add fs2mds $(mkfs_opts mds1 ${fs2mdsdev} ) --nomgs --mgsnode=$MGSNID \
968 --fsname=${FSNAME2} --reformat $fs2mdsdev $fs2mdsvdev || exit 10
970 add fs2ost $(mkfs_opts ost1 ${fs2ostdev}) --fsname=${FSNAME2} \
971 --reformat $fs2ostdev $fs2ostvdev || exit 10
974 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS && trap cleanup_fs2 EXIT INT
975 start fs2ost $fs2ostdev $OST_MOUNT_OPTS
976 mkdir -p $MOUNT2 || error "mkdir $MOUNT2 failed"
977 $MOUNT_CMD $MGSNID:/${FSNAME2} $MOUNT2 || error "$MOUNT_CMD failed"
979 check_mount || error "check_mount failed"
980 # files written on 1 should not show up on 2
981 cp /etc/passwd $DIR/$tfile
983 [ -e $MOUNT2/$tfile ] && error "File bleed"
986 cp /etc/passwd $MOUNT2/$tfile ||
987 error "cp /etc/passwd $MOUNT2/$tfile failed"
988 rm $MOUNT2/$tfile || error "remove $MOUNT2/$tfile failed"
989 # 2 is actually mounted
990 grep $MOUNT2' ' /proc/mounts > /dev/null || error "$MOUNT2 not mounted"
992 facet_failover fs2mds
993 facet_failover fs2ost
996 # the MDS must remain up until last MDT
998 MDS=$(do_facet $SINGLEMDS "$LCTL get_param -n devices" |
999 awk '($3 ~ "mdt" && $4 ~ "MDT") { print $4 }' | head -1)
1000 [ -z "$MDS" ] && error "No MDT"
1002 cleanup_nocli || error "cleanup_nocli failed with rc $?"
1004 run_test 24a "Multiple MDTs on a single node"
1007 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
1009 if [ -z "$fs2mds_DEV" ]; then
1010 local dev=${SINGLEMDS}_dev
1011 local MDSDEV=${!dev}
1012 is_blkdev $SINGLEMDS $MDSDEV &&
1013 skip_env "mixed loopback and real device not working" && return
1016 local fs2mdsdev=$(mdsdevname 1_2)
1017 local fs2mdsvdev=$(mdsvdevname 1_2)
1019 add fs2mds $(mkfs_opts mds1 ${fs2mdsdev} ) --mgs --fsname=${FSNAME}2 \
1020 --reformat $fs2mdsdev $fs2mdsvdev || exit 10
1022 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS &&
1023 error "start MDS should fail"
1025 cleanup || error "cleanup failed with rc $?"
1027 run_test 24b "Multiple MGSs on a single node (should return err)"
1031 check_mount || error "check_mount failed"
1032 local MODULES=$($LCTL modules | awk '{ print $2 }')
1033 rmmod $MODULES 2>/dev/null || true
1034 cleanup || error "cleanup failed with $?"
1036 run_test 25 "Verify modules are referenced"
1040 # we need modules before mount for sysctl, so make sure...
1041 do_facet $SINGLEMDS "lsmod | grep -q lustre || modprobe lustre"
1042 #define OBD_FAIL_MDS_FS_SETUP 0x135
1043 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0x80000135"
1044 start_mds && error "MDS started but should not have started"
1045 $LCTL get_param -n devices
1046 DEVS=$($LCTL get_param -n devices | egrep -v MG | wc -l)
1047 [ $DEVS -gt 0 ] && error "number of devices is $DEVS, should be zero"
1048 # start mds to drop writeconf setting
1049 start_mds || error "Unable to start MDS"
1050 stop_mds || error "Unable to stop MDS"
1051 unload_modules_conf || error "unload_modules_conf failed with $?"
1053 run_test 26 "MDT startup failure cleans LOV (should return errs)"
1056 start_ost || error "Unable to start OST1"
1057 start_mds || error "Unable to start MDS"
1058 echo "Requeue thread should have started: "
1059 ps -e | grep ll_cfg_requeue
1060 set_conf_param_and_check ost1 \
1061 "$LCTL get_param -n obdfilter.$FSNAME-OST0000.client_cache_seconds" \
1062 "$FSNAME-OST0000.ost.client_cache_seconds" ||
1063 error "set_conf_param_and_check ost1 failed"
1064 cleanup_nocli || error "cleanup_nocli failed with rc $?"
1066 run_test 27a "Reacquire MGS lock if OST started first"
1071 local device=$(do_facet $SINGLEMDS "$LCTL get_param -n devices" |
1072 awk '($3 ~ "mdt" && $4 ~ "MDT0000") { print $4 }')
1074 facet_failover $SINGLEMDS
1075 set_conf_param_and_check $SINGLEMDS \
1076 "$LCTL get_param -n mdt.$device.identity_acquire_expire" \
1077 "$device.mdt.identity_acquire_expire" ||
1078 error "set_conf_param_and_check $SINGLEMDS failed"
1079 set_conf_param_and_check client \
1080 "$LCTL get_param -n mdc.$device-mdc-*.max_rpcs_in_flight"\
1081 "$device.mdc.max_rpcs_in_flight" ||
1082 error "set_conf_param_and_check client failed"
1084 cleanup || error "cleanup failed with $?"
1086 run_test 27b "Reacquire MGS lock after failover"
1090 TEST="$LCTL get_param -n llite.$FSNAME-*.max_read_ahead_whole_mb"
1091 PARAM="$FSNAME.llite.max_read_ahead_whole_mb"
1093 FINAL=$(($ORIG + 1))
1094 set_conf_param_and_check client "$TEST" "$PARAM" $FINAL ||
1095 error "first set_conf_param_and_check client failed"
1096 FINAL=$(($FINAL + 1))
1097 set_conf_param_and_check client "$TEST" "$PARAM" $FINAL ||
1098 error "second set_conf_param_and_check client failed"
1099 umount_client $MOUNT || error "umount_client $MOUNT failed"
1100 mount_client $MOUNT || error "mount_client $MOUNT failed"
1102 if [ $RESULT -ne $FINAL ]; then
1103 error "New config not seen: wanted $FINAL got $RESULT"
1105 echo "New config success: got $RESULT"
1107 set_conf_param_and_check client "$TEST" "$PARAM" $ORIG ||
1108 error "third set_conf_param_and_check client failed"
1109 cleanup || error "cleanup failed with rc $?"
1111 run_test 28 "permanent parameter setting"
1113 test_28a() { # LU-4221
1114 [[ $(lustre_version_code ost1) -ge $(version_code 2.5.52) ]] ||
1115 { skip "Need OST version at least 2.5.52" && return 0; }
1116 [ "$(facet_fstype ost1)" = "zfs" ] &&
1117 skip "LU-4221: no such proc params for ZFS OSTs" && return
1124 local device="$FSNAME-OST0000"
1128 # In this test we will set three kinds of proc parameters with
1130 # 1. the ones moved from the OFD to the OSD, and only their
1131 # symlinks kept in obdfilter
1132 # 2. non-symlink ones in the OFD
1133 # 3. non-symlink ones in the OSD
1136 # prepare a symlink parameter in the OFD
1137 name="writethrough_cache_enable"
1138 param="$device.ost.$name"
1139 cmd="$LCTL get_param -n obdfilter.$device.$name"
1141 # conf_param the symlink parameter in the OFD
1142 old=$(do_facet ost1 $cmd)
1143 new=$(((old + 1) % 2))
1144 set_conf_param_and_check ost1 "$cmd" "$param" $new ||
1145 error "lctl conf_param $device.ost.$param=$new failed"
1147 # conf_param the target parameter in the OSD
1148 param="$device.osd.$name"
1149 cmd="$LCTL get_param -n osd-*.$device.$name"
1150 set_conf_param_and_check ost1 "$cmd" "$param" $old ||
1151 error "lctl conf_param $device.osd.$param=$old failed"
1154 # prepare a non-symlink parameter in the OFD
1155 name="client_cache_seconds"
1156 param="$device.ost.$name"
1157 cmd="$LCTL get_param -n obdfilter.$device.$name"
1159 # conf_param the parameter in the OFD
1160 old=$(do_facet ost1 $cmd)
1162 set_conf_param_and_check ost1 "$cmd" "$param" $new ||
1163 error "lctl conf_param $device.ost.$param=$new failed"
1164 set_conf_param_and_check ost1 "$cmd" "$param" $old ||
1165 error "lctl conf_param $device.ost.$param=$old failed"
1168 # prepare a non-symlink parameter in the OSD
1170 param="$device.osd.$name"
1171 cmd="$LCTL get_param -n osd-*.$device.$name"
1173 # conf_param the parameter in the OSD
1174 old=$(do_facet ost1 $cmd)
1175 new=$(((old + 1) % 2))
1176 set_conf_param_and_check ost1 "$cmd" "$param" $new ||
1177 error "lctl conf_param $device.osd.$param=$new failed"
1178 set_conf_param_and_check ost1 "$cmd" "$param" $old ||
1179 error "lctl conf_param $device.osd.$param=$old failed"
1181 cleanup || error "cleanup failed with $?"
1183 run_test 28a "set symlink parameters permanently with conf_param"
1186 [ "$OSTCOUNT" -lt "2" ] && skip_env "$OSTCOUNT < 2, skipping" && return
1187 setup > /dev/null 2>&1
1188 start_ost2 || error "Unable to start OST2"
1191 local PARAM="$FSNAME-OST0001.osc.active"
1192 local PROC_ACT="osc.$FSNAME-OST0001-osc-[^M]*.active"
1193 local PROC_UUID="osc.$FSNAME-OST0001-osc-[^M]*.ost_server_uuid"
1195 ACTV=$($LCTL get_param -n $PROC_ACT)
1197 set_conf_param_and_check client \
1198 "$LCTL get_param -n $PROC_ACT" "$PARAM" $DEAC ||
1199 error "set_conf_param_and_check client failed"
1200 # also check ost_server_uuid status
1201 RESULT=$($LCTL get_param -n $PROC_UUID | grep DEACTIV)
1202 if [ -z "$RESULT" ]; then
1203 error "Client not deactivated: $($LCTL get_param \
1206 echo "Live client success: got $RESULT"
1210 wait_osp_active ost ${FSNAME}-OST0001 1 0
1212 # test new client starts deactivated
1213 umount_client $MOUNT || error "umount_client $MOUNT failed"
1214 mount_client $MOUNT || error "mount_client $MOUNT failed"
1215 RESULT=$($LCTL get_param -n $PROC_UUID | grep DEACTIV | grep NEW)
1216 if [ -z "$RESULT" ]; then
1217 error "New client start active: $(lctl get_param -n $PROC_UUID)"
1219 echo "New client success: got $RESULT"
1222 # make sure it reactivates
1223 set_conf_param_and_check client \
1224 "$LCTL get_param -n $PROC_ACT" "$PARAM" $ACTV ||
1225 error "lctl get_param $PROC_ACT $PARAM $ACTV failed"
1227 umount_client $MOUNT
1228 stop_ost2 || error "Unable to stop OST2"
1229 cleanup_nocli || error "cleanup_nocli failed with $?"
1230 #writeconf to remove all ost2 traces for subsequent tests
1231 writeconf_or_reformat
1233 run_test 29 "permanently remove an OST"
1238 echo Big config llog
1239 TEST="$LCTL get_param -n llite.$FSNAME-*.max_read_ahead_whole_mb"
1241 LIST=(1 2 3 4 5 4 3 2 1 2 3 4 5 4 3 2 1 2 3 4 5)
1242 for i in ${LIST[@]}; do
1243 set_conf_param_and_check client "$TEST" \
1244 "$FSNAME.llite.max_read_ahead_whole_mb" $i ||
1245 error "Set $FSNAME.llite.max_read_ahead_whole_mb failed"
1247 # make sure client restart still works
1248 umount_client $MOUNT
1249 mount_client $MOUNT || error "mount_client $MOUNT failed"
1250 [ "$($TEST)" -ne "$i" ] &&
1251 error "Param didn't stick across restart $($TEST) != $i"
1254 echo Erase parameter setting
1255 do_facet mgs "$LCTL conf_param \
1256 -d $FSNAME.llite.max_read_ahead_whole_mb" ||
1257 error "Erase param $FSNAME.llite.max_read_ahead_whole_mb failed"
1258 umount_client $MOUNT
1259 mount_client $MOUNT || error "mount_client $MOUNT failed"
1261 echo "deleted (default) value=$FINAL, orig=$ORIG"
1262 # assumes this parameter started at the default value
1263 [ "$FINAL" -eq "$ORIG" ] || fail "Deleted value=$FINAL, orig=$ORIG"
1265 cleanup || error "cleanup failed with rc $?"
1267 run_test 30a "Big config llog and conf_param deletion"
1272 local orignids=$($LCTL get_param -n \
1273 osc.$FSNAME-OST0000-osc-[^M]*.import | grep failover_nids)
1275 local orignidcount=$(echo "$orignids" | wc -w)
1277 # Make a fake nid. Use the OST nid, and add 20 to the least significant
1278 # numerical part of it. Hopefully that's not already a failover address
1280 local OSTNID=$(do_facet ost1 "$LCTL get_param nis" | tail -1 | \
1282 local ORIGVAL=$(echo $OSTNID | egrep -oi "[0-9]*@")
1283 local NEWVAL=$((($(echo $ORIGVAL | egrep -oi "[0-9]*") + 20) % 256))
1284 local NEW=$(echo $OSTNID | sed "s/$ORIGVAL/$NEWVAL@/")
1285 echo "Using fake nid $NEW"
1287 local TEST="$LCTL get_param -n osc.$FSNAME-OST0000-osc-[^M]*.import |
1288 grep failover_nids | sed -n 's/.*\($NEW\).*/\1/p'"
1289 set_conf_param_and_check client "$TEST" \
1290 "$FSNAME-OST0000.failover.node" $NEW ||
1291 error "didn't add failover nid $NEW"
1292 local NIDS=$($LCTL get_param -n osc.$FSNAME-OST0000-osc-[^M]*.import |
1295 local NIDCOUNT=$(echo "$NIDS" | wc -w)
1296 echo "should have $((orignidcount + 1)) entries \
1297 in failover nids string, have $NIDCOUNT"
1298 [ $NIDCOUNT -eq $((orignidcount + 1)) ] ||
1299 error "Failover nid not added"
1301 do_facet mgs "$LCTL conf_param -d $FSNAME-OST0000.failover.node" ||
1302 error "conf_param delete failed"
1303 umount_client $MOUNT
1304 mount_client $MOUNT || error "mount_client $MOUNT failed"
1306 NIDS=$($LCTL get_param -n osc.$FSNAME-OST0000-osc-[^M]*.import |
1309 NIDCOUNT=$(echo "$NIDS" | wc -w)
1310 echo "only $orignidcount final entries should remain \
1311 in failover nids string, have $NIDCOUNT"
1312 [ $NIDCOUNT -eq $orignidcount ] || error "Failover nids not removed"
1314 cleanup || error "cleanup failed with rc $?"
1316 run_test 30b "Remove failover nids"
1318 test_31() { # bug 10734
1319 # ipaddr must not exist
1320 $MOUNT_CMD 4.3.2.1@tcp:/lustre $MOUNT || true
1321 cleanup || error "cleanup failed with rc $?"
1323 run_test 31 "Connect to non-existent node (shouldn't crash)"
1327 T32_BLIMIT=40960 # Kbytes
1331 # This is not really a test but a tool to create new disk
1332 # image tarballs for the upgrade tests.
1334 # Disk image tarballs should be created on single-node
1335 # clusters by running this test with default configurations
1336 # plus a few mandatory environment settings that are verified
1337 # at the beginning of the test.
1339 test_32newtarball() {
1343 local tmp=$TMP/t32_image_create
1344 local server_version=$(lustre_version_code $SINGLEMDS)
1349 if [ $FSNAME != t32fs -o \( -z "$MDSDEV" -a -z "$MDSDEV1" \) -o \
1350 $OSTCOUNT -ne 1 -o -z "$OSTDEV1" ]; then
1351 error "Needs FSNAME=t32fs MDSCOUNT=2 " \
1352 "MDSDEV1=<nonexistent_file>" \
1353 "MDSDEV2=<nonexistent_file>" \
1354 "(or MDSDEV, in the case of b1_8)" \
1355 "OSTCOUNT=1 OSTDEV1=<nonexistent_file>"
1359 echo "Found stale $tmp"
1363 mkdir $tmp/src || return 1
1364 tar cf - -C $src . | tar xf - -C $tmp/src
1365 dd if=/dev/zero of=$tmp/src/t32_qf_old bs=1M \
1366 count=$(($T32_BLIMIT / 1024 / 4))
1367 chown $T32_QID.$T32_QID $tmp/src/t32_qf_old
1369 # format ost with comma-separated NIDs to verify LU-4460
1370 local failnid="$(h2$NETTYPE 1.2.3.4),$(h2$NETTYPE 4.3.2.1)"
1371 MGSNID="$MGSNID,$MGSNID" OSTOPT="--failnode=$failnid" formatall
1375 [[ $server_version -ge $(version_code 2.3.50) ]] ||
1376 $LFS quotacheck -ug /mnt/$FSNAME
1377 $LFS setquota -u $T32_QID -b 0 -B $T32_BLIMIT -i 0 -I $T32_ILIMIT \
1380 tar cf - -C $tmp/src . | tar xf - -C /mnt/$FSNAME
1382 if [[ $MDSCOUNT -ge 2 ]]; then
1383 remote_dir=/mnt/$FSNAME/remote_dir
1384 $LFS mkdir -i 1 $remote_dir
1385 tar cf - -C $tmp/src . | tar xf - -C $remote_dir
1387 if [[ $server_version -ge $(version_code 2.7.0) ]]; then
1388 striped_dir=/mnt/$FSNAME/striped_dir_old
1389 $LFS mkdir -i 1 -c 2 $striped_dir
1390 tar cf - -C $tmp/src . | tar xf - -C $striped_dir
1396 mkdir $tmp/img || return 1
1400 pushd_dir=/mnt/$FSNAME
1401 if [[ $MDSCOUNT -ge 2 ]]; then
1402 pushd_dir=$remote_dir
1403 if [[ $server_version -ge $(version_code 2.7.0) ]]; then
1405 ls -Rni --time-style=+%s >$tmp/img/list2
1411 ls -Rni --time-style=+%s >$tmp/img/list
1412 find ! -name .lustre -type f -exec sha1sum {} \; |
1413 sort -k 2 >$tmp/img/sha1sums
1415 $LCTL get_param -n version | head -n 1 |
1416 sed -e 's/^lustre: *//' >$tmp/img/commit
1418 [[ $server_version -ge $(version_code 2.3.50) ]] ||
1419 $LFS quotaon -ug /mnt/$FSNAME
1420 $LFS quota -u $T32_QID -v /mnt/$FSNAME
1421 $LFS quota -v -u $T32_QID /mnt/$FSNAME |
1422 awk 'BEGIN { num='1' } { if ($1 == "'/mnt/$FSNAME'") \
1423 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1424 | tr -d "*" > $tmp/img/bspace
1425 $LFS quota -v -u $T32_QID /mnt/$FSNAME |
1426 awk 'BEGIN { num='5' } { if ($1 == "'/mnt/$FSNAME'") \
1427 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1428 | tr -d "*" > $tmp/img/ispace
1429 echo $T32_BLIMIT > $tmp/img/blimit
1430 echo $T32_ILIMIT > $tmp/img/ilimit
1435 find -type f -exec sha1sum {} \; | sort -k 2 >$tmp/sha1sums.src
1438 if ! diff -u $tmp/sha1sums.src $tmp/img/sha1sums; then
1439 echo "Data verification failed"
1442 uname -r >$tmp/img/kernel
1443 uname -m >$tmp/img/arch
1445 mv ${MDSDEV1:-$MDSDEV} $tmp/img
1446 for num in $(seq 2 $MDSCOUNT); do
1447 local devname=$(mdsdevname $num)
1449 [[ $(facet_fstype $facet) != zfs ]] ||
1450 devname=$(mdsvdevname $num)
1451 mv $devname $tmp/img
1453 mv $OSTDEV1 $tmp/img
1455 version=$(sed -e 's/\(^[0-9]\+\.[0-9]\+\)\(.*$\)/\1/' $tmp/img/commit |
1456 sed -e 's/\./_/g') # E.g., "1.8.7" -> "1_8"
1459 tar cjvf $dst/disk$version-$(facet_fstype $SINGLEMDS).tar.bz2 -S *
1464 #run_test 32newtarball "Create a new test_32 disk image tarball for this version"
1467 # The list of applicable tarballs is returned via the caller's
1468 # variable "tarballs".
1471 local node=$(facet_active_host $SINGLEMDS)
1472 local r="do_node $node"
1474 if [ "$CLIENTONLY" ]; then
1475 skip "Client-only testing"
1479 if ! $r which $TUNEFS; then
1480 skip_env "tunefs.lustre required on $node"
1484 local IMGTYPE=$(facet_fstype $SINGLEMDS)
1486 tarballs=$($r find $RLUSTRE/tests -maxdepth 1 -name \'disk*-$IMGTYPE.tar.bz2\' | grep -v "2_9")
1488 if [ -z "$tarballs" ]; then
1489 skip "No applicable tarballs found"
1494 t32_test_cleanup() {
1496 local facet=$SINGLEMDS
1497 local fstype=$(facet_fstype $facet)
1500 if $shall_cleanup_lustre; then
1501 umount $tmp/mnt/lustre || rc=$?
1503 if $shall_cleanup_mdt; then
1504 $r $UMOUNT $tmp/mnt/mdt || rc=$?
1506 if $shall_cleanup_mdt1; then
1507 $r $UMOUNT $tmp/mnt/mdt1 || rc=$?
1509 if $shall_cleanup_ost; then
1510 $r $UMOUNT $tmp/mnt/ost || rc=$?
1515 if [[ $fstype == zfs ]]; then
1517 local poolname_list="t32fs-mdt1 t32fs-ost1"
1519 ! $mdt2_is_available || poolname_list+=" t32fs-mdt2"
1521 for poolname in $poolname_list; do
1522 destroy_zpool $facet $poolname
1525 combined_mgs_mds || start_mgs || rc=$?
1529 t32_bits_per_long() {
1531 # Yes, this is not meant to be perfect.
1541 t32_reload_modules() {
1543 local all_removed=false
1546 while ((i < 20)); do
1547 echo "Unloading modules on $node: Attempt $i"
1548 do_rpc_nodes $node $LUSTRE_RMMOD $(facet_fstype $SINGLEMDS) &&
1550 do_rpc_nodes $node check_mem_leak || return 1
1551 if $all_removed; then
1552 do_rpc_nodes $node load_modules
1558 echo "Unloading modules on $node: Given up"
1562 t32_wait_til_devices_gone() {
1568 echo wait for devices to go
1569 while ((i < 20)); do
1570 devices=$(do_rpc_nodes $node $LCTL device_list | wc -l)
1571 loops=$(do_rpc_nodes $node losetup -a | grep -c t32)
1572 ((devices == 0 && loops == 0)) && return 0
1576 echo "waiting for dev on $node: dev $devices loop $loops given up"
1577 do_rpc_nodes $node "losetup -a"
1578 do_rpc_nodes $node "$LCTL device_list"
1582 t32_verify_quota() {
1586 local fstype=$(facet_fstype $SINGLEMDS)
1590 # LU-2435: if the underlying zfs doesn't support userobj_accounting,
1591 # lustre will estimate the object count usage. This fails quota
1592 # verification in 32b. The object quota usage should be accurate after
1593 # zfs-0.7.0 is released.
1594 [ $fstype == "zfs" ] && {
1595 local zfs_version=$(do_node $node cat /sys/module/zfs/version)
1597 [ $(version_code $zfs_version) -lt $(version_code 0.7.0) ] && {
1598 echo "Skip quota verify for zfs: $zfs_version"
1603 $LFS quota -u $T32_QID -v $mnt
1605 qval=$($LFS quota -v -u $T32_QID $mnt |
1606 awk 'BEGIN { num='1' } { if ($1 == "'$mnt'") \
1607 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1609 [ $qval -eq $img_bspace ] || {
1610 echo "bspace, act:$qval, exp:$img_bspace"
1614 qval=$($LFS quota -v -u $T32_QID $mnt |
1615 awk 'BEGIN { num='5' } { if ($1 == "'$mnt'") \
1616 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1618 [ $qval -eq $img_ispace ] || {
1619 echo "ispace, act:$qval, exp:$img_ispace"
1623 qval=$($LFS quota -v -u $T32_QID $mnt |
1624 awk 'BEGIN { num='3' } { if ($1 == "'$mnt'") \
1625 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1627 [ $qval -eq $img_blimit ] || {
1628 echo "blimit, act:$qval, exp:$img_blimit"
1632 qval=$($LFS quota -v -u $T32_QID $mnt |
1633 awk 'BEGIN { num='7' } { if ($1 == "'$mnt'") \
1634 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1636 [ $qval -eq $img_ilimit ] || {
1637 echo "ilimit, act:$qval, exp:$img_ilimit"
1641 do_node $node $LCTL conf_param $fsname.quota.mdt=ug
1642 cmd="$LCTL get_param -n osd-$fstype.$fsname-MDT0000"
1643 cmd=$cmd.quota_slave.enabled
1644 wait_update $node "$cmd" "ug" || {
1645 echo "Enable mdt quota failed"
1649 do_node $node $LCTL conf_param $fsname.quota.ost=ug
1650 cmd="$LCTL get_param -n osd-$fstype.$fsname-OST0000"
1651 cmd=$cmd.quota_slave.enabled
1652 wait_update $node "$cmd" "ug" || {
1653 echo "Enable ost quota failed"
1658 runas -u $T32_QID -g $T32_QID dd if=/dev/zero of=$mnt/t32_qf_new \
1659 bs=1M count=$((img_blimit / 1024)) oflag=sync && {
1660 echo "Write succeed, but expect -EDQUOT"
1663 rm -f $mnt/t32_qf_new
1665 runas -u $T32_QID -g $T32_QID createmany -m $mnt/t32_qf_ \
1667 echo "Create succeed, but expect -EDQUOT"
1670 unlinkmany $mnt/t32_qf_ $img_ilimit
1678 local dne_upgrade=${dne_upgrade:-"no"}
1679 local ff_convert=${ff_convert:-"no"}
1680 local shall_cleanup_mdt=false
1681 local shall_cleanup_mdt1=false
1682 local shall_cleanup_ost=false
1683 local shall_cleanup_lustre=false
1684 local mdt2_is_available=false
1685 local node=$(facet_active_host $SINGLEMDS)
1686 local r="do_node $node"
1702 local fstype=$(facet_fstype $SINGLEMDS)
1703 local mdt_dev=$tmp/mdt
1704 local mdt2_dev=$tmp/mdt2
1705 local ost_dev=$tmp/ost
1710 combined_mgs_mds || stop_mgs || error "Unable to stop MGS"
1711 trap 'trap - RETURN; t32_test_cleanup' RETURN
1714 nid=$($r $LCTL list_nids | head -1)
1716 mkdir -p $tmp/mnt/lustre || error "mkdir $tmp/mnt/lustre failed"
1717 $r mkdir -p $tmp/mnt/{mdt,mdt1,ost}
1718 $r tar xjvf $tarball -S -C $tmp || {
1719 error_noexit "Unpacking the disk image tarball"
1722 img_commit=$($r cat $tmp/commit)
1723 img_kernel=$($r cat $tmp/kernel)
1724 img_arch=$($r cat $tmp/arch)
1725 img_bspace=$($r cat $tmp/bspace)
1726 img_ispace=$($r cat $tmp/ispace)
1728 # older images did not have "blimit" and "ilimit" files
1729 # use old values for T32_BLIMIT and T32_ILIMIT
1730 $r test -f $tmp/blimit && img_blimit=$($r cat $tmp/blimit) ||
1732 $r test -f $tmp/ilimit && img_ilimit=$($r cat $tmp/ilimit) ||
1735 echo "Upgrading from $(basename $tarball), created with:"
1736 echo " Commit: $img_commit"
1737 echo " Kernel: $img_kernel"
1738 echo " Arch: $img_arch"
1739 echo "OST version: $(lustre_build_version ost1)"
1741 # The conversion can be made only when both of the following
1742 # conditions are satisfied:
1743 # - ost device img version < 2.3.64
1744 # - ost server version >= 2.5
1745 [ $(version_code $img_commit) -ge $(version_code 2.3.64) -o \
1746 $(lustre_version_code ost1) -lt $(version_code 2.5.0) ] &&
1749 ! $r test -f $mdt2_dev || mdt2_is_available=true
1751 if [[ $fstype == zfs ]]; then
1754 local poolname_list="t32fs-mdt1 t32fs-ost1"
1756 ! $mdt2_is_available || poolname_list+=" t32fs-mdt2"
1758 for poolname in $poolname_list; do
1760 $ZPOOL list -H $poolname >/dev/null 2>&1 ||
1761 $ZPOOL import -f -d $tmp $poolname"
1764 # upgrade zpool to latest supported features, including
1765 # dnode quota accounting in 0.7.0
1766 $r "$ZPOOL upgrade -a"
1768 mdt_dev=t32fs-mdt1/mdt1
1769 ost_dev=t32fs-ost1/ost1
1770 ! $mdt2_is_available || mdt2_dev=t32fs-mdt2/mdt2
1771 wait_update_facet $SINGLEMDS "$ZPOOL list |
1772 awk '/^t32fs-mdt1/ { print \\\$1 }'" "t32fs-mdt1" || {
1773 error_noexit "import zfs pool failed"
1778 $r $LCTL set_param debug="$PTLDEBUG"
1780 $r $TUNEFS --dryrun $mdt_dev || {
1782 error_noexit "tunefs.lustre before mounting the MDT"
1786 if $mdt2_is_available; then
1787 $r $TUNEFS --dryrun $mdt2_dev || {
1789 error_noexit "tunefs.lustre before mounting the MDT"
1794 if [ "$writeconf" ]; then
1796 if [ $fstype == "ldiskfs" ]; then
1798 $r $TUNEFS --quota $mdt_dev || {
1800 error_noexit "Enable mdt quota feature"
1803 if $mdt2_is_available; then
1804 $r $TUNEFS --quota $mdt2_dev || {
1806 error_noexit "Enable mdt quota feature"
1812 if [ -n "$($LCTL list_nids | grep -v '\(tcp\|lo\)[[:digit:]]*$')" ]; then
1813 [[ $(lustre_version_code mgs) -ge $(version_code 2.3.59) ]] ||
1814 { skip "LU-2200: Cannot run over Inifiniband w/o lctl replace_nids "
1815 "(Need MGS version at least 2.3.59)"; return 0; }
1817 local osthost=$(facet_active_host ost1)
1818 local ostnid=$(do_node $osthost $LCTL list_nids | head -1)
1821 if [ $fstype == "ldiskfs" ]; then
1824 $r $MOUNT_CMD -o $mopts $mdt_dev $tmp/mnt/mdt
1825 $r $LCTL replace_nids $fsname-OST0000 $ostnid
1826 $r $LCTL replace_nids $fsname-MDT0000 $nid
1827 $r $UMOUNT $tmp/mnt/mdt
1830 mopts=exclude=$fsname-OST0000
1831 if [ $fstype == "ldiskfs" ]; then
1836 t32_wait_til_devices_gone $node
1838 $r $MOUNT_CMD -o $mopts $mdt_dev $tmp/mnt/mdt || {
1840 error_noexit "Mounting the MDT"
1843 shall_cleanup_mdt=true
1845 if $mdt2_is_available; then
1846 mopts=mgsnode=$nid,$mopts
1847 $r $MOUNT_CMD -o $mopts $mdt2_dev $tmp/mnt/mdt1 || {
1849 error_noexit "Mounting the MDT"
1853 echo "mount new MDT....$mdt2_dev"
1854 $r $LCTL set_param -n mdt.${fsname}*.enable_remote_dir=1 ||
1855 error_noexit "enable remote dir create failed"
1857 shall_cleanup_mdt1=true
1858 elif [ "$dne_upgrade" != "no" ]; then
1859 local fs2mdsdev=$(mdsdevname 1_2)
1860 local fs2mdsvdev=$(mdsvdevname 1_2)
1862 echo "mkfs new MDT on ${fs2mdsdev}...."
