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 for num in $(seq $MDSCOUNT); do
1211 local mdtosc=$(get_mdtosc_proc_path mds${num} $FSNAME-OST0001)
1212 local MPROC="osc.$mdtosc.active"
1217 RESULT=$(do_facet mds${num} "$LCTL get_param -n $MPROC")
1218 [ ${PIPESTATUS[0]} = 0 ] || error "Can't read $MPROC"
1219 if [ $RESULT -eq $DEAC ]; then
1220 echo -n "MDT deactivated also after"
1221 echo "$WAIT sec (got $RESULT)"
1225 if [ $WAIT -eq $MAX ]; then
1226 error "MDT active: wanted $DEAC got $RESULT"
1228 echo "Waiting $(($MAX - $WAIT))secs for MDT deactivated"
1231 # test new client starts deactivated
1232 umount_client $MOUNT || error "umount_client $MOUNT failed"
1233 mount_client $MOUNT || error "mount_client $MOUNT failed"
1234 RESULT=$($LCTL get_param -n $PROC_UUID | grep DEACTIV | grep NEW)
1235 if [ -z "$RESULT" ]; then
1236 error "New client start active: $(lctl get_param -n $PROC_UUID)"
1238 echo "New client success: got $RESULT"
1241 # make sure it reactivates
1242 set_conf_param_and_check client \
1243 "$LCTL get_param -n $PROC_ACT" "$PARAM" $ACTV ||
1244 error "lctl get_param $PROC_ACT $PARAM $ACTV failed"
1246 umount_client $MOUNT
1247 stop_ost2 || error "Unable to stop OST2"
1248 cleanup_nocli || error "cleanup_nocli failed with $?"
1249 #writeconf to remove all ost2 traces for subsequent tests
1250 writeconf_or_reformat
1252 run_test 29 "permanently remove an OST"
1257 echo Big config llog
1258 TEST="$LCTL get_param -n llite.$FSNAME-*.max_read_ahead_whole_mb"
1260 LIST=(1 2 3 4 5 4 3 2 1 2 3 4 5 4 3 2 1 2 3 4 5)
1261 for i in ${LIST[@]}; do
1262 set_conf_param_and_check client "$TEST" \
1263 "$FSNAME.llite.max_read_ahead_whole_mb" $i ||
1264 error "Set $FSNAME.llite.max_read_ahead_whole_mb failed"
1266 # make sure client restart still works
1267 umount_client $MOUNT
1268 mount_client $MOUNT || error "mount_client $MOUNT failed"
1269 [ "$($TEST)" -ne "$i" ] &&
1270 error "Param didn't stick across restart $($TEST) != $i"
1273 echo Erase parameter setting
1274 do_facet mgs "$LCTL conf_param \
1275 -d $FSNAME.llite.max_read_ahead_whole_mb" ||
1276 error "Erase param $FSNAME.llite.max_read_ahead_whole_mb failed"
1277 umount_client $MOUNT
1278 mount_client $MOUNT || error "mount_client $MOUNT failed"
1280 echo "deleted (default) value=$FINAL, orig=$ORIG"
1281 # assumes this parameter started at the default value
1282 [ "$FINAL" -eq "$ORIG" ] || fail "Deleted value=$FINAL, orig=$ORIG"
1284 cleanup || error "cleanup failed with rc $?"
1286 run_test 30a "Big config llog and conf_param deletion"
1291 local orignids=$($LCTL get_param -n \
1292 osc.$FSNAME-OST0000-osc-[^M]*.import | grep failover_nids)
1294 local orignidcount=$(echo "$orignids" | wc -w)
1296 # Make a fake nid. Use the OST nid, and add 20 to the least significant
1297 # numerical part of it. Hopefully that's not already a failover address
1299 local OSTNID=$(do_facet ost1 "$LCTL get_param nis" | tail -1 | \
1301 local ORIGVAL=$(echo $OSTNID | egrep -oi "[0-9]*@")
1302 local NEWVAL=$((($(echo $ORIGVAL | egrep -oi "[0-9]*") + 20) % 256))
1303 local NEW=$(echo $OSTNID | sed "s/$ORIGVAL/$NEWVAL@/")
1304 echo "Using fake nid $NEW"
1306 local TEST="$LCTL get_param -n osc.$FSNAME-OST0000-osc-[^M]*.import |
1307 grep failover_nids | sed -n 's/.*\($NEW\).*/\1/p'"
1308 set_conf_param_and_check client "$TEST" \
1309 "$FSNAME-OST0000.failover.node" $NEW ||
1310 error "didn't add failover nid $NEW"
1311 local NIDS=$($LCTL get_param -n osc.$FSNAME-OST0000-osc-[^M]*.import |
1314 local NIDCOUNT=$(echo "$NIDS" | wc -w)
1315 echo "should have $((orignidcount + 1)) entries \
1316 in failover nids string, have $NIDCOUNT"
1317 [ $NIDCOUNT -eq $((orignidcount + 1)) ] ||
1318 error "Failover nid not added"
1320 do_facet mgs "$LCTL conf_param -d $FSNAME-OST0000.failover.node" ||
1321 error "conf_param delete failed"
1322 umount_client $MOUNT
1323 mount_client $MOUNT || error "mount_client $MOUNT failed"
1325 NIDS=$($LCTL get_param -n osc.$FSNAME-OST0000-osc-[^M]*.import |
1328 NIDCOUNT=$(echo "$NIDS" | wc -w)
1329 echo "only $orignidcount final entries should remain \
1330 in failover nids string, have $NIDCOUNT"
1331 [ $NIDCOUNT -eq $orignidcount ] || error "Failover nids not removed"
1333 cleanup || error "cleanup failed with rc $?"
1335 run_test 30b "Remove failover nids"
1337 test_31() { # bug 10734
1338 # ipaddr must not exist
1339 $MOUNT_CMD 4.3.2.1@tcp:/lustre $MOUNT || true
1340 cleanup || error "cleanup failed with rc $?"
1342 run_test 31 "Connect to non-existent node (shouldn't crash)"
1346 T32_BLIMIT=40960 # Kbytes
1350 # This is not really a test but a tool to create new disk
1351 # image tarballs for the upgrade tests.
1353 # Disk image tarballs should be created on single-node
1354 # clusters by running this test with default configurations
1355 # plus a few mandatory environment settings that are verified
1356 # at the beginning of the test.
1358 test_32newtarball() {
1362 local tmp=$TMP/t32_image_create
1363 local server_version=$(lustre_version_code $SINGLEMDS)
1368 if [ $FSNAME != t32fs -o \( -z "$MDSDEV" -a -z "$MDSDEV1" \) -o \
1369 $OSTCOUNT -ne 1 -o -z "$OSTDEV1" ]; then
1370 error "Needs FSNAME=t32fs MDSCOUNT=2 " \
1371 "MDSDEV1=<nonexistent_file>" \
1372 "MDSDEV2=<nonexistent_file>" \
1373 "(or MDSDEV, in the case of b1_8)" \
1374 "OSTCOUNT=1 OSTDEV1=<nonexistent_file>"
1378 echo "Found stale $tmp"
1382 mkdir $tmp/src || return 1
1383 tar cf - -C $src . | tar xf - -C $tmp/src
1384 dd if=/dev/zero of=$tmp/src/t32_qf_old bs=1M \
1385 count=$(($T32_BLIMIT / 1024 / 4))
1386 chown $T32_QID.$T32_QID $tmp/src/t32_qf_old
1388 # format ost with comma-separated NIDs to verify LU-4460
1389 local failnid="$(h2$NETTYPE 1.2.3.4),$(h2$NETTYPE 4.3.2.1)"
1390 MGSNID="$MGSNID,$MGSNID" OSTOPT="--failnode=$failnid" formatall
1394 [[ $server_version -ge $(version_code 2.3.50) ]] ||
1395 $LFS quotacheck -ug /mnt/$FSNAME
1396 $LFS setquota -u $T32_QID -b 0 -B $T32_BLIMIT -i 0 -I $T32_ILIMIT \
1399 tar cf - -C $tmp/src . | tar xf - -C /mnt/$FSNAME
1401 if [[ $MDSCOUNT -ge 2 ]]; then
1402 remote_dir=/mnt/$FSNAME/remote_dir
1403 $LFS mkdir -i 1 $remote_dir
1404 tar cf - -C $tmp/src . | tar xf - -C $remote_dir
1406 if [[ $server_version -ge $(version_code 2.7.0) ]]; then
1407 striped_dir=/mnt/$FSNAME/striped_dir_old
1408 $LFS mkdir -i 1 -c 2 $striped_dir
1409 tar cf - -C $tmp/src . | tar xf - -C $striped_dir
1415 mkdir $tmp/img || return 1
1419 pushd_dir=/mnt/$FSNAME
1420 if [[ $MDSCOUNT -ge 2 ]]; then
1421 pushd_dir=$remote_dir
1422 if [[ $server_version -ge $(version_code 2.7.0) ]]; then
1424 ls -Rni --time-style=+%s >$tmp/img/list2
1430 ls -Rni --time-style=+%s >$tmp/img/list
1431 find ! -name .lustre -type f -exec sha1sum {} \; |
1432 sort -k 2 >$tmp/img/sha1sums
1434 $LCTL get_param -n version | head -n 1 |
1435 sed -e 's/^lustre: *//' >$tmp/img/commit
1437 [[ $server_version -ge $(version_code 2.3.50) ]] ||
1438 $LFS quotaon -ug /mnt/$FSNAME
1439 $LFS quota -u $T32_QID -v /mnt/$FSNAME
1440 $LFS quota -v -u $T32_QID /mnt/$FSNAME |
1441 awk 'BEGIN { num='1' } { if ($1 == "'/mnt/$FSNAME'") \
1442 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1443 | tr -d "*" > $tmp/img/bspace
1444 $LFS quota -v -u $T32_QID /mnt/$FSNAME |
1445 awk 'BEGIN { num='5' } { if ($1 == "'/mnt/$FSNAME'") \
1446 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1447 | tr -d "*" > $tmp/img/ispace
1448 echo $T32_BLIMIT > $tmp/img/blimit
1449 echo $T32_ILIMIT > $tmp/img/ilimit
1454 find -type f -exec sha1sum {} \; | sort -k 2 >$tmp/sha1sums.src
1457 if ! diff -u $tmp/sha1sums.src $tmp/img/sha1sums; then
1458 echo "Data verification failed"
1461 uname -r >$tmp/img/kernel
1462 uname -m >$tmp/img/arch
1464 mv ${MDSDEV1:-$MDSDEV} $tmp/img
1465 for num in $(seq 2 $MDSCOUNT); do
1466 local devname=$(mdsdevname $num)
1468 [[ $(facet_fstype $facet) != zfs ]] ||
1469 devname=$(mdsvdevname $num)
1470 mv $devname $tmp/img
1472 mv $OSTDEV1 $tmp/img
1474 version=$(sed -e 's/\(^[0-9]\+\.[0-9]\+\)\(.*$\)/\1/' $tmp/img/commit |
1475 sed -e 's/\./_/g') # E.g., "1.8.7" -> "1_8"
1478 tar cjvf $dst/disk$version-$(facet_fstype $SINGLEMDS).tar.bz2 -S *
1483 #run_test 32newtarball "Create a new test_32 disk image tarball for this version"
1486 # The list of applicable tarballs is returned via the caller's
1487 # variable "tarballs".
1490 local node=$(facet_active_host $SINGLEMDS)
1491 local r="do_node $node"
1493 if [ "$CLIENTONLY" ]; then
1494 skip "Client-only testing"
1498 if ! $r which $TUNEFS; then
1499 skip_env "tunefs.lustre required on $node"
1503 local IMGTYPE=$(facet_fstype $SINGLEMDS)
1505 tarballs=$($r find $RLUSTRE/tests -maxdepth 1 -name \'disk*-$IMGTYPE.tar.bz2\')
1507 if [ -z "$tarballs" ]; then
1508 skip "No applicable tarballs found"
1513 t32_test_cleanup() {
1515 local facet=$SINGLEMDS
1516 local fstype=$(facet_fstype $facet)
1519 if $shall_cleanup_lustre; then
1520 umount $tmp/mnt/lustre || rc=$?
1522 if $shall_cleanup_mdt; then
1523 $r $UMOUNT $tmp/mnt/mdt || rc=$?
1525 if $shall_cleanup_mdt1; then
1526 $r $UMOUNT $tmp/mnt/mdt1 || rc=$?
1528 if $shall_cleanup_ost; then
1529 $r $UMOUNT $tmp/mnt/ost || rc=$?
1534 if [[ $fstype == zfs ]]; then
1536 local poolname_list="t32fs-mdt1 t32fs-ost1"
1538 ! $mdt2_is_available || poolname_list+=" t32fs-mdt2"
1540 for poolname in $poolname_list; do
1541 destroy_zpool $facet $poolname
1544 combined_mgs_mds || start_mgs || rc=$?
1548 t32_bits_per_long() {
1550 # Yes, this is not meant to be perfect.
1560 t32_reload_modules() {
1562 local all_removed=false
1565 while ((i < 20)); do
1566 echo "Unloading modules on $node: Attempt $i"
1567 do_rpc_nodes $node $LUSTRE_RMMOD $(facet_fstype $SINGLEMDS) &&
1569 do_rpc_nodes $node check_mem_leak || return 1
1570 if $all_removed; then
1571 do_rpc_nodes $node load_modules
1577 echo "Unloading modules on $node: Given up"
1581 t32_wait_til_devices_gone() {
1587 echo wait for devices to go
1588 while ((i < 20)); do
1589 devices=$(do_rpc_nodes $node $LCTL device_list | wc -l)
1590 loops=$(do_rpc_nodes $node losetup -a | grep -c t32)
1591 ((devices == 0 && loops == 0)) && return 0
1595 echo "waiting for dev on $node: dev $devices loop $loops given up"
1596 do_rpc_nodes $node "losetup -a"
1597 do_rpc_nodes $node "$LCTL device_list"
1601 t32_verify_quota() {
1605 local fstype=$(facet_fstype $SINGLEMDS)
1609 $LFS quota -u $T32_QID -v $mnt
1611 qval=$($LFS quota -v -u $T32_QID $mnt |
1612 awk 'BEGIN { num='1' } { if ($1 == "'$mnt'") \
1613 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1615 [ $qval -eq $img_bspace ] || {
1616 echo "bspace, act:$qval, exp:$img_bspace"
1620 qval=$($LFS quota -v -u $T32_QID $mnt |
1621 awk 'BEGIN { num='5' } { if ($1 == "'$mnt'") \
1622 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1624 [ $qval -eq $img_ispace ] || {
1625 echo "ispace, act:$qval, exp:$img_ispace"
1629 qval=$($LFS quota -v -u $T32_QID $mnt |
1630 awk 'BEGIN { num='3' } { if ($1 == "'$mnt'") \
1631 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1633 [ $qval -eq $img_blimit ] || {
1634 echo "blimit, act:$qval, exp:$img_blimit"
1638 qval=$($LFS quota -v -u $T32_QID $mnt |
1639 awk 'BEGIN { num='7' } { if ($1 == "'$mnt'") \
1640 { if (NF == 1) { getline } else { num++ } ; print $num;} }' \
1642 [ $qval -eq $img_ilimit ] || {
1643 echo "ilimit, act:$qval, exp:$img_ilimit"
1647 do_node $node $LCTL conf_param $fsname.quota.mdt=ug
1648 cmd="$LCTL get_param -n osd-$fstype.$fsname-MDT0000"
1649 cmd=$cmd.quota_slave.enabled
1650 wait_update $node "$cmd" "ug" || {
1651 echo "Enable mdt quota failed"
1655 do_node $node $LCTL conf_param $fsname.quota.ost=ug
1656 cmd="$LCTL get_param -n osd-$fstype.$fsname-OST0000"
1657 cmd=$cmd.quota_slave.enabled
1658 wait_update $node "$cmd" "ug" || {
1659 echo "Enable ost quota failed"
1664 runas -u $T32_QID -g $T32_QID dd if=/dev/zero of=$mnt/t32_qf_new \
1665 bs=1M count=$((img_blimit / 1024)) oflag=sync && {
1666 echo "Write succeed, but expect -EDQUOT"
1669 rm -f $mnt/t32_qf_new
1671 runas -u $T32_QID -g $T32_QID createmany -m $mnt/t32_qf_ \
1673 echo "Create succeed, but expect -EDQUOT"
1676 unlinkmany $mnt/t32_qf_ $img_ilimit
1684 local dne_upgrade=${dne_upgrade:-"no"}
1685 local ff_convert=${ff_convert:-"no"}
1686 local shall_cleanup_mdt=false
1687 local shall_cleanup_mdt1=false
1688 local shall_cleanup_ost=false
1689 local shall_cleanup_lustre=false
1690 local mdt2_is_available=false
1691 local node=$(facet_active_host $SINGLEMDS)
1692 local r="do_node $node"
1708 local fstype=$(facet_fstype $SINGLEMDS)
1709 local mdt_dev=$tmp/mdt
1710 local mdt2_dev=$tmp/mdt2
1711 local ost_dev=$tmp/ost
1716 combined_mgs_mds || stop_mgs || error "Unable to stop MGS"
1717 trap 'trap - RETURN; t32_test_cleanup' RETURN
1720 nid=$($r $LCTL list_nids | head -1)
1722 mkdir -p $tmp/mnt/lustre || error "mkdir $tmp/mnt/lustre failed"
1723 $r mkdir -p $tmp/mnt/{mdt,mdt1,ost}
1724 $r tar xjvf $tarball -S -C $tmp || {
1725 error_noexit "Unpacking the disk image tarball"
1728 img_commit=$($r cat $tmp/commit)
1729 img_kernel=$($r cat $tmp/kernel)
1730 img_arch=$($r cat $tmp/arch)
1731 img_bspace=$($r cat $tmp/bspace)
1732 img_ispace=$($r cat $tmp/ispace)
1734 # older images did not have "blimit" and "ilimit" files
1735 # use old values for T32_BLIMIT and T32_ILIMIT
1736 $r test -f $tmp/blimit && img_blimit=$($r cat $tmp/blimit) ||
1738 $r test -f $tmp/ilimit && img_ilimit=$($r cat $tmp/ilimit) ||
1741 echo "Upgrading from $(basename $tarball), created with:"
1742 echo " Commit: $img_commit"
1743 echo " Kernel: $img_kernel"
1744 echo " Arch: $img_arch"
1745 echo "OST version: $(lustre_build_version ost1)"
1747 # The conversion can be made only when both of the following
1748 # conditions are satisfied:
1749 # - ost device img version < 2.3.64
1750 # - ost server version >= 2.5
1751 [ $(version_code $img_commit) -ge $(version_code 2.3.64) -o \
1752 $(lustre_version_code ost1) -lt $(version_code 2.5.0) ] &&
1755 ! $r test -f $mdt2_dev || mdt2_is_available=true
1757 if [[ $fstype == zfs ]]; then
1760 local poolname_list="t32fs-mdt1 t32fs-ost1"
1762 ! $mdt2_is_available || poolname_list+=" t32fs-mdt2"
1764 for poolname in $poolname_list; do
1766 $ZPOOL list -H $poolname >/dev/null 2>&1 ||
1767 $ZPOOL import -f -d $tmp $poolname"
1770 mdt_dev=t32fs-mdt1/mdt1
1771 ost_dev=t32fs-ost1/ost1
1772 ! $mdt2_is_available || mdt2_dev=t32fs-mdt2/mdt2
1773 wait_update_facet $SINGLEMDS "$ZPOOL list |
1774 awk '/^t32fs-mdt1/ { print \\\$1 }'" "t32fs-mdt1" || {
1775 error_noexit "import zfs pool failed"
1780 $r $LCTL set_param debug="$PTLDEBUG"
1782 $r $TUNEFS --dryrun $mdt_dev || {
1784 error_noexit "tunefs.lustre before mounting the MDT"
1788 if $mdt2_is_available; then
1789 $r $TUNEFS --dryrun $mdt2_dev || {
1791 error_noexit "tunefs.lustre before mounting the MDT"
1796 if [ "$writeconf" ]; then
1798 if [ $fstype == "ldiskfs" ]; then
1800 $r $TUNEFS --quota $mdt_dev || {
1802 error_noexit "Enable mdt quota feature"
1805 if $mdt2_is_available; then
1806 $r $TUNEFS --quota $mdt2_dev || {
1808 error_noexit "Enable mdt quota feature"
1814 if [ -n "$($LCTL list_nids | grep -v '\(tcp\|lo\)[[:digit:]]*$')" ]; then
1815 [[ $(lustre_version_code mgs) -ge $(version_code 2.3.59) ]] ||
1816 { skip "LU-2200: Cannot run over Inifiniband w/o lctl replace_nids "
1817 "(Need MGS version at least 2.3.59)"; return 0; }
1819 local osthost=$(facet_active_host ost1)
1820 local ostnid=$(do_node $osthost $LCTL list_nids | head -1)
1823 if [ $fstype == "ldiskfs" ]; then
1826 $r $MOUNT_CMD -o $mopts $mdt_dev $tmp/mnt/mdt
1827 $r $LCTL replace_nids $fsname-OST0000 $ostnid
1828 $r $LCTL replace_nids $fsname-MDT0000 $nid
1829 $r $UMOUNT $tmp/mnt/mdt
1832 mopts=exclude=$fsname-OST0000
1833 if [ $fstype == "ldiskfs" ]; then
1838 t32_wait_til_devices_gone $node
1840 $r $MOUNT_CMD -o $mopts $mdt_dev $tmp/mnt/mdt || {
1842 error_noexit "Mounting the MDT"
1845 shall_cleanup_mdt=true
1847 if $mdt2_is_available; then
1848 mopts=mgsnode=$nid,$mopts
1849 $r $MOUNT_CMD -o $mopts $mdt2_dev $tmp/mnt/mdt1 || {
1851 error_noexit "Mounting the MDT"
1855 echo "mount new MDT....$mdt2_dev"
1856 $r $LCTL set_param -n mdt.${fsname}*.enable_remote_dir=1 ||
1857 error_noexit "enable remote dir create failed"
1859 shall_cleanup_mdt1=true
1860 elif [ "$dne_upgrade" != "no" ]; then
1861 local fs2mdsdev=$(mdsdevname 1_2)
1862 local fs2mdsvdev=$(mdsvdevname 1_2)
1864 echo "mkfs new MDT on ${fs2mdsdev}...."
1865 if [ $(facet_fstype mds1) == ldiskfs ]; then
1866 mkfsoptions="--mkfsoptions=\\\"-J size=8\\\""
1869 add $SINGLEMDS $(mkfs_opts mds2 $fs2mdsdev $fsname) --reformat \
1870 $mkfsoptions $fs2mdsdev $fs2mdsvdev > /dev/null || {
1871 error_noexit "Mkfs new MDT failed"
1875 [[ $(facet_fstype mds1) != zfs ]] || import_zpool mds1
1877 $r $TUNEFS --dryrun $fs2mdsdev || {
1878 error_noexit "tunefs.lustre before mounting the MDT"
1882 echo "mount new MDT....$fs2mdsdev"
1883 $r $MOUNT_CMD -o $mopts $fs2mdsdev $tmp/mnt/mdt1 || {
1884 error_noexit "mount mdt1 failed"
1888 $r $LCTL set_param -n mdt.${fsname}*.enable_remote_dir=1 ||
1889 error_noexit "enable remote dir create failed"
1891 shall_cleanup_mdt1=true
1894 uuid=$($r $LCTL get_param -n mdt.$fsname-MDT0000.uuid) || {
1895 error_noexit "Getting MDT UUID"
1898 if [ "$uuid" != $fsname-MDT0000_UUID ]; then
1899 error_noexit "Unexpected MDT UUID: \"$uuid\""
1903 $r $TUNEFS --dryrun $ost_dev || {
1904 error_noexit "tunefs.lustre before mounting the OST"
1907 if [ "$writeconf" ]; then
1908 mopts=mgsnode=$nid,$writeconf
1909 if [ $fstype == "ldiskfs" ]; then
1911 $r $TUNEFS --quota $ost_dev || {
1913 error_noexit "Enable ost quota feature"
1919 if [ $fstype == "ldiskfs" ]; then
1924 $r $MOUNT_CMD -onomgs -o$mopts $ost_dev $tmp/mnt/ost || {
1925 error_noexit "Mounting the OST"
1928 shall_cleanup_ost=true
1930 uuid=$($r $LCTL get_param -n obdfilter.$fsname-OST0000.uuid) || {
1931 error_noexit "Getting OST UUID"
1934 if [ "$uuid" != $fsname-OST0000_UUID ]; then
1935 error_noexit "Unexpected OST UUID: \"$uuid\""
1939 $r $LCTL conf_param $fsname-OST0000.osc.max_dirty_mb=15 || {
1940 error_noexit "Setting \"max_dirty_mb\""
1943 $r $LCTL conf_param $fsname-OST0000.failover.node=$nid || {
1944 error_noexit "Setting OST \"failover.node\""
1947 $r $LCTL conf_param $fsname-MDT0000.mdc.max_rpcs_in_flight=9 || {
1948 error_noexit "Setting \"max_rpcs_in_flight\""
1951 $r $LCTL conf_param $fsname-MDT0000.failover.node=$nid || {
1952 error_noexit "Setting MDT \"failover.node\""
1955 $r $LCTL pool_new $fsname.interop || {
1956 error_noexit "Setting \"interop\""
1959 $r $LCTL conf_param $fsname-MDT0000.lov.stripesize=4M || {
1960 error_noexit "Setting \"lov.stripesize\""
1963 $r $LCTL conf_param $fsname-MDT0000.mdd.atime_diff=70 || {
1964 error_noexit "Setting \"mdd.atime_diff\""
1968 if [ "$ff_convert" != "no" -a $(facet_fstype ost1) == "ldiskfs" ]; then
1969 $r $LCTL lfsck_start -M $fsname-OST0000 || {
1970 error_noexit "Start OI scrub on OST0"
1974 # The oi_scrub should be on ost1, but for test_32(),
1975 # all on the SINGLEMDS.
1976 wait_update_facet $SINGLEMDS "$LCTL get_param -n \
1977 osd-ldiskfs.$fsname-OST0000.oi_scrub |
1978 awk '/^status/ { print \\\$2 }'" "completed" 30 || {
1979 error_noexit "Failed to get the expected 'completed'"
1983 local UPDATED=$($r $LCTL get_param -n \
1984 osd-ldiskfs.$fsname-OST0000.oi_scrub |
1985 awk '/^updated/ { print $2 }')
1986 [ $UPDATED -ge 1 ] || {
1987 error_noexit "Only $UPDATED objects have been converted"
1992 if [ "$dne_upgrade" != "no" ]; then
1993 $r $LCTL conf_param \
1994 $fsname-MDT0001.mdc.max_rpcs_in_flight=9 || {
1995 error_noexit "Setting MDT1 \"max_rpcs_in_flight\""
1998 $r $LCTL conf_param $fsname-MDT0001.failover.node=$nid || {
1999 error_noexit "Setting MDT1 \"failover.node\""
2002 $r $LCTL conf_param $fsname-MDT0001.lov.stripesize=4M || {
2003 error_noexit "Setting MDT1 \"lov.stripesize\""
2008 if [ "$writeconf" ]; then
2009 $MOUNT_CMD $nid:/$fsname $tmp/mnt/lustre || {
2010 error_noexit "Mounting the client"
2014 shall_cleanup_lustre=true
2015 $r $LCTL set_param debug="$PTLDEBUG"
2017 t32_verify_quota $node $fsname $tmp/mnt/lustre || {
2018 error_noexit "verify quota failed"
2022 if $r test -f $tmp/list; then
2024 # There is not a Test Framework API to copy files to or
2025 # from a remote node.
