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 $node"
1702 local nid=$($r $LCTL list_nids | head -1)
1708 local fstype=$(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 mkdir -p $tmp/mnt/lustre || error "mkdir $tmp/mnt/lustre failed"
1721 $r mkdir -p $tmp/mnt/{mdt,mdt1,ost}
1722 $r tar xjvf $tarball -S -C $tmp || {
1723 error_noexit "Unpacking the disk image tarball"
1726 img_commit=$($r cat $tmp/commit)
1727 img_kernel=$($r cat $tmp/kernel)
1728 img_arch=$($r cat $tmp/arch)
1729 img_bspace=$($r cat $tmp/bspace)
1730 img_ispace=$($r cat $tmp/ispace)
1732 # older images did not have "blimit" and "ilimit" files
1733 # use old values for T32_BLIMIT and T32_ILIMIT
1734 $r test -f $tmp/blimit && img_blimit=$($r cat $tmp/blimit) ||
1736 $r test -f $tmp/ilimit && img_ilimit=$($r cat $tmp/ilimit) ||
1739 echo "Upgrading from $(basename $tarball), created with:"
1740 echo " Commit: $img_commit"
1741 echo " Kernel: $img_kernel"
1742 echo " Arch: $img_arch"
1743 echo "OST version: $(lustre_build_version ost1)"
1745 # The conversion can be made only when both of the following
1746 # conditions are satisfied:
1747 # - ost device img version < 2.3.64
1748 # - ost server version >= 2.5
1749 [ $(version_code $img_commit) -ge $(version_code 2.3.64) -o \
1750 $(lustre_version_code ost1) -lt $(version_code 2.5.0) ] &&
1753 ! $r test -f $mdt2_dev || mdt2_is_available=true
1755 if [[ $fstype == zfs ]]; then
1758 local poolname_list="t32fs-mdt1 t32fs-ost1"
1760 ! $mdt2_is_available || poolname_list+=" t32fs-mdt2"
1762 for poolname in $poolname_list; do
1764 $ZPOOL list -H $poolname >/dev/null 2>&1 ||
1765 $ZPOOL import -f -d $tmp $poolname"
1768 mdt_dev=t32fs-mdt1/mdt1
1769 ost_dev=t32fs-ost1/ost1
1770 ! $mdt2_is_available || mdt2_dev=t32fs-mdt2/mdt2
1771 wait_update_facet $SINGLEMDS "$ZPOOL list |
1772 awk '/^t32fs-mdt1/ { print \\\$1 }'" "t32fs-mdt1" || {
1773 error_noexit "import zfs pool failed"
1778 $r $LCTL set_param debug="$PTLDEBUG"
1780 $r $TUNEFS --dryrun $mdt_dev || {
1782 error_noexit "tunefs.lustre before mounting the MDT"
1786 if $mdt2_is_available; then
1787 $r $TUNEFS --dryrun $mdt2_dev || {
1789 error_noexit "tunefs.lustre before mounting the MDT"
1794 if [ "$writeconf" ]; then
1796 if [ $fstype == "ldiskfs" ]; then
1798 $r $TUNEFS --quota $mdt_dev || {
1800 error_noexit "Enable mdt quota feature"
1803 if $mdt2_is_available; then
1804 $r $TUNEFS --quota $mdt2_dev || {
1806 error_noexit "Enable mdt quota feature"
1812 if [ -n "$($LCTL list_nids | grep -v '\(tcp\|lo\)[[:digit:]]*$')" ]; then
1813 [[ $(lustre_version_code mgs) -ge $(version_code 2.3.59) ]] ||
1814 { skip "LU-2200: Cannot run over Inifiniband w/o lctl replace_nids "
1815 "(Need MGS version at least 2.3.59)"; return 0; }
1817 local osthost=$(facet_active_host ost1)
1818 local ostnid=$(do_node $osthost $LCTL list_nids | head -1)
1821 if [ $fstype == "ldiskfs" ]; then
1824 $r $MOUNT_CMD -o $mopts $mdt_dev $tmp/mnt/mdt
1825 $r $LCTL replace_nids $fsname-OST0000 $ostnid
1826 $r $LCTL replace_nids $fsname-MDT0000 $nid
1827 $r $UMOUNT $tmp/mnt/mdt
1830 mopts=exclude=$fsname-OST0000
1831 if [ $fstype == "ldiskfs" ]; then
1836 t32_wait_til_devices_gone $node
1838 $r $MOUNT_CMD -o $mopts $mdt_dev $tmp/mnt/mdt || {
1840 error_noexit "Mounting the MDT"
1843 shall_cleanup_mdt=true
1845 if $mdt2_is_available; then
1846 mopts=mgsnode=$nid,$mopts
1847 $r $MOUNT_CMD -o $mopts $mdt2_dev $tmp/mnt/mdt1 || {
1849 error_noexit "Mounting the MDT"
1853 echo "mount new MDT....$mdt2_dev"
1854 $r $LCTL set_param -n mdt.${fsname}*.enable_remote_dir=1 ||
1855 error_noexit "enable remote dir create failed"
1857 shall_cleanup_mdt1=true
1858 elif [ "$dne_upgrade" != "no" ]; then
1859 local fs2mdsdev=$(mdsdevname 1_2)
1860 local fs2mdsvdev=$(mdsvdevname 1_2)
1862 echo "mkfs new MDT on ${fs2mdsdev}...."
1863 if [ $(facet_fstype mds1) == ldiskfs ]; then
1864 mkfsoptions="--mkfsoptions=\\\"-J size=8\\\""
1867 add $SINGLEMDS $(mkfs_opts mds2 $fs2mdsdev $fsname) --reformat \
1868 $mkfsoptions $fs2mdsdev $fs2mdsvdev > /dev/null || {
1869 error_noexit "Mkfs new MDT failed"
1873 [[ $(facet_fstype mds1) != zfs ]] || import_zpool mds1
1875 $r $TUNEFS --dryrun $fs2mdsdev || {
1876 error_noexit "tunefs.lustre before mounting the MDT"
1880 echo "mount new MDT....$fs2mdsdev"
1881 $r $MOUNT_CMD -o $mopts $fs2mdsdev $tmp/mnt/mdt1 || {
1882 error_noexit "mount mdt1 failed"
1886 $r $LCTL set_param -n mdt.${fsname}*.enable_remote_dir=1 ||
1887 error_noexit "enable remote dir create failed"
1889 shall_cleanup_mdt1=true
1892 uuid=$($r $LCTL get_param -n mdt.$fsname-MDT0000.uuid) || {
1893 error_noexit "Getting MDT UUID"
1896 if [ "$uuid" != $fsname-MDT0000_UUID ]; then
1897 error_noexit "Unexpected MDT UUID: \"$uuid\""
1901 $r $TUNEFS --dryrun $ost_dev || {
1902 error_noexit "tunefs.lustre before mounting the OST"
1905 if [ "$writeconf" ]; then
1906 mopts=mgsnode=$nid,$writeconf
1907 if [ $fstype == "ldiskfs" ]; then
1909 $r $TUNEFS --quota $ost_dev || {
1911 error_noexit "Enable ost quota feature"
1917 if [ $fstype == "ldiskfs" ]; then
1921 $r $MOUNT_CMD -o $mopts $ost_dev $tmp/mnt/ost || {
1922 error_noexit "Mounting the OST"
1925 shall_cleanup_ost=true
1927 uuid=$($r $LCTL get_param -n obdfilter.$fsname-OST0000.uuid) || {
1928 error_noexit "Getting OST UUID"
1931 if [ "$uuid" != $fsname-OST0000_UUID ]; then
1932 error_noexit "Unexpected OST UUID: \"$uuid\""
1936 $r $LCTL conf_param $fsname-OST0000.osc.max_dirty_mb=15 || {
1937 error_noexit "Setting \"max_dirty_mb\""
1940 $r $LCTL conf_param $fsname-OST0000.failover.node=$nid || {
1941 error_noexit "Setting OST \"failover.node\""
1944 $r $LCTL conf_param $fsname-MDT0000.mdc.max_rpcs_in_flight=9 || {
1945 error_noexit "Setting \"max_rpcs_in_flight\""
1948 $r $LCTL conf_param $fsname-MDT0000.failover.node=$nid || {
1949 error_noexit "Setting MDT \"failover.node\""
1952 $r $LCTL pool_new $fsname.interop || {
1953 error_noexit "Setting \"interop\""
1956 $r $LCTL conf_param $fsname-MDT0000.lov.stripesize=4M || {
1957 error_noexit "Setting \"lov.stripesize\""
1960 $r $LCTL conf_param $fsname-MDT0000.mdd.atime_diff=70 || {
1961 error_noexit "Setting \"mdd.atime_diff\""
1965 if [ "$ff_convert" != "no" -a $(facet_fstype ost1) == "ldiskfs" ]; then
1966 $r $LCTL lfsck_start -M $fsname-OST0000 || {
1967 error_noexit "Start OI scrub on OST0"
1971 # The oi_scrub should be on ost1, but for test_32(),
1972 # all on the SINGLEMDS.
1973 wait_update_facet $SINGLEMDS "$LCTL get_param -n \
1974 osd-ldiskfs.$fsname-OST0000.oi_scrub |
1975 awk '/^status/ { print \\\$2 }'" "completed" 30 || {
1976 error_noexit "Failed to get the expected 'completed'"
1980 local UPDATED=$($r $LCTL get_param -n \
1981 osd-ldiskfs.$fsname-OST0000.oi_scrub |
1982 awk '/^updated/ { print $2 }')
1983 [ $UPDATED -ge 1 ] || {
1984 error_noexit "Only $UPDATED objects have been converted"
1989 if [ "$dne_upgrade" != "no" ]; then
1990 $r $LCTL conf_param \
1991 $fsname-MDT0001.mdc.max_rpcs_in_flight=9 || {
1992 error_noexit "Setting MDT1 \"max_rpcs_in_flight\""
1995 $r $LCTL conf_param $fsname-MDT0001.failover.node=$nid || {
1996 error_noexit "Setting MDT1 \"failover.node\""
1999 $r $LCTL conf_param $fsname-MDT0001.lov.stripesize=4M || {
2000 error_noexit "Setting MDT1 \"lov.stripesize\""
2005 if [ "$writeconf" ]; then
2006 $MOUNT_CMD $nid:/$fsname $tmp/mnt/lustre || {
2007 error_noexit "Mounting the client"
2011 shall_cleanup_lustre=true
2012 $r $LCTL set_param debug="$PTLDEBUG"
2014 t32_verify_quota $node $fsname $tmp/mnt/lustre || {
2015 error_noexit "verify quota failed"
2019 if $r test -f $tmp/list; then
2021 # There is not a Test Framework API to copy files to or
2022 # from a remote node.
2024 # LU-2393 - do both sorts on same node to ensure locale
2026 local list_file=$tmp/list
2028 if $mdt2_is_available; then
2029 if [[ -d $tmp/mnt/lustre/striped_dir_old ]] &&
2030 $r test -f $tmp/list2; then
2031 list_file=$tmp/list2
2032 pushd $tmp/mnt/lustre/striped_dir_old
2034 pushd $tmp/mnt/lustre/remote_dir
2037 pushd $tmp/mnt/lustre
2039 $r cat $list_file | sort -k 6 >$tmp/list.orig
2040 ls -Rni --time-style=+%s | sort -k 6 |
2041 sed 's/\. / /' >$tmp/list || {
2047 # 32-bit and 64-bit clients use different algorithms to
2048 # convert FIDs into inode numbers. Hence, remove the
2049 # inode numbers from the lists, if the original list was
2050 # created on an architecture with different number of
2053 if [ $(t32_bits_per_long $(uname -m)) != \
2054 $(t32_bits_per_long $img_arch) ]; then
2055 echo "Different number of bits per \"long\"" \
2056 "from the disk image"
2057 for list in list.orig list; do
2058 sed -i -e 's/^[0-9]\+[ \t]\+//' \
2062 if ! diff -ub $tmp/list.orig $tmp/list; then
2063 error_noexit "list verification failed"
2067 echo "list verification skipped"
2070 if [ "$dne_upgrade" != "no" ]; then
2071 $LFS mkdir -i 1 -c2 $tmp/mnt/lustre/striped_dir || {
2072 error_noexit "set striped dir failed"
2076 $LFS setdirstripe -D -c2 $tmp/mnt/lustre/striped_dir
2078 pushd $tmp/mnt/lustre
2079 tar -cf - . --exclude=./striped_dir \
2080 --exclude=./striped_dir_old \
2081 --exclude=./remote_dir |
2082 tar -xvf - -C striped_dir 1>/dev/null || {
2083 error_noexit "cp to striped dir failed"
2089 # If it is upgrade from DNE (2.5), then rename the remote dir,
2090 # which is created in 2.5 to striped dir.
2091 if $mdt2_is_available && [[ "$dne_upgrade" != "no" ]]; then
2092 stripe_index=$($LFS getdirstripe -i \
2093 $tmp/mnt/lustre/remote_dir)
2095 [[ $stripe_index -eq 1 ]] || {
2096 error_noexit "get index \"$stripe_index\"" \
2097 "from remote dir failed"
2100 mv $tmp/mnt/lustre/remote_dir \
2101 $tmp/mnt/lustre/striped_dir/ || {
2102 error_noexit "mv remote dir failed"
2107 # If it is upgraded from DNE (2.7), then move the striped dir
2108 # which was created in 2.7 to the new striped dir.
2109 if $mdt2_is_available && [[ "$dne_upgrade" != "no" ]] &&
2110 [[ -d $tmp/mnt/lustre/striped_dir_old ]]; then
2111 stripe_count=$($LFS getdirstripe -c \
2112 $tmp/mnt/lustre/striped_dir_old)
2113 [[ $stripe_count -eq 2 ]] || {
2114 error_noexit "get count $stripe_count" \
2115 "from striped dir failed"
2118 mv $tmp/mnt/lustre/striped_dir_old \
2119 $tmp/mnt/lustre/striped_dir/ || {
2120 error_noexit "mv striped dir failed"
2126 $r $LCTL set_param -n osd*.*.force_sync=1
2127 dd if=/dev/zero of=$tmp/mnt/lustre/tmp_file bs=10k count=10 || {
2128 error_noexit "dd failed"
2131 rm -rf $tmp/mnt/lustre/tmp_file || {
2132 error_noexit "rm failed"
2136 if $r test -f $tmp/sha1sums; then
2137 # LU-2393 - do both sorts on same node to ensure locale
2139 $r cat $tmp/sha1sums | sort -k 2 >$tmp/sha1sums.orig
2140 if [ "$dne_upgrade" != "no" ]; then
2141 pushd $tmp/mnt/lustre/striped_dir
2143 pushd $tmp/mnt/lustre
2146 find ! -path "*remote_dir*" ! -path "*striped_dir*" \
2147 ! -name .lustre -type f -exec sha1sum {} \; |
2148 sort -k 2 >$tmp/sha1sums || {
2150 error_noexit "sha1sum"
2154 if ! diff -ub $tmp/sha1sums.orig $tmp/sha1sums; then
2155 error_noexit "sha1sum verification failed"
2159 # if upgrade from DNE(2.5), then check remote directory
2160 # if upgrade from DNE(2.7), then check striped directory
2161 if $mdt2_is_available &&
2162 [[ "$dne_upgrade" != "no" ]]; then
2163 local new_dir="$tmp/mnt/lustre/striped_dir"
2164 local striped_dir_old="$new_dir/striped_dir_old"
2166 local dir_list="$new_dir/remote_dir"
2167 [[ ! -d $triped_dir_old ]] ||
2168 dir_list+=" $striped_dir_old"
2170 for dir in $dir_list; do
2172 find ! -name .lustre -type f \
2173 -exec sha1sum {} \; |
2174 sort -k 2 >$tmp/sha1sums || {
2176 error_noexit "sha1sum"
2180 if ! diff -ub $tmp/sha1sums.orig \
2182 error_noexit "sha1sum $dir" \
2189 echo "sha1sum verification skipped"
2192 if [ "$dne_upgrade" != "no" ]; then
2193 rm -rf $tmp/mnt/lustre/striped_dir || {
2194 error_noexit "remove remote dir failed"
2199 # migrate files/dirs to remote MDT, then move them back
2200 if [ $(lustre_version_code mds1) -ge $(version_code 2.7.50) -a \
2201 $dne_upgrade != "no" ]; then
2202 $r $LCTL set_param -n \
2203 mdt.${fsname}*.enable_remote_dir=1 2>/dev/null
2205 echo "test migration"
2206 pushd $tmp/mnt/lustre
2207 for dir in $(find ! -name .lustre ! -name . -type d); do
2208 mdt_index=$($LFS getdirstripe -i $dir)
2209 stripe_cnt=$($LFS getdirstripe -c $dir)
2210 if [ $mdt_index = 0 -a $stripe_cnt -le 1 ]; then
2211 $LFS mv -M 1 $dir || {
2213 error_noexit "migrate MDT1 failed"
2219 for dir in $(find ! -name . ! -name .lustre -type d); do
2220 mdt_index=$($LFS getdirstripe -i $dir)
2221 stripe_cnt=$($LFS getdirstripe -c $dir)
2222 if [ $mdt_index = 1 -a $stripe_cnt -le 1 ]; then
2223 $LFS mv -M 0 $dir || {
2225 error_noexit "migrate MDT0 failed"
2234 # When adding new data verification tests, please check for
2235 # the presence of the required reference files first, like
2236 # the "sha1sums" and "list" tests above, to avoid the need to
2237 # regenerate every image for each test addition.
2240 nrpcs_orig=$($LCTL get_param \
2241 -n mdc.*MDT0000*.max_rpcs_in_flight) || {
2242 error_noexit "Getting \"max_rpcs_in_flight\""
2245 nrpcs=$((nrpcs_orig + 5))
2246 $r $LCTL conf_param $fsname-MDT0000.mdc.max_rpcs_in_flight=$nrpcs || {
2247 error_noexit "Changing \"max_rpcs_in_flight\""
2250 wait_update $HOSTNAME "$LCTL get_param \
2251 -n mdc.*MDT0000*.max_rpcs_in_flight" $nrpcs || {
2252 error_noexit "Verifying \"max_rpcs_in_flight\""
2256 umount $tmp/mnt/lustre || {
2257 error_noexit "Unmounting the client"
2260 shall_cleanup_lustre=false
2262 if [[ "$dne_upgrade" != "no" ]] || $mdt2_is_available; then
2263 $r $UMOUNT $tmp/mnt/mdt1 || {
2264 error_noexit "Unmounting the MDT2"
2267 shall_cleanup_mdt1=false
2270 $r $UMOUNT $tmp/mnt/mdt || {
2271 error_noexit "Unmounting the MDT"
2274 shall_cleanup_mdt=false
2276 $r $UMOUNT $tmp/mnt/ost || {
2277 error_noexit "Unmounting the OST"
2280 shall_cleanup_ost=false
2282 t32_reload_modules $node || {
2283 error_noexit "Reloading modules"
2287 if [[ $fstype == zfs ]]; then
2288 local poolname=t32fs-mdt1
2290 $ZPOOL list -H $poolname >/dev/null 2>&1 ||
2291 $ZPOOL import -f -d $tmp $poolname"
2294 # mount a second time to make sure we didnt leave upgrade flag on
2295 $r $TUNEFS --dryrun $mdt_dev || {
2297 error_noexit "tunefs.lustre before remounting the MDT"
2301 mopts=exclude=$fsname-OST0000
2302 if [ $fstype == "ldiskfs" ]; then
2305 $r $MOUNT_CMD -o $mopts $mdt_dev $tmp/mnt/mdt || {
2306 error_noexit "Remounting the MDT"
2309 shall_cleanup_mdt=true
2319 for tarball in $tarballs; do
2320 t32_test $tarball || let "rc += $?"
2324 run_test 32a "Upgrade (not live)"
2332 for tarball in $tarballs; do
2333 t32_test $tarball writeconf || let "rc += $?"
2337 run_test 32b "Upgrade with writeconf"
2344 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
2346 for tarball in $tarballs; do
2347 # Do not support 1_8 and 2_1 direct upgrade to DNE2 anymore */
2348 echo $tarball | grep "1_8" && continue
2349 echo $tarball | grep "2_1" && continue
2351 dne_upgrade=yes t32_test $tarball writeconf || rc=$?
2355 run_test 32c "dne upgrade test"
2363 for tarball in $tarballs; do
2364 ff_convert=yes t32_test $tarball || rc=$?
2368 run_test 32d "convert ff test"
2370 test_33a() { # bug 12333, was test_33
2371 local FSNAME2=test-123
2372 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
2375 [ -n "$ost1_HOST" ] && fs2ost_HOST=$ost1_HOST
2377 if [ -z "$fs2ost_DEV" -o -z "$fs2mds_DEV" ]; then
2378 local dev=${SINGLEMDS}_dev
2379 local MDSDEV=${!dev}
2380 is_blkdev $SINGLEMDS $MDSDEV &&
2381 skip_env "mixed loopback and real device not working" &&
2385 local fs2mdsdev=$(mdsdevname 1_2)
2386 local fs2ostdev=$(ostdevname 1_2)
2387 local fs2mdsvdev=$(mdsvdevname 1_2)
2388 local fs2ostvdev=$(ostvdevname 1_2)
2390 if [ $(facet_fstype mds1) == ldiskfs ]; then
2391 mkfsoptions="--mkfsoptions=\\\"-J size=8\\\"" # See bug 17931.
2394 add fs2mds $(mkfs_opts mds1 ${fs2mdsdev}) --mgs --fsname=${FSNAME2} \
2395 --reformat $mkfsoptions $fs2mdsdev $fs2mdsvdev || exit 10
2396 add fs2ost $(mkfs_opts ost1 ${fs2ostdev}) --mgsnode=$MGSNID \
2397 --fsname=${FSNAME2} --index=8191 --reformat $fs2ostdev \
2398 $fs2ostvdev || exit 10
2400 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS && trap cleanup_fs2 EXIT INT
2401 start fs2ost $fs2ostdev $OST_MOUNT_OPTS
2402 do_facet $SINGLEMDS "$LCTL conf_param $FSNAME2.sys.timeout=200" ||
2403 error "$LCTL conf_param $FSNAME2.sys.timeout=200 failed"
2404 mkdir -p $MOUNT2 || error "mkdir $MOUNT2 failed"
2405 $MOUNT_CMD $MGSNID:/${FSNAME2} $MOUNT2 || error "$MOUNT_CMD failed"
2408 cp /etc/hosts $MOUNT2/ || error "copy /etc/hosts $MOUNT2/ failed"
2409 $GETSTRIPE $MOUNT2/hosts || error "$GETSTRIPE $MOUNT2/hosts failed"
2414 cleanup_nocli || error "cleanup_nocli failed with $?"
2416 run_test 33a "Mount ost with a large index number"
2418 test_33b() { # was test_34
2421 do_facet client dd if=/dev/zero of=$MOUNT/24 bs=1024k count=1
2422 # Drop lock cancelation reply during umount
2423 #define OBD_FAIL_LDLM_CANCEL_NET 0x304
2424 do_facet client $LCTL set_param fail_loc=0x80000304
2425 #lctl set_param debug=-1
2426 umount_client $MOUNT
2427 cleanup || error "cleanup failed with $?"
2429 run_test 33b "Drop cancel during umount"
2433 do_facet client "sh runmultiop_bg_pause $DIR/file O_c"
2434 manual_umount_client
2436 do_facet client killall -USR1 multiop
2437 if [ $rc -eq 0 ]; then
2438 error "umount not fail!"
2441 cleanup || error "cleanup failed with rc $?"
2443 run_test 34a "umount with opened file should be fail"
2447 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
2448 stop_mds || error "Unable to stop MDS"
2450 manual_umount_client --force || error "mtab after failed umount with $?"
2452 cleanup || error "cleanup failed with $?"
2454 run_test 34b "force umount with failed mds should be normal"
2458 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
2459 stop_ost || error "Unable to stop OST1"
2461 manual_umount_client --force || error "mtab after failed umount with $?"
2463 cleanup || error "cleanup failed with $?"
2465 run_test 34c "force umount with failed ost should be normal"
2467 test_35a() { # bug 12459
2470 DBG_SAVE="`$LCTL get_param -n debug`"
2471 $LCTL set_param debug="ha"
2473 log "Set up a fake failnode for the MDS"
2475 local device=$(do_facet $SINGLEMDS "$LCTL get_param -n devices" |
2476 awk '($3 ~ "mdt" && $4 ~ "MDT") { print $4 }' | head -1)
2477 do_facet mgs "$LCTL conf_param \
2478 ${device}.failover.node=$(h2$NETTYPE $FAKENID)" ||
2479 error "Setting ${device}.failover.node=\
2480 $(h2$NETTYPE $FAKENID) failed."