1863 if [ $(facet_fstype mds1) == ldiskfs ]; then
1864 mkfsoptions="--mkfsoptions=\\\"-J size=8\\\""
1867 add $SINGLEMDS $(mkfs_opts mds2 $fs2mdsdev $fsname) --reformat \
1868 $mkfsoptions $fs2mdsdev $fs2mdsvdev > /dev/null || {
1869 error_noexit "Mkfs new MDT failed"
1873 [[ $(facet_fstype mds1) != zfs ]] || import_zpool mds1
1875 $r $TUNEFS --dryrun $fs2mdsdev || {
1876 error_noexit "tunefs.lustre before mounting the MDT"
1880 echo "mount new MDT....$fs2mdsdev"
1881 $r $MOUNT_CMD -o $mopts $fs2mdsdev $tmp/mnt/mdt1 || {
1882 error_noexit "mount mdt1 failed"
1886 $r $LCTL set_param -n mdt.${fsname}*.enable_remote_dir=1 ||
1887 error_noexit "enable remote dir create failed"
1889 shall_cleanup_mdt1=true
1892 uuid=$($r $LCTL get_param -n mdt.$fsname-MDT0000.uuid) || {
1893 error_noexit "Getting MDT UUID"
1896 if [ "$uuid" != $fsname-MDT0000_UUID ]; then
1897 error_noexit "Unexpected MDT UUID: \"$uuid\""
1901 $r $TUNEFS --dryrun $ost_dev || {
1902 error_noexit "tunefs.lustre before mounting the OST"
1905 if [ "$writeconf" ]; then
1906 mopts=mgsnode=$nid,$writeconf
1907 if [ $fstype == "ldiskfs" ]; then
1909 $r $TUNEFS --quota $ost_dev || {
1911 error_noexit "Enable ost quota feature"
1917 if [ $fstype == "ldiskfs" ]; then
1922 $r $MOUNT_CMD -onomgs -o$mopts $ost_dev $tmp/mnt/ost || {
1923 error_noexit "Mounting the OST"
1926 shall_cleanup_ost=true
1928 uuid=$($r $LCTL get_param -n obdfilter.$fsname-OST0000.uuid) || {
1929 error_noexit "Getting OST UUID"
1932 if [ "$uuid" != $fsname-OST0000_UUID ]; then
1933 error_noexit "Unexpected OST UUID: \"$uuid\""
1937 $r $LCTL conf_param $fsname-OST0000.osc.max_dirty_mb=15 || {
1938 error_noexit "Setting \"max_dirty_mb\""
1941 $r $LCTL conf_param $fsname-OST0000.failover.node=$nid || {
1942 error_noexit "Setting OST \"failover.node\""
1945 $r $LCTL conf_param $fsname-MDT0000.mdc.max_rpcs_in_flight=9 || {
1946 error_noexit "Setting \"max_rpcs_in_flight\""
1949 $r $LCTL conf_param $fsname-MDT0000.failover.node=$nid || {
1950 error_noexit "Setting MDT \"failover.node\""
1953 $r $LCTL pool_new $fsname.interop || {
1954 error_noexit "Setting \"interop\""
1957 $r $LCTL conf_param $fsname-MDT0000.lov.stripesize=4M || {
1958 error_noexit "Setting \"lov.stripesize\""
1961 $r $LCTL conf_param $fsname-MDT0000.mdd.atime_diff=70 || {
1962 error_noexit "Setting \"mdd.atime_diff\""
1966 if [ "$ff_convert" != "no" -a $(facet_fstype ost1) == "ldiskfs" ]; then
1967 $r $LCTL lfsck_start -M $fsname-OST0000 || {
1968 error_noexit "Start OI scrub on OST0"
1972 # The oi_scrub should be on ost1, but for test_32(),
1973 # all on the SINGLEMDS.
1974 wait_update_facet $SINGLEMDS "$LCTL get_param -n \
1975 osd-ldiskfs.$fsname-OST0000.oi_scrub |
1976 awk '/^status/ { print \\\$2 }'" "completed" 30 || {
1977 error_noexit "Failed to get the expected 'completed'"
1981 local UPDATED=$($r $LCTL get_param -n \
1982 osd-ldiskfs.$fsname-OST0000.oi_scrub |
1983 awk '/^updated/ { print $2 }')
1984 [ $UPDATED -ge 1 ] || {
1985 error_noexit "Only $UPDATED objects have been converted"
1990 if [ "$dne_upgrade" != "no" ]; then
1991 $r $LCTL conf_param \
1992 $fsname-MDT0001.mdc.max_rpcs_in_flight=9 || {
1993 error_noexit "Setting MDT1 \"max_rpcs_in_flight\""
1996 $r $LCTL conf_param $fsname-MDT0001.failover.node=$nid || {
1997 error_noexit "Setting MDT1 \"failover.node\""
2000 $r $LCTL conf_param $fsname-MDT0001.lov.stripesize=4M || {
2001 error_noexit "Setting MDT1 \"lov.stripesize\""
2006 if [ "$writeconf" ]; then
2007 $MOUNT_CMD $nid:/$fsname $tmp/mnt/lustre || {
2008 error_noexit "Mounting the client"
2012 shall_cleanup_lustre=true
2013 $r $LCTL set_param debug="$PTLDEBUG"
2015 t32_verify_quota $node $fsname $tmp/mnt/lustre || {
2016 error_noexit "verify quota failed"
2020 if $r test -f $tmp/list; then
2022 # There is not a Test Framework API to copy files to or
2023 # from a remote node.
2025 # LU-2393 - do both sorts on same node to ensure locale
2027 local list_file=$tmp/list
2029 if $mdt2_is_available; then
2030 if [[ -d $tmp/mnt/lustre/striped_dir_old ]] &&
2031 $r test -f $tmp/list2; then
2032 list_file=$tmp/list2
2033 pushd $tmp/mnt/lustre/striped_dir_old
2035 pushd $tmp/mnt/lustre/remote_dir
2038 pushd $tmp/mnt/lustre
2040 $r cat $list_file | sort -k 6 >$tmp/list.orig
2041 ls -Rni --time-style=+%s | sort -k 6 |
2042 sed 's/\. / /' >$tmp/list || {
2048 # 32-bit and 64-bit clients use different algorithms to
2049 # convert FIDs into inode numbers. Hence, remove the
2050 # inode numbers from the lists, if the original list was
2051 # created on an architecture with different number of
2054 if [ $(t32_bits_per_long $(uname -m)) != \
2055 $(t32_bits_per_long $img_arch) ]; then
2056 echo "Different number of bits per \"long\"" \
2057 "from the disk image"
2058 for list in list.orig list; do
2059 sed -i -e 's/^[0-9]\+[ \t]\+//' \
2063 if ! diff -ub $tmp/list.orig $tmp/list; then
2064 error_noexit "list verification failed"
2068 echo "list verification skipped"
2071 if [ "$dne_upgrade" != "no" ]; then
2072 $LFS mkdir -i 1 -c2 $tmp/mnt/lustre/striped_dir || {
2073 error_noexit "set striped dir failed"
2077 $LFS setdirstripe -D -c2 $tmp/mnt/lustre/striped_dir
2079 pushd $tmp/mnt/lustre
2080 tar -cf - . --exclude=./striped_dir \
2081 --exclude=./striped_dir_old \
2082 --exclude=./remote_dir |
2083 tar -xvf - -C striped_dir 1>/dev/null || {
2084 error_noexit "cp to striped dir failed"
2090 # If it is upgrade from DNE (2.5), then rename the remote dir,
2091 # which is created in 2.5 to striped dir.
2092 if $mdt2_is_available && [[ "$dne_upgrade" != "no" ]]; then
2093 stripe_index=$($LFS getdirstripe -i \
2094 $tmp/mnt/lustre/remote_dir)
2096 [[ $stripe_index -eq 1 ]] || {
2097 error_noexit "get index \"$stripe_index\"" \
2098 "from remote dir failed"
2101 mv $tmp/mnt/lustre/remote_dir \
2102 $tmp/mnt/lustre/striped_dir/ || {
2103 error_noexit "mv remote dir failed"
2108 # If it is upgraded from DNE (2.7), then move the striped dir
2109 # which was created in 2.7 to the new striped dir.
2110 if $mdt2_is_available && [[ "$dne_upgrade" != "no" ]] &&
2111 [[ -d $tmp/mnt/lustre/striped_dir_old ]]; then
2112 stripe_count=$($LFS getdirstripe -c \
2113 $tmp/mnt/lustre/striped_dir_old)
2114 [[ $stripe_count -eq 2 ]] || {
2115 error_noexit "get count $stripe_count" \
2116 "from striped dir failed"
2119 mv $tmp/mnt/lustre/striped_dir_old \
2120 $tmp/mnt/lustre/striped_dir/ || {
2121 error_noexit "mv striped dir failed"
2127 $r $LCTL set_param -n osd*.*.force_sync=1
2128 dd if=/dev/zero of=$tmp/mnt/lustre/tmp_file bs=10k count=10 || {
2129 error_noexit "dd failed"
2132 rm -rf $tmp/mnt/lustre/tmp_file || {
2133 error_noexit "rm failed"
2137 if $r test -f $tmp/sha1sums; then
2138 # LU-2393 - do both sorts on same node to ensure locale
2140 $r cat $tmp/sha1sums | sort -k 2 >$tmp/sha1sums.orig
2141 if [ "$dne_upgrade" != "no" ]; then
2142 pushd $tmp/mnt/lustre/striped_dir
2144 pushd $tmp/mnt/lustre
2147 find ! -path "*remote_dir*" ! -path "*striped_dir*" \
2148 ! -name .lustre -type f -exec sha1sum {} \; |
2149 sort -k 2 >$tmp/sha1sums || {
2151 error_noexit "sha1sum"
2155 if ! diff -ub $tmp/sha1sums.orig $tmp/sha1sums; then
2156 error_noexit "sha1sum verification failed"
2160 # if upgrade from DNE(2.5), then check remote directory
2161 # if upgrade from DNE(2.7), then check striped directory
2162 if $mdt2_is_available &&
2163 [[ "$dne_upgrade" != "no" ]]; then
2164 local new_dir="$tmp/mnt/lustre/striped_dir"
2165 local striped_dir_old="$new_dir/striped_dir_old"
2167 local dir_list="$new_dir/remote_dir"
2168 [[ ! -d $triped_dir_old ]] ||
2169 dir_list+=" $striped_dir_old"
2171 for dir in $dir_list; do
2173 find ! -name .lustre -type f \
2174 -exec sha1sum {} \; |
2175 sort -k 2 >$tmp/sha1sums || {
2177 error_noexit "sha1sum"
2181 if ! diff -ub $tmp/sha1sums.orig \
2183 error_noexit "sha1sum $dir" \
2190 echo "sha1sum verification skipped"
2193 if [ "$dne_upgrade" != "no" ]; then
2194 rm -rf $tmp/mnt/lustre/striped_dir || {
2195 error_noexit "remove remote dir failed"
2200 # migrate files/dirs to remote MDT, then move them back
2201 if [ $(lustre_version_code mds1) -ge $(version_code 2.7.50) -a \
2202 $dne_upgrade != "no" ]; then
2203 $r $LCTL set_param -n \
2204 mdt.${fsname}*.enable_remote_dir=1 2>/dev/null
2206 echo "test migration"
2207 pushd $tmp/mnt/lustre
2208 for dir in $(find ! -name .lustre ! -name . -type d); do
2209 mdt_index=$($LFS getdirstripe -i $dir)
2210 stripe_cnt=$($LFS getdirstripe -c $dir)
2211 if [ $mdt_index = 0 -a $stripe_cnt -le 1 ]; then
2212 $LFS mv -M 1 $dir || {
2214 error_noexit "migrate MDT1 failed"
2220 for dir in $(find ! -name . ! -name .lustre -type d); do
2221 mdt_index=$($LFS getdirstripe -i $dir)
2222 stripe_cnt=$($LFS getdirstripe -c $dir)
2223 if [ $mdt_index = 1 -a $stripe_cnt -le 1 ]; then
2224 $LFS mv -M 0 $dir || {
2226 error_noexit "migrate MDT0 failed"
2235 # When adding new data verification tests, please check for
2236 # the presence of the required reference files first, like
2237 # the "sha1sums" and "list" tests above, to avoid the need to
2238 # regenerate every image for each test addition.
2241 nrpcs_orig=$($LCTL get_param \
2242 -n mdc.*MDT0000*.max_rpcs_in_flight) || {
2243 error_noexit "Getting \"max_rpcs_in_flight\""
2246 nrpcs=$((nrpcs_orig + 5))
2247 $r $LCTL conf_param $fsname-MDT0000.mdc.max_rpcs_in_flight=$nrpcs || {
2248 error_noexit "Changing \"max_rpcs_in_flight\""
2251 wait_update $HOSTNAME "$LCTL get_param \
2252 -n mdc.*MDT0000*.max_rpcs_in_flight" $nrpcs || {
2253 error_noexit "Verifying \"max_rpcs_in_flight\""
2257 umount $tmp/mnt/lustre || {
2258 error_noexit "Unmounting the client"
2261 shall_cleanup_lustre=false
2263 if [[ "$dne_upgrade" != "no" ]] || $mdt2_is_available; then
2264 $r $UMOUNT $tmp/mnt/mdt1 || {
2265 error_noexit "Unmounting the MDT2"
2268 shall_cleanup_mdt1=false
2271 $r $UMOUNT $tmp/mnt/mdt || {
2272 error_noexit "Unmounting the MDT"
2275 shall_cleanup_mdt=false
2277 $r $UMOUNT $tmp/mnt/ost || {
2278 error_noexit "Unmounting the OST"
2281 shall_cleanup_ost=false
2283 t32_reload_modules $node || {
2284 error_noexit "Reloading modules"
2288 if [[ $fstype == zfs ]]; then
2289 local poolname=t32fs-mdt1
2291 $ZPOOL list -H $poolname >/dev/null 2>&1 ||
2292 $ZPOOL import -f -d $tmp $poolname"
2294 # upgrade zpool to latest supported features,
2295 # including dnode quota accounting in 0.7.0
2296 $r "$ZPOOL upgrade $poolname"
2299 # mount a second time to make sure we didnt leave upgrade flag on
2300 $r $TUNEFS --dryrun $mdt_dev || {
2302 error_noexit "tunefs.lustre before remounting the MDT"
2306 mopts=exclude=$fsname-OST0000
2307 if [ $fstype == "ldiskfs" ]; then
2310 $r $MOUNT_CMD -o $mopts $mdt_dev $tmp/mnt/mdt || {
2311 error_noexit "Remounting the MDT"
2314 shall_cleanup_mdt=true
2324 for tarball in $tarballs; do
2325 t32_test $tarball || let "rc += $?"
2329 run_test 32a "Upgrade (not live)"
2337 for tarball in $tarballs; do
2338 t32_test $tarball writeconf || let "rc += $?"
2342 run_test 32b "Upgrade with writeconf"
2349 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
2351 for tarball in $tarballs; do
2352 # Do not support 1_8 and 2_1 direct upgrade to DNE2 anymore */
2353 echo $tarball | grep "1_8" && continue
2354 echo $tarball | grep "2_1" && continue
2356 dne_upgrade=yes t32_test $tarball writeconf || rc=$?
2360 run_test 32c "dne upgrade test"
2368 for tarball in $tarballs; do
2369 ff_convert=yes t32_test $tarball || rc=$?
2373 run_test 32d "convert ff test"
2375 test_33a() { # bug 12333, was test_33
2376 local FSNAME2=test-123
2377 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
2380 [ -n "$ost1_HOST" ] && fs2ost_HOST=$ost1_HOST
2382 if [ -z "$fs2ost_DEV" -o -z "$fs2mds_DEV" ]; then
2383 local dev=${SINGLEMDS}_dev
2384 local MDSDEV=${!dev}
2385 is_blkdev $SINGLEMDS $MDSDEV &&
2386 skip_env "mixed loopback and real device not working" &&
2390 local fs2mdsdev=$(mdsdevname 1_2)
2391 local fs2ostdev=$(ostdevname 1_2)
2392 local fs2mdsvdev=$(mdsvdevname 1_2)
2393 local fs2ostvdev=$(ostvdevname 1_2)
2395 if [ $(facet_fstype mds1) == ldiskfs ]; then
2396 mkfsoptions="--mkfsoptions=\\\"-J size=8\\\"" # See bug 17931.
2399 add fs2mds $(mkfs_opts mds1 ${fs2mdsdev}) --mgs --fsname=${FSNAME2} \
2400 --reformat $mkfsoptions $fs2mdsdev $fs2mdsvdev || exit 10
2401 add fs2ost $(mkfs_opts ost1 ${fs2ostdev}) --mgsnode=$MGSNID \
2402 --fsname=${FSNAME2} --index=8191 --reformat $fs2ostdev \
2403 $fs2ostvdev || exit 10
2405 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS && trap cleanup_fs2 EXIT INT
2406 start fs2ost $fs2ostdev $OST_MOUNT_OPTS
2407 do_facet $SINGLEMDS "$LCTL conf_param $FSNAME2.sys.timeout=200" ||
2408 error "$LCTL conf_param $FSNAME2.sys.timeout=200 failed"
2409 mkdir -p $MOUNT2 || error "mkdir $MOUNT2 failed"
2410 $MOUNT_CMD $MGSNID:/${FSNAME2} $MOUNT2 || error "$MOUNT_CMD failed"
2413 cp /etc/hosts $MOUNT2/ || error "copy /etc/hosts $MOUNT2/ failed"
2414 $GETSTRIPE $MOUNT2/hosts || error "$GETSTRIPE $MOUNT2/hosts failed"
2419 cleanup_nocli || error "cleanup_nocli failed with $?"
2421 run_test 33a "Mount ost with a large index number"
2423 test_33b() { # was test_34
2426 do_facet client dd if=/dev/zero of=$MOUNT/24 bs=1024k count=1
2427 # Drop lock cancelation reply during umount
2428 #define OBD_FAIL_LDLM_CANCEL_NET 0x304
2429 do_facet client $LCTL set_param fail_loc=0x80000304
2430 #lctl set_param debug=-1
2431 umount_client $MOUNT
2432 cleanup || error "cleanup failed with $?"
2434 run_test 33b "Drop cancel during umount"
2438 do_facet client "sh runmultiop_bg_pause $DIR/file O_c"
2439 manual_umount_client
2441 do_facet client killall -USR1 multiop
2442 if [ $rc -eq 0 ]; then
2443 error "umount not fail!"
2446 cleanup || error "cleanup failed with rc $?"
2448 run_test 34a "umount with opened file should be fail"
2452 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
2453 stop_mds || error "Unable to stop MDS"
2455 manual_umount_client --force || error "mtab after failed umount with $?"
2457 cleanup || error "cleanup failed with $?"
2459 run_test 34b "force umount with failed mds should be normal"
2463 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
2464 stop_ost || error "Unable to stop OST1"
2466 manual_umount_client --force || error "mtab after failed umount with $?"
2468 cleanup || error "cleanup failed with $?"
2470 run_test 34c "force umount with failed ost should be normal"
2472 test_35a() { # bug 12459
2475 DBG_SAVE="`$LCTL get_param -n debug`"
2476 $LCTL set_param debug="ha"
2478 log "Set up a fake failnode for the MDS"
2480 local device=$(do_facet $SINGLEMDS "$LCTL get_param -n devices" |
2481 awk '($3 ~ "mdt" && $4 ~ "MDT") { print $4 }' | head -1)
2482 do_facet mgs "$LCTL conf_param \
2483 ${device}.failover.node=$(h2$NETTYPE $FAKENID)" ||
2484 error "Setting ${device}.failover.node=\
2485 $(h2$NETTYPE $FAKENID) failed."
2487 log "Wait for RECONNECT_INTERVAL seconds (10s)"
2490 MSG="conf-sanity.sh test_35a `date +%F%kh%Mm%Ss`"
2493 log "Stopping the MDT: $device"
2494 stop_mdt 1 || error "MDT0 stop fail"
2496 df $MOUNT > /dev/null 2>&1 &
2498 log "Restarting the MDT: $device"
2499 start_mdt 1 || error "MDT0 start fail"
2500 log "Wait for df ($DFPID) ... "
2503 $LCTL set_param debug="$DBG_SAVE"
2505 # retrieve from the log the first server that the client tried to
2506 # contact after the connection loss
2507 $LCTL dk $TMP/lustre-log-$TESTNAME.log
2508 NEXTCONN=`awk "/${MSG}/ {start = 1;}
2509 /import_select_connection.*$device-mdc.* using connection/ {
2511 if (\\\$NF ~ /$FAKENID/)
2517 }" $TMP/lustre-log-$TESTNAME.log`
2518 [ "$NEXTCONN" != "0" ] &&
2519 error "Tried to connect to ${NEXTCONN} not last active server"
2520 cleanup || error "cleanup failed with $?"
2521 # remove nid settings
2522 writeconf_or_reformat
2524 run_test 35a "Reconnect to the last active server first"
2526 test_35b() { # bug 18674
2527 remote_mds || { skip "local MDS" && return 0; }
2531 $LCTL set_param debug="ha"
2533 MSG="conf-sanity.sh test_35b `date +%F%kh%Mm%Ss`"
2536 log "Set up a fake failnode for the MDS"
2538 local device=$(do_facet $SINGLEMDS "$LCTL get_param -n devices" |
2539 awk '($3 ~ "mdt" && $4 ~ "MDT") { print $4 }' | head -1)
2540 do_facet mgs "$LCTL conf_param \
2541 ${device}.failover.node=$(h2$NETTYPE $FAKENID)" ||
2542 error "Set ${device}.failover.node=\
2543 $(h2$NETTYPE $FAKENID) failed"
2545 local at_max_saved=0
2546 # adaptive timeouts may prevent seeing the issue
2547 if at_is_enabled; then
2548 at_max_saved=$(at_max_get mds)
2549 at_max_set 0 mds client
2552 mkdir $MOUNT/$tdir || error "mkdir $MOUNT/$tdir failed"
2554 log "Injecting EBUSY on MDS"
2555 # Setting OBD_FAIL_MDS_RESEND=0x136
2556 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0x80000136" ||
2557 error "unable to set param fail_loc=0x80000136"
2559 $LCTL set_param mdc.${FSNAME}*.stats=clear
2561 log "Creating a test file and stat it"
2562 touch $MOUNT/$tdir/$tfile || error "touch $MOUNT/$tdir/$tfile failed"
2563 stat $MOUNT/$tdir/$tfile
2565 log "Stop injecting EBUSY on MDS"
2566 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0" ||
2567 error "unable to set param fail_loc=0"
2568 rm -f $MOUNT/$tdir/$tfile || error "remove $MOUNT/$tdir/$tfile failed"
2571 # restore adaptive timeout
2572 [ $at_max_saved -ne 0 ] && at_max_set $at_max_saved mds client
2574 $LCTL dk $TMP/lustre-log-$TESTNAME.log
2576 CONNCNT=$($LCTL get_param mdc.${FSNAME}*.stats |
2577 awk '/mds_connect/{print $2}')
2579 # retrieve from the log if the client has ever tried to
2580 # contact the fake server after the loss of connection
2581 FAILCONN=`awk "BEGIN {ret = 0;}
2582 /import_select_connection.*${FSNAME}-MDT0000-mdc.* using connection/ {
2584 if (\\\$NF ~ /$FAKENID/) {
2589 END {print ret}" $TMP/lustre-log-$TESTNAME.log`
2591 [ "$FAILCONN" == "0" ] &&
2592 error "The client reconnection has not been triggered"
2593 [ "$FAILCONN" == "2" ] &&
2594 error "Primary server busy, client reconnect to failover failed"
2597 # When OBD_FAIL_MDS_RESEND is hit, we sleep for 2 * obd_timeout
2598 # Reconnects are supposed to be rate limited to one every 5s
2599 [ $CONNCNT -gt $((2 * $TIMEOUT / 5 + 1)) ] &&
2600 error "Too many reconnects $CONNCNT"
2602 cleanup || error "cleanup failed with $?"
2603 # remove nid settings
2604 writeconf_or_reformat
2606 run_test 35b "Continue reconnection retries, if the active server is busy"
2609 [ $OSTCOUNT -lt 2 ] && skip_env "skipping test for single OST" && return
2611 [ "$ost_HOST" = "`hostname`" -o "$ost1_HOST" = "`hostname`" ] ||
2612 { skip "remote OST" && return 0; }
2615 local FSNAME2=test1234
2616 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
2618 [ -n "$ost1_HOST" ] && fs2ost_HOST=$ost1_HOST && fs3ost_HOST=$ost1_HOST
2620 if [ -z "$fs2ost_DEV" -o -z "$fs2mds_DEV" -o -z "$fs3ost_DEV" ]; then
2621 is_blkdev $SINGLEMDS $MDSDEV &&
2622 skip_env "mixed loopback and real device not working" && return
2625 local fs2mdsdev=$(mdsdevname 1_2)
2626 local fs2ostdev=$(ostdevname 1_2)
2627 local fs3ostdev=$(ostdevname 2_2)
2628 local fs2mdsvdev=$(mdsvdevname 1_2)
2629 local fs2ostvdev=$(ostvdevname 1_2)
2630 local fs3ostvdev=$(ostvdevname 2_2)
2632 add fs2mds $(mkfs_opts mds1 ${fs2mdsdev}) --mgs --fsname=${FSNAME2} \
2633 --reformat $fs2mdsdev $fs2mdsvdev || exit 10
2634 # XXX after we support non 4K disk blocksize in ldiskfs, specify a
2635 # different one than the default value here.