2027 # LU-2393 - do both sorts on same node to ensure locale
2029 local list_file=$tmp/list
2031 if $mdt2_is_available; then
2032 if [[ -d $tmp/mnt/lustre/striped_dir_old ]] &&
2033 $r test -f $tmp/list2; then
2034 list_file=$tmp/list2
2035 pushd $tmp/mnt/lustre/striped_dir_old
2037 pushd $tmp/mnt/lustre/remote_dir
2040 pushd $tmp/mnt/lustre
2042 $r cat $list_file | sort -k 6 >$tmp/list.orig
2043 ls -Rni --time-style=+%s | sort -k 6 |
2044 sed 's/\. / /' >$tmp/list || {
2050 # 32-bit and 64-bit clients use different algorithms to
2051 # convert FIDs into inode numbers. Hence, remove the
2052 # inode numbers from the lists, if the original list was
2053 # created on an architecture with different number of
2056 if [ $(t32_bits_per_long $(uname -m)) != \
2057 $(t32_bits_per_long $img_arch) ]; then
2058 echo "Different number of bits per \"long\"" \
2059 "from the disk image"
2060 for list in list.orig list; do
2061 sed -i -e 's/^[0-9]\+[ \t]\+//' \
2065 if ! diff -ub $tmp/list.orig $tmp/list; then
2066 error_noexit "list verification failed"
2070 echo "list verification skipped"
2073 if [ "$dne_upgrade" != "no" ]; then
2074 $LFS mkdir -i 1 -c2 $tmp/mnt/lustre/striped_dir || {
2075 error_noexit "set striped dir failed"
2079 $LFS setdirstripe -D -c2 $tmp/mnt/lustre/striped_dir
2081 pushd $tmp/mnt/lustre
2082 tar -cf - . --exclude=./striped_dir \
2083 --exclude=./striped_dir_old \
2084 --exclude=./remote_dir |
2085 tar -xvf - -C striped_dir 1>/dev/null || {
2086 error_noexit "cp to striped dir failed"
2092 # If it is upgrade from DNE (2.5), then rename the remote dir,
2093 # which is created in 2.5 to striped dir.
2094 if $mdt2_is_available && [[ "$dne_upgrade" != "no" ]]; then
2095 stripe_index=$($LFS getdirstripe -i \
2096 $tmp/mnt/lustre/remote_dir)
2098 [[ $stripe_index -eq 1 ]] || {
2099 error_noexit "get index \"$stripe_index\"" \
2100 "from remote dir failed"
2103 mv $tmp/mnt/lustre/remote_dir \
2104 $tmp/mnt/lustre/striped_dir/ || {
2105 error_noexit "mv remote dir failed"
2110 # If it is upgraded from DNE (2.7), then move the striped dir
2111 # which was created in 2.7 to the new striped dir.
2112 if $mdt2_is_available && [[ "$dne_upgrade" != "no" ]] &&
2113 [[ -d $tmp/mnt/lustre/striped_dir_old ]]; then
2114 stripe_count=$($LFS getdirstripe -c \
2115 $tmp/mnt/lustre/striped_dir_old)
2116 [[ $stripe_count -eq 2 ]] || {
2117 error_noexit "get count $stripe_count" \
2118 "from striped dir failed"
2121 mv $tmp/mnt/lustre/striped_dir_old \
2122 $tmp/mnt/lustre/striped_dir/ || {
2123 error_noexit "mv striped dir failed"
2129 $r $LCTL set_param -n osd*.*.force_sync=1
2130 dd if=/dev/zero of=$tmp/mnt/lustre/tmp_file bs=10k count=10 || {
2131 error_noexit "dd failed"
2134 rm -rf $tmp/mnt/lustre/tmp_file || {
2135 error_noexit "rm failed"
2139 if $r test -f $tmp/sha1sums; then
2140 # LU-2393 - do both sorts on same node to ensure locale
2142 $r cat $tmp/sha1sums | sort -k 2 >$tmp/sha1sums.orig
2143 if [ "$dne_upgrade" != "no" ]; then
2144 pushd $tmp/mnt/lustre/striped_dir
2146 pushd $tmp/mnt/lustre
2149 find ! -path "*remote_dir*" ! -path "*striped_dir*" \
2150 ! -name .lustre -type f -exec sha1sum {} \; |
2151 sort -k 2 >$tmp/sha1sums || {
2153 error_noexit "sha1sum"
2157 if ! diff -ub $tmp/sha1sums.orig $tmp/sha1sums; then
2158 error_noexit "sha1sum verification failed"
2162 # if upgrade from DNE(2.5), then check remote directory
2163 # if upgrade from DNE(2.7), then check striped directory
2164 if $mdt2_is_available &&
2165 [[ "$dne_upgrade" != "no" ]]; then
2166 local new_dir="$tmp/mnt/lustre/striped_dir"
2167 local striped_dir_old="$new_dir/striped_dir_old"
2169 local dir_list="$new_dir/remote_dir"
2170 [[ ! -d $triped_dir_old ]] ||
2171 dir_list+=" $striped_dir_old"
2173 for dir in $dir_list; do
2175 find ! -name .lustre -type f \
2176 -exec sha1sum {} \; |
2177 sort -k 2 >$tmp/sha1sums || {
2179 error_noexit "sha1sum"
2183 if ! diff -ub $tmp/sha1sums.orig \
2185 error_noexit "sha1sum $dir" \
2192 echo "sha1sum verification skipped"
2195 if [ "$dne_upgrade" != "no" ]; then
2196 rm -rf $tmp/mnt/lustre/striped_dir || {
2197 error_noexit "remove remote dir failed"
2202 # migrate files/dirs to remote MDT, then move them back
2203 if [ $(lustre_version_code mds1) -ge $(version_code 2.7.50) -a \
2204 $dne_upgrade != "no" ]; then
2205 $r $LCTL set_param -n \
2206 mdt.${fsname}*.enable_remote_dir=1 2>/dev/null
2208 echo "test migration"
2209 pushd $tmp/mnt/lustre
2210 for dir in $(find ! -name .lustre ! -name . -type d); do
2211 mdt_index=$($LFS getdirstripe -i $dir)
2212 stripe_cnt=$($LFS getdirstripe -c $dir)
2213 if [ $mdt_index = 0 -a $stripe_cnt -le 1 ]; then
2214 $LFS mv -M 1 $dir || {
2216 error_noexit "migrate MDT1 failed"
2222 for dir in $(find ! -name . ! -name .lustre -type d); do
2223 mdt_index=$($LFS getdirstripe -i $dir)
2224 stripe_cnt=$($LFS getdirstripe -c $dir)
2225 if [ $mdt_index = 1 -a $stripe_cnt -le 1 ]; then
2226 $LFS mv -M 0 $dir || {
2228 error_noexit "migrate MDT0 failed"
2237 # When adding new data verification tests, please check for
2238 # the presence of the required reference files first, like
2239 # the "sha1sums" and "list" tests above, to avoid the need to
2240 # regenerate every image for each test addition.
2243 nrpcs_orig=$($LCTL get_param \
2244 -n mdc.*MDT0000*.max_rpcs_in_flight) || {
2245 error_noexit "Getting \"max_rpcs_in_flight\""
2248 nrpcs=$((nrpcs_orig + 5))
2249 $r $LCTL conf_param $fsname-MDT0000.mdc.max_rpcs_in_flight=$nrpcs || {
2250 error_noexit "Changing \"max_rpcs_in_flight\""
2253 wait_update $HOSTNAME "$LCTL get_param \
2254 -n mdc.*MDT0000*.max_rpcs_in_flight" $nrpcs || {
2255 error_noexit "Verifying \"max_rpcs_in_flight\""
2259 umount $tmp/mnt/lustre || {
2260 error_noexit "Unmounting the client"
2263 shall_cleanup_lustre=false
2265 if [[ "$dne_upgrade" != "no" ]] || $mdt2_is_available; then
2266 $r $UMOUNT $tmp/mnt/mdt1 || {
2267 error_noexit "Unmounting the MDT2"
2270 shall_cleanup_mdt1=false
2273 $r $UMOUNT $tmp/mnt/mdt || {
2274 error_noexit "Unmounting the MDT"
2277 shall_cleanup_mdt=false
2279 $r $UMOUNT $tmp/mnt/ost || {
2280 error_noexit "Unmounting the OST"
2283 shall_cleanup_ost=false
2285 t32_reload_modules $node || {
2286 error_noexit "Reloading modules"
2290 if [[ $fstype == zfs ]]; then
2291 local poolname=t32fs-mdt1
2293 $ZPOOL list -H $poolname >/dev/null 2>&1 ||
2294 $ZPOOL import -f -d $tmp $poolname"
2297 # mount a second time to make sure we didnt leave upgrade flag on
2298 $r $TUNEFS --dryrun $mdt_dev || {
2300 error_noexit "tunefs.lustre before remounting the MDT"
2304 mopts=exclude=$fsname-OST0000
2305 if [ $fstype == "ldiskfs" ]; then
2308 $r $MOUNT_CMD -o $mopts $mdt_dev $tmp/mnt/mdt || {
2309 error_noexit "Remounting the MDT"
2312 shall_cleanup_mdt=true
2322 for tarball in $tarballs; do
2323 t32_test $tarball || let "rc += $?"
2327 run_test 32a "Upgrade (not live)"
2335 for tarball in $tarballs; do
2336 t32_test $tarball writeconf || let "rc += $?"
2340 run_test 32b "Upgrade with writeconf"
2347 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
2349 for tarball in $tarballs; do
2350 # Do not support 1_8 and 2_1 direct upgrade to DNE2 anymore */
2351 echo $tarball | grep "1_8" && continue
2352 echo $tarball | grep "2_1" && continue
2354 dne_upgrade=yes t32_test $tarball writeconf || rc=$?
2358 run_test 32c "dne upgrade test"
2366 for tarball in $tarballs; do
2367 ff_convert=yes t32_test $tarball || rc=$?
2371 run_test 32d "convert ff test"
2373 test_33a() { # bug 12333, was test_33
2374 local FSNAME2=test-123
2375 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
2378 [ -n "$ost1_HOST" ] && fs2ost_HOST=$ost1_HOST
2380 if [ -z "$fs2ost_DEV" -o -z "$fs2mds_DEV" ]; then
2381 local dev=${SINGLEMDS}_dev
2382 local MDSDEV=${!dev}
2383 is_blkdev $SINGLEMDS $MDSDEV &&
2384 skip_env "mixed loopback and real device not working" &&
2388 local fs2mdsdev=$(mdsdevname 1_2)
2389 local fs2ostdev=$(ostdevname 1_2)
2390 local fs2mdsvdev=$(mdsvdevname 1_2)
2391 local fs2ostvdev=$(ostvdevname 1_2)
2393 if [ $(facet_fstype mds1) == ldiskfs ]; then
2394 mkfsoptions="--mkfsoptions=\\\"-J size=8\\\"" # See bug 17931.
2397 add fs2mds $(mkfs_opts mds1 ${fs2mdsdev}) --mgs --fsname=${FSNAME2} \
2398 --reformat $mkfsoptions $fs2mdsdev $fs2mdsvdev || exit 10
2399 add fs2ost $(mkfs_opts ost1 ${fs2ostdev}) --mgsnode=$MGSNID \
2400 --fsname=${FSNAME2} --index=8191 --reformat $fs2ostdev \
2401 $fs2ostvdev || exit 10
2403 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS && trap cleanup_fs2 EXIT INT
2404 start fs2ost $fs2ostdev $OST_MOUNT_OPTS
2405 do_facet $SINGLEMDS "$LCTL conf_param $FSNAME2.sys.timeout=200" ||
2406 error "$LCTL conf_param $FSNAME2.sys.timeout=200 failed"
2407 mkdir -p $MOUNT2 || error "mkdir $MOUNT2 failed"
2408 $MOUNT_CMD $MGSNID:/${FSNAME2} $MOUNT2 || error "$MOUNT_CMD failed"
2411 cp /etc/hosts $MOUNT2/ || error "copy /etc/hosts $MOUNT2/ failed"
2412 $GETSTRIPE $MOUNT2/hosts || error "$GETSTRIPE $MOUNT2/hosts failed"
2417 cleanup_nocli || error "cleanup_nocli failed with $?"
2419 run_test 33a "Mount ost with a large index number"
2421 test_33b() { # was test_34
2424 do_facet client dd if=/dev/zero of=$MOUNT/24 bs=1024k count=1
2425 # Drop lock cancelation reply during umount
2426 #define OBD_FAIL_LDLM_CANCEL_NET 0x304
2427 do_facet client $LCTL set_param fail_loc=0x80000304
2428 #lctl set_param debug=-1
2429 umount_client $MOUNT
2430 cleanup || error "cleanup failed with $?"
2432 run_test 33b "Drop cancel during umount"
2436 do_facet client "sh runmultiop_bg_pause $DIR/file O_c"
2437 manual_umount_client
2439 do_facet client killall -USR1 multiop
2440 if [ $rc -eq 0 ]; then
2441 error "umount not fail!"
2444 cleanup || error "cleanup failed with rc $?"
2446 run_test 34a "umount with opened file should be fail"
2450 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
2451 stop_mds || error "Unable to stop MDS"
2453 manual_umount_client --force || error "mtab after failed umount with $?"
2455 cleanup || error "cleanup failed with $?"
2457 run_test 34b "force umount with failed mds should be normal"
2461 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
2462 stop_ost || error "Unable to stop OST1"
2464 manual_umount_client --force || error "mtab after failed umount with $?"
2466 cleanup || error "cleanup failed with $?"
2468 run_test 34c "force umount with failed ost should be normal"
2470 test_35a() { # bug 12459
2473 DBG_SAVE="`$LCTL get_param -n debug`"
2474 $LCTL set_param debug="ha"
2476 log "Set up a fake failnode for the MDS"
2478 local device=$(do_facet $SINGLEMDS "$LCTL get_param -n devices" |
2479 awk '($3 ~ "mdt" && $4 ~ "MDT") { print $4 }' | head -1)
2480 do_facet mgs "$LCTL conf_param \
2481 ${device}.failover.node=$(h2$NETTYPE $FAKENID)" ||
2482 error "Setting ${device}.failover.node=\
2483 $(h2$NETTYPE $FAKENID) failed."
2485 log "Wait for RECONNECT_INTERVAL seconds (10s)"
2488 MSG="conf-sanity.sh test_35a `date +%F%kh%Mm%Ss`"
2491 log "Stopping the MDT: $device"
2492 stop_mdt 1 || error "MDT0 stop fail"
2494 df $MOUNT > /dev/null 2>&1 &
2496 log "Restarting the MDT: $device"
2497 start_mdt 1 || error "MDT0 start fail"
2498 log "Wait for df ($DFPID) ... "
2501 $LCTL set_param debug="$DBG_SAVE"
2503 # retrieve from the log the first server that the client tried to
2504 # contact after the connection loss
2505 $LCTL dk $TMP/lustre-log-$TESTNAME.log
2506 NEXTCONN=`awk "/${MSG}/ {start = 1;}
2507 /import_select_connection.*$device-mdc.* using connection/ {
2509 if (\\\$NF ~ /$FAKENID/)
2515 }" $TMP/lustre-log-$TESTNAME.log`
2516 [ "$NEXTCONN" != "0" ] &&
2517 error "Tried to connect to ${NEXTCONN} not last active server"
2518 cleanup || error "cleanup failed with $?"
2519 # remove nid settings
2520 writeconf_or_reformat
2522 run_test 35a "Reconnect to the last active server first"
2524 test_35b() { # bug 18674
2525 remote_mds || { skip "local MDS" && return 0; }
2529 $LCTL set_param debug="ha"
2531 MSG="conf-sanity.sh test_35b `date +%F%kh%Mm%Ss`"
2534 log "Set up a fake failnode for the MDS"
2536 local device=$(do_facet $SINGLEMDS "$LCTL get_param -n devices" |
2537 awk '($3 ~ "mdt" && $4 ~ "MDT") { print $4 }' | head -1)
2538 do_facet mgs "$LCTL conf_param \
2539 ${device}.failover.node=$(h2$NETTYPE $FAKENID)" ||
2540 error "Set ${device}.failover.node=\
2541 $(h2$NETTYPE $FAKENID) failed"
2543 local at_max_saved=0
2544 # adaptive timeouts may prevent seeing the issue
2545 if at_is_enabled; then
2546 at_max_saved=$(at_max_get mds)
2547 at_max_set 0 mds client
2550 mkdir $MOUNT/$tdir || error "mkdir $MOUNT/$tdir failed"
2552 log "Injecting EBUSY on MDS"
2553 # Setting OBD_FAIL_MDS_RESEND=0x136
2554 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0x80000136" ||
2555 error "unable to set param fail_loc=0x80000136"
2557 $LCTL set_param mdc.${FSNAME}*.stats=clear
2559 log "Creating a test file and stat it"
2560 touch $MOUNT/$tdir/$tfile || error "touch $MOUNT/$tdir/$tfile failed"
2561 stat $MOUNT/$tdir/$tfile
2563 log "Stop injecting EBUSY on MDS"
2564 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0" ||
2565 error "unable to set param fail_loc=0"
2566 rm -f $MOUNT/$tdir/$tfile || error "remove $MOUNT/$tdir/$tfile failed"
2569 # restore adaptive timeout
2570 [ $at_max_saved -ne 0 ] && at_max_set $at_max_saved mds client
2572 $LCTL dk $TMP/lustre-log-$TESTNAME.log
2574 CONNCNT=$($LCTL get_param mdc.${FSNAME}*.stats |
2575 awk '/mds_connect/{print $2}')
2577 # retrieve from the log if the client has ever tried to
2578 # contact the fake server after the loss of connection
2579 FAILCONN=`awk "BEGIN {ret = 0;}
2580 /import_select_connection.*${FSNAME}-MDT0000-mdc.* using connection/ {
2582 if (\\\$NF ~ /$FAKENID/) {
2587 END {print ret}" $TMP/lustre-log-$TESTNAME.log`
2589 [ "$FAILCONN" == "0" ] &&
2590 error "The client reconnection has not been triggered"
2591 [ "$FAILCONN" == "2" ] &&
2592 error "Primary server busy, client reconnect to failover failed"
2595 # When OBD_FAIL_MDS_RESEND is hit, we sleep for 2 * obd_timeout
2596 # Reconnects are supposed to be rate limited to one every 5s
2597 [ $CONNCNT -gt $((2 * $TIMEOUT / 5 + 1)) ] &&
2598 error "Too many reconnects $CONNCNT"
2600 cleanup || error "cleanup failed with $?"
2601 # remove nid settings
2602 writeconf_or_reformat
2604 run_test 35b "Continue reconnection retries, if the active server is busy"
2607 [ $OSTCOUNT -lt 2 ] && skip_env "skipping test for single OST" && return
2609 [ "$ost_HOST" = "`hostname`" -o "$ost1_HOST" = "`hostname`" ] ||
2610 { skip "remote OST" && return 0; }
2613 local FSNAME2=test1234
2614 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
2616 [ -n "$ost1_HOST" ] && fs2ost_HOST=$ost1_HOST && fs3ost_HOST=$ost1_HOST
2618 if [ -z "$fs2ost_DEV" -o -z "$fs2mds_DEV" -o -z "$fs3ost_DEV" ]; then
2619 is_blkdev $SINGLEMDS $MDSDEV &&
2620 skip_env "mixed loopback and real device not working" && return
2623 local fs2mdsdev=$(mdsdevname 1_2)
2624 local fs2ostdev=$(ostdevname 1_2)
2625 local fs3ostdev=$(ostdevname 2_2)
2626 local fs2mdsvdev=$(mdsvdevname 1_2)
2627 local fs2ostvdev=$(ostvdevname 1_2)
2628 local fs3ostvdev=$(ostvdevname 2_2)
2630 add fs2mds $(mkfs_opts mds1 ${fs2mdsdev}) --mgs --fsname=${FSNAME2} \
2631 --reformat $fs2mdsdev $fs2mdsvdev || exit 10
2632 # XXX after we support non 4K disk blocksize in ldiskfs, specify a
2633 # different one than the default value here.
2634 add fs2ost $(mkfs_opts ost1 ${fs2ostdev}) --mgsnode=$MGSNID \
2635 --fsname=${FSNAME2} --reformat $fs2ostdev $fs2ostvdev || exit 10
2636 add fs3ost $(mkfs_opts ost2 ${fs3ostdev}) --mgsnode=$MGSNID \
2637 --fsname=${FSNAME2} --reformat $fs3ostdev $fs3ostvdev || exit 10
2639 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS
2640 start fs2ost $fs2ostdev $OST_MOUNT_OPTS
2641 start fs3ost $fs3ostdev $OST_MOUNT_OPTS
2642 mkdir -p $MOUNT2 || error "mkdir $MOUNT2 failed"
2643 $MOUNT_CMD $MGSNID:/${FSNAME2} $MOUNT2 || error "$MOUNT_CMD failed"
2645 sleep 5 # until 11778 fixed
2647 dd if=/dev/zero of=$MOUNT2/$tfile bs=1M count=7 || error "dd failed"
2649 BKTOTAL=$($LCTL get_param -n obdfilter.*.kbytestotal |
2650 awk 'BEGIN{total=0}; {total+=$1}; END{print total}')
2651 BKFREE=$($LCTL get_param -n obdfilter.*.kbytesfree |
2652 awk 'BEGIN{free=0}; {free+=$1}; END{print free}')
2653 BKAVAIL=$($LCTL get_param -n obdfilter.*.kbytesavail |
2654 awk 'BEGIN{avail=0}; {avail+=$1}; END{print avail}')
2655 STRING=$(df -P $MOUNT2 | tail -n 1 | awk '{print $2","$3","$4}')
2656 DFTOTAL=$(echo $STRING | cut -d, -f1)
2657 DFUSED=$(echo $STRING | cut -d, -f2)
2658 DFAVAIL=$(echo $STRING | cut -d, -f3)
2659 DFFREE=$(($DFTOTAL - $DFUSED))
2661 ALLOWANCE=$((64 * $OSTCOUNT))
2663 if [ $DFTOTAL -lt $(($BKTOTAL - $ALLOWANCE)) ] ||
2664 [ $DFTOTAL -gt $(($BKTOTAL + $ALLOWANCE)) ] ; then
2665 echo "**** FAIL: df total($DFTOTAL) mismatch OST total($BKTOTAL)"
2668 if [ $DFFREE -lt $(($BKFREE - $ALLOWANCE)) ] ||
2669 [ $DFFREE -gt $(($BKFREE + $ALLOWANCE)) ] ; then
2670 echo "**** FAIL: df free($DFFREE) mismatch OST free($BKFREE)"
2673 if [ $DFAVAIL -lt $(($BKAVAIL - $ALLOWANCE)) ] ||
2674 [ $DFAVAIL -gt $(($BKAVAIL + $ALLOWANCE)) ] ; then
2675 echo "**** FAIL: df avail($DFAVAIL) mismatch OST avail($BKAVAIL)"
2680 stop fs3ost -f || error "unable to stop OST3"
2681 stop fs2ost -f || error "unable to stop OST2"
2682 stop fs2mds -f || error "unable to stop second MDS"
2683 unload_modules_conf || error "unable unload modules"
2686 run_test 36 "df report consistency on OSTs with different block size"
2689 local mntpt=$(facet_mntpt $SINGLEMDS)
2690 local mdsdev=$(mdsdevname ${SINGLEMDS//mds/})
2691 local mdsdev_sym="$TMP/sym_mdt.img"
2692 local opts=$MDS_MOUNT_OPTS
2695 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
2696 skip "Currently only applicable to ldiskfs-based MDTs"
2700 echo "MDS : $mdsdev"
2701 echo "SYMLINK : $mdsdev_sym"
2702 do_facet $SINGLEMDS rm -f $mdsdev_sym
2704 do_facet $SINGLEMDS ln -s $mdsdev $mdsdev_sym
2706 echo "mount symlink device - $mdsdev_sym"
2708 if ! do_facet $SINGLEMDS test -b $mdsdev; then
2709 opts=$(csa_add "$opts" -o loop)
2711 mount_op=$(do_facet $SINGLEMDS mount -v -t lustre $opts \
2712 $mdsdev_sym $mntpt 2>&1)
2715 echo mount_op=$mount_op
2717 do_facet $SINGLEMDS "$UMOUNT $mntpt && rm -f $mdsdev_sym"
2719 if $(echo $mount_op | grep -q "unable to set tunable"); then
2720 error "set tunables failed for symlink device"
2723 [ $rc -eq 0 ] || error "mount symlink $mdsdev_sym failed! rc=$rc"
2725 run_test 37 "verify set tunables works for symlink device"
2727 test_38() { # bug 14222
2728 local fstype=$(facet_fstype $SINGLEMDS)
2729 local mntpt=$(facet_mntpt $SINGLEMDS)
2734 local SRC="/etc /bin"
2735 local FILES=$(find $SRC -type f -mtime +1 | head -n $COUNT)
2736 log "copying $(echo $FILES | wc -w) files to $DIR/$tdir"
2737 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
2738 tar cf - $FILES | tar xf - -C $DIR/$tdir ||
2739 error "copying $SRC to $DIR/$tdir"
2741 umount_client $MOUNT || error "umount_client $MOUNT failed"
2742 do_facet $SINGLEMDS "$LCTL get_param osp.*.prealloc_next_id"
2743 stop_mds || error "Unable to stop MDS"
2744 log "delete lov_objid file on MDS"
2746 mount_fstype $SINGLEMDS || error "mount MDS failed (1)"
2748 do_facet $SINGLEMDS "od -Ax -td8 $mntpt/lov_objid; rm $mntpt/lov_objid"
2750 unmount_fstype $SINGLEMDS || error "umount failed (1)"
2752 # check create in mds_lov_connect
2753 start_mds || error "unable to start MDS"
2754 mount_client $MOUNT || error "mount_client $MOUNT failed"
2756 [ $V ] && log "verifying $DIR/$tdir/$f"
2757 diff -q $f $DIR/$tdir/$f || ERROR=y
2759 do_facet $SINGLEMDS "$LCTL get_param osp.*.prealloc_next_id"
2760 if [ "$ERROR" = "y" ]; then
2761 # check it's updates in sync
2762 umount_client $MOUNT
2764 mount_fstype $SIGNLEMDS
2765 do_facet $SINGLEMDS "od -Ax -td8 $mntpt/lov_objid"
2766 unmount_fstype $SINGLEMDS
2767 error "old and new files are different after connect" || true
2769 touch $DIR/$tdir/f2 || error "f2 file create failed"
2771 # check it's updates in sync
2772 umount_client $MOUNT || error "second umount_client $MOUNT failed"
2775 mount_fstype $SINGLEMDS || error "mount MDS failed (3)"
2777 do_facet $SINGLEMDS "od -Ax -td8 $mntpt/lov_objid"
2778 do_facet $SINGLEMDS dd if=/dev/zero of=$mntpt/lov_objid.clear count=8
2780 unmount_fstype $SINGLEMDS || error "umount failed (3)"
2782 start_mds || error "unable to start MDS"
2783 mount_client $MOUNT || error "mount_client $MOUNT failed"
2785 [ $V ] && log "verifying $DIR/$tdir/$f"
2786 diff -q $f $DIR/$tdir/$f || ERROR=y
2788 touch $DIR/$tdir/f3 || error "f3 file create failed"
2789 do_facet $SINGLEMDS "$LCTL get_param osp.*.prealloc_next_id"
2790 umount_client $MOUNT || error "third umount_client $MOUNT failed"
2792 mount_fstype $SINGLEMDS || error "mount MDS failed (4)"
2793 do_facet $SINGLEMDS "od -Ax -td8 $mntpt/lov_objid"
2794 unmount_fstype $SINGLEMDS || error "umount failed (4)"
2796 [ "$ERROR" = "y" ] &&
2797 error "old and new files are different after sync" || true
2799 log "files compared the same"
2800 cleanup || error "cleanup failed with $?"