2482 log "Wait for RECONNECT_INTERVAL seconds (10s)"
2485 MSG="conf-sanity.sh test_35a `date +%F%kh%Mm%Ss`"
2488 log "Stopping the MDT: $device"
2489 stop_mdt 1 || error "MDT0 stop fail"
2491 df $MOUNT > /dev/null 2>&1 &
2493 log "Restarting the MDT: $device"
2494 start_mdt 1 || error "MDT0 start fail"
2495 log "Wait for df ($DFPID) ... "
2498 $LCTL set_param debug="$DBG_SAVE"
2500 # retrieve from the log the first server that the client tried to
2501 # contact after the connection loss
2502 $LCTL dk $TMP/lustre-log-$TESTNAME.log
2503 NEXTCONN=`awk "/${MSG}/ {start = 1;}
2504 /import_select_connection.*$device-mdc.* using connection/ {
2506 if (\\\$NF ~ /$FAKENID/)
2512 }" $TMP/lustre-log-$TESTNAME.log`
2513 [ "$NEXTCONN" != "0" ] &&
2514 error "Tried to connect to ${NEXTCONN} not last active server"
2515 cleanup || error "cleanup failed with $?"
2516 # remove nid settings
2517 writeconf_or_reformat
2519 run_test 35a "Reconnect to the last active server first"
2521 test_35b() { # bug 18674
2522 remote_mds || { skip "local MDS" && return 0; }
2526 $LCTL set_param debug="ha"
2528 MSG="conf-sanity.sh test_35b `date +%F%kh%Mm%Ss`"
2531 log "Set up a fake failnode for the MDS"
2533 local device=$(do_facet $SINGLEMDS "$LCTL get_param -n devices" |
2534 awk '($3 ~ "mdt" && $4 ~ "MDT") { print $4 }' | head -1)
2535 do_facet mgs "$LCTL conf_param \
2536 ${device}.failover.node=$(h2$NETTYPE $FAKENID)" ||
2537 error "Set ${device}.failover.node=\
2538 $(h2$NETTYPE $FAKENID) failed"
2540 local at_max_saved=0
2541 # adaptive timeouts may prevent seeing the issue
2542 if at_is_enabled; then
2543 at_max_saved=$(at_max_get mds)
2544 at_max_set 0 mds client
2547 mkdir $MOUNT/$tdir || error "mkdir $MOUNT/$tdir failed"
2549 log "Injecting EBUSY on MDS"
2550 # Setting OBD_FAIL_MDS_RESEND=0x136
2551 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0x80000136" ||
2552 error "unable to set param fail_loc=0x80000136"
2554 $LCTL set_param mdc.${FSNAME}*.stats=clear
2556 log "Creating a test file and stat it"
2557 touch $MOUNT/$tdir/$tfile || error "touch $MOUNT/$tdir/$tfile failed"
2558 stat $MOUNT/$tdir/$tfile
2560 log "Stop injecting EBUSY on MDS"
2561 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0" ||
2562 error "unable to set param fail_loc=0"
2563 rm -f $MOUNT/$tdir/$tfile || error "remove $MOUNT/$tdir/$tfile failed"
2566 # restore adaptive timeout
2567 [ $at_max_saved -ne 0 ] && at_max_set $at_max_saved mds client
2569 $LCTL dk $TMP/lustre-log-$TESTNAME.log
2571 CONNCNT=$($LCTL get_param mdc.${FSNAME}*.stats |
2572 awk '/mds_connect/{print $2}')
2574 # retrieve from the log if the client has ever tried to
2575 # contact the fake server after the loss of connection
2576 FAILCONN=`awk "BEGIN {ret = 0;}
2577 /import_select_connection.*${FSNAME}-MDT0000-mdc.* using connection/ {
2579 if (\\\$NF ~ /$FAKENID/) {
2584 END {print ret}" $TMP/lustre-log-$TESTNAME.log`
2586 [ "$FAILCONN" == "0" ] &&
2587 error "The client reconnection has not been triggered"
2588 [ "$FAILCONN" == "2" ] &&
2589 error "Primary server busy, client reconnect to failover failed"
2592 # When OBD_FAIL_MDS_RESEND is hit, we sleep for 2 * obd_timeout
2593 # Reconnects are supposed to be rate limited to one every 5s
2594 [ $CONNCNT -gt $((2 * $TIMEOUT / 5 + 1)) ] &&
2595 error "Too many reconnects $CONNCNT"
2597 cleanup || error "cleanup failed with $?"
2598 # remove nid settings
2599 writeconf_or_reformat
2601 run_test 35b "Continue reconnection retries, if the active server is busy"
2604 [ $OSTCOUNT -lt 2 ] && skip_env "skipping test for single OST" && return
2606 [ "$ost_HOST" = "`hostname`" -o "$ost1_HOST" = "`hostname`" ] ||
2607 { skip "remote OST" && return 0; }
2610 local FSNAME2=test1234
2611 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
2613 [ -n "$ost1_HOST" ] && fs2ost_HOST=$ost1_HOST && fs3ost_HOST=$ost1_HOST
2615 if [ -z "$fs2ost_DEV" -o -z "$fs2mds_DEV" -o -z "$fs3ost_DEV" ]; then
2616 is_blkdev $SINGLEMDS $MDSDEV &&
2617 skip_env "mixed loopback and real device not working" && return
2620 local fs2mdsdev=$(mdsdevname 1_2)
2621 local fs2ostdev=$(ostdevname 1_2)
2622 local fs3ostdev=$(ostdevname 2_2)
2623 local fs2mdsvdev=$(mdsvdevname 1_2)
2624 local fs2ostvdev=$(ostvdevname 1_2)
2625 local fs3ostvdev=$(ostvdevname 2_2)
2627 add fs2mds $(mkfs_opts mds1 ${fs2mdsdev}) --mgs --fsname=${FSNAME2} \
2628 --reformat $fs2mdsdev $fs2mdsvdev || exit 10
2629 # XXX after we support non 4K disk blocksize in ldiskfs, specify a
2630 # different one than the default value here.
2631 add fs2ost $(mkfs_opts ost1 ${fs2ostdev}) --mgsnode=$MGSNID \
2632 --fsname=${FSNAME2} --reformat $fs2ostdev $fs2ostvdev || exit 10
2633 add fs3ost $(mkfs_opts ost2 ${fs3ostdev}) --mgsnode=$MGSNID \
2634 --fsname=${FSNAME2} --reformat $fs3ostdev $fs3ostvdev || exit 10
2636 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS
2637 start fs2ost $fs2ostdev $OST_MOUNT_OPTS
2638 start fs3ost $fs3ostdev $OST_MOUNT_OPTS
2639 mkdir -p $MOUNT2 || error "mkdir $MOUNT2 failed"
2640 $MOUNT_CMD $MGSNID:/${FSNAME2} $MOUNT2 || error "$MOUNT_CMD failed"
2642 sleep 5 # until 11778 fixed
2644 dd if=/dev/zero of=$MOUNT2/$tfile bs=1M count=7 || error "dd failed"
2646 BKTOTAL=$($LCTL get_param -n obdfilter.*.kbytestotal |
2647 awk 'BEGIN{total=0}; {total+=$1}; END{print total}')
2648 BKFREE=$($LCTL get_param -n obdfilter.*.kbytesfree |
2649 awk 'BEGIN{free=0}; {free+=$1}; END{print free}')
2650 BKAVAIL=$($LCTL get_param -n obdfilter.*.kbytesavail |
2651 awk 'BEGIN{avail=0}; {avail+=$1}; END{print avail}')
2652 STRING=$(df -P $MOUNT2 | tail -n 1 | awk '{print $2","$3","$4}')
2653 DFTOTAL=$(echo $STRING | cut -d, -f1)
2654 DFUSED=$(echo $STRING | cut -d, -f2)
2655 DFAVAIL=$(echo $STRING | cut -d, -f3)
2656 DFFREE=$(($DFTOTAL - $DFUSED))
2658 ALLOWANCE=$((64 * $OSTCOUNT))
2660 if [ $DFTOTAL -lt $(($BKTOTAL - $ALLOWANCE)) ] ||
2661 [ $DFTOTAL -gt $(($BKTOTAL + $ALLOWANCE)) ] ; then
2662 echo "**** FAIL: df total($DFTOTAL) mismatch OST total($BKTOTAL)"
2665 if [ $DFFREE -lt $(($BKFREE - $ALLOWANCE)) ] ||
2666 [ $DFFREE -gt $(($BKFREE + $ALLOWANCE)) ] ; then
2667 echo "**** FAIL: df free($DFFREE) mismatch OST free($BKFREE)"
2670 if [ $DFAVAIL -lt $(($BKAVAIL - $ALLOWANCE)) ] ||
2671 [ $DFAVAIL -gt $(($BKAVAIL + $ALLOWANCE)) ] ; then
2672 echo "**** FAIL: df avail($DFAVAIL) mismatch OST avail($BKAVAIL)"
2677 stop fs3ost -f || error "unable to stop OST3"
2678 stop fs2ost -f || error "unable to stop OST2"
2679 stop fs2mds -f || error "unable to stop second MDS"
2680 unload_modules_conf || error "unable unload modules"
2683 run_test 36 "df report consistency on OSTs with different block size"
2686 local mntpt=$(facet_mntpt $SINGLEMDS)
2687 local mdsdev=$(mdsdevname ${SINGLEMDS//mds/})
2688 local mdsdev_sym="$TMP/sym_mdt.img"
2689 local opts=$MDS_MOUNT_OPTS
2692 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
2693 skip "Currently only applicable to ldiskfs-based MDTs"
2697 echo "MDS : $mdsdev"
2698 echo "SYMLINK : $mdsdev_sym"
2699 do_facet $SINGLEMDS rm -f $mdsdev_sym
2701 do_facet $SINGLEMDS ln -s $mdsdev $mdsdev_sym
2703 echo "mount symlink device - $mdsdev_sym"
2705 if ! do_facet $SINGLEMDS test -b $mdsdev; then
2706 opts=$(csa_add "$opts" -o loop)
2708 mount_op=$(do_facet $SINGLEMDS mount -v -t lustre $opts \
2709 $mdsdev_sym $mntpt 2>&1)
2712 echo mount_op=$mount_op
2714 do_facet $SINGLEMDS "$UMOUNT $mntpt && rm -f $mdsdev_sym"
2716 if $(echo $mount_op | grep -q "unable to set tunable"); then
2717 error "set tunables failed for symlink device"
2720 [ $rc -eq 0 ] || error "mount symlink $mdsdev_sym failed! rc=$rc"
2722 run_test 37 "verify set tunables works for symlink device"
2724 test_38() { # bug 14222
2725 local fstype=$(facet_fstype $SINGLEMDS)
2726 local mntpt=$(facet_mntpt $SINGLEMDS)
2731 local SRC="/etc /bin"
2732 local FILES=$(find $SRC -type f -mtime +1 | head -n $COUNT)
2733 log "copying $(echo $FILES | wc -w) files to $DIR/$tdir"
2734 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
2735 tar cf - $FILES | tar xf - -C $DIR/$tdir ||
2736 error "copying $SRC to $DIR/$tdir"
2738 umount_client $MOUNT || error "umount_client $MOUNT failed"
2739 do_facet $SINGLEMDS "$LCTL get_param osp.*.prealloc_next_id"
2740 stop_mds || error "Unable to stop MDS"
2741 log "delete lov_objid file on MDS"
2743 mount_fstype $SINGLEMDS || error "mount MDS failed (1)"
2745 do_facet $SINGLEMDS "od -Ax -td8 $mntpt/lov_objid; rm $mntpt/lov_objid"
2747 unmount_fstype $SINGLEMDS || error "umount failed (1)"
2749 # check create in mds_lov_connect
2750 start_mds || error "unable to start MDS"
2751 mount_client $MOUNT || error "mount_client $MOUNT failed"
2753 [ $V ] && log "verifying $DIR/$tdir/$f"
2754 diff -q $f $DIR/$tdir/$f || ERROR=y
2756 do_facet $SINGLEMDS "$LCTL get_param osp.*.prealloc_next_id"
2757 if [ "$ERROR" = "y" ]; then
2758 # check it's updates in sync
2759 umount_client $MOUNT
2761 mount_fstype $SIGNLEMDS
2762 do_facet $SINGLEMDS "od -Ax -td8 $mntpt/lov_objid"
2763 unmount_fstype $SINGLEMDS
2764 error "old and new files are different after connect" || true
2766 touch $DIR/$tdir/f2 || error "f2 file create failed"
2768 # check it's updates in sync
2769 umount_client $MOUNT || error "second umount_client $MOUNT failed"
2772 mount_fstype $SINGLEMDS || error "mount MDS failed (3)"
2774 do_facet $SINGLEMDS "od -Ax -td8 $mntpt/lov_objid"
2775 do_facet $SINGLEMDS dd if=/dev/zero of=$mntpt/lov_objid.clear count=8
2777 unmount_fstype $SINGLEMDS || error "umount failed (3)"
2779 start_mds || error "unable to start MDS"
2780 mount_client $MOUNT || error "mount_client $MOUNT failed"
2782 [ $V ] && log "verifying $DIR/$tdir/$f"
2783 diff -q $f $DIR/$tdir/$f || ERROR=y
2785 touch $DIR/$tdir/f3 || error "f3 file create failed"
2786 do_facet $SINGLEMDS "$LCTL get_param osp.*.prealloc_next_id"
2787 umount_client $MOUNT || error "third umount_client $MOUNT failed"
2789 mount_fstype $SINGLEMDS || error "mount MDS failed (4)"
2790 do_facet $SINGLEMDS "od -Ax -td8 $mntpt/lov_objid"
2791 unmount_fstype $SINGLEMDS || error "umount failed (4)"
2793 [ "$ERROR" = "y" ] &&
2794 error "old and new files are different after sync" || true
2796 log "files compared the same"
2797 cleanup || error "cleanup failed with $?"
2799 run_test 38 "MDS recreates missing lov_objid file from OST data"
2804 cleanup || error "cleanup failed with $?"
2805 perl $SRCDIR/leak_finder.pl $TMP/debug 2>&1 | egrep '*** Leak:' &&
2806 error "memory leak detected" || true
2808 run_test 39 "leak_finder recognizes both LUSTRE and LNET malloc messages"
2810 test_40() { # bug 15759
2811 start_ost || error "Unable to start OST1"
2812 #define OBD_FAIL_TGT_TOOMANY_THREADS 0x706
2813 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0x80000706"
2815 cleanup || error "cleanup failed with rc $?"
2817 run_test 40 "race during service thread startup"
2819 test_41a() { #bug 14134
2820 if [ $(facet_fstype $SINGLEMDS) == ldiskfs ] &&
2821 ! do_facet $SINGLEMDS test -b $(mdsdevname 1); then
2822 skip "Loop devices does not work with nosvc option"
2827 { skip "needs combined MGT and MDT device" && return 0; }
2829 start_mdt 1 -o nosvc -n
2830 if [ $MDSCOUNT -ge 2 ]; then
2831 for num in $(seq 2 $MDSCOUNT); do
2832 start_mdt $num || return
2835 start ost1 $(ostdevname 1) $OST_MOUNT_OPTS
2836 start_mdt 1 -o nomgs,force
2837 mount_client $MOUNT || error "mount_client $MOUNT failed"
2840 echo "blah blah" > $MOUNT/$tfile
2843 umount_client $MOUNT || error "umount_client $MOUNT failed"
2844 stop ost1 -f || error "unable to stop OST1"
2845 stop_mds || error "Unable to stop MDS"
2846 stop_mds || error "Unable to stop MDS on second try"
2848 run_test 41a "mount mds with --nosvc and --nomgs"
2851 if [ $(facet_fstype $SINGLEMDS) == ldiskfs ] &&
2852 ! do_facet $SINGLEMDS test -b $(mdsdevname 1); then
2853 skip "Loop devices does not work with nosvc option"
2857 ! combined_mgs_mds && skip "needs combined mgs device" && return 0
2861 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/})
2863 start_mdt 1 -o nosvc -n
2864 if [ $MDSCOUNT -ge 2 ]; then
2865 for num in $(seq 2 $MDSCOUNT); do
2866 start_mdt $num || return
2869 start_ost || error "Unable to start OST1"
2870 start_mdt 1 -o nomgs,force
2871 mount_client $MOUNT || error "mount_client $MOUNT failed"
2874 echo "blah blah" > $MOUNT/$tfile
2875 cat $MOUNT/$tfile || error "cat $MOUNT/$tfile failed"
2877 umount_client $MOUNT || error "umount_client $MOUNT failed"
2878 stop_ost || error "Unable to stop OST1"
2879 stop_mds || error "Unable to stop MDS"
2880 stop_mds || error "Unable to stop MDS on second try"
2882 run_test 41b "mount mds with --nosvc and --nomgs on first mount"
2885 local server_version=$(lustre_version_code $SINGLEMDS)
2886 local oss_list=$(comma_list $(osts_nodes))
2888 [[ $server_version -ge $(version_code 2.6.52) ]] ||
2889 [[ $server_version -ge $(version_code 2.5.26) &&
2890 $server_version -lt $(version_code 2.5.50) ]] ||
2891 [[ $server_version -ge $(version_code 2.5.4) &&
2892 $server_version -lt $(version_code 2.5.11) ]] ||
2893 { skip "Need MDS version 2.5.4+ or 2.5.26+ or 2.6.52+"; return; }
2895 # ensure mds1 ost1 have been created even if running sub-test standalone
2898 cleanup || error "cleanup failed"
2900 # using directly mount command instead of start() function to avoid
2901 # any side effect of // with others/externals tools/features
2902 # ("zpool import", ...)
2904 # MDT concurrent start
2906 LOAD_MODULES_REMOTE=true load_modules
2907 do_facet $SINGLEMDS "lsmod | grep -q libcfs" ||
2908 error "MDT concurrent start: libcfs module not loaded"
2910 local mds1dev=$(mdsdevname 1)
2911 local mds1mnt=$(facet_mntpt mds1)
2912 local mds1fstype=$(facet_fstype mds1)
2913 local mds1opts=$MDS_MOUNT_OPTS
2915 if [ $mds1fstype == ldiskfs ] &&
2916 ! do_facet mds1 test -b $mds1dev; then
2917 mds1opts=$(csa_add "$mds1opts" -o loop)
2919 if [[ $mds1fstype == zfs ]]; then
2920 import_zpool mds1 || return ${PIPESTATUS[0]}
2923 #define OBD_FAIL_TGT_MOUNT_RACE 0x716
2924 do_facet mds1 "$LCTL set_param fail_loc=0x80000716"
2926 do_facet mds1 mount -t lustre $mds1dev $mds1mnt $mds1opts &
2929 do_facet mds1 mount -t lustre $mds1dev $mds1mnt $mds1opts
2933 do_facet mds1 "$LCTL set_param fail_loc=0x0"
2934 if [ $rc -eq 0 ] && [ $rc2 -ne 0 ]; then
2935 echo "1st MDT start succeed"
2936 echo "2nd MDT start failed with $rc2"
2937 elif [ $rc2 -eq 0 ] && [ $rc -ne 0 ]; then
2938 echo "1st MDT start failed with $rc"
2939 echo "2nd MDT start succeed"
2942 error "unexpected concurrent MDT mounts result, rc=$rc rc2=$rc2"
2945 if [ $MDSCOUNT -ge 2 ]; then
2946 for num in $(seq 2 $MDSCOUNT); do
2947 start_mdt $num || return
2951 # OST concurrent start
2953 do_rpc_nodes $oss_list "lsmod | grep -q libcfs" ||
2954 error "OST concurrent start: libcfs module not loaded"
2956 local ost1dev=$(ostdevname 1)
2957 local ost1mnt=$(facet_mntpt ost1)
2958 local ost1fstype=$(facet_fstype ost1)
2959 local ost1opts=$OST_MOUNT_OPTS
2961 if [ $ost1fstype == ldiskfs ] &&
2962 ! do_facet ost1 test -b $ost1dev; then
2963 ost1opts=$(csa_add "$ost1opts" -o loop)
2965 if [[ $ost1fstype == zfs ]]; then
2966 import_zpool ost1 || return ${PIPESTATUS[0]}
2969 #define OBD_FAIL_TGT_MOUNT_RACE 0x716
2970 do_facet ost1 "$LCTL set_param fail_loc=0x80000716"
2972 do_facet ost1 mount -t lustre $ost1dev $ost1mnt $ost1opts &
2975 do_facet ost1 mount -t lustre $ost1dev $ost1mnt $ost1opts
2979 do_facet ost1 "$LCTL set_param fail_loc=0x0"
2980 if [ $rc -eq 0 ] && [ $rc2 -ne 0 ]; then
2981 echo "1st OST start succeed"
2982 echo "2nd OST start failed with $rc2"
2983 elif [ $rc2 -eq 0 ] && [ $rc -ne 0 ]; then
2984 echo "1st OST start failed with $rc"
2985 echo "2nd OST start succeed"
2989 error "unexpected concurrent OST mounts result, rc=$rc rc2=$rc2"
2995 # verify everything ok
3000 error "MDT(s) start failed"
3008 error "OST(s) start failed"
3016 error "client start failed"
3023 error "client mount failed"
3027 run_test 41c "concurrent mounts of MDT/OST should all fail but one"
3029 test_42() { #bug 14693
3031 check_mount || error "client was not mounted"
3033 do_facet mgs $LCTL conf_param $FSNAME.llite.some_wrong_param=10
3034 umount_client $MOUNT ||
3035 error "unmounting client failed with invalid llite param"
3036 mount_client $MOUNT ||
3037 error "mounting client failed with invalid llite param"
3039 do_facet mgs $LCTL conf_param $FSNAME.sys.some_wrong_param=20
3040 cleanup || error "stopping $FSNAME failed with invalid sys param"
3042 check_mount || error "client was not mounted with invalid sys param"
3043 cleanup || error "stopping $FSNAME failed with invalid sys param"
3045 run_test 42 "allow client/server mount/unmount with invalid config param"
3048 [[ $(lustre_version_code mgs) -ge $(version_code 2.5.58) ]] ||
3049 { skip "Need MDS version at least 2.5.58" && return 0; }
3050 [ $UID -ne 0 -o $RUNAS_ID -eq 0 ] && skip_env "run as root"
3053 USER1=$(getent passwd | grep :$ID1:$ID1: | cut -d: -f1)
3054 [ -z "$USER1" ] && skip_env "missing user with uid=$ID1 gid=$ID1" &&
3058 chmod ugo+x $DIR || error "chmod 0 failed"
3059 set_conf_param_and_check mds1 \
3060 "$LCTL get_param -n mdt.$FSNAME-MDT0000.root_squash" \
3061 "$FSNAME.mdt.root_squash" \
3063 wait_update $HOSTNAME \
3064 "$LCTL get_param -n llite.${FSNAME}*.root_squash" \
3066 error "check llite root_squash failed!"
3067 set_conf_param_and_check mds1 \
3068 "$LCTL get_param -n mdt.$FSNAME-MDT0000.nosquash_nids" \
3069 "$FSNAME.mdt.nosquash_nids" \
3071 wait_update $HOSTNAME \
3072 "$LCTL get_param -n llite.${FSNAME}*.nosquash_nids" \
3074 error "check llite nosquash_nids failed!"
3077 # create set of test files
3079 echo "111" > $DIR/$tfile-userfile || error "write 1 failed"
3080 chmod go-rw $DIR/$tfile-userfile || error "chmod 1 failed"
3081 chown $RUNAS_ID.$RUNAS_ID $DIR/$tfile-userfile || error "chown failed"
3083 echo "222" > $DIR/$tfile-rootfile || error "write 2 failed"
3084 chmod go-rw $DIR/$tfile-rootfile || error "chmod 2 faield"
3086 mkdir $DIR/$tdir-rootdir || error "mkdir failed"
3087 chmod go-rwx $DIR/$tdir-rootdir || error "chmod 3 failed"
3088 touch $DIR/$tdir-rootdir/tfile-1 || error "touch failed"
3090 echo "777" > $DIR/$tfile-user1file || error "write 7 failed"
3091 chmod go-rw $DIR/$tfile-user1file || error "chmod 7 failed"
3092 chown $ID1.$ID1 $DIR/$tfile-user1file || error "chown failed"
3095 # check root_squash:
3096 # set root squash UID:GID to RUNAS_ID
3097 # root should be able to access only files owned by RUNAS_ID
3099 set_conf_param_and_check mds1 \
3100 "$LCTL get_param -n mdt.$FSNAME-MDT0000.root_squash" \
3101 "$FSNAME.mdt.root_squash" \
3102 "$RUNAS_ID:$RUNAS_ID"
3103 wait_update $HOSTNAME \
3104 "$LCTL get_param -n llite.${FSNAME}*.root_squash" \
3105 "$RUNAS_ID:$RUNAS_ID" ||
3106 error "check llite root_squash failed!"