2636 add fs2ost $(mkfs_opts ost1 ${fs2ostdev}) --mgsnode=$MGSNID \
2637 --fsname=${FSNAME2} --reformat $fs2ostdev $fs2ostvdev || exit 10
2638 add fs3ost $(mkfs_opts ost2 ${fs3ostdev}) --mgsnode=$MGSNID \
2639 --fsname=${FSNAME2} --reformat $fs3ostdev $fs3ostvdev || exit 10
2641 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS
2642 start fs2ost $fs2ostdev $OST_MOUNT_OPTS
2643 start fs3ost $fs3ostdev $OST_MOUNT_OPTS
2644 mkdir -p $MOUNT2 || error "mkdir $MOUNT2 failed"
2645 $MOUNT_CMD $MGSNID:/${FSNAME2} $MOUNT2 || error "$MOUNT_CMD failed"
2647 sleep 5 # until 11778 fixed
2649 dd if=/dev/zero of=$MOUNT2/$tfile bs=1M count=7 || error "dd failed"
2651 BKTOTAL=$($LCTL get_param -n obdfilter.*.kbytestotal |
2652 awk 'BEGIN{total=0}; {total+=$1}; END{print total}')
2653 BKFREE=$($LCTL get_param -n obdfilter.*.kbytesfree |
2654 awk 'BEGIN{free=0}; {free+=$1}; END{print free}')
2655 BKAVAIL=$($LCTL get_param -n obdfilter.*.kbytesavail |
2656 awk 'BEGIN{avail=0}; {avail+=$1}; END{print avail}')
2657 STRING=$(df -P $MOUNT2 | tail -n 1 | awk '{print $2","$3","$4}')
2658 DFTOTAL=$(echo $STRING | cut -d, -f1)
2659 DFUSED=$(echo $STRING | cut -d, -f2)
2660 DFAVAIL=$(echo $STRING | cut -d, -f3)
2661 DFFREE=$(($DFTOTAL - $DFUSED))
2663 ALLOWANCE=$((64 * $OSTCOUNT))
2665 if [ $DFTOTAL -lt $(($BKTOTAL - $ALLOWANCE)) ] ||
2666 [ $DFTOTAL -gt $(($BKTOTAL + $ALLOWANCE)) ] ; then
2667 echo "**** FAIL: df total($DFTOTAL) mismatch OST total($BKTOTAL)"
2670 if [ $DFFREE -lt $(($BKFREE - $ALLOWANCE)) ] ||
2671 [ $DFFREE -gt $(($BKFREE + $ALLOWANCE)) ] ; then
2672 echo "**** FAIL: df free($DFFREE) mismatch OST free($BKFREE)"
2675 if [ $DFAVAIL -lt $(($BKAVAIL - $ALLOWANCE)) ] ||
2676 [ $DFAVAIL -gt $(($BKAVAIL + $ALLOWANCE)) ] ; then
2677 echo "**** FAIL: df avail($DFAVAIL) mismatch OST avail($BKAVAIL)"
2682 stop fs3ost -f || error "unable to stop OST3"
2683 stop fs2ost -f || error "unable to stop OST2"
2684 stop fs2mds -f || error "unable to stop second MDS"
2685 unload_modules_conf || error "unable unload modules"
2688 run_test 36 "df report consistency on OSTs with different block size"
2691 local mntpt=$(facet_mntpt $SINGLEMDS)
2692 local mdsdev=$(mdsdevname ${SINGLEMDS//mds/})
2693 local mdsdev_sym="$TMP/sym_mdt.img"
2694 local opts=$MDS_MOUNT_OPTS
2697 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
2698 skip "Currently only applicable to ldiskfs-based MDTs"
2702 echo "MDS : $mdsdev"
2703 echo "SYMLINK : $mdsdev_sym"
2704 do_facet $SINGLEMDS rm -f $mdsdev_sym
2706 do_facet $SINGLEMDS ln -s $mdsdev $mdsdev_sym
2708 echo "mount symlink device - $mdsdev_sym"
2710 if ! do_facet $SINGLEMDS test -b $mdsdev; then
2711 opts=$(csa_add "$opts" -o loop)
2713 mount_op=$(do_facet $SINGLEMDS mount -v -t lustre $opts \
2714 $mdsdev_sym $mntpt 2>&1)
2717 echo mount_op=$mount_op
2719 do_facet $SINGLEMDS "$UMOUNT $mntpt && rm -f $mdsdev_sym"
2721 if $(echo $mount_op | grep -q "unable to set tunable"); then
2722 error "set tunables failed for symlink device"
2725 [ $rc -eq 0 ] || error "mount symlink $mdsdev_sym failed! rc=$rc"
2727 run_test 37 "verify set tunables works for symlink device"
2729 test_38() { # bug 14222
2730 local fstype=$(facet_fstype $SINGLEMDS)
2731 local mntpt=$(facet_mntpt $SINGLEMDS)
2736 local SRC="/etc /bin"
2737 local FILES=$(find $SRC -type f -mtime +1 | head -n $COUNT)
2738 log "copying $(echo $FILES | wc -w) files to $DIR/$tdir"
2739 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
2740 tar cf - $FILES | tar xf - -C $DIR/$tdir ||
2741 error "copying $SRC to $DIR/$tdir"
2743 umount_client $MOUNT || error "umount_client $MOUNT failed"
2744 do_facet $SINGLEMDS "$LCTL get_param osp.*.prealloc_next_id"
2745 stop_mds || error "Unable to stop MDS"
2746 log "delete lov_objid file on MDS"
2748 mount_fstype $SINGLEMDS || error "mount MDS failed (1)"
2750 do_facet $SINGLEMDS "od -Ax -td8 $mntpt/lov_objid; rm $mntpt/lov_objid"
2752 unmount_fstype $SINGLEMDS || error "umount failed (1)"
2754 # check create in mds_lov_connect
2755 start_mds || error "unable to start MDS"
2756 mount_client $MOUNT || error "mount_client $MOUNT failed"
2758 [ $V ] && log "verifying $DIR/$tdir/$f"
2759 diff -q $f $DIR/$tdir/$f || ERROR=y
2761 do_facet $SINGLEMDS "$LCTL get_param osp.*.prealloc_next_id"
2762 if [ "$ERROR" = "y" ]; then
2763 # check it's updates in sync
2764 umount_client $MOUNT
2766 mount_fstype $SIGNLEMDS
2767 do_facet $SINGLEMDS "od -Ax -td8 $mntpt/lov_objid"
2768 unmount_fstype $SINGLEMDS
2769 error "old and new files are different after connect" || true
2771 touch $DIR/$tdir/f2 || error "f2 file create failed"
2773 # check it's updates in sync
2774 umount_client $MOUNT || error "second umount_client $MOUNT failed"
2777 mount_fstype $SINGLEMDS || error "mount MDS failed (3)"
2779 do_facet $SINGLEMDS "od -Ax -td8 $mntpt/lov_objid"
2780 do_facet $SINGLEMDS dd if=/dev/zero of=$mntpt/lov_objid.clear count=8
2782 unmount_fstype $SINGLEMDS || error "umount failed (3)"
2784 start_mds || error "unable to start MDS"
2785 mount_client $MOUNT || error "mount_client $MOUNT failed"
2787 [ $V ] && log "verifying $DIR/$tdir/$f"
2788 diff -q $f $DIR/$tdir/$f || ERROR=y
2790 touch $DIR/$tdir/f3 || error "f3 file create failed"
2791 do_facet $SINGLEMDS "$LCTL get_param osp.*.prealloc_next_id"
2792 umount_client $MOUNT || error "third umount_client $MOUNT failed"
2794 mount_fstype $SINGLEMDS || error "mount MDS failed (4)"
2795 do_facet $SINGLEMDS "od -Ax -td8 $mntpt/lov_objid"
2796 unmount_fstype $SINGLEMDS || error "umount failed (4)"
2798 [ "$ERROR" = "y" ] &&
2799 error "old and new files are different after sync" || true
2801 log "files compared the same"
2802 cleanup || error "cleanup failed with $?"
2804 run_test 38 "MDS recreates missing lov_objid file from OST data"
2809 cleanup || error "cleanup failed with $?"
2810 perl $SRCDIR/leak_finder.pl $TMP/debug 2>&1 | egrep '*** Leak:' &&
2811 error "memory leak detected" || true
2813 run_test 39 "leak_finder recognizes both LUSTRE and LNET malloc messages"
2815 test_40() { # bug 15759
2816 start_ost || error "Unable to start OST1"
2817 #define OBD_FAIL_TGT_TOOMANY_THREADS 0x706
2818 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0x80000706"
2820 cleanup || error "cleanup failed with rc $?"
2822 run_test 40 "race during service thread startup"
2824 test_41a() { #bug 14134
2825 if [ $(facet_fstype $SINGLEMDS) == ldiskfs ] &&
2826 ! do_facet $SINGLEMDS test -b $(mdsdevname 1); then
2827 skip "Loop devices does not work with nosvc option"
2832 { skip "needs combined MGT and MDT device" && return 0; }
2834 start_mdt 1 -o nosvc -n
2835 if [ $MDSCOUNT -ge 2 ]; then
2836 for num in $(seq 2 $MDSCOUNT); do
2837 start_mdt $num || return
2840 start ost1 $(ostdevname 1) $OST_MOUNT_OPTS
2841 start_mdt 1 -o nomgs,force
2842 mount_client $MOUNT || error "mount_client $MOUNT failed"
2845 echo "blah blah" > $MOUNT/$tfile
2848 umount_client $MOUNT || error "umount_client $MOUNT failed"
2849 stop ost1 -f || error "unable to stop OST1"
2850 stop_mds || error "Unable to stop MDS"
2851 stop_mds || error "Unable to stop MDS on second try"
2853 run_test 41a "mount mds with --nosvc and --nomgs"
2856 if [ $(facet_fstype $SINGLEMDS) == ldiskfs ] &&
2857 ! do_facet $SINGLEMDS test -b $(mdsdevname 1); then
2858 skip "Loop devices does not work with nosvc option"
2862 ! combined_mgs_mds && skip "needs combined mgs device" && return 0
2866 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
2868 start_mdt 1 -o nosvc -n
2869 if [ $MDSCOUNT -ge 2 ]; then
2870 for num in $(seq 2 $MDSCOUNT); do
2871 start_mdt $num || return
2874 start_ost || error "Unable to start OST1"
2875 start_mdt 1 -o nomgs,force
2876 mount_client $MOUNT || error "mount_client $MOUNT failed"
2879 echo "blah blah" > $MOUNT/$tfile
2880 cat $MOUNT/$tfile || error "cat $MOUNT/$tfile failed"
2882 umount_client $MOUNT || error "umount_client $MOUNT failed"
2883 stop_ost || error "Unable to stop OST1"
2884 stop_mds || error "Unable to stop MDS"
2885 stop_mds || error "Unable to stop MDS on second try"
2887 run_test 41b "mount mds with --nosvc and --nomgs on first mount"
2890 local server_version=$(lustre_version_code $SINGLEMDS)
2891 local oss_list=$(comma_list $(osts_nodes))
2893 [[ $server_version -ge $(version_code 2.6.52) ]] ||
2894 [[ $server_version -ge $(version_code 2.5.26) &&
2895 $server_version -lt $(version_code 2.5.50) ]] ||
2896 [[ $server_version -ge $(version_code 2.5.4) &&
2897 $server_version -lt $(version_code 2.5.11) ]] ||
2898 { skip "Need MDS version 2.5.4+ or 2.5.26+ or 2.6.52+"; return; }
2900 # ensure mds1 ost1 have been created even if running sub-test standalone
2903 cleanup || error "cleanup failed"
2905 # using directly mount command instead of start() function to avoid
2906 # any side effect of // with others/externals tools/features
2907 # ("zpool import", ...)
2909 # MDT concurrent start
2911 LOAD_MODULES_REMOTE=true load_modules
2912 do_facet $SINGLEMDS "lsmod | grep -q libcfs" ||
2913 error "MDT concurrent start: libcfs module not loaded"
2915 local mds1dev=$(mdsdevname 1)
2916 local mds1mnt=$(facet_mntpt mds1)
2917 local mds1fstype=$(facet_fstype mds1)
2918 local mds1opts=$MDS_MOUNT_OPTS
2920 if [ $mds1fstype == ldiskfs ] &&
2921 ! do_facet mds1 test -b $mds1dev; then
2922 mds1opts=$(csa_add "$mds1opts" -o loop)
2924 if [[ $mds1fstype == zfs ]]; then
2925 import_zpool mds1 || return ${PIPESTATUS[0]}
2928 #define OBD_FAIL_TGT_MOUNT_RACE 0x716
2929 do_facet mds1 "$LCTL set_param fail_loc=0x80000716"
2931 do_facet mds1 mount -t lustre $mds1dev $mds1mnt $mds1opts &
2934 do_facet mds1 mount -t lustre $mds1dev $mds1mnt $mds1opts
2938 do_facet mds1 "$LCTL set_param fail_loc=0x0"
2939 if [ $rc -eq 0 ] && [ $rc2 -ne 0 ]; then
2940 echo "1st MDT start succeed"
2941 echo "2nd MDT start failed with $rc2"
2942 elif [ $rc2 -eq 0 ] && [ $rc -ne 0 ]; then
2943 echo "1st MDT start failed with $rc"
2944 echo "2nd MDT start succeed"
2947 error "unexpected concurrent MDT mounts result, rc=$rc rc2=$rc2"
2950 if [ $MDSCOUNT -ge 2 ]; then
2951 for num in $(seq 2 $MDSCOUNT); do
2952 start_mdt $num || return
2956 # OST concurrent start
2958 do_rpc_nodes $oss_list "lsmod | grep -q libcfs" ||
2959 error "OST concurrent start: libcfs module not loaded"
2961 local ost1dev=$(ostdevname 1)
2962 local ost1mnt=$(facet_mntpt ost1)
2963 local ost1fstype=$(facet_fstype ost1)
2964 local ost1opts=$OST_MOUNT_OPTS
2966 if [ $ost1fstype == ldiskfs ] &&
2967 ! do_facet ost1 test -b $ost1dev; then
2968 ost1opts=$(csa_add "$ost1opts" -o loop)
2970 if [[ $ost1fstype == zfs ]]; then
2971 import_zpool ost1 || return ${PIPESTATUS[0]}
2974 #define OBD_FAIL_TGT_MOUNT_RACE 0x716
2975 do_facet ost1 "$LCTL set_param fail_loc=0x80000716"
2977 do_facet ost1 mount -t lustre $ost1dev $ost1mnt $ost1opts &
2980 do_facet ost1 mount -t lustre $ost1dev $ost1mnt $ost1opts
2984 do_facet ost1 "$LCTL set_param fail_loc=0x0"
2985 if [ $rc -eq 0 ] && [ $rc2 -ne 0 ]; then
2986 echo "1st OST start succeed"
2987 echo "2nd OST start failed with $rc2"
2988 elif [ $rc2 -eq 0 ] && [ $rc -ne 0 ]; then
2989 echo "1st OST start failed with $rc"
2990 echo "2nd OST start succeed"
2994 error "unexpected concurrent OST mounts result, rc=$rc rc2=$rc2"
3000 # verify everything ok
3005 error "MDT(s) start failed"
3013 error "OST(s) start failed"
3021 error "client start failed"
3028 error "client mount failed"
3032 run_test 41c "concurrent mounts of MDT/OST should all fail but one"
3034 test_42() { #bug 14693
3036 check_mount || error "client was not mounted"
3038 do_facet mgs $LCTL conf_param $FSNAME.llite.some_wrong_param=10
3039 umount_client $MOUNT ||
3040 error "unmounting client failed with invalid llite param"
3041 mount_client $MOUNT ||
3042 error "mounting client failed with invalid llite param"
3044 do_facet mgs $LCTL conf_param $FSNAME.sys.some_wrong_param=20
3045 cleanup || error "stopping $FSNAME failed with invalid sys param"
3047 check_mount || error "client was not mounted with invalid sys param"
3048 cleanup || error "stopping $FSNAME failed with invalid sys param"
3050 run_test 42 "allow client/server mount/unmount with invalid config param"
3053 [[ $(lustre_version_code mgs) -ge $(version_code 2.5.58) ]] ||
3054 { skip "Need MDS version at least 2.5.58" && return 0; }
3055 [ $UID -ne 0 -o $RUNAS_ID -eq 0 ] && skip_env "run as root"
3058 USER1=$(getent passwd | grep :$ID1:$ID1: | cut -d: -f1)
3059 [ -z "$USER1" ] && skip_env "missing user with uid=$ID1 gid=$ID1" &&
3063 chmod ugo+x $DIR || error "chmod 0 failed"
3064 set_conf_param_and_check mds1 \
3065 "$LCTL get_param -n mdt.$FSNAME-MDT0000.root_squash" \
3066 "$FSNAME.mdt.root_squash" \
3068 wait_update $HOSTNAME \
3069 "$LCTL get_param -n llite.${FSNAME}*.root_squash" \
3071 error "check llite root_squash failed!"
3072 set_conf_param_and_check mds1 \
3073 "$LCTL get_param -n mdt.$FSNAME-MDT0000.nosquash_nids" \
3074 "$FSNAME.mdt.nosquash_nids" \
3076 wait_update $HOSTNAME \
3077 "$LCTL get_param -n llite.${FSNAME}*.nosquash_nids" \
3079 error "check llite nosquash_nids failed!"
3082 # create set of test files
3084 echo "111" > $DIR/$tfile-userfile || error "write 1 failed"
3085 chmod go-rw $DIR/$tfile-userfile || error "chmod 1 failed"
3086 chown $RUNAS_ID.$RUNAS_ID $DIR/$tfile-userfile || error "chown failed"
3088 echo "222" > $DIR/$tfile-rootfile || error "write 2 failed"
3089 chmod go-rw $DIR/$tfile-rootfile || error "chmod 2 faield"
3091 mkdir $DIR/$tdir-rootdir || error "mkdir failed"
3092 chmod go-rwx $DIR/$tdir-rootdir || error "chmod 3 failed"
3093 touch $DIR/$tdir-rootdir/tfile-1 || error "touch failed"
3095 echo "777" > $DIR/$tfile-user1file || error "write 7 failed"
3096 chmod go-rw $DIR/$tfile-user1file || error "chmod 7 failed"
3097 chown $ID1.$ID1 $DIR/$tfile-user1file || error "chown failed"
3100 # check root_squash:
3101 # set root squash UID:GID to RUNAS_ID
3102 # root should be able to access only files owned by RUNAS_ID
3104 set_conf_param_and_check mds1 \
3105 "$LCTL get_param -n mdt.$FSNAME-MDT0000.root_squash" \
3106 "$FSNAME.mdt.root_squash" \
3107 "$RUNAS_ID:$RUNAS_ID"
3108 wait_update $HOSTNAME \
3109 "$LCTL get_param -n llite.${FSNAME}*.root_squash" \
3110 "$RUNAS_ID:$RUNAS_ID" ||
3111 error "check llite root_squash failed!"
3113 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-userfile)
3114 dd if=$DIR/$tfile-userfile 1>/dev/null 2>/dev/null ||
3115 error "$ST: root read permission is denied"
3116 echo "$ST: root read permission is granted - ok"
3119 dd conv=notrunc of=$DIR/$tfile-userfile 1>/dev/null 2>/dev/null ||
3120 error "$ST: root write permission is denied"
3121 echo "$ST: root write permission is granted - ok"
3123 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-rootfile)
3124 dd if=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null &&
3125 error "$ST: root read permission is granted"
3126 echo "$ST: root read permission is denied - ok"
3129 dd conv=notrunc of=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null &&
3130 error "$ST: root write permission is granted"
3131 echo "$ST: root write permission is denied - ok"
3133 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tdir-rootdir)
3134 rm $DIR/$tdir-rootdir/tfile-1 1>/dev/null 2>/dev/null &&
3135 error "$ST: root unlink permission is granted"
3136 echo "$ST: root unlink permission is denied - ok"
3138 touch $DIR/tdir-rootdir/tfile-2 1>/dev/null 2>/dev/null &&
3139 error "$ST: root create permission is granted"
3140 echo "$ST: root create permission is denied - ok"
3143 # check root_squash is enforced independently
3144 # of client cache content
3146 # access file by USER1, keep access open
3147 # root should be denied access to user file
3149 runas -u $ID1 tail -f $DIR/$tfile-user1file 1>/dev/null 2>&1 &
3153 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-user1file)
3154 dd if=$DIR/$tfile-user1file 1>/dev/null 2>&1 &&
3155 { kill $pid; error "$ST: root read permission is granted"; }
3156 echo "$ST: root read permission is denied - ok"
3159 dd conv=notrunc of=$DIR/$tfile-user1file 1>/dev/null 2>&1 &&
3160 { kill $pid; error "$ST: root write permission is granted"; }
3161 echo "$ST: root write permission is denied - ok"
3167 # check nosquash_nids:
3168 # put client's NID into nosquash_nids list,
3169 # root should be able to access root file after that
3171 local NIDLIST=$($LCTL list_nids all | tr '\n' ' ')
3172 NIDLIST="2@gni $NIDLIST 192.168.0.[2,10]@tcp"
3173 NIDLIST=$(echo $NIDLIST | tr -s ' ' ' ')
3174 set_conf_param_and_check mds1 \
3175 "$LCTL get_param -n mdt.$FSNAME-MDT0000.nosquash_nids" \
3176 "$FSNAME-MDTall.mdt.nosquash_nids" \
3178 wait_update $HOSTNAME \
3179 "$LCTL get_param -n llite.${FSNAME}*.nosquash_nids" \
3181 error "check llite nosquash_nids failed!"
3183 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-rootfile)
3184 dd if=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null ||
3185 error "$ST: root read permission is denied"
3186 echo "$ST: root read permission is granted - ok"
3189 dd conv=notrunc of=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null ||
3190 error "$ST: root write permission is denied"
3191 echo "$ST: root write permission is granted - ok"
3193 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tdir-rootdir)
3194 rm $DIR/$tdir-rootdir/tfile-1 ||
3195 error "$ST: root unlink permission is denied"
3196 echo "$ST: root unlink permission is granted - ok"
3197 touch $DIR/$tdir-rootdir/tfile-2 ||
3198 error "$ST: root create permission is denied"
3199 echo "$ST: root create permission is granted - ok"
3200 cleanup || error "cleanup failed with $?"
3202 run_test 43a "check root_squash and nosquash_nids"
3204 test_43b() { # LU-5690
3205 [[ $(lustre_version_code mgs) -ge $(version_code 2.7.62) ]] ||
3206 { skip "Need MGS version 2.7.62+"; return; }
3208 if [[ -z "$fs2mds_DEV" ]]; then
3209 is_blkdev $SINGLEMDS $(mdsdevname ${SINGLEMDS//mds/}) &&
3210 skip_env "mixed loopback and real device not working" && return
3213 local fs2mdsdev=$(mdsdevname 1_2)
3214 local fs2mdsvdev=$(mdsvdevname 1_2)
3216 # temporarily use fs2mds as fs2mgs
3218 local fs2mgsdev=$fs2mdsdev
3219 local fs2mgsvdev=$fs2mdsvdev
3221 local fsname=test1234
3223 load_module llite/lustre
3224 local client_ip=$(host_nids_address $HOSTNAME $NETTYPE)
3225 local host=${client_ip//*./}
3226 local net=${client_ip/%$host/}
3227 local nosquash_nids=$(h2$NETTYPE $net[$host,$host,$host])
3229 add $fs2mgs $(mkfs_opts mgs $fs2mgsdev) --fsname=$fsname \
3230 --param mdt.root_squash=$RUNAS_ID:$RUNAS_ID \
3231 --param mdt.nosquash_nids=$nosquash_nids \
3232 --reformat $fs2mgsdev $fs2mgsvdev || error "add fs2mgs failed"
3233 start $fs2mgs $fs2mgsdev $MGS_MOUNT_OPTS || error "start fs2mgs failed"
3234 stop $fs2mgs -f || error "stop fs2mgs failed"
3236 run_test 43b "parse nosquash_nids with commas in expr_list"
3238 umount_client $MOUNT
3243 check_mount || error "check_mount"
3244 UUID=$($LCTL get_param llite.${FSNAME}*.uuid | cut -d= -f2)
3246 UUIDS=$(do_facet $SINGLEMDS "$LCTL get_param mdt.${FSNAME}*.exports.*.uuid")
3247 for VAL in $UUIDS; do
3248 NID=$(echo $VAL | cut -d= -f1)
3249 CLUUID=$(echo $VAL | cut -d= -f2)
3250 [ "$UUID" = "$CLUUID" ] && STATS_FOUND=yes && break
3252 [ "$STATS_FOUND" = "no" ] && error "stats not found for client"
3253 cleanup || error "cleanup failed with $?"
3255 run_test 44 "mounted client proc entry exists"
3259 check_mount || error "check_mount"
3260 stop_mds || error "Unable to stop MDS"
3264 #define OBD_FAIL_PTLRPC_LONG_REPL_UNLINK 0x50f
3265 do_facet client "$LCTL set_param fail_loc=0x8000050f"
3268 manual_umount_client --force || error "manual_umount_client failed"
3269 do_facet client "$LCTL set_param fail_loc=0x0"
3270 start_mds || error "unable to start MDS"
3271 mount_client $MOUNT || error "mount_client $MOUNT failed"
3272 cleanup || error "cleanup failed with $?"
3274 run_test 45 "long unlink handling in ptlrpcd"
3281 umount_client $MOUNT2 || rc=$?
3282 umount_client $MOUNT || rc=$?
3283 while [ $count -gt 0 ]; do
3284 stop ost${count} -f || rc=$?
3288 cleanup_nocli || rc=$?
3289 #writeconf to remove all ost2 traces for subsequent tests
3290 writeconf_or_reformat
3295 echo "Testing with $OSTCOUNT OSTs"
3297 start_mds || error "unable to start MDS"
3298 #first client should see only one ost
3299 start_ost || error "Unable to start OST1"
3300 wait_osc_import_state mds ost FULL
3302 mount_client $MOUNT || error "mount_client $MOUNT failed"
3303 trap "cleanup_46a $OSTCOUNT" EXIT ERR
3306 for (( i=2; i<=$OSTCOUNT; i++ )); do
3307 start ost$i $(ostdevname $i) $OST_MOUNT_OPTS ||
3308 error "start_ost$i $(ostdevname $i) failed"
3311 # wait until osts in sync
3312 for (( i=2; i<=$OSTCOUNT; i++ )); do
3313 wait_osc_import_state mds ost$i FULL
3314 wait_osc_import_state client ost$i FULL
3317 #second client see all ost's
3319 mount_client $MOUNT2 || error "mount_client failed"
3320 $SETSTRIPE -c -1 $MOUNT2 || error "$SETSTRIPE -c -1 $MOUNT2 failed"
3321 $GETSTRIPE $MOUNT2 || error "$GETSTRIPE $MOUNT2 failed"
3323 echo "ok" > $MOUNT2/widestripe
3324 $GETSTRIPE $MOUNT2/widestripe ||
3325 error "$GETSTRIPE $MOUNT2/widestripe failed"
3326 # fill acl buffer for avoid expand lsm to them
3327 awk -F : '{if (FNR < 25) { print "u:"$1":rwx" }}' /etc/passwd |
3329 setfacl -m $acl $MOUNT2/widestripe
3333 stat $MOUNT/widestripe || error "stat $MOUNT/widestripe failed"
3335 cleanup_46a $OSTCOUNT || error "cleanup_46a failed"
3337 run_test 46a "handle ost additional - wide striped file"
3342 check_mount || error "check_mount failed"
3343 $LCTL set_param ldlm.namespaces.$FSNAME-*-*-*.lru_size=100
3347 for ns in $($LCTL get_param ldlm.namespaces.$FSNAME-*-*-*.lru_size); do
3348 if echo $ns | grep "MDT[[:digit:]]*"; then
3351 lrs=$(echo $ns | sed 's/.*lru_size=//')
3352 lru_size[count]=$lrs
3357 facet_failover $SINGLEMDS
3358 client_up || error "client_up failed"
3361 for ns in $($LCTL get_param ldlm.namespaces.$FSNAME-*-*-*.lru_size); do
3362 if echo $ns | grep "MDT[[:digit:]]*"; then
3365 lrs=$(echo $ns | sed 's/.*lru_size=//')
3366 if ! test "$lrs" -eq "${lru_size[count]}"; then
3367 n=$(echo $ns | sed -e 's/ldlm.namespaces.//' -e 's/.lru_size=.*//')
3368 error "$n has lost lru_size: $lrs vs. ${lru_size[count]}"
3373 cleanup || error "cleanup failed with $?"