2802 run_test 38 "MDS recreates missing lov_objid file from OST data"
2807 cleanup || error "cleanup failed with $?"
2808 perl $SRCDIR/leak_finder.pl $TMP/debug 2>&1 | egrep '*** Leak:' &&
2809 error "memory leak detected" || true
2811 run_test 39 "leak_finder recognizes both LUSTRE and LNET malloc messages"
2813 test_40() { # bug 15759
2814 start_ost || error "Unable to start OST1"
2815 #define OBD_FAIL_TGT_TOOMANY_THREADS 0x706
2816 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0x80000706"
2818 cleanup || error "cleanup failed with rc $?"
2820 run_test 40 "race during service thread startup"
2822 test_41a() { #bug 14134
2823 if [ $(facet_fstype $SINGLEMDS) == ldiskfs ] &&
2824 ! do_facet $SINGLEMDS test -b $(mdsdevname 1); then
2825 skip "Loop devices does not work with nosvc option"
2830 { skip "needs combined MGT and MDT device" && return 0; }
2832 start_mdt 1 -o nosvc -n
2833 if [ $MDSCOUNT -ge 2 ]; then
2834 for num in $(seq 2 $MDSCOUNT); do
2835 start_mdt $num || return
2838 start ost1 $(ostdevname 1) $OST_MOUNT_OPTS
2839 start_mdt 1 -o nomgs,force
2840 mount_client $MOUNT || error "mount_client $MOUNT failed"
2843 echo "blah blah" > $MOUNT/$tfile
2846 umount_client $MOUNT || error "umount_client $MOUNT failed"
2847 stop ost1 -f || error "unable to stop OST1"
2848 stop_mds || error "Unable to stop MDS"
2849 stop_mds || error "Unable to stop MDS on second try"
2851 run_test 41a "mount mds with --nosvc and --nomgs"
2854 if [ $(facet_fstype $SINGLEMDS) == ldiskfs ] &&
2855 ! do_facet $SINGLEMDS test -b $(mdsdevname 1); then
2856 skip "Loop devices does not work with nosvc option"
2860 ! combined_mgs_mds && skip "needs combined mgs device" && return 0
2864 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
2866 start_mdt 1 -o nosvc -n
2867 if [ $MDSCOUNT -ge 2 ]; then
2868 for num in $(seq 2 $MDSCOUNT); do
2869 start_mdt $num || return
2872 start_ost || error "Unable to start OST1"
2873 start_mdt 1 -o nomgs,force
2874 mount_client $MOUNT || error "mount_client $MOUNT failed"
2877 echo "blah blah" > $MOUNT/$tfile
2878 cat $MOUNT/$tfile || error "cat $MOUNT/$tfile failed"
2880 umount_client $MOUNT || error "umount_client $MOUNT failed"
2881 stop_ost || error "Unable to stop OST1"
2882 stop_mds || error "Unable to stop MDS"
2883 stop_mds || error "Unable to stop MDS on second try"
2885 run_test 41b "mount mds with --nosvc and --nomgs on first mount"
2888 local server_version=$(lustre_version_code $SINGLEMDS)
2889 local oss_list=$(comma_list $(osts_nodes))
2891 [[ $server_version -ge $(version_code 2.6.52) ]] ||
2892 [[ $server_version -ge $(version_code 2.5.26) &&
2893 $server_version -lt $(version_code 2.5.50) ]] ||
2894 [[ $server_version -ge $(version_code 2.5.4) &&
2895 $server_version -lt $(version_code 2.5.11) ]] ||
2896 { skip "Need MDS version 2.5.4+ or 2.5.26+ or 2.6.52+"; return; }
2898 # ensure mds1 ost1 have been created even if running sub-test standalone
2901 cleanup || error "cleanup failed"
2903 # using directly mount command instead of start() function to avoid
2904 # any side effect of // with others/externals tools/features
2905 # ("zpool import", ...)
2907 # MDT concurrent start
2909 LOAD_MODULES_REMOTE=true load_modules
2910 do_facet $SINGLEMDS "lsmod | grep -q libcfs" ||
2911 error "MDT concurrent start: libcfs module not loaded"
2913 local mds1dev=$(mdsdevname 1)
2914 local mds1mnt=$(facet_mntpt mds1)
2915 local mds1fstype=$(facet_fstype mds1)
2916 local mds1opts=$MDS_MOUNT_OPTS
2918 if [ $mds1fstype == ldiskfs ] &&
2919 ! do_facet mds1 test -b $mds1dev; then
2920 mds1opts=$(csa_add "$mds1opts" -o loop)
2922 if [[ $mds1fstype == zfs ]]; then
2923 import_zpool mds1 || return ${PIPESTATUS[0]}
2926 #define OBD_FAIL_TGT_MOUNT_RACE 0x716
2927 do_facet mds1 "$LCTL set_param fail_loc=0x80000716"
2929 do_facet mds1 mount -t lustre $mds1dev $mds1mnt $mds1opts &
2932 do_facet mds1 mount -t lustre $mds1dev $mds1mnt $mds1opts
2936 do_facet mds1 "$LCTL set_param fail_loc=0x0"
2937 if [ $rc -eq 0 ] && [ $rc2 -ne 0 ]; then
2938 echo "1st MDT start succeed"
2939 echo "2nd MDT start failed with $rc2"
2940 elif [ $rc2 -eq 0 ] && [ $rc -ne 0 ]; then
2941 echo "1st MDT start failed with $rc"
2942 echo "2nd MDT start succeed"
2945 error "unexpected concurrent MDT mounts result, rc=$rc rc2=$rc2"
2948 if [ $MDSCOUNT -ge 2 ]; then
2949 for num in $(seq 2 $MDSCOUNT); do
2950 start_mdt $num || return
2954 # OST concurrent start
2956 do_rpc_nodes $oss_list "lsmod | grep -q libcfs" ||
2957 error "OST concurrent start: libcfs module not loaded"
2959 local ost1dev=$(ostdevname 1)
2960 local ost1mnt=$(facet_mntpt ost1)
2961 local ost1fstype=$(facet_fstype ost1)
2962 local ost1opts=$OST_MOUNT_OPTS
2964 if [ $ost1fstype == ldiskfs ] &&
2965 ! do_facet ost1 test -b $ost1dev; then
2966 ost1opts=$(csa_add "$ost1opts" -o loop)
2968 if [[ $ost1fstype == zfs ]]; then
2969 import_zpool ost1 || return ${PIPESTATUS[0]}
2972 #define OBD_FAIL_TGT_MOUNT_RACE 0x716
2973 do_facet ost1 "$LCTL set_param fail_loc=0x80000716"
2975 do_facet ost1 mount -t lustre $ost1dev $ost1mnt $ost1opts &
2978 do_facet ost1 mount -t lustre $ost1dev $ost1mnt $ost1opts
2982 do_facet ost1 "$LCTL set_param fail_loc=0x0"
2983 if [ $rc -eq 0 ] && [ $rc2 -ne 0 ]; then
2984 echo "1st OST start succeed"
2985 echo "2nd OST start failed with $rc2"
2986 elif [ $rc2 -eq 0 ] && [ $rc -ne 0 ]; then
2987 echo "1st OST start failed with $rc"
2988 echo "2nd OST start succeed"
2992 error "unexpected concurrent OST mounts result, rc=$rc rc2=$rc2"
2998 # verify everything ok
3003 error "MDT(s) start failed"
3011 error "OST(s) start failed"
3019 error "client start failed"
3026 error "client mount failed"
3030 run_test 41c "concurrent mounts of MDT/OST should all fail but one"
3032 test_42() { #bug 14693
3034 check_mount || error "client was not mounted"
3036 do_facet mgs $LCTL conf_param $FSNAME.llite.some_wrong_param=10
3037 umount_client $MOUNT ||
3038 error "unmounting client failed with invalid llite param"
3039 mount_client $MOUNT ||
3040 error "mounting client failed with invalid llite param"
3042 do_facet mgs $LCTL conf_param $FSNAME.sys.some_wrong_param=20
3043 cleanup || error "stopping $FSNAME failed with invalid sys param"
3045 check_mount || error "client was not mounted with invalid sys param"
3046 cleanup || error "stopping $FSNAME failed with invalid sys param"
3048 run_test 42 "allow client/server mount/unmount with invalid config param"
3051 [[ $(lustre_version_code mgs) -ge $(version_code 2.5.58) ]] ||
3052 { skip "Need MDS version at least 2.5.58" && return 0; }
3053 [ $UID -ne 0 -o $RUNAS_ID -eq 0 ] && skip_env "run as root"
3056 USER1=$(getent passwd | grep :$ID1:$ID1: | cut -d: -f1)
3057 [ -z "$USER1" ] && skip_env "missing user with uid=$ID1 gid=$ID1" &&
3061 chmod ugo+x $DIR || error "chmod 0 failed"
3062 set_conf_param_and_check mds1 \
3063 "$LCTL get_param -n mdt.$FSNAME-MDT0000.root_squash" \
3064 "$FSNAME.mdt.root_squash" \
3066 wait_update $HOSTNAME \
3067 "$LCTL get_param -n llite.${FSNAME}*.root_squash" \
3069 error "check llite root_squash failed!"
3070 set_conf_param_and_check mds1 \
3071 "$LCTL get_param -n mdt.$FSNAME-MDT0000.nosquash_nids" \
3072 "$FSNAME.mdt.nosquash_nids" \
3074 wait_update $HOSTNAME \
3075 "$LCTL get_param -n llite.${FSNAME}*.nosquash_nids" \
3077 error "check llite nosquash_nids failed!"
3080 # create set of test files
3082 echo "111" > $DIR/$tfile-userfile || error "write 1 failed"
3083 chmod go-rw $DIR/$tfile-userfile || error "chmod 1 failed"
3084 chown $RUNAS_ID.$RUNAS_ID $DIR/$tfile-userfile || error "chown failed"
3086 echo "222" > $DIR/$tfile-rootfile || error "write 2 failed"
3087 chmod go-rw $DIR/$tfile-rootfile || error "chmod 2 faield"
3089 mkdir $DIR/$tdir-rootdir || error "mkdir failed"
3090 chmod go-rwx $DIR/$tdir-rootdir || error "chmod 3 failed"
3091 touch $DIR/$tdir-rootdir/tfile-1 || error "touch failed"
3093 echo "777" > $DIR/$tfile-user1file || error "write 7 failed"
3094 chmod go-rw $DIR/$tfile-user1file || error "chmod 7 failed"
3095 chown $ID1.$ID1 $DIR/$tfile-user1file || error "chown failed"
3098 # check root_squash:
3099 # set root squash UID:GID to RUNAS_ID
3100 # root should be able to access only files owned by RUNAS_ID
3102 set_conf_param_and_check mds1 \
3103 "$LCTL get_param -n mdt.$FSNAME-MDT0000.root_squash" \
3104 "$FSNAME.mdt.root_squash" \
3105 "$RUNAS_ID:$RUNAS_ID"
3106 wait_update $HOSTNAME \
3107 "$LCTL get_param -n llite.${FSNAME}*.root_squash" \
3108 "$RUNAS_ID:$RUNAS_ID" ||
3109 error "check llite root_squash failed!"
3111 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-userfile)
3112 dd if=$DIR/$tfile-userfile 1>/dev/null 2>/dev/null ||
3113 error "$ST: root read permission is denied"
3114 echo "$ST: root read permission is granted - ok"
3117 dd conv=notrunc of=$DIR/$tfile-userfile 1>/dev/null 2>/dev/null ||
3118 error "$ST: root write permission is denied"
3119 echo "$ST: root write permission is granted - ok"
3121 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-rootfile)
3122 dd if=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null &&
3123 error "$ST: root read permission is granted"
3124 echo "$ST: root read permission is denied - ok"
3127 dd conv=notrunc of=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null &&
3128 error "$ST: root write permission is granted"
3129 echo "$ST: root write permission is denied - ok"
3131 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tdir-rootdir)
3132 rm $DIR/$tdir-rootdir/tfile-1 1>/dev/null 2>/dev/null &&
3133 error "$ST: root unlink permission is granted"
3134 echo "$ST: root unlink permission is denied - ok"
3136 touch $DIR/tdir-rootdir/tfile-2 1>/dev/null 2>/dev/null &&
3137 error "$ST: root create permission is granted"
3138 echo "$ST: root create permission is denied - ok"
3141 # check root_squash is enforced independently
3142 # of client cache content
3144 # access file by USER1, keep access open
3145 # root should be denied access to user file
3147 runas -u $ID1 tail -f $DIR/$tfile-user1file 1>/dev/null 2>&1 &
3151 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-user1file)
3152 dd if=$DIR/$tfile-user1file 1>/dev/null 2>&1 &&
3153 { kill $pid; error "$ST: root read permission is granted"; }
3154 echo "$ST: root read permission is denied - ok"
3157 dd conv=notrunc of=$DIR/$tfile-user1file 1>/dev/null 2>&1 &&
3158 { kill $pid; error "$ST: root write permission is granted"; }
3159 echo "$ST: root write permission is denied - ok"
3165 # check nosquash_nids:
3166 # put client's NID into nosquash_nids list,
3167 # root should be able to access root file after that
3169 local NIDLIST=$($LCTL list_nids all | tr '\n' ' ')
3170 NIDLIST="2@gni $NIDLIST 192.168.0.[2,10]@tcp"
3171 NIDLIST=$(echo $NIDLIST | tr -s ' ' ' ')
3172 set_conf_param_and_check mds1 \
3173 "$LCTL get_param -n mdt.$FSNAME-MDT0000.nosquash_nids" \
3174 "$FSNAME-MDTall.mdt.nosquash_nids" \
3176 wait_update $HOSTNAME \
3177 "$LCTL get_param -n llite.${FSNAME}*.nosquash_nids" \
3179 error "check llite nosquash_nids failed!"
3181 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-rootfile)
3182 dd if=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null ||
3183 error "$ST: root read permission is denied"
3184 echo "$ST: root read permission is granted - ok"
3187 dd conv=notrunc of=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null ||
3188 error "$ST: root write permission is denied"
3189 echo "$ST: root write permission is granted - ok"
3191 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tdir-rootdir)
3192 rm $DIR/$tdir-rootdir/tfile-1 ||
3193 error "$ST: root unlink permission is denied"
3194 echo "$ST: root unlink permission is granted - ok"
3195 touch $DIR/$tdir-rootdir/tfile-2 ||
3196 error "$ST: root create permission is denied"
3197 echo "$ST: root create permission is granted - ok"
3198 cleanup || error "cleanup failed with $?"
3200 run_test 43a "check root_squash and nosquash_nids"
3202 test_43b() { # LU-5690
3203 [[ $(lustre_version_code mgs) -ge $(version_code 2.7.62) ]] ||
3204 { skip "Need MGS version 2.7.62+"; return; }
3206 if [[ -z "$fs2mds_DEV" ]]; then
3207 is_blkdev $SINGLEMDS $(mdsdevname ${SINGLEMDS//mds/}) &&
3208 skip_env "mixed loopback and real device not working" && return
3211 local fs2mdsdev=$(mdsdevname 1_2)
3212 local fs2mdsvdev=$(mdsvdevname 1_2)
3214 # temporarily use fs2mds as fs2mgs
3216 local fs2mgsdev=$fs2mdsdev
3217 local fs2mgsvdev=$fs2mdsvdev
3219 local fsname=test1234
3221 load_module llite/lustre
3222 local client_ip=$(host_nids_address $HOSTNAME $NETTYPE)
3223 local host=${client_ip//*./}
3224 local net=${client_ip/%$host/}
3225 local nosquash_nids=$(h2$NETTYPE $net[$host,$host,$host])
3227 add $fs2mgs $(mkfs_opts mgs $fs2mgsdev) --fsname=$fsname \
3228 --param mdt.root_squash=$RUNAS_ID:$RUNAS_ID \
3229 --param mdt.nosquash_nids=$nosquash_nids \
3230 --reformat $fs2mgsdev $fs2mgsvdev || error "add fs2mgs failed"
3231 start $fs2mgs $fs2mgsdev $MGS_MOUNT_OPTS || error "start fs2mgs failed"
3232 stop $fs2mgs -f || error "stop fs2mgs failed"
3234 run_test 43b "parse nosquash_nids with commas in expr_list"
3236 umount_client $MOUNT
3241 check_mount || error "check_mount"
3242 UUID=$($LCTL get_param llite.${FSNAME}*.uuid | cut -d= -f2)
3244 UUIDS=$(do_facet $SINGLEMDS "$LCTL get_param mdt.${FSNAME}*.exports.*.uuid")
3245 for VAL in $UUIDS; do
3246 NID=$(echo $VAL | cut -d= -f1)
3247 CLUUID=$(echo $VAL | cut -d= -f2)
3248 [ "$UUID" = "$CLUUID" ] && STATS_FOUND=yes && break
3250 [ "$STATS_FOUND" = "no" ] && error "stats not found for client"
3251 cleanup || error "cleanup failed with $?"
3253 run_test 44 "mounted client proc entry exists"
3257 check_mount || error "check_mount"
3258 stop_mds || error "Unable to stop MDS"
3262 #define OBD_FAIL_PTLRPC_LONG_REPL_UNLINK 0x50f
3263 do_facet client "$LCTL set_param fail_loc=0x8000050f"
3266 manual_umount_client --force || error "manual_umount_client failed"
3267 do_facet client "$LCTL set_param fail_loc=0x0"
3268 start_mds || error "unable to start MDS"
3269 mount_client $MOUNT || error "mount_client $MOUNT failed"
3270 cleanup || error "cleanup failed with $?"
3272 run_test 45 "long unlink handling in ptlrpcd"
3279 umount_client $MOUNT2 || rc=$?
3280 umount_client $MOUNT || rc=$?
3281 while [ $count -gt 0 ]; do
3282 stop ost${count} -f || rc=$?
3286 cleanup_nocli || rc=$?
3287 #writeconf to remove all ost2 traces for subsequent tests
3288 writeconf_or_reformat
3293 echo "Testing with $OSTCOUNT OSTs"
3295 start_mds || error "unable to start MDS"
3296 #first client should see only one ost
3297 start_ost || error "Unable to start OST1"
3298 wait_osc_import_state mds ost FULL
3300 mount_client $MOUNT || error "mount_client $MOUNT failed"
3301 trap "cleanup_46a $OSTCOUNT" EXIT ERR
3304 for (( i=2; i<=$OSTCOUNT; i++ )); do
3305 start ost$i $(ostdevname $i) $OST_MOUNT_OPTS ||
3306 error "start_ost$i $(ostdevname $i) failed"
3309 # wait until osts in sync
3310 for (( i=2; i<=$OSTCOUNT; i++ )); do
3311 wait_osc_import_state mds ost$i FULL
3312 wait_osc_import_state client ost$i FULL
3315 #second client see all ost's
3317 mount_client $MOUNT2 || error "mount_client failed"
3318 $SETSTRIPE -c -1 $MOUNT2 || error "$SETSTRIPE -c -1 $MOUNT2 failed"
3319 $GETSTRIPE $MOUNT2 || error "$GETSTRIPE $MOUNT2 failed"
3321 echo "ok" > $MOUNT2/widestripe
3322 $GETSTRIPE $MOUNT2/widestripe ||
3323 error "$GETSTRIPE $MOUNT2/widestripe failed"
3324 # fill acl buffer for avoid expand lsm to them
3325 awk -F : '{if (FNR < 25) { print "u:"$1":rwx" }}' /etc/passwd |
3327 setfacl -m $acl $MOUNT2/widestripe
3331 stat $MOUNT/widestripe || error "stat $MOUNT/widestripe failed"
3333 cleanup_46a $OSTCOUNT || error "cleanup_46a failed"
3335 run_test 46a "handle ost additional - wide striped file"
3340 check_mount || error "check_mount failed"
3341 $LCTL set_param ldlm.namespaces.$FSNAME-*-*-*.lru_size=100
3345 for ns in $($LCTL get_param ldlm.namespaces.$FSNAME-*-*-*.lru_size); do
3346 if echo $ns | grep "MDT[[:digit:]]*"; then
3349 lrs=$(echo $ns | sed 's/.*lru_size=//')
3350 lru_size[count]=$lrs
3355 facet_failover $SINGLEMDS
3356 client_up || error "client_up failed"
3359 for ns in $($LCTL get_param ldlm.namespaces.$FSNAME-*-*-*.lru_size); do
3360 if echo $ns | grep "MDT[[:digit:]]*"; then
3363 lrs=$(echo $ns | sed 's/.*lru_size=//')
3364 if ! test "$lrs" -eq "${lru_size[count]}"; then
3365 n=$(echo $ns | sed -e 's/ldlm.namespaces.//' -e 's/.lru_size=.*//')
3366 error "$n has lost lru_size: $lrs vs. ${lru_size[count]}"
3371 cleanup || error "cleanup failed with $?"
3373 run_test 47 "server restart does not make client loss lru_resize settings"
3378 # reformat after this test is needed - if the test fails,
3379 # we will have unkillable file at FS
3383 test_48() { # bug 17636
3386 check_mount || error "check_mount failed"
3388 $SETSTRIPE -c -1 $MOUNT || error "$SETSTRIPE -c -1 $MOUNT failed"
3389 $GETSTRIPE $MOUNT || error "$GETSTRIPE $MOUNT failed"
3391 echo "ok" > $MOUNT/widestripe
3392 $GETSTRIPE $MOUNT/widestripe ||
3393 error "$GETSTRIPE $MOUNT/widestripe failed"
3395 trap cleanup_48 EXIT ERR
3397 # fill acl buffer for avoid expand lsm to them
3398 getent passwd | awk -F : '{ print "u:"$1":rwx" }' | while read acl; do
3399 setfacl -m $acl $MOUNT/widestripe
3402 stat $MOUNT/widestripe || error "stat $MOUNT/widestripe failed"
3406 run_test 48 "too many acls on file"
3408 # check PARAM_SYS_LDLM_TIMEOUT option of MKFS.LUSTRE
3409 test_49a() { # bug 17710
3410 local timeout_orig=$TIMEOUT
3411 local ldlm_timeout_orig=$LDLM_TIMEOUT
3412 local LOCAL_TIMEOUT=20
3414 LDLM_TIMEOUT=$LOCAL_TIMEOUT
3415 TIMEOUT=$LOCAL_TIMEOUT
3419 check_mount || error "client mount failed"
3421 echo "check ldlm_timout..."
3422 local LDLM_MDS="$(do_facet $SINGLEMDS $LCTL get_param -n ldlm_timeout)"
3423 local LDLM_OST1="$(do_facet ost1 $LCTL get_param -n ldlm_timeout)"
3424 local LDLM_CLIENT="$(do_facet client $LCTL get_param -n ldlm_timeout)"
3426 if [ $LDLM_MDS -ne $LDLM_OST1 -o $LDLM_MDS -ne $LDLM_CLIENT ]; then
3427 error "Different LDLM_TIMEOUT:$LDLM_MDS $LDLM_OST1 $LDLM_CLIENT"
3430 if [ $LDLM_MDS -ne $((LOCAL_TIMEOUT / 3)) ]; then
3431 error "LDLM_TIMEOUT($LDLM_MDS) is not $((LOCAL_TIMEOUT / 3))"
3434 umount_client $MOUNT || error "umount_client $MOUNT failed"
3435 stop_ost || error "problem stopping OSS"
3436 stop_mds || error "problem stopping MDS"
3438 LDLM_TIMEOUT=$ldlm_timeout_orig
3439 TIMEOUT=$timeout_orig
3441 run_test 49a "check PARAM_SYS_LDLM_TIMEOUT option of mkfs.lustre"
3443 test_49b() { # bug 17710
3444 local timeout_orig=$TIMEOUT
3445 local ldlm_timeout_orig=$LDLM_TIMEOUT
3446 local LOCAL_TIMEOUT=20
3448 LDLM_TIMEOUT=$((LOCAL_TIMEOUT - 1))
3449 TIMEOUT=$LOCAL_TIMEOUT
3453 check_mount || error "client mount failed"
3455 local LDLM_MDS="$(do_facet $SINGLEMDS $LCTL get_param -n ldlm_timeout)"
3456 local LDLM_OST1="$(do_facet ost1 $LCTL get_param -n ldlm_timeout)"
3457 local LDLM_CLIENT="$(do_facet client $LCTL get_param -n ldlm_timeout)"
3459 if [ $LDLM_MDS -ne $LDLM_OST1 -o $LDLM_MDS -ne $LDLM_CLIENT ]; then
3460 error "Different LDLM_TIMEOUT:$LDLM_MDS $LDLM_OST1 $LDLM_CLIENT"
3463 if [ $LDLM_MDS -ne $((LOCAL_TIMEOUT - 1)) ]; then
3464 error "LDLM_TIMEOUT($LDLM_MDS) is not $((LOCAL_TIMEOUT - 1))"
3467 cleanup || error "cleanup failed"
3469 LDLM_TIMEOUT=$ldlm_timeout_orig
3470 TIMEOUT=$timeout_orig
3472 run_test 49b "check PARAM_SYS_LDLM_TIMEOUT option of mkfs.lustre"
3475 # Test both statfs and lfs df and fail if either one fails
3476 multiop_bg_pause $1 f_
3479 killall -USR1 multiop
3480 [ $RC1 -ne 0 ] && log "lazystatfs multiop failed"
3481 wait $PID || { RC1=$?; log "multiop return error "; }
3488 if [ $RC2 -eq 0 ]; then
3490 log "lazystatfs df failed"
3494 [[ $RC1 -ne 0 || $RC2 -eq 0 ]] && RC=1
3500 $LCTL set_param llite.$FSNAME-*.lazystatfs=1
3501 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
3503 lazystatfs $MOUNT || error "lazystatfs failed but no down servers"
3505 cleanup || error "cleanup failed with rc $?"
3507 run_test 50a "lazystatfs all servers available"
3511 $LCTL set_param llite.$FSNAME-*.lazystatfs=1
3512 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
3514 # Wait for client to detect down OST
3515 stop_ost || error "Unable to stop OST1"
3516 wait_osc_import_state mds ost DISCONN
3518 lazystatfs $MOUNT || error "lazystatfs should not return EIO"
3520 umount_client $MOUNT || error "Unable to unmount client"
3521 stop_mds || error "Unable to stop MDS"
3523 run_test 50b "lazystatfs all servers down"
3526 start_mds || error "Unable to start MDS"
3527 start_ost || error "Unable to start OST1"
3528 start_ost2 || error "Unable to start OST2"
3529 mount_client $MOUNT || error "Unable to mount client"
3530 $LCTL set_param llite.$FSNAME-*.lazystatfs=1
3531 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
3533 # Wait for client to detect down OST
3534 stop_ost || error "Unable to stop OST1"
3535 wait_osc_import_state mds ost DISCONN
3536 lazystatfs $MOUNT || error "lazystatfs failed with one down server"
3538 umount_client $MOUNT || error "Unable to unmount client"
3539 stop_ost2 || error "Unable to stop OST2"
3540 stop_mds || error "Unable to stop MDS"
3541 #writeconf to remove all ost2 traces for subsequent tests
3542 writeconf_or_reformat
3544 run_test 50c "lazystatfs one server down"
3547 start_mds || error "Unable to start MDS"
3548 start_ost || error "Unable to start OST1"
3549 start_ost2 || error "Unable to start OST2"
3550 mount_client $MOUNT || error "Unable to mount client"
3551 $LCTL set_param llite.$FSNAME-*.lazystatfs=1
3552 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
3554 # Issue the statfs during the window where the client still
3555 # belives the OST to be available but it is in fact down.