3108 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-userfile)
3109 dd if=$DIR/$tfile-userfile 1>/dev/null 2>/dev/null ||
3110 error "$ST: root read permission is denied"
3111 echo "$ST: root read permission is granted - ok"
3114 dd conv=notrunc of=$DIR/$tfile-userfile 1>/dev/null 2>/dev/null ||
3115 error "$ST: root write permission is denied"
3116 echo "$ST: root write permission is granted - ok"
3118 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-rootfile)
3119 dd if=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null &&
3120 error "$ST: root read permission is granted"
3121 echo "$ST: root read permission is denied - ok"
3124 dd conv=notrunc of=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null &&
3125 error "$ST: root write permission is granted"
3126 echo "$ST: root write permission is denied - ok"
3128 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tdir-rootdir)
3129 rm $DIR/$tdir-rootdir/tfile-1 1>/dev/null 2>/dev/null &&
3130 error "$ST: root unlink permission is granted"
3131 echo "$ST: root unlink permission is denied - ok"
3133 touch $DIR/tdir-rootdir/tfile-2 1>/dev/null 2>/dev/null &&
3134 error "$ST: root create permission is granted"
3135 echo "$ST: root create permission is denied - ok"
3138 # check root_squash is enforced independently
3139 # of client cache content
3141 # access file by USER1, keep access open
3142 # root should be denied access to user file
3144 runas -u $ID1 tail -f $DIR/$tfile-user1file 1>/dev/null 2>&1 &
3148 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-user1file)
3149 dd if=$DIR/$tfile-user1file 1>/dev/null 2>&1 &&
3150 { kill $pid; error "$ST: root read permission is granted"; }
3151 echo "$ST: root read permission is denied - ok"
3154 dd conv=notrunc of=$DIR/$tfile-user1file 1>/dev/null 2>&1 &&
3155 { kill $pid; error "$ST: root write permission is granted"; }
3156 echo "$ST: root write permission is denied - ok"
3162 # check nosquash_nids:
3163 # put client's NID into nosquash_nids list,
3164 # root should be able to access root file after that
3166 local NIDLIST=$($LCTL list_nids all | tr '\n' ' ')
3167 NIDLIST="2@gni $NIDLIST 192.168.0.[2,10]@tcp"
3168 NIDLIST=$(echo $NIDLIST | tr -s ' ' ' ')
3169 set_conf_param_and_check mds1 \
3170 "$LCTL get_param -n mdt.$FSNAME-MDT0000.nosquash_nids" \
3171 "$FSNAME-MDTall.mdt.nosquash_nids" \
3173 wait_update $HOSTNAME \
3174 "$LCTL get_param -n llite.${FSNAME}*.nosquash_nids" \
3176 error "check llite nosquash_nids failed!"
3178 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-rootfile)
3179 dd if=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null ||
3180 error "$ST: root read permission is denied"
3181 echo "$ST: root read permission is granted - ok"
3184 dd conv=notrunc of=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null ||
3185 error "$ST: root write permission is denied"
3186 echo "$ST: root write permission is granted - ok"
3188 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tdir-rootdir)
3189 rm $DIR/$tdir-rootdir/tfile-1 ||
3190 error "$ST: root unlink permission is denied"
3191 echo "$ST: root unlink permission is granted - ok"
3192 touch $DIR/$tdir-rootdir/tfile-2 ||
3193 error "$ST: root create permission is denied"
3194 echo "$ST: root create permission is granted - ok"
3195 cleanup || error "cleanup failed with $?"
3197 run_test 43a "check root_squash and nosquash_nids"
3199 test_43b() { # LU-5690
3200 [[ $(lustre_version_code mgs) -ge $(version_code 2.7.62) ]] ||
3201 { skip "Need MGS version 2.7.62+"; return; }
3203 if [[ -z "$fs2mds_DEV" ]]; then
3204 is_blkdev $SINGLEMDS $(mdsdevname ${SINGLEMDS//mds/}) &&
3205 skip_env "mixed loopback and real device not working" && return
3208 local fs2mdsdev=$(mdsdevname 1_2)
3209 local fs2mdsvdev=$(mdsvdevname 1_2)
3211 # temporarily use fs2mds as fs2mgs
3213 local fs2mgsdev=$fs2mdsdev
3214 local fs2mgsvdev=$fs2mdsvdev
3216 local fsname=test1234
3218 load_module llite/lustre
3219 local client_ip=$(host_nids_address $HOSTNAME $NETTYPE)
3220 local host=${client_ip//*./}
3221 local net=${client_ip/%$host/}
3222 local nosquash_nids=$(h2$NETTYPE $net[$host,$host,$host])
3224 add $fs2mgs $(mkfs_opts mgs $fs2mgsdev) --fsname=$fsname \
3225 --param mdt.root_squash=$RUNAS_ID:$RUNAS_ID \
3226 --param mdt.nosquash_nids=$nosquash_nids \
3227 --reformat $fs2mgsdev $fs2mgsvdev || error "add fs2mgs failed"
3228 start $fs2mgs $fs2mgsdev $MGS_MOUNT_OPTS || error "start fs2mgs failed"
3229 stop $fs2mgs -f || error "stop fs2mgs failed"
3231 run_test 43b "parse nosquash_nids with commas in expr_list"
3233 umount_client $MOUNT
3238 check_mount || error "check_mount"
3239 UUID=$($LCTL get_param llite.${FSNAME}*.uuid | cut -d= -f2)
3241 UUIDS=$(do_facet $SINGLEMDS "$LCTL get_param mdt.${FSNAME}*.exports.*.uuid")
3242 for VAL in $UUIDS; do
3243 NID=$(echo $VAL | cut -d= -f1)
3244 CLUUID=$(echo $VAL | cut -d= -f2)
3245 [ "$UUID" = "$CLUUID" ] && STATS_FOUND=yes && break
3247 [ "$STATS_FOUND" = "no" ] && error "stats not found for client"
3248 cleanup || error "cleanup failed with $?"
3250 run_test 44 "mounted client proc entry exists"
3254 check_mount || error "check_mount"
3255 stop_mds || error "Unable to stop MDS"
3259 #define OBD_FAIL_PTLRPC_LONG_REPL_UNLINK 0x50f
3260 do_facet client "$LCTL set_param fail_loc=0x8000050f"
3263 manual_umount_client --force || error "manual_umount_client failed"
3264 do_facet client "$LCTL set_param fail_loc=0x0"
3265 start_mds || error "unable to start MDS"
3266 mount_client $MOUNT || error "mount_client $MOUNT failed"
3267 cleanup || error "cleanup failed with $?"
3269 run_test 45 "long unlink handling in ptlrpcd"
3276 umount_client $MOUNT2 || rc=$?
3277 umount_client $MOUNT || rc=$?
3278 while [ $count -gt 0 ]; do
3279 stop ost${count} -f || rc=$?
3283 cleanup_nocli || rc=$?
3284 #writeconf to remove all ost2 traces for subsequent tests
3285 writeconf_or_reformat
3290 echo "Testing with $OSTCOUNT OSTs"
3292 start_mds || error "unable to start MDS"
3293 #first client should see only one ost
3294 start_ost || error "Unable to start OST1"
3295 wait_osc_import_state mds ost FULL
3297 mount_client $MOUNT || error "mount_client $MOUNT failed"
3298 trap "cleanup_46a $OSTCOUNT" EXIT ERR
3301 for (( i=2; i<=$OSTCOUNT; i++ )); do
3302 start ost$i $(ostdevname $i) $OST_MOUNT_OPTS ||
3303 error "start_ost$i $(ostdevname $i) failed"
3306 # wait until osts in sync
3307 for (( i=2; i<=$OSTCOUNT; i++ )); do
3308 wait_osc_import_state mds ost$i FULL
3309 wait_osc_import_state client ost$i FULL
3312 #second client see all ost's
3314 mount_client $MOUNT2 || error "mount_client failed"
3315 $SETSTRIPE -c -1 $MOUNT2 || error "$SETSTRIPE -c -1 $MOUNT2 failed"
3316 $GETSTRIPE $MOUNT2 || error "$GETSTRIPE $MOUNT2 failed"
3318 echo "ok" > $MOUNT2/widestripe
3319 $GETSTRIPE $MOUNT2/widestripe ||
3320 error "$GETSTRIPE $MOUNT2/widestripe failed"
3321 # fill acl buffer for avoid expand lsm to them
3322 awk -F : '{if (FNR < 25) { print "u:"$1":rwx" }}' /etc/passwd |
3324 setfacl -m $acl $MOUNT2/widestripe
3328 stat $MOUNT/widestripe || error "stat $MOUNT/widestripe failed"
3330 cleanup_46a $OSTCOUNT || error "cleanup_46a failed"
3332 run_test 46a "handle ost additional - wide striped file"
3337 check_mount || error "check_mount failed"
3338 $LCTL set_param ldlm.namespaces.$FSNAME-*-*-*.lru_size=100
3342 for ns in $($LCTL get_param ldlm.namespaces.$FSNAME-*-*-*.lru_size); do
3343 if echo $ns | grep "MDT[[:digit:]]*"; then
3346 lrs=$(echo $ns | sed 's/.*lru_size=//')
3347 lru_size[count]=$lrs
3352 facet_failover $SINGLEMDS
3353 client_up || error "client_up failed"
3356 for ns in $($LCTL get_param ldlm.namespaces.$FSNAME-*-*-*.lru_size); do
3357 if echo $ns | grep "MDT[[:digit:]]*"; then
3360 lrs=$(echo $ns | sed 's/.*lru_size=//')
3361 if ! test "$lrs" -eq "${lru_size[count]}"; then
3362 n=$(echo $ns | sed -e 's/ldlm.namespaces.//' -e 's/.lru_size=.*//')
3363 error "$n has lost lru_size: $lrs vs. ${lru_size[count]}"
3368 cleanup || error "cleanup failed with $?"
3370 run_test 47 "server restart does not make client loss lru_resize settings"
3375 # reformat after this test is needed - if the test fails,
3376 # we will have unkillable file at FS
3380 test_48() { # bug 17636
3383 check_mount || error "check_mount failed"
3385 $SETSTRIPE -c -1 $MOUNT || error "$SETSTRIPE -c -1 $MOUNT failed"
3386 $GETSTRIPE $MOUNT || error "$GETSTRIPE $MOUNT failed"
3388 echo "ok" > $MOUNT/widestripe
3389 $GETSTRIPE $MOUNT/widestripe ||
3390 error "$GETSTRIPE $MOUNT/widestripe failed"
3392 trap cleanup_48 EXIT ERR
3394 # fill acl buffer for avoid expand lsm to them
3395 getent passwd | awk -F : '{ print "u:"$1":rwx" }' | while read acl; do
3396 setfacl -m $acl $MOUNT/widestripe
3399 stat $MOUNT/widestripe || error "stat $MOUNT/widestripe failed"
3403 run_test 48 "too many acls on file"
3405 # check PARAM_SYS_LDLM_TIMEOUT option of MKFS.LUSTRE
3406 test_49a() { # bug 17710
3407 local timeout_orig=$TIMEOUT
3408 local ldlm_timeout_orig=$LDLM_TIMEOUT
3409 local LOCAL_TIMEOUT=20
3411 LDLM_TIMEOUT=$LOCAL_TIMEOUT
3412 TIMEOUT=$LOCAL_TIMEOUT
3416 check_mount || error "client mount failed"
3418 echo "check ldlm_timout..."
3419 local LDLM_MDS="$(do_facet $SINGLEMDS $LCTL get_param -n ldlm_timeout)"
3420 local LDLM_OST1="$(do_facet ost1 $LCTL get_param -n ldlm_timeout)"
3421 local LDLM_CLIENT="$(do_facet client $LCTL get_param -n ldlm_timeout)"
3423 if [ $LDLM_MDS -ne $LDLM_OST1 -o $LDLM_MDS -ne $LDLM_CLIENT ]; then
3424 error "Different LDLM_TIMEOUT:$LDLM_MDS $LDLM_OST1 $LDLM_CLIENT"
3427 if [ $LDLM_MDS -ne $((LOCAL_TIMEOUT / 3)) ]; then
3428 error "LDLM_TIMEOUT($LDLM_MDS) is not $((LOCAL_TIMEOUT / 3))"
3431 umount_client $MOUNT || error "umount_client $MOUNT failed"
3432 stop_ost || error "problem stopping OSS"
3433 stop_mds || error "problem stopping MDS"
3435 LDLM_TIMEOUT=$ldlm_timeout_orig
3436 TIMEOUT=$timeout_orig
3438 run_test 49a "check PARAM_SYS_LDLM_TIMEOUT option of mkfs.lustre"
3440 test_49b() { # bug 17710
3441 local timeout_orig=$TIMEOUT
3442 local ldlm_timeout_orig=$LDLM_TIMEOUT
3443 local LOCAL_TIMEOUT=20
3445 LDLM_TIMEOUT=$((LOCAL_TIMEOUT - 1))
3446 TIMEOUT=$LOCAL_TIMEOUT
3450 check_mount || error "client mount failed"
3452 local LDLM_MDS="$(do_facet $SINGLEMDS $LCTL get_param -n ldlm_timeout)"
3453 local LDLM_OST1="$(do_facet ost1 $LCTL get_param -n ldlm_timeout)"
3454 local LDLM_CLIENT="$(do_facet client $LCTL get_param -n ldlm_timeout)"
3456 if [ $LDLM_MDS -ne $LDLM_OST1 -o $LDLM_MDS -ne $LDLM_CLIENT ]; then
3457 error "Different LDLM_TIMEOUT:$LDLM_MDS $LDLM_OST1 $LDLM_CLIENT"
3460 if [ $LDLM_MDS -ne $((LOCAL_TIMEOUT - 1)) ]; then
3461 error "LDLM_TIMEOUT($LDLM_MDS) is not $((LOCAL_TIMEOUT - 1))"
3464 cleanup || error "cleanup failed"
3466 LDLM_TIMEOUT=$ldlm_timeout_orig
3467 TIMEOUT=$timeout_orig
3469 run_test 49b "check PARAM_SYS_LDLM_TIMEOUT option of mkfs.lustre"
3472 # Test both statfs and lfs df and fail if either one fails
3473 multiop_bg_pause $1 f_
3476 killall -USR1 multiop
3477 [ $RC1 -ne 0 ] && log "lazystatfs multiop failed"
3478 wait $PID || { RC1=$?; log "multiop return error "; }
3485 if [ $RC2 -eq 0 ]; then
3487 log "lazystatfs df failed"
3491 [[ $RC1 -ne 0 || $RC2 -eq 0 ]] && RC=1
3497 $LCTL set_param llite.$FSNAME-*.lazystatfs=1
3498 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
3500 lazystatfs $MOUNT || error "lazystatfs failed but no down servers"
3502 cleanup || error "cleanup failed with rc $?"
3504 run_test 50a "lazystatfs all servers available"
3508 $LCTL set_param llite.$FSNAME-*.lazystatfs=1
3509 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
3511 # Wait for client to detect down OST
3512 stop_ost || error "Unable to stop OST1"
3513 wait_osc_import_state mds ost DISCONN
3515 lazystatfs $MOUNT || error "lazystatfs should not return EIO"
3517 umount_client $MOUNT || error "Unable to unmount client"
3518 stop_mds || error "Unable to stop MDS"
3520 run_test 50b "lazystatfs all servers down"
3523 start_mds || error "Unable to start MDS"
3524 start_ost || error "Unable to start OST1"
3525 start_ost2 || error "Unable to start OST2"
3526 mount_client $MOUNT || error "Unable to mount client"
3527 $LCTL set_param llite.$FSNAME-*.lazystatfs=1
3528 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
3530 # Wait for client to detect down OST
3531 stop_ost || error "Unable to stop OST1"
3532 wait_osc_import_state mds ost DISCONN
3533 lazystatfs $MOUNT || error "lazystatfs failed with one down server"
3535 umount_client $MOUNT || error "Unable to unmount client"
3536 stop_ost2 || error "Unable to stop OST2"
3537 stop_mds || error "Unable to stop MDS"
3538 #writeconf to remove all ost2 traces for subsequent tests
3539 writeconf_or_reformat
3541 run_test 50c "lazystatfs one server down"
3544 start_mds || error "Unable to start MDS"
3545 start_ost || error "Unable to start OST1"
3546 start_ost2 || error "Unable to start OST2"
3547 mount_client $MOUNT || error "Unable to mount client"
3548 $LCTL set_param llite.$FSNAME-*.lazystatfs=1
3549 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
3551 # Issue the statfs during the window where the client still
3552 # belives the OST to be available but it is in fact down.
3553 # No failure just a statfs which hangs for a timeout interval.
3554 stop_ost || error "Unable to stop OST1"
3555 lazystatfs $MOUNT || error "lazystatfs failed with one down server"
3557 umount_client $MOUNT || error "Unable to unmount client"
3558 stop_ost2 || error "Unable to stop OST2"
3559 stop_mds || error "Unable to stop MDS"
3560 #writeconf to remove all ost2 traces for subsequent tests
3561 writeconf_or_reformat
3563 run_test 50d "lazystatfs client/server conn race"
3570 start_mds || error "Unable to start MDS"
3571 #first client should see only one ost
3572 start_ost || error "Unable to start OST1"
3573 wait_osc_import_state mds ost FULL
3575 # Wait for client to detect down OST
3576 stop_ost || error "Unable to stop OST1"
3577 wait_osc_import_state mds ost DISCONN
3579 mount_client $MOUNT || error "Unable to mount client"
3580 $LCTL set_param llite.$FSNAME-*.lazystatfs=0
3582 multiop_bg_pause $MOUNT _f
3586 if [ $RC1 -ne 0 ]; then
3587 log "multiop failed $RC1"
3590 sleep $(( $TIMEOUT+1 ))
3592 [ $? -ne 0 ] && error "process isn't sleep"
3593 start_ost || error "Unable to start OST1"
3594 wait $pid || error "statfs failed"
3597 umount_client $MOUNT || error "Unable to unmount client"
3598 stop_ost || error "Unable to stop OST1"
3599 stop_mds || error "Unable to stop MDS"
3601 run_test 50e "normal statfs all servers down"
3606 CONN_PROC="osc.$FSNAME-OST0001-osc-[M]*.ost_server_uuid"
3608 start_mds || error "Unable to start mds"
3609 #first client should see only one ost
3610 start_ost || error "Unable to start OST1"
3611 wait_osc_import_state mds ost FULL
3613 start_ost2 || error "Unable to start OST2"
3614 wait_osc_import_state mds ost2 FULL
3616 # Wait for client to detect down OST
3617 stop_ost2 || error "Unable to stop OST2"
3619 wait_osc_import_state mds ost2 DISCONN
3620 mount_client $MOUNT || error "Unable to mount client"
3621 $LCTL set_param llite.$FSNAME-*.lazystatfs=0
3623 multiop_bg_pause $MOUNT _f
3627 if [ $RC1 -ne 0 ]; then
3628 log "lazystatfs multiop failed $RC1"
3631 sleep $(( $TIMEOUT+1 ))
3633 [ $? -ne 0 ] && error "process isn't sleep"
3634 start_ost2 || error "Unable to start OST2"
3635 wait $pid || error "statfs failed"
3636 stop_ost2 || error "Unable to stop OST2"
3639 umount_client $MOUNT -f || error "Unable to unmount client"
3640 stop_ost || error "Unable to stop OST1"
3641 stop_mds || error "Unable to stop MDS"
3642 #writeconf to remove all ost2 traces for subsequent tests
3643 writeconf_or_reformat
3645 run_test 50f "normal statfs one server in down"
3648 [ "$OSTCOUNT" -lt "2" ] && skip_env "$OSTCOUNT < 2, skipping" && return
3650 start_ost2 || error "Unable to start OST2"
3651 wait_osc_import_state mds ost2 FULL
3652 wait_osc_import_state client ost2 FULL
3654 local PARAM="${FSNAME}-OST0001.osc.active"
3656 $SETSTRIPE -c -1 $DIR/$tfile || error "$SETSTRIPE failed"
3657 do_facet mgs $LCTL conf_param $PARAM=0 ||
3658 error "Unable to deactivate OST"
3660 umount_client $MOUNT || error "Unable to unmount client"
3661 mount_client $MOUNT || error "Unable to mount client"
3662 # This df should not cause a panic
3665 do_facet mgs $LCTL conf_param $PARAM=1 || error "Unable to activate OST"
3666 rm -f $DIR/$tfile || error "unable to remove file $DIR/$tfile"
3667 umount_client $MOUNT || error "Unable to unmount client"
3668 stop_ost2 || error "Unable to stop OST2"
3669 stop_ost || error "Unable to stop OST1"
3670 stop_mds || error "Unable to stop MDS"
3671 #writeconf to remove all ost2 traces for subsequent tests
3672 writeconf_or_reformat
3674 run_test 50g "deactivated OST should not cause panic"
3678 # prepare MDT/OST, make OSC inactive for OST1
3679 [ "$OSTCOUNT" -lt "2" ] && skip_env "$OSTCOUNT < 2, skipping" && return
3681 [ $(facet_fstype ost1) == zfs ] && import_zpool ost1
3682 do_facet ost1 "$TUNEFS --param osc.active=0 `ostdevname 1`" ||
3683 error "tunefs OST1 failed"
3684 start_mds || error "Unable to start MDT"
3685 start_ost || error "Unable to start OST1"
3686 start_ost2 || error "Unable to start OST2"
3687 mount_client $MOUNT || error "client start failed"
3689 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
3691 # activatate OSC for OST1
3692 local TEST="$LCTL get_param -n osc.${FSNAME}-OST0000-osc-[!M]*.active"
3693 set_conf_param_and_check client \
3694 "$TEST" "${FSNAME}-OST0000.osc.active" 1 ||
3695 error "Unable to activate OST1"
3697 mkdir $DIR/$tdir/2 || error "mkdir $DIR/$tdir/2 failed"
3698 $SETSTRIPE -c -1 -i 0 $DIR/$tdir/2 ||
3699 error "$SETSTRIPE $DIR/$tdir/2 failed"
3700 sleep 1 && echo "create a file after OST1 is activated"
3702 createmany -o $DIR/$tdir/2/$tfile-%d 1
3704 # check OSC import is working
3705 stat $DIR/$tdir/2/* >/dev/null 2>&1 ||
3706 error "some OSC imports are still not connected"
3709 umount_client $MOUNT || error "Unable to umount client"
3710 stop_ost2 || error "Unable to stop OST2"
3711 cleanup_nocli || error "cleanup_nocli failed with $?"