3375 run_test 47 "server restart does not make client loss lru_resize settings"
3380 # reformat after this test is needed - if the test fails,
3381 # we will have unkillable file at FS
3385 test_48() { # bug 17636
3388 check_mount || error "check_mount failed"
3390 $SETSTRIPE -c -1 $MOUNT || error "$SETSTRIPE -c -1 $MOUNT failed"
3391 $GETSTRIPE $MOUNT || error "$GETSTRIPE $MOUNT failed"
3393 echo "ok" > $MOUNT/widestripe
3394 $GETSTRIPE $MOUNT/widestripe ||
3395 error "$GETSTRIPE $MOUNT/widestripe failed"
3397 trap cleanup_48 EXIT ERR
3399 # fill acl buffer for avoid expand lsm to them
3400 getent passwd | awk -F : '{ print "u:"$1":rwx" }' | while read acl; do
3401 setfacl -m $acl $MOUNT/widestripe
3404 stat $MOUNT/widestripe || error "stat $MOUNT/widestripe failed"
3408 run_test 48 "too many acls on file"
3410 # check PARAM_SYS_LDLM_TIMEOUT option of MKFS.LUSTRE
3411 test_49a() { # bug 17710
3412 local timeout_orig=$TIMEOUT
3413 local ldlm_timeout_orig=$LDLM_TIMEOUT
3414 local LOCAL_TIMEOUT=20
3416 LDLM_TIMEOUT=$LOCAL_TIMEOUT
3417 TIMEOUT=$LOCAL_TIMEOUT
3421 check_mount || error "client mount failed"
3423 echo "check ldlm_timout..."
3424 local LDLM_MDS="$(do_facet $SINGLEMDS $LCTL get_param -n ldlm_timeout)"
3425 local LDLM_OST1="$(do_facet ost1 $LCTL get_param -n ldlm_timeout)"
3426 local LDLM_CLIENT="$(do_facet client $LCTL get_param -n ldlm_timeout)"
3428 if [ $LDLM_MDS -ne $LDLM_OST1 -o $LDLM_MDS -ne $LDLM_CLIENT ]; then
3429 error "Different LDLM_TIMEOUT:$LDLM_MDS $LDLM_OST1 $LDLM_CLIENT"
3432 if [ $LDLM_MDS -ne $((LOCAL_TIMEOUT / 3)) ]; then
3433 error "LDLM_TIMEOUT($LDLM_MDS) is not $((LOCAL_TIMEOUT / 3))"
3436 umount_client $MOUNT || error "umount_client $MOUNT failed"
3437 stop_ost || error "problem stopping OSS"
3438 stop_mds || error "problem stopping MDS"
3440 LDLM_TIMEOUT=$ldlm_timeout_orig
3441 TIMEOUT=$timeout_orig
3443 run_test 49a "check PARAM_SYS_LDLM_TIMEOUT option of mkfs.lustre"
3445 test_49b() { # bug 17710
3446 local timeout_orig=$TIMEOUT
3447 local ldlm_timeout_orig=$LDLM_TIMEOUT
3448 local LOCAL_TIMEOUT=20
3450 LDLM_TIMEOUT=$((LOCAL_TIMEOUT - 1))
3451 TIMEOUT=$LOCAL_TIMEOUT
3455 check_mount || error "client mount failed"
3457 local LDLM_MDS="$(do_facet $SINGLEMDS $LCTL get_param -n ldlm_timeout)"
3458 local LDLM_OST1="$(do_facet ost1 $LCTL get_param -n ldlm_timeout)"
3459 local LDLM_CLIENT="$(do_facet client $LCTL get_param -n ldlm_timeout)"
3461 if [ $LDLM_MDS -ne $LDLM_OST1 -o $LDLM_MDS -ne $LDLM_CLIENT ]; then
3462 error "Different LDLM_TIMEOUT:$LDLM_MDS $LDLM_OST1 $LDLM_CLIENT"
3465 if [ $LDLM_MDS -ne $((LOCAL_TIMEOUT - 1)) ]; then
3466 error "LDLM_TIMEOUT($LDLM_MDS) is not $((LOCAL_TIMEOUT - 1))"
3469 cleanup || error "cleanup failed"
3471 LDLM_TIMEOUT=$ldlm_timeout_orig
3472 TIMEOUT=$timeout_orig
3474 run_test 49b "check PARAM_SYS_LDLM_TIMEOUT option of mkfs.lustre"
3477 # Test both statfs and lfs df and fail if either one fails
3478 multiop_bg_pause $1 f_
3481 killall -USR1 multiop
3482 [ $RC1 -ne 0 ] && log "lazystatfs multiop failed"
3483 wait $PID || { RC1=$?; log "multiop return error "; }
3490 if [ $RC2 -eq 0 ]; then
3492 log "lazystatfs df failed"
3496 [[ $RC1 -ne 0 || $RC2 -eq 0 ]] && RC=1
3502 $LCTL set_param llite.$FSNAME-*.lazystatfs=1
3503 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
3505 lazystatfs $MOUNT || error "lazystatfs failed but no down servers"
3507 cleanup || error "cleanup failed with rc $?"
3509 run_test 50a "lazystatfs all servers available"
3513 $LCTL set_param llite.$FSNAME-*.lazystatfs=1
3514 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
3516 # Wait for client to detect down OST
3517 stop_ost || error "Unable to stop OST1"
3518 wait_osc_import_state mds ost DISCONN
3520 lazystatfs $MOUNT || error "lazystatfs should not return EIO"
3522 umount_client $MOUNT || error "Unable to unmount client"
3523 stop_mds || error "Unable to stop MDS"
3525 run_test 50b "lazystatfs all servers down"
3528 start_mds || error "Unable to start MDS"
3529 start_ost || error "Unable to start OST1"
3530 start_ost2 || error "Unable to start OST2"
3531 mount_client $MOUNT || error "Unable to mount client"
3532 $LCTL set_param llite.$FSNAME-*.lazystatfs=1
3533 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
3535 # Wait for client to detect down OST
3536 stop_ost || error "Unable to stop OST1"
3537 wait_osc_import_state mds ost DISCONN
3538 lazystatfs $MOUNT || error "lazystatfs failed with one down server"
3540 umount_client $MOUNT || error "Unable to unmount client"
3541 stop_ost2 || error "Unable to stop OST2"
3542 stop_mds || error "Unable to stop MDS"
3543 #writeconf to remove all ost2 traces for subsequent tests
3544 writeconf_or_reformat
3546 run_test 50c "lazystatfs one server down"
3549 start_mds || error "Unable to start MDS"
3550 start_ost || error "Unable to start OST1"
3551 start_ost2 || error "Unable to start OST2"
3552 mount_client $MOUNT || error "Unable to mount client"
3553 $LCTL set_param llite.$FSNAME-*.lazystatfs=1
3554 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
3556 # Issue the statfs during the window where the client still
3557 # belives the OST to be available but it is in fact down.
3558 # No failure just a statfs which hangs for a timeout interval.
3559 stop_ost || error "Unable to stop OST1"
3560 lazystatfs $MOUNT || error "lazystatfs failed with one down server"
3562 umount_client $MOUNT || error "Unable to unmount client"
3563 stop_ost2 || error "Unable to stop OST2"
3564 stop_mds || error "Unable to stop MDS"
3565 #writeconf to remove all ost2 traces for subsequent tests
3566 writeconf_or_reformat
3568 run_test 50d "lazystatfs client/server conn race"
3575 start_mds || error "Unable to start MDS"
3576 #first client should see only one ost
3577 start_ost || error "Unable to start OST1"
3578 wait_osc_import_state mds ost FULL
3580 # Wait for client to detect down OST
3581 stop_ost || error "Unable to stop OST1"
3582 wait_osc_import_state mds ost DISCONN
3584 mount_client $MOUNT || error "Unable to mount client"
3585 $LCTL set_param llite.$FSNAME-*.lazystatfs=0
3587 multiop_bg_pause $MOUNT _f
3591 if [ $RC1 -ne 0 ]; then
3592 log "multiop failed $RC1"
3595 sleep $(( $TIMEOUT+1 ))
3597 [ $? -ne 0 ] && error "process isn't sleep"
3598 start_ost || error "Unable to start OST1"
3599 wait $pid || error "statfs failed"
3602 umount_client $MOUNT || error "Unable to unmount client"
3603 stop_ost || error "Unable to stop OST1"
3604 stop_mds || error "Unable to stop MDS"
3606 run_test 50e "normal statfs all servers down"
3611 CONN_PROC="osc.$FSNAME-OST0001-osc-[M]*.ost_server_uuid"
3613 start_mds || error "Unable to start mds"
3614 #first client should see only one ost
3615 start_ost || error "Unable to start OST1"
3616 wait_osc_import_state mds ost FULL
3618 start_ost2 || error "Unable to start OST2"
3619 wait_osc_import_state mds ost2 FULL
3621 # Wait for client to detect down OST
3622 stop_ost2 || error "Unable to stop OST2"
3624 wait_osc_import_state mds ost2 DISCONN
3625 mount_client $MOUNT || error "Unable to mount client"
3626 $LCTL set_param llite.$FSNAME-*.lazystatfs=0
3628 multiop_bg_pause $MOUNT _f
3632 if [ $RC1 -ne 0 ]; then
3633 log "lazystatfs multiop failed $RC1"
3636 sleep $(( $TIMEOUT+1 ))
3638 [ $? -ne 0 ] && error "process isn't sleep"
3639 start_ost2 || error "Unable to start OST2"
3640 wait $pid || error "statfs failed"
3641 stop_ost2 || error "Unable to stop OST2"
3644 umount_client $MOUNT -f || error "Unable to unmount client"
3645 stop_ost || error "Unable to stop OST1"
3646 stop_mds || error "Unable to stop MDS"
3647 #writeconf to remove all ost2 traces for subsequent tests
3648 writeconf_or_reformat
3650 run_test 50f "normal statfs one server in down"
3653 [ "$OSTCOUNT" -lt "2" ] && skip_env "$OSTCOUNT < 2, skipping" && return
3655 start_ost2 || error "Unable to start OST2"
3656 wait_osc_import_state mds ost2 FULL
3657 wait_osc_import_state client ost2 FULL
3659 local PARAM="${FSNAME}-OST0001.osc.active"
3661 $SETSTRIPE -c -1 $DIR/$tfile || error "$SETSTRIPE failed"
3662 do_facet mgs $LCTL conf_param $PARAM=0 ||
3663 error "Unable to deactivate OST"
3665 umount_client $MOUNT || error "Unable to unmount client"
3666 mount_client $MOUNT || error "Unable to mount client"
3667 # This df should not cause a panic
3670 do_facet mgs $LCTL conf_param $PARAM=1 || error "Unable to activate OST"
3671 rm -f $DIR/$tfile || error "unable to remove file $DIR/$tfile"
3672 umount_client $MOUNT || error "Unable to unmount client"
3673 stop_ost2 || error "Unable to stop OST2"
3674 stop_ost || error "Unable to stop OST1"
3675 stop_mds || error "Unable to stop MDS"
3676 #writeconf to remove all ost2 traces for subsequent tests
3677 writeconf_or_reformat
3679 run_test 50g "deactivated OST should not cause panic"
3683 # prepare MDT/OST, make OSC inactive for OST1
3684 [ "$OSTCOUNT" -lt "2" ] && skip_env "$OSTCOUNT < 2, skipping" && return
3686 [ $(facet_fstype ost1) == zfs ] && import_zpool ost1
3687 do_facet ost1 "$TUNEFS --param osc.active=0 `ostdevname 1`" ||
3688 error "tunefs OST1 failed"
3689 start_mds || error "Unable to start MDT"
3690 start_ost || error "Unable to start OST1"
3691 start_ost2 || error "Unable to start OST2"
3692 mount_client $MOUNT || error "client start failed"
3694 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
3696 # activatate OSC for OST1
3697 local TEST="$LCTL get_param -n osc.${FSNAME}-OST0000-osc-[!M]*.active"
3698 set_conf_param_and_check client \
3699 "$TEST" "${FSNAME}-OST0000.osc.active" 1 ||
3700 error "Unable to activate OST1"
3702 mkdir $DIR/$tdir/2 || error "mkdir $DIR/$tdir/2 failed"
3703 $SETSTRIPE -c -1 -i 0 $DIR/$tdir/2 ||
3704 error "$SETSTRIPE $DIR/$tdir/2 failed"
3705 sleep 1 && echo "create a file after OST1 is activated"
3707 createmany -o $DIR/$tdir/2/$tfile-%d 1
3709 # check OSC import is working
3710 stat $DIR/$tdir/2/* >/dev/null 2>&1 ||
3711 error "some OSC imports are still not connected"
3714 umount_client $MOUNT || error "Unable to umount client"
3715 stop_ost2 || error "Unable to stop OST2"
3716 cleanup_nocli || error "cleanup_nocli failed with $?"
3718 run_test 50h "LU-642: activate deactivated OST"
3721 # prepare MDT/OST, make OSC inactive for OST1
3722 [ "$MDSCOUNT" -lt "2" ] && skip_env "$MDSCOUNT < 2, skipping" && return
3725 [ $(facet_fstype mds2) == zfs ] && import_zpool mds2
3726 do_facet mds2 "$TUNEFS --param mdc.active=0 $(mdsdevname 2)" ||
3727 error "tunefs MDT2 failed"
3728 start_mds || error "Unable to start MDT"
3729 start_ost || error "Unable to start OST1"
3730 start_ost2 || error "Unable to start OST2"
3731 mount_client $MOUNT || error "client start failed"
3733 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
3735 $LCTL conf_param ${FSNAME}-MDT0000.mdc.active=0 &&
3736 error "deactive MDC0 succeeds"
3737 # activate MDC for MDT2
3738 local TEST="$LCTL get_param -n mdc.${FSNAME}-MDT0001-mdc-[!M]*.active"
3739 set_conf_param_and_check client \
3740 "$TEST" "${FSNAME}-MDT0001.mdc.active" 1 ||
3741 error "Unable to activate MDT2"
3743 wait_clients_import_state ${CLIENTS:-$HOSTNAME} mds2 FULL
3744 if [ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.7.60) ]
3746 wait_dne_interconnect
3748 $LFS mkdir -i1 $DIR/$tdir/2 || error "mkdir $DIR/$tdir/2 failed"
3750 createmany -o $DIR/$tdir/2/$tfile-%d 1 || error "create files failed"
3752 rm -rf $DIR/$tdir/2 || error "unlink dir failed"
3754 # deactivate MDC for MDT2
3755 local TEST="$LCTL get_param -n mdc.${FSNAME}-MDT0001-mdc-[!M]*.active"
3756 set_conf_param_and_check client \
3757 "$TEST" "${FSNAME}-MDT0001.mdc.active" 0 ||
3758 error "Unable to deactivate MDT2"
3760 wait_osp_active mds ${FSNAME}-MDT0001 1 0
3762 $LFS mkdir -i1 $DIR/$tdir/2 &&
3763 error "mkdir $DIR/$tdir/2 succeeds after deactive MDT"
3765 $LFS mkdir -i0 -c$MDSCOUNT $DIR/$tdir/striped_dir ||
3766 error "mkdir $DIR/$tdir/striped_dir fails after deactive MDT2"
3768 local stripe_count=$($LFS getdirstripe -c $DIR/$tdir/striped_dir)
3769 [ $stripe_count -eq $((MDSCOUNT - 1)) ] ||
3770 error "wrong $stripe_count != $((MDSCOUNT -1)) for striped_dir"
3773 umount_client $MOUNT || error "Unable to umount client"
3778 run_test 50i "activate deactivated MDT"
3781 local LOCAL_TIMEOUT=20
3785 check_mount || error "check_mount failed"
3787 mkdir $MOUNT/$tdir || error "mkdir $MOUNT/$tdir failed"
3788 $SETSTRIPE -c -1 $MOUNT/$tdir ||
3789 error "$SETSTRIPE -c -1 $MOUNT/$tdir failed"
3790 #define OBD_FAIL_MDS_REINT_DELAY 0x142
3791 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0x142"
3792 touch $MOUNT/$tdir/$tfile &
3795 start_ost2 || error "Unable to start OST1"
3797 stop_ost2 || error "Unable to stop OST1"
3798 umount_client $MOUNT -f || error “unmount $MOUNT failed”
3799 cleanup_nocli || error “stop server failed”
3800 #writeconf to remove all ost2 traces for subsequent tests
3801 writeconf_or_reformat
3803 run_test 51 "Verify that mdt_reint handles RMF_MDT_MD correctly when an OST is added"
3812 do_node $node mkdir -p $dest
3813 [ $? -eq 0 ] || { error "Unable to create directory"; return 1; }
3815 do_node $node 'tar cf - '$@' | tar xf - -C '$dest';
3816 [ \"\${PIPESTATUS[*]}\" = \"0 0\" ] || exit 1'
3817 [ $? -eq 0 ] || { error "Unable to tar files"; return 2; }
3819 do_node $node 'getfattr -d -m "[a-z]*\\." '$@' > '$xattrs
3820 [ $? -eq 0 ] || { error "Unable to read xattrs"; return 3; }
3830 local backup2=${TMP}/backup2
3832 do_node $node mkdir -p $backup2
3833 [ $? -eq 0 ] || { error "Unable to create directory"; return 1; }
3835 do_node $node 'tar cf - '$@' | tar xf - -C '$backup2';
3836 [ \"\${PIPESTATUS[*]}\" = \"0 0\" ] || exit 1'
3837 [ $? -eq 0 ] || { error "Unable to tar files to diff"; return 2; }
3839 do_node $node "diff -rq $backup $backup2"
3840 [ $? -eq 0 ] || { error "contents differ"; return 3; }
3842 local xattrs2=${TMP}/xattrs2
3843 do_node $node 'getfattr -d -m "[a-z]*\\." '$@' > '$xattrs2
3844 [ $? -eq 0 ] || { error "Unable to read xattrs to diff"; return 4; }
3846 do_node $node "diff $xattrs $xattrs2"
3847 [ $? -eq 0 ] || { error "xattrs differ"; return 5; }
3849 do_node $node "rm -rf $backup2 $xattrs2"
3850 [ $? -eq 0 ] || { error "Unable to delete temporary files"; return 6; }
3854 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
3855 skip "Only applicable to ldiskfs-based MDTs"
3859 start_mds || error "Unable to start MDS"
3860 start_ost || error "Unable to start OST1"
3861 mount_client $MOUNT || error "Unable to mount client"
3864 local ost1mnt=$(facet_mntpt ost1)
3865 local ost1node=$(facet_active_host ost1)
3866 local ost1tmp=$TMP/conf52
3869 mkdir $DIR/$tdir || error "Unable to create $DIR/$tdir"
3870 touch $TMP/modified_first || error "Unable to create temporary file"
3871 local mtime=$(stat -c %Y $TMP/modified_first)
3872 do_node $ost1node "mkdir -p $ost1tmp &&
3873 touch -m -d @$mtime $ost1tmp/modified_first" ||
3874 error "Unable to create temporary file"
3877 $SETSTRIPE -c -1 -S 1M $DIR/$tdir || error "$SETSTRIPE failed"
3879 for (( i=0; i < nrfiles; i++ )); do
3880 multiop $DIR/$tdir/$tfile-$i Ow1048576w1048576w524288c ||
3881 error "multiop failed"
3887 echo backup files to $TMP/$tdir
3888 local files=$(find $DIR/$tdir -type f -newer $TMP/modified_first)
3889 copy_files_xattrs $(hostname) $TMP/$tdir $TMP/file_xattrs $files ||
3890 error "Unable to copy files"
3892 umount_client $MOUNT || error "Unable to umount client"
3893 stop_ost || error "Unable to stop ost1"
3895 echo mount ost1 as ldiskfs
3896 do_node $ost1node mkdir -p $ost1mnt || error "Unable to create $ost1mnt"
3897 if ! do_node $ost1node test -b $ost1_dev; then
3900 do_node $ost1node mount -t $(facet_fstype ost1) $loop $ost1_dev \
3902 error "Unable to mount ost1 as ldiskfs"
3905 echo backup objects to $ost1tmp/objects
3906 local objects=$(do_node $ost1node 'find '$ost1mnt'/O/[0-9]* -type f'\
3907 '-size +0 -newer '$ost1tmp'/modified_first -regex ".*\/[0-9]+"')
3908 copy_files_xattrs $ost1node $ost1tmp/objects $ost1tmp/object_xattrs \
3910 error "Unable to copy objects"
3912 # move objects to lost+found
3913 do_node $ost1node 'mv '$objects' '${ost1mnt}'/lost+found'
3914 [ $? -eq 0 ] || { error "Unable to move objects"; return 14; }
3916 do_node $ost1node "umount $ost1mnt" ||
3917 error "Unable to umount ost1 as ldiskfs"
3919 start_ost || error "Unable to start OST1"
3920 mount_client $MOUNT || error "Unable to mount client"
3922 local REPAIRED=$(do_node $ost1node "$LCTL get_param \
3923 -n osd-ldiskfs.$FSNAME-OST0000.oi_scrub" |
3924 awk '/^lf_repa[ri]*ed/ { print $2 }')
3925 [ $REPAIRED -gt 0 ] ||
3926 error "Some entry under /lost+found should be repaired"
3929 diff_files_xattrs $(hostname) $TMP/$tdir $TMP/file_xattrs $files ||
3930 error "Unable to diff files"
3932 rm -rf $TMP/$tdir $TMP/file_xattrs ||
3933 error "Unable to delete temporary files"
3934 do_node $ost1node "rm -rf $ost1tmp" ||
3935 error "Unable to delete temporary files"
3936 cleanup || error "cleanup failed with $?"
3938 run_test 52 "check recovering objects from lost+found"
3940 # Checks threads_min/max/started for some service
3942 # Arguments: service name (OST or MDT), facet (e.g., ost1, $SINGLEMDS), and a
3943 # parameter pattern prefix like 'ost.*.ost'.
3956 local msg="Insane $modname thread counts"
3957 local ncpts=$(check_cpt_number $facet)
3961 check_mount || return 41
3963 # We need to expand $parampat, but it may match multiple parameters, so
3964 # we'll pick the first one
3965 if ! paramp=$(do_facet $facet "lctl get_param -N ${parampat}.threads_min"|head -1); then
3966 error "Couldn't expand ${parampat}.threads_min parameter name"
3970 # Remove the .threads_min part
3971 paramp=${paramp%.threads_min}
3973 # Check for sanity in defaults
3974 tmin=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min" ||
3976 tmax=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_max" ||
3978 tstarted=$(do_facet $facet "$LCTL get_param \
3979 -n ${paramp}.threads_started" || echo 0)
3980 lassert 23 "$msg (PDSH problems?)" '(($tstarted && $tmin && $tmax))' ||
3982 lassert 24 "$msg" '(($tstarted >= $tmin && $tstarted <= $tmax ))' ||
3984 nthrs=$(expr $tmax - $tmin)
3985 if [ $nthrs -lt $ncpts ]; then
3991 [ $tmin -eq $tmax -a $tmin -eq $tstarted ] &&
3992 skip_env "module parameter forced $facet thread count" &&
3993 tmin=3 && tmax=$((3 * tmax))
3995 # Check that we can change min/max
3996 do_facet $facet "$LCTL set_param \
3997 ${paramp}.threads_min=$((tmin + nthrs))"
3998 do_facet $facet "$LCTL set_param \
3999 ${paramp}.threads_max=$((tmax - nthrs))"
4000 tmin2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min" ||
4002 tmax2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_max" ||
4004 lassert 25 "$msg" '(($tmin2 == ($tmin + $nthrs) &&
4005 $tmax2 == ($tmax - $nthrs)))' || return $?
4007 # Check that we can set min/max to the same value
4008 tmin=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min" ||
4010 do_facet $facet "$LCTL set_param ${paramp}.threads_max=$tmin"
4011 tmin2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min" ||
4013 tmax2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_max" ||
4015 lassert 26 "$msg" '(($tmin2 == $tmin && $tmax2 == $tmin))' || return $?
4017 # Check that we can't set max < min
4018 do_facet $facet "$LCTL set_param ${paramp}.threads_max=$((tmin - 1))"
4019 tmin2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min" ||
4021 tmax2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_max" ||
4023 lassert 27 "$msg" '(($tmin2 <= $tmax2))' || return $?
4025 # We need to ensure that we get the module options desired; to do this
4026 # we set LOAD_MODULES_REMOTE=true and we call setmodopts below.
4027 LOAD_MODULES_REMOTE=true
4030 local newvalue="${opts}=$(expr $basethr \* $ncpts)"
4031 setmodopts -a $modname "$newvalue" oldvalue
4034 check_mount || return 41
4036 # Restore previous setting of MODOPTS_*
4037 setmodopts $modname "$oldvalue"
4039 # Check that $opts took
4040 tmin=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min")
4041 tmax=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_max")
4042 tstarted=$(do_facet $facet \
4043 "$LCTL get_param -n ${paramp}.threads_started")
4044 lassert 28 "$msg" '(($tstarted >= $tmin && $tstarted <= $tmax ))' ||
4053 thread_sanity OST ost1 'ost.*.ost' 'oss_num_threads' '16'
4054 cleanup || error "cleanup failed with rc $?"
4056 run_test 53a "check OSS thread count params"
4060 local mds=$(do_facet $SINGLEMDS "$LCTL get_param \
4061 -N mds.*.*.threads_max 2>/dev/null")
4062 if [ -z "$mds" ]; then
4063 #running this on an old MDT
4064 thread_sanity MDT $SINGLEMDS 'mdt.*.*.' 'mdt_num_threads' 16
4066 thread_sanity MDT $SINGLEMDS 'mds.*.*.' 'mds_num_threads' 16
4068 cleanup || error "cleanup failed with $?"
4070 run_test 53b "check MDS thread count params"
4073 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
4074 skip "Only applicable to ldiskfs-based MDTs"
4078 do_rpc_nodes $(facet_host ost1) run_llverdev $(ostdevname 1) -p ||
4079 error "llverdev failed with rc=$?"
4082 run_test 54a "test llverdev and partial verify of device"
4085 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
4086 skip "Only applicable to ldiskfs-based MDTs"
4091 run_llverfs $MOUNT -p || error "llverfs failed with rc=$?"
4092 cleanup || error "cleanup failed with rc=$?"