3556 # No failure just a statfs which hangs for a timeout interval.
3557 stop_ost || error "Unable to stop OST1"
3558 lazystatfs $MOUNT || error "lazystatfs failed with one down server"
3560 umount_client $MOUNT || error "Unable to unmount client"
3561 stop_ost2 || error "Unable to stop OST2"
3562 stop_mds || error "Unable to stop MDS"
3563 #writeconf to remove all ost2 traces for subsequent tests
3564 writeconf_or_reformat
3566 run_test 50d "lazystatfs client/server conn race"
3573 start_mds || error "Unable to start MDS"
3574 #first client should see only one ost
3575 start_ost || error "Unable to start OST1"
3576 wait_osc_import_state mds ost FULL
3578 # Wait for client to detect down OST
3579 stop_ost || error "Unable to stop OST1"
3580 wait_osc_import_state mds ost DISCONN
3582 mount_client $MOUNT || error "Unable to mount client"
3583 $LCTL set_param llite.$FSNAME-*.lazystatfs=0
3585 multiop_bg_pause $MOUNT _f
3589 if [ $RC1 -ne 0 ]; then
3590 log "multiop failed $RC1"
3593 sleep $(( $TIMEOUT+1 ))
3595 [ $? -ne 0 ] && error "process isn't sleep"
3596 start_ost || error "Unable to start OST1"
3597 wait $pid || error "statfs failed"
3600 umount_client $MOUNT || error "Unable to unmount client"
3601 stop_ost || error "Unable to stop OST1"
3602 stop_mds || error "Unable to stop MDS"
3604 run_test 50e "normal statfs all servers down"
3609 CONN_PROC="osc.$FSNAME-OST0001-osc-[M]*.ost_server_uuid"
3611 start_mds || error "Unable to start mds"
3612 #first client should see only one ost
3613 start_ost || error "Unable to start OST1"
3614 wait_osc_import_state mds ost FULL
3616 start_ost2 || error "Unable to start OST2"
3617 wait_osc_import_state mds ost2 FULL
3619 # Wait for client to detect down OST
3620 stop_ost2 || error "Unable to stop OST2"
3622 wait_osc_import_state mds ost2 DISCONN
3623 mount_client $MOUNT || error "Unable to mount client"
3624 $LCTL set_param llite.$FSNAME-*.lazystatfs=0
3626 multiop_bg_pause $MOUNT _f
3630 if [ $RC1 -ne 0 ]; then
3631 log "lazystatfs multiop failed $RC1"
3634 sleep $(( $TIMEOUT+1 ))
3636 [ $? -ne 0 ] && error "process isn't sleep"
3637 start_ost2 || error "Unable to start OST2"
3638 wait $pid || error "statfs failed"
3639 stop_ost2 || error "Unable to stop OST2"
3642 umount_client $MOUNT -f || error "Unable to unmount client"
3643 stop_ost || error "Unable to stop OST1"
3644 stop_mds || error "Unable to stop MDS"
3645 #writeconf to remove all ost2 traces for subsequent tests
3646 writeconf_or_reformat
3648 run_test 50f "normal statfs one server in down"
3651 [ "$OSTCOUNT" -lt "2" ] && skip_env "$OSTCOUNT < 2, skipping" && return
3653 start_ost2 || error "Unable to start OST2"
3654 wait_osc_import_state mds ost2 FULL
3655 wait_osc_import_state client ost2 FULL
3657 local PARAM="${FSNAME}-OST0001.osc.active"
3659 $SETSTRIPE -c -1 $DIR/$tfile || error "$SETSTRIPE failed"
3660 do_facet mgs $LCTL conf_param $PARAM=0 ||
3661 error "Unable to deactivate OST"
3663 umount_client $MOUNT || error "Unable to unmount client"
3664 mount_client $MOUNT || error "Unable to mount client"
3665 # This df should not cause a panic
3668 do_facet mgs $LCTL conf_param $PARAM=1 || error "Unable to activate OST"
3669 rm -f $DIR/$tfile || error "unable to remove file $DIR/$tfile"
3670 umount_client $MOUNT || error "Unable to unmount client"
3671 stop_ost2 || error "Unable to stop OST2"
3672 stop_ost || error "Unable to stop OST1"
3673 stop_mds || error "Unable to stop MDS"
3674 #writeconf to remove all ost2 traces for subsequent tests
3675 writeconf_or_reformat
3677 run_test 50g "deactivated OST should not cause panic"
3681 # prepare MDT/OST, make OSC inactive for OST1
3682 [ "$OSTCOUNT" -lt "2" ] && skip_env "$OSTCOUNT < 2, skipping" && return
3684 [ $(facet_fstype ost1) == zfs ] && import_zpool ost1
3685 do_facet ost1 "$TUNEFS --param osc.active=0 `ostdevname 1`" ||
3686 error "tunefs OST1 failed"
3687 start_mds || error "Unable to start MDT"
3688 start_ost || error "Unable to start OST1"
3689 start_ost2 || error "Unable to start OST2"
3690 mount_client $MOUNT || error "client start failed"
3692 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
3694 # activatate OSC for OST1
3695 local TEST="$LCTL get_param -n osc.${FSNAME}-OST0000-osc-[!M]*.active"
3696 set_conf_param_and_check client \
3697 "$TEST" "${FSNAME}-OST0000.osc.active" 1 ||
3698 error "Unable to activate OST1"
3700 mkdir $DIR/$tdir/2 || error "mkdir $DIR/$tdir/2 failed"
3701 $SETSTRIPE -c -1 -i 0 $DIR/$tdir/2 ||
3702 error "$SETSTRIPE $DIR/$tdir/2 failed"
3703 sleep 1 && echo "create a file after OST1 is activated"
3705 createmany -o $DIR/$tdir/2/$tfile-%d 1
3707 # check OSC import is working
3708 stat $DIR/$tdir/2/* >/dev/null 2>&1 ||
3709 error "some OSC imports are still not connected"
3712 umount_client $MOUNT || error "Unable to umount client"
3713 stop_ost2 || error "Unable to stop OST2"
3714 cleanup_nocli || error "cleanup_nocli failed with $?"
3716 run_test 50h "LU-642: activate deactivated OST"
3719 # prepare MDT/OST, make OSC inactive for OST1
3720 [ "$MDSCOUNT" -lt "2" ] && skip_env "$MDSCOUNT < 2, skipping" && return
3722 [ $(facet_fstype mds2) == zfs ] && import_zpool mds2
3723 do_facet mds2 "$TUNEFS --param mdc.active=0 $(mdsdevname 2)" ||
3724 error "tunefs MDT2 failed"
3725 start_mds || error "Unable to start MDT"
3726 start_ost || error "Unable to start OST1"
3727 start_ost2 || error "Unable to start OST2"
3728 mount_client $MOUNT || error "client start failed"
3730 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
3732 $LCTL conf_param ${FSNAME}-MDT0000.mdc.active=0 &&
3733 error "deactive MDC0 succeeds"
3734 # activate MDC for MDT2
3735 local TEST="$LCTL get_param -n mdc.${FSNAME}-MDT0001-mdc-[!M]*.active"
3736 set_conf_param_and_check client \
3737 "$TEST" "${FSNAME}-MDT0001.mdc.active" 1 ||
3738 error "Unable to activate MDT2"
3740 wait_clients_import_state ${CLIENTS:-$HOSTNAME} mds2 FULL
3741 if [ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.7.60) ]
3743 wait_dne_interconnect
3745 $LFS mkdir -i1 $DIR/$tdir/2 || error "mkdir $DIR/$tdir/2 failed"
3747 createmany -o $DIR/$tdir/2/$tfile-%d 1 || error "create files failed"
3749 rm -rf $DIR/$tdir/2 || error "unlink dir failed"
3751 # deactivate MDC for MDT2
3752 local TEST="$LCTL get_param -n mdc.${FSNAME}-MDT0001-mdc-[!M]*.active"
3753 set_conf_param_and_check client \
3754 "$TEST" "${FSNAME}-MDT0001.mdc.active" 0 ||
3755 error "Unable to deactivate MDT2"
3757 $LFS mkdir -i1 $DIR/$tdir/2 &&
3758 error "mkdir $DIR/$tdir/2 succeeds after deactive MDT"
3761 umount_client $MOUNT || error "Unable to umount client"
3766 run_test 50i "activate deactivated MDT"
3769 local LOCAL_TIMEOUT=20
3773 check_mount || error "check_mount failed"
3775 mkdir $MOUNT/$tdir || error "mkdir $MOUNT/$tdir failed"
3776 $SETSTRIPE -c -1 $MOUNT/$tdir ||
3777 error "$SETSTRIPE -c -1 $MOUNT/$tdir failed"
3778 #define OBD_FAIL_MDS_REINT_DELAY 0x142
3779 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0x142"
3780 touch $MOUNT/$tdir/$tfile &
3783 start_ost2 || error "Unable to start OST1"
3785 stop_ost2 || error "Unable to stop OST1"
3786 umount_client $MOUNT -f || error “unmount $MOUNT failed”
3787 cleanup_nocli || error “stop server failed”
3788 #writeconf to remove all ost2 traces for subsequent tests
3789 writeconf_or_reformat
3791 run_test 51 "Verify that mdt_reint handles RMF_MDT_MD correctly when an OST is added"
3800 do_node $node mkdir -p $dest
3801 [ $? -eq 0 ] || { error "Unable to create directory"; return 1; }
3803 do_node $node 'tar cf - '$@' | tar xf - -C '$dest';
3804 [ \"\${PIPESTATUS[*]}\" = \"0 0\" ] || exit 1'
3805 [ $? -eq 0 ] || { error "Unable to tar files"; return 2; }
3807 do_node $node 'getfattr -d -m "[a-z]*\\." '$@' > '$xattrs
3808 [ $? -eq 0 ] || { error "Unable to read xattrs"; return 3; }
3818 local backup2=${TMP}/backup2
3820 do_node $node mkdir -p $backup2
3821 [ $? -eq 0 ] || { error "Unable to create directory"; return 1; }
3823 do_node $node 'tar cf - '$@' | tar xf - -C '$backup2';
3824 [ \"\${PIPESTATUS[*]}\" = \"0 0\" ] || exit 1'
3825 [ $? -eq 0 ] || { error "Unable to tar files to diff"; return 2; }
3827 do_node $node "diff -rq $backup $backup2"
3828 [ $? -eq 0 ] || { error "contents differ"; return 3; }
3830 local xattrs2=${TMP}/xattrs2
3831 do_node $node 'getfattr -d -m "[a-z]*\\." '$@' > '$xattrs2
3832 [ $? -eq 0 ] || { error "Unable to read xattrs to diff"; return 4; }
3834 do_node $node "diff $xattrs $xattrs2"
3835 [ $? -eq 0 ] || { error "xattrs differ"; return 5; }
3837 do_node $node "rm -rf $backup2 $xattrs2"
3838 [ $? -eq 0 ] || { error "Unable to delete temporary files"; return 6; }
3842 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
3843 skip "Only applicable to ldiskfs-based MDTs"
3847 start_mds || error "Unable to start MDS"
3848 start_ost || error "Unable to start OST1"
3849 mount_client $MOUNT || error "Unable to mount client"
3852 local ost1mnt=$(facet_mntpt ost1)
3853 local ost1node=$(facet_active_host ost1)
3854 local ost1tmp=$TMP/conf52
3857 mkdir $DIR/$tdir || error "Unable to create $DIR/$tdir"
3858 touch $TMP/modified_first || error "Unable to create temporary file"
3859 local mtime=$(stat -c %Y $TMP/modified_first)
3860 do_node $ost1node "mkdir -p $ost1tmp &&
3861 touch -m -d @$mtime $ost1tmp/modified_first" ||
3862 error "Unable to create temporary file"
3865 $SETSTRIPE -c -1 -S 1M $DIR/$tdir || error "$SETSTRIPE failed"
3867 for (( i=0; i < nrfiles; i++ )); do
3868 multiop $DIR/$tdir/$tfile-$i Ow1048576w1048576w524288c ||
3869 error "multiop failed"
3875 echo backup files to $TMP/$tdir
3876 local files=$(find $DIR/$tdir -type f -newer $TMP/modified_first)
3877 copy_files_xattrs $(hostname) $TMP/$tdir $TMP/file_xattrs $files ||
3878 error "Unable to copy files"
3880 umount_client $MOUNT || error "Unable to umount client"
3881 stop_ost || error "Unable to stop ost1"
3883 echo mount ost1 as ldiskfs
3884 do_node $ost1node mkdir -p $ost1mnt || error "Unable to create $ost1mnt"
3885 if ! do_node $ost1node test -b $ost1_dev; then
3888 do_node $ost1node mount -t $(facet_fstype ost1) $loop $ost1_dev \
3890 error "Unable to mount ost1 as ldiskfs"
3893 echo backup objects to $ost1tmp/objects
3894 local objects=$(do_node $ost1node 'find '$ost1mnt'/O/[0-9]* -type f'\
3895 '-size +0 -newer '$ost1tmp'/modified_first -regex ".*\/[0-9]+"')
3896 copy_files_xattrs $ost1node $ost1tmp/objects $ost1tmp/object_xattrs \
3898 error "Unable to copy objects"
3900 # move objects to lost+found
3901 do_node $ost1node 'mv '$objects' '${ost1mnt}'/lost+found'
3902 [ $? -eq 0 ] || { error "Unable to move objects"; return 14; }
3904 do_node $ost1node "umount $ost1mnt" ||
3905 error "Unable to umount ost1 as ldiskfs"
3907 start_ost || error "Unable to start OST1"
3908 mount_client $MOUNT || error "Unable to mount client"
3910 local REPAIRED=$(do_node $ost1node "$LCTL get_param \
3911 -n osd-ldiskfs.$FSNAME-OST0000.oi_scrub" |
3912 awk '/^lf_repa[ri]*ed/ { print $2 }')
3913 [ $REPAIRED -gt 0 ] ||
3914 error "Some entry under /lost+found should be repaired"
3917 diff_files_xattrs $(hostname) $TMP/$tdir $TMP/file_xattrs $files ||
3918 error "Unable to diff files"
3920 rm -rf $TMP/$tdir $TMP/file_xattrs ||
3921 error "Unable to delete temporary files"
3922 do_node $ost1node "rm -rf $ost1tmp" ||
3923 error "Unable to delete temporary files"
3924 cleanup || error "cleanup failed with $?"
3926 run_test 52 "check recovering objects from lost+found"
3928 # Checks threads_min/max/started for some service
3930 # Arguments: service name (OST or MDT), facet (e.g., ost1, $SINGLEMDS), and a
3931 # parameter pattern prefix like 'ost.*.ost'.
3944 local msg="Insane $modname thread counts"
3945 local ncpts=$(check_cpt_number $facet)
3949 check_mount || return 41
3951 # We need to expand $parampat, but it may match multiple parameters, so
3952 # we'll pick the first one
3953 if ! paramp=$(do_facet $facet "lctl get_param -N ${parampat}.threads_min"|head -1); then
3954 error "Couldn't expand ${parampat}.threads_min parameter name"
3958 # Remove the .threads_min part
3959 paramp=${paramp%.threads_min}
3961 # Check for sanity in defaults
3962 tmin=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min" ||
3964 tmax=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_max" ||
3966 tstarted=$(do_facet $facet "$LCTL get_param \
3967 -n ${paramp}.threads_started" || echo 0)
3968 lassert 23 "$msg (PDSH problems?)" '(($tstarted && $tmin && $tmax))' ||
3970 lassert 24 "$msg" '(($tstarted >= $tmin && $tstarted <= $tmax ))' ||
3972 nthrs=$(expr $tmax - $tmin)
3973 if [ $nthrs -lt $ncpts ]; then
3979 [ $tmin -eq $tmax -a $tmin -eq $tstarted ] &&
3980 skip_env "module parameter forced $facet thread count" &&
3981 tmin=3 && tmax=$((3 * tmax))
3983 # Check that we can change min/max
3984 do_facet $facet "$LCTL set_param \
3985 ${paramp}.threads_min=$((tmin + nthrs))"
3986 do_facet $facet "$LCTL set_param \
3987 ${paramp}.threads_max=$((tmax - nthrs))"
3988 tmin2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min" ||
3990 tmax2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_max" ||
3992 lassert 25 "$msg" '(($tmin2 == ($tmin + $nthrs) &&
3993 $tmax2 == ($tmax - $nthrs)))' || return $?
3995 # Check that we can set min/max to the same value
3996 tmin=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min" ||
3998 do_facet $facet "$LCTL set_param ${paramp}.threads_max=$tmin"
3999 tmin2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min" ||
4001 tmax2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_max" ||
4003 lassert 26 "$msg" '(($tmin2 == $tmin && $tmax2 == $tmin))' || return $?
4005 # Check that we can't set max < min
4006 do_facet $facet "$LCTL set_param ${paramp}.threads_max=$((tmin - 1))"
4007 tmin2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min" ||
4009 tmax2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_max" ||
4011 lassert 27 "$msg" '(($tmin2 <= $tmax2))' || return $?
4013 # We need to ensure that we get the module options desired; to do this
4014 # we set LOAD_MODULES_REMOTE=true and we call setmodopts below.
4015 LOAD_MODULES_REMOTE=true
4018 local newvalue="${opts}=$(expr $basethr \* $ncpts)"
4019 setmodopts -a $modname "$newvalue" oldvalue
4022 check_mount || return 41
4024 # Restore previous setting of MODOPTS_*
4025 setmodopts $modname "$oldvalue"
4027 # Check that $opts took
4028 tmin=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min")
4029 tmax=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_max")
4030 tstarted=$(do_facet $facet \
4031 "$LCTL get_param -n ${paramp}.threads_started")
4032 lassert 28 "$msg" '(($tstarted >= $tmin && $tstarted <= $tmax ))' ||
4041 thread_sanity OST ost1 'ost.*.ost' 'oss_num_threads' '16'
4042 cleanup || error "cleanup failed with rc $?"
4044 run_test 53a "check OSS thread count params"
4048 local mds=$(do_facet $SINGLEMDS "$LCTL get_param \
4049 -N mds.*.*.threads_max 2>/dev/null")
4050 if [ -z "$mds" ]; then
4051 #running this on an old MDT
4052 thread_sanity MDT $SINGLEMDS 'mdt.*.*.' 'mdt_num_threads' 16
4054 thread_sanity MDT $SINGLEMDS 'mds.*.*.' 'mds_num_threads' 16
4056 cleanup || error "cleanup failed with $?"
4058 run_test 53b "check MDS thread count params"
4061 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
4062 skip "Only applicable to ldiskfs-based MDTs"
4066 do_rpc_nodes $(facet_host ost1) run_llverdev $(ostdevname 1) -p ||
4067 error "llverdev failed with rc=$?"
4070 run_test 54a "test llverdev and partial verify of device"
4073 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
4074 skip "Only applicable to ldiskfs-based MDTs"
4079 run_llverfs $MOUNT -p || error "llverfs failed with rc=$?"
4080 cleanup || error "cleanup failed with rc=$?"
4082 run_test 54b "test llverfs and partial verify of filesystem"
4086 local max_ost_index=$1
4087 echo -n $(((max_ost_index + 1) * 8))
4091 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
4092 skip "Only applicable to ldiskfs-based MDTs"
4096 local mdsdev=$(mdsdevname 1)
4097 local mdsvdev=$(mdsvdevname 1)
4101 if ! combined_mgs_mds; then
4102 stop_mgs || error "stopping MGS service failed"
4103 format_mgs || error "formatting MGT failed"
4105 add mds1 $(mkfs_opts mds1 ${mdsdev}) --reformat $mdsdev \
4107 add ost1 $(mkfs_opts ost1 $(ostdevname 1)) --index=$i \
4108 --reformat $(ostdevname 1) $(ostvdevname 1)
4114 echo checking size of lov_objid for ost index $i
4115 LOV_OBJID_SIZE=$(do_facet mds1 "$DEBUGFS -R 'stat lov_objid' $mdsdev 2>/dev/null" | grep ^User | awk '{print $6}')
4116 if [ "$LOV_OBJID_SIZE" != $(lov_objid_size $i) ]; then
4117 error "lov_objid size has to be $(lov_objid_size $i), not $LOV_OBJID_SIZE"
4119 echo ok, lov_objid size is correct: $LOV_OBJID_SIZE
4126 run_test 55 "check lov_objid size"
4129 local server_version=$(lustre_version_code $SINGLEMDS)
4130 local mds_journal_size_orig=$MDSJOURNALSIZE
4135 for num in $(seq 1 $MDSCOUNT); do
4138 add ost1 $(mkfs_opts ost1 $(ostdevname 1)) --index=10000 --reformat \
4139 $(ostdevname 1) $(ostvdevname 1)
4140 add ost2 $(mkfs_opts ost2 $(ostdevname 2)) --index=1000 --reformat \
4141 $(ostdevname 2) $(ostvdevname 2)
4144 start_ost || error "Unable to start first ost (idx 10000)"
4145 start_ost2 || error "Unable to start second ost (idx 1000)"
4146 mount_client $MOUNT || error "Unable to mount client"
4150 if [[ $server_version -ge $(version_code 2.6.54) ]] ||
4151 [[ $server_version -ge $(version_code 2.5.4) &&
4152 $server_version -lt $(version_code 2.5.11) ]]; then
4153 wait_osc_import_state mds ost1 FULL
4154 wait_osc_import_state mds ost2 FULL
4155 $SETSTRIPE --stripe-count=-1 $DIR/$tfile ||
4156 error "Unable to setstripe $DIR/$tfile"
4157 n=$($LFS getstripe --stripe-count $DIR/$tfile)
4158 [ "$n" -eq 2 ] || error "Stripe count not two: $n"
4163 MDSJOURNALSIZE=$mds_journal_size_orig
4166 run_test 56a "check big OST indexes and out-of-index-order start"
4171 umount_client $MOUNT -f || error "unmount client failed"
4180 [ $MDSCOUNT -lt 3 ] && skip "needs >= 3 MDTs" && return
4182 trap cleanup_56b EXIT RETURN ERR
4185 if ! combined_mgs_mds ; then
4190 add mds1 $(mkfs_opts mds1 $(mdsdevname 1)) --index=0 --reformat \
4191 $(mdsdevname 1) $(mdsvdevname 1)
4192 add mds2 $(mkfs_opts mds2 $(mdsdevname 2)) --index=1 --reformat \
4193 $(mdsdevname 2) $(mdsvdevname 2)
4194 add mds3 $(mkfs_opts mds3 $(mdsdevname 3)) --index=1000 --reformat \
4195 $(mdsdevname 3) $(mdsvdevname 3)
4199 start_mdt 1 || error "MDT 1 (idx 0) start failed"
4200 start_mdt 2 || error "MDT 2 (idx 1) start failed"
4201 start_mdt 3 || error "MDT 3 (idx 1000) start failed"
4202 start_ost || error "Unable to start first ost"
4203 start_ost2 || error "Unable to start second ost"
4205 do_nodes $(comma_list $(mdts_nodes)) \
4206 "$LCTL set_param mdt.*.enable_remote_dir=1 \
4207 mdt.*.enable_remote_dir_gid=-1"
4209 mount_client $MOUNT || error "Unable to mount client"
4211 $LFS mkdir -c3 $MOUNT/$tdir || error "failed to make testdir"
4213 echo "This is test file 1!" > $MOUNT/$tdir/$tfile.1 ||
4214 error "failed to make test file 1"
4215 echo "This is test file 2!" > $MOUNT/$tdir/$tfile.2 ||
4216 error "failed to make test file 2"
4217 echo "This is test file 1000!" > $MOUNT/$tdir/$tfile.1000 ||
4218 error "failed to make test file 1000"
4220 rm -rf $MOUNT/$tdir || error "failed to remove testdir"
4222 $LFS mkdir -i1000 $MOUNT/$tdir.1000 ||
4223 error "create remote dir at idx 1000 failed"
4226 echo "=== START lfs df OUTPUT ==="
4228 echo "==== END lfs df OUTPUT ===="
4230 mdtcnt=$(echo -e "$output" | grep $FSNAME-MDT | wc -l)
4231 ostcnt=$(echo -e "$output" | grep $FSNAME-OST | wc -l)
4233 echo "lfs df returned mdt count $mdtcnt and ost count $ostcnt"
4234 [ $mdtcnt -eq 3 ] || error "lfs df returned wrong mdt count"
4235 [ $ostcnt -eq 2 ] || error "lfs df returned wrong ost count"
4237 echo "This is test file 1!" > $MOUNT/$tdir.1000/$tfile.1 ||
4238 error "failed to make test file 1"
4239 echo "This is test file 2!" > $MOUNT/$tdir.1000/$tfile.2 ||
4240 error "failed to make test file 2"
4241 echo "This is test file 1000!" > $MOUNT/$tdir.1000/$tfile.1000 ||
4242 error "failed to make test file 1000"
4243 rm -rf $MOUNT/$tdir.1000 || error "failed to remove remote_dir"
4246 echo "=== START lfs mdts OUTPUT ==="
4248 echo "==== END lfs mdts OUTPUT ===="
4250 echo -e "$output" | grep -v "MDTS:" | awk '{print $1}' |
4251 sed 's/://g' > $TMP/mdts-actual.txt
4252 sort $TMP/mdts-actual.txt -o $TMP/mdts-actual.txt
4254 echo -e "0\n1\n1000" > $TMP/mdts-expected.txt
4256 diff $TMP/mdts-expected.txt $TMP/mdts-actual.txt
4259 rm $TMP/mdts-expected.txt $TMP/mdts-actual.txt
4261 [ $result -eq 0 ] || error "target_obd proc file is incorrect!"