3713 run_test 50h "LU-642: activate deactivated OST"
3716 # prepare MDT/OST, make OSC inactive for OST1
3717 [ "$MDSCOUNT" -lt "2" ] && skip_env "$MDSCOUNT < 2, skipping" && return
3719 [ $(facet_fstype mds2) == zfs ] && import_zpool mds2
3720 do_facet mds2 "$TUNEFS --param mdc.active=0 $(mdsdevname 2)" ||
3721 error "tunefs MDT2 failed"
3722 start_mds || error "Unable to start MDT"
3723 start_ost || error "Unable to start OST1"
3724 start_ost2 || error "Unable to start OST2"
3725 mount_client $MOUNT || error "client start failed"
3727 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
3729 $LCTL conf_param ${FSNAME}-MDT0000.mdc.active=0 &&
3730 error "deactive MDC0 succeeds"
3731 # activate MDC for MDT2
3732 local TEST="$LCTL get_param -n mdc.${FSNAME}-MDT0001-mdc-[!M]*.active"
3733 set_conf_param_and_check client \
3734 "$TEST" "${FSNAME}-MDT0001.mdc.active" 1 ||
3735 error "Unable to activate MDT2"
3737 wait_clients_import_state ${CLIENTS:-$HOSTNAME} mds2 FULL
3738 if [ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.7.60) ]
3740 wait_dne_interconnect
3742 $LFS mkdir -i1 $DIR/$tdir/2 || error "mkdir $DIR/$tdir/2 failed"
3744 createmany -o $DIR/$tdir/2/$tfile-%d 1 || error "create files failed"
3746 rm -rf $DIR/$tdir/2 || error "unlink dir failed"
3748 # deactivate MDC for MDT2
3749 local TEST="$LCTL get_param -n mdc.${FSNAME}-MDT0001-mdc-[!M]*.active"
3750 set_conf_param_and_check client \
3751 "$TEST" "${FSNAME}-MDT0001.mdc.active" 0 ||
3752 error "Unable to deactivate MDT2"
3754 $LFS mkdir -i1 $DIR/$tdir/2 &&
3755 error "mkdir $DIR/$tdir/2 succeeds after deactive MDT"
3758 umount_client $MOUNT || error "Unable to umount client"
3763 run_test 50i "activate deactivated MDT"
3766 local LOCAL_TIMEOUT=20
3770 check_mount || error "check_mount failed"
3772 mkdir $MOUNT/$tdir || error "mkdir $MOUNT/$tdir failed"
3773 $SETSTRIPE -c -1 $MOUNT/$tdir ||
3774 error "$SETSTRIPE -c -1 $MOUNT/$tdir failed"
3775 #define OBD_FAIL_MDS_REINT_DELAY 0x142
3776 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0x142"
3777 touch $MOUNT/$tdir/$tfile &
3780 start_ost2 || error "Unable to start OST1"
3782 stop_ost2 || error "Unable to stop OST1"
3783 umount_client $MOUNT -f || error “unmount $MOUNT failed”
3784 cleanup_nocli || error “stop server failed”
3785 #writeconf to remove all ost2 traces for subsequent tests
3786 writeconf_or_reformat
3788 run_test 51 "Verify that mdt_reint handles RMF_MDT_MD correctly when an OST is added"
3797 do_node $node mkdir -p $dest
3798 [ $? -eq 0 ] || { error "Unable to create directory"; return 1; }
3800 do_node $node 'tar cf - '$@' | tar xf - -C '$dest';
3801 [ \"\${PIPESTATUS[*]}\" = \"0 0\" ] || exit 1'
3802 [ $? -eq 0 ] || { error "Unable to tar files"; return 2; }
3804 do_node $node 'getfattr -d -m "[a-z]*\\." '$@' > '$xattrs
3805 [ $? -eq 0 ] || { error "Unable to read xattrs"; return 3; }
3815 local backup2=${TMP}/backup2
3817 do_node $node mkdir -p $backup2
3818 [ $? -eq 0 ] || { error "Unable to create directory"; return 1; }
3820 do_node $node 'tar cf - '$@' | tar xf - -C '$backup2';
3821 [ \"\${PIPESTATUS[*]}\" = \"0 0\" ] || exit 1'
3822 [ $? -eq 0 ] || { error "Unable to tar files to diff"; return 2; }
3824 do_node $node "diff -rq $backup $backup2"
3825 [ $? -eq 0 ] || { error "contents differ"; return 3; }
3827 local xattrs2=${TMP}/xattrs2
3828 do_node $node 'getfattr -d -m "[a-z]*\\." '$@' > '$xattrs2
3829 [ $? -eq 0 ] || { error "Unable to read xattrs to diff"; return 4; }
3831 do_node $node "diff $xattrs $xattrs2"
3832 [ $? -eq 0 ] || { error "xattrs differ"; return 5; }
3834 do_node $node "rm -rf $backup2 $xattrs2"
3835 [ $? -eq 0 ] || { error "Unable to delete temporary files"; return 6; }
3839 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
3840 skip "Only applicable to ldiskfs-based MDTs"
3844 start_mds || error "Unable to start MDS"
3845 start_ost || error "Unable to start OST1"
3846 mount_client $MOUNT || error "Unable to mount client"
3849 local ost1mnt=$(facet_mntpt ost1)
3850 local ost1node=$(facet_active_host ost1)
3851 local ost1tmp=$TMP/conf52
3854 mkdir $DIR/$tdir || error "Unable to create $DIR/$tdir"
3855 touch $TMP/modified_first || error "Unable to create temporary file"
3856 local mtime=$(stat -c %Y $TMP/modified_first)
3857 do_node $ost1node "mkdir -p $ost1tmp &&
3858 touch -m -d @$mtime $ost1tmp/modified_first" ||
3859 error "Unable to create temporary file"
3862 $SETSTRIPE -c -1 -S 1M $DIR/$tdir || error "$SETSTRIPE failed"
3864 for (( i=0; i < nrfiles; i++ )); do
3865 multiop $DIR/$tdir/$tfile-$i Ow1048576w1048576w524288c ||
3866 error "multiop failed"
3872 echo backup files to $TMP/$tdir
3873 local files=$(find $DIR/$tdir -type f -newer $TMP/modified_first)
3874 copy_files_xattrs $(hostname) $TMP/$tdir $TMP/file_xattrs $files ||
3875 error "Unable to copy files"
3877 umount_client $MOUNT || error "Unable to umount client"
3878 stop_ost || error "Unable to stop ost1"
3880 echo mount ost1 as ldiskfs
3881 do_node $ost1node mkdir -p $ost1mnt || error "Unable to create $ost1mnt"
3882 if ! do_node $ost1node test -b $ost1_dev; then
3885 do_node $ost1node mount -t $(facet_fstype ost1) $loop $ost1_dev \
3887 error "Unable to mount ost1 as ldiskfs"
3890 echo backup objects to $ost1tmp/objects
3891 local objects=$(do_node $ost1node 'find '$ost1mnt'/O/[0-9]* -type f'\
3892 '-size +0 -newer '$ost1tmp'/modified_first -regex ".*\/[0-9]+"')
3893 copy_files_xattrs $ost1node $ost1tmp/objects $ost1tmp/object_xattrs \
3895 error "Unable to copy objects"
3897 # move objects to lost+found
3898 do_node $ost1node 'mv '$objects' '${ost1mnt}'/lost+found'
3899 [ $? -eq 0 ] || { error "Unable to move objects"; return 14; }
3901 do_node $ost1node "umount $ost1mnt" ||
3902 error "Unable to umount ost1 as ldiskfs"
3904 start_ost || error "Unable to start OST1"
3905 mount_client $MOUNT || error "Unable to mount client"
3907 local REPAIRED=$(do_node $ost1node "$LCTL get_param \
3908 -n osd-ldiskfs.$FSNAME-OST0000.oi_scrub" |
3909 awk '/^lf_repa[ri]*ed/ { print $2 }')
3910 [ $REPAIRED -gt 0 ] ||
3911 error "Some entry under /lost+found should be repaired"
3914 diff_files_xattrs $(hostname) $TMP/$tdir $TMP/file_xattrs $files ||
3915 error "Unable to diff files"
3917 rm -rf $TMP/$tdir $TMP/file_xattrs ||
3918 error "Unable to delete temporary files"
3919 do_node $ost1node "rm -rf $ost1tmp" ||
3920 error "Unable to delete temporary files"
3921 cleanup || error "cleanup failed with $?"
3923 run_test 52 "check recovering objects from lost+found"
3925 # Checks threads_min/max/started for some service
3927 # Arguments: service name (OST or MDT), facet (e.g., ost1, $SINGLEMDS), and a
3928 # parameter pattern prefix like 'ost.*.ost'.
3941 local msg="Insane $modname thread counts"
3942 local ncpts=$(check_cpt_number $facet)
3946 check_mount || return 41
3948 # We need to expand $parampat, but it may match multiple parameters, so
3949 # we'll pick the first one
3950 if ! paramp=$(do_facet $facet "lctl get_param -N ${parampat}.threads_min"|head -1); then
3951 error "Couldn't expand ${parampat}.threads_min parameter name"
3955 # Remove the .threads_min part
3956 paramp=${paramp%.threads_min}
3958 # Check for sanity in defaults
3959 tmin=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min" ||
3961 tmax=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_max" ||
3963 tstarted=$(do_facet $facet "$LCTL get_param \
3964 -n ${paramp}.threads_started" || echo 0)
3965 lassert 23 "$msg (PDSH problems?)" '(($tstarted && $tmin && $tmax))' ||
3967 lassert 24 "$msg" '(($tstarted >= $tmin && $tstarted <= $tmax ))' ||
3969 nthrs=$(expr $tmax - $tmin)
3970 if [ $nthrs -lt $ncpts ]; then
3976 [ $tmin -eq $tmax -a $tmin -eq $tstarted ] &&
3977 skip_env "module parameter forced $facet thread count" &&
3978 tmin=3 && tmax=$((3 * tmax))
3980 # Check that we can change min/max
3981 do_facet $facet "$LCTL set_param \
3982 ${paramp}.threads_min=$((tmin + nthrs))"
3983 do_facet $facet "$LCTL set_param \
3984 ${paramp}.threads_max=$((tmax - nthrs))"
3985 tmin2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min" ||
3987 tmax2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_max" ||
3989 lassert 25 "$msg" '(($tmin2 == ($tmin + $nthrs) &&
3990 $tmax2 == ($tmax - $nthrs)))' || return $?
3992 # Check that we can set min/max to the same value
3993 tmin=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min" ||
3995 do_facet $facet "$LCTL set_param ${paramp}.threads_max=$tmin"
3996 tmin2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min" ||
3998 tmax2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_max" ||
4000 lassert 26 "$msg" '(($tmin2 == $tmin && $tmax2 == $tmin))' || return $?
4002 # Check that we can't set max < min
4003 do_facet $facet "$LCTL set_param ${paramp}.threads_max=$((tmin - 1))"
4004 tmin2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min" ||
4006 tmax2=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_max" ||
4008 lassert 27 "$msg" '(($tmin2 <= $tmax2))' || return $?
4010 # We need to ensure that we get the module options desired; to do this
4011 # we set LOAD_MODULES_REMOTE=true and we call setmodopts below.
4012 LOAD_MODULES_REMOTE=true
4015 local newvalue="${opts}=$(expr $basethr \* $ncpts)"
4016 setmodopts -a $modname "$newvalue" oldvalue
4019 check_mount || return 41
4021 # Restore previous setting of MODOPTS_*
4022 setmodopts $modname "$oldvalue"
4024 # Check that $opts took
4025 tmin=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_min")
4026 tmax=$(do_facet $facet "$LCTL get_param -n ${paramp}.threads_max")
4027 tstarted=$(do_facet $facet \
4028 "$LCTL get_param -n ${paramp}.threads_started")
4029 lassert 28 "$msg" '(($tstarted >= $tmin && $tstarted <= $tmax ))' ||
4038 thread_sanity OST ost1 'ost.*.ost' 'oss_num_threads' '16'
4039 cleanup || error "cleanup failed with rc $?"
4041 run_test 53a "check OSS thread count params"
4045 local mds=$(do_facet $SINGLEMDS "$LCTL get_param \
4046 -N mds.*.*.threads_max 2>/dev/null")
4047 if [ -z "$mds" ]; then
4048 #running this on an old MDT
4049 thread_sanity MDT $SINGLEMDS 'mdt.*.*.' 'mdt_num_threads' 16
4051 thread_sanity MDT $SINGLEMDS 'mds.*.*.' 'mds_num_threads' 16
4053 cleanup || error "cleanup failed with $?"
4055 run_test 53b "check MDS thread count params"
4058 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
4059 skip "Only applicable to ldiskfs-based MDTs"
4063 do_rpc_nodes $(facet_host ost1) run_llverdev $(ostdevname 1) -p ||
4064 error "llverdev failed with rc=$?"
4067 run_test 54a "test llverdev and partial verify of device"
4070 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
4071 skip "Only applicable to ldiskfs-based MDTs"
4076 run_llverfs $MOUNT -p || error "llverfs failed with rc=$?"
4077 cleanup || error "cleanup failed with rc=$?"
4079 run_test 54b "test llverfs and partial verify of filesystem"
4083 local max_ost_index=$1
4084 echo -n $(((max_ost_index + 1) * 8))
4088 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
4089 skip "Only applicable to ldiskfs-based MDTs"
4093 local mdsdev=$(mdsdevname 1)
4094 local mdsvdev=$(mdsvdevname 1)
4098 if ! combined_mgs_mds; then
4099 stop_mgs || error "stopping MGS service failed"
4100 format_mgs || error "formatting MGT failed"
4102 add mds1 $(mkfs_opts mds1 ${mdsdev}) --reformat $mdsdev \
4104 add ost1 $(mkfs_opts ost1 $(ostdevname 1)) --index=$i \
4105 --reformat $(ostdevname 1) $(ostvdevname 1)
4111 echo checking size of lov_objid for ost index $i
4112 LOV_OBJID_SIZE=$(do_facet mds1 "$DEBUGFS -R 'stat lov_objid' $mdsdev 2>/dev/null" | grep ^User | awk '{print $6}')
4113 if [ "$LOV_OBJID_SIZE" != $(lov_objid_size $i) ]; then
4114 error "lov_objid size has to be $(lov_objid_size $i), not $LOV_OBJID_SIZE"
4116 echo ok, lov_objid size is correct: $LOV_OBJID_SIZE
4123 run_test 55 "check lov_objid size"
4126 local server_version=$(lustre_version_code $SINGLEMDS)
4127 local mds_journal_size_orig=$MDSJOURNALSIZE
4132 for num in $(seq 1 $MDSCOUNT); do
4135 add ost1 $(mkfs_opts ost1 $(ostdevname 1)) --index=10000 --reformat \
4136 $(ostdevname 1) $(ostvdevname 1)
4137 add ost2 $(mkfs_opts ost2 $(ostdevname 2)) --index=1000 --reformat \
4138 $(ostdevname 2) $(ostvdevname 2)
4141 start_ost || error "Unable to start first ost (idx 10000)"
4142 start_ost2 || error "Unable to start second ost (idx 1000)"
4143 mount_client $MOUNT || error "Unable to mount client"
4147 if [[ $server_version -ge $(version_code 2.6.54) ]] ||
4148 [[ $server_version -ge $(version_code 2.5.4) &&
4149 $server_version -lt $(version_code 2.5.11) ]]; then
4150 wait_osc_import_state mds ost1 FULL
4151 wait_osc_import_state mds ost2 FULL
4152 $SETSTRIPE --stripe-count=-1 $DIR/$tfile ||
4153 error "Unable to setstripe $DIR/$tfile"
4154 n=$($LFS getstripe --stripe-count $DIR/$tfile)
4155 [ "$n" -eq 2 ] || error "Stripe count not two: $n"
4160 MDSJOURNALSIZE=$mds_journal_size_orig
4163 run_test 56a "check big OST indexes and out-of-index-order start"
4168 umount_client $MOUNT -f || error "unmount client failed"
4177 [ $MDSCOUNT -lt 3 ] && skip "needs >= 3 MDTs" && return
4179 trap cleanup_56b EXIT RETURN ERR
4182 if ! combined_mgs_mds ; then
4187 add mds1 $(mkfs_opts mds1 $(mdsdevname 1)) --index=0 --reformat \
4188 $(mdsdevname 1) $(mdsvdevname 1)
4189 add mds2 $(mkfs_opts mds2 $(mdsdevname 2)) --index=1 --reformat \
4190 $(mdsdevname 2) $(mdsvdevname 2)
4191 add mds3 $(mkfs_opts mds3 $(mdsdevname 3)) --index=1000 --reformat \
4192 $(mdsdevname 3) $(mdsvdevname 3)
4196 start_mdt 1 || error "MDT 1 (idx 0) start failed"
4197 start_mdt 2 || error "MDT 2 (idx 1) start failed"
4198 start_mdt 3 || error "MDT 3 (idx 1000) start failed"
4199 start_ost || error "Unable to start first ost"
4200 start_ost2 || error "Unable to start second ost"
4202 do_nodes $(comma_list $(mdts_nodes)) \
4203 "$LCTL set_param mdt.*.enable_remote_dir=1 \
4204 mdt.*.enable_remote_dir_gid=-1"
4206 mount_client $MOUNT || error "Unable to mount client"
4208 $LFS mkdir -c3 $MOUNT/$tdir || error "failed to make testdir"
4210 echo "This is test file 1!" > $MOUNT/$tdir/$tfile.1 ||
4211 error "failed to make test file 1"
4212 echo "This is test file 2!" > $MOUNT/$tdir/$tfile.2 ||
4213 error "failed to make test file 2"
4214 echo "This is test file 1000!" > $MOUNT/$tdir/$tfile.1000 ||
4215 error "failed to make test file 1000"
4217 rm -rf $MOUNT/$tdir || error "failed to remove testdir"
4219 $LFS mkdir -i1000 $MOUNT/$tdir.1000 ||
4220 error "create remote dir at idx 1000 failed"
4223 echo "=== START lfs df OUTPUT ==="
4225 echo "==== END lfs df OUTPUT ===="
4227 mdtcnt=$(echo -e "$output" | grep $FSNAME-MDT | wc -l)
4228 ostcnt=$(echo -e "$output" | grep $FSNAME-OST | wc -l)
4230 echo "lfs df returned mdt count $mdtcnt and ost count $ostcnt"
4231 [ $mdtcnt -eq 3 ] || error "lfs df returned wrong mdt count"
4232 [ $ostcnt -eq 2 ] || error "lfs df returned wrong ost count"
4234 echo "This is test file 1!" > $MOUNT/$tdir.1000/$tfile.1 ||
4235 error "failed to make test file 1"
4236 echo "This is test file 2!" > $MOUNT/$tdir.1000/$tfile.2 ||
4237 error "failed to make test file 2"
4238 echo "This is test file 1000!" > $MOUNT/$tdir.1000/$tfile.1000 ||
4239 error "failed to make test file 1000"
4240 rm -rf $MOUNT/$tdir.1000 || error "failed to remove remote_dir"
4243 echo "=== START lfs mdts OUTPUT ==="
4245 echo "==== END lfs mdts OUTPUT ===="
4247 echo -e "$output" | grep -v "MDTS:" | awk '{print $1}' |
4248 sed 's/://g' > $TMP/mdts-actual.txt
4249 sort $TMP/mdts-actual.txt -o $TMP/mdts-actual.txt
4251 echo -e "0\n1\n1000" > $TMP/mdts-expected.txt
4253 diff $TMP/mdts-expected.txt $TMP/mdts-actual.txt
4256 rm $TMP/mdts-expected.txt $TMP/mdts-actual.txt
4258 [ $result -eq 0 ] || error "target_obd proc file is incorrect!"
4260 run_test 56b "test target_obd correctness with nonconsecutive MDTs"
4262 test_57a() { # bug 22656
4263 do_rpc_nodes $(facet_active_host ost1) load_modules_local
4264 local NID=$(do_facet ost1 "$LCTL get_param nis" |
4265 tail -1 | awk '{print $1}')
4266 writeconf_or_reformat
4267 [ $(facet_fstype ost1) == zfs ] && import_zpool ost1
4268 do_facet ost1 "$TUNEFS --failnode=$NID `ostdevname 1`" ||
4269 error "tunefs failed"
4271 start_ost && error "OST registration from failnode should fail"
4274 run_test 57a "initial registration from failnode should fail (should return errs)"
4277 do_rpc_nodes $(facet_active_host ost1) load_modules_local
4278 local NID=$(do_facet ost1 "$LCTL get_param nis" |
4279 tail -1 | awk '{print $1}')
4280 writeconf_or_reformat
4281 [ $(facet_fstype ost1) == zfs ] && import_zpool ost1
4282 do_facet ost1 "$TUNEFS --servicenode=$NID `ostdevname 1`" ||
4283 error "tunefs failed"
4285 start_ost || error "OST registration from servicenode should not fail"
4288 run_test 57b "initial registration from servicenode should not fail"
4291 do_facet mgs $LCTL get_param mgs.MGS.live.$FSNAME | grep OST | wc -l
4294 test_58() { # bug 22658
4295 combined_mgs_mds || stop_mgs || error "stopping MGS service failed"
4297 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
4298 createmany -o $DIR/$tdir/$tfile-%d 100
4299 # make sure that OSTs do not cancel llog cookies before we unmount the MDS
4300 #define OBD_FAIL_OBD_LOG_CANCEL_NET 0x601
4301 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0x601"
4302 unlinkmany $DIR/$tdir/$tfile-%d 100
4303 stop_mds || error "Unable to stop MDS"
4305 local MNTDIR=$(facet_mntpt $SINGLEMDS)
4306 local devname=$(mdsdevname ${SINGLEMDS//mds/})
4308 # remove all files from the OBJECTS dir
4309 mount_fstype $SINGLEMDS
4311 do_facet $SINGLEMDS "find $MNTDIR/O/1/d* -type f -delete"
4313 unmount_fstype $SINGLEMDS
4314 # restart MDS with missing llog files
4315 start_mds || error "unable to start MDS"
4316 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0"
4319 run_test 58 "missing llog files must not prevent MDT from mounting"
4322 start_mgsmds >> /dev/null
4323 local C1=$(count_osts)
4324 if [ $C1 -eq 0 ]; then
4325 start_ost >> /dev/null
4329 echo "original ost count: $C1 (expect > 0)"
4330 [ $C1 -gt 0 ] || error "No OSTs in $FSNAME log"
4331 start_mgsmds -o writeconf >> /dev/null || error "MDT start failed"
4332 local C2=$(count_osts)
4333 echo "after mdt writeconf count: $C2 (expect 0)"
4334 [ $C2 -gt 0 ] && error "MDT writeconf should erase OST logs"
4335 echo "OST start without writeconf should fail:"
4336 start_ost >> /dev/null &&
4337 error "OST start without writeconf didn't fail"
4338 echo "OST start with writeconf should succeed:"
4339 start_ost -o writeconf >> /dev/null || error "OST1 start failed"
4340 local C3=$(count_osts)
4341 echo "after ost writeconf count: $C3 (expect 1)"
4342 [ $C3 -eq 1 ] || error "new OST writeconf should add:"
4343 start_ost2 -o writeconf >> /dev/null || error "OST2 start failed"
4344 local C4=$(count_osts)
4345 echo "after ost2 writeconf count: $C4 (expect 2)"
4346 [ $C4 -eq 2 ] || error "OST2 writeconf should add log"
4347 stop_ost2 >> /dev/null
4348 cleanup_nocli >> /dev/null
4349 #writeconf to remove all ost2 traces for subsequent tests
4350 writeconf_or_reformat
4352 run_test 59 "writeconf mount option"
4354 test_60() { # LU-471
4357 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
4358 skip "Only applicable to ldiskfs-based MDTs"
4362 for num in $(seq $MDSCOUNT); do
4363 add mds${num} $(mkfs_opts mds${num} $(mdsdevname $num)) \
4364 --mkfsoptions='\" -E stride=64 -O ^uninit_bg\"' \
4365 --reformat $(mdsdevname $num) $(mdsvdevname $num) ||
4369 dump=$(do_facet $SINGLEMDS dumpe2fs $(mdsdevname 1))
4370 [ ${PIPESTATUS[0]} -eq 0 ] || error "dumpe2fs $(mdsdevname 1) failed"
4372 # MDT default has dirdata feature
4373 echo $dump | grep dirdata > /dev/null || error "dirdata is not set"
4374 # we disable uninit_bg feature
4375 echo $dump | grep uninit_bg > /dev/null && error "uninit_bg is set"
4376 # we set stride extended options
4377 echo $dump | grep stride > /dev/null || error "stride is not set"
4380 run_test 60 "check mkfs.lustre --mkfsoptions -E -O options setting"
4385 [ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.1.53) ] ||
4386 { skip "Need MDS version at least 2.1.53"; return 0; }
4388 if [ $(facet_fstype $SINGLEMDS) == ldiskfs ] &&
4389 ! large_xattr_enabled; then
4392 for num in $(seq $MDSCOUNT); do
4393 do_facet mds${num} $TUNE2FS -O large_xattr \
4394 $(mdsdevname $num) ||
4395 error "tune2fs on mds $num failed"
4399 combined_mgs_mds || stop_mgs || error "stopping MGS service failed"
4400 setup_noconfig || error "setting up the filesystem failed"
4401 client_up || error "starting client failed"
4403 local file=$DIR/$tfile
4404 touch $file || error "touch $file failed"
4406 local large_value="$(generate_string $(max_xattr_size))"
4407 local small_value="bar"
4409 local name="trusted.big"
4410 log "save large xattr $name on $file"
4411 setfattr -n $name -v $large_value $file ||
4412 error "saving $name on $file failed"
4414 local new_value=$(get_xattr_value $name $file)
4415 [[ "$new_value" != "$large_value" ]] &&
4416 error "$name different after saving"
4418 log "shrink value of $name on $file"
4419 setfattr -n $name -v $small_value $file ||
4420 error "shrinking value of $name on $file failed"
4422 new_value=$(get_xattr_value $name $file)
4423 [[ "$new_value" != "$small_value" ]] &&
4424 error "$name different after shrinking"
4426 log "grow value of $name on $file"
4427 setfattr -n $name -v $large_value $file ||
4428 error "growing value of $name on $file failed"
4430 new_value=$(get_xattr_value $name $file)
4431 [[ "$new_value" != "$large_value" ]] &&
4432 error "$name different after growing"
4434 log "check value of $name on $file after remounting MDS"
4436 new_value=$(get_xattr_value $name $file)
4437 [[ "$new_value" != "$large_value" ]] &&
4438 error "$name different after remounting MDS"
4440 log "remove large xattr $name from $file"
4441 setfattr -x $name $file || error "removing $name from $file failed"
4444 stopall || error "stopping for e2fsck run"
4445 for num in $(seq $MDSCOUNT); do
4446 run_e2fsck $(facet_active_host mds$num) \
4447 $(mdsdevname $num) "-y" ||
4448 error "e2fsck MDT$num failed"
4450 setup_noconfig || error "remounting the filesystem failed"
4453 # need to delete this file to avoid problems in other tests
4455 stopall || error "stopping systems to turn off large_xattr"
4457 for num in $(seq $MDSCOUNT); do
4458 do_facet mds${num} $TUNE2FS -O ^large_xattr \
4459 $(mdsdevname $num) ||
4460 error "tune2fs on mds $num failed"
4464 run_test 61 "large xattr"
4467 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
4468 skip "Only applicable to ldiskfs-based MDTs"
4473 local mdsdev=$(mdsdevname 1)
4474 local ostdev=$(ostdevname 1)
4476 [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.2.51) ]] ||
4477 { skip "Need MDS version at least 2.2.51"; return 0; }
4479 echo "disable journal for mds"
4480 do_facet mds1 $TUNE2FS -O ^has_journal $mdsdev || error "tune2fs failed"
4481 start_mds && error "MDT start should fail"
4482 echo "disable journal for ost"
4483 do_facet ost1 $TUNE2FS -O ^has_journal $ostdev || error "tune2fs failed"
4484 start_ost && error "OST start should fail"
4485 cleanup || error "cleanup failed with rc $?"