4094 run_test 54b "test llverfs and partial verify of filesystem"
4098 local max_ost_index=$1
4099 echo -n $(((max_ost_index + 1) * 8))
4103 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
4104 skip "Only applicable to ldiskfs-based MDTs"
4108 local mdsdev=$(mdsdevname 1)
4109 local mdsvdev=$(mdsvdevname 1)
4113 if ! combined_mgs_mds; then
4114 stop_mgs || error "stopping MGS service failed"
4115 format_mgs || error "formatting MGT failed"
4117 add mds1 $(mkfs_opts mds1 ${mdsdev}) --reformat $mdsdev \
4119 add ost1 $(mkfs_opts ost1 $(ostdevname 1)) --index=$i \
4120 --reformat $(ostdevname 1) $(ostvdevname 1)
4126 echo checking size of lov_objid for ost index $i
4127 LOV_OBJID_SIZE=$(do_facet mds1 "$DEBUGFS -R 'stat lov_objid' $mdsdev 2>/dev/null" |
4128 grep ^User | awk -F 'Size: ' '{print $2}')
4129 if [ "$LOV_OBJID_SIZE" != $(lov_objid_size $i) ]; then
4130 error "lov_objid size has to be $(lov_objid_size $i), not $LOV_OBJID_SIZE"
4132 echo ok, lov_objid size is correct: $LOV_OBJID_SIZE
4139 run_test 55 "check lov_objid size"
4142 local server_version=$(lustre_version_code $SINGLEMDS)
4143 local mds_journal_size_orig=$MDSJOURNALSIZE
4148 for num in $(seq 1 $MDSCOUNT); do
4151 add ost1 $(mkfs_opts ost1 $(ostdevname 1)) --index=10000 --reformat \
4152 $(ostdevname 1) $(ostvdevname 1)
4153 add ost2 $(mkfs_opts ost2 $(ostdevname 2)) --index=1000 --reformat \
4154 $(ostdevname 2) $(ostvdevname 2)
4157 start_ost || error "Unable to start first ost (idx 10000)"
4158 start_ost2 || error "Unable to start second ost (idx 1000)"
4159 mount_client $MOUNT || error "Unable to mount client"
4163 if [[ $server_version -ge $(version_code 2.6.54) ]] ||
4164 [[ $server_version -ge $(version_code 2.5.4) &&
4165 $server_version -lt $(version_code 2.5.11) ]]; then
4166 wait_osc_import_state mds ost1 FULL
4167 wait_osc_import_state mds ost2 FULL
4168 $SETSTRIPE --stripe-count=-1 $DIR/$tfile ||
4169 error "Unable to setstripe $DIR/$tfile"
4170 n=$($LFS getstripe --stripe-count $DIR/$tfile)
4171 [ "$n" -eq 2 ] || error "Stripe count not two: $n"
4176 MDSJOURNALSIZE=$mds_journal_size_orig
4179 run_test 56a "check big OST indexes and out-of-index-order start"
4184 umount_client $MOUNT -f || error "unmount client failed"
4193 [ $MDSCOUNT -lt 3 ] && skip "needs >= 3 MDTs" && return
4195 trap cleanup_56b EXIT RETURN ERR
4198 if ! combined_mgs_mds ; then
4203 add mds1 $(mkfs_opts mds1 $(mdsdevname 1)) --index=0 --reformat \
4204 $(mdsdevname 1) $(mdsvdevname 1)
4205 add mds2 $(mkfs_opts mds2 $(mdsdevname 2)) --index=1 --reformat \
4206 $(mdsdevname 2) $(mdsvdevname 2)
4207 add mds3 $(mkfs_opts mds3 $(mdsdevname 3)) --index=1000 --reformat \
4208 $(mdsdevname 3) $(mdsvdevname 3)
4212 start_mdt 1 || error "MDT 1 (idx 0) start failed"
4213 start_mdt 2 || error "MDT 2 (idx 1) start failed"
4214 start_mdt 3 || error "MDT 3 (idx 1000) start failed"
4215 start_ost || error "Unable to start first ost"
4216 start_ost2 || error "Unable to start second ost"
4218 do_nodes $(comma_list $(mdts_nodes)) \
4219 "$LCTL set_param mdt.*.enable_remote_dir=1 \
4220 mdt.*.enable_remote_dir_gid=-1"
4222 mount_client $MOUNT || error "Unable to mount client"
4224 $LFS mkdir -c3 $MOUNT/$tdir || error "failed to make testdir"
4226 echo "This is test file 1!" > $MOUNT/$tdir/$tfile.1 ||
4227 error "failed to make test file 1"
4228 echo "This is test file 2!" > $MOUNT/$tdir/$tfile.2 ||
4229 error "failed to make test file 2"
4230 echo "This is test file 1000!" > $MOUNT/$tdir/$tfile.1000 ||
4231 error "failed to make test file 1000"
4233 rm -rf $MOUNT/$tdir || error "failed to remove testdir"
4235 $LFS mkdir -i1000 $MOUNT/$tdir.1000 ||
4236 error "create remote dir at idx 1000 failed"
4239 echo "=== START lfs df OUTPUT ==="
4241 echo "==== END lfs df OUTPUT ===="
4243 mdtcnt=$(echo -e "$output" | grep $FSNAME-MDT | wc -l)
4244 ostcnt=$(echo -e "$output" | grep $FSNAME-OST | wc -l)
4246 echo "lfs df returned mdt count $mdtcnt and ost count $ostcnt"
4247 [ $mdtcnt -eq 3 ] || error "lfs df returned wrong mdt count"
4248 [ $ostcnt -eq 2 ] || error "lfs df returned wrong ost count"
4250 echo "This is test file 1!" > $MOUNT/$tdir.1000/$tfile.1 ||
4251 error "failed to make test file 1"
4252 echo "This is test file 2!" > $MOUNT/$tdir.1000/$tfile.2 ||
4253 error "failed to make test file 2"
4254 echo "This is test file 1000!" > $MOUNT/$tdir.1000/$tfile.1000 ||
4255 error "failed to make test file 1000"
4256 rm -rf $MOUNT/$tdir.1000 || error "failed to remove remote_dir"
4259 echo "=== START lfs mdts OUTPUT ==="
4261 echo "==== END lfs mdts OUTPUT ===="
4263 echo -e "$output" | grep -v "MDTS:" | awk '{print $1}' |
4264 sed 's/://g' > $TMP/mdts-actual.txt
4265 sort $TMP/mdts-actual.txt -o $TMP/mdts-actual.txt
4267 echo -e "0\n1\n1000" > $TMP/mdts-expected.txt
4269 diff $TMP/mdts-expected.txt $TMP/mdts-actual.txt
4272 rm $TMP/mdts-expected.txt $TMP/mdts-actual.txt
4274 [ $result -eq 0 ] || error "target_obd proc file is incorrect!"
4276 run_test 56b "test target_obd correctness with nonconsecutive MDTs"
4278 test_57a() { # bug 22656
4279 do_rpc_nodes $(facet_active_host ost1) load_modules_local
4280 local NID=$(do_facet ost1 "$LCTL get_param nis" |
4281 tail -1 | awk '{print $1}')
4282 writeconf_or_reformat
4283 [ $(facet_fstype ost1) == zfs ] && import_zpool ost1
4284 do_facet ost1 "$TUNEFS --failnode=$NID `ostdevname 1`" ||
4285 error "tunefs failed"
4287 start_ost && error "OST registration from failnode should fail"
4290 run_test 57a "initial registration from failnode should fail (should return errs)"
4293 do_rpc_nodes $(facet_active_host ost1) load_modules_local
4294 local NID=$(do_facet ost1 "$LCTL get_param nis" |
4295 tail -1 | awk '{print $1}')
4296 writeconf_or_reformat
4297 [ $(facet_fstype ost1) == zfs ] && import_zpool ost1
4298 do_facet ost1 "$TUNEFS --servicenode=$NID `ostdevname 1`" ||
4299 error "tunefs failed"
4301 start_ost || error "OST registration from servicenode should not fail"
4304 run_test 57b "initial registration from servicenode should not fail"
4307 do_facet mgs $LCTL get_param mgs.MGS.live.$FSNAME | grep OST | wc -l
4310 test_58() { # bug 22658
4311 combined_mgs_mds || stop_mgs || error "stopping MGS service failed"
4313 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
4314 createmany -o $DIR/$tdir/$tfile-%d 100
4315 # make sure that OSTs do not cancel llog cookies before we unmount the MDS
4316 #define OBD_FAIL_OBD_LOG_CANCEL_NET 0x601
4317 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0x601"
4318 unlinkmany $DIR/$tdir/$tfile-%d 100
4319 stop_mds || error "Unable to stop MDS"
4321 local MNTDIR=$(facet_mntpt $SINGLEMDS)
4322 local devname=$(mdsdevname ${SINGLEMDS//mds/})
4324 # remove all files from the OBJECTS dir
4325 mount_fstype $SINGLEMDS
4327 do_facet $SINGLEMDS "find $MNTDIR/O/1/d* -type f -delete"
4329 unmount_fstype $SINGLEMDS
4330 # restart MDS with missing llog files
4331 start_mds || error "unable to start MDS"
4332 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0"
4335 run_test 58 "missing llog files must not prevent MDT from mounting"
4338 start_mgsmds >> /dev/null
4339 local C1=$(count_osts)
4340 if [ $C1 -eq 0 ]; then
4341 start_ost >> /dev/null
4345 echo "original ost count: $C1 (expect > 0)"
4346 [ $C1 -gt 0 ] || error "No OSTs in $FSNAME log"
4347 start_mgsmds -o writeconf >> /dev/null || error "MDT start failed"
4348 local C2=$(count_osts)
4349 echo "after mdt writeconf count: $C2 (expect 0)"
4350 [ $C2 -gt 0 ] && error "MDT writeconf should erase OST logs"
4351 echo "OST start without writeconf should fail:"
4352 start_ost >> /dev/null &&
4353 error "OST start without writeconf didn't fail"
4354 echo "OST start with writeconf should succeed:"
4355 start_ost -o writeconf >> /dev/null || error "OST1 start failed"
4356 local C3=$(count_osts)
4357 echo "after ost writeconf count: $C3 (expect 1)"
4358 [ $C3 -eq 1 ] || error "new OST writeconf should add:"
4359 start_ost2 -o writeconf >> /dev/null || error "OST2 start failed"
4360 local C4=$(count_osts)
4361 echo "after ost2 writeconf count: $C4 (expect 2)"
4362 [ $C4 -eq 2 ] || error "OST2 writeconf should add log"
4363 stop_ost2 >> /dev/null
4364 cleanup_nocli >> /dev/null
4365 #writeconf to remove all ost2 traces for subsequent tests
4366 writeconf_or_reformat
4368 run_test 59 "writeconf mount option"
4370 test_60() { # LU-471
4373 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
4374 skip "Only applicable to ldiskfs-based MDTs"
4378 for num in $(seq $MDSCOUNT); do
4379 add mds${num} $(mkfs_opts mds${num} $(mdsdevname $num)) \
4380 --mkfsoptions='\" -E stride=64 -O ^uninit_bg\"' \
4381 --reformat $(mdsdevname $num) $(mdsvdevname $num) ||
4385 dump=$(do_facet $SINGLEMDS dumpe2fs $(mdsdevname 1))
4386 [ ${PIPESTATUS[0]} -eq 0 ] || error "dumpe2fs $(mdsdevname 1) failed"
4388 # MDT default has dirdata feature
4389 echo $dump | grep dirdata > /dev/null || error "dirdata is not set"
4390 # we disable uninit_bg feature
4391 echo $dump | grep uninit_bg > /dev/null && error "uninit_bg is set"
4392 # we set stride extended options
4393 echo $dump | grep stride > /dev/null || error "stride is not set"
4396 run_test 60 "check mkfs.lustre --mkfsoptions -E -O options setting"
4401 [ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.1.53) ] ||
4402 { skip "Need MDS version at least 2.1.53"; return 0; }
4404 if [ $(facet_fstype $SINGLEMDS) == ldiskfs ] &&
4405 ! large_xattr_enabled; then
4408 for num in $(seq $MDSCOUNT); do
4409 do_facet mds${num} $TUNE2FS -O large_xattr \
4410 $(mdsdevname $num) ||
4411 error "tune2fs on mds $num failed"
4415 combined_mgs_mds || stop_mgs || error "stopping MGS service failed"
4416 setup_noconfig || error "setting up the filesystem failed"
4417 client_up || error "starting client failed"
4419 local file=$DIR/$tfile
4420 touch $file || error "touch $file failed"
4422 local large_value="$(generate_string $(max_xattr_size))"
4423 local small_value="bar"
4425 local name="trusted.big"
4426 log "save large xattr $name on $file"
4427 setfattr -n $name -v $large_value $file ||
4428 error "saving $name on $file failed"
4430 local new_value=$(get_xattr_value $name $file)
4431 [[ "$new_value" != "$large_value" ]] &&
4432 error "$name different after saving"
4434 log "shrink value of $name on $file"
4435 setfattr -n $name -v $small_value $file ||
4436 error "shrinking value of $name on $file failed"
4438 new_value=$(get_xattr_value $name $file)
4439 [[ "$new_value" != "$small_value" ]] &&
4440 error "$name different after shrinking"
4442 log "grow value of $name on $file"
4443 setfattr -n $name -v $large_value $file ||
4444 error "growing value of $name on $file failed"
4446 new_value=$(get_xattr_value $name $file)
4447 [[ "$new_value" != "$large_value" ]] &&
4448 error "$name different after growing"
4450 log "check value of $name on $file after remounting MDS"
4452 new_value=$(get_xattr_value $name $file)
4453 [[ "$new_value" != "$large_value" ]] &&
4454 error "$name different after remounting MDS"
4456 log "remove large xattr $name from $file"
4457 setfattr -x $name $file || error "removing $name from $file failed"
4460 stopall || error "stopping for e2fsck run"
4461 for num in $(seq $MDSCOUNT); do
4462 run_e2fsck $(facet_active_host mds$num) \
4463 $(mdsdevname $num) "-y" ||
4464 error "e2fsck MDT$num failed"
4466 setup_noconfig || error "remounting the filesystem failed"
4469 # need to delete this file to avoid problems in other tests
4471 stopall || error "stopping systems to turn off large_xattr"
4473 for num in $(seq $MDSCOUNT); do
4474 do_facet mds${num} $TUNE2FS -O ^large_xattr \
4475 $(mdsdevname $num) ||
4476 error "tune2fs on mds $num failed"
4480 run_test 61 "large xattr"
4483 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
4484 skip "Only applicable to ldiskfs-based MDTs"
4489 local mdsdev=$(mdsdevname 1)
4490 local ostdev=$(ostdevname 1)
4492 [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.2.51) ]] ||
4493 { skip "Need MDS version at least 2.2.51"; return 0; }
4495 echo "disable journal for mds"
4496 do_facet mds1 $TUNE2FS -O ^has_journal $mdsdev || error "tune2fs failed"
4497 start_mds && error "MDT start should fail"
4498 echo "disable journal for ost"
4499 do_facet ost1 $TUNE2FS -O ^has_journal $ostdev || error "tune2fs failed"
4500 start_ost && error "OST start should fail"
4501 cleanup || error "cleanup failed with rc $?"
4504 run_test 62 "start with disabled journal"
4507 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
4508 skip "Only applicable to ldiskfs-based MDTs"
4512 do_rpc_nodes $(facet_active_host $SINGLEMDS) load_module ldiskfs
4513 local inode_slab=$(do_facet $SINGLEMDS "cat /proc/slabinfo" |
4514 awk '/ldiskfs_inode_cache/ { print $5 / $6 }')
4515 if [ -z "$inode_slab" ]; then
4516 skip "ldiskfs module has not been loaded"
4520 echo "$inode_slab ldiskfs inodes per page"
4521 [ "${inode_slab%.*}" -ge "3" ] && return 0
4523 # If kmalloc-128 is also 1 per page - this is a debug kernel
4524 # and so this is not an error.
4525 local kmalloc128=$(do_facet $SINGLEMDS "cat /proc/slabinfo" |
4526 awk '/^(kmalloc|size)-128 / { print $5 / $6 }')
4527 # 32 128-byte chunks in 4k
4528 [ "${kmalloc128%.*}" -lt "32" ] ||
4529 error "ldiskfs inode too big, only $inode_slab objs/page, " \
4530 "kmalloc128 = $kmalloc128 objs/page"
4532 run_test 63 "Verify each page can at least hold 3 ldiskfs inodes"
4535 start_mds || error "unable to start MDS"
4536 start_ost || error "Unable to start OST1"
4537 start_ost2 || error "Unable to start second ost"
4538 mount_client $MOUNT || error "Unable to mount client"
4539 stop_ost2 || error "Unable to stop second ost"
4542 umount_client $MOUNT -f || error “unmount $MOUNT failed”
4543 cleanup_nocli || error "cleanup_nocli failed with $?"
4544 #writeconf to remove all ost2 traces for subsequent tests
4545 writeconf_or_reformat
4547 run_test 64 "check lfs df --lazy "
4549 test_65() { # LU-2237
4550 # Currently, the test is only valid for ldiskfs backend
4551 [ "$(facet_fstype $SINGLEMDS)" != "ldiskfs" ] &&
4552 skip "non-ldiskfs backend" && return
4554 local devname=$(mdsdevname ${SINGLEMDS//mds/})
4555 local brpt=$(facet_mntpt brpt)
4558 if ! do_facet $SINGLEMDS "test -b $devname"; then
4562 stop_mds || error "Unable to stop MDS"
4563 local obj=$(do_facet $SINGLEMDS \
4564 "$DEBUGFS -c -R \\\"stat last_rcvd\\\" $devname" |
4566 if [ -z "$obj" ]; then
4567 # The MDT may be just re-formatted, mount the MDT for the
4568 # first time to guarantee the "last_rcvd" file is there.
4569 start_mds || error "fail to mount the MDS for the first time"
4570 stop_mds || error "Unable to stop MDS"
4573 # remove the "last_rcvd" file
4574 do_facet $SINGLEMDS "mkdir -p $brpt"
4575 do_facet $SINGLEMDS \
4576 "mount -t $(facet_fstype $SINGLEMDS) $opts $devname $brpt"
4577 do_facet $SINGLEMDS "rm -f ${brpt}/last_rcvd"
4578 do_facet $SINGLEMDS "$UMOUNT $brpt"
4580 # restart MDS, the "last_rcvd" file should be recreated.
4581 start_mds || error "fail to restart the MDS"
4582 stop_mds || error "Unable to stop MDS"
4583 obj=$(do_facet $SINGLEMDS \
4584 "$DEBUGFS -c -R \\\"stat last_rcvd\\\" $devname" | grep Inode)
4585 [ -n "$obj" ] || error "fail to re-create the last_rcvd"
4587 run_test 65 "re-create the lost last_rcvd file when server mount"
4590 [[ $(lustre_version_code mgs) -ge $(version_code 2.3.59) ]] ||
4591 { skip "Need MGS version at least 2.3.59"; return 0; }
4594 local OST1_NID=$(do_facet ost1 $LCTL list_nids | head -1)
4595 local MDS_NID=$(do_facet $SINGLEMDS $LCTL list_nids | head -1)
4597 echo "replace_nids should fail if MDS, OSTs and clients are UP"
4598 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 $OST1_NID &&
4599 error "replace_nids fail"
4601 umount_client $MOUNT || error "unmounting client failed"
4602 echo "replace_nids should fail if MDS and OSTs are UP"
4603 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 $OST1_NID &&
4604 error "replace_nids fail"
4606 stop_ost || error "Unable to stop OST1"
4607 echo "replace_nids should fail if MDS is UP"
4608 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 $OST1_NID &&
4609 error "replace_nids fail"
4611 stop_mds || error "stopping mds failed"
4613 if combined_mgs_mds; then
4614 start_mdt 1 "-o nosvc" ||
4615 error "starting mds with nosvc option failed"
4618 echo "command should accept two parameters"
4619 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 &&
4620 error "command should accept two params"
4622 echo "correct device name should be passed"
4623 do_facet mgs $LCTL replace_nids $FSNAME-WRONG0000 $OST1_NID &&
4624 error "wrong devname"
4626 echo "wrong nids list should not destroy the system"
4627 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 "wrong nids list" &&
4630 echo "replace OST nid"
4631 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 $OST1_NID ||
4632 error "replace nids failed"
4634 echo "command should accept two parameters"
4635 do_facet mgs $LCTL replace_nids $FSNAME-MDT0000 &&
4636 error "command should accept two params"
4638 echo "wrong nids list should not destroy the system"
4639 do_facet mgs $LCTL replace_nids $FSNAME-MDT0000 "wrong nids list" &&
4642 echo "replace MDS nid"
4643 do_facet mgs $LCTL replace_nids $FSNAME-MDT0000 $MDS_NID ||
4644 error "replace nids failed"
4646 if ! combined_mgs_mds ; then
4649 stop_mds || error "Unable to stop MDS"
4653 check_mount || error "error after nid replace"
4654 cleanup || error "cleanup failed"
4657 run_test 66 "replace nids"
4659 test_67() { #LU-2950
4660 local legacy="$TMP/legacy_lnet_config"
4661 local new="$TMP/new_routes_test"
4662 local out="$TMP/config_out_file"
4663 local verify="$TMP/conv_verify"
4664 local verify_conf="$TMP/conf_verify"
4666 # Create the legacy file that will be run through the
4667 # lustre_routes_conversion script
4668 cat <<- LEGACY_LNET_CONFIG > $legacy
4669 tcp1 23 192.168.213.1@tcp:1; tcp5 34 193.30.4.3@tcp:4;
4670 tcp2 54 10.1.3.2@tcp;
4671 tcp3 10.3.4.3@tcp:3;
4675 # Create the verification file to verify the output of
4676 # lustre_routes_conversion script against.
4677 cat <<- VERIFY_LNET_CONFIG > $verify
4678 tcp1: { gateway: 192.168.213.1@tcp, hop: 23, priority: 1 }
4679 tcp5: { gateway: 193.30.4.3@tcp, hop: 34, priority: 4 }
4680 tcp2: { gateway: 10.1.3.2@tcp, hop: 54 }
4681 tcp3: { gateway: 10.3.4.3@tcp, priority: 3 }
4682 tcp4: { gateway: 10.3.3.4@tcp }
4685 # Create the verification file to verify the output of
4686 # lustre_routes_config script against
4687 cat <<- VERIFY_LNET_CONFIG > $verify_conf
4688 lctl --net tcp1 add_route 192.168.213.1@tcp 23 1
4689 lctl --net tcp5 add_route 193.30.4.3@tcp 34 4
4690 lctl --net tcp2 add_route 10.1.3.2@tcp 54 4
4691 lctl --net tcp3 add_route 10.3.4.3@tcp 1 3
4692 lctl --net tcp4 add_route 10.3.3.4@tcp 1 3
4695 lustre_routes_conversion $legacy $new > /dev/null
4696 if [ -f $new ]; then
4697 # verify the conversion output
4698 cmp -s $new $verify > /dev/null
4699 if [ $? -eq 1 ]; then
4700 error "routes conversion failed"
4703 lustre_routes_config --dry-run --verbose $new > $out
4704 # check that the script succeeded
4705 cmp -s $out $verify_conf > /dev/null
4706 if [ $? -eq 1 ]; then
4707 error "routes config failed"
4710 error "routes conversion test failed"
4712 # remove generated files
4713 rm -f $new $legacy $verify $verify_conf $out
4715 run_test 67 "test routes conversion and configuration"
4723 [ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.4.53) ] ||
4724 { skip "Need MDS version at least 2.4.53"; return 0; }
4726 umount_client $MOUNT || error "umount client failed"
4728 if ! combined_mgs_mds; then
4729 start_mgs || error "start mgs failed"
4732 start_mdt 1 || error "MDT start failed"
4733 start_ost || error "Unable to start OST1"
4735 # START-END - the sequences we'll be reserving
4736 START=$(do_facet $SINGLEMDS \
4737 $LCTL get_param -n seq.ctl*.space | awk -F'[[ ]' '{print $2}')
4738 END=$((START + (1 << 30)))
4739 do_facet $SINGLEMDS \
4740 $LCTL set_param seq.ctl*.fldb="[$START-$END\):0:mdt"
4742 # reset the sequences MDT0000 has already assigned
4743 do_facet $SINGLEMDS \
4744 $LCTL set_param seq.srv*MDT0000.space=clear
4746 # remount to let the client allocate new sequence
4747 mount_client $MOUNT || error "mount client failed"
4749 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
4750 do_facet $SINGLEMDS \
4751 $LCTL get_param seq.srv*MDT0000.space
4752 $LFS path2fid $DIR/$tfile
4754 local old_ifs="$IFS"
4756 fid=($($LFS path2fid $DIR/$tfile))
4760 if [[ $seq < $END ]]; then
4761 error "used reserved sequence $seq?"
4763 cleanup || error "cleanup failed with $?"
4765 run_test 68 "be able to reserve specific sequences in FLDB"
4767 # Test 69: is about the total number of objects ever created on an OST.
4768 # so that when it is reformatted the normal MDS->OST orphan recovery won't
4769 # just "precreate" the missing objects. In the past it might try to recreate
4770 # millions of objects after an OST was reformatted
4772 local server_version=$(lustre_version_code $SINGLEMDS)
4774 [[ $server_version -lt $(version_code 2.4.2) ]] &&
4775 skip "Need MDS version at least 2.4.2" && return
4777 [[ $server_version -ge $(version_code 2.4.50) ]] &&
4778 [[ $server_version -lt $(version_code 2.5.0) ]] &&
4779 skip "Need MDS version at least 2.5.0" && return
4782 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
4784 # use OST0000 since it probably has the most creations
4785 local OSTNAME=$(ostname_from_index 0)
4786 local mdtosc_proc1=$(get_mdtosc_proc_path mds1 $OSTNAME)
4787 local last_id=$(do_facet mds1 $LCTL get_param -n \
4788 osc.$mdtosc_proc1.prealloc_last_id)
4790 # Want to have OST LAST_ID over 5 * OST_MAX_PRECREATE to
4791 # verify that the LAST_ID recovery is working properly. If
4792 # not, then the OST will refuse to allow the MDS connect
4793 # because the LAST_ID value is too different from the MDS
4794 #define OST_MAX_PRECREATE=20000
4795 local ost_max_pre=20000
4796 local num_create=$(( ost_max_pre * 5 + 1 - last_id))
4798 # If the LAST_ID is already over 5 * OST_MAX_PRECREATE, we don't
4799 # need to create any files. So, skip this section.
4800 if [ $num_create -gt 0 ]; then
4801 # Check the number of inodes available on OST0
4803 local ifree=$($LFS df -i $MOUNT | awk '/OST0000/ { print $4 }')
4804 log "On OST0, $ifree inodes available. Want $num_create."
4806 $SETSTRIPE -i 0 $DIR/$tdir ||
4807 error "$SETSTRIPE -i 0 $DIR/$tdir failed"
4808 if [ $ifree -lt 10000 ]; then
4809 files=$(( ifree - 50 ))
4814 local j=$((num_create / files + 1))
4815 for i in $(seq 1 $j); do
4816 createmany -o $DIR/$tdir/$tfile-$i- $files ||
4817 error "createmany fail create $files files: $?"
4818 unlinkmany $DIR/$tdir/$tfile-$i- $files ||
4819 error "unlinkmany failed unlink $files files"
4823 # delete all of the files with objects on OST0 so the
4824 # filesystem is not inconsistent later on
4825 $LFS find $MOUNT --ost 0 -print0 | xargs -0 rm
4827 umount_client $MOUNT || error "umount client failed"
4828 stop_ost || error "OST0 stop failure"
4829 add ost1 $(mkfs_opts ost1 $(ostdevname 1)) --reformat --replace \
4830 $(ostdevname 1) $(ostvdevname 1) ||
4831 error "reformat and replace $ostdev failed"
4832 start_ost || error "OST0 restart failure"
4833 wait_osc_import_state mds ost FULL
4835 mount_client $MOUNT || error "mount client failed"
4836 touch $DIR/$tdir/$tfile-last || error "create file after reformat"
4837 local idx=$($GETSTRIPE -i $DIR/$tdir/$tfile-last)
4838 [ $idx -ne 0 ] && error "$DIR/$tdir/$tfile-last on $idx not 0" || true
4840 local iused=$($LFS df -i $MOUNT | awk '/OST0000/ { print $3 }')
4841 log "On OST0, $iused used inodes"
4842 [ $iused -ge $((ost_max_pre/2 + 1000)) ] &&
4843 error "OST replacement created too many inodes; $iused"
4844 cleanup || error "cleanup failed with $?"
4846 run_test 69 "replace an OST with the same index"
4849 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
4852 cleanup || error "cleanup failed with $?"
4854 start_mdt 1 || error "MDT0 start fail"
4856 start_ost || error "OST0 start fail"
4857 for num in $(seq 2 $MDSCOUNT); do
4858 start_mdt $num || return
4861 mount_client $MOUNT || error "mount client fails"
4863 mkdir $DIR/$tdir || error "create $DIR/$tdir failed"
4865 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
4866 error "create remote dir fail"
4868 rm -rf $DIR/$tdir || error "delete dir fail"
4869 cleanup || error "cleanup failed with $?"
4871 run_test 70a "start MDT0, then OST, then MDT1"
4874 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
4877 start_ost || error "OST0 start fail"
4879 start_mds || error "MDS start fail"
4881 mount_client $MOUNT || error "mount client fails"
4883 mkdir $DIR/$tdir || error "create $DIR/$tdir failed"
4885 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
4886 error "create remote dir fail"
4888 rm -rf $DIR/$tdir || error "delete dir fail"
4890 cleanup || error "cleanup failed with $?"
4892 run_test 70b "start OST, MDT1, MDT0"
4895 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
4898 start_mds || error "MDS start fail"
4899 start_ost || error "OST0 start fail"
4901 mount_client $MOUNT || error "mount client fails"
4902 stop_mdt 1 || error "MDT1 start fail"
4904 local mdc_for_mdt1=$($LCTL dl | grep MDT0000-mdc | awk '{print $4}')
4905 echo "deactivate $mdc_for_mdt1"
4906 $LCTL --device $mdc_for_mdt1 deactivate ||
4907 error "set $mdc_for_mdt1 deactivate failed"
4909 mkdir $DIR/$tdir && error "mkdir succeed"
4911 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir &&
4912 error "create remote dir succeed"
4914 cleanup || error "cleanup failed with $?"
4916 run_test 70c "stop MDT0, mkdir fail, create remote dir fail"
4919 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
4922 start_mds || error "MDS start fail"
4923 start_ost || error "OST0 start fail"
4925 mount_client $MOUNT || error "mount client fails"
4927 stop_mdt 2 || error "MDT1 start fail"
4929 local mdc_for_mdt2=$($LCTL dl | grep MDT0001-mdc |
4931 echo "deactivate $mdc_for_mdt2"
4932 $LCTL --device $mdc_for_mdt2 deactivate ||
4933 error "set $mdc_for_mdt2 deactivate failed"
4935 mkdir $DIR/$tdir || error "mkdir fail"
4936 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir &&
4937 error "create remote dir succeed"
4939 rm -rf $DIR/$tdir || error "delete dir fail"
4941 cleanup || error "cleanup failed with $?"
4943 run_test 70d "stop MDT1, mkdir succeed, create remote dir fail"
4946 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
4948 [ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.7.62) ] ||
4949 { skip "Need MDS version at least 2.7.62"; return 0; }
4951 cleanup || error "cleanup failed with $?"