4263 run_test 56b "test target_obd correctness with nonconsecutive MDTs"
4265 test_57a() { # bug 22656
4266 do_rpc_nodes $(facet_active_host ost1) load_modules_local
4267 local NID=$(do_facet ost1 "$LCTL get_param nis" |
4268 tail -1 | awk '{print $1}')
4269 writeconf_or_reformat
4270 [ $(facet_fstype ost1) == zfs ] && import_zpool ost1
4271 do_facet ost1 "$TUNEFS --failnode=$NID `ostdevname 1`" ||
4272 error "tunefs failed"
4274 start_ost && error "OST registration from failnode should fail"
4277 run_test 57a "initial registration from failnode should fail (should return errs)"
4280 do_rpc_nodes $(facet_active_host ost1) load_modules_local
4281 local NID=$(do_facet ost1 "$LCTL get_param nis" |
4282 tail -1 | awk '{print $1}')
4283 writeconf_or_reformat
4284 [ $(facet_fstype ost1) == zfs ] && import_zpool ost1
4285 do_facet ost1 "$TUNEFS --servicenode=$NID `ostdevname 1`" ||
4286 error "tunefs failed"
4288 start_ost || error "OST registration from servicenode should not fail"
4291 run_test 57b "initial registration from servicenode should not fail"
4294 do_facet mgs $LCTL get_param mgs.MGS.live.$FSNAME | grep OST | wc -l
4297 test_58() { # bug 22658
4298 combined_mgs_mds || stop_mgs || error "stopping MGS service failed"
4300 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
4301 createmany -o $DIR/$tdir/$tfile-%d 100
4302 # make sure that OSTs do not cancel llog cookies before we unmount the MDS
4303 #define OBD_FAIL_OBD_LOG_CANCEL_NET 0x601
4304 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0x601"
4305 unlinkmany $DIR/$tdir/$tfile-%d 100
4306 stop_mds || error "Unable to stop MDS"
4308 local MNTDIR=$(facet_mntpt $SINGLEMDS)
4309 local devname=$(mdsdevname ${SINGLEMDS//mds/})
4311 # remove all files from the OBJECTS dir
4312 mount_fstype $SINGLEMDS
4314 do_facet $SINGLEMDS "find $MNTDIR/O/1/d* -type f -delete"
4316 unmount_fstype $SINGLEMDS
4317 # restart MDS with missing llog files
4318 start_mds || error "unable to start MDS"
4319 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0"
4322 run_test 58 "missing llog files must not prevent MDT from mounting"
4325 start_mgsmds >> /dev/null
4326 local C1=$(count_osts)
4327 if [ $C1 -eq 0 ]; then
4328 start_ost >> /dev/null
4332 echo "original ost count: $C1 (expect > 0)"
4333 [ $C1 -gt 0 ] || error "No OSTs in $FSNAME log"
4334 start_mgsmds -o writeconf >> /dev/null || error "MDT start failed"
4335 local C2=$(count_osts)
4336 echo "after mdt writeconf count: $C2 (expect 0)"
4337 [ $C2 -gt 0 ] && error "MDT writeconf should erase OST logs"
4338 echo "OST start without writeconf should fail:"
4339 start_ost >> /dev/null &&
4340 error "OST start without writeconf didn't fail"
4341 echo "OST start with writeconf should succeed:"
4342 start_ost -o writeconf >> /dev/null || error "OST1 start failed"
4343 local C3=$(count_osts)
4344 echo "after ost writeconf count: $C3 (expect 1)"
4345 [ $C3 -eq 1 ] || error "new OST writeconf should add:"
4346 start_ost2 -o writeconf >> /dev/null || error "OST2 start failed"
4347 local C4=$(count_osts)
4348 echo "after ost2 writeconf count: $C4 (expect 2)"
4349 [ $C4 -eq 2 ] || error "OST2 writeconf should add log"
4350 stop_ost2 >> /dev/null
4351 cleanup_nocli >> /dev/null
4352 #writeconf to remove all ost2 traces for subsequent tests
4353 writeconf_or_reformat
4355 run_test 59 "writeconf mount option"
4357 test_60() { # LU-471
4360 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
4361 skip "Only applicable to ldiskfs-based MDTs"
4365 for num in $(seq $MDSCOUNT); do
4366 add mds${num} $(mkfs_opts mds${num} $(mdsdevname $num)) \
4367 --mkfsoptions='\" -E stride=64 -O ^uninit_bg\"' \
4368 --reformat $(mdsdevname $num) $(mdsvdevname $num) ||
4372 dump=$(do_facet $SINGLEMDS dumpe2fs $(mdsdevname 1))
4373 [ ${PIPESTATUS[0]} -eq 0 ] || error "dumpe2fs $(mdsdevname 1) failed"
4375 # MDT default has dirdata feature
4376 echo $dump | grep dirdata > /dev/null || error "dirdata is not set"
4377 # we disable uninit_bg feature
4378 echo $dump | grep uninit_bg > /dev/null && error "uninit_bg is set"
4379 # we set stride extended options
4380 echo $dump | grep stride > /dev/null || error "stride is not set"
4383 run_test 60 "check mkfs.lustre --mkfsoptions -E -O options setting"
4388 [ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.1.53) ] ||
4389 { skip "Need MDS version at least 2.1.53"; return 0; }
4391 if [ $(facet_fstype $SINGLEMDS) == ldiskfs ] &&
4392 ! large_xattr_enabled; then
4395 for num in $(seq $MDSCOUNT); do
4396 do_facet mds${num} $TUNE2FS -O large_xattr \
4397 $(mdsdevname $num) ||
4398 error "tune2fs on mds $num failed"
4402 combined_mgs_mds || stop_mgs || error "stopping MGS service failed"
4403 setup_noconfig || error "setting up the filesystem failed"
4404 client_up || error "starting client failed"
4406 local file=$DIR/$tfile
4407 touch $file || error "touch $file failed"
4409 local large_value="$(generate_string $(max_xattr_size))"
4410 local small_value="bar"
4412 local name="trusted.big"
4413 log "save large xattr $name on $file"
4414 setfattr -n $name -v $large_value $file ||
4415 error "saving $name on $file failed"
4417 local new_value=$(get_xattr_value $name $file)
4418 [[ "$new_value" != "$large_value" ]] &&
4419 error "$name different after saving"
4421 log "shrink value of $name on $file"
4422 setfattr -n $name -v $small_value $file ||
4423 error "shrinking value of $name on $file failed"
4425 new_value=$(get_xattr_value $name $file)
4426 [[ "$new_value" != "$small_value" ]] &&
4427 error "$name different after shrinking"
4429 log "grow value of $name on $file"
4430 setfattr -n $name -v $large_value $file ||
4431 error "growing value of $name on $file failed"
4433 new_value=$(get_xattr_value $name $file)
4434 [[ "$new_value" != "$large_value" ]] &&
4435 error "$name different after growing"
4437 log "check value of $name on $file after remounting MDS"
4439 new_value=$(get_xattr_value $name $file)
4440 [[ "$new_value" != "$large_value" ]] &&
4441 error "$name different after remounting MDS"
4443 log "remove large xattr $name from $file"
4444 setfattr -x $name $file || error "removing $name from $file failed"
4447 stopall || error "stopping for e2fsck run"
4448 for num in $(seq $MDSCOUNT); do
4449 run_e2fsck $(facet_active_host mds$num) \
4450 $(mdsdevname $num) "-y" ||
4451 error "e2fsck MDT$num failed"
4453 setup_noconfig || error "remounting the filesystem failed"
4456 # need to delete this file to avoid problems in other tests
4458 stopall || error "stopping systems to turn off large_xattr"
4460 for num in $(seq $MDSCOUNT); do
4461 do_facet mds${num} $TUNE2FS -O ^large_xattr \
4462 $(mdsdevname $num) ||
4463 error "tune2fs on mds $num failed"
4467 run_test 61 "large xattr"
4470 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
4471 skip "Only applicable to ldiskfs-based MDTs"
4476 local mdsdev=$(mdsdevname 1)
4477 local ostdev=$(ostdevname 1)
4479 [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.2.51) ]] ||
4480 { skip "Need MDS version at least 2.2.51"; return 0; }
4482 echo "disable journal for mds"
4483 do_facet mds1 $TUNE2FS -O ^has_journal $mdsdev || error "tune2fs failed"
4484 start_mds && error "MDT start should fail"
4485 echo "disable journal for ost"
4486 do_facet ost1 $TUNE2FS -O ^has_journal $ostdev || error "tune2fs failed"
4487 start_ost && error "OST start should fail"
4488 cleanup || error "cleanup failed with rc $?"
4491 run_test 62 "start with disabled journal"
4494 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
4495 skip "Only applicable to ldiskfs-based MDTs"
4499 do_rpc_nodes $(facet_active_host $SINGLEMDS) load_module ldiskfs
4500 local inode_slab=$(do_facet $SINGLEMDS "cat /proc/slabinfo" |
4501 awk '/ldiskfs_inode_cache/ { print $5 / $6 }')
4502 if [ -z "$inode_slab" ]; then
4503 skip "ldiskfs module has not been loaded"
4507 echo "$inode_slab ldiskfs inodes per page"
4508 [ "${inode_slab%.*}" -ge "3" ] && return 0
4510 # If kmalloc-128 is also 1 per page - this is a debug kernel
4511 # and so this is not an error.
4512 local kmalloc128=$(do_facet $SINGLEMDS "cat /proc/slabinfo" |
4513 awk '/^(kmalloc|size)-128 / { print $5 / $6 }')
4514 # 32 128-byte chunks in 4k
4515 [ "${kmalloc128%.*}" -lt "32" ] ||
4516 error "ldiskfs inode too big, only $inode_slab objs/page, " \
4517 "kmalloc128 = $kmalloc128 objs/page"
4519 run_test 63 "Verify each page can at least hold 3 ldiskfs inodes"
4522 start_mds || error "unable to start MDS"
4523 start_ost || error "Unable to start OST1"
4524 start_ost2 || error "Unable to start second ost"
4525 mount_client $MOUNT || error "Unable to mount client"
4526 stop_ost2 || error "Unable to stop second ost"
4529 umount_client $MOUNT -f || error “unmount $MOUNT failed”
4530 cleanup_nocli || error "cleanup_nocli failed with $?"
4531 #writeconf to remove all ost2 traces for subsequent tests
4532 writeconf_or_reformat
4534 run_test 64 "check lfs df --lazy "
4536 test_65() { # LU-2237
4537 # Currently, the test is only valid for ldiskfs backend
4538 [ "$(facet_fstype $SINGLEMDS)" != "ldiskfs" ] &&
4539 skip "non-ldiskfs backend" && return
4541 local devname=$(mdsdevname ${SINGLEMDS//mds/})
4542 local brpt=$(facet_mntpt brpt)
4545 if ! do_facet $SINGLEMDS "test -b $devname"; then
4549 stop_mds || error "Unable to stop MDS"
4550 local obj=$(do_facet $SINGLEMDS \
4551 "$DEBUGFS -c -R \\\"stat last_rcvd\\\" $devname" |
4553 if [ -z "$obj" ]; then
4554 # The MDT may be just re-formatted, mount the MDT for the
4555 # first time to guarantee the "last_rcvd" file is there.
4556 start_mds || error "fail to mount the MDS for the first time"
4557 stop_mds || error "Unable to stop MDS"
4560 # remove the "last_rcvd" file
4561 do_facet $SINGLEMDS "mkdir -p $brpt"
4562 do_facet $SINGLEMDS \
4563 "mount -t $(facet_fstype $SINGLEMDS) $opts $devname $brpt"
4564 do_facet $SINGLEMDS "rm -f ${brpt}/last_rcvd"
4565 do_facet $SINGLEMDS "$UMOUNT $brpt"
4567 # restart MDS, the "last_rcvd" file should be recreated.
4568 start_mds || error "fail to restart the MDS"
4569 stop_mds || error "Unable to stop MDS"
4570 obj=$(do_facet $SINGLEMDS \
4571 "$DEBUGFS -c -R \\\"stat last_rcvd\\\" $devname" | grep Inode)
4572 [ -n "$obj" ] || error "fail to re-create the last_rcvd"
4574 run_test 65 "re-create the lost last_rcvd file when server mount"
4577 [[ $(lustre_version_code mgs) -ge $(version_code 2.3.59) ]] ||
4578 { skip "Need MGS version at least 2.3.59"; return 0; }
4581 local OST1_NID=$(do_facet ost1 $LCTL list_nids | head -1)
4582 local MDS_NID=$(do_facet $SINGLEMDS $LCTL list_nids | head -1)
4584 echo "replace_nids should fail if MDS, OSTs and clients are UP"
4585 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 $OST1_NID &&
4586 error "replace_nids fail"
4588 umount_client $MOUNT || error "unmounting client failed"
4589 echo "replace_nids should fail if MDS and OSTs are UP"
4590 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 $OST1_NID &&
4591 error "replace_nids fail"
4593 stop_ost || error "Unable to stop OST1"
4594 echo "replace_nids should fail if MDS is UP"
4595 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 $OST1_NID &&
4596 error "replace_nids fail"
4598 stop_mds || error "stopping mds failed"
4600 if combined_mgs_mds; then
4601 start_mdt 1 "-o nosvc" ||
4602 error "starting mds with nosvc option failed"
4605 echo "command should accept two parameters"
4606 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 &&
4607 error "command should accept two params"
4609 echo "correct device name should be passed"
4610 do_facet mgs $LCTL replace_nids $FSNAME-WRONG0000 $OST1_NID &&
4611 error "wrong devname"
4613 echo "wrong nids list should not destroy the system"
4614 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 "wrong nids list" &&
4617 echo "replace OST nid"
4618 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 $OST1_NID ||
4619 error "replace nids failed"
4621 echo "command should accept two parameters"
4622 do_facet mgs $LCTL replace_nids $FSNAME-MDT0000 &&
4623 error "command should accept two params"
4625 echo "wrong nids list should not destroy the system"
4626 do_facet mgs $LCTL replace_nids $FSNAME-MDT0000 "wrong nids list" &&
4629 echo "replace MDS nid"
4630 do_facet mgs $LCTL replace_nids $FSNAME-MDT0000 $MDS_NID ||
4631 error "replace nids failed"
4633 if ! combined_mgs_mds ; then
4636 stop_mds || error "Unable to stop MDS"
4640 check_mount || error "error after nid replace"
4641 cleanup || error "cleanup failed"
4644 run_test 66 "replace nids"
4646 test_67() { #LU-2950
4647 local legacy="$TMP/legacy_lnet_config"
4648 local new="$TMP/new_routes_test"
4649 local out="$TMP/config_out_file"
4650 local verify="$TMP/conv_verify"
4651 local verify_conf="$TMP/conf_verify"
4653 # Create the legacy file that will be run through the
4654 # lustre_routes_conversion script
4655 cat <<- LEGACY_LNET_CONFIG > $legacy
4656 tcp1 23 192.168.213.1@tcp:1; tcp5 34 193.30.4.3@tcp:4;
4657 tcp2 54 10.1.3.2@tcp;
4658 tcp3 10.3.4.3@tcp:3;
4662 # Create the verification file to verify the output of
4663 # lustre_routes_conversion script against.
4664 cat <<- VERIFY_LNET_CONFIG > $verify
4665 tcp1: { gateway: 192.168.213.1@tcp, hop: 23, priority: 1 }
4666 tcp5: { gateway: 193.30.4.3@tcp, hop: 34, priority: 4 }
4667 tcp2: { gateway: 10.1.3.2@tcp, hop: 54 }
4668 tcp3: { gateway: 10.3.4.3@tcp, priority: 3 }
4669 tcp4: { gateway: 10.3.3.4@tcp }
4672 # Create the verification file to verify the output of
4673 # lustre_routes_config script against
4674 cat <<- VERIFY_LNET_CONFIG > $verify_conf
4675 lctl --net tcp1 add_route 192.168.213.1@tcp 23 1
4676 lctl --net tcp5 add_route 193.30.4.3@tcp 34 4
4677 lctl --net tcp2 add_route 10.1.3.2@tcp 54 4
4678 lctl --net tcp3 add_route 10.3.4.3@tcp 1 3
4679 lctl --net tcp4 add_route 10.3.3.4@tcp 1 3
4682 lustre_routes_conversion $legacy $new > /dev/null
4683 if [ -f $new ]; then
4684 # verify the conversion output
4685 cmp -s $new $verify > /dev/null
4686 if [ $? -eq 1 ]; then
4687 error "routes conversion failed"
4690 lustre_routes_config --dry-run --verbose $new > $out
4691 # check that the script succeeded
4692 cmp -s $out $verify_conf > /dev/null
4693 if [ $? -eq 1 ]; then
4694 error "routes config failed"
4697 error "routes conversion test failed"
4699 # remove generated files
4700 rm -f $new $legacy $verify $verify_conf $out
4702 run_test 67 "test routes conversion and configuration"
4710 [ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.4.53) ] ||
4711 { skip "Need MDS version at least 2.4.53"; return 0; }
4713 umount_client $MOUNT || error "umount client failed"
4715 start_mdt 1 || error "MDT start failed"
4716 start_ost || error "Unable to start OST1"
4718 # START-END - the sequences we'll be reserving
4719 START=$(do_facet $SINGLEMDS \
4720 $LCTL get_param -n seq.ctl*.space | awk -F'[[ ]' '{print $2}')
4721 END=$((START + (1 << 30)))
4722 do_facet $SINGLEMDS \
4723 $LCTL set_param seq.ctl*.fldb="[$START-$END\):0:mdt"
4725 # reset the sequences MDT0000 has already assigned
4726 do_facet $SINGLEMDS \
4727 $LCTL set_param seq.srv*MDT0000.space=clear
4729 # remount to let the client allocate new sequence
4730 mount_client $MOUNT || error "mount client failed"
4732 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
4733 do_facet $SINGLEMDS \
4734 $LCTL get_param seq.srv*MDT0000.space
4735 $LFS path2fid $DIR/$tfile
4737 local old_ifs="$IFS"
4739 fid=($($LFS path2fid $DIR/$tfile))
4743 if [[ $seq < $END ]]; then
4744 error "used reserved sequence $seq?"
4746 cleanup || error "cleanup failed with $?"
4748 run_test 68 "be able to reserve specific sequences in FLDB"
4750 # Test 69: is about the total number of objects ever created on an OST.
4751 # so that when it is reformatted the normal MDS->OST orphan recovery won't
4752 # just "precreate" the missing objects. In the past it might try to recreate
4753 # millions of objects after an OST was reformatted
4755 local server_version=$(lustre_version_code $SINGLEMDS)
4757 [[ $server_version -lt $(version_code 2.4.2) ]] &&
4758 skip "Need MDS version at least 2.4.2" && return
4760 [[ $server_version -ge $(version_code 2.4.50) ]] &&
4761 [[ $server_version -lt $(version_code 2.5.0) ]] &&
4762 skip "Need MDS version at least 2.5.0" && return
4765 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
4767 # use OST0000 since it probably has the most creations
4768 local OSTNAME=$(ostname_from_index 0)
4769 local mdtosc_proc1=$(get_mdtosc_proc_path mds1 $OSTNAME)
4770 local last_id=$(do_facet mds1 $LCTL get_param -n \
4771 osc.$mdtosc_proc1.prealloc_last_id)
4773 # Want to have OST LAST_ID over 5 * OST_MAX_PRECREATE to
4774 # verify that the LAST_ID recovery is working properly. If
4775 # not, then the OST will refuse to allow the MDS connect
4776 # because the LAST_ID value is too different from the MDS
4777 #define OST_MAX_PRECREATE=20000
4778 local ost_max_pre=20000
4779 local num_create=$(( ost_max_pre * 5 + 1 - last_id))
4781 # If the LAST_ID is already over 5 * OST_MAX_PRECREATE, we don't
4782 # need to create any files. So, skip this section.
4783 if [ $num_create -gt 0 ]; then
4784 # Check the number of inodes available on OST0
4786 local ifree=$($LFS df -i $MOUNT | awk '/OST0000/ { print $4 }')
4787 log "On OST0, $ifree inodes available. Want $num_create."
4789 $SETSTRIPE -i 0 $DIR/$tdir ||
4790 error "$SETSTRIPE -i 0 $DIR/$tdir failed"
4791 if [ $ifree -lt 10000 ]; then
4792 files=$(( ifree - 50 ))
4797 local j=$((num_create / files + 1))
4798 for i in $(seq 1 $j); do
4799 createmany -o $DIR/$tdir/$tfile-$i- $files ||
4800 error "createmany fail create $files files: $?"
4801 unlinkmany $DIR/$tdir/$tfile-$i- $files ||
4802 error "unlinkmany failed unlink $files files"
4806 # delete all of the files with objects on OST0 so the
4807 # filesystem is not inconsistent later on
4808 $LFS find $MOUNT --ost 0 -print0 | xargs -0 rm
4810 umount_client $MOUNT || error "umount client failed"
4811 stop_ost || error "OST0 stop failure"
4812 add ost1 $(mkfs_opts ost1 $(ostdevname 1)) --reformat --replace \
4813 $(ostdevname 1) $(ostvdevname 1) ||
4814 error "reformat and replace $ostdev failed"
4815 start_ost || error "OST0 restart failure"
4816 wait_osc_import_state mds ost FULL
4818 mount_client $MOUNT || error "mount client failed"
4819 touch $DIR/$tdir/$tfile-last || error "create file after reformat"
4820 local idx=$($GETSTRIPE -i $DIR/$tdir/$tfile-last)
4821 [ $idx -ne 0 ] && error "$DIR/$tdir/$tfile-last on $idx not 0" || true
4823 local iused=$($LFS df -i $MOUNT | awk '/OST0000/ { print $3 }')
4824 log "On OST0, $iused used inodes"
4825 [ $iused -ge $((ost_max_pre/2 + 1000)) ] &&
4826 error "OST replacement created too many inodes; $iused"
4827 cleanup || error "cleanup failed with $?"
4829 run_test 69 "replace an OST with the same index"
4832 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
4835 cleanup || error "cleanup failed with $?"
4837 start_mdt 1 || error "MDT0 start fail"
4839 start_ost || error "OST0 start fail"
4840 for num in $(seq 2 $MDSCOUNT); do
4841 start_mdt $num || return
4844 mount_client $MOUNT || error "mount client fails"
4846 mkdir $DIR/$tdir || error "create $DIR/$tdir failed"
4848 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
4849 error "create remote dir fail"
4851 rm -rf $DIR/$tdir || error "delete dir fail"
4852 cleanup || error "cleanup failed with $?"
4854 run_test 70a "start MDT0, then OST, then MDT1"
4857 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
4860 start_ost || error "OST0 start fail"
4862 start_mds || error "MDS start fail"
4864 mount_client $MOUNT || error "mount client fails"
4866 mkdir $DIR/$tdir || error "create $DIR/$tdir failed"
4868 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
4869 error "create remote dir fail"
4871 rm -rf $DIR/$tdir || error "delete dir fail"
4873 cleanup || error "cleanup failed with $?"
4875 run_test 70b "start OST, MDT1, MDT0"
4878 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
4881 start_mds || error "MDS start fail"
4882 start_ost || error "OST0 start fail"
4884 mount_client $MOUNT || error "mount client fails"
4885 stop_mdt 1 || error "MDT1 start fail"
4887 local mdc_for_mdt1=$($LCTL dl | grep MDT0000-mdc | awk '{print $4}')
4888 echo "deactivate $mdc_for_mdt1"
4889 $LCTL --device $mdc_for_mdt1 deactivate ||
4890 error "set $mdc_for_mdt1 deactivate failed"
4892 mkdir $DIR/$tdir && error "mkdir succeed"
4894 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir &&
4895 error "create remote dir succeed"
4897 cleanup || error "cleanup failed with $?"
4899 run_test 70c "stop MDT0, mkdir fail, create remote dir fail"
4902 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
4905 start_mds || error "MDS start fail"
4906 start_ost || error "OST0 start fail"
4908 mount_client $MOUNT || error "mount client fails"
4910 stop_mdt 2 || error "MDT1 start fail"
4912 local mdc_for_mdt2=$($LCTL dl | grep MDT0001-mdc |
4914 echo "deactivate $mdc_for_mdt2"
4915 $LCTL --device $mdc_for_mdt2 deactivate ||
4916 error "set $mdc_for_mdt2 deactivate failed"
4918 mkdir $DIR/$tdir || error "mkdir fail"
4919 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir &&
4920 error "create remote dir succeed"
4922 rm -rf $DIR/$tdir || error "delete dir fail"
4924 cleanup || error "cleanup failed with $?"
4926 run_test 70d "stop MDT1, mkdir succeed, create remote dir fail"
4929 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
4931 [ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.7.62) ] ||
4932 { skip "Need MDS version at least 2.7.62"; return 0; }
4934 cleanup || error "cleanup failed with $?"
4936 local mdsdev=$(mdsdevname 1)
4937 local ostdev=$(ostdevname 1)
4938 local mdsvdev=$(mdsvdevname 1)
4939 local ostvdev=$(ostvdevname 1)
4940 local opts_mds="$(mkfs_opts mds1 $mdsdev) --reformat $mdsdev $mdsvdev"
4941 local opts_ost="$(mkfs_opts ost1 $ostdev) --reformat $ostdev $ostvdev"
4943 add mds1 $opts_mds || error "add mds1 failed"
4944 start_mdt 1 || error "start mdt1 failed"
4945 add ost1 $opts_ost || error "add ost1 failed"
4946 start_ost || error "start ost failed"
4947 mount_client $MOUNT > /dev/null || error "mount client $MOUNT failed"
4949 local soc=$(do_facet mds1 "$LCTL get_param -n \
4950 mdt.*MDT0000.sync_lock_cancel")
4951 [ $soc == "never" ] || error "SoC enabled on single MDS"
4953 for i in $(seq 2 $MDSCOUNT); do
4954 mdsdev=$(mdsdevname $i)
4955 mdsvdev=$(mdsvdevname $i)
4956 opts_mds="$(mkfs_opts mds$i $mdsdev) --reformat $mdsdev \
4958 add mds$i $opts_mds || error "add mds$i failed"
4959 start_mdt $i || error "start mdt$i fail"
4962 wait_dne_interconnect
4964 for i in $(seq $MDSCOUNT); do
4965 soc=$(do_facet mds$i "$LCTL get_param -n \
4966 mdt.*MDT000$((i - 1)).sync_lock_cancel")
4967 [ $soc == "blocking" ] || error "SoC not enabled on DNE"
4970 for i in $(seq 2 $MDSCOUNT); do
4971 stop_mdt $i || error "stop mdt$i fail"
4973 soc=$(do_facet mds1 "$LCTL get_param -n \
4974 mdt.*MDT0000.sync_lock_cancel")
4975 [ $soc == "never" ] || error "SoC enabled on single MDS"
4977 cleanup || error "cleanup failed with $?"