4488 run_test 62 "start with disabled journal"
4491 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
4492 skip "Only applicable to ldiskfs-based MDTs"
4496 do_rpc_nodes $(facet_active_host $SINGLEMDS) load_module ldiskfs
4497 local inode_slab=$(do_facet $SINGLEMDS "cat /proc/slabinfo" |
4498 awk '/ldiskfs_inode_cache/ { print $5 / $6 }')
4499 if [ -z "$inode_slab" ]; then
4500 skip "ldiskfs module has not been loaded"
4504 echo "$inode_slab ldiskfs inodes per page"
4505 [ "${inode_slab%.*}" -ge "3" ] && return 0
4507 # If kmalloc-128 is also 1 per page - this is a debug kernel
4508 # and so this is not an error.
4509 local kmalloc128=$(do_facet $SINGLEMDS "cat /proc/slabinfo" |
4510 awk '/^(kmalloc|size)-128 / { print $5 / $6 }')
4511 # 32 128-byte chunks in 4k
4512 [ "${kmalloc128%.*}" -lt "32" ] ||
4513 error "ldiskfs inode too big, only $inode_slab objs/page, " \
4514 "kmalloc128 = $kmalloc128 objs/page"
4516 run_test 63 "Verify each page can at least hold 3 ldiskfs inodes"
4519 start_mds || error "unable to start MDS"
4520 start_ost || error "Unable to start OST1"
4521 start_ost2 || error "Unable to start second ost"
4522 mount_client $MOUNT || error "Unable to mount client"
4523 stop_ost2 || error "Unable to stop second ost"
4526 umount_client $MOUNT -f || error “unmount $MOUNT failed”
4527 cleanup_nocli || error "cleanup_nocli failed with $?"
4528 #writeconf to remove all ost2 traces for subsequent tests
4529 writeconf_or_reformat
4531 run_test 64 "check lfs df --lazy "
4533 test_65() { # LU-2237
4534 # Currently, the test is only valid for ldiskfs backend
4535 [ "$(facet_fstype $SINGLEMDS)" != "ldiskfs" ] &&
4536 skip "non-ldiskfs backend" && return
4538 local devname=$(mdsdevname ${SINGLEMDS//mds/})
4539 local brpt=$(facet_mntpt brpt)
4542 if ! do_facet $SINGLEMDS "test -b $devname"; then
4546 stop_mds || error "Unable to stop MDS"
4547 local obj=$(do_facet $SINGLEMDS \
4548 "$DEBUGFS -c -R \\\"stat last_rcvd\\\" $devname" |
4550 if [ -z "$obj" ]; then
4551 # The MDT may be just re-formatted, mount the MDT for the
4552 # first time to guarantee the "last_rcvd" file is there.
4553 start_mds || error "fail to mount the MDS for the first time"
4554 stop_mds || error "Unable to stop MDS"
4557 # remove the "last_rcvd" file
4558 do_facet $SINGLEMDS "mkdir -p $brpt"
4559 do_facet $SINGLEMDS \
4560 "mount -t $(facet_fstype $SINGLEMDS) $opts $devname $brpt"
4561 do_facet $SINGLEMDS "rm -f ${brpt}/last_rcvd"
4562 do_facet $SINGLEMDS "$UMOUNT $brpt"
4564 # restart MDS, the "last_rcvd" file should be recreated.
4565 start_mds || error "fail to restart the MDS"
4566 stop_mds || error "Unable to stop MDS"
4567 obj=$(do_facet $SINGLEMDS \
4568 "$DEBUGFS -c -R \\\"stat last_rcvd\\\" $devname" | grep Inode)
4569 [ -n "$obj" ] || error "fail to re-create the last_rcvd"
4571 run_test 65 "re-create the lost last_rcvd file when server mount"
4574 [[ $(lustre_version_code mgs) -ge $(version_code 2.3.59) ]] ||
4575 { skip "Need MGS version at least 2.3.59"; return 0; }
4578 local OST1_NID=$(do_facet ost1 $LCTL list_nids | head -1)
4579 local MDS_NID=$(do_facet $SINGLEMDS $LCTL list_nids | head -1)
4581 echo "replace_nids should fail if MDS, OSTs and clients are UP"
4582 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 $OST1_NID &&
4583 error "replace_nids fail"
4585 umount_client $MOUNT || error "unmounting client failed"
4586 echo "replace_nids should fail if MDS and OSTs are UP"
4587 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 $OST1_NID &&
4588 error "replace_nids fail"
4590 stop_ost || error "Unable to stop OST1"
4591 echo "replace_nids should fail if MDS is UP"
4592 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 $OST1_NID &&
4593 error "replace_nids fail"
4595 stop_mds || error "stopping mds failed"
4597 if combined_mgs_mds; then
4598 start_mdt 1 "-o nosvc" ||
4599 error "starting mds with nosvc option failed"
4602 echo "command should accept two parameters"
4603 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 &&
4604 error "command should accept two params"
4606 echo "correct device name should be passed"
4607 do_facet mgs $LCTL replace_nids $FSNAME-WRONG0000 $OST1_NID &&
4608 error "wrong devname"
4610 echo "wrong nids list should not destroy the system"
4611 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 "wrong nids list" &&
4614 echo "replace OST nid"
4615 do_facet mgs $LCTL replace_nids $FSNAME-OST0000 $OST1_NID ||
4616 error "replace nids failed"
4618 echo "command should accept two parameters"
4619 do_facet mgs $LCTL replace_nids $FSNAME-MDT0000 &&
4620 error "command should accept two params"
4622 echo "wrong nids list should not destroy the system"
4623 do_facet mgs $LCTL replace_nids $FSNAME-MDT0000 "wrong nids list" &&
4626 echo "replace MDS nid"
4627 do_facet mgs $LCTL replace_nids $FSNAME-MDT0000 $MDS_NID ||
4628 error "replace nids failed"
4630 if ! combined_mgs_mds ; then
4633 stop_mds || error "Unable to stop MDS"
4637 check_mount || error "error after nid replace"
4638 cleanup || error "cleanup failed"
4641 run_test 66 "replace nids"
4643 test_67() { #LU-2950
4644 local legacy="$TMP/legacy_lnet_config"
4645 local new="$TMP/new_routes_test"
4646 local out="$TMP/config_out_file"
4647 local verify="$TMP/conv_verify"
4648 local verify_conf="$TMP/conf_verify"
4650 # Create the legacy file that will be run through the
4651 # lustre_routes_conversion script
4652 cat <<- LEGACY_LNET_CONFIG > $legacy
4653 tcp1 23 192.168.213.1@tcp:1; tcp5 34 193.30.4.3@tcp:4;
4654 tcp2 54 10.1.3.2@tcp;
4655 tcp3 10.3.4.3@tcp:3;
4659 # Create the verification file to verify the output of
4660 # lustre_routes_conversion script against.
4661 cat <<- VERIFY_LNET_CONFIG > $verify
4662 tcp1: { gateway: 192.168.213.1@tcp, hop: 23, priority: 1 }
4663 tcp5: { gateway: 193.30.4.3@tcp, hop: 34, priority: 4 }
4664 tcp2: { gateway: 10.1.3.2@tcp, hop: 54 }
4665 tcp3: { gateway: 10.3.4.3@tcp, priority: 3 }
4666 tcp4: { gateway: 10.3.3.4@tcp }
4669 # Create the verification file to verify the output of
4670 # lustre_routes_config script against
4671 cat <<- VERIFY_LNET_CONFIG > $verify_conf
4672 lctl --net tcp1 add_route 192.168.213.1@tcp 23 1
4673 lctl --net tcp5 add_route 193.30.4.3@tcp 34 4
4674 lctl --net tcp2 add_route 10.1.3.2@tcp 54 4
4675 lctl --net tcp3 add_route 10.3.4.3@tcp 1 3
4676 lctl --net tcp4 add_route 10.3.3.4@tcp 1 3
4679 lustre_routes_conversion $legacy $new > /dev/null
4680 if [ -f $new ]; then
4681 # verify the conversion output
4682 cmp -s $new $verify > /dev/null
4683 if [ $? -eq 1 ]; then
4684 error "routes conversion failed"
4687 lustre_routes_config --dry-run --verbose $new > $out
4688 # check that the script succeeded
4689 cmp -s $out $verify_conf > /dev/null
4690 if [ $? -eq 1 ]; then
4691 error "routes config failed"
4694 error "routes conversion test failed"
4696 # remove generated files
4697 rm -f $new $legacy $verify $verify_conf $out
4699 run_test 67 "test routes conversion and configuration"
4707 [ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.4.53) ] ||
4708 { skip "Need MDS version at least 2.4.53"; return 0; }
4710 umount_client $MOUNT || error "umount client failed"
4712 start_mdt 1 || error "MDT start failed"
4713 start_ost || error "Unable to start OST1"
4715 # START-END - the sequences we'll be reserving
4716 START=$(do_facet $SINGLEMDS \
4717 $LCTL get_param -n seq.ctl*.space | awk -F'[[ ]' '{print $2}')
4718 END=$((START + (1 << 30)))
4719 do_facet $SINGLEMDS \
4720 $LCTL set_param seq.ctl*.fldb="[$START-$END\):0:mdt"
4722 # reset the sequences MDT0000 has already assigned
4723 do_facet $SINGLEMDS \
4724 $LCTL set_param seq.srv*MDT0000.space=clear
4726 # remount to let the client allocate new sequence
4727 mount_client $MOUNT || error "mount client failed"
4729 touch $DIR/$tfile || error "touch $DIR/$tfile failed"
4730 do_facet $SINGLEMDS \
4731 $LCTL get_param seq.srv*MDT0000.space
4732 $LFS path2fid $DIR/$tfile
4734 local old_ifs="$IFS"
4736 fid=($($LFS path2fid $DIR/$tfile))
4740 if [[ $seq < $END ]]; then
4741 error "used reserved sequence $seq?"
4743 cleanup || error "cleanup failed with $?"
4745 run_test 68 "be able to reserve specific sequences in FLDB"
4747 # Test 69: is about the total number of objects ever created on an OST.
4748 # so that when it is reformatted the normal MDS->OST orphan recovery won't
4749 # just "precreate" the missing objects. In the past it might try to recreate
4750 # millions of objects after an OST was reformatted
4752 local server_version=$(lustre_version_code $SINGLEMDS)
4754 [[ $server_version -lt $(version_code 2.4.2) ]] &&
4755 skip "Need MDS version at least 2.4.2" && return
4757 [[ $server_version -ge $(version_code 2.4.50) ]] &&
4758 [[ $server_version -lt $(version_code 2.5.0) ]] &&
4759 skip "Need MDS version at least 2.5.0" && return
4762 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
4764 # use OST0000 since it probably has the most creations
4765 local OSTNAME=$(ostname_from_index 0)
4766 local mdtosc_proc1=$(get_mdtosc_proc_path mds1 $OSTNAME)
4767 local last_id=$(do_facet mds1 $LCTL get_param -n \
4768 osc.$mdtosc_proc1.prealloc_last_id)
4770 # Want to have OST LAST_ID over 5 * OST_MAX_PRECREATE to
4771 # verify that the LAST_ID recovery is working properly. If
4772 # not, then the OST will refuse to allow the MDS connect
4773 # because the LAST_ID value is too different from the MDS
4774 #define OST_MAX_PRECREATE=20000
4775 local ost_max_pre=20000
4776 local num_create=$(( ost_max_pre * 5 + 1 - last_id))
4778 # If the LAST_ID is already over 5 * OST_MAX_PRECREATE, we don't
4779 # need to create any files. So, skip this section.
4780 if [ $num_create -gt 0 ]; then
4781 # Check the number of inodes available on OST0
4783 local ifree=$($LFS df -i $MOUNT | awk '/OST0000/ { print $4 }')
4784 log "On OST0, $ifree inodes available. Want $num_create."
4786 $SETSTRIPE -i 0 $DIR/$tdir ||
4787 error "$SETSTRIPE -i 0 $DIR/$tdir failed"
4788 if [ $ifree -lt 10000 ]; then
4789 files=$(( ifree - 50 ))
4794 local j=$((num_create / files + 1))
4795 for i in $(seq 1 $j); do
4796 createmany -o $DIR/$tdir/$tfile-$i- $files ||
4797 error "createmany fail create $files files: $?"
4798 unlinkmany $DIR/$tdir/$tfile-$i- $files ||
4799 error "unlinkmany failed unlink $files files"
4803 # delete all of the files with objects on OST0 so the
4804 # filesystem is not inconsistent later on
4805 $LFS find $MOUNT --ost 0 -print0 | xargs -0 rm
4807 umount_client $MOUNT || error "umount client failed"
4808 stop_ost || error "OST0 stop failure"
4809 add ost1 $(mkfs_opts ost1 $(ostdevname 1)) --reformat --replace \
4810 $(ostdevname 1) $(ostvdevname 1) ||
4811 error "reformat and replace $ostdev failed"
4812 start_ost || error "OST0 restart failure"
4813 wait_osc_import_state mds ost FULL
4815 mount_client $MOUNT || error "mount client failed"
4816 touch $DIR/$tdir/$tfile-last || error "create file after reformat"
4817 local idx=$($GETSTRIPE -i $DIR/$tdir/$tfile-last)
4818 [ $idx -ne 0 ] && error "$DIR/$tdir/$tfile-last on $idx not 0" || true
4820 local iused=$($LFS df -i $MOUNT | awk '/OST0000/ { print $3 }')
4821 log "On OST0, $iused used inodes"
4822 [ $iused -ge $((ost_max_pre/2 + 1000)) ] &&
4823 error "OST replacement created too many inodes; $iused"
4824 cleanup || error "cleanup failed with $?"
4826 run_test 69 "replace an OST with the same index"
4829 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
4832 cleanup || error "cleanup failed with $?"
4834 start_mdt 1 || error "MDT0 start fail"
4836 start_ost || error "OST0 start fail"
4837 for num in $(seq 2 $MDSCOUNT); do
4838 start_mdt $num || return
4841 mount_client $MOUNT || error "mount client fails"
4843 mkdir $DIR/$tdir || error "create $DIR/$tdir failed"
4845 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
4846 error "create remote dir fail"
4848 rm -rf $DIR/$tdir || error "delete dir fail"
4849 cleanup || error "cleanup failed with $?"
4851 run_test 70a "start MDT0, then OST, then MDT1"
4854 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
4857 start_ost || error "OST0 start fail"
4859 start_mds || error "MDS start fail"
4861 mount_client $MOUNT || error "mount client fails"
4863 mkdir $DIR/$tdir || error "create $DIR/$tdir failed"
4865 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
4866 error "create remote dir fail"
4868 rm -rf $DIR/$tdir || error "delete dir fail"
4870 cleanup || error "cleanup failed with $?"
4872 run_test 70b "start OST, MDT1, MDT0"
4875 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
4878 start_mds || error "MDS start fail"
4879 start_ost || error "OST0 start fail"
4881 mount_client $MOUNT || error "mount client fails"
4882 stop_mdt 1 || error "MDT1 start fail"
4884 local mdc_for_mdt1=$($LCTL dl | grep MDT0000-mdc | awk '{print $4}')
4885 echo "deactivate $mdc_for_mdt1"
4886 $LCTL --device $mdc_for_mdt1 deactivate ||
4887 error "set $mdc_for_mdt1 deactivate failed"
4889 mkdir $DIR/$tdir && error "mkdir succeed"
4891 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir &&
4892 error "create remote dir succeed"
4894 cleanup || error "cleanup failed with $?"
4896 run_test 70c "stop MDT0, mkdir fail, create remote dir fail"
4899 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
4902 start_mds || error "MDS start fail"
4903 start_ost || error "OST0 start fail"
4905 mount_client $MOUNT || error "mount client fails"
4907 stop_mdt 2 || error "MDT1 start fail"
4909 local mdc_for_mdt2=$($LCTL dl | grep MDT0001-mdc |
4911 echo "deactivate $mdc_for_mdt2"
4912 $LCTL --device $mdc_for_mdt2 deactivate ||
4913 error "set $mdc_for_mdt2 deactivate failed"
4915 mkdir $DIR/$tdir || error "mkdir fail"
4916 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir &&
4917 error "create remote dir succeed"
4919 rm -rf $DIR/$tdir || error "delete dir fail"
4921 cleanup || error "cleanup failed with $?"
4923 run_test 70d "stop MDT1, mkdir succeed, create remote dir fail"
4926 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
4928 [ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.7.62) ] ||
4929 { skip "Need MDS version at least 2.7.62"; return 0; }
4931 cleanup || error "cleanup failed with $?"
4933 local mdsdev=$(mdsdevname 1)
4934 local ostdev=$(ostdevname 1)
4935 local mdsvdev=$(mdsvdevname 1)
4936 local ostvdev=$(ostvdevname 1)
4937 local opts_mds="$(mkfs_opts mds1 $mdsdev) --reformat $mdsdev $mdsvdev"
4938 local opts_ost="$(mkfs_opts ost1 $ostdev) --reformat $ostdev $ostvdev"
4940 add mds1 $opts_mds || error "add mds1 failed"
4941 start_mdt 1 || error "start mdt1 failed"
4942 add ost1 $opts_ost || error "add ost1 failed"
4943 start_ost || error "start ost failed"
4944 mount_client $MOUNT > /dev/null || error "mount client $MOUNT failed"
4946 local soc=$(do_facet mds1 "$LCTL get_param -n \
4947 mdt.*MDT0000.sync_lock_cancel")
4948 [ $soc == "never" ] || error "SoC enabled on single MDS"
4950 for i in $(seq 2 $MDSCOUNT); do
4951 mdsdev=$(mdsdevname $i)
4952 mdsvdev=$(mdsvdevname $i)
4953 opts_mds="$(mkfs_opts mds$i $mdsdev) --reformat $mdsdev \
4955 add mds$i $opts_mds || error "add mds$i failed"
4956 start_mdt $i || error "start mdt$i fail"
4959 wait_dne_interconnect
4961 for i in $(seq $MDSCOUNT); do
4962 soc=$(do_facet mds$i "$LCTL get_param -n \
4963 mdt.*MDT000$((i - 1)).sync_lock_cancel")
4964 [ $soc == "blocking" ] || error "SoC not enabled on DNE"
4967 for i in $(seq 2 $MDSCOUNT); do
4968 stop_mdt $i || error "stop mdt$i fail"
4970 soc=$(do_facet mds1 "$LCTL get_param -n \
4971 mdt.*MDT0000.sync_lock_cancel")
4972 [ $soc == "never" ] || error "SoC enabled on single MDS"
4974 cleanup || error "cleanup failed with $?"