4953 local mdsdev=$(mdsdevname 1)
4954 local ostdev=$(ostdevname 1)
4955 local mdsvdev=$(mdsvdevname 1)
4956 local ostvdev=$(ostvdevname 1)
4957 local opts_mds="$(mkfs_opts mds1 $mdsdev) --reformat $mdsdev $mdsvdev"
4958 local opts_ost="$(mkfs_opts ost1 $ostdev) --reformat $ostdev $ostvdev"
4960 add mds1 $opts_mds || error "add mds1 failed"
4961 start_mdt 1 || error "start mdt1 failed"
4962 add ost1 $opts_ost || error "add ost1 failed"
4963 start_ost || error "start ost failed"
4964 mount_client $MOUNT > /dev/null || error "mount client $MOUNT failed"
4966 local soc=$(do_facet mds1 "$LCTL get_param -n \
4967 mdt.*MDT0000.sync_lock_cancel")
4968 [ $soc == "never" ] || error "SoC enabled on single MDS"
4970 for i in $(seq 2 $MDSCOUNT); do
4971 mdsdev=$(mdsdevname $i)
4972 mdsvdev=$(mdsvdevname $i)
4973 opts_mds="$(mkfs_opts mds$i $mdsdev) --reformat $mdsdev \
4975 add mds$i $opts_mds || error "add mds$i failed"
4976 start_mdt $i || error "start mdt$i fail"
4979 wait_dne_interconnect
4981 for i in $(seq $MDSCOUNT); do
4982 soc=$(do_facet mds$i "$LCTL get_param -n \
4983 mdt.*MDT000$((i - 1)).sync_lock_cancel")
4984 [ $soc == "blocking" ] || error "SoC not enabled on DNE"
4987 for i in $(seq 2 $MDSCOUNT); do
4988 stop_mdt $i || error "stop mdt$i fail"
4990 soc=$(do_facet mds1 "$LCTL get_param -n \
4991 mdt.*MDT0000.sync_lock_cancel")
4992 [ $soc == "never" ] || error "SoC enabled on single MDS"
4994 cleanup || error "cleanup failed with $?"
4996 run_test 70e "Sync-on-Cancel will be enabled by default on DNE"
4999 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
5000 if combined_mgs_mds; then
5001 skip "needs separate MGS/MDT" && return
5005 start_mdt 1 || error "MDT0 start fail"
5006 start_ost || error "OST0 start fail"
5007 for num in $(seq 2 $MDSCOUNT); do
5008 start_mdt $num || return
5011 start_ost2 || error "OST1 start fail"
5013 mount_client $MOUNT || error "mount client fails"
5015 mkdir $DIR/$tdir || error "mkdir fail"
5016 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
5017 error "create remote dir succeed"
5019 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
5020 rm -rf $DIR/$tdir || error "delete dir fail"
5022 umount_client $MOUNT || error "umount_client failed"
5023 stop_mds || error "MDS stop fail"
5024 stop_ost || error "OST0 stop fail"
5025 stop_ost2 || error "OST1 stop fail"
5027 run_test 71a "start MDT0 OST0, MDT1, OST1"
5030 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
5031 if combined_mgs_mds; then
5032 skip "needs separate MGS/MDT" && return
5036 for num in $(seq 2 $MDSCOUNT); do
5037 start_mdt $num || return
5039 start_ost || error "OST0 start fail"
5040 start_mdt 1 || error "MDT0 start fail"
5041 start_ost2 || error "OST1 start fail"
5043 mount_client $MOUNT || error "mount client fails"
5045 mkdir $DIR/$tdir || error "mkdir fail"
5046 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
5047 error "create remote dir succeed"
5049 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
5050 rm -rf $DIR/$tdir || error "delete dir fail"
5052 umount_client $MOUNT || error "umount_client failed"
5053 stop_mds || error "MDT0 stop fail"
5054 stop_ost || error "OST0 stop fail"
5055 stop_ost2 || error "OST1 stop fail"
5057 run_test 71b "start MDT1, OST0, MDT0, OST1"
5060 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
5061 if combined_mgs_mds; then
5062 skip "needs separate MGS/MDT" && return
5066 start_ost || error "OST0 start fail"
5067 start_ost2 || error "OST1 start fail"
5068 for num in $(seq 2 $MDSCOUNT); do
5069 start_mdt $num || return
5071 start_mdt 1 || error "MDT0 start fail"
5073 mount_client $MOUNT || error "mount client fails"
5075 mkdir $DIR/$tdir || error "mkdir fail"
5076 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
5077 error "create remote dir succeed"
5079 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
5080 rm -rf $DIR/$tdir || error "delete dir fail"
5082 umount_client $MOUNT || error "umount_client failed"
5083 stop_mds || error "MDS stop fail"
5084 stop_ost || error "OST0 stop fail"
5085 stop_ost2 || error "OST1 stop fail"
5088 run_test 71c "start OST0, OST1, MDT1, MDT0"
5091 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
5092 if combined_mgs_mds; then
5093 skip "needs separate MGS/MDT" && return
5097 start_ost || error "OST0 start fail"
5098 for num in $(seq 2 $MDSCOUNT); do
5099 start_mdt $num || return
5101 start_mdt 1 || error "MDT0 start fail"
5102 start_ost2 || error "OST1 start fail"
5104 mount_client $MOUNT || error "mount client fails"
5106 mkdir $DIR/$tdir || error "mkdir fail"
5107 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
5108 error "create remote dir succeed"
5110 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
5111 rm -rf $DIR/$tdir || error "delete dir fail"
5113 umount_client $MOUNT || error "umount_client failed"
5114 stop_mds || error "MDS stop fail"
5115 stop_ost || error "OST0 stop fail"
5116 stop_ost2 || error "OST1 stop fail"
5119 run_test 71d "start OST0, MDT1, MDT0, OST1"
5122 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
5123 if combined_mgs_mds; then
5124 skip "needs separate MGS/MDT" && return
5128 start_ost || error "OST0 start fail"
5129 for num in $(seq 2 $MDSCOUNT); do
5130 start_mdt $num || return
5132 start_ost2 || error "OST1 start fail"
5133 start_mdt 1 || error "MDT0 start fail"
5135 mount_client $MOUNT || error "mount client fails"
5137 mkdir $DIR/$tdir || error "mkdir fail"
5138 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
5139 error "create remote dir succeed"
5141 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
5142 rm -rf $DIR/$tdir || error "delete dir fail"
5144 umount_client $MOUNT || error "umount_client failed"
5145 stop_mds || error "MDS stop fail"
5146 stop_ost || error "OST0 stop fail"
5147 stop_ost2 || error "OST1 stop fail"
5150 run_test 71e "start OST0, MDT1, OST1, MDT0"
5152 test_72() { #LU-2634
5153 local mdsdev=$(mdsdevname 1)
5154 local ostdev=$(ostdevname 1)
5155 local cmd="$E2FSCK -fnvd $mdsdev"
5159 [ "$(facet_fstype $SINGLEMDS)" != "ldiskfs" ] &&
5160 skip "ldiskfs only test" && return
5162 if combined_mgs_mds; then
5163 add_options='--reformat'
5165 add_options='--reformat --replace'
5168 #tune MDT with "-O extents"
5170 for num in $(seq $MDSCOUNT); do
5171 add mds${num} $(mkfs_opts mds$num $(mdsdevname $num)) \
5172 $add_options $(mdsdevname $num) $(mdsvdevname $num) ||
5173 error "add mds $num failed"
5174 do_facet mds${num} "$TUNE2FS -O extents $(mdsdevname $num)" ||
5175 error "$TUNE2FS failed on mds${num}"
5178 add ost1 $(mkfs_opts ost1 $ostdev) $add_options $ostdev ||
5179 error "add $ostdev failed"
5180 start_mds || error "start mds failed"
5181 start_ost || error "start ost failed"
5182 mount_client $MOUNT || error "mount client failed"
5184 #create some short symlinks
5185 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
5186 createmany -o $DIR/$tdir/$tfile-%d $fn
5187 echo "create $fn short symlinks"
5188 for i in $(seq -w 1 $fn); do
5189 ln -s $DIR/$tdir/$tfile-$i $MOUNT/$tfile-$i
5194 umount_client $MOUNT || error "umount client failed"
5195 stop_mds || error "stop mds failed"
5196 stop_ost || error "stop ost failed"
5199 run_e2fsck $(facet_active_host $SINGLEMDS) $mdsdev "-n"
5201 run_test 72 "test fast symlink with extents flag enabled"
5203 test_73() { #LU-3006
5204 [ $(facet_fstype ost1) == zfs ] && import_zpool ost1
5205 do_facet ost1 "$TUNEFS --failnode=1.2.3.4@$NETTYPE $(ostdevname 1)" ||
5206 error "1st tunefs failed"
5207 start_mgsmds || error "start mds failed"
5208 start_ost || error "start ost failed"
5209 mount_client $MOUNT || error "mount client failed"
5210 $LCTL get_param -n osc.*OST0000-osc-[^M]*.import | grep failover_nids |
5211 grep 1.2.3.4@$NETTYPE || error "failover nids haven't changed"
5212 umount_client $MOUNT || error "umount client failed"
5216 run_test 73 "failnode to update from mountdata properly"
5218 test_75() { # LU-2374
5219 [[ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.4.1) ]] &&
5220 skip "Need MDS version at least 2.4.1" && return
5223 local opts_mds="$(mkfs_opts mds1 $(mdsdevname 1)) \
5224 --reformat $(mdsdevname 1) $(mdsvdevname 1)"
5225 local opts_ost="$(mkfs_opts ost1 $(ostdevname 1)) \
5226 --reformat $(ostdevname 1) $(ostvdevname 1)"
5228 #check with default parameters
5229 add mds1 $opts_mds || error "add mds1 failed for default params"
5230 add ost1 $opts_ost || error "add ost1 failed for default params"
5232 opts_mds=$(echo $opts_mds | sed -e "s/--mdt//")
5233 opts_mds=$(echo $opts_mds |
5234 sed -e "s/--index=$index/--index=$index --mdt/")
5235 opts_ost=$(echo $opts_ost | sed -e "s/--ost//")
5236 opts_ost=$(echo $opts_ost |
5237 sed -e "s/--index=$index/--index=$index --ost/")
5239 add mds1 $opts_mds || error "add mds1 failed for new params"
5240 add ost1 $opts_ost || error "add ost1 failed for new params"
5241 if ! combined_mgs_mds; then
5242 stop_mgs || error "stop mgs failed"
5247 run_test 75 "The order of --index should be irrelevant"
5250 [[ $(lustre_version_code mgs) -ge $(version_code 2.4.52) ]] ||
5251 { skip "Need MDS version at least 2.4.52" && return 0; }
5253 if ! combined_mgs_mds; then
5254 start_mgs || error "start mgs failed"
5257 local MDMB_PARAM="osc.*.max_dirty_mb"
5258 echo "Change MGS params"
5259 local MAX_DIRTY_MB=$($LCTL get_param -n $MDMB_PARAM |
5261 echo "max_dirty_mb: $MAX_DIRTY_MB"
5262 local NEW_MAX_DIRTY_MB=$((MAX_DIRTY_MB + MAX_DIRTY_MB))
5263 echo "new_max_dirty_mb: $NEW_MAX_DIRTY_MB"
5264 do_facet mgs $LCTL set_param -P $MDMB_PARAM=$NEW_MAX_DIRTY_MB
5265 wait_update $HOSTNAME "$LCTL get_param -n $MDMB_PARAM |
5266 head -1" $NEW_MAX_DIRTY_MB
5267 MAX_DIRTY_MB=$($LCTL get_param -n $MDMB_PARAM | head -1)
5268 echo "$MAX_DIRTY_MB"
5269 [ $MAX_DIRTY_MB = $NEW_MAX_DIRTY_MB ] ||
5270 error "error while apply max_dirty_mb"
5272 echo "Check the value is stored after remount"
5275 wait_update $HOSTNAME "$LCTL get_param -n $MDMB_PARAM |
5276 head -1" $NEW_MAX_DIRTY_MB
5277 MAX_DIRTY_MB=$($LCTL get_param -n $MDMB_PARAM | head -1)
5278 [ $MAX_DIRTY_MB = $NEW_MAX_DIRTY_MB ] ||
5279 error "max_dirty_mb is not saved after remount"
5281 echo "Change OST params"
5282 CLIENT_PARAM="obdfilter.*.client_cache_count"
5283 local CLIENT_CACHE_COUNT
5284 CLIENT_CACHE_COUNT=$(do_facet ost1 $LCTL get_param -n $CLIENT_PARAM |
5286 echo "client_cache_count: $CLIENT_CACHE_COUNT"
5287 NEW_CLIENT_CACHE_COUNT=$((CLIENT_CACHE_COUNT+CLIENT_CACHE_COUNT))
5288 echo "new_client_cache_count: $NEW_CLIENT_CACHE_COUNT"
5289 do_facet mgs $LCTL set_param -P $CLIENT_PARAM=$NEW_CLIENT_CACHE_COUNT
5290 wait_update $(facet_host ost1) "$LCTL get_param -n $CLIENT_PARAM |
5291 head -1" $NEW_CLIENT_CACHE_COUNT
5292 CLIENT_CACHE_COUNT=$(do_facet ost1 $LCTL get_param -n $CLIENT_PARAM |
5294 echo "$CLIENT_CACHE_COUNT"
5295 [ $CLIENT_CACHE_COUNT = $NEW_CLIENT_CACHE_COUNT ] ||
5296 error "error while apply client_cache_count"
5298 echo "Check the value is stored after remount"
5301 wait_update $(facet_host ost1) "$LCTL get_param -n $CLIENT_PARAM |
5302 head -1" $NEW_CLIENT_CACHE_COUNT
5303 CLIENT_CACHE_COUNT=$(do_facet ost1 $LCTL get_param -n $CLIENT_PARAM |
5305 echo "$CLIENT_CACHE_COUNT"
5306 [ $CLIENT_CACHE_COUNT = $NEW_CLIENT_CACHE_COUNT ] ||
5307 error "client_cache_count is not saved after remount"
5310 run_test 76a "set permanent params set_param -P"
5312 test_76b() { # LU-4783
5313 [[ $(lustre_version_code mgs) -ge $(version_code 2.5.57) ]] ||
5314 { skip "Need MGS version at least 2.5.57" && return 0; }
5317 do_facet mgs $LCTL get_param mgs.MGS.live.params ||
5318 error "start params log failed"
5321 run_test 76b "verify params log setup correctly"
5324 [[ $(lustre_version_code mgs) -ge $(version_code 2.8.54) ]] ||
5325 { skip "Need MDS version at least 2.4.52" && return 0; }
5327 local MASK_PARAM="mdd.*.changelog_mask"
5328 echo "Change changelog_mask"
5329 do_facet mgs $LCTL set_param -P $MASK_PARAM=-CLOSE ||
5330 error "Can't change changlog_mask"
5331 wait_update $(facet_host mds) "$LCTL get_param -n $MASK_PARAM |
5334 echo "Check the value is stored after mds remount"
5335 stop_mds || error "Failed to stop MDS"
5336 start_mds || error "Failed to start MDS"
5337 local CHANGELOG_MASK=$(do_facet mgs $LCTL get_param -n $MASK_PARAM)
5338 echo $CHANGELOG_MASK | grep CLOSE > /dev/null &&
5339 error "changelog_mask is not changed"
5343 run_test 76c "verify changelog_mask is applied with set_param -P"
5345 test_77() { # LU-3445
5346 local server_version=$(lustre_version_code $SINGLEMDS)
5347 [[ $server_version -ge $(version_code 2.8.55) ]] ||
5348 { skip "Need MDS version 2.8.55+ "; return; }
5350 if [[ -z "$fs2ost_DEV" || -z "$fs2mds_DEV" ]]; then
5351 is_blkdev $SINGLEMDS $(mdsdevname ${SINGLEMDS//mds/}) &&
5352 skip_env "mixed loopback and real device not working" && return
5355 local fs2mdsdev=$(mdsdevname 1_2)
5356 local fs2ostdev=$(ostdevname 1_2)
5357 local fs2mdsvdev=$(mdsvdevname 1_2)
5358 local fs2ostvdev=$(ostvdevname 1_2)
5359 local fsname=test1234
5361 local failnid="$(h2$NETTYPE 1.2.3.4),$(h2$NETTYPE 4.3.2.1)"
5363 combined_mgs_mds || stop_mgs || error "stopping MGS service failed"
5365 add fs2mds $(mkfs_opts mds1 $fs2mdsdev) --mgs --fsname=$fsname \
5366 --reformat $fs2mdsdev $fs2mdsvdev || error "add fs2mds failed"
5367 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS && trap cleanup_fs2 EXIT INT ||
5368 error "start fs2mds failed"
5370 mgsnid=$(do_facet fs2mds $LCTL list_nids | xargs | tr ' ' ,)
5371 mgsnid="0.0.0.0@tcp,$mgsnid,$mgsnid:$mgsnid"
5373 add fs2ost --mgsnode=$mgsnid $(mkfs_opts ost1 $fs2ostdev) \
5374 --failnode=$failnid --fsname=$fsname \
5375 --reformat $fs2ostdev $fs2ostvdev ||
5376 error "add fs2ost failed"
5377 start fs2ost $fs2ostdev $OST_MOUNT_OPTS || error "start fs2ost failed"
5379 mkdir -p $MOUNT2 || error "mkdir $MOUNT2 failed"
5380 $MOUNT_CMD $mgsnid:/$fsname $MOUNT2 || error "mount $MOUNT2 failed"
5381 DIR=$MOUNT2 MOUNT=$MOUNT2 check_mount || error "check $MOUNT2 failed"
5384 run_test 77 "comma-separated MGS NIDs and failover node NIDs"
5387 [[ $(facet_fstype $SINGLEMDS) != ldiskfs ||
5388 $(facet_fstype ost1) != ldiskfs ]] &&
5389 skip "only applicable to ldiskfs-based MDTs and OSTs" && return
5391 # reformat the Lustre filesystem with a smaller size
5392 local saved_MDSCOUNT=$MDSCOUNT
5393 local saved_MDSSIZE=$MDSSIZE
5394 local saved_OSTCOUNT=$OSTCOUNT
5395 local saved_OSTSIZE=$OSTSIZE
5398 MDSSIZE=$((MDSSIZE - 20000))
5399 OSTSIZE=$((OSTSIZE - 20000))
5400 reformat || error "(1) reformat Lustre filesystem failed"
5401 MDSSIZE=$saved_MDSSIZE
5402 OSTSIZE=$saved_OSTSIZE
5404 # mount the Lustre filesystem
5405 setup_noconfig || error "(2) setup Lustre filesystem failed"
5408 log "create test files"
5413 mkdir $MOUNT/$tdir || error "(3) mkdir $MOUNT/$tdir failed"
5415 for i in $(seq $num_files); do
5416 file=$MOUNT/$tdir/$tfile-$i
5417 dd if=/dev/urandom of=$file count=1 bs=1M || {
5418 $LCTL get_param osc.*.cur*grant*
5419 $LFS df; $LFS df -i;
5420 # stop creating files if there is no more space
5421 if [ ! -e $file ]; then
5422 num_files=$((i - 1))
5426 $LFS getstripe -v $file
5427 local ost_idx=$(LFS getstripe -i $file)
5428 do_facet ost$((ost_idx + 1)) \
5429 $LCTL get_param obdfilter.*.*grant*
5430 error "(4) create $file failed"
5434 # unmount the Lustre filesystem
5435 cleanup || error "(5) cleanup Lustre filesystem failed"
5437 # run e2fsck on the MDT and OST devices
5438 local mds_host=$(facet_active_host $SINGLEMDS)
5439 local ost_host=$(facet_active_host ost1)
5440 local mds_dev=$(mdsdevname ${SINGLEMDS//mds/})
5441 local ost_dev=$(ostdevname 1)
5443 run_e2fsck $mds_host $mds_dev "-y"
5444 run_e2fsck $ost_host $ost_dev "-y"
5446 # get the original block count of the MDT and OST filesystems
5447 local mds_orig_blks=$(get_block_count $SINGLEMDS $mds_dev)
5448 local ost_orig_blks=$(get_block_count ost1 $ost_dev)
5450 # expand the MDT and OST filesystems to the device size
5451 run_resize2fs $SINGLEMDS $mds_dev "" || error "expand $SINGLEMDS failed"
5452 run_resize2fs ost1 $ost_dev "" || error "expand ost1 failed"
5454 # run e2fsck on the MDT and OST devices again
5455 run_e2fsck $mds_host $mds_dev "-y"
5456 run_e2fsck $ost_host $ost_dev "-y"
5458 # mount the Lustre filesystem
5462 log "check files after expanding the MDT and OST filesystems"
5463 for i in $(seq $num_files); do
5464 file=$MOUNT/$tdir/$tfile-$i
5465 $CHECKSTAT -t file -s 1048576 $file ||
5466 error "(6) checkstat $file failed"
5470 log "create more files after expanding the MDT and OST filesystems"
5471 for i in $(seq $((num_files + 1)) $((num_files + 10))); do
5472 file=$MOUNT/$tdir/$tfile-$i
5473 dd if=/dev/urandom of=$file count=1 bs=1M ||
5474 error "(7) create $file failed"
5477 # unmount the Lustre filesystem
5478 cleanup || error "(8) cleanup Lustre filesystem failed"
5480 # run e2fsck on the MDT and OST devices
5481 run_e2fsck $mds_host $mds_dev "-y"
5482 run_e2fsck $ost_host $ost_dev "-y"
5484 # get the maximum block count of the MDT and OST filesystems
5485 local mds_max_blks=$(get_block_count $SINGLEMDS $mds_dev)
5486 local ost_max_blks=$(get_block_count ost1 $ost_dev)
5488 # get the minimum block count of the MDT and OST filesystems
5489 local mds_min_blks=$(run_resize2fs $SINGLEMDS $mds_dev "" "-P" 2>&1 |
5490 grep minimum | sed -e 's/^.*filesystem: //g')
5491 local ost_min_blks=$(run_resize2fs ost1 $ost_dev "" "-P" 2>&1 |
5492 grep minimum | sed -e 's/^.*filesystem: //g')
5494 # shrink the MDT and OST filesystems to a smaller size
5498 if [[ $mds_max_blks -gt $mds_min_blks &&
5499 $mds_max_blks -gt $mds_orig_blks ]]; then
5500 [[ $mds_orig_blks -gt $mds_min_blks ]] &&
5501 base_blks=$mds_orig_blks || base_blks=$mds_min_blks
5502 new_blks=$(( (mds_max_blks - base_blks) / 2 + base_blks ))
5503 run_resize2fs $SINGLEMDS $mds_dev $new_blks ||
5504 error "shrink $SINGLEMDS to $new_blks failed"
5508 if [[ $ost_max_blks -gt $ost_min_blks &&
5509 $ost_max_blks -gt $ost_orig_blks ]]; then
5510 [[ $ost_orig_blks -gt $ost_min_blks ]] &&
5511 base_blks=$ost_orig_blks || base_blks=$ost_min_blks
5512 new_blks=$(( (ost_max_blks - base_blks) / 2 + base_blks ))
5513 run_resize2fs ost1 $ost_dev $new_blks ||
5514 error "shrink ost1 to $new_blks failed"
5518 # check whether the MDT or OST filesystem was shrunk or not
5520 combined_mgs_mds || stop_mgs || error "(9) stop mgs failed"
5521 reformat || error "(10) reformat Lustre filesystem failed"
5525 # run e2fsck on the MDT and OST devices again
5526 run_e2fsck $mds_host $mds_dev "-y"
5527 run_e2fsck $ost_host $ost_dev "-y"
5529 # mount the Lustre filesystem again
5533 log "check files after shrinking the MDT and OST filesystems"
5534 for i in $(seq $((num_files + 10))); do
5535 file=$MOUNT/$tdir/$tfile-$i
5536 $CHECKSTAT -t file -s 1048576 $file ||
5537 error "(11) checkstat $file failed"
5540 # unmount and reformat the Lustre filesystem
5541 cleanup || error "(12) cleanup Lustre filesystem failed"
5542 combined_mgs_mds || stop_mgs || error "(13) stop mgs failed"
5544 MDSCOUNT=$saved_MDSCOUNT
5545 OSTCOUNT=$saved_OSTCOUNT
5546 reformat || error "(14) reformat Lustre filesystem failed"
5548 run_test 78 "run resize2fs on MDT and OST filesystems"
5550 test_79() { # LU-4227
5551 [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.5.59) ]] ||
5552 { skip "Need MDS version at least 2.5.59"; return 0; }
5554 local mdsdev1=$(mdsdevname 1)
5555 local mdsvdev1=$(mdsvdevname 1)
5556 local mdsdev2=$(mdsdevname 2)
5557 local mdsvdev2=$(mdsvdevname 2)
5558 local ostdev1=$(ostdevname 1)
5559 local ostvdev1=$(ostvdevname 1)
5560 local opts_mds1="$(mkfs_opts mds1 $mdsdev1) --reformat"
5561 local opts_mds2="$(mkfs_opts mds2 $mdsdev2) --reformat"
5562 local opts_ost1="$(mkfs_opts ost1 $ostdev1) --reformat"
5565 # remove --mgs/--mgsnode from mkfs.lustre options
5566 opts_mds1=$(echo $opts_mds1 | sed -e "s/--mgs//")
5568 mgsnode_opt=$(echo $opts_mds2 |
5569 awk '{ for ( i = 1; i < NF; i++ )
5570 if ( $i ~ "--mgsnode" ) { print $i; break } }')
5571 [ -n $mgsnode_opt ] &&
5572 opts_mds2=$(echo $opts_mds2 | sed -e "s/$mgsnode_opt//")
5574 mgsnode_opt=$(echo $opts_ost1 |
5575 awk '{ for ( i = 1; i < NF; i++ )
5576 if ( $i ~ "--mgsnode" ) { print $i; break } }')
5577 [ -n $mgsnode_opt ] &&
5578 opts_ost1=$(echo $opts_ost1 | sed -e "s/$mgsnode_opt//")
5580 # -MGS, format a mdt without --mgs option
5581 add mds1 $opts_mds1 $mdsdev1 $mdsvdev1 &&
5582 error "Must specify --mgs when formatting mdt combined with mgs"
5584 # +MGS, format a mdt/ost without --mgsnode option
5585 add mds1 $(mkfs_opts mds1 $mdsdev1) --reformat $mdsdev1 $mdsvdev1 \
5586 > /dev/null || error "start mds1 failed"
5587 add mds2 $opts_mds2 $mdsdev2 $mdsvdev2 &&
5588 error "Must specify --mgsnode when formatting a mdt"
5589 add ost1 $opts_ost1 $ostdev1 $ostvdev2 &&
5590 error "Must specify --mgsnode when formatting an ost"
5594 run_test 79 "format MDT/OST without mgs option (should return errors)"
5597 start_mds || error "Failed to start MDT"
5598 start_ost || error "Failed to start OST1"
5599 uuid=$(do_facet ost1 $LCTL get_param -n mgc.*.uuid)
5600 #define OBD_FAIL_MGS_PAUSE_TARGET_CON 0x906
5601 do_facet ost1 "$LCTL set_param fail_val=10 fail_loc=0x906"
5602 do_facet mgs "$LCTL set_param fail_val=10 fail_loc=0x906"
5603 do_facet mgs "$LCTL set_param -n mgs/MGS/evict_client $uuid"
5605 start_ost2 || error "Failed to start OST2"
5607 do_facet ost1 "$LCTL set_param fail_loc=0"
5610 run_test 80 "mgc import reconnect race"
5612 #Save the original values of $OSTCOUNT and $OSTINDEX$i.
5614 local new_ostcount=$1
5615 saved_ostcount=$OSTCOUNT
5616 OSTCOUNT=$new_ostcount
5620 for i in $(seq $OSTCOUNT); do
5622 eval saved_ostindex$i=${!index}
5627 # Restore the original values of $OSTCOUNT and $OSTINDEX$i.
5628 restore_ostindex() {
5633 for i in $(seq $OSTCOUNT); do
5634 index=saved_ostindex$i
5635 eval OSTINDEX$i=${!index}
5637 OSTCOUNT=$saved_ostcount
5642 # The main purpose of this test is to ensure the OST_INDEX_LIST functions as
5643 # expected. This test uses OST_INDEX_LIST to format OSTs with a randomly
5644 # assigned index and ensures we can mount such a formatted file system
5645 test_81() { # LU-4665
5646 [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.6.54) ]] ||
5647 { skip "Need MDS version at least 2.6.54" && return; }
5648 [[ $OSTCOUNT -ge 3 ]] || { skip_env "Need at least 3 OSTs" && return; }
5652 # Each time RANDOM is referenced, a random integer between 0 and 32767
5655 local saved_ostindex1=$OSTINDEX1
5656 for i in 65535 $((RANDOM + 65536)); do
5657 echo -e "\nFormat ost1 with --index=$i, should fail"
5659 if add ost1 $(mkfs_opts ost1 $(ostdevname 1)) --reformat \
5660 $(ostdevname 1) $(ostvdevname 1); then
5661 OSTINDEX1=$saved_ostindex1
5662 error "format ost1 with --index=$i should fail"
5665 OSTINDEX1=$saved_ostindex1
5669 # Format OSTs with random sparse indices.
5670 trap "restore_ostindex" EXIT
5671 echo -e "\nFormat $OSTCOUNT OSTs with sparse indices"
5672 OST_INDEX_LIST=[0,$((RANDOM * 2 % 65533 + 1)),65534] formatall
5674 # Setup and check Lustre filesystem.