4979 run_test 70e "Sync-on-Cancel will be enabled by default on DNE"
4982 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
4983 if combined_mgs_mds; then
4984 skip "needs separate MGS/MDT" && return
4988 start_mdt 1 || error "MDT0 start fail"
4989 start_ost || error "OST0 start fail"
4990 for num in $(seq 2 $MDSCOUNT); do
4991 start_mdt $num || return
4994 start_ost2 || error "OST1 start fail"
4996 mount_client $MOUNT || error "mount client fails"
4998 mkdir $DIR/$tdir || error "mkdir fail"
4999 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
5000 error "create remote dir succeed"
5002 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
5003 rm -rf $DIR/$tdir || error "delete dir fail"
5005 umount_client $MOUNT || error "umount_client failed"
5006 stop_mds || error "MDS stop fail"
5007 stop_ost || error "OST0 stop fail"
5008 stop_ost2 || error "OST1 stop fail"
5010 run_test 71a "start MDT0 OST0, MDT1, OST1"
5013 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
5014 if combined_mgs_mds; then
5015 skip "needs separate MGS/MDT" && return
5019 for num in $(seq 2 $MDSCOUNT); do
5020 start_mdt $num || return
5022 start_ost || error "OST0 start fail"
5023 start_mdt 1 || error "MDT0 start fail"
5024 start_ost2 || error "OST1 start fail"
5026 mount_client $MOUNT || error "mount client fails"
5028 mkdir $DIR/$tdir || error "mkdir fail"
5029 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
5030 error "create remote dir succeed"
5032 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
5033 rm -rf $DIR/$tdir || error "delete dir fail"
5035 umount_client $MOUNT || error "umount_client failed"
5036 stop_mds || error "MDT0 stop fail"
5037 stop_ost || error "OST0 stop fail"
5038 stop_ost2 || error "OST1 stop fail"
5040 run_test 71b "start MDT1, OST0, MDT0, OST1"
5043 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
5044 if combined_mgs_mds; then
5045 skip "needs separate MGS/MDT" && return
5049 start_ost || error "OST0 start fail"
5050 start_ost2 || error "OST1 start fail"
5051 for num in $(seq 2 $MDSCOUNT); do
5052 start_mdt $num || return
5054 start_mdt 1 || error "MDT0 start fail"
5056 mount_client $MOUNT || error "mount client fails"
5058 mkdir $DIR/$tdir || error "mkdir fail"
5059 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
5060 error "create remote dir succeed"
5062 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
5063 rm -rf $DIR/$tdir || error "delete dir fail"
5065 umount_client $MOUNT || error "umount_client failed"
5066 stop_mds || error "MDS stop fail"
5067 stop_ost || error "OST0 stop fail"
5068 stop_ost2 || error "OST1 stop fail"
5071 run_test 71c "start OST0, OST1, MDT1, MDT0"
5074 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
5075 if combined_mgs_mds; then
5076 skip "needs separate MGS/MDT" && return
5080 start_ost || error "OST0 start fail"
5081 for num in $(seq 2 $MDSCOUNT); do
5082 start_mdt $num || return
5084 start_mdt 1 || error "MDT0 start fail"
5085 start_ost2 || error "OST1 start fail"
5087 mount_client $MOUNT || error "mount client fails"
5089 mkdir $DIR/$tdir || error "mkdir fail"
5090 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
5091 error "create remote dir succeed"
5093 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
5094 rm -rf $DIR/$tdir || error "delete dir fail"
5096 umount_client $MOUNT || error "umount_client failed"
5097 stop_mds || error "MDS stop fail"
5098 stop_ost || error "OST0 stop fail"
5099 stop_ost2 || error "OST1 stop fail"
5102 run_test 71d "start OST0, MDT1, MDT0, OST1"
5105 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
5106 if combined_mgs_mds; then
5107 skip "needs separate MGS/MDT" && return
5111 start_ost || error "OST0 start fail"
5112 for num in $(seq 2 $MDSCOUNT); do
5113 start_mdt $num || return
5115 start_ost2 || error "OST1 start fail"
5116 start_mdt 1 || error "MDT0 start fail"
5118 mount_client $MOUNT || error "mount client fails"
5120 mkdir $DIR/$tdir || error "mkdir fail"
5121 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
5122 error "create remote dir succeed"
5124 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
5125 rm -rf $DIR/$tdir || error "delete dir fail"
5127 umount_client $MOUNT || error "umount_client failed"
5128 stop_mds || error "MDS stop fail"
5129 stop_ost || error "OST0 stop fail"
5130 stop_ost2 || error "OST1 stop fail"
5133 run_test 71e "start OST0, MDT1, OST1, MDT0"
5135 test_72() { #LU-2634
5136 local mdsdev=$(mdsdevname 1)
5137 local ostdev=$(ostdevname 1)
5138 local cmd="$E2FSCK -fnvd $mdsdev"
5142 [ "$(facet_fstype $SINGLEMDS)" != "ldiskfs" ] &&
5143 skip "ldiskfs only test" && return
5145 if combined_mgs_mds; then
5146 add_options='--reformat'
5148 add_options='--reformat --replace'
5151 #tune MDT with "-O extents"
5153 for num in $(seq $MDSCOUNT); do
5154 add mds${num} $(mkfs_opts mds$num $(mdsdevname $num)) \
5155 $add_options $(mdsdevname $num) $(mdsvdevname $num) ||
5156 error "add mds $num failed"
5157 do_facet mds${num} "$TUNE2FS -O extents $(mdsdevname $num)" ||
5158 error "$TUNE2FS failed on mds${num}"
5161 add ost1 $(mkfs_opts ost1 $ostdev) $add_options $ostdev ||
5162 error "add $ostdev failed"
5163 start_mds || error "start mds failed"
5164 start_ost || error "start ost failed"
5165 mount_client $MOUNT || error "mount client failed"
5167 #create some short symlinks
5168 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
5169 createmany -o $DIR/$tdir/$tfile-%d $fn
5170 echo "create $fn short symlinks"
5171 for i in $(seq -w 1 $fn); do
5172 ln -s $DIR/$tdir/$tfile-$i $MOUNT/$tfile-$i
5177 umount_client $MOUNT || error "umount client failed"
5178 stop_mds || error "stop mds failed"
5179 stop_ost || error "stop ost failed"
5182 run_e2fsck $(facet_active_host $SINGLEMDS) $mdsdev "-n"
5184 run_test 72 "test fast symlink with extents flag enabled"
5186 test_73() { #LU-3006
5187 [ $(facet_fstype ost1) == zfs ] && import_zpool ost1
5188 do_facet ost1 "$TUNEFS --failnode=1.2.3.4@$NETTYPE $(ostdevname 1)" ||
5189 error "1st tunefs failed"
5190 start_mgsmds || error "start mds failed"
5191 start_ost || error "start ost failed"
5192 mount_client $MOUNT || error "mount client failed"
5193 $LCTL get_param -n osc.*OST0000-osc-[^M]*.import | grep failover_nids |
5194 grep 1.2.3.4@$NETTYPE || error "failover nids haven't changed"
5195 umount_client $MOUNT || error "umount client failed"
5199 run_test 73 "failnode to update from mountdata properly"
5201 test_75() { # LU-2374
5202 [[ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.4.1) ]] &&
5203 skip "Need MDS version at least 2.4.1" && return
5206 local opts_mds="$(mkfs_opts mds1 $(mdsdevname 1)) \
5207 --reformat $(mdsdevname 1) $(mdsvdevname 1)"
5208 local opts_ost="$(mkfs_opts ost1 $(ostdevname 1)) \
5209 --reformat $(ostdevname 1) $(ostvdevname 1)"
5211 #check with default parameters
5212 add mds1 $opts_mds || error "add mds1 failed for default params"
5213 add ost1 $opts_ost || error "add ost1 failed for default params"
5215 opts_mds=$(echo $opts_mds | sed -e "s/--mdt//")
5216 opts_mds=$(echo $opts_mds |
5217 sed -e "s/--index=$index/--index=$index --mdt/")
5218 opts_ost=$(echo $opts_ost | sed -e "s/--ost//")
5219 opts_ost=$(echo $opts_ost |
5220 sed -e "s/--index=$index/--index=$index --ost/")
5222 add mds1 $opts_mds || error "add mds1 failed for new params"
5223 add ost1 $opts_ost || error "add ost1 failed for new params"
5226 run_test 75 "The order of --index should be irrelevant"
5229 [[ $(lustre_version_code mgs) -ge $(version_code 2.4.52) ]] ||
5230 { skip "Need MDS version at least 2.4.52" && return 0; }
5232 local MDMB_PARAM="osc.*.max_dirty_mb"
5233 echo "Change MGS params"
5234 local MAX_DIRTY_MB=$($LCTL get_param -n $MDMB_PARAM |
5236 echo "max_dirty_mb: $MAX_DIRTY_MB"
5237 local NEW_MAX_DIRTY_MB=$((MAX_DIRTY_MB + MAX_DIRTY_MB))
5238 echo "new_max_dirty_mb: $NEW_MAX_DIRTY_MB"
5239 do_facet mgs $LCTL set_param -P $MDMB_PARAM=$NEW_MAX_DIRTY_MB
5240 wait_update $HOSTNAME "$LCTL get_param -n $MDMB_PARAM |
5241 head -1" $NEW_MAX_DIRTY_MB
5242 MAX_DIRTY_MB=$($LCTL get_param -n $MDMB_PARAM | head -1)
5243 echo "$MAX_DIRTY_MB"
5244 [ $MAX_DIRTY_MB = $NEW_MAX_DIRTY_MB ] ||
5245 error "error while apply max_dirty_mb"
5247 echo "Check the value is stored after remount"
5250 wait_update $HOSTNAME "$LCTL get_param -n $MDMB_PARAM |
5251 head -1" $NEW_MAX_DIRTY_MB
5252 MAX_DIRTY_MB=$($LCTL get_param -n $MDMB_PARAM | head -1)
5253 [ $MAX_DIRTY_MB = $NEW_MAX_DIRTY_MB ] ||
5254 error "max_dirty_mb is not saved after remount"
5256 echo "Change OST params"
5257 CLIENT_PARAM="obdfilter.*.client_cache_count"
5258 local CLIENT_CACHE_COUNT
5259 CLIENT_CACHE_COUNT=$(do_facet ost1 $LCTL get_param -n $CLIENT_PARAM |
5261 echo "client_cache_count: $CLIENT_CACHE_COUNT"
5262 NEW_CLIENT_CACHE_COUNT=$((CLIENT_CACHE_COUNT+CLIENT_CACHE_COUNT))
5263 echo "new_client_cache_count: $NEW_CLIENT_CACHE_COUNT"
5264 do_facet mgs $LCTL set_param -P $CLIENT_PARAM=$NEW_CLIENT_CACHE_COUNT
5265 wait_update $(facet_host ost1) "$LCTL get_param -n $CLIENT_PARAM |
5266 head -1" $NEW_CLIENT_CACHE_COUNT
5267 CLIENT_CACHE_COUNT=$(do_facet ost1 $LCTL get_param -n $CLIENT_PARAM |
5269 echo "$CLIENT_CACHE_COUNT"
5270 [ $CLIENT_CACHE_COUNT = $NEW_CLIENT_CACHE_COUNT ] ||
5271 error "error while apply client_cache_count"
5273 echo "Check the value is stored after remount"
5276 wait_update $(facet_host ost1) "$LCTL get_param -n $CLIENT_PARAM |
5277 head -1" $NEW_CLIENT_CACHE_COUNT
5278 CLIENT_CACHE_COUNT=$(do_facet ost1 $LCTL get_param -n $CLIENT_PARAM |
5280 echo "$CLIENT_CACHE_COUNT"
5281 [ $CLIENT_CACHE_COUNT = $NEW_CLIENT_CACHE_COUNT ] ||
5282 error "client_cache_count is not saved after remount"
5285 run_test 76a "set permanent params set_param -P"
5287 test_76b() { # LU-4783
5288 [[ $(lustre_version_code mgs) -ge $(version_code 2.5.57) ]] ||
5289 { skip "Need MGS version at least 2.5.57" && return 0; }
5292 do_facet mgs $LCTL get_param mgs.MGS.live.params ||
5293 error "start params log failed"
5296 run_test 76b "verify params log setup correctly"
5299 [[ $(lustre_version_code mgs) -ge $(version_code 2.8.54) ]] ||
5300 { skip "Need MDS version at least 2.4.52" && return 0; }
5302 local MASK_PARAM="mdd.*.changelog_mask"
5303 echo "Change changelog_mask"
5304 do_facet mgs $LCTL set_param -P $MASK_PARAM=-CLOSE ||
5305 error "Can't change changlog_mask"
5306 wait_update $(facet_host mds) "$LCTL get_param -n $MASK_PARAM |
5309 echo "Check the value is stored after mds remount"
5310 stop_mds || error "Failed to stop MDS"
5311 start_mds || error "Failed to start MDS"
5312 local CHANGELOG_MASK=$(do_facet mgs $LCTL get_param -n $MASK_PARAM)
5313 echo $CHANGELOG_MASK | grep CLOSE > /dev/null &&
5314 error "changelog_mask is not changed"
5318 run_test 76c "verify changelog_mask is applied with set_param -P"
5320 test_77() { # LU-3445
5321 local server_version=$(lustre_version_code $SINGLEMDS)
5322 [[ $server_version -ge $(version_code 2.8.55) ]] ||
5323 { skip "Need MDS version 2.8.55+ "; return; }
5325 if [[ -z "$fs2ost_DEV" || -z "$fs2mds_DEV" ]]; then
5326 is_blkdev $SINGLEMDS $(mdsdevname ${SINGLEMDS//mds/}) &&
5327 skip_env "mixed loopback and real device not working" && return
5330 local fs2mdsdev=$(mdsdevname 1_2)
5331 local fs2ostdev=$(ostdevname 1_2)
5332 local fs2mdsvdev=$(mdsvdevname 1_2)
5333 local fs2ostvdev=$(ostvdevname 1_2)
5334 local fsname=test1234
5336 local failnid="$(h2$NETTYPE 1.2.3.4),$(h2$NETTYPE 4.3.2.1)"
5338 add fs2mds $(mkfs_opts mds1 $fs2mdsdev) --mgs --fsname=$fsname \
5339 --reformat $fs2mdsdev $fs2mdsvdev || error "add fs2mds failed"
5340 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS && trap cleanup_fs2 EXIT INT ||
5341 error "start fs2mds failed"
5343 mgsnid=$(do_facet fs2mds $LCTL list_nids | xargs | tr ' ' ,)
5344 mgsnid="$mgsnid,$mgsnid:$mgsnid"
5346 add fs2ost $(mkfs_opts ost1 $fs2ostdev) --mgsnode=$mgsnid \
5347 --failnode=$failnid --fsname=$fsname \
5348 --reformat $fs2ostdev $fs2ostvdev ||
5349 error "add fs2ost failed"
5350 start fs2ost $fs2ostdev $OST_MOUNT_OPTS || error "start fs2ost failed"
5352 mkdir -p $MOUNT2 || error "mkdir $MOUNT2 failed"
5353 $MOUNT_CMD $mgsnid:/$fsname $MOUNT2 || error "mount $MOUNT2 failed"
5354 DIR=$MOUNT2 MOUNT=$MOUNT2 check_mount || error "check $MOUNT2 failed"
5357 run_test 77 "comma-separated MGS NIDs and failover node NIDs"
5360 [[ $(facet_fstype $SINGLEMDS) != ldiskfs ||
5361 $(facet_fstype ost1) != ldiskfs ]] &&
5362 skip "only applicable to ldiskfs-based MDTs and OSTs" && return
5364 # reformat the Lustre filesystem with a smaller size
5365 local saved_MDSCOUNT=$MDSCOUNT
5366 local saved_MDSSIZE=$MDSSIZE
5367 local saved_OSTCOUNT=$OSTCOUNT
5368 local saved_OSTSIZE=$OSTSIZE
5371 MDSSIZE=$((MDSSIZE - 20000))
5372 OSTSIZE=$((OSTSIZE - 20000))
5373 reformat || error "(1) reformat Lustre filesystem failed"
5374 MDSSIZE=$saved_MDSSIZE
5375 OSTSIZE=$saved_OSTSIZE
5377 # mount the Lustre filesystem
5378 setup_noconfig || error "(2) setup Lustre filesystem failed"
5381 log "create test files"
5386 mkdir $MOUNT/$tdir || error "(3) mkdir $MOUNT/$tdir failed"
5388 for i in $(seq $num_files); do
5389 file=$MOUNT/$tdir/$tfile-$i
5390 dd if=/dev/urandom of=$file count=1 bs=1M || {
5391 $LCTL get_param osc.*.cur*grant*
5392 $LFS df; $LFS df -i;
5393 # stop creating files if there is no more space
5394 if [ ! -e $file ]; then
5395 num_files=$((i - 1))
5399 $LFS getstripe -v $file
5400 local ost_idx=$(LFS getstripe -i $file)
5401 do_facet ost$((ost_idx + 1)) \
5402 $LCTL get_param obdfilter.*.*grant*
5403 error "(4) create $file failed"
5407 # unmount the Lustre filesystem
5408 cleanup || error "(5) cleanup Lustre filesystem failed"
5410 # run e2fsck on the MDT and OST devices
5411 local mds_host=$(facet_active_host $SINGLEMDS)
5412 local ost_host=$(facet_active_host ost1)
5413 local mds_dev=$(mdsdevname ${SINGLEMDS//mds/})
5414 local ost_dev=$(ostdevname 1)
5416 run_e2fsck $mds_host $mds_dev "-y"
5417 run_e2fsck $ost_host $ost_dev "-y"
5419 # get the original block count of the MDT and OST filesystems
5420 local mds_orig_blks=$(get_block_count $SINGLEMDS $mds_dev)
5421 local ost_orig_blks=$(get_block_count ost1 $ost_dev)
5423 # expand the MDT and OST filesystems to the device size
5424 run_resize2fs $SINGLEMDS $mds_dev "" || error "expand $SINGLEMDS failed"
5425 run_resize2fs ost1 $ost_dev "" || error "expand ost1 failed"
5427 # run e2fsck on the MDT and OST devices again
5428 run_e2fsck $mds_host $mds_dev "-y"
5429 run_e2fsck $ost_host $ost_dev "-y"
5431 # mount the Lustre filesystem
5435 log "check files after expanding the MDT and OST filesystems"
5436 for i in $(seq $num_files); do
5437 file=$MOUNT/$tdir/$tfile-$i
5438 $CHECKSTAT -t file -s 1048576 $file ||
5439 error "(6) checkstat $file failed"
5443 log "create more files after expanding the MDT and OST filesystems"
5444 for i in $(seq $((num_files + 1)) $((num_files + 10))); do
5445 file=$MOUNT/$tdir/$tfile-$i
5446 dd if=/dev/urandom of=$file count=1 bs=1M ||
5447 error "(7) create $file failed"
5450 # unmount the Lustre filesystem
5451 cleanup || error "(8) cleanup Lustre filesystem failed"
5453 # run e2fsck on the MDT and OST devices
5454 run_e2fsck $mds_host $mds_dev "-y"
5455 run_e2fsck $ost_host $ost_dev "-y"
5457 # get the maximum block count of the MDT and OST filesystems
5458 local mds_max_blks=$(get_block_count $SINGLEMDS $mds_dev)
5459 local ost_max_blks=$(get_block_count ost1 $ost_dev)
5461 # get the minimum block count of the MDT and OST filesystems
5462 local mds_min_blks=$(run_resize2fs $SINGLEMDS $mds_dev "" "-P" 2>&1 |
5463 grep minimum | sed -e 's/^.*filesystem: //g')
5464 local ost_min_blks=$(run_resize2fs ost1 $ost_dev "" "-P" 2>&1 |
5465 grep minimum | sed -e 's/^.*filesystem: //g')
5467 # shrink the MDT and OST filesystems to a smaller size
5471 if [[ $mds_max_blks -gt $mds_min_blks &&
5472 $mds_max_blks -gt $mds_orig_blks ]]; then
5473 [[ $mds_orig_blks -gt $mds_min_blks ]] &&
5474 base_blks=$mds_orig_blks || base_blks=$mds_min_blks
5475 new_blks=$(( (mds_max_blks - base_blks) / 2 + base_blks ))
5476 run_resize2fs $SINGLEMDS $mds_dev $new_blks ||
5477 error "shrink $SINGLEMDS to $new_blks failed"
5481 if [[ $ost_max_blks -gt $ost_min_blks &&
5482 $ost_max_blks -gt $ost_orig_blks ]]; then
5483 [[ $ost_orig_blks -gt $ost_min_blks ]] &&
5484 base_blks=$ost_orig_blks || base_blks=$ost_min_blks
5485 new_blks=$(( (ost_max_blks - base_blks) / 2 + base_blks ))
5486 run_resize2fs ost1 $ost_dev $new_blks ||
5487 error "shrink ost1 to $new_blks failed"
5491 # check whether the MDT or OST filesystem was shrunk or not
5493 combined_mgs_mds || stop_mgs || error "(9) stop mgs failed"
5494 reformat || error "(10) reformat Lustre filesystem failed"
5498 # run e2fsck on the MDT and OST devices again
5499 run_e2fsck $mds_host $mds_dev "-y"
5500 run_e2fsck $ost_host $ost_dev "-y"
5502 # mount the Lustre filesystem again
5506 log "check files after shrinking the MDT and OST filesystems"
5507 for i in $(seq $((num_files + 10))); do
5508 file=$MOUNT/$tdir/$tfile-$i
5509 $CHECKSTAT -t file -s 1048576 $file ||
5510 error "(11) checkstat $file failed"
5513 # unmount and reformat the Lustre filesystem
5514 cleanup || error "(12) cleanup Lustre filesystem failed"
5515 combined_mgs_mds || stop_mgs || error "(13) stop mgs failed"
5517 MDSCOUNT=$saved_MDSCOUNT
5518 OSTCOUNT=$saved_OSTCOUNT
5519 reformat || error "(14) reformat Lustre filesystem failed"
5521 run_test 78 "run resize2fs on MDT and OST filesystems"
5523 test_79() { # LU-4227
5524 [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.5.59) ]] ||
5525 { skip "Need MDS version at least 2.5.59"; return 0; }
5527 local mdsdev1=$(mdsdevname 1)
5528 local mdsvdev1=$(mdsvdevname 1)
5529 local mdsdev2=$(mdsdevname 2)
5530 local mdsvdev2=$(mdsvdevname 2)
5531 local ostdev1=$(ostdevname 1)
5532 local ostvdev1=$(ostvdevname 1)
5533 local opts_mds1="$(mkfs_opts mds1 $mdsdev1) --reformat"
5534 local opts_mds2="$(mkfs_opts mds2 $mdsdev2) --reformat"
5535 local opts_ost1="$(mkfs_opts ost1 $ostdev1) --reformat"
5538 # remove --mgs/--mgsnode from mkfs.lustre options
5539 opts_mds1=$(echo $opts_mds1 | sed -e "s/--mgs//")
5541 mgsnode_opt=$(echo $opts_mds2 |
5542 awk '{ for ( i = 1; i < NF; i++ )
5543 if ( $i ~ "--mgsnode" ) { print $i; break } }')
5544 [ -n $mgsnode_opt ] &&
5545 opts_mds2=$(echo $opts_mds2 | sed -e "s/$mgsnode_opt//")
5547 mgsnode_opt=$(echo $opts_ost1 |
5548 awk '{ for ( i = 1; i < NF; i++ )
5549 if ( $i ~ "--mgsnode" ) { print $i; break } }')
5550 [ -n $mgsnode_opt ] &&
5551 opts_ost1=$(echo $opts_ost1 | sed -e "s/$mgsnode_opt//")
5553 # -MGS, format a mdt without --mgs option
5554 add mds1 $opts_mds1 $mdsdev1 $mdsvdev1 &&
5555 error "Must specify --mgs when formatting mdt combined with mgs"
5557 # +MGS, format a mdt/ost without --mgsnode option
5558 add mds1 $(mkfs_opts mds1 $mdsdev1) --reformat $mdsdev1 $mdsvdev1 \
5559 > /dev/null || error "start mds1 failed"
5560 add mds2 $opts_mds2 $mdsdev2 $mdsvdev2 &&
5561 error "Must specify --mgsnode when formatting a mdt"
5562 add ost1 $opts_ost1 $ostdev1 $ostvdev2 &&
5563 error "Must specify --mgsnode when formatting an ost"
5567 run_test 79 "format MDT/OST without mgs option (should return errors)"
5570 start_mds || error "Failed to start MDT"
5571 start_ost || error "Failed to start OST1"
5572 uuid=$(do_facet ost1 $LCTL get_param -n mgc.*.uuid)
5573 #define OBD_FAIL_MGS_PAUSE_TARGET_CON 0x906
5574 do_facet ost1 "$LCTL set_param fail_val=10 fail_loc=0x906"
5575 do_facet mgs "$LCTL set_param fail_val=10 fail_loc=0x906"
5576 do_facet mgs "$LCTL set_param -n mgs/MGS/evict_client $uuid"
5578 start_ost2 || error "Failed to start OST2"
5580 do_facet ost1 "$LCTL set_param fail_loc=0"
5583 run_test 80 "mgc import reconnect race"
5585 #Save the original values of $OSTCOUNT and $OSTINDEX$i.
5587 local new_ostcount=$1
5588 saved_ostcount=$OSTCOUNT
5589 OSTCOUNT=$new_ostcount
5593 for i in $(seq $OSTCOUNT); do
5595 eval saved_ostindex$i=${!index}
5600 # Restore the original values of $OSTCOUNT and $OSTINDEX$i.
5601 restore_ostindex() {
5606 for i in $(seq $OSTCOUNT); do
5607 index=saved_ostindex$i
5608 eval OSTINDEX$i=${!index}
5610 OSTCOUNT=$saved_ostcount
5615 # The main purpose of this test is to ensure the OST_INDEX_LIST functions as
5616 # expected. This test uses OST_INDEX_LIST to format OSTs with a randomly
5617 # assigned index and ensures we can mount such a formatted file system
5618 test_81() { # LU-4665
5619 [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.6.54) ]] ||
5620 { skip "Need MDS version at least 2.6.54" && return; }
5621 [[ $OSTCOUNT -ge 3 ]] || { skip_env "Need at least 3 OSTs" && return; }
5625 # Each time RANDOM is referenced, a random integer between 0 and 32767
5628 local saved_ostindex1=$OSTINDEX1
5629 for i in 65535 $((RANDOM + 65536)); do
5630 echo -e "\nFormat ost1 with --index=$i, should fail"
5632 if add ost1 $(mkfs_opts ost1 $(ostdevname 1)) --reformat \
5633 $(ostdevname 1) $(ostvdevname 1); then
5634 OSTINDEX1=$saved_ostindex1
5635 error "format ost1 with --index=$i should fail"
5638 OSTINDEX1=$saved_ostindex1
5642 # Format OSTs with random sparse indices.
5643 trap "restore_ostindex" EXIT
5644 echo -e "\nFormat $OSTCOUNT OSTs with sparse indices"
5645 OST_INDEX_LIST=[0,$((RANDOM * 2 % 65533 + 1)),65534] formatall
5647 # Setup and check Lustre filesystem.
5648 start_mgsmds || error "start_mgsmds failed"
5649 for i in $(seq $OSTCOUNT); do
5650 start ost$i $(ostdevname $i) $OST_MOUNT_OPTS ||
5651 error "start ost$i failed"
5654 mount_client $MOUNT || error "mount client $MOUNT failed"
5655 check_mount || error "check client $MOUNT failed"
5658 local max_easize=$($LCTL get_param -n llite.*.max_easize)
5659 [[ $max_easize -eq 128 ]] ||
5660 error "max_easize is $max_easize, should be 128 bytes"
5664 run_test 81 "sparse OST indexing"
5666 # Here we exercise the stripe placement functionality on a file system that
5667 # has formatted the OST with a random index. With the file system the following
5668 # functionality is tested:
5670 # 1. Creating a new file with a specific stripe layout.
5672 # 2. Modifiy a existing empty file with a specific stripe layout.
5674 # 3. Ensure we fail to set the stripe layout of a file that already has one.
5676 # 4. If ost-index is defined we need to ensure it is the first entry in the
5677 # ost index list returned by lfs getstripe.
5679 # 5. Lastly ensure this functionality fails with directories.
5680 test_82a() { # LU-4665
5681 [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.6.54) ]] ||
5682 { skip "Need MDS version at least 2.6.54" && return; }
5683 [[ $OSTCOUNT -ge 3 ]] || { skip_env "Need at least 3 OSTs" && return; }
5689 # Format OSTs with random sparse indices.
5693 local LOV_V1_INSANE_STRIPE_COUNT=65532
5694 for i in $(seq $OSTCOUNT); do
5695 index=$(((RANDOM * 2) % LOV_V1_INSANE_STRIPE_COUNT))
5696 ost_indices+=" $index"
5698 ost_indices=$(comma_list $ost_indices)
5700 trap "restore_ostindex" EXIT
5701 echo -e "\nFormat $OSTCOUNT OSTs with sparse indices $ost_indices"
5702 OST_INDEX_LIST=[$ost_indices] formatall
5704 # Setup Lustre filesystem.
5705 start_mgsmds || error "start_mgsmds failed"
5706 for i in $(seq $OSTCOUNT); do
5707 start ost$i $(ostdevname $i) $OST_MOUNT_OPTS ||
5708 error "start ost$i failed"
5711 mount_client $MOUNT || error "mount client $MOUNT failed"
5714 $LFS df $MOUNT || error "$LFS df $MOUNT failed"
5715 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
5717 # 1. If the file does not exist, new file will be created
5718 # with specified OSTs.
5719 local file=$DIR/$tdir/$tfile-1
5720 local cmd="$SETSTRIPE -o $ost_indices $file"
5722 eval $cmd || error "$cmd failed"
5723 check_stripe_count $file $OSTCOUNT
5724 check_obdidx $file $ost_indices
5725 dd if=/dev/urandom of=$file count=1 bs=1M > /dev/null 2>&1 ||
5726 error "write $file failed"
5728 # 2. If the file already exists and is an empty file, the file
5729 # will be attached with specified layout.