4976 run_test 70e "Sync-on-Cancel will be enabled by default on DNE"
4979 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
4980 if combined_mgs_mds; then
4981 skip "needs separate MGS/MDT" && return
4985 start_mdt 1 || error "MDT0 start fail"
4986 start_ost || error "OST0 start fail"
4987 for num in $(seq 2 $MDSCOUNT); do
4988 start_mdt $num || return
4991 start_ost2 || error "OST1 start fail"
4993 mount_client $MOUNT || error "mount client fails"
4995 mkdir $DIR/$tdir || error "mkdir fail"
4996 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
4997 error "create remote dir succeed"
4999 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
5000 rm -rf $DIR/$tdir || error "delete dir fail"
5002 umount_client $MOUNT || error "umount_client failed"
5003 stop_mds || error "MDS stop fail"
5004 stop_ost || error "OST0 stop fail"
5005 stop_ost2 || error "OST1 stop fail"
5007 run_test 71a "start MDT0 OST0, MDT1, OST1"
5010 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
5011 if combined_mgs_mds; then
5012 skip "needs separate MGS/MDT" && return
5016 for num in $(seq 2 $MDSCOUNT); do
5017 start_mdt $num || return
5019 start_ost || error "OST0 start fail"
5020 start_mdt 1 || error "MDT0 start fail"
5021 start_ost2 || error "OST1 start fail"
5023 mount_client $MOUNT || error "mount client fails"
5025 mkdir $DIR/$tdir || error "mkdir fail"
5026 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
5027 error "create remote dir succeed"
5029 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
5030 rm -rf $DIR/$tdir || error "delete dir fail"
5032 umount_client $MOUNT || error "umount_client failed"
5033 stop_mds || error "MDT0 stop fail"
5034 stop_ost || error "OST0 stop fail"
5035 stop_ost2 || error "OST1 stop fail"
5037 run_test 71b "start MDT1, OST0, MDT0, OST1"
5040 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
5041 if combined_mgs_mds; then
5042 skip "needs separate MGS/MDT" && return
5046 start_ost || error "OST0 start fail"
5047 start_ost2 || error "OST1 start fail"
5048 for num in $(seq 2 $MDSCOUNT); do
5049 start_mdt $num || return
5051 start_mdt 1 || error "MDT0 start fail"
5053 mount_client $MOUNT || error "mount client fails"
5055 mkdir $DIR/$tdir || error "mkdir fail"
5056 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
5057 error "create remote dir succeed"
5059 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
5060 rm -rf $DIR/$tdir || error "delete dir fail"
5062 umount_client $MOUNT || error "umount_client failed"
5063 stop_mds || error "MDS stop fail"
5064 stop_ost || error "OST0 stop fail"
5065 stop_ost2 || error "OST1 stop fail"
5068 run_test 71c "start OST0, OST1, MDT1, MDT0"
5071 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
5072 if combined_mgs_mds; then
5073 skip "needs separate MGS/MDT" && return
5077 start_ost || error "OST0 start fail"
5078 for num in $(seq 2 $MDSCOUNT); do
5079 start_mdt $num || return
5081 start_mdt 1 || error "MDT0 start fail"
5082 start_ost2 || error "OST1 start fail"
5084 mount_client $MOUNT || error "mount client fails"
5086 mkdir $DIR/$tdir || error "mkdir fail"
5087 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
5088 error "create remote dir succeed"
5090 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
5091 rm -rf $DIR/$tdir || error "delete dir fail"
5093 umount_client $MOUNT || error "umount_client failed"
5094 stop_mds || error "MDS stop fail"
5095 stop_ost || error "OST0 stop fail"
5096 stop_ost2 || error "OST1 stop fail"
5099 run_test 71d "start OST0, MDT1, MDT0, OST1"
5102 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
5103 if combined_mgs_mds; then
5104 skip "needs separate MGS/MDT" && return
5108 start_ost || error "OST0 start fail"
5109 for num in $(seq 2 $MDSCOUNT); do
5110 start_mdt $num || return
5112 start_ost2 || error "OST1 start fail"
5113 start_mdt 1 || error "MDT0 start fail"
5115 mount_client $MOUNT || error "mount client fails"
5117 mkdir $DIR/$tdir || error "mkdir fail"
5118 $LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
5119 error "create remote dir succeed"
5121 mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
5122 rm -rf $DIR/$tdir || error "delete dir fail"
5124 umount_client $MOUNT || error "umount_client failed"
5125 stop_mds || error "MDS stop fail"
5126 stop_ost || error "OST0 stop fail"
5127 stop_ost2 || error "OST1 stop fail"
5130 run_test 71e "start OST0, MDT1, OST1, MDT0"
5132 test_72() { #LU-2634
5133 local mdsdev=$(mdsdevname 1)
5134 local ostdev=$(ostdevname 1)
5135 local cmd="$E2FSCK -fnvd $mdsdev"
5139 [ "$(facet_fstype $SINGLEMDS)" != "ldiskfs" ] &&
5140 skip "ldiskfs only test" && return
5142 if combined_mgs_mds; then
5143 add_options='--reformat'
5145 add_options='--reformat --replace'
5148 #tune MDT with "-O extents"
5150 for num in $(seq $MDSCOUNT); do
5151 add mds${num} $(mkfs_opts mds$num $(mdsdevname $num)) \
5152 $add_options $(mdsdevname $num) $(mdsvdevname $num) ||
5153 error "add mds $num failed"
5154 do_facet mds${num} "$TUNE2FS -O extents $(mdsdevname $num)" ||
5155 error "$TUNE2FS failed on mds${num}"
5158 add ost1 $(mkfs_opts ost1 $ostdev) $add_options $ostdev ||
5159 error "add $ostdev failed"
5160 start_mds || error "start mds failed"
5161 start_ost || error "start ost failed"
5162 mount_client $MOUNT || error "mount client failed"
5164 #create some short symlinks
5165 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
5166 createmany -o $DIR/$tdir/$tfile-%d $fn
5167 echo "create $fn short symlinks"
5168 for i in $(seq -w 1 $fn); do
5169 ln -s $DIR/$tdir/$tfile-$i $MOUNT/$tfile-$i
5174 umount_client $MOUNT || error "umount client failed"
5175 stop_mds || error "stop mds failed"
5176 stop_ost || error "stop ost failed"
5179 run_e2fsck $(facet_active_host $SINGLEMDS) $mdsdev "-n"
5181 run_test 72 "test fast symlink with extents flag enabled"
5183 test_73() { #LU-3006
5184 [ $(facet_fstype ost1) == zfs ] && import_zpool ost1
5185 do_facet ost1 "$TUNEFS --failnode=1.2.3.4@$NETTYPE $(ostdevname 1)" ||
5186 error "1st tunefs failed"
5187 start_mgsmds || error "start mds failed"
5188 start_ost || error "start ost failed"
5189 mount_client $MOUNT || error "mount client failed"
5190 $LCTL get_param -n osc.*OST0000-osc-[^M]*.import | grep failover_nids |
5191 grep 1.2.3.4@$NETTYPE || error "failover nids haven't changed"
5192 umount_client $MOUNT || error "umount client failed"
5196 run_test 73 "failnode to update from mountdata properly"
5198 test_75() { # LU-2374
5199 [[ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.4.1) ]] &&
5200 skip "Need MDS version at least 2.4.1" && return
5203 local opts_mds="$(mkfs_opts mds1 $(mdsdevname 1)) \
5204 --reformat $(mdsdevname 1) $(mdsvdevname 1)"
5205 local opts_ost="$(mkfs_opts ost1 $(ostdevname 1)) \
5206 --reformat $(ostdevname 1) $(ostvdevname 1)"
5208 #check with default parameters
5209 add mds1 $opts_mds || error "add mds1 failed for default params"
5210 add ost1 $opts_ost || error "add ost1 failed for default params"
5212 opts_mds=$(echo $opts_mds | sed -e "s/--mdt//")
5213 opts_mds=$(echo $opts_mds |
5214 sed -e "s/--index=$index/--index=$index --mdt/")
5215 opts_ost=$(echo $opts_ost | sed -e "s/--ost//")
5216 opts_ost=$(echo $opts_ost |
5217 sed -e "s/--index=$index/--index=$index --ost/")
5219 add mds1 $opts_mds || error "add mds1 failed for new params"
5220 add ost1 $opts_ost || error "add ost1 failed for new params"
5223 run_test 75 "The order of --index should be irrelevant"
5226 [[ $(lustre_version_code mgs) -ge $(version_code 2.4.52) ]] ||
5227 { skip "Need MDS version at least 2.4.52" && return 0; }
5229 local MDMB_PARAM="osc.*.max_dirty_mb"
5230 echo "Change MGS params"
5231 local MAX_DIRTY_MB=$($LCTL get_param -n $MDMB_PARAM |
5233 echo "max_dirty_mb: $MAX_DIRTY_MB"
5234 local NEW_MAX_DIRTY_MB=$((MAX_DIRTY_MB + MAX_DIRTY_MB))
5235 echo "new_max_dirty_mb: $NEW_MAX_DIRTY_MB"
5236 do_facet mgs $LCTL set_param -P $MDMB_PARAM=$NEW_MAX_DIRTY_MB
5237 wait_update $HOSTNAME "$LCTL get_param -n $MDMB_PARAM |
5238 head -1" $NEW_MAX_DIRTY_MB
5239 MAX_DIRTY_MB=$($LCTL get_param -n $MDMB_PARAM | head -1)
5240 echo "$MAX_DIRTY_MB"
5241 [ $MAX_DIRTY_MB = $NEW_MAX_DIRTY_MB ] ||
5242 error "error while apply max_dirty_mb"
5244 echo "Check the value is stored after remount"
5247 wait_update $HOSTNAME "$LCTL get_param -n $MDMB_PARAM |
5248 head -1" $NEW_MAX_DIRTY_MB
5249 MAX_DIRTY_MB=$($LCTL get_param -n $MDMB_PARAM | head -1)
5250 [ $MAX_DIRTY_MB = $NEW_MAX_DIRTY_MB ] ||
5251 error "max_dirty_mb is not saved after remount"
5253 echo "Change OST params"
5254 CLIENT_PARAM="obdfilter.*.client_cache_count"
5255 local CLIENT_CACHE_COUNT
5256 CLIENT_CACHE_COUNT=$(do_facet ost1 $LCTL get_param -n $CLIENT_PARAM |
5258 echo "client_cache_count: $CLIENT_CACHE_COUNT"
5259 NEW_CLIENT_CACHE_COUNT=$((CLIENT_CACHE_COUNT+CLIENT_CACHE_COUNT))
5260 echo "new_client_cache_count: $NEW_CLIENT_CACHE_COUNT"
5261 do_facet mgs $LCTL set_param -P $CLIENT_PARAM=$NEW_CLIENT_CACHE_COUNT
5262 wait_update $(facet_host ost1) "$LCTL get_param -n $CLIENT_PARAM |
5263 head -1" $NEW_CLIENT_CACHE_COUNT
5264 CLIENT_CACHE_COUNT=$(do_facet ost1 $LCTL get_param -n $CLIENT_PARAM |
5266 echo "$CLIENT_CACHE_COUNT"
5267 [ $CLIENT_CACHE_COUNT = $NEW_CLIENT_CACHE_COUNT ] ||
5268 error "error while apply client_cache_count"
5270 echo "Check the value is stored after remount"
5273 wait_update $(facet_host ost1) "$LCTL get_param -n $CLIENT_PARAM |
5274 head -1" $NEW_CLIENT_CACHE_COUNT
5275 CLIENT_CACHE_COUNT=$(do_facet ost1 $LCTL get_param -n $CLIENT_PARAM |
5277 echo "$CLIENT_CACHE_COUNT"
5278 [ $CLIENT_CACHE_COUNT = $NEW_CLIENT_CACHE_COUNT ] ||
5279 error "client_cache_count is not saved after remount"
5282 run_test 76a "set permanent params set_param -P"
5284 test_76b() { # LU-4783
5285 [[ $(lustre_version_code mgs) -ge $(version_code 2.5.57) ]] ||
5286 { skip "Need MGS version at least 2.5.57" && return 0; }
5289 do_facet mgs $LCTL get_param mgs.MGS.live.params ||
5290 error "start params log failed"
5293 run_test 76b "verify params log setup correctly"
5296 [[ $(lustre_version_code mgs) -ge $(version_code 2.8.54) ]] ||
5297 { skip "Need MDS version at least 2.4.52" && return 0; }
5299 local MASK_PARAM="mdd.*.changelog_mask"
5300 echo "Change changelog_mask"
5301 do_facet mgs $LCTL set_param -P $MASK_PARAM=-CLOSE ||
5302 error "Can't change changlog_mask"
5303 wait_update $(facet_host mds) "$LCTL get_param -n $MASK_PARAM |
5306 echo "Check the value is stored after mds remount"
5307 stop_mds || error "Failed to stop MDS"
5308 start_mds || error "Failed to start MDS"
5309 local CHANGELOG_MASK=$(do_facet mgs $LCTL get_param -n $MASK_PARAM)
5310 echo $CHANGELOG_MASK | grep CLOSE > /dev/null &&
5311 error "changelog_mask is not changed"
5315 run_test 76c "verify changelog_mask is applied with set_param -P"
5317 test_77() { # LU-3445
5318 local server_version=$(lustre_version_code $SINGLEMDS)
5319 [[ $server_version -ge $(version_code 2.8.55) ]] ||
5320 { skip "Need MDS version 2.8.55+ "; return; }
5322 if [[ -z "$fs2ost_DEV" || -z "$fs2mds_DEV" ]]; then
5323 is_blkdev $SINGLEMDS $(mdsdevname ${SINGLEMDS//mds/}) &&
5324 skip_env "mixed loopback and real device not working" && return
5327 local fs2mdsdev=$(mdsdevname 1_2)
5328 local fs2ostdev=$(ostdevname 1_2)
5329 local fs2mdsvdev=$(mdsvdevname 1_2)
5330 local fs2ostvdev=$(ostvdevname 1_2)
5331 local fsname=test1234
5333 local failnid="$(h2$NETTYPE 1.2.3.4),$(h2$NETTYPE 4.3.2.1)"
5335 add fs2mds $(mkfs_opts mds1 $fs2mdsdev) --mgs --fsname=$fsname \
5336 --reformat $fs2mdsdev $fs2mdsvdev || error "add fs2mds failed"
5337 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS && trap cleanup_fs2 EXIT INT ||
5338 error "start fs2mds failed"
5340 mgsnid=$(do_facet fs2mds $LCTL list_nids | xargs | tr ' ' ,)
5341 mgsnid="$mgsnid,$mgsnid:$mgsnid"
5343 add fs2ost $(mkfs_opts ost1 $fs2ostdev) --mgsnode=$mgsnid \
5344 --failnode=$failnid --fsname=$fsname \
5345 --reformat $fs2ostdev $fs2ostvdev ||
5346 error "add fs2ost failed"
5347 start fs2ost $fs2ostdev $OST_MOUNT_OPTS || error "start fs2ost failed"
5349 mkdir -p $MOUNT2 || error "mkdir $MOUNT2 failed"
5350 $MOUNT_CMD $mgsnid:/$fsname $MOUNT2 || error "mount $MOUNT2 failed"
5351 DIR=$MOUNT2 MOUNT=$MOUNT2 check_mount || error "check $MOUNT2 failed"
5354 run_test 77 "comma-separated MGS NIDs and failover node NIDs"
5357 [[ $(facet_fstype $SINGLEMDS) != ldiskfs ||
5358 $(facet_fstype ost1) != ldiskfs ]] &&
5359 skip "only applicable to ldiskfs-based MDTs and OSTs" && return
5361 # reformat the Lustre filesystem with a smaller size
5362 local saved_MDSCOUNT=$MDSCOUNT
5363 local saved_MDSSIZE=$MDSSIZE
5364 local saved_OSTCOUNT=$OSTCOUNT
5365 local saved_OSTSIZE=$OSTSIZE
5368 MDSSIZE=$((MDSSIZE - 20000))
5369 OSTSIZE=$((OSTSIZE - 20000))
5370 reformat || error "(1) reformat Lustre filesystem failed"
5371 MDSSIZE=$saved_MDSSIZE
5372 OSTSIZE=$saved_OSTSIZE
5374 # mount the Lustre filesystem
5375 setup_noconfig || error "(2) setup Lustre filesystem failed"
5378 log "create test files"
5383 mkdir $MOUNT/$tdir || error "(3) mkdir $MOUNT/$tdir failed"
5385 for i in $(seq $num_files); do
5386 file=$MOUNT/$tdir/$tfile-$i
5387 dd if=/dev/urandom of=$file count=1 bs=1M || {
5388 $LCTL get_param osc.*.cur*grant*
5389 $LFS df; $LFS df -i;
5390 # stop creating files if there is no more space
5391 if [ ! -e $file ]; then
5392 num_files=$((i - 1))
5396 $LFS getstripe -v $file
5397 local ost_idx=$(LFS getstripe -i $file)
5398 do_facet ost$((ost_idx + 1)) \
5399 $LCTL get_param obdfilter.*.*grant*
5400 error "(4) create $file failed"
5404 # unmount the Lustre filesystem
5405 cleanup || error "(5) cleanup Lustre filesystem failed"
5407 # run e2fsck on the MDT and OST devices
5408 local mds_host=$(facet_active_host $SINGLEMDS)
5409 local ost_host=$(facet_active_host ost1)
5410 local mds_dev=$(mdsdevname ${SINGLEMDS//mds/})
5411 local ost_dev=$(ostdevname 1)
5413 run_e2fsck $mds_host $mds_dev "-y"
5414 run_e2fsck $ost_host $ost_dev "-y"
5416 # get the original block count of the MDT and OST filesystems
5417 local mds_orig_blks=$(get_block_count $SINGLEMDS $mds_dev)
5418 local ost_orig_blks=$(get_block_count ost1 $ost_dev)
5420 # expand the MDT and OST filesystems to the device size
5421 run_resize2fs $SINGLEMDS $mds_dev "" || error "expand $SINGLEMDS failed"
5422 run_resize2fs ost1 $ost_dev "" || error "expand ost1 failed"
5424 # run e2fsck on the MDT and OST devices again
5425 run_e2fsck $mds_host $mds_dev "-y"
5426 run_e2fsck $ost_host $ost_dev "-y"
5428 # mount the Lustre filesystem
5432 log "check files after expanding the MDT and OST filesystems"
5433 for i in $(seq $num_files); do
5434 file=$MOUNT/$tdir/$tfile-$i
5435 $CHECKSTAT -t file -s 1048576 $file ||
5436 error "(6) checkstat $file failed"
5440 log "create more files after expanding the MDT and OST filesystems"
5441 for i in $(seq $((num_files + 1)) $((num_files + 10))); do
5442 file=$MOUNT/$tdir/$tfile-$i
5443 dd if=/dev/urandom of=$file count=1 bs=1M ||
5444 error "(7) create $file failed"
5447 # unmount the Lustre filesystem
5448 cleanup || error "(8) cleanup Lustre filesystem failed"
5450 # run e2fsck on the MDT and OST devices
5451 run_e2fsck $mds_host $mds_dev "-y"
5452 run_e2fsck $ost_host $ost_dev "-y"
5454 # get the maximum block count of the MDT and OST filesystems
5455 local mds_max_blks=$(get_block_count $SINGLEMDS $mds_dev)
5456 local ost_max_blks=$(get_block_count ost1 $ost_dev)
5458 # get the minimum block count of the MDT and OST filesystems
5459 local mds_min_blks=$(run_resize2fs $SINGLEMDS $mds_dev "" "-P" 2>&1 |
5460 grep minimum | sed -e 's/^.*filesystem: //g')
5461 local ost_min_blks=$(run_resize2fs ost1 $ost_dev "" "-P" 2>&1 |
5462 grep minimum | sed -e 's/^.*filesystem: //g')
5464 # shrink the MDT and OST filesystems to a smaller size
5468 if [[ $mds_max_blks -gt $mds_min_blks &&
5469 $mds_max_blks -gt $mds_orig_blks ]]; then
5470 [[ $mds_orig_blks -gt $mds_min_blks ]] &&
5471 base_blks=$mds_orig_blks || base_blks=$mds_min_blks
5472 new_blks=$(( (mds_max_blks - base_blks) / 2 + base_blks ))
5473 run_resize2fs $SINGLEMDS $mds_dev $new_blks ||
5474 error "shrink $SINGLEMDS to $new_blks failed"
5478 if [[ $ost_max_blks -gt $ost_min_blks &&
5479 $ost_max_blks -gt $ost_orig_blks ]]; then
5480 [[ $ost_orig_blks -gt $ost_min_blks ]] &&
5481 base_blks=$ost_orig_blks || base_blks=$ost_min_blks
5482 new_blks=$(( (ost_max_blks - base_blks) / 2 + base_blks ))
5483 run_resize2fs ost1 $ost_dev $new_blks ||
5484 error "shrink ost1 to $new_blks failed"
5488 # check whether the MDT or OST filesystem was shrunk or not
5490 combined_mgs_mds || stop_mgs || error "(9) stop mgs failed"
5491 reformat || error "(10) reformat Lustre filesystem failed"
5495 # run e2fsck on the MDT and OST devices again
5496 run_e2fsck $mds_host $mds_dev "-y"
5497 run_e2fsck $ost_host $ost_dev "-y"
5499 # mount the Lustre filesystem again
5503 log "check files after shrinking the MDT and OST filesystems"
5504 for i in $(seq $((num_files + 10))); do
5505 file=$MOUNT/$tdir/$tfile-$i
5506 $CHECKSTAT -t file -s 1048576 $file ||
5507 error "(11) checkstat $file failed"
5510 # unmount and reformat the Lustre filesystem
5511 cleanup || error "(12) cleanup Lustre filesystem failed"
5512 combined_mgs_mds || stop_mgs || error "(13) stop mgs failed"
5514 MDSCOUNT=$saved_MDSCOUNT
5515 OSTCOUNT=$saved_OSTCOUNT
5516 reformat || error "(14) reformat Lustre filesystem failed"
5518 run_test 78 "run resize2fs on MDT and OST filesystems"
5520 test_79() { # LU-4227
5521 [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.5.59) ]] ||
5522 { skip "Need MDS version at least 2.5.59"; return 0; }
5524 local mdsdev1=$(mdsdevname 1)
5525 local mdsvdev1=$(mdsvdevname 1)
5526 local mdsdev2=$(mdsdevname 2)
5527 local mdsvdev2=$(mdsvdevname 2)
5528 local ostdev1=$(ostdevname 1)
5529 local ostvdev1=$(ostvdevname 1)
5530 local opts_mds1="$(mkfs_opts mds1 $mdsdev1) --reformat"
5531 local opts_mds2="$(mkfs_opts mds2 $mdsdev2) --reformat"
5532 local opts_ost1="$(mkfs_opts ost1 $ostdev1) --reformat"
5535 # remove --mgs/--mgsnode from mkfs.lustre options
5536 opts_mds1=$(echo $opts_mds1 | sed -e "s/--mgs//")
5538 mgsnode_opt=$(echo $opts_mds2 |
5539 awk '{ for ( i = 1; i < NF; i++ )
5540 if ( $i ~ "--mgsnode" ) { print $i; break } }')
5541 [ -n $mgsnode_opt ] &&
5542 opts_mds2=$(echo $opts_mds2 | sed -e "s/$mgsnode_opt//")
5544 mgsnode_opt=$(echo $opts_ost1 |
5545 awk '{ for ( i = 1; i < NF; i++ )
5546 if ( $i ~ "--mgsnode" ) { print $i; break } }')
5547 [ -n $mgsnode_opt ] &&
5548 opts_ost1=$(echo $opts_ost1 | sed -e "s/$mgsnode_opt//")
5550 # -MGS, format a mdt without --mgs option
5551 add mds1 $opts_mds1 $mdsdev1 $mdsvdev1 &&
5552 error "Must specify --mgs when formatting mdt combined with mgs"
5554 # +MGS, format a mdt/ost without --mgsnode option
5555 add mds1 $(mkfs_opts mds1 $mdsdev1) --reformat $mdsdev1 $mdsvdev1 \
5556 > /dev/null || error "start mds1 failed"
5557 add mds2 $opts_mds2 $mdsdev2 $mdsvdev2 &&
5558 error "Must specify --mgsnode when formatting a mdt"
5559 add ost1 $opts_ost1 $ostdev1 $ostvdev2 &&
5560 error "Must specify --mgsnode when formatting an ost"
5564 run_test 79 "format MDT/OST without mgs option (should return errors)"
5567 start_mds || error "Failed to start MDT"
5568 start_ost || error "Failed to start OST1"
5569 uuid=$(do_facet ost1 $LCTL get_param -n mgc.*.uuid)
5570 #define OBD_FAIL_MGS_PAUSE_TARGET_CON 0x906
5571 do_facet ost1 "$LCTL set_param fail_val=10 fail_loc=0x906"
5572 do_facet mgs "$LCTL set_param fail_val=10 fail_loc=0x906"
5573 do_facet mgs "$LCTL set_param -n mgs/MGS/evict_client $uuid"
5575 start_ost2 || error "Failed to start OST2"
5577 do_facet ost1 "$LCTL set_param fail_loc=0"
5580 run_test 80 "mgc import reconnect race"
5582 #Save the original values of $OSTCOUNT and $OSTINDEX$i.
5584 local new_ostcount=$1
5585 saved_ostcount=$OSTCOUNT
5586 OSTCOUNT=$new_ostcount
5590 for i in $(seq $OSTCOUNT); do
5592 eval saved_ostindex$i=${!index}
5597 # Restore the original values of $OSTCOUNT and $OSTINDEX$i.
5598 restore_ostindex() {
5603 for i in $(seq $OSTCOUNT); do
5604 index=saved_ostindex$i
5605 eval OSTINDEX$i=${!index}
5607 OSTCOUNT=$saved_ostcount
5612 # The main purpose of this test is to ensure the OST_INDEX_LIST functions as
5613 # expected. This test uses OST_INDEX_LIST to format OSTs with a randomly
5614 # assigned index and ensures we can mount such a formatted file system
5615 test_81() { # LU-4665
5616 [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.6.54) ]] ||
5617 { skip "Need MDS version at least 2.6.54" && return; }
5618 [[ $OSTCOUNT -ge 3 ]] || { skip_env "Need at least 3 OSTs" && return; }
5622 # Each time RANDOM is referenced, a random integer between 0 and 32767
5625 local saved_ostindex1=$OSTINDEX1
5626 for i in 65535 $((RANDOM + 65536)); do
5627 echo -e "\nFormat ost1 with --index=$i, should fail"
5629 if add ost1 $(mkfs_opts ost1 $(ostdevname 1)) --reformat \
5630 $(ostdevname 1) $(ostvdevname 1); then
5631 OSTINDEX1=$saved_ostindex1
5632 error "format ost1 with --index=$i should fail"
5635 OSTINDEX1=$saved_ostindex1
5639 # Format OSTs with random sparse indices.
5640 trap "restore_ostindex" EXIT
5641 echo -e "\nFormat $OSTCOUNT OSTs with sparse indices"
5642 OST_INDEX_LIST=[0,$((RANDOM * 2 % 65533 + 1)),65534] formatall
5644 # Setup and check Lustre filesystem.
5645 start_mgsmds || error "start_mgsmds failed"
5646 for i in $(seq $OSTCOUNT); do
5647 start ost$i $(ostdevname $i) $OST_MOUNT_OPTS ||
5648 error "start ost$i failed"
5651 mount_client $MOUNT || error "mount client $MOUNT failed"
5652 check_mount || error "check client $MOUNT failed"
5655 local max_easize=$($LCTL get_param -n llite.*.max_easize)
5656 [[ $max_easize -eq 128 ]] ||
5657 error "max_easize is $max_easize, should be 128 bytes"
5661 run_test 81 "sparse OST indexing"
5663 # Here we exercise the stripe placement functionality on a file system that
5664 # has formatted the OST with a random index. With the file system the following
5665 # functionality is tested:
5667 # 1. Creating a new file with a specific stripe layout.
5669 # 2. Modifiy a existing empty file with a specific stripe layout.
5671 # 3. Ensure we fail to set the stripe layout of a file that already has one.
5673 # 4. If ost-index is defined we need to ensure it is the first entry in the
5674 # ost index list returned by lfs getstripe.
5676 # 5. Lastly ensure this functionality fails with directories.
5677 test_82a() { # LU-4665
5678 [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.6.54) ]] ||
5679 { skip "Need MDS version at least 2.6.54" && return; }
5680 [[ $OSTCOUNT -ge 3 ]] || { skip_env "Need at least 3 OSTs" && return; }
5686 # Format OSTs with random sparse indices.
5690 local LOV_V1_INSANE_STRIPE_COUNT=65532
5691 for i in $(seq $OSTCOUNT); do
5692 index=$(((RANDOM * 2) % LOV_V1_INSANE_STRIPE_COUNT))
5693 ost_indices+=" $index"
5695 ost_indices=$(comma_list $ost_indices)
5697 trap "restore_ostindex" EXIT
5698 echo -e "\nFormat $OSTCOUNT OSTs with sparse indices $ost_indices"
5699 OST_INDEX_LIST=[$ost_indices] formatall
5701 # Setup Lustre filesystem.
5702 start_mgsmds || error "start_mgsmds failed"
5703 for i in $(seq $OSTCOUNT); do
5704 start ost$i $(ostdevname $i) $OST_MOUNT_OPTS ||
5705 error "start ost$i failed"
5708 mount_client $MOUNT || error "mount client $MOUNT failed"
5711 $LFS df $MOUNT || error "$LFS df $MOUNT failed"
5712 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
5714 # 1. If the file does not exist, new file will be created
5715 # with specified OSTs.
5716 local file=$DIR/$tdir/$tfile-1
5717 local cmd="$SETSTRIPE -o $ost_indices $file"
5719 eval $cmd || error "$cmd failed"
5720 check_stripe_count $file $OSTCOUNT
5721 check_obdidx $file $ost_indices
5722 dd if=/dev/urandom of=$file count=1 bs=1M > /dev/null 2>&1 ||
5723 error "write $file failed"
5725 # 2. If the file already exists and is an empty file, the file
5726 # will be attached with specified layout.
5727 file=$DIR/$tdir/$tfile-2
5728 mcreate $file || error "mcreate $file failed"
5729 cmd="$SETSTRIPE -o $ost_indices $file"
5731 eval $cmd || error "$cmd failed"
5732 dd if=/dev/urandom of=$file count=1 bs=1M > /dev/null 2>&1 ||
5733 error "write $file failed"
5734 check_stripe_count $file $OSTCOUNT
5735 check_obdidx $file $ost_indices
5737 # 3. If the file already has a valid layout attached, the command
5738 # should fail with EBUSY.
5740 eval $cmd && error "stripe is already set on $file, $cmd should fail"
5742 # 4. If [--stripe-index|-i <start_ost_idx>] is used, the index must
5743 # be in the OST indices list.
5744 local start_ost_idx=${ost_indices##*,}
5745 file=$DIR/$tdir/$tfile-3
5746 cmd="$SETSTRIPE -o $ost_indices -i $start_ost_idx $file"
5748 eval $cmd || error "$cmd failed"
5749 check_stripe_count $file $OSTCOUNT
5750 check_obdidx $file $ost_indices
5751 check_start_ost_idx $file $start_ost_idx
5753 file=$DIR/$tdir/$tfile-4
5755 cmd+=" -o $(exclude_items_from_list $ost_indices $start_ost_idx)"
5756 cmd+=" -i $start_ost_idx $file"
5758 eval $cmd && error "index $start_ost_idx should be in $ost_indices"
5760 # 5. Specifying OST indices for directory should fail with ENOSUPP.