5675 start_mgsmds || error "start_mgsmds failed"
5676 for i in $(seq $OSTCOUNT); do
5677 start ost$i $(ostdevname $i) $OST_MOUNT_OPTS ||
5678 error "start ost$i failed"
5681 mount_client $MOUNT || error "mount client $MOUNT failed"
5682 check_mount || error "check client $MOUNT failed"
5685 local max_easize=$($LCTL get_param -n llite.*.max_easize)
5686 [[ $max_easize -eq 128 ]] ||
5687 error "max_easize is $max_easize, should be 128 bytes"
5691 run_test 81 "sparse OST indexing"
5693 # Here we exercise the stripe placement functionality on a file system that
5694 # has formatted the OST with a random index. With the file system the following
5695 # functionality is tested:
5697 # 1. Creating a new file with a specific stripe layout.
5699 # 2. Modifiy a existing empty file with a specific stripe layout.
5701 # 3. Ensure we fail to set the stripe layout of a file that already has one.
5703 # 4. If ost-index is defined we need to ensure it is the first entry in the
5704 # ost index list returned by lfs getstripe.
5706 # 5. Lastly ensure this functionality fails with directories.
5707 test_82a() { # LU-4665
5708 [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.6.54) ]] ||
5709 { skip "Need MDS version at least 2.6.54" && return; }
5710 [[ $OSTCOUNT -ge 3 ]] || { skip_env "Need at least 3 OSTs" && return; }
5716 # Format OSTs with random sparse indices.
5720 local LOV_V1_INSANE_STRIPE_COUNT=65532
5721 for i in $(seq $OSTCOUNT); do
5722 index=$(((RANDOM * 2) % LOV_V1_INSANE_STRIPE_COUNT))
5723 ost_indices+=" $index"
5725 ost_indices=$(comma_list $ost_indices)
5727 trap "restore_ostindex" EXIT
5728 echo -e "\nFormat $OSTCOUNT OSTs with sparse indices $ost_indices"
5729 OST_INDEX_LIST=[$ost_indices] formatall
5731 # Setup Lustre filesystem.
5732 start_mgsmds || error "start_mgsmds failed"
5733 for i in $(seq $OSTCOUNT); do
5734 start ost$i $(ostdevname $i) $OST_MOUNT_OPTS ||
5735 error "start ost$i failed"
5738 mount_client $MOUNT || error "mount client $MOUNT failed"
5741 $LFS df $MOUNT || error "$LFS df $MOUNT failed"
5742 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
5744 # 1. If the file does not exist, new file will be created
5745 # with specified OSTs.
5746 local file=$DIR/$tdir/$tfile-1
5747 local cmd="$SETSTRIPE -o $ost_indices $file"
5749 eval $cmd || error "$cmd failed"
5750 check_stripe_count $file $OSTCOUNT
5751 check_obdidx $file $ost_indices
5752 dd if=/dev/urandom of=$file count=1 bs=1M > /dev/null 2>&1 ||
5753 error "write $file failed"
5755 # 2. If the file already exists and is an empty file, the file
5756 # will be attached with specified layout.
5757 file=$DIR/$tdir/$tfile-2
5758 mcreate $file || error "mcreate $file failed"
5759 cmd="$SETSTRIPE -o $ost_indices $file"
5761 eval $cmd || error "$cmd failed"
5762 dd if=/dev/urandom of=$file count=1 bs=1M > /dev/null 2>&1 ||
5763 error "write $file failed"
5764 check_stripe_count $file $OSTCOUNT
5765 check_obdidx $file $ost_indices
5767 # 3. If the file already has a valid layout attached, the command
5768 # should fail with EBUSY.
5770 eval $cmd && error "stripe is already set on $file, $cmd should fail"
5772 # 4. If [--stripe-index|-i <start_ost_idx>] is used, the index must
5773 # be in the OST indices list.
5774 local start_ost_idx=${ost_indices##*,}
5775 file=$DIR/$tdir/$tfile-3
5776 cmd="$SETSTRIPE -o $ost_indices -i $start_ost_idx $file"
5778 eval $cmd || error "$cmd failed"
5779 check_stripe_count $file $OSTCOUNT
5780 check_obdidx $file $ost_indices
5781 check_start_ost_idx $file $start_ost_idx
5783 file=$DIR/$tdir/$tfile-4
5785 cmd+=" -o $(exclude_items_from_list $ost_indices $start_ost_idx)"
5786 cmd+=" -i $start_ost_idx $file"
5788 eval $cmd && error "index $start_ost_idx should be in $ost_indices"
5790 # 5. Specifying OST indices for directory should fail with ENOSUPP.
5791 local dir=$DIR/$tdir/$tdir
5792 mkdir $dir || error "mkdir $dir failed"
5793 cmd="$SETSTRIPE -o $ost_indices $dir"
5795 eval $cmd && error "$cmd should fail, specifying OST indices" \
5796 "for directory is not supported"
5800 run_test 82a "specify OSTs for file (succeed) or directory (fail)"
5805 # Remove OSTs from a pool and destroy the pool.
5806 destroy_pool $ost_pool || true
5811 # Test 82b is run to ensure that if the user supplies a pool with a specific
5812 # stripe layout that it behaves proprerly. It should fail in the case that
5813 # the supplied OST index list points to OSTs not contained in the user
5815 test_82b() { # LU-4665
5816 [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.6.54) ]] ||
5817 { skip "Need MDS version at least 2.6.54" && return; }
5818 [[ $OSTCOUNT -ge 4 ]] || { skip_env "Need at least 4 OSTs" && return; }
5824 # Format OSTs with random sparse indices.
5828 local LOV_V1_INSANE_STRIPE_COUNT=65532
5829 for i in $(seq $OSTCOUNT); do
5830 index=$(((RANDOM * 2) % LOV_V1_INSANE_STRIPE_COUNT))
5831 ost_indices+=" $index"
5833 ost_indices=$(comma_list $ost_indices)
5835 trap "restore_ostindex" EXIT
5836 echo -e "\nFormat $OSTCOUNT OSTs with sparse indices $ost_indices"
5837 OST_INDEX_LIST=[$ost_indices] formatall
5839 # Setup Lustre filesystem.
5840 start_mgsmds || error "start_mgsmds failed"
5841 for i in $(seq $OSTCOUNT); do
5842 start ost$i $(ostdevname $i) $OST_MOUNT_OPTS ||
5843 error "start ost$i failed"
5846 mount_client $MOUNT || error "mount client $MOUNT failed"
5848 $LFS df $MOUNT || error "$LFS df $MOUNT failed"
5849 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
5851 # Create a new pool and add OSTs into it.
5852 local ost_pool=$FSNAME.$TESTNAME
5853 create_pool $ost_pool || error "create OST pool $ost_pool failed"
5856 trap "cleanup_82b" EXIT
5858 local ost_idx_in_list=${ost_indices##*,}
5859 local ost_idx_in_pool=$(exclude_items_from_list $ost_indices \
5862 local ost_targets="$FSNAME-OST["
5863 for i in ${ost_idx_in_pool//,/ }; do
5864 ost_targets=$ost_targets$(printf "%04x," $i)
5866 ost_targets="${ost_targets%,}]"
5868 local ost_targets_uuid=$(for i in ${ost_idx_in_pool//,/ }; \
5869 do printf "$FSNAME-OST%04x_UUID\n" $i; done |
5870 sort -u | tr '\n' ' ')
5872 local cmd="$LCTL pool_add $ost_pool $ost_targets"
5873 do_facet mgs $cmd || error "$cmd failed"
5874 wait_update $HOSTNAME "$LCTL get_param -n lov.$FSNAME-*.pools.$TESTNAME|
5875 sort -u | tr '\n' ' ' " "$ost_targets_uuid" ||
5876 error "wait_update $ost_pool failed"
5877 wait_update_facet $SINGLEMDS "$LCTL pool_list $ost_pool | wc -l" 4 ||
5878 error "wait_update pool_list $ost_pool failed"
5880 # If [--pool|-p <pool_name>] is set with [--ost-list|-o <ost_indices>],
5881 # then the OSTs must be the members of the pool.
5882 local file=$DIR/$tdir/$tfile
5883 cmd="$SETSTRIPE -p $ost_pool -o $ost_idx_in_list $file"
5885 eval $cmd && error "OST with index $ost_idx_in_list should be" \
5886 "in OST pool $ost_pool"
5888 # Only select OST $ost_idx_in_list from $ost_pool for file.
5889 ost_idx_in_list=${ost_idx_in_pool#*,}
5890 cmd="$SETSTRIPE -p $ost_pool -o $ost_idx_in_list $file"
5892 eval $cmd || error "$cmd failed"
5893 cmd="$GETSTRIPE $file"
5895 eval $cmd || error "$cmd failed"
5896 check_stripe_count $file 2
5897 check_obdidx $file $ost_idx_in_list
5898 dd if=/dev/urandom of=$file count=1 bs=1M > /dev/null 2>&1 ||
5899 error "write $file failed"
5903 run_test 82b "specify OSTs for file with --pool and --ost-list options"
5906 [[ $(lustre_version_code ost1) -ge $(version_code 2.6.91) ]] ||
5907 { skip "Need OST version at least 2.6.91" && return 0; }
5908 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
5909 skip "Only applicable to ldiskfs-based MDTs"
5919 ostmnt=$(facet_mntpt ost1)
5920 fstype=$(facet_fstype ost1)
5922 # Mount the OST as an ldiskfs filesystem.
5923 log "mount the OST $dev as a $fstype filesystem"
5924 add ost1 $(mkfs_opts ost1 $dev) $FSTYPE_OPT \
5925 --reformat $dev $dev > /dev/null ||
5926 error "format ost1 error"
5928 if ! test -b $dev; then
5929 mnt_opts=$(csa_add "$OST_MOUNT_OPTS" -o loop)
5931 echo "mnt_opts $mnt_opts"
5932 do_facet ost1 mount -t $fstype $dev \
5934 # Run llverfs on the mounted ldiskfs filesystem.
5935 # It is needed to get ENOSPACE.
5936 log "run llverfs in partial mode on the OST $fstype $ostmnt"
5937 do_rpc_nodes $(facet_host ost1) run_llverfs $ostmnt -vpl \
5938 "no" || error "run_llverfs error on $fstype"
5941 log "unmount the OST $dev"
5944 # Delete file IO_scrub. Later osd_scrub_setup will try to
5945 # create "IO_scrub" but will get ENOSPACE.
5947 echo "start ost1 service on `facet_active_host ost1`"
5948 start ost1 `ostdevname 1` $OST_MOUNT_OPTS
5951 err=$(do_facet ost1 dmesg | grep "VFS: Busy inodes after unmount of")
5952 echo "string err $err"
5953 [ -z "$err" ] || error $err
5956 run_test 83 "ENOSPACE on OST doesn't cause message VFS: \
5957 Busy inodes after unmount ..."
5960 local facet=$SINGLEMDS
5961 local num=$(echo $facet | tr -d "mds")
5962 local dev=$(mdsdevname $num)
5963 local time_min=$(recovery_time_min)
5964 local recovery_duration
5965 local completed_clients
5966 local correct_clients
5969 echo "start mds service on $(facet_active_host $facet)"
5971 "-o recovery_time_hard=$time_min,recovery_time_soft=$time_min" $@ ||
5972 error "start MDS failed"
5974 start_ost || error "start OST0000 failed"
5975 start_ost2 || error "start OST0001 failed"
5977 echo "recovery_time=$time_min, timeout=$TIMEOUT, wrap_up=$wrap_up"
5979 mount_client $MOUNT1 || error "mount $MOUNT1 failed"
5980 mount_client $MOUNT2 || error "mount $MOUNT2 failed"
5981 # make sure new superblock labels are sync'd before disabling writes
5985 replay_barrier $SINGLEMDS
5986 createmany -o $DIR1/$tfile-%d 1000
5988 # We need to catch the end of recovery window to extend it.
5989 # Skip 5 requests and add delay to request handling.
5990 #define OBD_FAIL_TGT_REPLAY_DELAY 0x709 | FAIL_SKIP
5991 do_facet $SINGLEMDS "lctl set_param fail_loc=0x20000709 fail_val=5"
5993 facet_failover --fsck $SINGLEMDS || error "failover: $?"
5996 echo "recovery status"
5997 do_facet $SINGLEMDS \
5998 "$LCTL get_param -n mdt.$FSNAME-MDT0000.recovery_status"
6000 recovery_duration=$(do_facet $SINGLEMDS \
6001 "$LCTL get_param -n mdt.$FSNAME-MDT0000.recovery_status" |
6002 awk '/recovery_duration/ { print $2 }')
6003 (( $recovery_duration > $time_min + $wrap_up )) &&
6004 error "recovery_duration > recovery_time_hard + wrap up"
6005 completed_clients=$(do_facet $SINGLEMDS \
6006 "$LCTL get_param -n mdt.$FSNAME-MDT0000.recovery_status" |
6007 awk '/completed_clients/ { print $2 }')
6009 correct_clients="$MDSCOUNT/$((MDSCOUNT+1))"
6010 [ "$completed_clients" = "${correct_clients}" ] ||
6011 error "$completed_clients != $correct_clients"
6013 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
6014 umount_client $MOUNT1
6015 umount_client $MOUNT2
6021 run_test 84 "check recovery_hard_time"
6024 [[ $(lustre_version_code ost1) -ge $(version_code 2.7.55) ]] ||
6025 { skip "Need OST version at least 2.7.55" && return 0; }
6026 ##define OBD_FAIL_OSD_OST_EA_FID_SET 0x197
6027 do_facet ost1 "lctl set_param fail_loc=0x197"
6031 run_test 85 "osd_ost init: fail ea_fid_set"
6036 # ost1 has already registered to the MGS before the reformat.
6037 # So after reformatting it with option "-G", it could not be
6038 # mounted to the MGS. Cleanup the system for subsequent tests.
6043 local server_version=$(lustre_version_code $SINGLEMDS)
6044 [ "$(facet_fstype ost1)" = "zfs" ] &&
6045 skip "LU-6442: no such mkfs params for ZFS OSTs" && return
6046 [[ $server_version -ge $(version_code 2.7.56) ]] ||
6047 { skip "Need server version newer than 2.7.55"; return 0; }
6049 local OST_OPTS="$(mkfs_opts ost1 $(ostdevname 1)) \
6050 --reformat $(ostdevname 1) $(ostvdevname 1)"
6053 local OLDSIZE=$(do_facet ost1 "$DEBUGFS -c -R stats $(ostdevname 1)" |
6054 awk '/Flex block group size: / { print $NF; exit; }')
6057 if [[ ${!opts} != *mkfsoptions* ]]; then
6058 eval opts=\"${!opts} \
6059 --mkfsoptions='\\\"-O flex_bg -G $NEWSIZE\\\"'\"
6061 val=${!opts//--mkfsoptions=\\\"/ \
6062 --mkfsoptions=\\\"-O flex_bg -G $NEWSIZE }
6066 echo "params: $opts"
6068 trap cleanup_86 EXIT ERR
6071 add ost1 $opts || error "add ost1 failed with new params"
6073 local FOUNDSIZE=$(do_facet ost1 "$DEBUGFS -c -R stats $(ostdevname 1)" |
6074 awk '/Flex block group size: / { print $NF; exit; }')
6076 [[ $FOUNDSIZE == $NEWSIZE ]] ||
6077 error "Flex block group size: $FOUNDSIZE, expected: $NEWSIZE"
6081 run_test 86 "Replacing mkfs.lustre -G option"
6083 test_87() { #LU-6544
6084 [[ $(lustre_version_code $SINGLEMDS1) -ge $(version_code 2.9.51) ]] ||
6085 { skip "Need MDS version at least 2.9.51" && return; }
6086 [[ $(facet_fstype $SINGLEMDS) != ldiskfs ]] &&
6087 { skip "Only applicable to ldiskfs-based MDTs" && return; }
6088 [[ $OSTCOUNT -gt 59 ]] &&
6089 { skip "Ignore wide striping situation" && return; }
6091 local mdsdev=$(mdsdevname 1)
6092 local mdsvdev=$(mdsvdevname 1)
6093 local file=$DIR/$tfile
6094 local mntpt=$(facet_mntpt $SINGLEMDS)
6095 local used_xattr_blk=0
6096 local inode_size=${1:-1024}
6098 local xtest="trusted.test"
6102 local stripe_cnt=$(($OSTCOUNT + 2))
6104 #Please see ldiskfs_make_lustre() for MDT inode size calculation
6105 if [ $stripe_cnt -gt 16 ]; then
6108 left_size=$(expr $inode_size - \
6111 32 - 40 \* 3 - 32 \* 3 - $stripe_cnt \* 24 - 16 - 3 - \
6113 24 - 18 - $(expr length $tfile) - 16 - 4)
6114 if [ $left_size -le 0 ]; then
6115 echo "No space($left_size) is expected in inode."
6116 echo "Try 1-byte xattr instead to verify this."
6119 echo "Estimate: at most $left_size-byte space left in inode."
6125 add mds1 $(mkfs_opts mds1 ${mdsdev}) --stripe-count-hint=$stripe_cnt \
6126 --reformat $mdsdev $mdsvdev || error "add mds1 failed"
6127 start_mdt 1 > /dev/null || error "start mdt1 failed"
6128 for i in $(seq $OSTCOUNT); do
6129 start ost$i $(ostdevname $i) $OST_MOUNT_OPTS > /dev/null ||
6130 error "start ost$i failed"
6132 mount_client $MOUNT > /dev/null || error "mount client $MOUNT failed"
6133 check_mount || error "check client $MOUNT failed"
6136 $SETSTRIPE -E 1M -c 1 -E 64M -c 1 -E -1 -c -1 $file ||
6137 error "Create file with 3 components failed"
6138 i=$($GETSTRIPE -I 3 -c $file)
6139 if [ $i -ne $OSTCOUNT ]; then
6140 left_size=$(expr $left_size + $(expr $OSTCOUNT - $i) \* 24)
6141 echo -n "Since only $i out $OSTCOUNT OSTs are used, "
6142 echo -n "the expected left space is changed to "
6143 echo "$left_size bytes at most."
6145 value=$(generate_string $left_size)
6146 setfattr -n $xtest -v $value $file
6147 orig=$(get_xattr_value $xtest $file)
6148 [[ "$orig" != "$value" ]] && error "$xtest changed"
6150 #Verify if inode has some expected space left
6151 umount $MOUNT > /dev/null || error "umount $MOUNT failed"
6152 stop_mdt 1 > /dev/null || error "stop mdt1 failed"
6153 mount_ldiskfs $SINGLEMDS || error "mount -t ldiskfs $SINGLEMDS failed"
6155 do_facet $SINGLEMDS ls -sal $mntpt/ROOT/$tfile
6156 used_xattr_blk=$(do_facet $SINGLEMDS ls -s $mntpt/ROOT/$tfile |
6158 [[ $used_xattr_blk -eq 0 ]] &&
6159 error "Please check MDS inode size calculation: \
6160 more than $left_size-byte space left in inode."
6161 echo "Verified: at most $left_size-byte space left in inode."
6165 run_test 87 "check if MDT inode can hold EAs with N stripes properly"
6168 [ "$(facet_fstype mds1)" == "zfs" ] &&
6169 skip "LU-6662: no implementation for ZFS" && return
6173 add mds1 $(mkfs_opts mds1 $(mdsdevname 1)) \
6174 --reformat $(mdsdevname 1) || error "add mds1 failed"
6176 do_facet mds1 "$TUNEFS $(mdsdevname 1) |
6177 grep -e \".*opts:.*errors=remount-ro.*\"" ||
6178 error "default mount options is missing"
6180 add mds1 $(mkfs_opts mds1 $(mdsdevname 1)) \
6181 --mountfsoptions="user_xattr,errors=panic" \
6182 --reformat $(mdsdevname 1) || error "add mds1 failed"
6184 do_facet mds1 "$TUNEFS $(mdsdevname 1) |
6185 grep -e \".*opts:.*errors=panic.*\"" ||
6186 error "user can't override default mount options"
6188 run_test 88 "check the default mount options can be overridden"
6190 test_89() { # LU-7131
6191 [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.9.54) ]] ||
6192 { skip "Need MDT version at least 2.9.54" && return 0; }
6194 local key=failover.node
6195 local val1=192.0.2.254@tcp0 # Reserved IPs, see RFC 5735
6196 local val2=192.0.2.255@tcp0
6197 local mdsdev=$(mdsdevname 1)
6202 [ $(facet_fstype mds1) == zfs ] && import_zpool mds1
6203 # Check that parameters are added correctly
6204 echo "tunefs --param $key=$val1"
6205 do_facet mds "$TUNEFS --param $key=$val1 $mdsdev >/dev/null" ||
6206 error "tunefs --param $key=$val1 failed"
6207 params=$(do_facet mds $TUNEFS --dryrun $mdsdev) ||
6208 error "tunefs --dryrun failed"
6209 params=${params##*Parameters:}
6210 params=${params%%exiting*}
6211 [ $(echo $params | tr ' ' '\n' | grep -c $key=$val1) = "1" ] ||
6212 error "on-disk parameter not added correctly via tunefs"
6214 # Check that parameters replace existing instances when added
6215 echo "tunefs --param $key=$val2"
6216 do_facet mds "$TUNEFS --param $key=$val2 $mdsdev >/dev/null" ||
6217 error "tunefs --param $key=$val2 failed"
6218 params=$(do_facet mds $TUNEFS --dryrun $mdsdev) ||
6219 error "tunefs --dryrun failed"
6220 params=${params##*Parameters:}
6221 params=${params%%exiting*}
6222 [ $(echo $params | tr ' ' '\n' | grep -c $key=) = "1" ] ||
6223 error "on-disk parameter not replaced via tunefs"
6224 [ $(echo $params | tr ' ' '\n' | grep -c $key=$val2) = "1" ] ||
6225 error "on-disk parameter not replaced correctly via tunefs"
6227 # Check that a parameter is erased properly
6228 echo "tunefs --erase-param $key"
6229 do_facet mds "$TUNEFS --erase-param $key $mdsdev >/dev/null" ||
6230 error "tunefs --erase-param $key failed"
6231 params=$(do_facet mds $TUNEFS --dryrun $mdsdev) ||
6232 error "tunefs --dryrun failed"
6233 params=${params##*Parameters:}
6234 params=${params%%exiting*}
6235 [ $(echo $params | tr ' ' '\n' | grep -c $key=) = "0" ] ||
6236 error "on-disk parameter not erased correctly via tunefs"
6238 # Check that all the parameters are erased
6239 echo "tunefs --erase-params"
6240 do_facet mds "$TUNEFS --erase-params $mdsdev >/dev/null" ||
6241 error "tunefs --erase-params failed"
6242 params=$(do_facet mds $TUNEFS --dryrun $mdsdev) ||
6243 error "tunefs --dryrun failed"
6244 params=${params##*Parameters:}
6245 params=${params%%exiting*}
6247 error "all on-disk parameters not erased correctly via tunefs"
6249 # Check the order of options --erase-params and --param
6250 echo "tunefs --param $key=$val1 --erase-params"
6252 "$TUNEFS --param $key=$val1 --erase-params $mdsdev >/dev/null"||
6253 error "tunefs --param $key=$val1 --erase-params failed"
6254 params=$(do_facet mds $TUNEFS --dryrun $mdsdev) ||
6255 error "tunefs --dryrun failed"
6256 params=${params##*Parameters:}
6257 params=${params%%exiting*}
6258 [ $(echo $params | tr ' ' '\n') == "$key=$val1" ] ||
6259 error "on-disk param not added correctly with --erase-params"
6263 run_test 89 "check tunefs --param and --erase-param{s} options"
6266 # $2 (optional) value of max_mod_rpcs_in_flight to set
6267 check_max_mod_rpcs_in_flight() {
6275 idx=$(printf "%04x" $($LFS getdirstripe -i $dir))
6276 facet="mds$((0x$idx + 1))"
6278 if [ -z "$mmr" ]; then
6279 # get value of max_mod_rcps_in_flight
6280 mmr=$($LCTL get_param -n \
6281 mdc.$FSNAME-MDT$idx-mdc-*.max_mod_rpcs_in_flight) ||
6282 error "Unable to get max_mod_rpcs_in_flight"
6283 echo "max_mod_rcps_in_flight is $mmr"
6285 # set value of max_mod_rpcs_in_flight
6287 mdc.$FSNAME-MDT$idx-mdc-*.max_mod_rpcs_in_flight=$mmr ||
6288 error "Unable to set max_mod_rpcs_in_flight to $mmr"
6289 echo "max_mod_rpcs_in_flight set to $mmr"
6292 # create mmr+1 files
6293 echo "creating $((mmr + 1)) files ..."
6295 for i in $(seq $((mmr + 1))); do
6301 # consumes mmr-1 modify RPC slots
6302 #define OBD_FAIL_MDS_REINT_MULTI_NET 0x159
6303 # drop requests on MDT so that RPC slots are consumed
6304 # during all the request resend interval
6305 do_facet $facet "$LCTL set_param fail_loc=0x159"
6306 echo "launch $((mmr - 1)) chmod in parallel ..."
6307 for i in $(seq $((mmr - 1))); do
6308 chmod 0600 $dir/file-$i &
6312 # send one additional modify RPC
6313 do_facet $facet "$LCTL set_param fail_loc=0"
6314 echo "launch 1 additional chmod in parallel ..."
6315 chmod 0600 $dir/file-$mmr &
6318 # check this additional modify RPC get a modify RPC slot
6319 # and succeed its operation
6320 checkstat -vp 0600 $dir/file-$mmr ||
6321 error "Unable to send $mmr modify RPCs in parallel"
6326 # consumes mmr modify RPC slots
6327 #define OBD_FAIL_MDS_REINT_MULTI_NET 0x159
6328 # drop requests on MDT so that RPC slots are consumed
6329 # during all the request resend interval
6330 do_facet $facet "$LCTL set_param fail_loc=0x159"
6331 echo "launch $mmr chmod in parallel ..."
6332 for i in $(seq $mmr); do
6333 chmod 0666 $dir/file-$i &
6337 # send one additional modify RPC
6338 do_facet $facet "$LCTL set_param fail_loc=0"
6339 echo "launch 1 additional chmod in parallel ..."
6340 chmod 0666 $dir/file-$((mmr + 1)) &
6343 # check this additional modify RPC blocked getting a modify RPC slot
6344 checkstat -vp 0644 $dir/file-$((mmr + 1)) ||
6345 error "Unexpectedly send $(($mmr + 1)) modify RPCs in parallel"
6351 if ! combined_mgs_mds ; then
6356 [[ $($LCTL get_param mdc.*.import |
6357 grep "connect_flags:.*multi_mod_rpc") ]] ||
6358 { skip "Need MDC with 'multi_mod_rpcs' feature"; return 0; }
6360 # check default value
6361 $LFS mkdir -c1 $DIR/$tdir || error "mkdir $DIR/$tdir failed"
6362 check_max_mod_rpcs_in_flight $DIR/$tdir
6366 run_test 90a "check max_mod_rpcs_in_flight is enforced"
6376 [[ $($LCTL get_param mdc.*.import |
6377 grep "connect_flags:.*multi_mod_rpc") ]] ||
6378 { skip "Need MDC with 'multi_mod_rpcs' feature"; return 0; }
6381 # update max_mod_rpcs_in_flight
6382 $LFS mkdir -c1 $DIR/${tdir}1 || error "mkdir $DIR/${tdir}1 failed"
6383 check_max_mod_rpcs_in_flight $DIR/${tdir}1 1
6386 # check client is able to send multiple modify RPCs in paralell
6387 tmp=$($LCTL get_param -n mdc.$FSNAME-MDT*-mdc-*.import |
6388 grep -c "multi_mod_rpcs")
6389 if [ "$tmp" -ne $MDSCOUNT ]; then
6390 echo "Client not able to send multiple modify RPCs in parallel"
6395 # update max_mod_rpcs_in_flight
6396 $LFS mkdir -c1 $DIR/${tdir}2 || error "mkdir $DIR/${tdir}2 failed"
6397 check_max_mod_rpcs_in_flight $DIR/${tdir}2 5
6400 $LFS mkdir -c1 $DIR/${tdir}3 || error "mkdir $DIR/${tdir}3 failed"
6401 idx=$(printf "%04x" $($LFS getdirstripe -i $DIR/${tdir}3))
6402 facet="mds$((0x$idx + 1))"
6404 # save MDT max_mod_rpcs_per_client
6405 mmrpc=$(do_facet $facet \
6406 cat /sys/module/mdt/parameters/max_mod_rpcs_per_client)
6408 # update max_mod_rpcs_in_flight
6409 umount_client $MOUNT
6411 "echo 16 > /sys/module/mdt/parameters/max_mod_rpcs_per_client"
6413 $LCTL set_param mdc.$FSNAME-MDT$idx-mdc-*.max_rpcs_in_flight=17
6414 check_max_mod_rpcs_in_flight $DIR/${tdir}3 16
6416 # restore MDT max_mod_rpcs_per_client initial value
6418 "echo $mmrpc > /sys/module/mdt/parameters/max_mod_rpcs_per_client"
6420 rm -rf $DIR/${tdir}?