5730 file=$DIR/$tdir/$tfile-2
5731 mcreate $file || error "mcreate $file failed"
5732 cmd="$SETSTRIPE -o $ost_indices $file"
5734 eval $cmd || error "$cmd failed"
5735 dd if=/dev/urandom of=$file count=1 bs=1M > /dev/null 2>&1 ||
5736 error "write $file failed"
5737 check_stripe_count $file $OSTCOUNT
5738 check_obdidx $file $ost_indices
5740 # 3. If the file already has a valid layout attached, the command
5741 # should fail with EBUSY.
5743 eval $cmd && error "stripe is already set on $file, $cmd should fail"
5745 # 4. If [--stripe-index|-i <start_ost_idx>] is used, the index must
5746 # be in the OST indices list.
5747 local start_ost_idx=${ost_indices##*,}
5748 file=$DIR/$tdir/$tfile-3
5749 cmd="$SETSTRIPE -o $ost_indices -i $start_ost_idx $file"
5751 eval $cmd || error "$cmd failed"
5752 check_stripe_count $file $OSTCOUNT
5753 check_obdidx $file $ost_indices
5754 check_start_ost_idx $file $start_ost_idx
5756 file=$DIR/$tdir/$tfile-4
5758 cmd+=" -o $(exclude_items_from_list $ost_indices $start_ost_idx)"
5759 cmd+=" -i $start_ost_idx $file"
5761 eval $cmd && error "index $start_ost_idx should be in $ost_indices"
5763 # 5. Specifying OST indices for directory should fail with ENOSUPP.
5764 local dir=$DIR/$tdir/$tdir
5765 mkdir $dir || error "mkdir $dir failed"
5766 cmd="$SETSTRIPE -o $ost_indices $dir"
5768 eval $cmd && error "$cmd should fail, specifying OST indices" \
5769 "for directory is not supported"
5773 run_test 82a "specify OSTs for file (succeed) or directory (fail)"
5778 # Remove OSTs from a pool and destroy the pool.
5779 destroy_pool $ost_pool || true
5784 # Test 82b is run to ensure that if the user supplies a pool with a specific
5785 # stripe layout that it behaves proprerly. It should fail in the case that
5786 # the supplied OST index list points to OSTs not contained in the user
5788 test_82b() { # LU-4665
5789 [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.6.54) ]] ||
5790 { skip "Need MDS version at least 2.6.54" && return; }
5791 [[ $OSTCOUNT -ge 4 ]] || { skip_env "Need at least 4 OSTs" && return; }
5797 # Format OSTs with random sparse indices.
5801 local LOV_V1_INSANE_STRIPE_COUNT=65532
5802 for i in $(seq $OSTCOUNT); do
5803 index=$(((RANDOM * 2) % LOV_V1_INSANE_STRIPE_COUNT))
5804 ost_indices+=" $index"
5806 ost_indices=$(comma_list $ost_indices)
5808 trap "restore_ostindex" EXIT
5809 echo -e "\nFormat $OSTCOUNT OSTs with sparse indices $ost_indices"
5810 OST_INDEX_LIST=[$ost_indices] formatall
5812 # Setup Lustre filesystem.
5813 start_mgsmds || error "start_mgsmds failed"
5814 for i in $(seq $OSTCOUNT); do
5815 start ost$i $(ostdevname $i) $OST_MOUNT_OPTS ||
5816 error "start ost$i failed"
5819 mount_client $MOUNT || error "mount client $MOUNT failed"
5821 $LFS df $MOUNT || error "$LFS df $MOUNT failed"
5822 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
5824 # Create a new pool and add OSTs into it.
5825 local ost_pool=$FSNAME.$TESTNAME
5826 create_pool $ost_pool || error "create OST pool $ost_pool failed"
5829 trap "cleanup_82b" EXIT
5831 local ost_idx_in_list=${ost_indices##*,}
5832 local ost_idx_in_pool=$(exclude_items_from_list $ost_indices \
5835 local ost_targets="$FSNAME-OST["
5836 for i in ${ost_idx_in_pool//,/ }; do
5837 ost_targets=$ost_targets$(printf "%04x," $i)
5839 ost_targets="${ost_targets%,}]"
5841 local ost_targets_uuid=$(for i in ${ost_idx_in_pool//,/ }; \
5842 do printf "$FSNAME-OST%04x_UUID\n" $i; done |
5843 sort -u | tr '\n' ' ')
5845 local cmd="$LCTL pool_add $ost_pool $ost_targets"
5846 do_facet mgs $cmd || error "$cmd failed"
5847 wait_update $HOSTNAME "$LCTL get_param -n lov.$FSNAME-*.pools.$TESTNAME|
5848 sort -u | tr '\n' ' ' " "$ost_targets_uuid" ||
5849 error "wait_update $ost_pool failed"
5850 [[ -z $(list_pool $ost_pool) ]] &&
5851 error "list OST pool $ost_pool failed"
5853 # If [--pool|-p <pool_name>] is set with [--ost-list|-o <ost_indices>],
5854 # then the OSTs must be the members of the pool.
5855 local file=$DIR/$tdir/$tfile
5856 cmd="$SETSTRIPE -p $ost_pool -o $ost_idx_in_list $file"
5858 eval $cmd && error "OST with index $ost_idx_in_list should be" \
5859 "in OST pool $ost_pool"
5861 # Only select OST $ost_idx_in_list from $ost_pool for file.
5862 ost_idx_in_list=${ost_idx_in_pool#*,}
5863 cmd="$SETSTRIPE -p $ost_pool -o $ost_idx_in_list $file"
5865 eval $cmd || error "$cmd failed"
5866 cmd="$GETSTRIPE $file"
5868 eval $cmd || error "$cmd failed"
5869 check_stripe_count $file 2
5870 check_obdidx $file $ost_idx_in_list
5871 dd if=/dev/urandom of=$file count=1 bs=1M > /dev/null 2>&1 ||
5872 error "write $file failed"
5876 run_test 82b "specify OSTs for file with --pool and --ost-list options"
5879 [[ $(lustre_version_code ost1) -ge $(version_code 2.6.91) ]] ||
5880 { skip "Need OST version at least 2.6.91" && return 0; }
5881 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
5882 skip "Only applicable to ldiskfs-based MDTs"
5892 ostmnt=$(facet_mntpt ost1)
5893 fstype=$(facet_fstype ost1)
5895 # Mount the OST as an ldiskfs filesystem.
5896 log "mount the OST $dev as a $fstype filesystem"
5897 add ost1 $(mkfs_opts ost1 $dev) $FSTYPE_OPT \
5898 --reformat $dev $dev > /dev/null ||
5899 error "format ost1 error"
5901 if ! test -b $dev; then
5902 mnt_opts=$(csa_add "$OST_MOUNT_OPTS" -o loop)
5904 echo "mnt_opts $mnt_opts"
5905 do_facet ost1 mount -t $fstype $dev \
5907 # Run llverfs on the mounted ldiskfs filesystem.
5908 # It is needed to get ENOSPACE.
5909 log "run llverfs in partial mode on the OST $fstype $ostmnt"
5910 do_rpc_nodes $(facet_host ost1) run_llverfs $ostmnt -vpl \
5911 "no" || error "run_llverfs error on $fstype"
5914 log "unmount the OST $dev"
5917 # Delete file IO_scrub. Later osd_scrub_setup will try to
5918 # create "IO_scrub" but will get ENOSPACE.
5920 echo "start ost1 service on `facet_active_host ost1`"
5921 start ost1 `ostdevname 1` $OST_MOUNT_OPTS
5924 err=$(do_facet ost1 dmesg | grep "VFS: Busy inodes after unmount of")
5925 echo "string err $err"
5926 [ -z "$err" ] || error $err
5929 run_test 83 "ENOSPACE on OST doesn't cause message VFS: \
5930 Busy inodes after unmount ..."
5933 local facet=$SINGLEMDS
5934 local num=$(echo $facet | tr -d "mds")
5935 local dev=$(mdsdevname $num)
5936 local time_min=$(recovery_time_min)
5937 local recovery_duration
5938 local completed_clients
5939 local correct_clients
5942 echo "start mds service on $(facet_active_host $facet)"
5944 "-o recovery_time_hard=$time_min,recovery_time_soft=$time_min" $@ ||
5945 error "start MDS failed"
5947 start_ost || error "start OST0000 failed"
5948 start_ost2 || error "start OST0001 failed"
5950 echo "recovery_time=$time_min, timeout=$TIMEOUT, wrap_up=$wrap_up"
5952 mount_client $MOUNT1 || error "mount $MOUNT1 failed"
5953 mount_client $MOUNT2 || error "mount $MOUNT2 failed"
5954 # make sure new superblock labels are sync'd before disabling writes
5958 replay_barrier $SINGLEMDS
5959 createmany -o $DIR1/$tfile-%d 1000
5961 # We need to catch the end of recovery window to extend it.
5962 # Skip 5 requests and add delay to request handling.
5963 #define OBD_FAIL_TGT_REPLAY_DELAY 0x709 | FAIL_SKIP
5964 do_facet $SINGLEMDS "lctl set_param fail_loc=0x20000709 fail_val=5"
5966 facet_failover --fsck $SINGLEMDS || error "failover: $?"
5969 echo "recovery status"
5970 do_facet $SINGLEMDS \
5971 "$LCTL get_param -n mdt.$FSNAME-MDT0000.recovery_status"
5973 recovery_duration=$(do_facet $SINGLEMDS \
5974 "$LCTL get_param -n mdt.$FSNAME-MDT0000.recovery_status" |
5975 awk '/recovery_duration/ { print $2 }')
5976 (( $recovery_duration > $time_min + $wrap_up )) &&
5977 error "recovery_duration > recovery_time_hard + wrap up"
5978 completed_clients=$(do_facet $SINGLEMDS \
5979 "$LCTL get_param -n mdt.$FSNAME-MDT0000.recovery_status" |
5980 awk '/completed_clients/ { print $2 }')
5982 correct_clients="$MDSCOUNT/$((MDSCOUNT+1))"
5983 [ "$completed_clients" = "${correct_clients}" ] ||
5984 error "$completed_clients != $correct_clients"
5986 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
5987 umount_client $MOUNT1
5988 umount_client $MOUNT2
5994 run_test 84 "check recovery_hard_time"
5997 [[ $(lustre_version_code ost1) -ge $(version_code 2.7.55) ]] ||
5998 { skip "Need OST version at least 2.7.55" && return 0; }
5999 ##define OBD_FAIL_OSD_OST_EA_FID_SET 0x197
6000 do_facet ost1 "lctl set_param fail_loc=0x197"
6004 run_test 85 "osd_ost init: fail ea_fid_set"
6009 # ost1 has already registered to the MGS before the reformat.
6010 # So after reformatting it with option "-G", it could not be
6011 # mounted to the MGS. Cleanup the system for subsequent tests.
6016 local server_version=$(lustre_version_code $SINGLEMDS)
6017 [ "$(facet_fstype ost1)" = "zfs" ] &&
6018 skip "LU-6442: no such mkfs params for ZFS OSTs" && return
6019 [[ $server_version -ge $(version_code 2.7.56) ]] ||
6020 { skip "Need server version newer than 2.7.55"; return 0; }
6022 local OST_OPTS="$(mkfs_opts ost1 $(ostdevname 1)) \
6023 --reformat $(ostdevname 1) $(ostvdevname 1)"
6026 local OLDSIZE=$(do_facet ost1 "$DEBUGFS -c -R stats $(ostdevname 1)" |
6027 awk '/Flex block group size: / { print $NF; exit; }')
6030 if [[ ${!opts} != *mkfsoptions* ]]; then
6031 eval opts=\"${!opts} \
6032 --mkfsoptions='\\\"-O flex_bg -G $NEWSIZE\\\"'\"
6034 val=${!opts//--mkfsoptions=\\\"/ \
6035 --mkfsoptions=\\\"-O flex_bg -G $NEWSIZE }
6039 echo "params: $opts"
6041 trap cleanup_86 EXIT ERR
6044 add ost1 $opts || error "add ost1 failed with new params"
6046 local FOUNDSIZE=$(do_facet ost1 "$DEBUGFS -c -R stats $(ostdevname 1)" |
6047 awk '/Flex block group size: / { print $NF; exit; }')
6049 [[ $FOUNDSIZE == $NEWSIZE ]] ||
6050 error "Flex block group size: $FOUNDSIZE, expected: $NEWSIZE"
6054 run_test 86 "Replacing mkfs.lustre -G option"
6056 test_87() { #LU-6544
6057 [[ $(lustre_version_code $SINGLEMDS1) -ge $(version_code 2.7.56) ]] ||
6058 { skip "Need MDS version at least 2.7.56" && return; }
6059 [[ $(facet_fstype $SINGLEMDS) != ldiskfs ]] &&
6060 { skip "Only applicable to ldiskfs-based MDTs" && return; }
6061 [[ $OSTCOUNT -gt 69 ]] &&
6062 { skip "Ignore wide striping situation" && return; }
6064 local mdsdev=$(mdsdevname 1)
6065 local mdsvdev=$(mdsvdevname 1)
6066 local file=$DIR/$tfile
6067 local mntpt=$(facet_mntpt $SINGLEMDS)
6068 local used_xattr_blk=0
6069 local inode_size=${1:-512}
6071 local xtest="trusted.test"
6076 #Please see LU-6544 for MDT inode size calculation
6077 if [ $OSTCOUNT -gt 26 ]; then
6079 elif [ $OSTCOUNT -gt 5 ]; then
6082 left_size=$(expr $inode_size - \
6085 32 - $OSTCOUNT \* 24 - 16 - 3 - \
6087 24 - 18 - $(expr length $tfile) - 16 - 4)
6088 if [ $left_size -le 0 ]; then
6089 echo "No space($left_size) is expected in inode."
6090 echo "Try 1-byte xattr instead to verify this."
6093 echo "Estimate: at most $left_size-byte space left in inode."
6099 add mds1 $(mkfs_opts mds1 ${mdsdev}) --stripe-count-hint=$OSTCOUNT \
6100 --reformat $mdsdev $mdsvdev || error "add mds1 failed"
6101 start_mdt 1 > /dev/null || error "start mdt1 failed"
6102 for i in $(seq $OSTCOUNT); do
6103 start ost$i $(ostdevname $i) $OST_MOUNT_OPTS > /dev/null ||
6104 error "start ost$i failed"
6106 mount_client $MOUNT > /dev/null || error "mount client $MOUNT failed"
6107 check_mount || error "check client $MOUNT failed"
6110 $SETSTRIPE -c -1 $file || error "$SETSTRIPE -c -1 $file failed"
6111 $GETSTRIPE $file || error "$GETSTRIPE $file failed"
6112 i=$($GETSTRIPE -c $file)
6113 if [ $i -ne $OSTCOUNT ]; then
6114 left_size=$(expr $left_size + $(expr $OSTCOUNT - $i) \* 24)
6115 echo -n "Since only $i out $OSTCOUNT OSTs are used, "
6116 echo -n "the expected left space is changed to "
6117 echo "$left_size bytes at most."
6119 value=$(generate_string $left_size)
6120 setfattr -n $xtest -v $value $file
6121 orig=$(get_xattr_value $xtest $file)
6122 [[ "$orig" != "$value" ]] && error "$xtest changed"
6124 #Verify if inode has some expected space left
6125 umount $MOUNT > /dev/null || error "umount $MOUNT failed"
6126 stop_mdt 1 > /dev/null || error "stop mdt1 failed"
6127 mount_ldiskfs $SINGLEMDS || error "mount -t ldiskfs $SINGLEMDS failed"
6129 do_facet $SINGLEMDS ls -sal $mntpt/ROOT/$tfile
6130 used_xattr_blk=$(do_facet $SINGLEMDS ls -s $mntpt/ROOT/$tfile |
6132 [[ $used_xattr_blk -eq 0 ]] &&
6133 error "Please check MDS inode size calculation: \
6134 more than $left_size-byte space left in inode."
6135 echo "Verified: at most $left_size-byte space left in inode."
6139 run_test 87 "check if MDT inode can hold EAs with N stripes properly"
6142 [ "$(facet_fstype mds1)" == "zfs" ] &&
6143 skip "LU-6662: no implementation for ZFS" && return
6147 add mds1 $(mkfs_opts mds1 $(mdsdevname 1)) \
6148 --reformat $(mdsdevname 1) || error "add mds1 failed"
6150 do_facet mds1 "$TUNEFS $(mdsdevname 1) |
6151 grep -e \".*opts:.*errors=remount-ro.*\"" ||
6152 error "default mount options is missing"
6154 add mds1 $(mkfs_opts mds1 $(mdsdevname 1)) \
6155 --mountfsoptions="user_xattr,errors=panic" \
6156 --reformat $(mdsdevname 1) || error "add mds1 failed"
6158 do_facet mds1 "$TUNEFS $(mdsdevname 1) |
6159 grep -e \".*opts:.*errors=panic.*\"" ||
6160 error "user can't override default mount options"
6162 run_test 88 "check the default mount options can be overridden"
6164 test_89() { # LU-7131
6165 [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.9.54) ]] ||
6166 { skip "Need MDT version at least 2.9.54" && return 0; }
6168 local key=failover.node
6169 local val1=192.0.2.254@tcp0 # Reserved IPs, see RFC 5735
6170 local val2=192.0.2.255@tcp0
6171 local mdsdev=$(mdsdevname 1)
6176 [ $(facet_fstype mds1) == zfs ] && import_zpool mds1
6177 # Check that parameters are added correctly
6178 echo "tunefs --param $key=$val1"
6179 do_facet mds "$TUNEFS --param $key=$val1 $mdsdev >/dev/null" ||
6180 error "tunefs --param $key=$val1 failed"
6181 params=$(do_facet mds $TUNEFS --dryrun $mdsdev) ||
6182 error "tunefs --dryrun failed"
6183 params=${params##*Parameters:}
6184 params=${params%%exiting*}
6185 [ $(echo $params | tr ' ' '\n' | grep -c $key=$val1) = "1" ] ||
6186 error "on-disk parameter not added correctly via tunefs"
6188 # Check that parameters replace existing instances when added
6189 echo "tunefs --param $key=$val2"
6190 do_facet mds "$TUNEFS --param $key=$val2 $mdsdev >/dev/null" ||
6191 error "tunefs --param $key=$val2 failed"
6192 params=$(do_facet mds $TUNEFS --dryrun $mdsdev) ||
6193 error "tunefs --dryrun failed"
6194 params=${params##*Parameters:}
6195 params=${params%%exiting*}
6196 [ $(echo $params | tr ' ' '\n' | grep -c $key=) = "1" ] ||
6197 error "on-disk parameter not replaced via tunefs"
6198 [ $(echo $params | tr ' ' '\n' | grep -c $key=$val2) = "1" ] ||
6199 error "on-disk parameter not replaced correctly via tunefs"
6201 # Check that a parameter is erased properly
6202 echo "tunefs --erase-param $key"
6203 do_facet mds "$TUNEFS --erase-param $key $mdsdev >/dev/null" ||
6204 error "tunefs --erase-param $key failed"
6205 params=$(do_facet mds $TUNEFS --dryrun $mdsdev) ||
6206 error "tunefs --dryrun failed"
6207 params=${params##*Parameters:}
6208 params=${params%%exiting*}
6209 [ $(echo $params | tr ' ' '\n' | grep -c $key=) = "0" ] ||
6210 error "on-disk parameter not erased correctly via tunefs"
6212 # Check that all the parameters are erased
6213 echo "tunefs --erase-params"
6214 do_facet mds "$TUNEFS --erase-params $mdsdev >/dev/null" ||
6215 error "tunefs --erase-params failed"
6216 params=$(do_facet mds $TUNEFS --dryrun $mdsdev) ||
6217 error "tunefs --dryrun failed"
6218 params=${params##*Parameters:}
6219 params=${params%%exiting*}
6221 error "all on-disk parameters not erased correctly via tunefs"
6223 # Check the order of options --erase-params and --param
6224 echo "tunefs --param $key=$val1 --erase-params"
6226 "$TUNEFS --param $key=$val1 --erase-params $mdsdev >/dev/null"||
6227 error "tunefs --param $key=$val1 --erase-params failed"
6228 params=$(do_facet mds $TUNEFS --dryrun $mdsdev) ||
6229 error "tunefs --dryrun failed"
6230 params=${params##*Parameters:}
6231 params=${params%%exiting*}
6232 [ $(echo $params | tr ' ' '\n') == "$key=$val1" ] ||
6233 error "on-disk param not added correctly with --erase-params"
6237 run_test 89 "check tunefs --param and --erase-param{s} options"
6240 # $2 (optional) value of max_mod_rpcs_in_flight to set
6241 check_max_mod_rpcs_in_flight() {
6249 idx=$(printf "%04x" $($LFS getdirstripe -i $dir))
6250 facet="mds$((0x$idx + 1))"
6252 if [ -z "$mmr" ]; then
6253 # get value of max_mod_rcps_in_flight
6254 mmr=$($LCTL get_param -n \
6255 mdc.$FSNAME-MDT$idx-mdc-*.max_mod_rpcs_in_flight) ||
6256 error "Unable to get max_mod_rpcs_in_flight"
6257 echo "max_mod_rcps_in_flight is $mmr"
6259 # set value of max_mod_rpcs_in_flight
6261 mdc.$FSNAME-MDT$idx-mdc-*.max_mod_rpcs_in_flight=$mmr ||
6262 error "Unable to set max_mod_rpcs_in_flight to $mmr"
6263 echo "max_mod_rpcs_in_flight set to $mmr"
6266 # create mmr+1 files
6267 echo "creating $((mmr + 1)) files ..."
6269 for i in $(seq $((mmr + 1))); do
6275 # consumes mmr-1 modify RPC slots
6276 #define OBD_FAIL_MDS_REINT_MULTI_NET 0x159
6277 # drop requests on MDT so that RPC slots are consumed
6278 # during all the request resend interval
6279 do_facet $facet "$LCTL set_param fail_loc=0x159"
6280 echo "launch $((mmr - 1)) chmod in parallel ..."
6281 for i in $(seq $((mmr - 1))); do
6282 chmod 0600 $dir/file-$i &
6286 # send one additional modify RPC
6287 do_facet $facet "$LCTL set_param fail_loc=0"
6288 echo "launch 1 additional chmod in parallel ..."
6289 chmod 0600 $dir/file-$mmr &
6292 # check this additional modify RPC get a modify RPC slot
6293 # and succeed its operation
6294 checkstat -vp 0600 $dir/file-$mmr ||
6295 error "Unable to send $mmr modify RPCs in parallel"
6300 # consumes mmr modify RPC slots
6301 #define OBD_FAIL_MDS_REINT_MULTI_NET 0x159
6302 # drop requests on MDT so that RPC slots are consumed
6303 # during all the request resend interval
6304 do_facet $facet "$LCTL set_param fail_loc=0x159"
6305 echo "launch $mmr chmod in parallel ..."
6306 for i in $(seq $mmr); do
6307 chmod 0666 $dir/file-$i &
6311 # send one additional modify RPC
6312 do_facet $facet "$LCTL set_param fail_loc=0"
6313 echo "launch 1 additional chmod in parallel ..."
6314 chmod 0666 $dir/file-$((mmr + 1)) &
6317 # check this additional modify RPC blocked getting a modify RPC slot
6318 checkstat -vp 0644 $dir/file-$((mmr + 1)) ||
6319 error "Unexpectedly send $(($mmr + 1)) modify RPCs in parallel"
6325 if ! combined_mgs_mds ; then
6330 [[ $($LCTL get_param mdc.*.import |
6331 grep "connect_flags:.*multi_mod_rpc") ]] ||
6332 { skip "Need MDC with 'multi_mod_rpcs' feature"; return 0; }
6334 # check default value
6335 $LFS mkdir -c1 $DIR/$tdir || error "mkdir $DIR/$tdir failed"
6336 check_max_mod_rpcs_in_flight $DIR/$tdir
6340 run_test 90a "check max_mod_rpcs_in_flight is enforced"
6350 [[ $($LCTL get_param mdc.*.import |
6351 grep "connect_flags:.*multi_mod_rpc") ]] ||
6352 { skip "Need MDC with 'multi_mod_rpcs' feature"; return 0; }
6355 # update max_mod_rpcs_in_flight
6356 $LFS mkdir -c1 $DIR/${tdir}1 || error "mkdir $DIR/${tdir}1 failed"
6357 check_max_mod_rpcs_in_flight $DIR/${tdir}1 1
6360 # check client is able to send multiple modify RPCs in paralell
6361 tmp=$($LCTL get_param -n mdc.$FSNAME-MDT*-mdc-*.import |
6362 grep -c "multi_mod_rpcs")
6363 if [ "$tmp" -ne $MDSCOUNT ]; then
6364 echo "Client not able to send multiple modify RPCs in parallel"
6369 # update max_mod_rpcs_in_flight
6370 $LFS mkdir -c1 $DIR/${tdir}2 || error "mkdir $DIR/${tdir}2 failed"
6371 check_max_mod_rpcs_in_flight $DIR/${tdir}2 5
6374 $LFS mkdir -c1 $DIR/${tdir}3 || error "mkdir $DIR/${tdir}3 failed"
6375 idx=$(printf "%04x" $($LFS getdirstripe -i $DIR/${tdir}3))
6376 facet="mds$((0x$idx + 1))"
6378 # save MDT max_mod_rpcs_per_client
6379 mmrpc=$(do_facet $facet \
6380 cat /sys/module/mdt/parameters/max_mod_rpcs_per_client)
6382 # update max_mod_rpcs_in_flight
6383 umount_client $MOUNT
6385 "echo 16 > /sys/module/mdt/parameters/max_mod_rpcs_per_client"
6387 $LCTL set_param mdc.$FSNAME-MDT$idx-mdc-*.max_rpcs_in_flight=17
6388 check_max_mod_rpcs_in_flight $DIR/${tdir}3 16
6390 # restore MDT max_mod_rpcs_per_client initial value
6392 "echo $mmrpc > /sys/module/mdt/parameters/max_mod_rpcs_per_client"
6394 rm -rf $DIR/${tdir}?