5761 local dir=$DIR/$tdir/$tdir
5762 mkdir $dir || error "mkdir $dir failed"
5763 cmd="$SETSTRIPE -o $ost_indices $dir"
5765 eval $cmd && error "$cmd should fail, specifying OST indices" \
5766 "for directory is not supported"
5770 run_test 82a "specify OSTs for file (succeed) or directory (fail)"
5775 # Remove OSTs from a pool and destroy the pool.
5776 destroy_pool $ost_pool || true
5781 # Test 82b is run to ensure that if the user supplies a pool with a specific
5782 # stripe layout that it behaves proprerly. It should fail in the case that
5783 # the supplied OST index list points to OSTs not contained in the user
5785 test_82b() { # LU-4665
5786 [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.6.54) ]] ||
5787 { skip "Need MDS version at least 2.6.54" && return; }
5788 [[ $OSTCOUNT -ge 4 ]] || { skip_env "Need at least 4 OSTs" && return; }
5794 # Format OSTs with random sparse indices.
5798 local LOV_V1_INSANE_STRIPE_COUNT=65532
5799 for i in $(seq $OSTCOUNT); do
5800 index=$(((RANDOM * 2) % LOV_V1_INSANE_STRIPE_COUNT))
5801 ost_indices+=" $index"
5803 ost_indices=$(comma_list $ost_indices)
5805 trap "restore_ostindex" EXIT
5806 echo -e "\nFormat $OSTCOUNT OSTs with sparse indices $ost_indices"
5807 OST_INDEX_LIST=[$ost_indices] formatall
5809 # Setup Lustre filesystem.
5810 start_mgsmds || error "start_mgsmds failed"
5811 for i in $(seq $OSTCOUNT); do
5812 start ost$i $(ostdevname $i) $OST_MOUNT_OPTS ||
5813 error "start ost$i failed"
5816 mount_client $MOUNT || error "mount client $MOUNT failed"
5818 $LFS df $MOUNT || error "$LFS df $MOUNT failed"
5819 mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
5821 # Create a new pool and add OSTs into it.
5822 local ost_pool=$FSNAME.$TESTNAME
5823 create_pool $ost_pool || error "create OST pool $ost_pool failed"
5826 trap "cleanup_82b" EXIT
5828 local ost_idx_in_list=${ost_indices##*,}
5829 local ost_idx_in_pool=$(exclude_items_from_list $ost_indices \
5832 local ost_targets="$FSNAME-OST["
5833 for i in ${ost_idx_in_pool//,/ }; do
5834 ost_targets=$ost_targets$(printf "%04x," $i)
5836 ost_targets="${ost_targets%,}]"
5838 local ost_targets_uuid=$(for i in ${ost_idx_in_pool//,/ }; \
5839 do printf "$FSNAME-OST%04x_UUID\n" $i; done |
5840 sort -u | tr '\n' ' ')
5842 local cmd="$LCTL pool_add $ost_pool $ost_targets"
5843 do_facet mgs $cmd || error "$cmd failed"
5844 wait_update $HOSTNAME "$LCTL get_param -n lov.$FSNAME-*.pools.$TESTNAME|
5845 sort -u | tr '\n' ' ' " "$ost_targets_uuid" ||
5846 error "wait_update $ost_pool failed"
5847 [[ -z $(list_pool $ost_pool) ]] &&
5848 error "list OST pool $ost_pool failed"
5850 # If [--pool|-p <pool_name>] is set with [--ost-list|-o <ost_indices>],
5851 # then the OSTs must be the members of the pool.
5852 local file=$DIR/$tdir/$tfile
5853 cmd="$SETSTRIPE -p $ost_pool -o $ost_idx_in_list $file"
5855 eval $cmd && error "OST with index $ost_idx_in_list should be" \
5856 "in OST pool $ost_pool"
5858 # Only select OST $ost_idx_in_list from $ost_pool for file.
5859 ost_idx_in_list=${ost_idx_in_pool#*,}
5860 cmd="$SETSTRIPE -p $ost_pool -o $ost_idx_in_list $file"
5862 eval $cmd || error "$cmd failed"
5863 cmd="$GETSTRIPE $file"
5865 eval $cmd || error "$cmd failed"
5866 check_stripe_count $file 2
5867 check_obdidx $file $ost_idx_in_list
5868 dd if=/dev/urandom of=$file count=1 bs=1M > /dev/null 2>&1 ||
5869 error "write $file failed"
5873 run_test 82b "specify OSTs for file with --pool and --ost-list options"
5876 [[ $(lustre_version_code ost1) -ge $(version_code 2.6.91) ]] ||
5877 { skip "Need OST version at least 2.6.91" && return 0; }
5878 if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
5879 skip "Only applicable to ldiskfs-based MDTs"
5889 ostmnt=$(facet_mntpt ost1)
5890 fstype=$(facet_fstype ost1)
5892 # Mount the OST as an ldiskfs filesystem.
5893 log "mount the OST $dev as a $fstype filesystem"
5894 add ost1 $(mkfs_opts ost1 $dev) $FSTYPE_OPT \
5895 --reformat $dev $dev > /dev/null ||
5896 error "format ost1 error"
5898 if ! test -b $dev; then
5899 mnt_opts=$(csa_add "$OST_MOUNT_OPTS" -o loop)
5901 echo "mnt_opts $mnt_opts"
5902 do_facet ost1 mount -t $fstype $dev \
5904 # Run llverfs on the mounted ldiskfs filesystem.
5905 # It is needed to get ENOSPACE.
5906 log "run llverfs in partial mode on the OST $fstype $ostmnt"
5907 do_rpc_nodes $(facet_host ost1) run_llverfs $ostmnt -vpl \
5908 "no" || error "run_llverfs error on $fstype"
5911 log "unmount the OST $dev"
5914 # Delete file IO_scrub. Later osd_scrub_setup will try to
5915 # create "IO_scrub" but will get ENOSPACE.
5917 echo "start ost1 service on `facet_active_host ost1`"
5918 start ost1 `ostdevname 1` $OST_MOUNT_OPTS
5921 err=$(do_facet ost1 dmesg | grep "VFS: Busy inodes after unmount of")
5922 echo "string err $err"
5923 [ -z "$err" ] || error $err
5926 run_test 83 "ENOSPACE on OST doesn't cause message VFS: \
5927 Busy inodes after unmount ..."
5930 local facet=$SINGLEMDS
5931 local num=$(echo $facet | tr -d "mds")
5932 local dev=$(mdsdevname $num)
5933 local time_min=$(recovery_time_min)
5934 local recovery_duration
5935 local completed_clients
5936 local correct_clients
5939 echo "start mds service on $(facet_active_host $facet)"
5941 "-o recovery_time_hard=$time_min,recovery_time_soft=$time_min" $@ ||
5942 error "start MDS failed"
5944 start_ost || error "start OST0000 failed"
5945 start_ost2 || error "start OST0001 failed"
5947 echo "recovery_time=$time_min, timeout=$TIMEOUT, wrap_up=$wrap_up"
5949 mount_client $MOUNT1 || error "mount $MOUNT1 failed"
5950 mount_client $MOUNT2 || error "mount $MOUNT2 failed"
5951 # make sure new superblock labels are sync'd before disabling writes
5955 replay_barrier $SINGLEMDS
5956 createmany -o $DIR1/$tfile-%d 1000
5958 # We need to catch the end of recovery window to extend it.
5959 # Skip 5 requests and add delay to request handling.
5960 #define OBD_FAIL_TGT_REPLAY_DELAY 0x709 | FAIL_SKIP
5961 do_facet $SINGLEMDS "lctl set_param fail_loc=0x20000709 fail_val=5"
5963 facet_failover --fsck $SINGLEMDS || error "failover: $?"
5966 echo "recovery status"
5967 do_facet $SINGLEMDS \
5968 "$LCTL get_param -n mdt.$FSNAME-MDT0000.recovery_status"
5970 recovery_duration=$(do_facet $SINGLEMDS \
5971 "$LCTL get_param -n mdt.$FSNAME-MDT0000.recovery_status" |
5972 awk '/recovery_duration/ { print $2 }')
5973 (( $recovery_duration > $time_min + $wrap_up )) &&
5974 error "recovery_duration > recovery_time_hard + wrap up"
5975 completed_clients=$(do_facet $SINGLEMDS \
5976 "$LCTL get_param -n mdt.$FSNAME-MDT0000.recovery_status" |
5977 awk '/completed_clients/ { print $2 }')
5979 correct_clients="$MDSCOUNT/$((MDSCOUNT+1))"
5980 [ "$completed_clients" = "${correct_clients}" ] ||
5981 error "$completed_clients != $correct_clients"
5983 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
5984 umount_client $MOUNT1
5985 umount_client $MOUNT2
5991 run_test 84 "check recovery_hard_time"
5994 [[ $(lustre_version_code ost1) -ge $(version_code 2.7.55) ]] ||
5995 { skip "Need OST version at least 2.7.55" && return 0; }
5996 ##define OBD_FAIL_OSD_OST_EA_FID_SET 0x197
5997 do_facet ost1 "lctl set_param fail_loc=0x197"
6001 run_test 85 "osd_ost init: fail ea_fid_set"
6006 # ost1 has already registered to the MGS before the reformat.
6007 # So after reformatting it with option "-G", it could not be
6008 # mounted to the MGS. Cleanup the system for subsequent tests.
6013 local server_version=$(lustre_version_code $SINGLEMDS)
6014 [ "$(facet_fstype ost1)" = "zfs" ] &&
6015 skip "LU-6442: no such mkfs params for ZFS OSTs" && return
6016 [[ $server_version -ge $(version_code 2.7.56) ]] ||
6017 { skip "Need server version newer than 2.7.55"; return 0; }
6019 local OST_OPTS="$(mkfs_opts ost1 $(ostdevname 1)) \
6020 --reformat $(ostdevname 1) $(ostvdevname 1)"
6023 local OLDSIZE=$(do_facet ost1 "$DEBUGFS -c -R stats $(ostdevname 1)" |
6024 awk '/Flex block group size: / { print $NF; exit; }')
6027 if [[ ${!opts} != *mkfsoptions* ]]; then
6028 eval opts=\"${!opts} \
6029 --mkfsoptions='\\\"-O flex_bg -G $NEWSIZE\\\"'\"
6031 val=${!opts//--mkfsoptions=\\\"/ \
6032 --mkfsoptions=\\\"-O flex_bg -G $NEWSIZE }
6036 echo "params: $opts"
6038 trap cleanup_86 EXIT ERR
6041 add ost1 $opts || error "add ost1 failed with new params"
6043 local FOUNDSIZE=$(do_facet ost1 "$DEBUGFS -c -R stats $(ostdevname 1)" |
6044 awk '/Flex block group size: / { print $NF; exit; }')
6046 [[ $FOUNDSIZE == $NEWSIZE ]] ||
6047 error "Flex block group size: $FOUNDSIZE, expected: $NEWSIZE"
6051 run_test 86 "Replacing mkfs.lustre -G option"
6053 test_87() { #LU-6544
6054 [[ $(lustre_version_code $SINGLEMDS1) -ge $(version_code 2.7.56) ]] ||
6055 { skip "Need MDS version at least 2.7.56" && return; }
6056 [[ $(facet_fstype $SINGLEMDS) != ldiskfs ]] &&
6057 { skip "Only applicable to ldiskfs-based MDTs" && return; }
6058 [[ $OSTCOUNT -gt 69 ]] &&
6059 { skip "Ignore wide striping situation" && return; }
6061 local mdsdev=$(mdsdevname 1)
6062 local mdsvdev=$(mdsvdevname 1)
6063 local file=$DIR/$tfile
6064 local mntpt=$(facet_mntpt $SINGLEMDS)
6065 local used_xattr_blk=0
6066 local inode_size=${1:-512}
6068 local xtest="trusted.test"
6073 #Please see LU-6544 for MDT inode size calculation
6074 if [ $OSTCOUNT -gt 26 ]; then
6076 elif [ $OSTCOUNT -gt 5 ]; then
6079 left_size=$(expr $inode_size - \
6082 32 - $OSTCOUNT \* 24 - 16 - 3 - \
6084 24 - 18 - $(expr length $tfile) - 16 - 4)
6085 if [ $left_size -le 0 ]; then
6086 echo "No space($left_size) is expected in inode."
6087 echo "Try 1-byte xattr instead to verify this."
6090 echo "Estimate: at most $left_size-byte space left in inode."
6096 add mds1 $(mkfs_opts mds1 ${mdsdev}) --stripe-count-hint=$OSTCOUNT \
6097 --reformat $mdsdev $mdsvdev || error "add mds1 failed"
6098 start_mdt 1 > /dev/null || error "start mdt1 failed"
6099 for i in $(seq $OSTCOUNT); do
6100 start ost$i $(ostdevname $i) $OST_MOUNT_OPTS > /dev/null ||
6101 error "start ost$i failed"
6103 mount_client $MOUNT > /dev/null || error "mount client $MOUNT failed"
6104 check_mount || error "check client $MOUNT failed"
6107 $SETSTRIPE -c -1 $file || error "$SETSTRIPE -c -1 $file failed"
6108 $GETSTRIPE $file || error "$GETSTRIPE $file failed"
6109 i=$($GETSTRIPE -c $file)
6110 if [ $i -ne $OSTCOUNT ]; then
6111 left_size=$(expr $left_size + $(expr $OSTCOUNT - $i) \* 24)
6112 echo -n "Since only $i out $OSTCOUNT OSTs are used, "
6113 echo -n "the expected left space is changed to "
6114 echo "$left_size bytes at most."
6116 value=$(generate_string $left_size)
6117 setfattr -n $xtest -v $value $file
6118 orig=$(get_xattr_value $xtest $file)
6119 [[ "$orig" != "$value" ]] && error "$xtest changed"
6121 #Verify if inode has some expected space left
6122 umount $MOUNT > /dev/null || error "umount $MOUNT failed"
6123 stop_mdt 1 > /dev/null || error "stop mdt1 failed"
6124 mount_ldiskfs $SINGLEMDS || error "mount -t ldiskfs $SINGLEMDS failed"
6126 do_facet $SINGLEMDS ls -sal $mntpt/ROOT/$tfile
6127 used_xattr_blk=$(do_facet $SINGLEMDS ls -s $mntpt/ROOT/$tfile |
6129 [[ $used_xattr_blk -eq 0 ]] &&
6130 error "Please check MDS inode size calculation: \
6131 more than $left_size-byte space left in inode."
6132 echo "Verified: at most $left_size-byte space left in inode."
6136 run_test 87 "check if MDT inode can hold EAs with N stripes properly"
6139 [ "$(facet_fstype mds1)" == "zfs" ] &&
6140 skip "LU-6662: no implementation for ZFS" && return
6144 add mds1 $(mkfs_opts mds1 $(mdsdevname 1)) \
6145 --reformat $(mdsdevname 1) || error "add mds1 failed"
6147 do_facet mds1 "$TUNEFS $(mdsdevname 1) |
6148 grep -e \".*opts:.*errors=remount-ro.*\"" ||
6149 error "default mount options is missing"
6151 add mds1 $(mkfs_opts mds1 $(mdsdevname 1)) \
6152 --mountfsoptions="user_xattr,errors=panic" \
6153 --reformat $(mdsdevname 1) || error "add mds1 failed"
6155 do_facet mds1 "$TUNEFS $(mdsdevname 1) |
6156 grep -e \".*opts:.*errors=panic.*\"" ||
6157 error "user can't override default mount options"
6159 run_test 88 "check the default mount options can be overridden"
6162 # $2 (optional) value of max_mod_rpcs_in_flight to set
6163 check_max_mod_rpcs_in_flight() {
6171 idx=$(printf "%04x" $($LFS getdirstripe -i $dir))
6172 facet="mds$((0x$idx + 1))"
6174 if [ -z "$mmr" ]; then
6175 # get value of max_mod_rcps_in_flight
6176 mmr=$($LCTL get_param -n \
6177 mdc.$FSNAME-MDT$idx-mdc-*.max_mod_rpcs_in_flight) ||
6178 error "Unable to get max_mod_rpcs_in_flight"
6179 echo "max_mod_rcps_in_flight is $mmr"
6181 # set value of max_mod_rpcs_in_flight
6183 mdc.$FSNAME-MDT$idx-mdc-*.max_mod_rpcs_in_flight=$mmr ||
6184 error "Unable to set max_mod_rpcs_in_flight to $mmr"
6185 echo "max_mod_rpcs_in_flight set to $mmr"
6188 # create mmr+1 files
6189 echo "creating $((mmr + 1)) files ..."
6191 for i in $(seq $((mmr + 1))); do
6197 # consumes mmr-1 modify RPC slots
6198 #define OBD_FAIL_MDS_REINT_MULTI_NET 0x159
6199 # drop requests on MDT so that RPC slots are consumed
6200 # during all the request resend interval
6201 do_facet $facet "$LCTL set_param fail_loc=0x159"
6202 echo "launch $((mmr - 1)) chmod in parallel ..."
6203 for i in $(seq $((mmr - 1))); do
6204 chmod 0600 $dir/file-$i &
6208 # send one additional modify RPC
6209 do_facet $facet "$LCTL set_param fail_loc=0"
6210 echo "launch 1 additional chmod in parallel ..."
6211 chmod 0600 $dir/file-$mmr &
6214 # check this additional modify RPC get a modify RPC slot
6215 # and succeed its operation
6216 checkstat -vp 0600 $dir/file-$mmr ||
6217 error "Unable to send $mmr modify RPCs in parallel"
6222 # consumes mmr modify RPC slots
6223 #define OBD_FAIL_MDS_REINT_MULTI_NET 0x159
6224 # drop requests on MDT so that RPC slots are consumed
6225 # during all the request resend interval
6226 do_facet $facet "$LCTL set_param fail_loc=0x159"
6227 echo "launch $mmr chmod in parallel ..."
6228 for i in $(seq $mmr); do
6229 chmod 0666 $dir/file-$i &
6233 # send one additional modify RPC
6234 do_facet $facet "$LCTL set_param fail_loc=0"
6235 echo "launch 1 additional chmod in parallel ..."
6236 chmod 0666 $dir/file-$((mmr + 1)) &
6239 # check this additional modify RPC blocked getting a modify RPC slot
6240 checkstat -vp 0644 $dir/file-$((mmr + 1)) ||
6241 error "Unexpectedly send $(($mmr + 1)) modify RPCs in parallel"
6247 if ! combined_mgs_mds ; then
6252 [[ $($LCTL get_param mdc.*.import |
6253 grep "connect_flags:.*multi_mod_rpc") ]] ||
6254 { skip "Need MDC with 'multi_mod_rpcs' feature"; return 0; }
6256 # check default value
6257 $LFS mkdir -c1 $DIR/$tdir || error "mkdir $DIR/$tdir failed"
6258 check_max_mod_rpcs_in_flight $DIR/$tdir
6262 run_test 90a "check max_mod_rpcs_in_flight is enforced"
6272 [[ $($LCTL get_param mdc.*.import |
6273 grep "connect_flags:.*multi_mod_rpc") ]] ||
6274 { skip "Need MDC with 'multi_mod_rpcs' feature"; return 0; }
6277 # update max_mod_rpcs_in_flight
6278 $LFS mkdir -c1 $DIR/${tdir}1 || error "mkdir $DIR/${tdir}1 failed"
6279 check_max_mod_rpcs_in_flight $DIR/${tdir}1 1
6282 # check client is able to send multiple modify RPCs in paralell
6283 tmp=$($LCTL get_param -n mdc.$FSNAME-MDT*-mdc-*.import |
6284 grep -c "multi_mod_rpcs")
6285 if [ "$tmp" -ne $MDSCOUNT ]; then
6286 echo "Client not able to send multiple modify RPCs in parallel"
6291 # update max_mod_rpcs_in_flight
6292 $LFS mkdir -c1 $DIR/${tdir}2 || error "mkdir $DIR/${tdir}2 failed"
6293 check_max_mod_rpcs_in_flight $DIR/${tdir}2 5
6296 $LFS mkdir -c1 $DIR/${tdir}3 || error "mkdir $DIR/${tdir}3 failed"
6297 idx=$(printf "%04x" $($LFS getdirstripe -i $DIR/${tdir}3))
6298 facet="mds$((0x$idx + 1))"
6300 # save MDT max_mod_rpcs_per_client
6301 mmrpc=$(do_facet $facet \
6302 cat /sys/module/mdt/parameters/max_mod_rpcs_per_client)
6304 # update max_mod_rpcs_in_flight
6305 umount_client $MOUNT
6307 "echo 16 > /sys/module/mdt/parameters/max_mod_rpcs_per_client"
6309 $LCTL set_param mdc.$FSNAME-MDT$idx-mdc-*.max_rpcs_in_flight=17
6310 check_max_mod_rpcs_in_flight $DIR/${tdir}3 16
6312 # restore MDT max_mod_rpcs_per_client initial value
6314 "echo $mmrpc > /sys/module/mdt/parameters/max_mod_rpcs_per_client"
6316 rm -rf $DIR/${tdir}?
6319 run_test 90b "check max_mod_rpcs_in_flight is enforced after update"
6328 [[ $($LCTL get_param mdc.*.import |
6329 grep "connect_flags:.*multi_mod_rpc") ]] ||
6330 { skip "Need MDC with 'multi_mod_rpcs' feature"; return 0; }
6332 # check client is able to send multiple modify RPCs in paralell
6333 tmp=$($LCTL get_param -n mdc.$FSNAME-MDT*-mdc-*.import |
6334 grep -c "multi_mod_rpcs")
6335 if [ "$tmp" -ne $MDSCOUNT ]; then
6336 skip "Client not able to send multiple modify RPCs in parallel"
6341 # get max_rpcs_in_flight value
6342 mrif=$($LCTL get_param -n mdc.$FSNAME-MDT0000-mdc-*.max_rpcs_in_flight)
6343 echo "max_rpcs_in_flight is $mrif"
6345 # get MDT max_mod_rpcs_per_client
6346 mmrpc=$(do_facet mds1 \
6347 cat /sys/module/mdt/parameters/max_mod_rpcs_per_client)
6348 echo "max_mod_rpcs_per_client is $mmrpc"
6351 # attempt to set max_mod_rpcs_in_flight to max_rpcs_in_flight value
6352 # prerequisite: set max_mod_rpcs_per_client to max_rpcs_in_flight value
6353 umount_client $MOUNT
6355 "echo $mrif > /sys/module/mdt/parameters/max_mod_rpcs_per_client"
6359 mdc.$FSNAME-MDT0000-mdc-*.max_mod_rpcs_in_flight=$mrif &&
6360 error "set max_mod_rpcs_in_flight to $mrif should fail"
6362 umount_client $MOUNT
6364 "echo $mmrpc > /sys/module/mdt/parameters/max_mod_rpcs_per_client"
6368 # attempt to set max_mod_rpcs_in_flight to max_mod_rpcs_per_client+1
6369 # prerequisite: set max_rpcs_in_flight to max_mod_rpcs_per_client+2
6371 mdc.$FSNAME-MDT0000-mdc-*.max_rpcs_in_flight=$((mmrpc + 2))
6374 mdc.$FSNAME-MDT0000-mdc-*.max_mod_rpcs_in_flight=$((mmrpc + 1)) &&
6375 error "set max_mod_rpcs_in_flight to $((mmrpc + 1)) should fail"
6379 run_test 90c "check max_mod_rpcs_in_flight update limits"
6390 [[ $($LCTL get_param mdc.*.import |
6391 grep "connect_flags:.*multi_mod_rpc") ]] ||
6392 { skip "Need MDC with 'multi_mod_rpcs' feature"; return 0; }
6394 $LFS mkdir -c1 $DIR/$tdir || error "mkdir $DIR/$tdir failed"
6395 idx=$(printf "%04x" $($LFS getdirstripe -i $DIR/$tdir))
6396 facet="mds$((0x$idx + 1))"
6398 # check client version supports multislots
6399 tmp=$($LCTL get_param -N \
6400 mdc.$FSNAME-MDT$idx-mdc-*.max_mod_rpcs_in_flight)
6401 if [ -z "$tmp" ]; then
6402 skip "Client does not support multiple modify RPCs in flight"
6407 # get current value of max_mod_rcps_in_flight
6408 mmr=$($LCTL get_param -n \
6409 mdc.$FSNAME-MDT$idx-mdc-*.max_mod_rpcs_in_flight)
6410 echo "max_mod_rcps_in_flight is $mmr"
6413 echo "creating $mmr files ..."
6415 for i in $(seq $mmr); do
6416 touch $DIR/$tdir/file-$i
6419 # prepare for close RPC
6420 multiop_bg_pause $DIR/$tdir/file-close O_c
6423 # consumes mmr modify RPC slots
6424 #define OBD_FAIL_MDS_REINT_MULTI_NET 0x159
6425 # drop requests on MDT so that RPC slots are consumed
6426 # during all the request resend interval
6427 do_facet $facet "$LCTL set_param fail_loc=0x159"
6428 echo "launch $mmr chmod in parallel ..."
6429 for i in $(seq $mmr); do
6430 chmod 0600 $DIR/$tdir/file-$i &
6433 # send one additional close RPC
6434 do_facet $facet "$LCTL set_param fail_loc=0"
6435 echo "launch 1 additional close in parallel ..."