6423 run_test 90b "check max_mod_rpcs_in_flight is enforced after update"
6432 [[ $($LCTL get_param mdc.*.import |
6433 grep "connect_flags:.*multi_mod_rpc") ]] ||
6434 { skip "Need MDC with 'multi_mod_rpcs' feature"; return 0; }
6436 # check client is able to send multiple modify RPCs in paralell
6437 tmp=$($LCTL get_param -n mdc.$FSNAME-MDT*-mdc-*.import |
6438 grep -c "multi_mod_rpcs")
6439 if [ "$tmp" -ne $MDSCOUNT ]; then
6440 skip "Client not able to send multiple modify RPCs in parallel"
6445 # get max_rpcs_in_flight value
6446 mrif=$($LCTL get_param -n mdc.$FSNAME-MDT0000-mdc-*.max_rpcs_in_flight)
6447 echo "max_rpcs_in_flight is $mrif"
6449 # get MDT max_mod_rpcs_per_client
6450 mmrpc=$(do_facet mds1 \
6451 cat /sys/module/mdt/parameters/max_mod_rpcs_per_client)
6452 echo "max_mod_rpcs_per_client is $mmrpc"
6455 # attempt to set max_mod_rpcs_in_flight to max_rpcs_in_flight value
6456 # prerequisite: set max_mod_rpcs_per_client to max_rpcs_in_flight value
6457 umount_client $MOUNT
6459 "echo $mrif > /sys/module/mdt/parameters/max_mod_rpcs_per_client"
6463 mdc.$FSNAME-MDT0000-mdc-*.max_mod_rpcs_in_flight=$mrif &&
6464 error "set max_mod_rpcs_in_flight to $mrif should fail"
6466 umount_client $MOUNT
6468 "echo $mmrpc > /sys/module/mdt/parameters/max_mod_rpcs_per_client"
6472 # attempt to set max_mod_rpcs_in_flight to max_mod_rpcs_per_client+1
6473 # prerequisite: set max_rpcs_in_flight to max_mod_rpcs_per_client+2
6475 mdc.$FSNAME-MDT0000-mdc-*.max_rpcs_in_flight=$((mmrpc + 2))
6478 mdc.$FSNAME-MDT0000-mdc-*.max_mod_rpcs_in_flight=$((mmrpc + 1)) &&
6479 error "set max_mod_rpcs_in_flight to $((mmrpc + 1)) should fail"
6483 run_test 90c "check max_mod_rpcs_in_flight update limits"
6494 [[ $($LCTL get_param mdc.*.import |
6495 grep "connect_flags:.*multi_mod_rpc") ]] ||
6496 { skip "Need MDC with 'multi_mod_rpcs' feature"; return 0; }
6498 $LFS mkdir -c1 $DIR/$tdir || error "mkdir $DIR/$tdir failed"
6499 idx=$(printf "%04x" $($LFS getdirstripe -i $DIR/$tdir))
6500 facet="mds$((0x$idx + 1))"
6502 # check client version supports multislots
6503 tmp=$($LCTL get_param -N \
6504 mdc.$FSNAME-MDT$idx-mdc-*.max_mod_rpcs_in_flight)
6505 if [ -z "$tmp" ]; then
6506 skip "Client does not support multiple modify RPCs in flight"
6511 # get current value of max_mod_rcps_in_flight
6512 mmr=$($LCTL get_param -n \
6513 mdc.$FSNAME-MDT$idx-mdc-*.max_mod_rpcs_in_flight)
6514 echo "max_mod_rcps_in_flight is $mmr"
6517 echo "creating $mmr files ..."
6519 for i in $(seq $mmr); do
6520 touch $DIR/$tdir/file-$i
6523 # prepare for close RPC
6524 multiop_bg_pause $DIR/$tdir/file-close O_c
6527 # consumes mmr modify RPC slots
6528 #define OBD_FAIL_MDS_REINT_MULTI_NET 0x159
6529 # drop requests on MDT so that RPC slots are consumed
6530 # during all the request resend interval
6531 do_facet $facet "$LCTL set_param fail_loc=0x159"
6532 echo "launch $mmr chmod in parallel ..."
6533 for i in $(seq $mmr); do
6534 chmod 0600 $DIR/$tdir/file-$i &
6537 # send one additional close RPC
6538 do_facet $facet "$LCTL set_param fail_loc=0"
6539 echo "launch 1 additional close in parallel ..."
6541 cancel_lru_locks mdc
6544 # check this additional close RPC get a modify RPC slot
6545 # and multiop process completed
6546 [ -d /proc/$pid ] &&
6547 error "Unable to send the additional close RPC in parallel"
6552 run_test 90d "check one close RPC is allowed above max_mod_rpcs_in_flight"
6554 check_uuid_on_ost() {
6556 do_facet ost1 "$LCTL get_param obdfilter.${FSNAME}*.exports.'$nid'.uuid"
6559 check_uuid_on_mdt() {
6561 do_facet $SINGLEMDS "$LCTL get_param mdt.${FSNAME}*.exports.'$nid'.uuid"
6569 [[ $(lustre_version_code ost1) -ge $(version_code 2.7.63) ]] ||
6570 { skip "Need OST version at least 2.7.63" && return 0; }
6571 [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.7.63) ]] ||
6572 { skip "Need MDT version at least 2.7.63" && return 0; }
6574 start_mds || error "MDS start failed"
6575 start_ost || error "unable to start OST"
6576 mount_client $MOUNT || error "client start failed"
6577 check_mount || error "check_mount failed"
6580 nid=$($LCTL list_nids | head -1 | sed "s/\./\\\./g")
6584 uuid=$(get_client_uuid $MOUNT)
6586 echo "list nids on mdt:"
6587 do_facet $SINGLEMDS "$LCTL list_param mdt.${FSNAME}*.exports.*"
6588 echo "uuid from $nid:"
6589 do_facet $SINGLEMDS "$LCTL get_param mdt.${FSNAME}*.exports.'$nid'.uuid"
6591 found=$(check_uuid_on_mdt $nid | grep $uuid)
6592 [ -z "$found" ] && error "can't find $uuid $nid on MDT"
6593 found=$(check_uuid_on_ost $nid | grep $uuid)
6594 [ -z "$found" ] && error "can't find $uuid $nid on OST"
6596 # umount the client so it won't reconnect
6597 manual_umount_client --force || error "failed to umount $?"
6598 # shouldn't disappear on MDS after forced umount
6599 found=$(check_uuid_on_mdt $nid | grep $uuid)
6600 [ -z "$found" ] && error "can't find $uuid $nid"
6603 do_facet $SINGLEMDS \
6604 "$LCTL set_param -n mdt.${mds1_svc}.evict_client nid:$nid"
6606 found=$(check_uuid_on_mdt $nid | grep $uuid)
6607 [ -n "$found" ] && error "found $uuid $nid on MDT"
6608 found=$(check_uuid_on_ost $nid | grep $uuid)
6609 [ -n "$found" ] && error "found $uuid $nid on OST"
6611 # check it didn't reconnect (being umounted)
6612 sleep $((TIMEOUT+1))
6613 found=$(check_uuid_on_mdt $nid | grep $uuid)
6614 [ -n "$found" ] && error "found $uuid $nid on MDT"
6615 found=$(check_uuid_on_ost $nid | grep $uuid)
6616 [ -n "$found" ] && error "found $uuid $nid on OST"
6620 run_test 91 "evict-by-nid support"
6622 generate_ldev_conf() {
6623 # generate an ldev.conf file
6624 local ldevconfpath=$1
6626 local fsldevformat=""
6629 fstype=$(facet_fstype mgs)
6630 if [ "$fstype" == "zfs" ]; then
6631 fsldevformat="$fstype:"
6636 printf "%s\t-\t%s-MGS0000\t%s%s\n" \
6640 $(mgsdevname) > $ldevconfpath
6642 local mdsfo_host=$mdsfailover_HOST;
6643 if [ -z "$mdsfo_host" ]; then
6647 for num in $(seq $MDSCOUNT); do
6648 fstype=$(facet_fstype mds$num)
6649 if [ "$fstype" == "zfs" ]; then
6650 fsldevformat="$fstype:"
6655 printf "%s\t%s\t%s-MDT%04d\t%s%s\n" \
6661 $(mdsdevname $num) >> $ldevconfpath
6664 local ostfo_host=$ostfailover_HOST;
6665 if [ -z "$ostfo_host" ]; then
6669 for num in $(seq $OSTCOUNT); do
6670 fstype=$(facet_fstype ost$num)
6671 if [ "$fstype" == "zfs" ]; then
6672 fsldevformat="$fstype:"
6677 printf "%s\t%s\t%s-OST%04d\t%s%s\n" \
6683 $(ostdevname $num) >> $ldevconfpath
6686 echo "----- $ldevconfpath -----"
6688 echo "--- END $ldevconfpath ---"
6693 # generate a nids file (mapping between hostname to nid)
6694 # looks like we only have the MGS nid available to us
6695 # so just echo that to a file
6697 echo -e "${mgs_HOST}\t${MGSNID}" > $nidspath
6699 echo "----- $nidspath -----"
6701 echo "--- END $nidspath ---"
6704 compare_ldev_output() {
6708 sort $expected_output -o $expected_output
6709 sort $ldev_output -o $ldev_output
6711 echo "-- START OF LDEV OUTPUT --"
6713 echo "--- END OF LDEV OUTPUT ---"
6715 echo "-- START OF EXPECTED OUTPUT --"
6716 cat $expected_output
6717 echo "--- END OF EXPECTED OUTPUT ---"
6719 diff $expected_output $ldev_output
6724 if [ -z "$LDEV" ]; then
6725 error "ldev is missing!"
6728 local LDEVCONFPATH=$TMP/ldev.conf
6729 local NIDSPATH=$TMP/nids
6731 echo "Host is $(hostname)"
6733 generate_ldev_conf $LDEVCONFPATH
6734 generate_nids $NIDSPATH
6736 # echo the mgs nid and compare it to environment variable MGSNID
6737 # also, ldev.conf and nids is a server side thing, use the OSS
6740 output=$($LDEV -c $LDEVCONFPATH -H $ost_HOST -n $NIDSPATH echo %m)
6742 echo "-- START OF LDEV OUTPUT --"
6744 echo "--- END OF LDEV OUTPUT ---"
6746 # ldev failed, error
6747 if [ $? -ne 0 ]; then
6748 rm $LDEVCONFPATH $NIDSPATH
6749 error "ldev failed to execute!"
6752 # need to process multiple lines because of combined MGS and MDS
6753 echo -e $output | awk '{ print $2 }' | while read -r line ; do
6754 if [ "$line" != "$MGSNID" ]; then
6755 rm $LDEVCONFPATH $NIDSPATH
6756 error "ldev failed mgs nid '$line', expected '$MGSNID'"
6760 rm $LDEVCONFPATH $NIDSPATH
6762 run_test 92 "ldev returns MGS NID correctly in command substitution"
6765 [ $MDSCOUNT -lt 3 ] && skip "needs >= 3 MDTs" && return
6768 #start mgs or mgs/mdt0
6769 if ! combined_mgs_mds ; then
6776 start_ost || error "OST0 start fail"
6778 #define OBD_FAIL_MGS_WRITE_TARGET_DELAY 0x90e
6779 do_facet mgs "$LCTL set_param fail_val = 10 fail_loc=0x8000090e"
6780 for num in $(seq 2 $MDSCOUNT); do
6784 mount_client $MOUNT || error "mount client fails"
6785 wait_osc_import_state mds ost FULL
6786 wait_osc_import_state client ost FULL
6787 check_mount || error "check_mount failed"
6789 cleanup || error "cleanup failed with $?"
6791 run_test 93 "register mulitple MDT at the same time"
6794 if [ -z "$LDEV" ]; then
6795 error "ldev is missing!"
6798 local LDEVCONFPATH=$TMP/ldev.conf
6799 local NIDSPATH=$TMP/nids
6801 generate_ldev_conf $LDEVCONFPATH
6802 generate_nids $NIDSPATH
6804 local LDEV_OUTPUT=$TMP/ldev-output.txt
6805 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -F $FSNAME > $LDEV_OUTPUT
6807 # ldev failed, error
6808 if [ $? -ne 0 ]; then
6809 rm $LDEVCONFPATH $NIDSPATH $LDEV_OUTPUT
6810 error "ldev failed to execute!"
6814 local EXPECTED_OUTPUT=$TMP/ldev-expected.txt
6816 printf "%s-MGS0000\n" $FSNAME > $EXPECTED_OUTPUT
6818 for num in $(seq $MDSCOUNT); do
6819 printf "%s-MDT%04d\n" $FSNAME $num >> $EXPECTED_OUTPUT
6822 for num in $(seq $OSTCOUNT); do
6823 printf "%s-OST%04d\n" $FSNAME $num >> $EXPECTED_OUTPUT
6826 compare_ldev_output $LDEV_OUTPUT $EXPECTED_OUTPUT
6828 if [ $? -ne 0 ]; then
6829 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
6830 error "ldev failed to produce the correct hostlist!"
6833 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
6835 run_test 94 "ldev outputs correct labels for file system name query"
6838 if [ -z "$LDEV" ]; then
6839 error "ldev is missing!"
6842 local LDEVCONFPATH=$TMP/ldev.conf
6843 local NIDSPATH=$TMP/nids
6845 generate_ldev_conf $LDEVCONFPATH
6846 generate_nids $NIDSPATH
6850 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -F $FSNAME &>/dev/null
6851 if [ $? -ne 0 ]; then
6852 rm $LDEVCONFPATH $NIDSPATH
6853 error "ldev label filtering w/ -F failed!"
6857 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -l &>/dev/null
6858 if [ $? -ne 0 ]; then
6859 rm $LDEVCONFPATH $NIDSPATH
6860 error "ldev label filtering w/ -l failed!"
6864 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -f &>/dev/null
6865 if [ $? -ne 0 ]; then
6866 rm $LDEVCONFPATH $NIDSPATH
6867 error "ldev label filtering w/ -f failed!"
6871 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -a &>/dev/null
6872 if [ $? -ne 0 ]; then
6873 rm $LDEVCONFPATH $NIDSPATH
6874 error "ldev label filtering w/ -a failed!"
6879 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -a -F $FSNAME &>/dev/null
6880 if [ $? -eq 0 ]; then
6881 rm $LDEVCONFPATH $NIDSPATH
6882 error "ldev label filtering w/ -a and -F incorrectly succeeded"
6886 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -a -f &>/dev/null
6887 if [ $? -eq 0 ]; then
6888 rm $LDEVCONFPATH $NIDSPATH
6889 error "ldev label filtering w/ -a and -f incorrectly succeeded"
6893 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -a -l &>/dev/null
6894 if [ $? -eq 0 ]; then
6895 rm $LDEVCONFPATH $NIDSPATH
6896 error "ldev label filtering w/ -a and -l incorrectly succeeded"
6900 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -f -l &>/dev/null
6901 if [ $? -eq 0 ]; then
6902 rm $LDEVCONFPATH $NIDSPATH
6903 error "ldev label filtering w/ -f and -l incorrectly succeeded"
6907 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -F $FSNAME -l &>/dev/null
6908 if [ $? -eq 0 ]; then
6909 rm $LDEVCONFPATH $NIDSPATH
6910 error "ldev label filtering w/ -F and -l incorrectly succeeded"
6913 # file sys & foreign
6914 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -F $FSNAME -f &>/dev/null
6915 if [ $? -eq 0 ]; then
6916 rm $LDEVCONFPATH $NIDSPATH
6917 error "ldev label filtering w/ -F and -f incorrectly succeeded"
6920 rm $LDEVCONFPATH $NIDSPATH
6922 run_test 95 "ldev should only allow one label filter"
6925 if [ -z "$LDEV" ]; then
6926 error "ldev is missing!"
6929 local LDEVCONFPATH=$TMP/ldev.conf
6930 local NIDSPATH=$TMP/nids
6932 generate_ldev_conf $LDEVCONFPATH
6933 generate_nids $NIDSPATH
6935 local LDEV_OUTPUT=$TMP/ldev-output.txt
6936 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -H $mgs_HOST \
6938 awk '{print $2}' > $LDEV_OUTPUT
6940 # ldev failed, error
6941 if [ $? -ne 0 ]; then
6942 rm $LDEVCONFPATH $NIDSPATH $LDEV_OUTPUT
6943 error "ldev failed to execute!"
6947 local EXPECTED_OUTPUT=$TMP/ldev-expected-output.txt
6949 echo "$mgs_HOST-$(facet_fstype mgs)" > $EXPECTED_OUTPUT
6951 if [ "$mgs_HOST" == "$mds_HOST" ]; then
6952 for num in $(seq $MDSCOUNT); do
6953 echo "$mds_HOST-$(facet_fstype mds$num)" \
6958 if [ "$mgs_HOST" == "$ost_HOST" ]; then
6959 for num in $(seq $OSTCOUNT); do
6960 echo "$ost_HOST-$(facet_fstype ost$num)" \
6965 compare_ldev_output $LDEV_OUTPUT $EXPECTED_OUTPUT
6967 if [ $? -ne 0 ]; then
6968 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
6969 error "ldev failed to produce the correct output!"
6972 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
6974 run_test 96 "ldev returns hostname and backend fs correctly in command sub"
6977 if [ -z "$LDEV" ]; then
6978 error "ldev is missing!"
6981 local LDEVCONFPATH=$TMP/ldev.conf
6982 local NIDSPATH=$TMP/nids
6984 generate_ldev_conf $LDEVCONFPATH
6985 generate_nids $NIDSPATH
6987 local LDEV_OUTPUT=$TMP/ldev-output.txt
6988 local EXPECTED_OUTPUT=$TMP/ldev-expected-output.txt
6990 echo -e "\nMDT role"
6991 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -F $FSNAME -R mdt > $LDEV_OUTPUT
6993 if [ $? -ne 0 ]; then
6994 rm $LDEVCONFPATH $NIDSPATH $LDEV_OUTPUT
6995 error "ldev failed to execute for mdt role!"
6998 for num in $(seq $MDSCOUNT); do
6999 printf "%s-MDT%04d\n" $FSNAME $num >> $EXPECTED_OUTPUT
7002 compare_ldev_output $LDEV_OUTPUT $EXPECTED_OUTPUT
7004 if [ $? -ne 0 ]; then
7005 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
7006 error "ldev failed to produce the correct output for mdt role!"
7009 echo -e "\nOST role"
7010 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -F $FSNAME -R ost > $LDEV_OUTPUT
7012 if [ $? -ne 0 ]; then
7013 rm $LDEVCONFPATH $NIDSPATH $LDEV_OUTPUT $EXPECTED_OUTPUT
7014 error "ldev failed to execute for ost role!"
7018 for num in $(seq $OSTCOUNT); do
7019 printf "%s-OST%04d\n" $FSNAME $num >> $EXPECTED_OUTPUT
7022 compare_ldev_output $LDEV_OUTPUT $EXPECTED_OUTPUT
7024 if [ $? -ne 0 ]; then
7025 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
7026 error "ldev failed to produce the correct output for ost role!"
7029 echo -e "\nMGS role"
7030 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -F $FSNAME -R mgs > $LDEV_OUTPUT
7032 if [ $? -ne 0 ]; then
7033 rm $LDEVCONFPATH $NIDSPATH $LDEV_OUTPUT $EXPECTED_OUTPUT
7034 error "ldev failed to execute for mgs role!"
7037 printf "%s-MGS0000\n" $FSNAME > $EXPECTED_OUTPUT
7039 compare_ldev_output $LDEV_OUTPUT $EXPECTED_OUTPUT
7041 if [ $? -ne 0 ]; then
7042 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
7043 error "ldev failed to produce the correct output for mgs role!"
7046 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
7048 run_test 97 "ldev returns correct ouput when querying based on role"
7053 local temp=$MDS_MOUNT_OPTS
7056 check_mount || error "mount failed"
7057 mountopt="user_xattr"
7058 for ((x = 1; x <= 400; x++)); do
7059 mountopt="$mountopt,user_xattr"
7061 remount_client $mountopt $MOUNT 2>&1 | grep "too long" ||
7062 error "Buffer overflow check failed"
7063 cleanup || error "cleanup failed"
7065 run_test 98 "Buffer-overflow check while parsing mount_opts"
7069 [[ $(facet_fstype ost1) != ldiskfs ]] &&
7070 { skip "Only applicable to ldiskfs-based OSTs" && return; }
7071 [[ $(lustre_version_code ost1) -ge $(version_code 2.8.57) ]] ||
7072 { skip "Need OST version at least 2.8.57" && return 0; }
7074 local ost_opts="$(mkfs_opts ost1 $(ostdevname 1)) \
7075 --reformat $(ostdevname 1) $(ostvdevname 1)"
7076 do_facet ost1 $DEBUGFS -c -R stats `ostdevname 1` | grep "meta_bg" &&
7077 skip "meta_bg already set" && return
7080 if [[ ${!opts} != *mkfsoptions* ]]; then
7081 eval opts=\"${!opts} \
7082 --mkfsoptions='\\\"-O ^resize_inode,meta_bg\\\"'\"
7084 local val=${!opts//--mkfsoptions=\\\"/ \
7085 --mkfsoptions=\\\"-O ^resize_inode,meta_bg }
7089 echo "params: $opts"
7091 add ost1 $opts || error "add ost1 failed with new params"
7093 do_facet ost1 $DEBUGFS -c -R stats `ostdevname 1` | grep "meta_bg" ||
7094 error "meta_bg is not set"
7098 run_test 99 "Adding meta_bg option"
7102 start_mds || error "MDS start failed"
7103 start_ost || error "unable to start OST"
7104 mount_client $MOUNT || error "client start failed"
7105 check_mount || error "check_mount failed"
7114 do_facet mgs 'lshowmount -v' | awk 'BEGIN {NR == 0; rc=1} /MGS:/ {rc=0}
7115 END {exit rc}' || error "lshowmount have no output MGS"
7117 do_facet mds1 'lshowmount -v' | awk 'BEGIN {NR == 2; rc=1} /-MDT0000:/
7118 {rc=0} END {exit rc}' || error "lshowmount have no output MDT0"
7120 do_facet ost1 'lshowmount -v' | awk 'BEGIN {NR == 4; rc=1} /-OST0000:/
7121 {rc=0} END {exit rc}' || error "lshowmount have no output OST0"
7123 cleanup || error "cleanup failed with $?"
7125 run_test 100 "check lshowmount lists MGS, MDT, OST and 0@lo"
7128 local createmany_oid
7129 local dev=$FSNAME-OST0000-osc-MDT0000
7132 createmany -o $DIR1/$tfile-%d 50000 &
7134 # MDT->OST reconnection causes MDT<->OST last_id synchornisation
7135 # via osp_precreate_cleanup_orphans.
7136 for ((i = 0; i < 100; i++)); do
7137 for ((k = 0; k < 10; k++)); do
7138 do_facet $SINGLEMDS "$LCTL --device $dev deactivate;" \
7139 "$LCTL --device $dev activate"
7142 ls -asl $MOUNT | grep '???' &&
7143 (kill -9 $createmany_oid &>/dev/null; \
7144 error "File hasn't object on OST")
7146 kill -s 0 $createmany_oid || break
7148 wait $createmany_oid
7151 run_test 101 "Race MDT->OST reconnection with create"
7154 cleanup || error "cleanup failed with $?"
7156 local mds1dev=$(mdsdevname 1)
7157 local mds1mnt=$(facet_mntpt mds1)
7158 local mds1fstype=$(facet_fstype mds1)
7159 local mds1opts=$MDS_MOUNT_OPTS
7161 if [ $mds1fstype == ldiskfs ] &&
7162 ! do_facet mds1 test -b $mds1dev; then
7163 mds1opts=$(csa_add "$mds1opts" -o loop)
7165 if [[ $mds1fstype == zfs ]]; then
7166 import_zpool mds1 || return ${PIPESTATUS[0]}
7169 # unload all and only load libcfs to allow fail_loc setting
7170 do_facet mds1 lustre_rmmod || error "unable to unload modules"
7171 do_facet mds1 modprobe libcfs || error "libcfs not loaded"
7172 do_facet mds1 lsmod \| grep libcfs || error "libcfs not loaded"
7174 #define OBD_FAIL_OBDCLASS_MODULE_LOAD 0x60a
7175 do_facet mds1 "$LCTL set_param fail_loc=0x8000060a"
7177 do_facet mds1 $MOUNT_CMD $mds1dev $mds1mnt $mds1opts &&
7178 error "mdt start must fail"
7179 do_facet mds1 lsmod \| grep obdclass && error "obdclass must not load"
7181 do_facet mds1 "$LCTL set_param fail_loc=0x0"
7183 do_facet mds1 $MOUNT_CMD $mds1dev $mds1mnt $mds1opts ||
7184 error "mdt start must not fail"
7186 cleanup || error "cleanup failed with $?"
7188 run_test 102 "obdclass module cleanup upon error"
7193 echo "rename $FSNAME to $newname"
7195 if [ ! combined_mgs_mds ]; then
7196 local facet=$(mgsdevname)
7199 "$TUNEFS --fsname=$newname --rename=$FSNAME -v $facet"||
7200 error "(7) Fail to rename MGS"
7201 if [ "$(facet_fstype $facet)" = "zfs" ]; then
7202 reimport_zpool mgs $newname-mgs
7206 for num in $(seq $MDSCOUNT); do
7207 local facet=$(mdsdevname $num)
7209 do_facet mds${num} \
7210 "$TUNEFS --fsname=$newname --rename=$FSNAME -v $facet"||
7211 error "(8) Fail to rename MDT $num"
7212 if [ "$(facet_fstype $facet)" = "zfs" ]; then
7213 reimport_zpool mds${num} $newname-mdt${num}
7217 for num in $(seq $OSTCOUNT); do
7218 local facet=$(ostdevname $num)
7220 do_facet ost${num} \
7221 "$TUNEFS --fsname=$newname --rename=$FSNAME -v $facet"||
7222 error "(9) Fail to rename OST $num"
7223 if [ "$(facet_fstype $facet)" = "zfs" ]; then
7224 reimport_zpool ost${num} $newname-ost${num}
7229 test_103_set_pool() {
7233 do_facet mgs $LCTL pool_add $FSNAME.$pname ${FSNAME}-$ost_x ||
7234 error "Fail to add $ost_x to $FSNAME.$pname"
7235 wait_update $HOSTNAME \
7236 "lctl get_param -n lov.$FSNAME-clilov-*.pools.$pname |
7237 grep $ost_x" "$FSNAME-${ost_x}_UUID" ||
7238 error "$ost_x is NOT in pool $FSNAME.$pname"
7241 test_103_check_pool() {
7242 local save_fsname=$1
7245 stat $DIR/$tdir/test-framework.sh ||
7246 error "($errno) Fail to stat"
7247 do_facet mgs $LCTL pool_list $FSNAME.pool1 ||
7248 error "($errno) Fail to list $FSNAME.pool1"
7249 do_facet mgs $LCTL pool_list $FSNAME.$save_fsname ||
7250 error "($errno) Fail to list $FSNAME.$save_fsname"
7251 do_facet mgs $LCTL pool_list $FSNAME.$save_fsname |
7252 grep ${FSNAME}-OST0000 ||
7253 error "($errno) List $FSNAME.$save_fsname is invalid"
7255 local pname=$($LFS getstripe --pool $DIR/$tdir/d0)
7256 [ "$pname" = "$save_fsname" ] ||
7257 error "($errno) Unexpected pool name $pname"
7261 check_mount_and_prep
7263 mkdir $DIR/$tdir || error "(1) Fail to mkdir $DIR/$tdir"
7264 cp $LUSTRE/tests/test-framework.sh $DIR/$tdir ||
7265 error "(2) Fail to copy test-framework.sh"
7267 do_facet mgs $LCTL pool_new $FSNAME.pool1 ||
7268 error "(3) Fail to create $FSNAME.pool1"
7269 # name the pool name as the fsname
7270 do_facet mgs $LCTL pool_new $FSNAME.$FSNAME ||
7271 error "(4) Fail to create $FSNAME.$FSNAME"
7273 test_103_set_pool $FSNAME OST0000
7275 $SETSTRIPE -p $FSNAME $DIR/$tdir/d0 ||
7276 error "(6) Fail to setstripe on $DIR/$tdir/d0"
7281 test_renamefs mylustre
7283 local save_fsname=$FSNAME
7287 test_103_check_pool $save_fsname 7
7289 if [ $OSTCOUNT -ge 2 ]; then
7290 test_103_set_pool $save_fsname OST0001
7293 $SETSTRIPE -p $save_fsname $DIR/$tdir/f0 ||
7294 error "(16) Fail to setstripe on $DIR/$tdir/f0"
7303 test_103_check_pool $save_fsname 17
7307 test_renamefs $save_fsname
7313 run_test 103 "rename filesystem name"
7315 if ! combined_mgs_mds ; then
7321 # restore the values of MDSSIZE and OSTSIZE
7322 MDSSIZE=$STORED_MDSSIZE
7323 OSTSIZE=$STORED_OSTSIZE