6397 run_test 90b "check max_mod_rpcs_in_flight is enforced after update"
6406 [[ $($LCTL get_param mdc.*.import |
6407 grep "connect_flags:.*multi_mod_rpc") ]] ||
6408 { skip "Need MDC with 'multi_mod_rpcs' feature"; return 0; }
6410 # check client is able to send multiple modify RPCs in paralell
6411 tmp=$($LCTL get_param -n mdc.$FSNAME-MDT*-mdc-*.import |
6412 grep -c "multi_mod_rpcs")
6413 if [ "$tmp" -ne $MDSCOUNT ]; then
6414 skip "Client not able to send multiple modify RPCs in parallel"
6419 # get max_rpcs_in_flight value
6420 mrif=$($LCTL get_param -n mdc.$FSNAME-MDT0000-mdc-*.max_rpcs_in_flight)
6421 echo "max_rpcs_in_flight is $mrif"
6423 # get MDT max_mod_rpcs_per_client
6424 mmrpc=$(do_facet mds1 \
6425 cat /sys/module/mdt/parameters/max_mod_rpcs_per_client)
6426 echo "max_mod_rpcs_per_client is $mmrpc"
6429 # attempt to set max_mod_rpcs_in_flight to max_rpcs_in_flight value
6430 # prerequisite: set max_mod_rpcs_per_client to max_rpcs_in_flight value
6431 umount_client $MOUNT
6433 "echo $mrif > /sys/module/mdt/parameters/max_mod_rpcs_per_client"
6437 mdc.$FSNAME-MDT0000-mdc-*.max_mod_rpcs_in_flight=$mrif &&
6438 error "set max_mod_rpcs_in_flight to $mrif should fail"
6440 umount_client $MOUNT
6442 "echo $mmrpc > /sys/module/mdt/parameters/max_mod_rpcs_per_client"
6446 # attempt to set max_mod_rpcs_in_flight to max_mod_rpcs_per_client+1
6447 # prerequisite: set max_rpcs_in_flight to max_mod_rpcs_per_client+2
6449 mdc.$FSNAME-MDT0000-mdc-*.max_rpcs_in_flight=$((mmrpc + 2))
6452 mdc.$FSNAME-MDT0000-mdc-*.max_mod_rpcs_in_flight=$((mmrpc + 1)) &&
6453 error "set max_mod_rpcs_in_flight to $((mmrpc + 1)) should fail"
6457 run_test 90c "check max_mod_rpcs_in_flight update limits"
6468 [[ $($LCTL get_param mdc.*.import |
6469 grep "connect_flags:.*multi_mod_rpc") ]] ||
6470 { skip "Need MDC with 'multi_mod_rpcs' feature"; return 0; }
6472 $LFS mkdir -c1 $DIR/$tdir || error "mkdir $DIR/$tdir failed"
6473 idx=$(printf "%04x" $($LFS getdirstripe -i $DIR/$tdir))
6474 facet="mds$((0x$idx + 1))"
6476 # check client version supports multislots
6477 tmp=$($LCTL get_param -N \
6478 mdc.$FSNAME-MDT$idx-mdc-*.max_mod_rpcs_in_flight)
6479 if [ -z "$tmp" ]; then
6480 skip "Client does not support multiple modify RPCs in flight"
6485 # get current value of max_mod_rcps_in_flight
6486 mmr=$($LCTL get_param -n \
6487 mdc.$FSNAME-MDT$idx-mdc-*.max_mod_rpcs_in_flight)
6488 echo "max_mod_rcps_in_flight is $mmr"
6491 echo "creating $mmr files ..."
6493 for i in $(seq $mmr); do
6494 touch $DIR/$tdir/file-$i
6497 # prepare for close RPC
6498 multiop_bg_pause $DIR/$tdir/file-close O_c
6501 # consumes mmr modify RPC slots
6502 #define OBD_FAIL_MDS_REINT_MULTI_NET 0x159
6503 # drop requests on MDT so that RPC slots are consumed
6504 # during all the request resend interval
6505 do_facet $facet "$LCTL set_param fail_loc=0x159"
6506 echo "launch $mmr chmod in parallel ..."
6507 for i in $(seq $mmr); do
6508 chmod 0600 $DIR/$tdir/file-$i &
6511 # send one additional close RPC
6512 do_facet $facet "$LCTL set_param fail_loc=0"
6513 echo "launch 1 additional close in parallel ..."
6515 cancel_lru_locks mdc
6518 # check this additional close RPC get a modify RPC slot
6519 # and multiop process completed
6520 [ -d /proc/$pid ] &&
6521 error "Unable to send the additional close RPC in parallel"
6526 run_test 90d "check one close RPC is allowed above max_mod_rpcs_in_flight"
6528 check_uuid_on_ost() {
6530 do_facet ost1 "$LCTL get_param obdfilter.${FSNAME}*.exports.'$nid'.uuid"
6533 check_uuid_on_mdt() {
6535 do_facet $SINGLEMDS "$LCTL get_param mdt.${FSNAME}*.exports.'$nid'.uuid"
6543 [[ $(lustre_version_code ost1) -ge $(version_code 2.7.63) ]] ||
6544 { skip "Need OST version at least 2.7.63" && return 0; }
6545 [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.7.63) ]] ||
6546 { skip "Need MDT version at least 2.7.63" && return 0; }
6548 start_mds || error "MDS start failed"
6549 start_ost || error "unable to start OST"
6550 mount_client $MOUNT || error "client start failed"
6551 check_mount || error "check_mount failed"
6554 nid=$($LCTL list_nids | head -1 | sed "s/\./\\\./g")
6558 uuid=$(get_client_uuid $MOUNT)
6560 echo "list nids on mdt:"
6561 do_facet $SINGLEMDS "$LCTL list_param mdt.${FSNAME}*.exports.*"
6562 echo "uuid from $nid:"
6563 do_facet $SINGLEMDS "$LCTL get_param mdt.${FSNAME}*.exports.'$nid'.uuid"
6565 found=$(check_uuid_on_mdt $nid | grep $uuid)
6566 [ -z "$found" ] && error "can't find $uuid $nid on MDT"
6567 found=$(check_uuid_on_ost $nid | grep $uuid)
6568 [ -z "$found" ] && error "can't find $uuid $nid on OST"
6570 # umount the client so it won't reconnect
6571 manual_umount_client --force || error "failed to umount $?"
6572 # shouldn't disappear on MDS after forced umount
6573 found=$(check_uuid_on_mdt $nid | grep $uuid)
6574 [ -z "$found" ] && error "can't find $uuid $nid"
6577 do_facet $SINGLEMDS \
6578 "$LCTL set_param -n mdt.${mds1_svc}.evict_client nid:$nid"
6580 found=$(check_uuid_on_mdt $nid | grep $uuid)
6581 [ -n "$found" ] && error "found $uuid $nid on MDT"
6582 found=$(check_uuid_on_ost $nid | grep $uuid)
6583 [ -n "$found" ] && error "found $uuid $nid on OST"
6585 # check it didn't reconnect (being umounted)
6586 sleep $((TIMEOUT+1))
6587 found=$(check_uuid_on_mdt $nid | grep $uuid)
6588 [ -n "$found" ] && error "found $uuid $nid on MDT"
6589 found=$(check_uuid_on_ost $nid | grep $uuid)
6590 [ -n "$found" ] && error "found $uuid $nid on OST"
6594 run_test 91 "evict-by-nid support"
6596 generate_ldev_conf() {
6597 # generate an ldev.conf file
6598 local ldevconfpath=$1
6600 local fsldevformat=""
6603 fstype=$(facet_fstype mgs)
6604 if [ "$fstype" == "zfs" ]; then
6605 fsldevformat="$fstype:"
6610 printf "%s\t-\t%s-MGS0000\t%s%s\n" \
6614 $(mgsdevname) > $ldevconfpath
6616 local mdsfo_host=$mdsfailover_HOST;
6617 if [ -z "$mdsfo_host" ]; then
6621 for num in $(seq $MDSCOUNT); do
6622 fstype=$(facet_fstype mds$num)
6623 if [ "$fstype" == "zfs" ]; then
6624 fsldevformat="$fstype:"
6629 printf "%s\t%s\t%s-MDT%04d\t%s%s\n" \
6635 $(mdsdevname $num) >> $ldevconfpath
6638 local ostfo_host=$ostfailover_HOST;
6639 if [ -z "$ostfo_host" ]; then
6643 for num in $(seq $OSTCOUNT); do
6644 fstype=$(facet_fstype ost$num)
6645 if [ "$fstype" == "zfs" ]; then
6646 fsldevformat="$fstype:"
6651 printf "%s\t%s\t%s-OST%04d\t%s%s\n" \
6657 $(ostdevname $num) >> $ldevconfpath
6660 echo "----- $ldevconfpath -----"
6662 echo "--- END $ldevconfpath ---"
6667 # generate a nids file (mapping between hostname to nid)
6668 # looks like we only have the MGS nid available to us
6669 # so just echo that to a file
6671 echo -e "${mgs_HOST}\t${MGSNID}" > $nidspath
6673 echo "----- $nidspath -----"
6675 echo "--- END $nidspath ---"
6678 compare_ldev_output() {
6682 sort $expected_output -o $expected_output
6683 sort $ldev_output -o $ldev_output
6685 echo "-- START OF LDEV OUTPUT --"
6687 echo "--- END OF LDEV OUTPUT ---"
6689 echo "-- START OF EXPECTED OUTPUT --"
6690 cat $expected_output
6691 echo "--- END OF EXPECTED OUTPUT ---"
6693 diff $expected_output $ldev_output
6698 if [ -z "$LDEV" ]; then
6699 error "ldev is missing!"
6702 local LDEVCONFPATH=$TMP/ldev.conf
6703 local NIDSPATH=$TMP/nids
6705 echo "Host is $(hostname)"
6707 generate_ldev_conf $LDEVCONFPATH
6708 generate_nids $NIDSPATH
6710 # echo the mgs nid and compare it to environment variable MGSNID
6711 # also, ldev.conf and nids is a server side thing, use the OSS
6714 output=$($LDEV -c $LDEVCONFPATH -H $ost_HOST -n $NIDSPATH echo %m)
6716 echo "-- START OF LDEV OUTPUT --"
6718 echo "--- END OF LDEV OUTPUT ---"
6720 # ldev failed, error
6721 if [ $? -ne 0 ]; then
6722 rm $LDEVCONFPATH $NIDSPATH
6723 error "ldev failed to execute!"
6726 # need to process multiple lines because of combined MGS and MDS
6727 echo -e $output | awk '{ print $2 }' | while read -r line ; do
6728 if [ "$line" != "$MGSNID" ]; then
6729 rm $LDEVCONFPATH $NIDSPATH
6730 error "ldev failed mgs nid '$line', expected '$MGSNID'"
6734 rm $LDEVCONFPATH $NIDSPATH
6736 run_test 92 "ldev returns MGS NID correctly in command substitution"
6739 [ $MDSCOUNT -lt 3 ] && skip "needs >= 3 MDTs" && return
6742 #start mgs or mgs/mdt0
6743 if ! combined_mgs_mds ; then
6750 start_ost || error "OST0 start fail"
6752 #define OBD_FAIL_MGS_WRITE_TARGET_DELAY 0x90e
6753 do_facet mgs "$LCTL set_param fail_val = 10 fail_loc=0x8000090e"
6754 for num in $(seq 2 $MDSCOUNT); do
6758 mount_client $MOUNT || error "mount client fails"
6759 wait_osc_import_state mds ost FULL
6760 wait_osc_import_state client ost FULL
6761 check_mount || error "check_mount failed"
6763 cleanup || error "cleanup failed with $?"
6765 run_test 93 "register mulitple MDT at the same time"
6768 if [ -z "$LDEV" ]; then
6769 error "ldev is missing!"
6772 local LDEVCONFPATH=$TMP/ldev.conf
6773 local NIDSPATH=$TMP/nids
6775 generate_ldev_conf $LDEVCONFPATH
6776 generate_nids $NIDSPATH
6778 local LDEV_OUTPUT=$TMP/ldev-output.txt
6779 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -F $FSNAME > $LDEV_OUTPUT
6781 # ldev failed, error
6782 if [ $? -ne 0 ]; then
6783 rm $LDEVCONFPATH $NIDSPATH $LDEV_OUTPUT
6784 error "ldev failed to execute!"
6788 local EXPECTED_OUTPUT=$TMP/ldev-expected.txt
6790 printf "%s-MGS0000\n" $FSNAME > $EXPECTED_OUTPUT
6792 for num in $(seq $MDSCOUNT); do
6793 printf "%s-MDT%04d\n" $FSNAME $num >> $EXPECTED_OUTPUT
6796 for num in $(seq $OSTCOUNT); do
6797 printf "%s-OST%04d\n" $FSNAME $num >> $EXPECTED_OUTPUT
6800 compare_ldev_output $LDEV_OUTPUT $EXPECTED_OUTPUT
6802 if [ $? -ne 0 ]; then
6803 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
6804 error "ldev failed to produce the correct hostlist!"
6807 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
6809 run_test 94 "ldev outputs correct labels for file system name query"
6812 if [ -z "$LDEV" ]; then
6813 error "ldev is missing!"
6816 local LDEVCONFPATH=$TMP/ldev.conf
6817 local NIDSPATH=$TMP/nids
6819 generate_ldev_conf $LDEVCONFPATH
6820 generate_nids $NIDSPATH
6824 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -F $FSNAME &>/dev/null
6825 if [ $? -ne 0 ]; then
6826 rm $LDEVCONFPATH $NIDSPATH
6827 error "ldev label filtering w/ -F failed!"
6831 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -l &>/dev/null
6832 if [ $? -ne 0 ]; then
6833 rm $LDEVCONFPATH $NIDSPATH
6834 error "ldev label filtering w/ -l failed!"
6838 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -f &>/dev/null
6839 if [ $? -ne 0 ]; then
6840 rm $LDEVCONFPATH $NIDSPATH
6841 error "ldev label filtering w/ -f failed!"
6845 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -a &>/dev/null
6846 if [ $? -ne 0 ]; then
6847 rm $LDEVCONFPATH $NIDSPATH
6848 error "ldev label filtering w/ -a failed!"
6853 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -a -F $FSNAME &>/dev/null
6854 if [ $? -eq 0 ]; then
6855 rm $LDEVCONFPATH $NIDSPATH
6856 error "ldev label filtering w/ -a and -F incorrectly succeeded"
6860 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -a -f &>/dev/null
6861 if [ $? -eq 0 ]; then
6862 rm $LDEVCONFPATH $NIDSPATH
6863 error "ldev label filtering w/ -a and -f incorrectly succeeded"
6867 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -a -l &>/dev/null
6868 if [ $? -eq 0 ]; then
6869 rm $LDEVCONFPATH $NIDSPATH
6870 error "ldev label filtering w/ -a and -l incorrectly succeeded"
6874 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -f -l &>/dev/null
6875 if [ $? -eq 0 ]; then
6876 rm $LDEVCONFPATH $NIDSPATH
6877 error "ldev label filtering w/ -f and -l incorrectly succeeded"
6881 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -F $FSNAME -l &>/dev/null
6882 if [ $? -eq 0 ]; then
6883 rm $LDEVCONFPATH $NIDSPATH
6884 error "ldev label filtering w/ -F and -l incorrectly succeeded"
6887 # file sys & foreign
6888 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -F $FSNAME -f &>/dev/null
6889 if [ $? -eq 0 ]; then
6890 rm $LDEVCONFPATH $NIDSPATH
6891 error "ldev label filtering w/ -F and -f incorrectly succeeded"
6894 rm $LDEVCONFPATH $NIDSPATH
6896 run_test 95 "ldev should only allow one label filter"
6899 if [ -z "$LDEV" ]; then
6900 error "ldev is missing!"
6903 local LDEVCONFPATH=$TMP/ldev.conf
6904 local NIDSPATH=$TMP/nids
6906 generate_ldev_conf $LDEVCONFPATH
6907 generate_nids $NIDSPATH
6909 local LDEV_OUTPUT=$TMP/ldev-output.txt
6910 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -H $mgs_HOST \
6912 awk '{print $2}' > $LDEV_OUTPUT
6914 # ldev failed, error
6915 if [ $? -ne 0 ]; then
6916 rm $LDEVCONFPATH $NIDSPATH $LDEV_OUTPUT
6917 error "ldev failed to execute!"
6921 local EXPECTED_OUTPUT=$TMP/ldev-expected-output.txt
6923 echo "$mgs_HOST-$(facet_fstype mgs)" > $EXPECTED_OUTPUT
6925 if [ "$mgs_HOST" == "$mds_HOST" ]; then
6926 for num in $(seq $MDSCOUNT); do
6927 echo "$mds_HOST-$(facet_fstype mds$num)" \
6932 if [ "$mgs_HOST" == "$ost_HOST" ]; then
6933 for num in $(seq $OSTCOUNT); do
6934 echo "$ost_HOST-$(facet_fstype ost$num)" \
6939 compare_ldev_output $LDEV_OUTPUT $EXPECTED_OUTPUT
6941 if [ $? -ne 0 ]; then
6942 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
6943 error "ldev failed to produce the correct output!"
6946 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
6948 run_test 96 "ldev returns hostname and backend fs correctly in command sub"
6951 if [ -z "$LDEV" ]; then
6952 error "ldev is missing!"
6955 local LDEVCONFPATH=$TMP/ldev.conf
6956 local NIDSPATH=$TMP/nids
6958 generate_ldev_conf $LDEVCONFPATH
6959 generate_nids $NIDSPATH
6961 local LDEV_OUTPUT=$TMP/ldev-output.txt
6962 local EXPECTED_OUTPUT=$TMP/ldev-expected-output.txt
6964 echo -e "\nMDT role"
6965 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -F $FSNAME -R mdt > $LDEV_OUTPUT
6967 if [ $? -ne 0 ]; then
6968 rm $LDEVCONFPATH $NIDSPATH $LDEV_OUTPUT
6969 error "ldev failed to execute for mdt role!"
6972 for num in $(seq $MDSCOUNT); do
6973 printf "%s-MDT%04d\n" $FSNAME $num >> $EXPECTED_OUTPUT
6976 compare_ldev_output $LDEV_OUTPUT $EXPECTED_OUTPUT
6978 if [ $? -ne 0 ]; then
6979 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
6980 error "ldev failed to produce the correct output for mdt role!"
6983 echo -e "\nOST role"
6984 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -F $FSNAME -R ost > $LDEV_OUTPUT
6986 if [ $? -ne 0 ]; then
6987 rm $LDEVCONFPATH $NIDSPATH $LDEV_OUTPUT $EXPECTED_OUTPUT
6988 error "ldev failed to execute for ost role!"
6992 for num in $(seq $OSTCOUNT); do
6993 printf "%s-OST%04d\n" $FSNAME $num >> $EXPECTED_OUTPUT
6996 compare_ldev_output $LDEV_OUTPUT $EXPECTED_OUTPUT
6998 if [ $? -ne 0 ]; then
6999 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
7000 error "ldev failed to produce the correct output for ost role!"
7003 echo -e "\nMGS role"
7004 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -F $FSNAME -R mgs > $LDEV_OUTPUT
7006 if [ $? -ne 0 ]; then
7007 rm $LDEVCONFPATH $NIDSPATH $LDEV_OUTPUT $EXPECTED_OUTPUT
7008 error "ldev failed to execute for mgs role!"
7011 printf "%s-MGS0000\n" $FSNAME > $EXPECTED_OUTPUT
7013 compare_ldev_output $LDEV_OUTPUT $EXPECTED_OUTPUT
7015 if [ $? -ne 0 ]; then
7016 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
7017 error "ldev failed to produce the correct output for mgs role!"
7020 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
7022 run_test 97 "ldev returns correct ouput when querying based on role"
7027 local temp=$MDS_MOUNT_OPTS
7030 check_mount || error "mount failed"
7031 mountopt="user_xattr"
7032 for ((x = 1; x <= 400; x++)); do
7033 mountopt="$mountopt,user_xattr"
7035 remount_client $mountopt $MOUNT 2>&1 | grep "too long" ||
7036 error "Buffer overflow check failed"
7037 cleanup || error "cleanup failed"
7039 run_test 98 "Buffer-overflow check while parsing mount_opts"
7043 [[ $(facet_fstype ost1) != ldiskfs ]] &&
7044 { skip "Only applicable to ldiskfs-based OSTs" && return; }
7045 [[ $(lustre_version_code ost1) -ge $(version_code 2.8.57) ]] ||
7046 { skip "Need OST version at least 2.8.57" && return 0; }
7048 local ost_opts="$(mkfs_opts ost1 $(ostdevname 1)) \
7049 --reformat $(ostdevname 1) $(ostvdevname 1)"
7050 do_facet ost1 $DEBUGFS -c -R stats `ostdevname 1` | grep "meta_bg" &&
7051 skip "meta_bg already set" && return
7054 if [[ ${!opts} != *mkfsoptions* ]]; then
7055 eval opts=\"${!opts} \
7056 --mkfsoptions='\\\"-O ^resize_inode,meta_bg\\\"'\"
7058 local val=${!opts//--mkfsoptions=\\\"/ \
7059 --mkfsoptions=\\\"-O ^resize_inode,meta_bg }
7063 echo "params: $opts"
7065 add ost1 $opts || error "add ost1 failed with new params"
7067 do_facet ost1 $DEBUGFS -c -R stats `ostdevname 1` | grep "meta_bg" ||
7068 error "meta_bg is not set"
7072 run_test 99 "Adding meta_bg option"
7076 start_mds || error "MDS start failed"
7077 start_ost || error "unable to start OST"
7078 mount_client $MOUNT || error "client start failed"
7079 check_mount || error "check_mount failed"
7088 do_facet mgs 'lshowmount -v' | awk 'BEGIN {NR == 0; rc=1} /MGS:/ {rc=0}
7089 END {exit rc}' || error "lshowmount have no output MGS"
7091 do_facet mds1 'lshowmount -v' | awk 'BEGIN {NR == 2; rc=1} /-MDT0000:/
7092 {rc=0} END {exit rc}' || error "lshowmount have no output MDT0"
7094 do_facet ost1 'lshowmount -v' | awk 'BEGIN {NR == 4; rc=1} /-OST0000:/
7095 {rc=0} END {exit rc}' || error "lshowmount have no output OST0"
7097 cleanup || error "cleanup failed with $?"
7099 run_test 100 "check lshowmount lists MGS, MDT, OST and 0@lo"
7102 local createmany_oid
7103 local dev=$FSNAME-OST0000-osc-MDT0000
7106 createmany -o $DIR1/$tfile-%d 50000 &
7108 # MDT->OST reconnection causes MDT<->OST last_id synchornisation
7109 # via osp_precreate_cleanup_orphans.
7110 for ((i = 0; i < 100; i++)); do
7111 for ((k = 0; k < 10; k++)); do
7112 do_facet $SINGLEMDS "$LCTL --device $dev deactivate;" \
7113 "$LCTL --device $dev activate"
7116 ls -asl $MOUNT | grep '???' &&
7117 (kill -9 $createmany_oid &>/dev/null; \
7118 error "File hasn't object on OST")
7120 kill -s 0 $createmany_oid || break
7122 wait $createmany_oid
7125 run_test 101 "Race MDT->OST reconnection with create"
7128 cleanup || error "cleanup failed with $?"
7130 local mds1dev=$(mdsdevname 1)
7131 local mds1mnt=$(facet_mntpt mds1)
7132 local mds1fstype=$(facet_fstype mds1)
7133 local mds1opts=$MDS_MOUNT_OPTS
7135 if [ $mds1fstype == ldiskfs ] &&
7136 ! do_facet mds1 test -b $mds1dev; then
7137 mds1opts=$(csa_add "$mds1opts" -o loop)
7139 if [[ $mds1fstype == zfs ]]; then
7140 import_zpool mds1 || return ${PIPESTATUS[0]}
7143 # unload all and only load libcfs to allow fail_loc setting
7144 do_facet mds1 lustre_rmmod || error "unable to unload modules"
7145 do_facet mds1 modprobe libcfs || error "libcfs not loaded"
7146 do_facet mds1 lsmod \| grep libcfs || error "libcfs not loaded"
7148 #define OBD_FAIL_OBDCLASS_MODULE_LOAD 0x60a
7149 do_facet mds1 "$LCTL set_param fail_loc=0x8000060a"
7151 do_facet mds1 $MOUNT_CMD $mds1dev $mds1mnt $mds1opts &&
7152 error "mdt start must fail"
7153 do_facet mds1 lsmod \| grep obdclass && error "obdclass must not load"
7155 do_facet mds1 "$LCTL set_param fail_loc=0x0"
7157 do_facet mds1 $MOUNT_CMD $mds1dev $mds1mnt $mds1opts ||
7158 error "mdt start must not fail"
7160 cleanup || error "cleanup failed with $?"
7162 run_test 102 "obdclass module cleanup upon error"
7167 echo "rename $FSNAME to $newname"
7169 if [ ! combined_mgs_mds ]; then
7170 local facet=$(mgsdevname)
7173 "$TUNEFS --fsname=$newname --rename=$FSNAME -v $facet"||
7174 error "(7) Fail to rename MGS"
7175 if [ "$(facet_fstype $facet)" = "zfs" ]; then
7176 reimport_zpool mgs $newname-mgs
7180 for num in $(seq $MDSCOUNT); do
7181 local facet=$(mdsdevname $num)
7183 do_facet mds${num} \
7184 "$TUNEFS --fsname=$newname --rename=$FSNAME -v $facet"||
7185 error "(8) Fail to rename MDT $num"
7186 if [ "$(facet_fstype $facet)" = "zfs" ]; then
7187 reimport_zpool mds${num} $newname-mdt${num}
7191 for num in $(seq $OSTCOUNT); do
7192 local facet=$(ostdevname $num)
7194 do_facet ost${num} \
7195 "$TUNEFS --fsname=$newname --rename=$FSNAME -v $facet"||
7196 error "(9) Fail to rename OST $num"
7197 if [ "$(facet_fstype $facet)" = "zfs" ]; then
7198 reimport_zpool ost${num} $newname-ost${num}
7203 test_103_set_pool() {
7207 do_facet mgs $LCTL pool_add $FSNAME.$pname ${FSNAME}-$ost_x ||
7208 error "Fail to add $ost_x to $FSNAME.$pname"
7209 wait_update $HOSTNAME \
7210 "lctl get_param -n lov.$FSNAME-clilov-*.pools.$pname |
7211 grep $ost_x" "$FSNAME-${ost_x}_UUID" ||
7212 error "$ost_x is NOT in pool $FSNAME.$pname"
7215 test_103_check_pool() {
7216 local save_fsname=$1
7219 stat $DIR/$tdir/test-framework.sh ||
7220 error "($errno) Fail to stat"
7221 do_facet mgs $LCTL pool_list $FSNAME.pool1 ||
7222 error "($errno) Fail to list $FSNAME.pool1"
7223 do_facet mgs $LCTL pool_list $FSNAME.$save_fsname ||
7224 error "($errno) Fail to list $FSNAME.$save_fsname"
7225 do_facet mgs $LCTL pool_list $FSNAME.$save_fsname |
7226 grep ${FSNAME}-OST0000 ||
7227 error "($errno) List $FSNAME.$save_fsname is invalid"
7229 local pname=$($LFS getstripe --pool $DIR/$tdir/d0)
7230 [ "$pname" = "$save_fsname" ] ||
7231 error "($errno) Unexpected pool name $pname"
7235 check_mount_and_prep
7237 mkdir $DIR/$tdir || error "(1) Fail to mkdir $DIR/$tdir"
7238 cp $LUSTRE/tests/test-framework.sh $DIR/$tdir ||
7239 error "(2) Fail to copy test-framework.sh"
7241 do_facet mgs $LCTL pool_new $FSNAME.pool1 ||
7242 error "(3) Fail to create $FSNAME.pool1"
7243 # name the pool name as the fsname
7244 do_facet mgs $LCTL pool_new $FSNAME.$FSNAME ||
7245 error "(4) Fail to create $FSNAME.$FSNAME"
7247 test_103_set_pool $FSNAME OST0000
7249 $SETSTRIPE -p $FSNAME $DIR/$tdir/d0 ||
7250 error "(6) Fail to setstripe on $DIR/$tdir/d0"
7255 test_renamefs mylustre
7257 local save_fsname=$FSNAME
7261 test_103_check_pool $save_fsname 7
7263 if [ $OSTCOUNT -ge 2 ]; then
7264 test_103_set_pool $save_fsname OST0001
7267 $SETSTRIPE -p $save_fsname $DIR/$tdir/f0 ||
7268 error "(16) Fail to setstripe on $DIR/$tdir/f0"
7277 test_103_check_pool $save_fsname 17
7281 test_renamefs $save_fsname
7287 run_test 103 "rename filesystem name"
7289 if ! combined_mgs_mds ; then
7295 # restore the values of MDSSIZE and OSTSIZE
7296 MDSSIZE=$STORED_MDSSIZE
7297 OSTSIZE=$STORED_OSTSIZE