6437 cancel_lru_locks mdc
6440 # check this additional close RPC get a modify RPC slot
6441 # and multiop process completed
6442 [ -d /proc/$pid ] &&
6443 error "Unable to send the additional close RPC in parallel"
6448 run_test 90d "check one close RPC is allowed above max_mod_rpcs_in_flight"
6450 check_uuid_on_ost() {
6452 do_facet ost1 "$LCTL get_param obdfilter.${FSNAME}*.exports.'$nid'.uuid"
6455 check_uuid_on_mdt() {
6457 do_facet $SINGLEMDS "$LCTL get_param mdt.${FSNAME}*.exports.'$nid'.uuid"
6465 [[ $(lustre_version_code ost1) -ge $(version_code 2.7.63) ]] ||
6466 { skip "Need OST version at least 2.7.63" && return 0; }
6467 [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.7.63) ]] ||
6468 { skip "Need MDT version at least 2.7.63" && return 0; }
6470 start_mds || error "MDS start failed"
6471 start_ost || error "unable to start OST"
6472 mount_client $MOUNT || error "client start failed"
6473 check_mount || error "check_mount failed"
6476 nid=$($LCTL list_nids | head -1 | sed "s/\./\\\./g")
6480 uuid=$(get_client_uuid $MOUNT)
6482 echo "list nids on mdt:"
6483 do_facet $SINGLEMDS "$LCTL list_param mdt.${FSNAME}*.exports.*"
6484 echo "uuid from $nid:"
6485 do_facet $SINGLEMDS "$LCTL get_param mdt.${FSNAME}*.exports.'$nid'.uuid"
6487 found=$(check_uuid_on_mdt $nid | grep $uuid)
6488 [ -z "$found" ] && error "can't find $uuid $nid on MDT"
6489 found=$(check_uuid_on_ost $nid | grep $uuid)
6490 [ -z "$found" ] && error "can't find $uuid $nid on OST"
6492 # umount the client so it won't reconnect
6493 manual_umount_client --force || error "failed to umount $?"
6494 # shouldn't disappear on MDS after forced umount
6495 found=$(check_uuid_on_mdt $nid | grep $uuid)
6496 [ -z "$found" ] && error "can't find $uuid $nid"
6499 do_facet $SINGLEMDS \
6500 "$LCTL set_param -n mdt.${mds1_svc}.evict_client nid:$nid"
6502 found=$(check_uuid_on_mdt $nid | grep $uuid)
6503 [ -n "$found" ] && error "found $uuid $nid on MDT"
6504 found=$(check_uuid_on_ost $nid | grep $uuid)
6505 [ -n "$found" ] && error "found $uuid $nid on OST"
6507 # check it didn't reconnect (being umounted)
6508 sleep $((TIMEOUT+1))
6509 found=$(check_uuid_on_mdt $nid | grep $uuid)
6510 [ -n "$found" ] && error "found $uuid $nid on MDT"
6511 found=$(check_uuid_on_ost $nid | grep $uuid)
6512 [ -n "$found" ] && error "found $uuid $nid on OST"
6516 run_test 91 "evict-by-nid support"
6518 generate_ldev_conf() {
6519 # generate an ldev.conf file
6520 local ldevconfpath=$1
6522 local fsldevformat=""
6525 fstype=$(facet_fstype mgs)
6526 if [ "$fstype" == "zfs" ]; then
6527 fsldevformat="$fstype:"
6532 printf "%s\t-\t%s-MGS0000\t%s%s\n" \
6536 $(mgsdevname) > $ldevconfpath
6538 local mdsfo_host=$mdsfailover_HOST;
6539 if [ -z "$mdsfo_host" ]; then
6543 for num in $(seq $MDSCOUNT); do
6544 fstype=$(facet_fstype mds$num)
6545 if [ "$fstype" == "zfs" ]; then
6546 fsldevformat="$fstype:"
6551 printf "%s\t%s\t%s-MDT%04d\t%s%s\n" \
6557 $(mdsdevname $num) >> $ldevconfpath
6560 local ostfo_host=$ostfailover_HOST;
6561 if [ -z "$ostfo_host" ]; then
6565 for num in $(seq $OSTCOUNT); do
6566 fstype=$(facet_fstype ost$num)
6567 if [ "$fstype" == "zfs" ]; then
6568 fsldevformat="$fstype:"
6573 printf "%s\t%s\t%s-OST%04d\t%s%s\n" \
6579 $(ostdevname $num) >> $ldevconfpath
6582 echo "----- $ldevconfpath -----"
6584 echo "--- END $ldevconfpath ---"
6589 # generate a nids file (mapping between hostname to nid)
6590 # looks like we only have the MGS nid available to us
6591 # so just echo that to a file
6593 echo -e "${mgs_HOST}\t${MGSNID}" > $nidspath
6595 echo "----- $nidspath -----"
6597 echo "--- END $nidspath ---"
6600 compare_ldev_output() {
6604 sort $expected_output -o $expected_output
6605 sort $ldev_output -o $ldev_output
6607 echo "-- START OF LDEV OUTPUT --"
6609 echo "--- END OF LDEV OUTPUT ---"
6611 echo "-- START OF EXPECTED OUTPUT --"
6612 cat $expected_output
6613 echo "--- END OF EXPECTED OUTPUT ---"
6615 diff $expected_output $ldev_output
6620 if [ -z "$LDEV" ]; then
6621 error "ldev is missing!"
6624 local LDEVCONFPATH=$TMP/ldev.conf
6625 local NIDSPATH=$TMP/nids
6627 echo "Host is $(hostname)"
6629 generate_ldev_conf $LDEVCONFPATH
6630 generate_nids $NIDSPATH
6632 # echo the mgs nid and compare it to environment variable MGSNID
6633 # also, ldev.conf and nids is a server side thing, use the OSS
6636 output=$($LDEV -c $LDEVCONFPATH -H $ost_HOST -n $NIDSPATH echo %m)
6638 echo "-- START OF LDEV OUTPUT --"
6640 echo "--- END OF LDEV OUTPUT ---"
6642 # ldev failed, error
6643 if [ $? -ne 0 ]; then
6644 rm $LDEVCONFPATH $NIDSPATH
6645 error "ldev failed to execute!"
6648 # need to process multiple lines because of combined MGS and MDS
6649 echo -e $output | awk '{ print $2 }' | while read -r line ; do
6650 if [ "$line" != "$MGSNID" ]; then
6651 rm $LDEVCONFPATH $NIDSPATH
6652 error "ldev failed mgs nid '$line', expected '$MGSNID'"
6656 rm $LDEVCONFPATH $NIDSPATH
6658 run_test 92 "ldev returns MGS NID correctly in command substitution"
6661 [ $MDSCOUNT -lt 3 ] && skip "needs >= 3 MDTs" && return
6664 #start mgs or mgs/mdt0
6665 if ! combined_mgs_mds ; then
6672 start_ost || error "OST0 start fail"
6674 #define OBD_FAIL_MGS_WRITE_TARGET_DELAY 0x90e
6675 do_facet mgs "$LCTL set_param fail_val = 10 fail_loc=0x8000090e"
6676 for num in $(seq 2 $MDSCOUNT); do
6680 mount_client $MOUNT || error "mount client fails"
6681 wait_osc_import_state mds ost FULL
6682 wait_osc_import_state client ost FULL
6683 check_mount || error "check_mount failed"
6685 cleanup || error "cleanup failed with $?"
6687 run_test 93 "register mulitple MDT at the same time"
6690 if [ -z "$LDEV" ]; then
6691 error "ldev is missing!"
6694 local LDEVCONFPATH=$TMP/ldev.conf
6695 local NIDSPATH=$TMP/nids
6697 generate_ldev_conf $LDEVCONFPATH
6698 generate_nids $NIDSPATH
6700 local LDEV_OUTPUT=$TMP/ldev-output.txt
6701 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -F $FSNAME > $LDEV_OUTPUT
6703 # ldev failed, error
6704 if [ $? -ne 0 ]; then
6705 rm $LDEVCONFPATH $NIDSPATH $LDEV_OUTPUT
6706 error "ldev failed to execute!"
6710 local EXPECTED_OUTPUT=$TMP/ldev-expected.txt
6712 printf "%s-MGS0000\n" $FSNAME > $EXPECTED_OUTPUT
6714 for num in $(seq $MDSCOUNT); do
6715 printf "%s-MDT%04d\n" $FSNAME $num >> $EXPECTED_OUTPUT
6718 for num in $(seq $OSTCOUNT); do
6719 printf "%s-OST%04d\n" $FSNAME $num >> $EXPECTED_OUTPUT
6722 compare_ldev_output $LDEV_OUTPUT $EXPECTED_OUTPUT
6724 if [ $? -ne 0 ]; then
6725 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
6726 error "ldev failed to produce the correct hostlist!"
6729 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
6731 run_test 94 "ldev outputs correct labels for file system name query"
6734 if [ -z "$LDEV" ]; then
6735 error "ldev is missing!"
6738 local LDEVCONFPATH=$TMP/ldev.conf
6739 local NIDSPATH=$TMP/nids
6741 generate_ldev_conf $LDEVCONFPATH
6742 generate_nids $NIDSPATH
6746 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -F $FSNAME &>/dev/null
6747 if [ $? -ne 0 ]; then
6748 rm $LDEVCONFPATH $NIDSPATH
6749 error "ldev label filtering w/ -F failed!"
6753 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -l &>/dev/null
6754 if [ $? -ne 0 ]; then
6755 rm $LDEVCONFPATH $NIDSPATH
6756 error "ldev label filtering w/ -l failed!"
6760 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -f &>/dev/null
6761 if [ $? -ne 0 ]; then
6762 rm $LDEVCONFPATH $NIDSPATH
6763 error "ldev label filtering w/ -f failed!"
6767 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -a &>/dev/null
6768 if [ $? -ne 0 ]; then
6769 rm $LDEVCONFPATH $NIDSPATH
6770 error "ldev label filtering w/ -a failed!"
6775 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -a -F $FSNAME &>/dev/null
6776 if [ $? -eq 0 ]; then
6777 rm $LDEVCONFPATH $NIDSPATH
6778 error "ldev label filtering w/ -a and -F incorrectly succeeded"
6782 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -a -f &>/dev/null
6783 if [ $? -eq 0 ]; then
6784 rm $LDEVCONFPATH $NIDSPATH
6785 error "ldev label filtering w/ -a and -f incorrectly succeeded"
6789 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -a -l &>/dev/null
6790 if [ $? -eq 0 ]; then
6791 rm $LDEVCONFPATH $NIDSPATH
6792 error "ldev label filtering w/ -a and -l incorrectly succeeded"
6796 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -f -l &>/dev/null
6797 if [ $? -eq 0 ]; then
6798 rm $LDEVCONFPATH $NIDSPATH
6799 error "ldev label filtering w/ -f and -l incorrectly succeeded"
6803 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -F $FSNAME -l &>/dev/null
6804 if [ $? -eq 0 ]; then
6805 rm $LDEVCONFPATH $NIDSPATH
6806 error "ldev label filtering w/ -F and -l incorrectly succeeded"
6809 # file sys & foreign
6810 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -F $FSNAME -f &>/dev/null
6811 if [ $? -eq 0 ]; then
6812 rm $LDEVCONFPATH $NIDSPATH
6813 error "ldev label filtering w/ -F and -f incorrectly succeeded"
6816 rm $LDEVCONFPATH $NIDSPATH
6818 run_test 95 "ldev should only allow one label filter"
6821 if [ -z "$LDEV" ]; then
6822 error "ldev is missing!"
6825 local LDEVCONFPATH=$TMP/ldev.conf
6826 local NIDSPATH=$TMP/nids
6828 generate_ldev_conf $LDEVCONFPATH
6829 generate_nids $NIDSPATH
6831 local LDEV_OUTPUT=$TMP/ldev-output.txt
6832 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -H $mgs_HOST \
6834 awk '{print $2}' > $LDEV_OUTPUT
6836 # ldev failed, error
6837 if [ $? -ne 0 ]; then
6838 rm $LDEVCONFPATH $NIDSPATH $LDEV_OUTPUT
6839 error "ldev failed to execute!"
6843 local EXPECTED_OUTPUT=$TMP/ldev-expected-output.txt
6845 echo "$mgs_HOST-$(facet_fstype mgs)" > $EXPECTED_OUTPUT
6847 if [ "$mgs_HOST" == "$mds_HOST" ]; then
6848 for num in $(seq $MDSCOUNT); do
6849 echo "$mds_HOST-$(facet_fstype mds$num)" \
6854 if [ "$mgs_HOST" == "$ost_HOST" ]; then
6855 for num in $(seq $OSTCOUNT); do
6856 echo "$ost_HOST-$(facet_fstype ost$num)" \
6861 compare_ldev_output $LDEV_OUTPUT $EXPECTED_OUTPUT
6863 if [ $? -ne 0 ]; then
6864 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
6865 error "ldev failed to produce the correct output!"
6868 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
6870 run_test 96 "ldev returns hostname and backend fs correctly in command sub"
6873 if [ -z "$LDEV" ]; then
6874 error "ldev is missing!"
6877 local LDEVCONFPATH=$TMP/ldev.conf
6878 local NIDSPATH=$TMP/nids
6880 generate_ldev_conf $LDEVCONFPATH
6881 generate_nids $NIDSPATH
6883 local LDEV_OUTPUT=$TMP/ldev-output.txt
6884 local EXPECTED_OUTPUT=$TMP/ldev-expected-output.txt
6886 echo -e "\nMDT role"
6887 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -F $FSNAME -R mdt > $LDEV_OUTPUT
6889 if [ $? -ne 0 ]; then
6890 rm $LDEVCONFPATH $NIDSPATH $LDEV_OUTPUT
6891 error "ldev failed to execute for mdt role!"
6894 for num in $(seq $MDSCOUNT); do
6895 printf "%s-MDT%04d\n" $FSNAME $num >> $EXPECTED_OUTPUT
6898 compare_ldev_output $LDEV_OUTPUT $EXPECTED_OUTPUT
6900 if [ $? -ne 0 ]; then
6901 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
6902 error "ldev failed to produce the correct output for mdt role!"
6905 echo -e "\nOST role"
6906 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -F $FSNAME -R ost > $LDEV_OUTPUT
6908 if [ $? -ne 0 ]; then
6909 rm $LDEVCONFPATH $NIDSPATH $LDEV_OUTPUT $EXPECTED_OUTPUT
6910 error "ldev failed to execute for ost role!"
6914 for num in $(seq $OSTCOUNT); do
6915 printf "%s-OST%04d\n" $FSNAME $num >> $EXPECTED_OUTPUT
6918 compare_ldev_output $LDEV_OUTPUT $EXPECTED_OUTPUT
6920 if [ $? -ne 0 ]; then
6921 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
6922 error "ldev failed to produce the correct output for ost role!"
6925 echo -e "\nMGS role"
6926 $LDEV -c $LDEVCONFPATH -n $NIDSPATH -F $FSNAME -R mgs > $LDEV_OUTPUT
6928 if [ $? -ne 0 ]; then
6929 rm $LDEVCONFPATH $NIDSPATH $LDEV_OUTPUT $EXPECTED_OUTPUT
6930 error "ldev failed to execute for mgs role!"
6933 printf "%s-MGS0000\n" $FSNAME > $EXPECTED_OUTPUT
6935 compare_ldev_output $LDEV_OUTPUT $EXPECTED_OUTPUT
6937 if [ $? -ne 0 ]; then
6938 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
6939 error "ldev failed to produce the correct output for mgs role!"
6942 rm $LDEVCONFPATH $NIDSPATH $EXPECTED_OUTPUT $LDEV_OUTPUT
6944 run_test 97 "ldev returns correct ouput when querying based on role"
6949 local temp=$MDS_MOUNT_OPTS
6952 check_mount || error "mount failed"
6953 mountopt="user_xattr"
6954 for ((x = 1; x <= 400; x++)); do
6955 mountopt="$mountopt,user_xattr"
6957 remount_client $mountopt $MOUNT 2>&1 | grep "too long" ||
6958 error "Buffer overflow check failed"
6959 cleanup || error "cleanup failed"
6961 run_test 98 "Buffer-overflow check while parsing mount_opts"
6965 [[ $(facet_fstype ost1) != ldiskfs ]] &&
6966 { skip "Only applicable to ldiskfs-based OSTs" && return; }
6967 [[ $(lustre_version_code ost1) -ge $(version_code 2.8.57) ]] ||
6968 { skip "Need OST version at least 2.8.57" && return 0; }
6970 local ost_opts="$(mkfs_opts ost1 $(ostdevname 1)) \
6971 --reformat $(ostdevname 1) $(ostvdevname 1)"
6972 do_facet ost1 $DEBUGFS -c -R stats `ostdevname 1` | grep "meta_bg" &&
6973 skip "meta_bg already set" && return
6976 if [[ ${!opts} != *mkfsoptions* ]]; then
6977 eval opts=\"${!opts} \
6978 --mkfsoptions='\\\"-O ^resize_inode,meta_bg\\\"'\"
6980 local val=${!opts//--mkfsoptions=\\\"/ \
6981 --mkfsoptions=\\\"-O ^resize_inode,meta_bg }
6985 echo "params: $opts"
6987 add ost1 $opts || error "add ost1 failed with new params"
6989 do_facet ost1 $DEBUGFS -c -R stats `ostdevname 1` | grep "meta_bg" ||
6990 error "meta_bg is not set"
6994 run_test 99 "Adding meta_bg option"
6998 start_mds || error "MDS start failed"
6999 start_ost || error "unable to start OST"
7000 mount_client $MOUNT || error "client start failed"
7001 check_mount || error "check_mount failed"
7010 do_facet mgs 'lshowmount -v' | awk 'BEGIN {NR == 0; rc=1} /MGS:/ {rc=0}
7011 END {exit rc}' || error "lshowmount have no output MGS"
7013 do_facet mds1 'lshowmount -v' | awk 'BEGIN {NR == 2; rc=1} /-MDT0000:/
7014 {rc=0} END {exit rc}' || error "lshowmount have no output MDT0"
7016 do_facet ost1 'lshowmount -v' | awk 'BEGIN {NR == 4; rc=1} /-OST0000:/
7017 {rc=0} END {exit rc}' || error "lshowmount have no output OST0"
7019 cleanup || error "cleanup failed with $?"
7021 run_test 100 "check lshowmount lists MGS, MDT, OST and 0@lo"
7024 local createmany_oid
7025 local dev=$FSNAME-OST0000-osc-MDT0000
7028 createmany -o $DIR1/$tfile-%d 50000 &
7030 # MDT->OST reconnection causes MDT<->OST last_id synchornisation
7031 # via osp_precreate_cleanup_orphans.
7032 for ((i = 0; i < 100; i++)); do
7033 for ((k = 0; k < 10; k++)); do
7034 do_facet $SINGLEMDS "$LCTL --device $dev deactivate;" \
7035 "$LCTL --device $dev activate"
7038 ls -asl $MOUNT | grep '???' &&
7039 (kill -9 $createmany_oid &>/dev/null; \
7040 error "File hasn't object on OST")
7042 kill -s 0 $createmany_oid || break
7044 wait $createmany_oid
7047 run_test 101 "Race MDT->OST reconnection with create"
7050 cleanup || error "cleanup failed with $?"
7052 local mds1dev=$(mdsdevname 1)
7053 local mds1mnt=$(facet_mntpt mds1)
7054 local mds1fstype=$(facet_fstype mds1)
7055 local mds1opts=$MDS_MOUNT_OPTS
7057 if [ $mds1fstype == ldiskfs ] &&
7058 ! do_facet mds1 test -b $mds1dev; then
7059 mds1opts=$(csa_add "$mds1opts" -o loop)
7061 if [[ $mds1fstype == zfs ]]; then
7062 import_zpool mds1 || return ${PIPESTATUS[0]}
7065 # unload all and only load libcfs to allow fail_loc setting
7066 do_facet mds1 lustre_rmmod || error "unable to unload modules"
7067 do_facet mds1 modprobe libcfs || error "libcfs not loaded"
7068 do_facet mds1 lsmod \| grep libcfs || error "libcfs not loaded"
7070 #define OBD_FAIL_OBDCLASS_MODULE_LOAD 0x60a
7071 do_facet mds1 "$LCTL set_param fail_loc=0x8000060a"
7073 do_facet mds1 $MOUNT_CMD $mds1dev $mds1mnt $mds1opts &&
7074 error "mdt start must fail"
7075 do_facet mds1 lsmod \| grep obdclass && error "obdclass must not load"
7077 do_facet mds1 "$LCTL set_param fail_loc=0x0"
7079 do_facet mds1 $MOUNT_CMD $mds1dev $mds1mnt $mds1opts ||
7080 error "mdt start must not fail"
7082 cleanup || error "cleanup failed with $?"
7084 run_test 102 "obdclass module cleanup upon error"
7089 echo "rename $FSNAME to $newname"
7091 if [ ! combined_mgs_mds ]; then
7092 local facet=$(mgsdevname)
7095 "$TUNEFS --fsname=$newname --rename=$FSNAME -v $facet"||
7096 error "(7) Fail to rename MGS"
7097 if [ "$(facet_fstype $facet)" = "zfs" ]; then
7098 reimport_zpool mgs $newname-mgs
7102 for num in $(seq $MDSCOUNT); do
7103 local facet=$(mdsdevname $num)
7105 do_facet mds${num} \
7106 "$TUNEFS --fsname=$newname --rename=$FSNAME -v $facet"||
7107 error "(8) Fail to rename MDT $num"
7108 if [ "$(facet_fstype $facet)" = "zfs" ]; then
7109 reimport_zpool mds${num} $newname-mdt${num}
7113 for num in $(seq $OSTCOUNT); do
7114 local facet=$(ostdevname $num)
7116 do_facet ost${num} \
7117 "$TUNEFS --fsname=$newname --rename=$FSNAME -v $facet"||
7118 error "(9) Fail to rename OST $num"
7119 if [ "$(facet_fstype $facet)" = "zfs" ]; then
7120 reimport_zpool ost${num} $newname-ost${num}
7125 test_103_set_pool() {
7129 do_facet mgs $LCTL pool_add $FSNAME.$pname ${FSNAME}-$ost_x ||
7130 error "Fail to add $ost_x to $FSNAME.$pname"
7131 wait_update $HOSTNAME \
7132 "lctl get_param -n lov.$FSNAME-clilov-*.pools.$pname |
7133 grep $ost_x" "$FSNAME-${ost_x}_UUID" ||
7134 error "$ost_x is NOT in pool $FSNAME.$pname"
7137 test_103_check_pool() {
7138 local save_fsname=$1
7141 stat $DIR/$tdir/test-framework.sh ||
7142 error "($errno) Fail to stat"
7143 do_facet mgs $LCTL pool_list $FSNAME.pool1 ||
7144 error "($errno) Fail to list $FSNAME.pool1"
7145 do_facet mgs $LCTL pool_list $FSNAME.$save_fsname ||
7146 error "($errno) Fail to list $FSNAME.$save_fsname"
7147 do_facet mgs $LCTL pool_list $FSNAME.$save_fsname |
7148 grep ${FSNAME}-OST0000 ||
7149 error "($errno) List $FSNAME.$save_fsname is invalid"
7151 local pname=$($LFS getstripe --pool $DIR/$tdir/d0)
7152 [ "$pname" = "$save_fsname" ] ||
7153 error "($errno) Unexpected pool name $pname"
7157 check_mount_and_prep
7159 mkdir $DIR/$tdir || error "(1) Fail to mkdir $DIR/$tdir"
7160 cp $LUSTRE/tests/test-framework.sh $DIR/$tdir ||
7161 error "(2) Fail to copy test-framework.sh"
7163 do_facet mgs $LCTL pool_new $FSNAME.pool1 ||
7164 error "(3) Fail to create $FSNAME.pool1"
7165 # name the pool name as the fsname
7166 do_facet mgs $LCTL pool_new $FSNAME.$FSNAME ||
7167 error "(4) Fail to create $FSNAME.$FSNAME"
7169 test_103_set_pool $FSNAME OST0000
7171 $SETSTRIPE -p $FSNAME $DIR/$tdir/d0 ||
7172 error "(6) Fail to setstripe on $DIR/$tdir/d0"
7177 test_renamefs mylustre
7179 local save_fsname=$FSNAME
7183 test_103_check_pool $save_fsname 7
7185 if [ $OSTCOUNT -ge 2 ]; then
7186 test_103_set_pool $save_fsname OST0001
7189 $SETSTRIPE -p $save_fsname $DIR/$tdir/f0 ||
7190 error "(16) Fail to setstripe on $DIR/$tdir/f0"
7199 test_103_check_pool $save_fsname 17
7203 test_renamefs $save_fsname
7209 run_test 103 "rename filesystem name"
7211 if ! combined_mgs_mds ; then
7217 # restore the values of MDSSIZE and OSTSIZE
7218 MDSSIZE=$STORED_MDSSIZE
7219 OSTSIZE=$STORED_OSTSIZE